Lectures on Analytic Geometry

Lectures on Analytic Geometry
Peter Scholze (all results joint with Dustin Clausen)

Contents
Analytic Geometry 5
Preface 5
1. Lecture I: Introduction 6
2. Lecture II: Solid modules 11
3. Lecture III: Condensed R-vector spaces 16
4. Lecture IV: M-complete condensed R-vector spaces 20
Appendix to Lecture IV: Quasiseparated condensed sets 26
5. Lecture V: Entropy and a real B+
dR 28
6. Lecture VI: Statement of main result 33
Appendix to Lecture VI: Recollections on analytic rings 39
7. Lecture VII: Z((T))>r is a principal ideal domain 42
8. Lecture VIII: Reduction to “Banach spaces” 47
Appendix to Lecture VIII: Completions of normed abelian groups 54
Appendix to Lecture VIII: Derived inverse limits 56
9. Lecture IX: End of proof 57
Appendix to Lecture IX: Some normed homological algebra 65
10. Lecture X: Some computations with liquid modules 69
11. Lecture XI: Towards localization 73
12. Lecture XII: Localizations 79
Appendix to Lecture XII: Topological invariance of analytic ring structures 86
Appendix to Lecture XII: Frobenius 89
Appendix to Lecture XII: Normalizations of analytic animated rings 93
13. Lecture XIII: Analytic spaces 95
14. Lecture XIV: Varia 103
Bibliography 109
3

Analytic Geometry
Preface
These are lectures notes for a course on analytic geometry taught in the winter term 2019/20
at the University of Bonn. The material presented is part of joint work with Dustin Clausen.
The goal of this course is to use the formalism of analytic rings as defined in the course on
condensed mathematics to define a category of analytic spaces that contains (for example) adic
spaces and complex-analytic spaces, and to adapt the basics of algebraic geometry to this context;
in particular, the theory of quasicoherent sheaves.
October 2019, Peter Scholze
5

6 ANALYTIC GEOMETRY
1. Lecture I: Introduction
Mumford writes in Curves and their Jacobians: “[Algebraic geometry] seems to have acquired
the reputation of being esoteric, exclusive, and very abstract, with adherents who are secretly
plotting to take over all the rest of mathematics. In one respect this last point is accurate.”
For some reason, this secret plot has so far stopped short of taking over analysis. The goal
of this course is to launch a new attack, turning functional analysis into a branch of commutative
algebra, and various types of analytic geometry (like manifolds) into algebraic geometry. Whether
this will make these subjects equally esoteric will be left to the reader’s judgement.
What do we mean by analytic geometry? We mean a version of algebraic geometry that
(1) instead of merely allowing polynomial rings as its basic building blocks, allows rings of
convergent power series as basic building blocks;
(2) instead of being able to define open subsets only by the nonvanishing of functions, one can
define open subsets by asking that a function is small, say less than 1;
(3) strictly generalizes algebraic geometry in the sense that the category of schemes, the theory
of quasicoherent sheaves over them, etc., all embed fully faithfully into the corresponding
analytic category.
The author always had the impression that the highly categorical techniques of algebraic geom-
etry could not possibly be applied in analytic situations; and certainly not over the real numbers.1
The goal of this course is to correct this impression.
How can one build algebraic geometry? One perspective is that one starts with the abelian
category Ab of abelian groups, with its symmetric monoidal tensor product. Then one can consider
rings R in this category (which are just usual rings), and over any ring R one can consider modules
M in that category (which are just usual R-modules). Now to any R, one associates the space
Spec R, essentially by declaring that (basic) open subsets of Spec R correspond to localizations
R[f−1] of R. One can then glue these Spec R’s together along open subsets to form schemes, and
accordingly the category of R-modules glues to form the category of quasicoherent sheaves.
Now how to build analytic geometry? We are basically stuck with the first step: We want an
abelian category of some kind of topological abelian groups (together with a symmetric monoidal
tensor product that behaves reasonably). Unfortunately, topological abelian groups do not form
an abelian category: A map of topological abelian groups that is an isomorphism of underlying
abelian groups but merely changes the topology, say
(R, discrete topology) → (R, natural topology),
has trivial kernel and cokernel, but is not an isomorphism.
This problem was solved in the course on condensed mathematics last semester, by replacing
the category of topological spaces with the much more algebraic category of condensed sets, and
accordingly topological abelian groups with abelian group objects in condensed sets, i.e. condensed
abelian groups. Let us recall the definition.
1In the nonarchimedean case, the theory of adic spaces goes a long way towards fulfilling these goals, but it
has its shortcomings: Notably, it lacks a theory of quasicoherent sheaves, and in the nonnoetherian case the general
formalism does not work well (e.g. the structure presheaf fails to be sheaf). Moreover, the language of adic spaces
cannot easily be modified to cover complex manifolds. This is possible by using Berkovich spaces, but again there is
no theory of quasicoherent sheaves etc.

1. LECTURE I: INTRODUCTION 7
Definition 1.1. Consider the pro-étale site ∗proét of the point ∗, i.e. the category ProFin ∼
=
Pro(Fin) of profinite sets S with covers given by finite families of jointly surjective maps. A
condensed set is a sheaf on ∗proét; similarly, a condensed abelian group, ring, etc. is a sheaf of
abelian groups, rings, etc. on ∗proét.2
As for sheaves on any site, it is then true that a condensed abelian group, ring, etc. is the
same as an abelian group object/ring object/etc. in the category of condensed sets: for example, a
condensed abelian group is a condensed set M together with a map M×M → M (the addition map)
of condensed sets making certain diagrams commute that codify commutativity and associativity,
and admits an inverse map.
More concretely, a condensed set is a functor
X : ProFinop
→ Sets
such that
(1) one has X(∅) = ∗, and for all profinite sets S1, S2, the map X(S1 t S2) → X(S1) × X(S2)
is bijective;
(2) for any surjection S0 → S of profinite sets, the map
X(S) → {x ∈ X(S0
) | p∗
1(x) = p∗
2(x) ∈ X(S0
×S S0
)}
is bijective, where p1, p2 : S0 ×S S0 → S0 are the two projections.
How to think about a condensed set? The value X(∗) should be thought of as the underlying
set, and intuitively X(S) is the space of continuous maps from S into X. For example, if T is a
topological space, one can define a condensed X = T via
T(S) = Cont(S, T),
the set of continuous maps from S into T. One can verify that this defines a condensed set.3
Part
(1) is clear, and for part (2) the key point is that any surjective map of profinite sets is a quotient
map.
For example, in analysis a central notion is that of a sequence x0, x1, . . . converging to x∞.
This is codified by maps from the profinite set S = {0, 1, . . . , ∞} (the one-point compactification
N ∪ {∞} of N) into X. Allowing more general profinite sets makes it possible to capture more
subtle convergence behaviour. For example, if T is a compact Hausdorff space and you have any
sequence of points x0, x1, . . . ∈ T, then it is not necessarily the case that it converges in T; but one
can always find convergent subsequences. More precisely, for each ultrafilter U on N, one can take
the limit along the ultrafilter. In fact, this gives a map βN → T from the space βN of ultrafilters
on N. The space βN is a profinite set, known as the Stone-Čech compactification of N; it is the
initial compact Hausdorff space to which N maps (by an argument along these lines).
2As discussed last semester, this definition has minor set-theoretic issues. We explained then how to resolve
them; we will mostly ignore the issues in these lectures.
3There are some set-theoretic subtleties with this assertion if T does not satisfy the separation axiom T1, i.e. its
points are not closed; we will only apply this functor under this assumption.

8 ANALYTIC GEOMETRY
We have the following result about the relation between topological spaces and condensed sets,
proved last semester except for the last assertion in part (4), which can be found for example in
[BS15, Lemma 4.3.7].4
Proposition 1.2. Consider the functor T 7→ T from topological spaces to condensed sets.
(1) The functor has a left adjoint X 7→ X(∗)top sending any condensed set X to its underlying
set X(∗) equipped with the quotient topology arising from the map
G
S,a∈X(S)
S → X(∗).
(2) Restricted to compactly generated (e.g., first-countable, e.g., metrizable) topological spaces,
the functor is fully faithful.
(3) The functor induces an equivalence between the category of compact Hausdorff spaces, and
quasicompact quasiseparated condensed sets.
(4) The functor induces a fully faithful functor from the category of compactly generated weak
Hausdorff spaces (the standard “convenient category of topological spaces” in algebraic
topology), to quasiseparated condensed sets. The category of quasiseparated condensed sets
is equivalent to the category of ind-compact Hausdorff spaces “ lim
−
→i
”Ti where all transition
maps Ti → Tj are closed immersions. If X0 ,→ X1 ,→ . . . is a sequence of compact
Hausdorff spaces with closed immersions and X = lim
−
→n
Xn as a topological space, the map
lim
−
→
n
Xn → X
is an isomorphism of condensed sets. In particular, lim
−
→n
Xn comes from a topological
space.
Here the notions of quasicompactness/quasiseparatedness are general notions applying to sheaves
on any (coherent) site. In our case, a condensed set X is quasicompact if there is some profinite S
with a surjective map S → X. A condensed set X is quasiseparated if for any two profinite sets
S1, S2 with maps to X, the fibre product S1 ×X S2 is quasicompact.
The functor in (4) is close to an equivalence, and in any case (4) asserts that quasiseparated
condensed sets can be described in very classical terms. Let us also mention the following related
result.
Lemma 1.3. Let X0 ,→ X1 ,→ . . . and Y0 ,→ Y1 ,→ . . . be two sequences of compact Hausdorff
spaces with closed immersions. Then, inside the category of topological spaces, the natural map
[
n
Xn × Yn → (
[
n
Xn) × (
[
n
Yn)
is a homeomorphism; i.e. the product on the right is equipped with its compactly generated topology.
Proof. The map is clearly a continuous bijection. In general, for a union like
S
n Xn, open
subsets U are the subsets of the form
S
n Un where each Un ⊂ Xn is open. Thus, let U ⊂
S
n Xn×Yn
be any open subset, written as a union of open subset Un ⊂ Xn ×Yn, and pick any point (x, y) ∈ U.
Then for any large enough n (so that (x, y) ∈ Xn × Yn), we can find open neighborhoods Vn ⊂ Xn
4Again, some of these assertions run into minor set-theoretic problems; we refer to the notes from last semester.
Importantly, everything is valid on the nose when restricted to topological spaces with closed points, and quasisepa-
rated condensed sets; in particular, points (3) and (4) are true as stated.

1. LECTURE I: INTRODUCTION 9
of x in Xn and Wn ⊂ Yn of y in Yn, such that Vn × Wn ⊂ Un. In fact, we can ensure that even
V n × Wn ⊂ Un by shrinking Vn and Wn. Constructing the Vn and Wn inductively, we may then
moreover ensure V n ⊂ Vn+1 and Wn ⊂ Wn+1. Then V =
S
n Vn ⊂
S
n Xn and W =
S
n Wn ⊂
S
n Yn are open, and V × W =
S
n Vn × Wn ⊂ U contains (x, y), showing that U is open in the
product topology.
We asserted above that one should think of X(∗) as the underlying set of X, and about X(S)
as the continous maps from S into X. Note however that in general the map
X(S) →
Y
s∈S
X({s}) =
Y
s∈S
X(∗) = Map(S, X(∗))
may not be injective, and in fact it may and does happen that X(∗) = ∗ but X(S) 6= ∗ for large
S. However, for quasiseparated condensed sets, this map is always injective (as follows from (4)
above), and in particular a quasiseparated condensed set X is trivial as soon as X(∗) is trivial.
The critical property of condensed sets is however exactly that they can also handle non-
Hausdorff situations (i.e., non-quasiseparated situations in the present technical jargon) well. For
example, condensed abelian groups, like sheaves of abelian groups on any site, form an abelian
category. Considering the example from above and passing to condensed abelian groups, we get a
short exact sequence
0 → (R, discrete topology) → (R, natural topology) → Q → 0
of condensed abelian groups, for a condensed abelian group Q satisfying
Q(S) = Cont(S, R)/{locally constant maps S → R}
for any profinite set S.5
In particular Q(∗) = 0 while Q(S) 6= 0 for general S: There are plenty
of non-locally constant maps S → R for profinite sets S, say any convergent sequence that is not
eventually constant. In particular, Q is not quasiseparated. We see that enlarging topological
abelian groups into an abelian category precisely forces us to include non-quasiseparated objects,
in such a way that a quotient M1/M2 still essentially remembers the topology on both M1 and M2.
At this point, we have our desired abelian category, the category Cond(Ab) of condensed abelian
groups. Again, like for sheaves of abelian groups on any site, it has a symmetric monoidal tensor
product (representing bilinear maps in the obvious way). We could then follow the same steps as
for schemes. However, the resulting theory does not yet achieve our stated goals:
(1) The basic building blocks in algebraic geometry, the polynomial rings, are exactly the free
rings on some set I. As sets are generated by finite sets, really the case of a finite set is
relevant, giving rise to the polynomial algebras Z[X1, . . . , Xn]. Similarly, the basic building
blocks are now the free rings on a condensed set; as these are generated by profinite sets,
one could also just take the free rings on a profinite set S. But for a profinite set S, the
corresponding free condensed ring generated by S is not anything like a ring of (convergent)
power series. In fact, the underlying ring is simply the free ring on the underlying set of
S, i.e. an infinite polynomial algebra.
To further illustrate this point, let us instead consider the free condensed ring A
equipped with an element T ∈ A with the sequence T, T2, . . . , Tn, . . . converging 0: This
5It is nontrivial that this formula for Q is correct: A priori, this describes the quotient on the level of presheaves,
and one might have to sheafify. However, using H1
(S, M) = 0 for any profinite set S and discrete abelian group M,
as proved last semester, one gets the result by using the long exact cohomology sequence.

10 ANALYTIC GEOMETRY
is given by A = Z[S]/([∞] = 0) for the profinite set S = N ∪ {∞}. The underlying ring
of A is then still the polynomial algebra Z[T]; it is merely equipped with a nonstandard
condensed ring structure.
(2) One cannot define open subsets by “putting bounds on functions”. If say over Qp we want
to make the locus {|T| ≤ 1} into an open subset of the affine line, and agree that this
closed unit disc should correspond to the algebra of convergent power series
QphTi = {
X
n≥0
anTn
| an → 0}
while the affine line corresponds to Qp[T], then being a localization should mean that the
multiplication map
QphTi ⊗Qp[T] QphTi → QphTi
should be an isomorphism. However, this is not an isomorphism of condensed abelian
groups: On underlying abelian groups, the tensor product is just the usual algebraic
tensor product.
These failures are not unexpected: We did not yet put in any nontrivial analysis. Somewhere
we have to specify which kinds of convergent power series we want to use. Concretely, in a con-
densed ring R with a sequence T, T2, . . . , Tn, . . . converging to 0, we want to allow certain power
series
P
n≥0 rnTn where not almost all of the coefficients rn are zero. Hopefully, this specification
maintains a nice abelian category that acquires its own tensor product, in such a way that now the
tensor product computation in point (2) above works out.
Last semester, we developed such a formalism that works very well in nonarchimedean geometry:
This is the formalism of solid abelian groups that we will recall in the next lecture. In particular,
the solidification of the ring A considered in (1) is exactly the power series ring Z[[T]] (with its
condensed ring structure coming from the usual topology on the power series ring), and after
solidification the tensor product equation in (2) becomes true. Unfortunately, the real numbers
R are not solid: Concretely, T = 1
2 ∈ R has the property that its powers go to 0, but not any
sum
P
rn(1
2)n with coefficients rn ∈ Z converges in R. Thus, the solid formalism breaks down
completely over R. Now maybe that was expected? After all, the whole formalism is based on the
paradigm of resolving nice compact Hausdorff spaces like the interval [0, 1] by profinite sets, and
this does not seem like a clever thing to do over the reals; over totally disconnected rings like Qp
it of course seems perfectly sensible.
However, last semester we stated a conjecture on how the formalism might be adapted to cover
the reals. The first main goal of this course will be to prove this conjecture.

2. LECTURE II: SOLID MODULES 11
2. Lecture II: Solid modules
In this lecture, we recall the theory of solid modules that was developed last semester.
Recall from the last lecture that we want to put some “completeness” condition on modules in
such a way that the free objects behave like some kind of power series. To understand the situation,
let us first analyze the structure of free condensed abelian groups.
Proposition 2.1. Let S = lim
←
−i
Si be a profinite set, written as an inverse limit of finite sets
Si. For any n, let Z[Si]≤n ⊂ Z[Si] be the subset of formal sums
P
s∈Si
ns[s] such that
P
|ns| ≤ n;
this is a finite set, and the natural transition maps Z[Sj] → Z[Si] preserve these subsets. There is
a natural isomorphism of condensed abelian groups
Z[S] ∼
=
[
n
lim
←
−
i
Z[Si]≤n ⊂ lim
←
−
i
Z[Si].
In particular, Z[S] is a countable union of the profinite sets Z[S]≤n := lim
←
−i
Z[Si]≤n.
Note that the right-hand side indeed defines a subgroup: The addition on lim
←
−i
Z[Si] takes
Z[Si]≤n × Z[Si]≤n0
into Z[Si]≤n+n0 . We also remark that the bound imposed is as an `1-bound, but in fact it is
equivalent to an `0-bound, as only finitely many ns (in fact, n of them) can be nonzero.
Proof. By definition, Z[S] is the free condensed abelian group on S, and this is formally given
by the sheafification of the functor T 7→ Z[Cont(T, S)]. First, we check that the map
Z[S] → lim
←
−
i
Z[Si]
is an injection. First, we observe that the map of underlying abelian groups is injective. This means
that given any finite formal sum
Pk
j=1 nj[sj] where the sj ∈ S are distinct elements and nj 6= 0 are
integers, one can find some projection S → Si such that the image in Z[Si] is nonzero. But we can
arrange that the images of the sj in Si are all distinct, giving the result.
Now, assume that f ∈ Z[Cont(T, S)] maps to 0 in lim
←
−i
Z[Si](T). In particular, for all t ∈ T, the
specialization f(t) ∈ Z[S] is zero by the injectivity on underlying abelian groups. We have to see
that there is some finite cover of T by profinite sets Tm → T such that the preimage of f in each
Z[Cont(Tm, S)] is zero. Write f =
Pk
j=1 nj[gj] where gj : T → S are distinct continuous functions
and nj 6= 0. We argue by induction on k. For each pair 1 ≤ j j0 ≤ k, let Tjj0 ⊂ T be the closed
subset where gj = gj0 . Then the Tjj0 cover T: Indeed, if t ∈ T does not lie in any Tjj0 , then all
gj(t) ∈ S are pairwise distinct, and then
Pk
j=1 nj[gj(t)] ∈ Z[S] is nontrivial. Thus, we may pass to
the cover by the Tjj0 and thereby assume that gj = gj0 for some j 6= j0. But this reduces k, so we
win by induction.
As observed before,
S
n lim
←
−i
Z[Si]≤n defines a condensed abelian group, and it admits a map
from S = lim
←
−i
Si; in particular, as Z[S] is the free condensed abelian group on S, the map Z[S] →
lim
←
−i
Z[Si] factors over
S
n lim
←
−i
Z[Si]≤n. It remains to see that the induced map
Z[S] →
[
n
lim
←
−
i
Z[Si]≤n

is surjective. For this, consider the map
Sn
× {−1, 0, 1}n
= lim
←
−
i
Sn
i × {−1, 0, 1}n
→ lim
←
−
i
Z[Si]≤n
given by (x1, . . . , xn, a1, . . . , an) 7→ a1[x1] + . . . + an[xn]. This is a cofiltered limit of surjections of
finite sets, and so a surjection of profinite sets. This map sits in a commutative diagram
Sn × {−1, 0, 1}n //

lim
←
−i
Z[Si]≤n

Z[S] //
S
m lim
←
−i
Z[Si]≤m
and so implies that the lower map contains Z[Si]≤n in its image. As this works for any n, we get
the desired result.
Remark 2.2. In particular, we see that the condensed set Z[S] is quasiseparated, and in fact
comes from a compactly generated weak Hausdorff topological space Z[S]top, by Proposition 1.2 (4).
Moreover, by Lemma 1.3, the addition
Z[S]top × Z[S]top → Z[S]top
is continuous (not only when the source is equipped with its compactly generated topology), so
Z[S] really comes from a topological abelian group Z[S]top.
Exercise 2.3. Prove that for any compact Hausdorff space S, the condensed abelian group
Z[S] can naturally be written as a countable union
S
n Z[S]≤n of compact Hausdorff spaces Z[S]≤n,
and comes from a topological abelian group Z[S]top.
The idea of solid modules is that we would want to enlarge Z[S], allowing more sums deemed
“convergent”, and an obvious possibility presents itself:
Z[S]
:= lim
←
−
i
Z[Si].
In particular, this ensures that
Z[N ∪ {∞}]
/([∞] = 0) = lim
←
−
n
Z[{0, 1, . . . , n − 1, ∞}]/([∞] = 0) = lim
←
−
n
Z[T]/Tn
= Z[[T]]
is the power series algebra.6
In other words, we want to pass to a subcategory Solid ⊂ Cond(Ab) with the property that the
free solid abelian group on a profinite set S is Z[S]
. This is codified in the following definition.
Definition 2.4. A condensed abelian group M is solid if for any profinite set S with a map
f : S → M, there is a unique map of condensed abelian groups e
f : Z[S]
→ M such that the
composite S → Z[S]
→ M is the given map f.
Let us immediately state the main theorem on solid abelian groups.
6Here, Z[T] denotes the polynomial algebra in a variable T, not the free condensed module on a profinite set T.

Theorem 2.5 ([Sch19, Theorem 5.8 (i), Theorem 6.2 (i)]). The category Solid of solid abelian
groups is, as a subcategory of the category Cond(Ab) of condensed abelian groups, closed under
all limits, colimits, and extensions. The inclusion functor admits a left adjoint M 7→ M
(called
solidification), which, as an endofunctor of Cond(Ab), commutes with all colimits and takes Z[S]
to Z[S]
. The category of solid abelian groups admits compact projective generators, which are
exactly the condensed abelian groups of the form
Q
I Z for some set I. There is a unique symmetric
monoidal tensor product ⊗
on Solid making Cond(Ab) → Solid : M 7→ M
symmetric monoidal.
The theorem is rather nontrivial. Indeed, a small fraction of it asserts that Z is solid. What
does this mean? Given a profinite set S = lim
←
−i
Si and a continuous map f : S → Z, there should
be a unique map e
f : Z[S]
→ Z with given restriction to S. Note that
Cont(S, Z) = lim
−
→
i
Cont(Si, Z) = lim
−
→
i
Hom(Z[Si], Z) → Hom(lim
←
−
i
Z[Si], Z) = Hom(Z[S]
, Z).
In particular, the existence of f is clear, but the uniqueness is not. We need to see that any
map Z[S]
→ Z of condensed abelian groups factors over Z[Si] for some i. This expresses some
“compactness” of Z[S]
that seems to be hard to prove by a direct attack (if S is a countable limit
of finite sets, it is however possible). We note in particular that we do not know whether the similar
result holds with Z replaced by any ring A (and in fact the naive translation of the theorem above
fails for general rings).
Let us analyze the structure of Z[S]
. It turns out that it can be regarded as a space of
measures. More precisely,
Z[S]
= lim
←
−
i
Z[Si] = lim
←
−
i
Hom(C(Si, Z), Z) = Hom(lim
−
→
i
C(Si, Z), Z) = Hom(C(S, Z), Z)
is the space dual to the (discrete) abelian group of continuous functions S → Z. Accordingly, we
will often denote elements of Z[S]
as µ ∈ Z[S]
and refer to them as measures.
Thus, in a solid abelian group, it holds true that whenever f : S → M is a continuous map,
and µ ∈ Z[S]
is a measure, one can form the integral
R
S fµ := e
f(µ) ∈ M. Again, an im-
portant special case is when S = N ∪ {∞}. In that case, f can be thought of as a convergent
sequence m0, m1, . . . , mn, . . . , m∞ in M, and a measure µ on S can be characterized by the masses
a0, a1, . . . , an, . . . it gives to the finite points (which are isolated in S), as well as the mass a it gives
to all of whose S. Then formally we have
Z
S
fµ = a0(m0 − m∞) + a1(m1 − m∞) + . . . + an(mn − m∞) + . . . + am∞ ;
in other words, the infinite sums on the right are defined in M. In particular, if m∞ = 0, then any
sum
P
aimi with coefficients ai ∈ Z is defined in M. Note that the sense in which it is defined is
a tricky one: It is not directly as any kind of limit of the finite sums; rather, it is characterized by
the uniqueness of the map Z[S]
→ M with given restriction to S. Roughly, this says that there
is only one way to consistently define such sums for all measures µ on S simultaneously; and then
one evaluates for any given µ.
Now we want to explain the proof of Theorem 2.5 when restricted to Fp-modules for some prime
p. (The arguments below could be adapted to Z/nZ or Zp with minor modifications.) This leads to
some important simplifications. Most critically, the underlying condensed set of Fp[S]
is profinite.

In other words, we set
Fp[S]
:= lim
←
−
i
Fp[Si] = Z[S]
/p
and define the category of solid Fp-modules Solid(Fp) ⊂ Cond(Fp) as the full subcategory of all
condensed Fp-modules such that for all profinite sets S with a map f : S → M, there is a unique
extension to a map Fp[S]
→ M of condensed Fp-modules. We note that equivalently we are simply
considering the p-torsion full subcategories inside Solid resp. Cond(Ab).
In the following, we use two simple facts about condensed sets:
(1) If M is a discrete set, considered as a condensed set, then for any profinite set S = lim
←
−i
Si,
M(S) = C(S, M) = lim
−
→
i
C(Si, M) = lim
−
→
i
M(Si).
(2) If Xi, i ∈ I, is any filtered system of condensed sets, then for any profinite set S,
(lim
−
→
i
Xi)(S) = lim
−
→
i
Xi(S).
For the latter, one has to see that the filtered colimit is still a sheaf on ∗proét, which follows
from the commutation of equalizers with filtered colimits.
Proposition 2.6. The discrete Fp-module Fp is solid.
Proof. We have to prove that
Hom(lim
←
−
i
Fp[Si], Fp) = lim
−
→
i
C(Si, Fp).
But any map lim
←
−i
Fp[Si] → Fp of condensed abelian groups can be regarded as a map of condensed
sets. As the source is profinite and the target is discrete, it follows that the map factors over Fp[Si]
for some i. A priori this factorization is only as condensed sets, but if we assume that the transition
maps are surjective, it is automatically a factorization as condensed abelian groups. This gives the
desired result.
Corollary 2.7. For any set I, the profinite Fp-vector space
Q
I Fp is solid.
Proof. It follows directly from the definition that the class of solid Fp-modules is stable under
all limits.
Let us say that a condensed Fp-vector space V is profinite if the underlying condensed set
is profinite. One checks easily that any such V can be written as a cofiltered limit of finite-
dimensional Fp-vector spaces, and then that V 7→ V ∗ = Hom(V, Fp) defines an anti-equivalence
between profinite Fp-vector spaces and discrete Fp-vector spaces. In particular, any profinite Fp-
vector space is isomorphic to
Q
I Fp for some set I. From here, it is not hard to verify the following
proposition:
Proposition 2.8. The class of profinite Fp-vector spaces forms an abelian subcategory of
Cond(Fp) stable under all limits, cokernels, and extensions.
Proof. Let f : V → W be a map of profinite Fp-vector spaces. This can be written as a
cofiltered limit of maps fi : Vi → Wi of finite-dimensional Fp-vector spaces. One can take kernels
and cokernels of each fi, and then pass to the limit to get the kernel and cokernel of f, which will
then again be profinite Fp-vector spaces. For stability under extensions, we only have to check that
the middle term is profinite as a condensed set, which is clear.

Theorem 2.9. A condensed Fp-module M is solid if and only if it is a filtered colimit of profinite
Fp-vector spaces. These form an abelian category stable under all limits and colimits.
One could also prove the rest of Theorem 2.5 in this setting.
Proof. We know that profinite Fp-vector spaces are solid. As Hom(Fp[S]
, −) (as a functor
from Cond(Fp) to sets) commutes with filtered colimits as Fp[S]
is profinite, the class of solid Fp-
vector spaces is stable under all filtered colimits, and as observed before it is stable under all limits.
Next, we observe that the class of filtered colimits of profinite Fp-vector spaces forms an abelian
subcategory of Cond(Fp). For this, note that any map f : V → W of such is a filtered colimit of
maps fi : Vi → Wi of profinite Fp-vector spaces, and one can form the kernel and cokernel of each fi
and then pass to the filtered colimit again. In particular, they are stable under all colimits (which
are generated by cokernels, finite direct sums, and filtered colimits).
It remains to see that any solid M is a filtered colimit of profinite Fp-vector spaces. Any M
admits a surjection
L
j Fp[Sj] → M for certain profinite sets Sj → M. As M is solid, this gives
a surjection V → M where V =
L
j Fp[Sj]
is a filtered colimit of profinite Fp-vector spaces. As
solid modules are stable under limits, in particular kernels, the kernel of this map is again solid,
and so by repeating the argument we find a presentation W → V → M → 0 where W and V are
filtered colimits of profinite Fp-vector spaces. As this class is stable under quotients, we see that
also M is such a filtered colimit, as desired.

3. Lecture III: Condensed R-vector spaces
The first goal of this course is to define an analogue of solid modules over the real numbers.
Roughly, we want to define a notion of liquid R-vector spaces that is close to the notion of complete
locally convex R-vector spaces, but is a nice abelian category (which topological vector spaces of
any kind are not).
To get started, we will translate some of the classical theory of topological vector spaces into
the condensed setup. The most familiar kind of topological R-vector space are the Banach spaces.
Definition 3.1. A Banach space is a topological R-vector space V that admits a norm || · ||,
i.e. a continuous function
|| · || : V → R≥0
with the following properties:
(1) For any v ∈ V , the norm ||v|| = 0 if and only if v = 0;
(2) For all v ∈ V and a ∈ R, one has ||av|| = |a|||v||;
(3) For all v, w ∈ V , one has ||v + w|| ≤ ||v|| + ||w||;
(4) The sets {v ∈ V | ||v|| } for varying ∈ R0 define a basis of open neighborhoods of 0;
(5) For any sequence v0, v1, . . . ∈ V with ||vi −vj|| → 0 as i, j → ∞, there exists a (necessarily
unique) v ∈ V with ||v − vi|| → 0.
We note that this notion is very natural from the topological point of view in the sense that
it is easy to say what the open subsets of V are – they are the unions of open balls with respect
to the given norm. On the other hand, V is clearly a metrizable topological space (with distance
d(x, y) = ||y − x||), so in particular first-countable, so in particular compactly generated, so the
passage to the condensed vector space V does not lose information. Let us try to understand V ;
in other words, we must understand how profinite sets map into V .
Proposition 3.2. Let V be a Banach space, or more generally a complete locally convex topo-
logical vector space. Let S be a profinite set, or more generally a compact Hausdorff space, and let
f : S → V be a continuous map. Then f factors over a compact absolutely convex subset K ⊂ V ,
i.e. a compact Hausdorff subspace of V such that for all x, y ∈ K also ax + by ∈ K whenever
|a| + |b| ≤ 1.
In other words, as a condensed set V is the union of its compact absolutely convex subsets.
Recall that a topological vector space V is locally convex if it has a basis of neighborhoods U
of 0 such that for all x, y ∈ U also ax + by ∈ U whenever |a| + |b| ≤ 1. It is complete if every
Cauchy net converges, i.e. for any directed index set I and any map I → V : i 7→ xi, if xi − xj
converges to 0, then there is a unique x ∈ V so that x − xi converges to 0. (Note that V may fail
to be metrizable, so one needs to pass to nets.)
To prove the proposition, we need to find K. It should definitely contain all convex combinations
of images of points in S; and all limit points of such. This quickly leads to the idea of integrating a
(suitably bounded) measure on S against f. The intuition is that if a profinite set S maps into V ,
then one can also integrate any measure on S against this map, to produce a map from the space
of measures on S towards V .
More precisely, in case S = lim
←
−i
Si is profinite, consider the space of (“signed Radon”) measures
of norm ≤ 1,
M(S)≤1 := lim
←
−
i
M(Si)≤1,

3. LECTURE III: CONDENSED R-VECTOR SPACES 17
where M(Si)≤1 ⊂ R[Si] is the subspace of `1-norm ≤ 1. More generally, increasing the norm, we
define
M(S) =
[
c0
M(S)≤c, M(S)≤c = lim
←
−
i
M(Si)≤c.
This can be defined either in the topological or condensed setting, and is known as the space of
“signed Radon” measures. We note that we do not regard M(S) as a Banach space with the norm
given by c. We also stress the strong similarity with the description of Z[S].
Exercise 3.3. For any compact Hausdorff space S, there is a notion of signed Radon measure
on S – look up the official definition. Show that the corresponding topological vector space M(S),
with the weak topology, is a union of compact Hausdorff subspaces M(S)≤c (but beware that the
weak topology is not the induced colimit topology). Moreover, show that if S is a profinite set,
then a signed Radon measure on S is equivalent to a map µ assigning to each open and closed
subset U ⊂ S of S a real number µ(U) ∈ R, so that µ(U t V ) = µ(U) + µ(V ) for two disjoint U, V ,
and there is some constant C = C(µ) such that for all disjoint decompositions S = U1 t . . . t Un,
n
X
i=1
|µ(Ui)| ≤ C.
Moreover, for a general compact Hausdorff space S, choose a surjection e
S → S from a profinite
set, so that S = e
S/R for the equivalence relation R = e
S ×S
e
S ⊂ e
S × e
S. Show that M(S) is the
coequalizer of M(R) ⇒ M(e
S). Thus, the rather complicated notion of signed Radon measures on
general compact Hausdorff spaces is simply a consequence of descent from profinite sets.
Proposition 3.4. Let V be a complete locally convex R-vector space. Then any continuous
map f : S → V from a profinite set S extends uniquely to a map of topological R-vector spaces
M(S) → V : µ 7→
R
S fµ. The image of M(S)≤1 is a compact absolutely convex subset of V
containing S.
Remark 3.5. The exercise implies that the same holds true for any compact Hausdorff S.
Also note that maps M(S) → V of topological vector spaces are equivalent to maps between the
corresponding condensed vector spaces, as the source is compactly generated.
Proof. The final sentence is clear as V is Hausdorff and so the image of any compact set is
compact; and clearly M(S)≤1 is absolutely convex (hence so is its image), and the image contains
S (as S ⊂ M(S)≤1 as Dirac measures).
We have to construct the map M(S) → V , so take µ ∈ M(S), and by rescaling assume that
µ ∈ M(S)≤1. Write S = lim
←
−i
Si as a limit of finite sets Si with surjective transition maps, and pick
any lift ti : Si → S of the projection π : S → Si (we ask for no compatibilities between different
ti). We define a net in V as follows: For each i, let
vi =
X
s∈Si
f(ti(s))µ(π−1
i (s)) ∈ V.
We have to see that this is a Cauchy net, so pick some absolutely convex neighborhood U of 0.
Choosing i large enough, we can (by continuity of f) ensure that for any two choices ti, t0
i : Si → S,
one has f(ti(s)) − f(t0
i(s)) ∈ U (in other words, f varies at most within a translate of U on the
preimages of S → Si). As
P
s∈Si
|µ(π−1
i (s))| ≤ 1, we get a convex combination of such differences,

which still lies in U. This verifies that each vi does not depend on the choice of the ti up to
translation by an element in U, and a slight refinement then proves that one gets a Cauchy net.
By completeness of V , we get a unique limit v ∈ V of the vi. The proof essentially also gives
continuity: Note that the choice of i depending on U only depended on f, not on µ.
Thus, from the condensed point of view, the following concepts seem fundamental: Compact
absolutely convex sets, and the spaces of measures M(S) with their compactly generated topology.
Note that the latter are ill-behaved from the topological point of view: It is hard to say what a
general open subset of M(S) looks like; one cannot do better than naively taking the increasing
union
S
c0 M(S)≤c of compact Hausdorff spaces.
These concepts lead one to the definition of a Smith space:
Definition 3.6. A Smith space is a complete locally convex topological R-vector space V that
admits a compact absolutely convex subset K ⊂ V such that V =
S
c0 cK with the induced
compactly generated topology on V .
One can verify that M(S) is a Smith space: One needs to check that it is complete and locally
convex. We will redefine Smith spaces in the next lecture, and check that it defines a Smith space
in the latter sense.
Corollary 3.7. Let V be a complete locally convex topological R-vector space, and consider
the category of Smith spaces W ⊂ V . Then this category is filtered, and as condensed R-vector
spaces
V = lim
−
→
W⊂V
W.
Proof. To see that the category is filtered, let W1, W2 ⊂ V be two Smith subspaces, with
compact generating subsets K1, K2. Then K1 t K2 → V factors over a compact absolutely convex
subset K ⊂ V by the proposition above, and W =
S
c0 cK ⊂ V (with the inductive limit topology)
is a Smith space containing both W1 and W2. By the same token, if f : S → V is any map from a
profinite set, then f factors over a compact absolutely convex subset K ⊂ V , and hence over the
Smith space W =
S
c0 cK ⊂ V , proving that
V = lim
−
→
W⊂V
W.

This realizes the idea that from the condensed point of view, Smith spaces are the basic building
blocks. On the other hand, it turns out that Smith spaces are closely related to Banach spaces:
Theorem 3.8 (Smith, [Smi52]). The categories of Smith spaces and Banach spaces are anti-
equivalent. More precisely, if V is a Banach space, then Hom(V, R) is a Smith space; and if W is
a Smith space, then Hom(W, R) is a Banach space, where in both cases we endow the dual space
with the compact-open topology. The corresponding biduality maps are isomorphisms.
This will be proved in the next lecture.
Remark 3.9. This gives one sense in which Banach spaces are always reflexive, i.e. isomorphic
to their bidual. Note, however, that this does not coincide with the usual notion of reflexivity: It
is customary to make the dual of a Banach space itself into a Banach space (by using the norm

3. LECTURE III: CONDENSED R-VECTOR SPACES 19
||f|| = supv∈V,||v||≤1 |f(v)|), and then reflexivity asks whether a Banach space is isomorphic to
its corresponding bidual. Even if that happens, the two notions of dual are still different: If a
topological vector space is both Banach and Smith, it is finite-dimensional.

4. Lecture IV: M-complete condensed R-vector spaces
The discussion of the last lecture motivates the following definition.
Definition 4.1. Let V be a condensed R-vector space. Then V is M-complete if (the under-
lying condensed set of) V is quasiseparated, and for all maps f : S → V from a profinite set, there
is an extension to a map
e
f : M(S) → V
of condensed R-vector spaces.
Definition 4.2 (slight return). A Smith space is an M-complete condensed R-vector space V
such that there exists some compact Hausdorff K ⊂ V with V =
S
c0 cK.
Exercise 4.3. Prove that this definition of Smith spaces is equivalent to the previous definition.
We note that it is clear that Smith spaces defined in this way form a full subcategory of topo-
logical R-vector spaces (using Proposition 1.2 and Lemma 1.3). Moreover, Proposition 3.4 implies
that any Smith space in the sense of the last lecture is a Smith space in the current sense. It
remains to see that for any Smith space in the current sense, the corresponding topological vector
space is complete and locally convex. Note that it follows easily that in the current definition one
can take K to be absolutely convex.
Let us check first that for any profinite set T, the condensed R-vector space M(T) is a Smith
space. We only need to see that it is M-complete. Any map f : S → M(T) factors over M(T)≤c
for some c. Then writing T = lim
←
−i
Ti as an inverse limit of finite sets Ti, f is an inverse limit of
maps fi : S → M(Ti)≤c. This reduces us to the case that T is finite (provided we can bound the
norm of the extension); but then we can apply Proposition 3.4 (and observing that the extension
maps M(S)≤1 into M(Ti)≤c).
Our initial hope was that M-complete condensed R-vector spaces, without the quasiseparation
condition, would behave as well as solid Z-modules. We will see in the next lecture that this is
not so. In this lecture, we however want to show that with the quasiseparation condition, the
category behaves as well as it can be hoped for: It is not an abelian category (because cokernels
are problematic under quasiseparatedness), but otherwise nice.
First, we check that the extension e
f is necessarily unique.
Proposition 4.4. Let V be a quasiseparated condensed R-vector space and let g : M(S) → V
be a map of condensed R-vector space for some profinite set S. If the restriction of g to S vanishes,
then g = 0.
Proof. The preimage g−1(0) ⊂ M(S) is a quasicompact injection of condensed sets. This
means that it is a closed subset in the topological sense, see the appendix to this lecture.
On the other hand, g−1(0) contains the R-vector space spanned by S. This contains a dense
subset of M(S)≤c for all c, and thus its closure by the preceding.
Exercise 4.5. Show that in the definition of M-completeness, one can restrict to extremally
disconnected S.
Proposition 4.6. Any M-complete condensed R-vector space V is the filtered colimit of the
Smith spaces W ⊂ V ; conversely, any filtered colimit of Smith spaces along injections is quasisepa-
rated and M-complete. For any map f : V → V 0 between M-complete condensed R-vector spaces,

4. LECTURE IV: M-COMPLETE CONDENSED R-VECTOR SPACES 21
the kernel and image of f (taken in condensed R-vector spaces) are M-complete condensed R-vector
spaces. Moreover, f admits a cokernel in the category of M-complete condensed R-vector spaces.
Proof. Let V be an M-complete condensed R-vector space and let f : S → V be a map
from a profinite set. Then f extends uniquely to e
f : M(S) → V . The subset M(S)≤1 ⊂ M(S)
is compact Hausdorff (i.e., a quasicompact quasiseparated condensed set); thus, its image in the
quasiseparated V is still quasicompact and quasiseparated, i.e. a compact Hausdorff K ⊂ V . Let
W =
S
c0 cK ⊂ V . This is a condensed R-vector space, and it is itself M-complete. Indeed,
assume that T is some profinite set with a map T → W; by rescaling, we can assume that T
maps into K. We get a unique extension to a map M(T) → V , and we need to see that it factors
over W. We may assume that T is extremally disconnected (as any surjection T0 → T from an
extremally disconnected T0 induces a surjection M(T0) → M(T)). Then the map T → K lifts to a
map T → M(S)≤1. Thus, we get a map M(T) → M(S) whose composite with M(S) → V must
agree with the original map M(T) → V by uniqueness. This shows that the image of M(T) → V
is contained in W, as desired. It follows that W is a Smith space.
It is now easy to see that the such of W is filtered, and then V is the filtered colimit, as we have
seen that any map from a profinite set S factors over one such W. It is clear that any filtered colimit
lim
−
→i
Wi of Smith spaces Wi along injections is quasiseparated (as quasiseparatedness is preserved
under filtered colimits of injections). It is also M-complete, as any map f : S → lim
−
→i
Wi factors
over one Wi, and then extends to a map e
f : M(S) → Wi → lim
−
→i
Wi.
For the assertions about a map f, the claim about the kernel is clear. For the image, we may
assume that V is a Smith space, by writing it as a filtered colimit of such; and then that V = M(S)
by picking a surjection S → K ⊂ V where K is a generating compact Hausdorff subset of V . But
then the image of f is exactly the Smith space W ⊂ V constructed in the first paragraph.
For the cokernel of f, we may first take the cokernel Q in the category of condensed R-vector
spaces. Now we may pass to the maximal quasiseparated quotient Qqs of Q, which is still a
condensed R-vector space, by the appendix to this lecture. It is then formal that this satisfies
the conditon of being M-complete when tested against extremally disconnected S, as any map
S → Qqs lifts to S → W, and so extends to M(S) → W → Qqs. The automatic uniqueness
of this extension then implies the result for general profinite S (by covering S by an extremally
disconnected e
S → S).
We will now prove the anti-equivalence between Smith spaces in the current sense, with Banach
spaces. (In particular, using the original form of Smith’s theorem, this also solves the exercise
above.)
Theorem 4.7. For any Smith space W, the internal dual HomR(W, R) is isomorphic to V for
a Banach space V . Conversely, for any Banach space V , the internal dual HomR(V , R) is a Smith
space. The corresponding biduality maps are isomorphisms.
Proof. Assume first that W = M(S, R) for some profinite set S. Then HomR(M(S, R), R) =
C(S, R) as we have seen that any continuous map S → R extends uniquely to a map of condensed
R-vector spaces M(S, R) → R. More generally, for any profinite set T,
HomR(M(S, R), R)(T) = HomR(M(S, R), HomR(R[T], R)),
and HomR(R[T], R) = C(T, R), for the Banach space C(T, R) with the sup-norm. Indeed, this
Banach space has the property that continuous maps T0 → C(T, R) are equivalent to continuous

maps T0 × T → R. Thus, using that Banach spaces are complete locally convex, we see
HomR(M(S, R), HomR(R[T], R)) = HomR(M(S, R), C(T, R)) = C(T, R)(S) = C(T × S, R),
showing that indeed HomR(M(S, R), R) = C(S, R).
Now assume that V = C(S, R). We need to see that for any profinite set T,
HomR(C(S, R), C(T, R)) = M(S)(T).
But the left-hand side can be computed in topological vector spaces, and any map of Banach spaces
has bounded norm. It suffices to see that maps C(S, R) → C(T, R) of Banach spaces of norm ≤ 1 are
in bijection with maps T → M(S)≤1. But any such map of Banach spaces is uniquely determined
by its restriction to lim
−
→i
C(Si, R), where S = lim
←
−i
Si is written as a limit of finite sets, where again
we assume that the operator norm is bounded by 1. Similarly, maps T → M(S)≤1 are limits of
maps T → M(Si)≤1. We can thus reduce to the case S is finite, where the claim is clear. We see
also that the biduality maps are isomorphisms in this case.
Now let W be a general Smith space, with a generating subset K ⊂ W. Pick a profinite set S
with a surjection S → K. We get a surjection M(S) → W. Its kernel W0 ⊂ M(S) is automatically
quasiseparated, and in fact a Smith space again: It is generated by W0 ∩M(S)≤1, which is compact
Hausdorff. Picking a further surjection M(S0) → W0, we get a resolution
M(S0
) → M(S) → W → 0.
Taking HomR(−, R), we get
0 → HomR(W, R) → C(S, R) → C(S0
, R).
The latter map corresponds to a map of Banach space C(S, R) → C(S0, R), whose kernel is a closed
subspace that is itself a Banach space V , and we find HomR(W, R) = V .
Conversely, if V is any Banach space, equipped with a norm, the unit ball B = {f ∈ Hom(V, R) |
||f|| ≤ 1} is compact Hausdorff when equipped with the weak topology.7
Picking a surjection
S → B from a profinite set S, we get a closed immersion V → C(S, R) (to check that it is a closed
immersion, use that it is an isometric embedding by the Hahn-Banach theorem), whose quotient
will then be another Banach space; thus, any Banach space admits a resolution
0 → V → C(S, R) → C(S0
, R)
We claim that taking HomR(−, R) gives an exact sequence
M(S0
) → M(S) → HomR(V , R) → 0.
For this, using that we already know about spaces of measures, we have to see that for any ex-
tremally disconnected T, the sequence
HomR(C(S0
, R), C(T, R)) → HomR(C(S, R), C(T, R)) → HomR(V, C(T, R)) → 0
7This is known as the Banach-Alaoglu theorem, and follows from Tychonoff by using the closed embedding
B ,→
Q
v∈V [−1, 1].

is exact. This follows from the fact that the Banach space C(T, R) is injective in the category of
Banach spaces, cf. [ASC+16, Section 1.3]: If V ⊂ V 0 is a closed embedding of Banach spaces, then
any map V → C(T, R) of Banach spaces extends to a map V 0 → C(T, R).8
Using these resolutions, it also follows that the biduality maps are isomorphisms in general.
We remark that if S is extremally disconnected, then M(S) is a projective object in the category
of Smith spaces (in fact, in the category of M-complete condensed R-vector spaces), as S is a
projective object in the category of condensed sets. Dually, this means that C(S, R) is an injective
object in the category of Banach spaces. For solid abelian groups, it turned out that Z[S]
is
projective in the category of solid abelian groups, for all profinite sets S. Could a similar thing
happen here? It turns out, that no:
Proposition 4.8. For “most” profinite sets S, for example S = N ∪ {∞}, or a product S =
S1 × S2 of two infinite profinite sets S1, S2, the Banach space C(S, R) is not injective; equivalently,
the Smith space M(S) is not projective.
Proof. Note that C(N∪{∞}, R) is the product of the Banach space c0 of null-sequences, with
R. One can show that C(S1 × S2, R) contains c0 as a direct factor, cf. [Cem84], so it is enough to
show that c0 is not injective, which is [ASC+16, Theorem 1.25].
In fact, cf. [ASC+16, Section 1.6.1], there is no known example of an injective Banach space
that is not isomorphic to C(S, R) where S is extremally disconnected!
Next, let us discuss tensor products. Before making the connection with the existing notions for
Banach spaces, let us follow our nose and define a tensor product on the category of M-complete
condensed R-vector spaces.
Proposition 4.9. Let V and W be M-complete condensed R-vector spaces. Then there is an
M-complete condensed R-vector space V ⊗π W equipped with a bilinear map
V × W → V ⊗π W
of condensed R-vector spaces, which is universal for bilinear maps; i.e. any bilinear map V ×W → L
to a M-complete condensed R-vector space extends uniquely to a map V ⊗π W → L of condensed
R-vector spaces.
The functor (V, W) 7→ V ⊗π W from pairs of M-complete condensed R-vector spaces to M-
complete condensed R-vector spaces commutes with colimits in each variable, and satisfies
M(S) ⊗π M(S0
) ∼
= M(S × S0
)
for any profinite sets S, S0.
Note that the case of S1 × S2 in Proposition 4.8 means that a tensor product M(S1 × S2) of
two projectives M(S1), M(S2) is not projective anymore. This is a subtlety that will persist, and
that we have to live with.
Proof. It is enough to show that M(S ⊗ S0) represents bilinear maps M(S) × M(S0) → L;
indeed, the other assertions will then follow via extending everything by colimits to the general
case.
8The case T = ∗ is the Hahn-Banach theorem. Part of the Hahn-Banach theorem is that the extension can be
bounded by the norm of the original map. This implies the injectivity of l∞
(S0, R) = C(βS0, R) for any set S0, and
then the result for general T follows by writing T as a retract of βS0 for some S0.

In other words, we have to show that bilinear maps M(S) × M(S0) → L are in bijection with
maps S × S0 → L. Any map S × S0 → L extends to M(S × S0) → L, and thus gives a bilinear map
M(S)×M(S0) → L, as one can directly define a natural bilinear map M(S)×M(S0) → M(S×S0).
Thus, it is enough to show that any bilinear map M(S) × M(S0) → L that vanishes on S × S0 is
actually zero. For this, we again observe that it vanishes on a dense subset of M(S)≤c×M(S0)≤c for
any c, and thus on the whole of it, as the kernel is a closed subspace when L is quasiseparated.
Tensor products of Banach spaces were defined by Grothendieck, [Gro55], and interestingly he
has defined several tensor products. There are two main examples. One is the projective tensor
product V1 ⊗π V2; this is a Banach space that represents bilinear maps V1 × V2 → W.
Proposition 4.10. Let V1 and V2 be two Banach spaces. Then V1 ⊗π V2
∼
= V1 ⊗π V2.
Remark 4.11. The proposition implies that any compact absolutely convex subset K ⊂ V1⊗πV2
is contained in the closed convex hull of K1 × K2 for compact absolutely convex subsets Ki ⊂ Vi.
Proof. Any Banach space admits a projective resolution by spaces of the form `1(I) for some
set I; this reduces us formally to that case. One can even reduce to the case that I is countable,
by writing `1(I) as the ω1-filtered colimit of `1(J) over all countable J ⊂ I. Thus, we may assume
that V1 = V2 = `1(N) is the space of `1-sequences. In that case V1 ⊗π V2 = `1(N × N).
In that case,
`1
(N) = lim
−
→
(λn)n
M(N ∪ {∞})/(R · [∞])
where the filtered colimit is over all null-sequences of positive real numbers 0 λn ≤ 1. Now the
result comes down to the observation that any null-sequence λn,m of positive real numbers in (0, 1]
parametrized by pairs n, m ∈ N can be bounded from above by a sequence of the form λnλ0
m where
λn and λ0
m are null-sequences of real numbers in (0, 1]. Indeed, one can take
λn = λ0
n =
r
max
n0,m0≤n
λn0,m0 .
On the other hand, there is the injective tensor product V1 ⊗ V2; this satisfies C(S1, R) ⊗
C(S2, R) ∼
= C(S1 × S2, R). There is a map V1 ⊗π V2 → V1 ⊗ V2, that is however far from an
isomorphism.
Proposition 4.12. Let Vi, i = 1, 2, be Banach spaces with dual Smith spaces Wi. Let W =
W1 ⊗π W2, which is itself a Smith space, and let V be the Banach space dual to W. Then there is
a natural map
V1 ⊗ V2 → V
that is a closed immersion of Banach spaces. It is an isomorphism if V1 or V2 satisfies the approx-
imation property.
We recall that most natural Banach spaces have the approximation property; in fact, it had
been a long-standing open problem whether all Banach spaces have the approximation property,
until a counterexample was found by Enflo, [Enf73] (for which he was awarded a live goose by
Mazur).

Proof. The injective tensor product has the property that if V1 ,→ V 0
1 is a closed immersion
of Banach spaces, then V1 ⊗ V2 ,→ V 0
1 ⊗ V2 is a closed immersion. However, it is not in general
true that if
0 → V1 → V 0
1 → V 00
1
is a resolution, then
0 → V1 ⊗ V2 → V 0
1 ⊗ V2 → V 00
1 ⊗ V2
is a resolution: Exactness in the middle may fail. The statement is true when V2 = C(S2, R) is the
space of continuous functions on some profinite set S2, as in that case W ⊗ V2 = C(S2, W) for any
Banach space W, and C(S2, −) preserves exact sequences of Banach spaces for profinite sets S2.
In any case, such resolutions prove the desired statement: Note that if Vi = C(Si, R), one has
Wi = M(Si), and then W = M(S1 × S2) and so V = C(S1 × S2, R), which is indeed V1 ⊗ V2; and
these identifications are functorial. Now the resolution above gives the same statement as long as
V2 = C(S2, R); and in general one sees that V is the kernel of the map V 0
1 ⊗ V2 → V 00
1 ⊗ V2.
It remains to see that the map V1 ⊗ V2 → V is an isomorphism if one of V1 and V2 has the
approximation property. Assume that V2 has the approximation property. Note that
V = HomR(W1 ⊗R W2, R) = HomR(W1, HomR(W2, R)) = HomR(W1, V2).
In the classical literature, this is known as the weak-∗-to-weak compact operators from V ∗
1 to V2.
On the other hand, V1 ⊗ V2 ⊂ V is the closed subspace generated by algebraic tensors V1 ⊗ V2;
equivalently, this is the closure of the space of maps W1 → V2 of finite rank. The condition that
V2 has the approximation property precisely ensures that this is all of HomR(W1, V2), cf. [DFS08,
Theorem 1.3.11].
In other words, when Banach spaces are covariantly embedded into M-complete condensed
R-spaces, one gets the projective tensor product; while under the duality with Smith spaces, one
(essentially) gets the injective tensor product. Thus, in the condensed setting, there is only one
tensor product, but it recovers both tensor products on Banach spaces.

Appendix to Lecture IV: Quasiseparated condensed sets
In this appendix, we make some remarks about the category of quasiseparated condensed sets.
We start with the following critical observation, giving an interpretation of the topological space
X(∗)top purely in condensed terms.
Proposition 4.13. Let X be a quasiseparated condensed set. Then quasicompact injections
i : Z ,→ X are equivalent to closed subspaces W ⊂ X(∗)top via Z 7→ Z(∗)top, resp. sending a closed
subspace W ⊂ X(∗)top to the subspace Z ⊂ X with
Z(S) = X(S) ×Map(S,X(∗)) Map(S, W).
In the following, we will sometimes refer to quasicompact injections i : Z ,→ X as closed
subspaces of X, noting that by the proposition this agrees with the topological notion.
Proof. The statement reduces formally to the case that X is quasicompact by writing X as the
filtered union of its quasicompact subspaces. In that case, X is equivalent to a compact Hausdorff
space. If i : Z ,→ X is a quasicompact injection, then Z is again quasicompact and quasiseparated,
so again (the condensed set associated to) a compact Hausdorff space. But injections of compact
Hausdorff spaces are closed immersions. In other words, the statement reduces to the assertion
that under the equivalence of quasicompact quasiseparated condensed sets with compact Hausdorff
spaces, injections correspond to closed subspaces, which is clear.
Lemma 4.14. The inclusion of the category of quasiseparated condensed sets into all condensed
sets admits a left adjoint X 7→ Xqs, with the unit X → Xqs being a surjection of condensed sets.
The functor X 7→ Xqs preserves finite products. In particular, it defines a similar left adjoint
for the inclusion of quasiseparated condensed A-modules into all condensed A-modules, for any
quasiseparated condensed ring A.
Proof. Choose a surjection X0 =
F
i Si → X from a disjoint union of profinite sets Si, giving
an equivalence relation R = X0 ×X X0 ⊂ X0 ×X0. For any map X → Y with Y quasiseparated, the
induced map X0 → Y has the property that X0 ×Y X0 ⊂ X0 × X0 is a quasicompact injection (i.e.,
is a closed subspace) and is an equivalence relation that contains R; thus, it contains the minimal
closed equivalence relation R ⊂ X0 × X0 generated by R (which exists, as any intersection of closed
equivalence relations is again a closed equivalence relation). This shows that Xqs = X0/R defines
the desired adjoint.
To check that it preserves finite products, we need to check that if R ⊂ X ×X and R0 ⊂ X0 ×X0
are two equivalence relations on quasiseparated condensed sets X, X0, then the minimal closed
equivalence relation R × R0 on X × X0 containing R × R0 is given by R × R0. To see this, note
first that for fixed x0 ∈ X0, it must contain R × (x0, x0) ⊂ X × X × X0 × X0. Similarly, for fixed
x ∈ X, it must contain (x, x)×R0. But now if (x1, x0
1) and (x2, x0
2) are two elements of X ×X0 such
that x1 is R-equivalent to x2 and x0
1 is R0-equivalent to x0
2, then (x1, x0
1) is R × R0-equivalent to
(x2, x0
1), which is R × R0-equivalent to (x2, x0
2). Thus, R × R0 ⊂ R × R0, and the reverse inclusion
is clear.
Corollary 4.15. The inclusion of the category of M-complete condensed R-vector spaces into
all condensed R-vector spaces admits a left adjoint, the “M-completion”.

APPENDIX TO LECTURE IV: QUASISEPARATED CONDENSED SETS 27
Proof. Let V be any condensed R-vector space, and pick a resolution
M
j
R[S0
j] →
M
i
R[Si] → V → 0.
The left adjoint exists for R[S], with S profinite, and takes the value M(S), essentially by definition
of M-completeness. It follows that the left adjoint for V is given by the quasiseparation of the
cokernel of M
j
M(S0
j) →
M
i
M(Si).

5. Lecture V: Entropy and a real B+
dR
in memoriam Jean-Marc Fontaine
The discussion of the last lecture along with the definition of solid abelian groups begs the
following question:
Question 5.1. Does the category of condensed R-vector spaces V such that for all profinite
sets S with a map f : S → V , there is a unique extension to a map e
f : M(S) → V of condensed
R-vector spaces, form an abelian category stable under all kernels, cokernels, and extensions?
We note that stability under kernels is easy to see. We will see that stability under cokernels
and stability under extensions fails, so the answer is no.
Note that the condition imposed is some form of local convexity. It is a known result in Banach
space theory that there are extensions of Banach spaces that are not themselves Banach spaces,
and in fact not locally convex. Let us recall the construction, due to Ribe, [Rib79].
Let V = `1(N) be the Banach space of `1-sequences of real numbers x0, x1, . . .. We will construct
a non-split extension
0 → R → V 0
→ V → 0.
The construction is based on the following two lemmas.
Lemma 5.2. Let V be a Banach space, let V0 ⊂ V be a dense subvectorspace and let φ : V0 → R
be a function that is almost linear in the sense that for some constant C, we have for all v, w ∈ V0
|φ(v + w) − φ(v) − φ(w)| ≤ C(||v|| + ||w||);
moreover, φ(av) = aφ(v) for all a ∈ R and v ∈ V0. Then one can turn the abstract R-vector space
V 0
0 = V0 × R into a topological vector space by declaring a system of open neighborhoods of 0 to be
{(v, r) | ||v|| + |r − φ(v)| }.
The completion V 0 of V 0
0 defines an extension
0 → R → V 0
→ V → 0
of topological vector spaces (which stays exact as a sequence of condensed vector spaces).
This extension of topological vector spaces is split if and only if there is a linear function
f : V0 → R such that |f(v) − φ(v)| ≤ C0||v|| for all v ∈ V0 and some constant C0.
One can in fact show that all extensions arise in this way.
Proof. All statements are immediately verified. Note that the extension V 0 → V splits as
topological spaces: Before completion we have the nonlinear splitting V0 → V 0
0 : v 7→ (v, φ(v)), and
this extends to completions.
We see that non-split extensions are related to functions that are almost linear locally, but not
almost linear globally. An example is given by entropy. (We note that a relation between entropy
and real analogues of p-adic Hodge-theoretic rings has been first proposed by Connes and Consani,
cf. e.g. [CC15], [Con11]. There is some relation between the constructions of this and the next
lecture, and their work.) Recall that if p1, . . . , pn are real numbers in [0, 1] with sum 1 (considered
as a probability distribution on the finite set {1, . . . , n}), its entropy is
H = −
n
X
i=1
pi log pi.

5. LECTURE V: ENTROPY AND A REAL B+
dR 29
The required local almost linearity comes from the following lemma.
Lemma 5.3. For all real numbers s and t, one has
|s log |s| + t log |t| − (s + t) log |s + t|| ≤ 2 log 2(|s| + |t|).
We note that 0 log 0 := 0 extends the function s 7→ s log |s| continuously to 0.
Proof. Rescaling both s and t by a positive scalar λ multiplies the left-hand side by λ (Check!).
We may thus assume that s, t ∈ [−1, 1] and at least one of them has absolute value equal to 1.
Changing sign and permuting, we assume that t = 1. It then suffices to see that the left-hand side
is bounded by 2 log 2 for all s ∈ [−1, 1]. Note that some bound is now clear, as the left-hand side is
continuous; to get 2 log 2, note that s log |s| and (s+1) log |s+1| take opposite signs for s ∈ [−1, 1],
so it suffices to bound both individually by 2 log 2, which is easy.
Corollary 5.4. Let V0 ⊂ V = `1(N) be the subspace spanned by sequences with finitely many
nonzero terms. The function
H : (x0, x1, . . .) ∈ V0 7→ s log |s| −
X
i≥0
xi log |xi|, where s =
X
i≥0
xi,
is locally almost linear but not globally almost linear, and so defines a nonsplit extension
0 → R → V 0
→ V → 0.
Proof. Local almost linearity follows from the lemma (the scaling invariance H(ax) = aH(x)
uses the addition of s log |s|). For global non-almost linearity, assume that H was close to
P
λixi
for certain λi ∈ R. Looking at the points (0, . . . , 0, 1, 0, . . .), one sees that the λi are bounded (as
H = 0 on such points). On the other hand, looking at ( 1
n, . . . , 1
n , 0, . . .) with n occurences of 1
n,
global almost linearity requires
|H(n) −
1
n
n−1
X
i=0
λi| ≤ C
for some constant C. This would require H(n) to be bounded (as the λi are), but one computes
H(n) = log n.
Now we translate this extension into the condensed picture. Ideally, we would like to show that
there is an extension
0 → M(S) → f
M(S) → M(S) → 0
of Smith spaces, functorial in the profinite set S. Taking S = N ∪ {∞} and writing M(S) =
W1 ⊕ R · [∞] by splitting off ∞, the space W1 is a Smith space containing `1(N) as a subspace
with the same underlying R-vector space (a compact convex generating set of W is the space of
sequences x0, x1, . . . ∈ [−1, 1] with
P
|xi| ≤ 1). Then we get an extension
0 → W → f
W → W → 0
and we can take the pullback along `1 → W and the pushout along W → R (summing all xi)9
to
get an extension
0 → R →? → `1
→ 0.
9Warning: This map is not well-defined, but after pullback to `1
the extension can be reduced to a self-extension
of `1
⊂ W by itself, and there is a well-defined map `1
→ R along which one can the pushout.

This will be the extension constructed above.
We would like to make the following definition.10
Definition 5.5 (does not work). For a finite set S, let
g
R[S]≤c = {(xs, ys) ∈ R[S] × R[S] |
X
s∈S
(|xs| + |ys − xs log |xs||) ≤ c}.
For a profinite set S = lim
←
−i
Si, let
^
M(S) =
[
c0
lim
←
−
i
g
R[S]≤c.
The problem with this definition is that the transition maps in the limit over i do not preserve
the subspaces ≤c. It would be enough if there was some universal constant C such that for any
map of finite set S → T, the set g
R[S]≤c maps into ]
R[T]≤Cc. Unfortunately, even this is not true.
We will see in the next lecture that replacing the `1-norm implicit in the definition of g
R[S]≤c with
the `p-norm for some p 1, this problem disappears.
One can salvage the definition for the Smith space W1 (the direct summand of M(N ∪ {∞}).
In this case, one simply directly builds the extension f
W1, as follows:
f
W1 =
[
c0
{(x0, x1, . . . , y0, y1, . . .) ∈
Y
N
[−c, c] ×
Y
N
[−c, c] |
X
n
(|xn| + |yn − xn log |xn||) ≤ c}.
Proposition 5.6. The condensed set f
W1 has a natural structure of a condensed R-vector space,
and sits in an exact sequence
0 → W1 → f
W1 → W1 → 0.
Proof. Surjectivity of f
W1 → W1 is clear by taking yn = xn log |xn|. To see that it is a
condensed R-vector space, note that stability under addition follows from Lemma 5.3, and one
similarly checks stability under scalar multiplication.
Exercise 5.7. Show that the extension 0 → W1 → f
W1 → W1 → 0 has a Banach analogue: An
extension 0 → `1 → e
`1 → `1 → 0, sitting inside the previous sequence.11
The proposition shows that the class of M-complete condensed R-vector spaces is not stable
under extensions. Indeed, if f
W1 were M-complete, then the map N ∪ {∞} → f
W1 given by sending
n to the element with xn = 1 and all other xi, yi = 0 would extend to a map M(N ∪ {∞}) → f
W1
vanishing on ∞, thus giving a section W1 → f
W1; but the extension is nonsplit.
We can now also show that cokernels are not well-behaved. Indeed, consider also the Smith
space version W∞ of `∞, given by
S
c0
Q
N[−c, c]. There is a natural inclusion W1 ⊂ W∞.
Proposition 5.8. The map of condensed sets f : W1 → W∞ given by
(x0, x1, . . .) 7→ (x0 log |x0|, x1 log |x1|, . . .)
induces, via projection W∞ → W∞/W1, a nonzero map of condensed R-vector spaces
W1 → W∞/W1.
10This was erroneously stated in the lecture.
11See [KP79].

5. LECTURE V: ENTROPY AND A REAL B+
dR 31
Proof. Let us check first that it is a map of condensed abelian groups. For this, we need to
see that two maps (of condensed sets) W1 ×W1 → W∞/W1 agree. Their difference is the projection
along W∞ → W∞/W1 of the map W1 × W1 → W∞ sending pairs of (x0, x1, . . .) and (y0, y1, . . .) to
(x0 log |x0| + y0 log |y0| − (x0 + y0) log |x0 + y0|, x1 log |x1| + y1 log |y1| − (x1 + y1) log |x1 + y1|, . . .),
which lies in W1 by Lemma 5.3. This gives additivity. One checks R-linearity similarly.
To see that it is nonzero, it suffices to see that the image of W1 under f is not contained in the
subspace W1 ⊂ W∞. But the sequence zn = ( 1
n , . . . , 1
n, 0, . . .) for varying n (with n occurences of
1
n ) defines a map N ∪ {∞} → W1, whose image under f is the sequence (−log n
n , . . . , −log n
n , 0, . . .)
which does not have bounded `1-norm.
Using this map, one can recover the extension 0 → W1 → f
W1 → W1 → 0 as the pullback of the
canonical extension 0 → W1 → W∞ → W∞/W1 → 0.
Moreover, this finally answers the question at the beginning of this lecture: The map M(N ∪
{∞}) → W1 → W∞/W1 is zero when restricted to N ∪ {∞}, so the cokernel W∞/W1 of a map of
Smith spaces does not satisfy the property that any map S → W∞/W1 extends uniquely to a map
M(S) → W∞/W1.
As a final topic of this lecture, we construct a kind of “real analogue of B+
dR”. We note that
the extension f
W1 is in fact a flat condensed R[]-module, where R[] = R[t]/t2, via letting map
sequences (x0, x1, . . . , y0, y1, . . .) to (0, 0, . . . , x0, x1, . . .).
One may wonder whether one can build an infinite self-extension of W1’s, leading to a flat
condensed R[[t]]-module W1,R[[t]] with W1,R[[t]] ⊗R[[t]] R = W1. Ideally, one would want to have a
similar self-extension of all M(S), functorial in the profinite set S; as before, this does not actually
work, but will work in an analogous setting as studied in the next lecture.
For the construction, consider the multiplicative map
R → R[[t]]
sending x ∈ R to the series
[x] := x|x|t
= x + x log |x|t + 1
2x log2
|x|t2
+ . . . + 1
n! x logn
|x|tn
+ . . . .
Any element of R[t]/tn can be written in a unique way in the form
[x0] + [x1]t + . . . + [xn−1]tn−1
with xi ∈ R. For n = 2, this writes a + bt in the form [a] + [b − a log |a|]t. For n = 3, the formula
is already rather complicated:
a + bt + ct2
= [a] + [b − a log |a|]t + [c − (b − a log |a|) log |b − a log |a|| − 1
2a log2
|a|]t2
.
Thus much complexity is hidden in this isomorphism.
Lemma 5.9. For any finite set S and real number c 0, let
(R[t]/tn
)[S]≤c = {(
n−1
X
i=0
[xi,s]ti
)s ∈ R[t]/tn
[S] |
X
i,s
|xi,s| ≤ c}.
Then (R[t]/tn)[S]≤c is stable under multiplication by [−, ] for some = (n), and there is a
constant C = C(n) such that
(R[t]/tn
)[S]≤c + (R[t]/tn
)[S]≤c ⊂ (R[t]/tn
)[S]≤Cc.

Proof. Let us explain the case of addition; we leave the case of scalar multiplication as an
exercise. In the given coordinates, there are functions S0, . . . , Sn−1, where Si is a continuous
function of x0, . . . , xi, y0, . . . , yi, such that
[x0]+[x1]t+. . .+[xn−1]tn−1
+[y0]+[y1]t+. . .+[yn−1]tn−1
= [S0]+[S1]t+. . .+[Sn−1]tn−1
∈ R[t]/tn
.
Using that multiplication by [λ] for any λ ∈ R commutes with all operations, one sees that
Si(λx0, . . . , λxi, λy0, . . . , λyi) = λSi(x0, . . . , xi, y0, . . . , yi).
This implies that
Si(x0, . . . , xi, y0, . . . , yi) ≤ Ci(|x0| + . . . + |xi| + |y0| + . . . + |yi|)
by taking Ci 0 to be the maximum absolute value taken by Si on the compact set where all
xj, yj ∈ [−1, 1] and at least one of them is ±1. This estimate easily implies the desired stability
under addition.
Ideally, one would like to use the lemma to see that for any profinite set S = lim
←
−i
Si, one can
define a condensed R[t]/tn-module
M(S, R[t]/tn
) :=
[
c0
lim
←
−
i
(R[t]/tn
)[Si]≤c.
However, as before this construction does not work because the transition maps do not preserve
the desired bounds. However, one can build the corresponding infinite self-extension of W1.12
12We believe this is related to the “complex interpolation of Banach spaces” as studied for example in [CCR+
82],
[CSCK15].

6. LECTURE VI: STATEMENT OF MAIN RESULT 33
6. Lecture VI: Statement of main result
In the last lecture, we saw that a naive adaptation of solid abelian groups to the case of the real
numbers, using the spaces M(S) of signed Radon measures, does not work; but its failure gives
rise to interesting phenomena.
In this lecture, however, we want to find a variant that does work. We want to find a full
subcategory of condensed R-vector spaces that is abelian, stable under kernels, cokernels, and
extensions, and contains all M-complete objects (but is still enforcing a nontrivial condition). This
forces us to include also all the extensions constructed in the previous lecture. These are not M-
complete; in the setting of extensions of Banach spaces, these extensions are not locally convex
anymore.
The following weakening of convexity has been studied in the Banach space literature.
Definition 6.1. For 0 p ≤ 1, a p-Banach space is a topological R-vector space V such that
there exists a p-norm, i.e. a continuous map
|| · || : V → R≥0
with the following properties:
(1) For any v ∈ V , the norm ||v|| = 0 if and only if v = 0;
(2) For all v ∈ V and a ∈ R, one has ||av|| = |a|p||v||;
(3) For all v, w ∈ V , one has ||v + w|| ≤ ||v|| + ||w||;
(4) The sets {v ∈ V | ||v|| } for varying ∈ R0 define a basis of open neighborhoods of 0;
(5) For any sequence v0, v1, . . . ∈ V with ||vi −vj|| → 0 as i, j → ∞, there exists a (necessarily
unique) v ∈ V with ||v − vi|| → 0.
Thus, the only difference is in the scaling behaviour. We note that V is a p-Banach then it is
a p0-Banach for all p0 ≤ p, as if || · || is a p-norm, then || · ||p0/p is a p0-norm. A quasi-Banach space
is a topological R-vector space that is a p-Banach for some p 0.
A key result is the following.
Theorem 6.2 ([Kal81]). Any extension of p-Banach spaces is a p0-Banach for all p0 p.
This suggests that the extension problems we encountered disappear if instead of convexity, we
ask for p-convexity13
for all p 1. Here p-convexity is the assertion that for all a1, . . . , an ∈ R with
|a1|p + . . . + |an|p ≤ 1, the ball {v ∈ V | ||v|| } is stable under (v1, . . . , vn) 7→ a1v1 + . . . + anvn.
Let us do the obvious translation into the condensed world.
Definition 6.3. Let 0 p ≤ 1 be a real number. For any finite set S and real number c 0,
let
R[S]`p≤c = {(as)s ∈ R[S] |
X
s∈S
|as|p
≤ c},
and for a profinite set S = lim
←
−i
Si, let
Mp(S) =
[
c0
lim
←
−
i
R[Si]`p≤c.
13Note that p-convexity for p 1 is weaker than convexity, and might also be called concavity.

For any fixed value of p, we run into the same problem with non-split self-extensions of Mp(S)
as in the last lecture, except that for p 1 things are better-behaved, in that one can define these
self-extensions functorially in the profinite set S: For any finite set S, and n ≥ 1, one can define
(R[T]/Tn
)[S]`p≤c := {(
n−1
X
i=0
[xi,s]Ti
)s |
X
i,s
|xi,s|p
≤ c}.
Then there is a constant C = C(n, p) ∞ such that for any map of finite sets S → S0, the image
of (R[T]/Tn)[S]`p≤c is contained in (R[T]/Tn)[S0]`p≤Cc, by Proposition 6.4 below. In particular,
the subset (R[T]/Tn)[S]`p≤c,stable ⊂ (R[T]/Tn)[S]`p≤c of all elements whose images under any map
S → S0 are still in the `p ≤ c-subspace, is cofinal with the full space, and now is compatible with
transition maps. In other words, for a profinite set S = lim
←
−i
Si, we can now functorially define
Mp(S, R[T]/Tn
) =
[
c0
lim
←
−
i
(R[T]/Tn
)[Si]`p≤c,stable
and thus, in the limit over n, define a B+
dR-theory for any p 1. As we will not have direct use for
it (and, just like Mp itself, it does not define an abelian category of modules), we will not go into
more details about this. We will just note that “the” real numbers are not well-defined anymore:
To define a category of modules over R, one needs to specify in addition 0 p 1; and in some
sense an infinitesimal deformation of p gives rise to B+
dR, so there is in some sense a 1-parameter
family of versions of the real numbers. We will make this picture more precise later.
We used the following proposition:
Proposition 6.4. There is a constant C = C(n, p) ∞ such that for all maps f : S → S0 of
finite sets,
f((R[T]/Tn
)[S]`p≤c) ⊂ (R[T]/Tn
)[S0
]`p≤Cc.
Proof. Decomposing into fibres over S0, we can assume that S0 is a point. Note that mul-
tiplication by [λ] for λ 0 maps the `p ≤ c-subspace isomorphically to the `p ≤ λpc-subspace.
Thus, by rescaling, we can assume that c = 1. It remains to see that there is some bounded subset
B ⊂ R[T]/Tn such that for all finite sets S and all (
Pn−1
i=0 [xi,s]Ti)s ∈ (R[T]/Tn)[S]`p≤1, one has
X
s∈S
n−1
X
i=0
[xi,s]Ti
∈ B ⊂ R[T]/Tn
.
By induction, such a statement holds true for the sum over i = 1, . . . , n − 1. Thus, it suffices to see
that for all integers m = |S| and all x1, . . . , xm with
Pm
j=1 |xj|p ≤ 1, one has
m
X
j=1
[xj] ∈ B0
⊂ R[T]/Tn
for some bounded subset B0. In other words, we need to see that for all i = 0, . . . , n − 1, one has
m
X
j=1
xj logi
|xj| ≤ Ci
for some constant Ci. But for all sufficiently small x, one has logi
|x| ≤ |x|p−1, so neglecting large
xj (which can only give a bounded contribution), one has
Pm
j=1 xj logi
|xj| ≤
Pm
j=1 |xj|p ≤ 1.

Without further ado, let us now state the first main theorem of this course. Define
Mp(S) = lim
−
→
p0p
Mp0 (S).
Theorem 6.5. Fix any 0 p ≤ 1. For a condensed R-vector space V , the following conditions
are equivalent.
(1) For all p0 p and all maps f : S → V from a profinite set S, there is a unique extension
to a map
e
fp0 : Mp0 (S) → V
(2) For all maps f : S → V from a profinite set S, there is a unique extension to a map
e
fp : Mp(S) → V
(3) One can write V as the cokernel of a map
M
i
Mp(Si) →
M
j
Mp(S0
j).
If V satisfies these conditions, we say that V is p-liquid. The class Liqp(R) of p-liquid R-vector
spaces is an abelian subcategory stable under all kernels, cokernels, and extensions. It is generated
by the compact projective objects Mp(S), S extremally disconnected. Moreover:
(1) The inclusion Liqp(R) → Cond(R) has a left adjoint V 7→ V liq
p , and there is a (necessarily
unique) symmetric monoidal tensor product ⊗liq
p on Liqp(R) making V 7→ V liq
p symmetric
monoidal.
(2) The functor D(Liqp(R)) → D(Cond(R)) is fully faithful, and admits a left adjoint, which
is the left derived functor C 7→ CLliq
p of V 7→ V liq
p . There is a unique symmetric monoidal
tensor product ⊗Lliq
p on D(Liqp(R)) making C 7→ CLliq
p symmetric monoidal; it is the left
derived functor of ⊗liq
p .
(3) An object C ∈ D(Cond(R)) lies in D(Liqp(R)) if and only if all Hi(C) ∈ Liqp(R); in that
case, for all profinite sets S and all p0 p, one has
RHomR(Mp0 (S), C) = RHomR(R[S], C).
In other words, one gets an extremely well-behaved theory of p-liquid R-vector spaces. In
the language of last semester, the present theory defines an analytic ring [Sch19, Definition 7.4,
Proposition 7.5]. In fact, by [Sch19, Lemma 5.9, Remark 5.11], all we need to prove is the following
result.
Theorem 6.6. Let 0 p ≤ 1, let
g :
M
i
Mp(Si) →
M
j
Mp(S0
j)
be a map of condensed R-vector spaces, where all Si and S0
j are extremally disconnected, and let V
be the kernel of g. Then for any p0 p and any profinite set S, the map
RHomR(Mp0 (S), V ) → RHomR(R[S], V )

is a quasi-isomorphism in D(Cond(Ab)).
Unfortunately, at this point, we have to pay a serious prize for going into the non-locally convex
setting. More precisely, in order to compute RHomR(Mp0 (S), V ), i.e. the complex of R-linear maps
Mp0 (S) → V , we first have to understand maps of condensed sets Mp0 (S) → V , i.e.
RΓ(Mp0 (S), V )
for V as above. We may restrict to the compact Hausdorff subspace Mp0 (S)≤1. In order for this
to be helpful, we need to be able to compute the cohomology
Hi
(S, V )
for compact Hausdorff S with coefficients in spaces V like above; in particular, for p-Banach spaces.
The proof of [Sch19, Theorem 3.3] gives the following result.
Theorem 6.7. Let V be a Banach space and let S be a compact Hausdorff space. Then
Hi(S, V ) = 0 for i 0. Similarly, if V is a p-Banach for any p ≤ 1 and S is a profinite set,
then Hi(S, V ) = 0 for i 0.
However, a key step in the proof is the existence of a partition of unity, the application of which
uses that V is locally convex critically. In particular, the first part fails if V is only a p-Banach for
p 1.14
This means that for the key computation, we have to further resolve Mp0 (S)≤1 by profinite
sets; and we should do so in a way that keeps the resulting resolution sufficiently explicit. In other
words, the condensed formalism now forces us to resolve the real numbers, and vector spaces over
it, in terms of profinite sets, explicitly!15
We are now ready for the key turn. We will generalize the present question about the real
numbers to an arithmetic ring that is a countable union of profinite subsets.
Let us define this ring. It depends on a real number 0 r 1. We define the condensed ring
Z((T))r whose S-valued points are
Z((T))r(S) = {
X
n∈Z
anTn
| an ∈ C(S, Z),
X
n∈Z
|an|rn
∞} ⊂ Z((T))(S).
More precisely, this is the increasing union of the subsets
Z((T))r,≤c(S) = {
X
n∈Z
anTn
| an ∈ C(S, Z),
X
n∈Z
|an|rn
≤ c}
for varying reals c 0; and here
P
n∈Z |an|rn ≤ c means that for all s ∈ S,
P
n∈Z |an(s)|rn ≤ c.
Proposition 6.8. The condensed set Z((T))r,≤c is a profinite set.
Proof. As multiplication by T is an isomorphism between Z((T))r,≤c and Z((T))r,≤rc, we can
assume that c ≤ 1. In that case Z((T))r,≤c ⊂ Z[[T]] =
Q
n≥0 Z · Tn. Now Z((T))r,≤c can be written
as the inverse limit of
{
m
X
n=0
anTn
|
m
X
n=0
|an|rn
≤ c} ⊂
m
Y
n=0
Z · Tn
,
14We believe that if S is finite-dimensional, one can prove the same result; but the relevant compact Hausdorff
spaces are highly infinite-dimensional.
15This is a difficulty that is not seen in the classical quasi-Banach space literature; and so Theorem 6.2 and its
techniques of proof cannot easily be extended to the present situation.

each of which is a finite set.
For any finite set S, we now write the free module Z((T))r[S] as the increasing union of
Z((T))r[S]≤c := {
X
n∈Z,s∈S
an,sTn
[s] | an,s ∈ C(S, Z),
X
n∈Z,s∈S
|an,s|rn
≤ c};
le each of these is again a profinite set. The addition defines maps
Z((T))r[S]≤c × Z((T))r[S]≤c0 → Z((T))r[S]≤c+c0
and everything is covariantly functorial in S. For any profinite set S = lim
←
−i
Si we define
M(S, Z((T))r)≤c := lim
←
−
i
Z((T))r[Si]≤c
which is a profinite set, functorial in S, equipped with an addition map
M(S, Z((T))r)≤c × M(S, Z((T))r)≤c0 → M(S, Z((T))r)≤c+c0 .
In particular, the colimit
M(S, Z((T))r) =
[
c0
M(S, Z((T))r)≤c
is a condensed abelian group, and in fact one easily sees that it is a Z((T))r-module.
In the next lecture, we will prove the following theorem, relating this theory to the various
`p-theories over R. Part (1) is due to Harbater, [Har84, Lemma 1.5].
Theorem 6.9. Let 0 r0 r and consider the map
θr0 : Z((T))r → R :
X
anTn
7→
X
an(r0
)n
.
This map is surjective. Moreover:
(1) The kernel of θr0 is generated by a nonzerodivisor fr0 ∈ Z((T))r.
(2) For any profinite set S, there is a canonical isomorphism
M(S, Z((T))r)/(fr0 ) ∼
= Mp(S)
of condensed Z((T))r/(fr0 ) = R-modules, where 0 p 1 is chosen so that (r0)p = r.
(3) More generally, for any profinite set S and any n ≥ 1, there is a canonical isomorphism
M(S, Z((T))r)/(fr0 )n ∼
= Mp(S, R[X]/Xn
)
of condensed Z((T))r/(fr0 )n ∼
= R[X]/Xn-modules.
In other words, the theory over Z((T))r specializes to all different Mp-theories for p 1, and
using infinitesimal variations one even recovers the B+
dR-theories for all p. In some sense, one could
regard Z((T))r as some version of Fontaine’s Ainf.
On the other hand, this again indicates that for a fixed r, the theory will not work, so we pass
to a colimit again. Another twist, necessary for the proof, will be to isolate the desired theory
within all condensed Z[T−1]-modules; this is related to the possibility to isolate p-liquid R-vector
spaces inside all condensed abelian groups (i.e., the forgetful functor is fully faithful).
Theorem 6.10. Fix any 0 r 1. For a condensed Z[T−1]-module M, the following conditions
are equivalent.

(1) For all 1 r0 r and all maps f : S → M from a profinite set S, there is a unique
extension to a map
e
fr0 : M(S, Z((T))r0 ) → M
of condensed Z[T−1]-modules.
(2) For all maps f : S → M from a profinite set S, there is a unique extension to a map
e
fr : M(S, Z((T))r) → M
of condensed Z[T−1]-modules.
(3) One can write M as the cokernel of a map
M
i
M(Si, Z((T))r) →
M
j
M(S0
j, Z((T))r).
The class Liqr(Z[T−1]) of such condensed Z[T−1]-modules is an abelian subcategory stable
under all kernels, cokernels, and extensions. It is generated by the compact projective objects
M(S, Z((T))r), S extremally disconnected. Moreover:
(1) The inclusion Liqr(Z[T−1]) → Cond(Z[T−1]) has a left adjoint M 7→ Mliq
r , and there is
a (necessarily unique) symmetric monoidal tensor product ⊗liq
r on Liqr(Z[T−1]) making
M 7→ Mliq
r symmetric monoidal.
(2) The functor D(Liqr(Z[T−1])) → D(Cond(Z[T−1])) is fully faithful, and admits a left ad-
joint, which is the left derived functor C 7→ CLliq
r of M 7→ Mliq
r . There is a unique sym-
metric monoidal tensor product ⊗Lliq
r on D(Liqr(Z[T−1])) making C 7→ CLliq
r symmetric
monoidal; it is the left derived functor of ⊗liq
r .
(3) An object C ∈ D(Cond(Z[T−1])) lies in D(Liqr(Z[T−1])) if and only if all Hi(C) ∈
Liqr(Z[T−1]); in that case, for all profinite sets S and all 1 r0 r, one has
RHomZ[T−1](M(S, Z((T))r0 ), C) = RHomZ[T−1](Z[T−1
][S], C).
It follows formally that Liqr(Z[T−1]) also sits fully faithfully in Cond(Z((T))r) (note that
the tensor unit of Liqr(Z[T−1]) is Z((T))r), and that everything holds true with Z[T−1] replaced
by Z((T))r; we will then also write Liqr(Z((T))r) or simply Liq(Z((T))r) for Liqr(Z[T−1]).
However, the proof will require us to prove the finer statement with base ring Z[T−1].
Again, it reduces formally to the following assertion.
Theorem 6.11. Let K be a condensed Z[T−1]-module that is the kernel of some map
f :
M
i∈I
M(Si, Z((T))r) →
M
j∈J
M(S0
j, Z((T))r)
where all Si and S0
j are extremally disconnected. Then for all 1 r0 r and all profinite sets S,
the map
RHomZ[T−1](M(S, Z((T))r0 ), K) → RHomZ[T−1](Z[T−1
][S], K)
in D(Cond(Ab)) is an isomorphism.
In the next lecture, we will show that Z((T))r is a principal ideal domain, and prove Theo-
rem 6.9. In the appendix to this lecture, we explain how the rest of the assertions of this lecture
reduce to Theorem 6.11.

APPENDIX TO LECTURE VI: RECOLLECTIONS ON ANALYTIC RINGS 39
Appendix to Lecture VI: Recollections on analytic rings
In [Sch19, Lecture VII], we defined analytic rings, abstracting the examples we have seen.
Definition 6.12 ([Sch19, Definition 7.1, 7.4]). A pre-analytic ring is a condensed ring A
together with a functor
{extremally disconnected sets} → Modcond
A : S 7→ A[S]
to the category of A-modules in condensed abelian groups, taking finite disjoint unions to finite
products, together with a natural transformation S → A[S].
An analytic ring is a pre-analytic ring A such that for all index sets I, J and extremally
disconnected sets Si, S0
j, and all maps
f :
M
i
A[Si] →
M
j
A[S0
j]
of condensed A-modules with kernel K, the map
RHomA(A[S], K) → RHomA(A[S], K)
is an isomorphism in D(Cond(Ab)), for all extremally disconnected sets S.
Remark 6.13. The definition of analytic rings was stated in a slightly different, but equivalent
way: the current condition implies the one in [Sch19, Definition 7.4] by writing C as the limit of
its Postnikov truncations; conversely, use the argument of [Sch19, Lemma 5.10].
Thus, Theorem 6.11 implies that A = Z[T−1] with A[S] = M(S, Z((T))r) is an analytic ring.
(It says slightly more as we allow S profinite there, and make a claim about all r0 r; we will
explain what extra information this gives.) This immediately gives a lot of information:
Proposition 6.14 ([Sch19, Proposition 7.5]). Let A be an analytic ring.
(i) The full subcategory
Modcond
A ⊂ Modcond
A
of all A-modules M in condensed abelian groups such that for all extremally disconnected
sets S, the map
HomA(A[S], M) → M(S)
is an isomorphism, is an abelian category stable under all limits, colimits, and extensions.
The objects A[S] for S extremally disconnected form a family of compact projective gen-
erators. It admits a left adjoint
Modcond
A → Modcond
A : M 7→ M ⊗A A
that is the unique colimit-preserving extension of A[S] 7→ A[S]. If A is commutative, there
is a unique symmetric monoidal tensor product ⊗A on Modcond
A making the functor
Modcond
A → Modcond
A : M 7→ M ⊗A A
symmetric monoidal.

(ii) The functor
D(Modcond
A ) → D(Modcond
A )
is fully faithful, and its essential image is stable under all limits and colimits and given by
those C ∈ D(Modcond
A ) such that for all extremally disconnected S, the map
R HomA(A[S], C) → R HomA(A[S], C)
is an isomorphism; in that case, also the RHom’s agree. An object C ∈ D(Modcond
A ) lies
in D(Modcond
A ) if and only if all Hi(C) are in Modcond
A . The inclusion D(Modcond
A ) ⊂
D(Modcond
A ) admits a left adjoint
D(Modcond
A ) → D(Modcond
A ) : C 7→ C ⊗L
A A
that is the left derived functor of M 7→ M ⊗A A. If A is commutative, there is a unique
symmetric monoidal tensor product ⊗L
A on D(Modcond
A ) making the functor
D(Modcond
A ) → D(Modcond
A ) : C 7→ C ⊗L
A A
symmetric monoidal.
This shows that Theorem 6.11 implies most of Theorem 6.10. We note one omission in the
statement of the proposition: It is not claimed that ⊗L
A is the left derived functor of ⊗A. This is
equivalent to the assertion that for all extremally disconnected S, S0, the complex
A[S] ⊗L
A A[T] = A[S × T] ⊗L
A A
sits in degree 0. But in the situation of Theorem 6.10, we defined A[S] for all profinite S, and
Theorem 6.11 ensures that
A[S] ⊗L
A A = A[S]
for all profinite sets S. Indeed, the right hand side lies in Modcond
A , and the R Hom into any complex
C ∈ D(Modcond
A ) agrees, by representing C by a complex of direct sums of projectives, taking the
limit of Postnikov truncations, and using Theorem 6.11.
In Theorem 6.10, we used the strange base ring Z[T−1]. We note that for analytic rings, one
can always change the base ring to its natural choice, and maintain an analytic ring:
Proposition 6.15. Let A be an analytic ring. Then A0 = A[∗] is naturally a condensed ring,
and A[S] is naturally a condensed A0-module for all extremally disconnected S, defining a pre-
analytic ring A0 with a map of pre-analytic rings A → A0. The pre-analytic ring A0 is analytic,
and the forgetful functor D(Modcond
A0 ) → D(Modcond
A ) is an equivalence.
Proof. Note that A[∗] is the tensor unit in Modcond
A (as the image of the tensor unit A[∗]
under the symmetric monoidal functors ⊗AA). One has A[∗] = HomModcond
A
(A[∗], A[∗]), which
is thus naturally a condensed ring, and every object of Modcond
A , in particular A[S], is naturally
a condensed module over it. Now for any K as in the definition of an analytic ring, we have
K ∈ Modcond
A . Then formally
RHomA[∗](A[S], K) = RHomA(A[∗] ⊗L
A A[S], K)
and as K ∈ D(Modcond
A ), we can replace A[∗] ⊗L
A A[S] by
(A[∗] ⊗L
A A[S]) ⊗L
A A.

APPENDIX TO LECTURE VI: RECOLLECTIONS ON ANALYTIC RINGS 41
As ⊗L
AA is symmetric monoidal, and both tensor factors on the left are already in D(Modcond
A ),
this agrees with
A[∗] ⊗L
A A[S] = A[S]
as A[∗] is the tensor unit. But finally
RHomA(A[S], K) = RHomA(A[S], K)
as A is an analytic ring, and the latter is formally the same as
RHomA0 (A0
[S], K).
It follows that A0 is an analytic ring. The present computation then also proves that D(Modcond
A0 ) →
D(Modcond
A ) is an equivalence.
In particular, A = Z((T))r with A[S] = M(S, Z((T))r) defines an analytic ring.
Finally, let us prove Theorem 6.6. Given 0 p0 p ≤ 1, pick x = 1
2, r = xp and r0 = xp0
. Then
we can take fx = 2 − T−1, and Theorem 6.9 (2) says that
M(S, Z((T))r0 ) = Mp0 (S)
while (by passage to a colimit over all r0 r)
M(S, Z((T))r) = Mp(S).
We regard R everywhere as Z[T−1]-algebra via T−1 7→ 2. Given V as in Theorem 6.6, it is automatic
that V ∈ Liqr(Z[T−1]). Using Theorem 6.11, this implies that
RHomZ[T−1](M(S, Z((T))r0 ), V ) = RHomZ[T−1](Z[T−1
][S], V )
for all profinite sets S. The right-hand side agrees with RHomZ(Z[S], V ), while the left-hand side
agrees with RHomZ(Mp0 (S), V ) by taking the quotient by fx = 2 − T−1. Thus, this proves a
version of Theorem 6.11 where we use the base ring Z in place of R.
From this discussion, by taking the colimit over p0 p, one deduces that Z with the modules
Mp(S) defines an analytic ring structure. Using the proposition above, this then also implies
that R with the same modules Mp(S) defines an analytic ring, and that the forgetful functor
D(Liqr(R)) → D(Cond(Ab)) is fully faithful. In particular, we see that a liquid R-vector space
admits a unique R-linear structure.
Finally, the exact statement of Theorem 6.6 follows by observing that
RHomR(Mp0 (S), V ) = RHomZ(R ⊗L
Z Mp0 (S), V )
and the p-liquidification of R ⊗L
Z Mp0 (S) is the same as the p-liquidification of Mp0 (S) (as p-
liquidification is symmetric monoidal and the first tensor factor is the p-liquid tensor unit).

7. Lecture VII: Z((T))r is a principal ideal domain
In this lecture, we prove the following theorem.
Theorem 7.1 ([Har84]). For any real number r with 0 r 1, the ring
Z((T))r = {
X
n−∞
anTn
| ∃r0
r, |an|(r0
)n
→ 0}
is a principal ideal domain. The nonzero prime ideals are the following:
(1) For any nonzero complex number x ∈ C with |x| ≤ r, the kernel of the map
Z((T))r → C :
X
anTn
7→
X
anxn
,
where this kernel depends only on (and determines) x up to complex conjugation;
(2) For any prime number p, the ideal (p);
(3) For any prime number p and any topologically nilpotent unit x ∈ Qp, the kernel of the map
Z((T))r → Qp :
X
anTn
7→
X
anxn
,
where this kernel depends only on (and determines) the Galois orbit of x.
Proof. It is clear that the ideals described are prime ideals. First, we check that all of them
are principal, and that the respective quotient rings are already fields (so that these ideals are also
maximal).
In case (1), assume first that x is real, so 0 x ≤ r or −r ≤ x 0. The second case reduces
to the first under the involution of Z((T))r taking T to −T, so we assume for concreteness that
0 x ≤ r. We start by verifying the surjectivity of
Z((T))r → R :
X
anTn
7→
X
anxn
.
Fix some integer N such that x ≥ 1
N . Then we claim that any element y ∈ R≥0 is represented by
a power series with coefficients in [0, N − 1]. Note that when x = 1
N , this is precisely an N-adic
expansion of y. For the proof, take the maximal n such that xn ≤ y, and the maximal integer an
such that anxn ≤ y. Then necessarily an ≤ N − 1 (otherwise xn+1 ≤ y). Now pass to y − anxn and
repeat, noting that the sequence of n’s is strictly decreasing.
To find a generator, we first find a polynomial gn ∈ 1 + TnR[T] such that inside the disc
{0 |y| ≤ r}, the only zero is x, with multiplicity 1. We argue by induction on n = 1; for n = 1,
we can take g1 = 1 − x−1T. Now write gn = 1 + anTn + . . .. Then, taking any integer m |an|, we
can take
gn+1 = gn(1 − an
m Tn
)m
.
Now choose n large enough so that rn 2(1 − r) and let g = gn. We can find some h =
1 + cnTn + cn+1Tn+1 + . . . ∈ 1 + TnR[[T]] such that all |ci| ≤ 1
2 and f = gh ∈ 1 + TnZ[[T]]. Note
that h is invertible on {y | 0 |y| ≤ r}, as
|
X
i≥n
ciyi
| ≤
X
i≥n
1
2
ri
=
1
2
rn
1 − r
1.

7. LECTURE VII: Z((T))r IS A PRINCIPAL IDEAL DOMAIN 43
We claim that f generates the kernel of evaluation at x. Indeed, if f0 ∈ Z((T))r vanishes at
x, then f0
f ∈ Z((T)) (as f ∈ 1 + TZ[[T]] ⊂ Z[[T]]×) and it still defines a holomorphic function on
{0 |y| r0} for some r0 r; thus, f0
f ∈ Z((T))r.
Still in case (1), if now x is complex, we argue similarly: First, surjectivity of
Z((T))r → C :
X
anTn
7→
X
anxn
can be proved in a similar way, this time allowing coefficients in [−N, N] such that for any element
y ∈ C with |y| ≤ 1 there are integers a0, a1 ∈ [−N, N] with |y−a0 −a1x| ≤ |x|2. To find a generator,
argue as above by finding first a polynomial gn ∈ 1 + TnR[T] whose only zeroes in {0 |y| ≤ r}
are x and its complex conjugate x with multiplicity 1, starting with g1 = (1 − x−1T)(1 − x−1T).
Multiplying by a power series h as before then produces the desired generator f = gh ∈ Z((T))r.
In case (2), it is clear that the ideals are principal. We claim that the map
Z((T))r/p → Fp((T))
is an isomorphism. It is clearly injective. For surjectivity, we use that there is a set-theoretic
section Fp → Z with image {0, 1, . . . , p − 1}, so we can always lift to a power series with bounded
coefficients an. But any power series with bounded coefficients lies in Z((T))r.
It remains to handle case (3). Let K ⊂ Qp be the field generated by x, which is a finite extension
of Qp; let d be its degree. First, we check that the map
fx : Z((T))r → K :
X
anTn
7→
X
anxn
is surjective. More precisely, if we denote by Λ ⊂ OK the Zp-lattice generated by x, we claim that
the map
Z[[T]]r → Λ :
X
anTn
7→
X
anxn
is surjective. For this, observe that if xd = upm for some unit u ∈ OK and integer m 0, then
any element of Λ/xd can be written uniquely as a sum
Pd−1
j=0 ajxj with all aj ∈ {0, . . . , pm − 1}.
It follows that one can reach any element of Λ uniquely by a power series
P
anTn where all
an ∈ {0, . . . , pm − 1}. Indeed, the projection to Λ/xd determines the first d coefficients uniquely,
and then we can divide by Td and induct.
Moreover, representing the element pm ∈ Λ in this way, we get an equation
pm
= fx(
X
n0
anTn
)
where all an ∈ {0, . . . , pm − 1}. Consider the element
gx = pm
−
X
anTn
in the kernel of fx. Note that gx may have additional archimedean zeroes. However, as we already
have generators for them, we can write gx = g0
xg00
x where g0
x has no archimedean zeroes, while
g00
x ∈ 1+TZ[[T]]r. We claim that g0
x generates the kernel of fx. Indeed, it clearly lies in the kernel
of fx; we have to see that if h ∈ Z((T))r is any element in the kernel of fx, then h
g0
x
∈ Q((T)) still
lies in Z((T))r. As g0
x ∈ pm + TZ[[T]], the inverse of g0
x has only powers of p in the denominator,
so h
g0
x
∈ Z[1
p ]((T)). But it also lies in Zp[[T]] as g0
x is a generator of the kernel of Zp[[T]] → OK. It
remains to see that the convergence condition is satisfied, but for this we simply note that g0
x has

no archimedean zeroes, so the quotient still defines a holomorphic function on {0 |y| r0} for
some r0 r.
Finally, we can finish the proof that Z((T))r is a principal ideal domain. Take any nonzero
element f ∈ Z((T))r. We want to see that, up to a unit, it is a product of the generators of the
principal ideals listed in the statement. Up to scaling by a unit, it is of the form
f = a0 +
X
n0
anTn
where a0 0. We can now see that it lies in only finitely many of the principal ideals listed above:
This is a standard result in complex analysis for the first points, and for the second and third type
of points, only primes p dividing a0 are relevant. There are only finitely many such, and for any
p these primes are all contracted from Zp((T)), where the relevant finiteness holds. (We also see
that it lies in any of these maximal ideals with finite multiplicity.) Dividing by the corresponding
generators, we can assume that f does not lie in any of the principal ideals listed in the statement.
This implies that a0 = 1 as otherwise there will be a zero at a p-adic place, by the similar result
for Zp((T)). Then f−1 ∈ 1 + TZ[[T]], and it still has convergence radius r as f has no zeroes on
the closed disc of radius r. Thus, f is invertible, as desired.
Now we can prove Theorem 6.9 from the last lecture.
Given any 0 r0 r 1, we consider the map Z((T))r → R : T 7→ r0. Applying the previous
lemma to x = r0, we get some f = fr0 ∈ Z((T))r that vanishes only at r0. We claim that the
sequence
0 → Z((T))r
fr0
−
−
→ Z((T))r → R → 0
is exact. Surjectivity was proved above (even Z((T))r surjects). On the other hand, if g ∈ Z((T))r
vanishes at x, then gf−1
r0 ∈ Z((T)) and it follows from a standard complex analysis consideration
that it still lies in Z((T))r.
We want to see that the sequence is still exact as condensed abelian groups, and we want to
identify the spaces of measures. For this, we prove the following quantitative version:
Proposition 7.2. There are constants C1, . . . , C4 with the following properties:
(1) If g ∈ Z((T))r,≤c then fr0 g ∈ Z((T))r,≤C1c.
(2) Conversely, if g ∈ Z((T))r such that fr0 g ∈ Z((T))r,≤c, then g ∈ Z((T))r,≤C2c.
(3) Let 0 p 1 such that (r0)p = r. If g ∈ Z((T))r,≤c, then g(x) ∈ R`p≤C3c.
(4) Conversely, for any z ∈ R`p≤c, there is some g ∈ Z((T))r,≤C4c such that g(x) = z.
The proposition easily implies exactness as condensed abelian groups (think in terms of ind-
(compact Hausdorff) sets, and use that for maps of compact Hausdorff spaces, surjectivity can be
checked on points).
Proof. Take C1 so that fr0 ∈ Z((T))r,≤C1 , then the claim about C1 is clear. For the claim
about C2, note that the norm can be expressed in terms of the restriction to a function on {y |
|r00| ≤ |y| ≤ |r|} for r0 r00 r. On that strip, fr0 is invertible, so picking a bound of the norm for
the inverse gives the claim.
For C3, let g =
P
anTn with
P
|an|rn ≤ c. Then
|g(x)|p
≤
X
|an|p
(r0
)np
=
X
|an|p
rn
≤
X
|an|rn
≤ c,

7. LECTURE VII: Z((T))r IS A PRINCIPAL IDEAL DOMAIN 45
so we can in fact take C3 = 1. Here, we used critically that all an are integers, which implies that
|an|p ≤ |an|: This is true whenever an = 0 or |an| ≥ 1 (when p 1).
For C4, take any z ∈ R with |z|p ≤ c and assume z 6= 0. Take n minimal so that (r0)n ≤ |z|,
and write z = an(r0)n + z0 where an ∈ Z and |z0| (r0)n. Note that |an| ≤ (r0)−1. Continuing, we
can write z = g(x) for some g = anTn + . . . ∈ Z((T)) with |ai| ≤ (r0)−1 for all i. Then
X
|ai|ri
≤ (r0
)−1 rn
1 − r
≤
1
r0(1 − r)
|z|p
,
so we can take C4 = 1
r0(1−r) .
From here, by observing that these bounds immediately generalize to finite free modules, one
easily deduces Theorem 6.9 (2). For part (3), we need a corresponding version. Note first that
Z((T))r/(fr0 )m ∼
= R[X]/Xm
where the isomorphism sends T to [r0] = (r0)1+X = r0 + r0 log r0X + . . .. In other words, we have a
short exact sequence
0 → Z((T))r
fm
r0
−
−
→ Z((T))r
θm
−
−
→ R[X]/Xm
→ 0.
Proposition 7.3. There are constants C1, . . . , C4 (depending on m) with the following prop-
erties:
(1) If g ∈ Z((T))r,≤c then fm
r0 g ∈ Z((T))r,≤C1c.
(2) Conversely, if g ∈ Z((T))r such that fm
r0 g ∈ Z((T))r,≤c, then g ∈ Z((T))r,≤C2c.
(3) If g ∈ Z((T))r,≤c, then θm(g) ∈ (R[X]/Xm)`p≤C3c.
(4) Conversely, for any z ∈ (R[X]/Xm)`p≤c, there is some g ∈ Z((T))r,≤C4c such that θm(g) =
z.
Proof. Parts (1) and (2) follows inductively from parts (1) and (2) of the previous proposition.
It remains to handle part (3): Part (4) then follows by successive approximation (using parts (3)
and (4) of the previous proposition, and the other claims already established).
Thus let g =
P
anTn ∈ Z((T))r with
P
|an|rn ≤ c. Then
θm(g) =
X
an[r0
]n
∈ R[X]/Xm
.
We need to bound this. Recall from Proposition 6.4 that there is some constant C such that for
all maps f : S → S0 of finite sets, one has
f((R[X]/Xm
)[S]`p≤c) ⊂ (R[X]/Xm
)[S0
]`p≤Cc.
We claim that we can take C3 = C; equivalently,
X
an[r0
]n
∈ (R[X]/Xm
)`p≤Cc.
To check this, we can assume that the sum is finite (as the target is closed). Let S be a finite set of
cardinality
P
|an|. Then we can define an element e
z ∈ (R[X]/Xm)[S] whose coefficients are of the
form ±[r0]n = [±r0]n, such that the image of e
z in R[X]/Xm (the sum of the coefficients) is
P
an[r0]n.
Thus, it suffices to see that e
z ∈ (R[X]/Xm)[S]`p≤c. But this is clear: One has to bound the sum of
(r0)np = rn, where there are |an| occurences of rn, giving in total at most
P
|an|rn ≤ c.

The lecturer finds the way in which the varying `p-norms, as well as the subtle B+
dR-type
deformations, arise from the very simple-minded spaces of measures for Z((T))r quite striking. The
discreteness of Z is critical for this behaviour: In some sense, the real B+
dR arises via some kind of
discretization.

8. LECTURE VIII: REDUCTION TO “BANACH SPACES” 47
8. Lecture VIII: Reduction to “Banach spaces”
Fix some 0 r 1. Recall that we want to prove Theorem 6.11:
Theorem 8.1. Let K be a condensed Z[T−1]-module that is the kernel of some map
f :
M
i∈I
M(Si, Z((T))r) →
M
j∈J
M(S0
j, Z((T))r)
where all Si and S0
j are extremally disconnected. Then for all 1 r0 r and all profinite sets S,
the map
RHomZ[T−1](M(S, Z((T))r0 ), K) → RHomZ[T−1](Z[T−1
][S], K)
Today, we make some preliminary reductions, and in particular reduce to the case where K is
a suitable analogue of a Banach space. First, we note the following simple proposition.
Proposition 8.2. Let S and S0 be two profinite sets and let f : S → M(S0, Z((T))r) be a map.
Then for any r0 ≥ r less than 1, there is a unique Z[T−1]-linear map
M(S, Z((T))r0 ) → M(S0
, Z((T))r)
extending the given map f.
Proof. First, we prove uniqueness. Thus assume that some map g : M(S, Z((T))r0 ) →
M(S0, Z((T))r) has trivial restriction to S. By Z[T−1]-linearity, it vanishes on Z[T±1][S], but
this is dense in M(S, Z((T))r0 ) (and everything is quasiseparated). Thus, the map g vanishes.
For existence, we may assume that r0 = r, and by rescaling that f maps S into M(S0, Z((T))r)≤1.
We want to show that there is a map of profinite sets
M(S, Z((T))r)≤1 → M(S0
, Z((T))r)≤1
extending the given map on the dense subset Z[T±1][S]. Writing S0 as an inverse limit of finite
sets, this reduces to the case that S0 is finite. Then
M(S0
, Z((T))r)≤1 ⊂
Y
n≥0,s0∈S0
Z · Tn
[s0
] = Z[[T]][S0
].
Moreover, it is defined as an inverse limit of finite subsets of (Z[T]/Tm)[S0], so we can work modulo
Tm for some m. Then the map S → M(S0, Z((T))r)≤1 → (Z[T]/Tm)[S0] factors through some
finite quotient S → Sm, and we can also reduce to the case that S is finite. In that case, the desired
map
M(S, Z((T))r) → M(S0
, Z((T))r)
clearly exists, and is given by
P
s∈S as[s] 7→
P
s∈S asf(s). One verifies that this indeed maps
M(S, Z((T))r)≤1 into M(S0, Z((T))r)≤1.
The following proposition is easy to prove by using that we choose exactly Z[T−1] as the base
ring. (It is also true that Mp(S) is a pseudocoherent R-module, but this is somewhat more tricky
to prove.)
Proposition 8.3. For any profinite set S, the condensed Z[T−1]-module M(S, Z((T))r) is
pseudocoherent.

Proof. Let M = M(S, Z((T))r). We have the Breen-Deligne resolution
. . . →
ni
M
j=1
Z[Mrij
] → . . . → Z[M2
] → Z[M] → M → 0.
Here ni and all rij are nonnegative integers. Note that M has an endomorphism (multiplication by
T−1), and by functoriality in M, the whole sequence is a sequence of condensed Z[T−1]-modules.
Thus, it is enough to see that each Z[Mr] is pseudocoherent as Z[T−1]-module. Replacing S by
r copies of itself, we can assume that r = 1. Now M has a profinite subset M≤1 ⊂ M and
M =
S
n≥0 T−nM≤1. In fact, we claim that the sequence
0 → Z[M≤r][T−1
]
T−1−[T−1]
−
−
−
−
−
−
−
→ Z[M≤1][T−1
] → Z[M] → 0
is an exact sequence of Z[T−1]-modules. This will give the desired result as the other two terms
are pseudocoherent Z[T−1]-modules. But the sequence is the filtered colimit of the sequences
0 →
n−1
M
i=0
Z[M≤r] · T−i T−1−[T−1]
−
−
−
−
−
−
−
→
n
M
i=0
Z[M≤1] · T−i
→ Z[T−n
M≤1] → 0
which are exact by induction on n: For n = 0, the first term is zero and the second map is
an equality, and for n 0 the quotient by the sequence for n − 1 (multiplied by T−1) is an
isomorphism Z[M≤r] ∼
= Z[M≤1] between the terms of degree 0 (given by multiplication by T−1 :
M≤r
∼
= M≤1).
It follows that RHomZ[T−1](M(S, Z((T))r0 ), K) commutes with filtered colimits in K. In par-
ticular, we can assume that K is the kernel of some map
M(S1, Z((T))r) →
M
j∈J
M(S0
j, Z((T))r).
Note that by Proposition 8.2 and Proposition 8.3, for any r0 r one has
Hom(M(S1, Z((T))r0 ),
M
j∈J
M(S0
j, Z((T))r)) =
M
j∈J
M(S0
j, Z((T))r)(S1).
Passing to the limit r0 → r, this shows that also
Hom(M(S1, Z((T))r),
M
j∈J
M(S0
j, Z((T))r)) =
M
j∈J
M(S0
j, Z((T))r)(S1).
In particular, the given map factors over a finite direct sum, and we may assume that K is the
kernel of some map
M(S1, Z((T))r) → M(S2, Z((T))r),
and we recall that any such map is uniquely induced by a map S1 → M(S2, Z((T))r). In fact, by
a triangle (and noting that M(S1, Z((T))r) itself is a special case of such an image, when taking
the identity map), it is enough to prove the similar assertion for the image of the map
M(S1, Z((T))r) → M(S2, Z((T))r).
In other words, we are reduced to the following assertion. (We could assume that S1 and S2 are
extremally disconnected instead of merely profinite, but this will be of no use.)

Theorem 8.4. Let S1 and S2 be profinite sets and let f : M(S1, Z((T))r) → M(S2, Z((T))r)
be a map of condensed Z[T−1]-modules; these are in bijection with maps f|S1 : S1 → M(S2, Z((T))r).
Let M be the image of f. Then for any r0 r, the map
RHomZ[T−1](M(S, Z((T))r0 ), M) → RHomZ[T−1](Z[T−1
][S], M)
We wish to reduce further. The datum is the map S1 → M(S2, Z((T))r). This factors
over M(S2, Z((T))r0 ) for some r0 r, and fixing such a factorization, we get modules Mr00
for all r0 ≥ r00 r as the image of M(S1, Z((T))r00 ) → M(S2, Z((T))r00 ), and M is the filtered
colimit of the Mr00 . Note that Mr00 can be endowed with subspaces Mr00,≤c by taking the image of
M(S1, Z((T))r00 )≤c. These are the inverse limits of the images Mr00,i,≤c in M(S2,i, Z((T))r00 ), where
we write S2 as an inverse limit of finite sets S2,i. We need the following proposition.
One can actually identify the module Mr00,i that appears, by applying the following proposition
with r00 in place of r.
Proposition 8.5. Let S1 be a profinite set and let S1 → Z((T))n
r be a map for some integer
n. Then the image M of
M(S1, Z((T))r) → Z((T))n
r
is isomorphic (as condensed Z[T−1]-module) to Z((T))n0
r for some n0 ≤ n.
Proof. By Theorem 7.1, the image of
M(S1, Z((T))r) → Z((T))n
r
on underlying modules is isomorphic to Z((T))n0
r for some n0 ≤ n. Let M0 = Z((T))n0
r as a
condensed Z[T−1]-module, which comes equipped with a map to Z((T))n
r. This map is injective
as both are quasiseparated as condensed sets and the underlying map is injective. Moreover, the
quotient of condensed Z[T−1]-modules
Z((T))n
r/M0
is still quasiseparated, by the structure of finitely generated Z((T))r-modules (using Theorem 7.1).
It follows that the map S1 → Z((T))n
r factors over M0, and replacing Z((T))n
r by M0, we may
assume that M(S1, Z((T))r) → Z((T))n
r is surjective (on underlying modules). In particular, the
image contains a basis, which implies that also
M(S, Z((T))r) → Z((T))n
r
is surjective as condensed modules, giving the desired result.
The next proposition expresses Mr00 in terms of the Mr00,i.
Proposition 8.6. For any r00, the complex
0 → Mr00 →
[
c0
Y
i
Mr00,i,≤c →
[
c0
Y
i0→i1
Mr00,i1,≤c → . . . →
[
c0
Y
i0→...→im
Mr00,im,≤c → . . .
of condensed Z[T−1]-modules is exact. More precisely, taking E-valued points for some extremally
disconnected set E, it is exact, and if
g ∈ ker(
Y
i0→...→im
Mr00,im,≤c(E) →
Y
i0→...→im+1
Mr00,im+1 (E))

then there exists some h ∈
Q
i0→...→im−1
Mr00,im−1,≤c(E) with d(h) = g.
We refer to the appendix for the construction of the complex.
Proof. The claim is equivalent to the surjectivity of some map of profinite sets, more precisely
that of Y
i0→...→im−1
Mr00,im−1,≤c ×Q
i0→...→im
Mr00,im
Y
i0→...→im
Mr00,im,≤c
→ ker(
Y
i0→...→im
Mr00,im,≤c →
Y
i0→...→im+1
Mr00,im+1 ).
We can write the index category I of i’s as a filtered union of finite categories Ij that admit an
initial object. The corresponding map is a cofiltered limit of similar maps for Ij in place of I.
Therefore it suffices to prove the similar assertion for each Ij. In that case, the presence of an
initial object gives a contracting homotopy, whose explicit description as given in the appendix to
this lecture shows that it preserves the ≤ c-subspace.
In particular, it is enough to prove the following statement.
Theorem 8.7. Fix radii 1 r0 r0 r 0. Let S1 be a profinite set and let fi : S1 →
Z((T))ni
r0
be maps for some index set i ∈ I. For r0 ≥ r00 r, let Mr00,i ⊂ Z((T))ni
r00 be the image of
the induced maps M(S1, Z((T))r00 ) → Z((T))ni
r00 , with the subspace Mr00,i,≤c ⊂ Mr00,i defined as the
image of M(S1, Z((T))r00 )≤c.
Then, letting
M = lim
−
→
r00r
[
c0
Y
i
Mr00,i,≤c,
the map
][S], M)
is an isomorphism in D(Cond(Ab)) for all profinite sets S.
At this point, we will make a change from Smith to Banach spaces: The Mr00,i are some kind
of Smith spaces, but their norm allows one to define canonical Banach spaces inside them.
Definition 8.8. Let N be a condensed abelian group written as an increasing union of compact
Hausdorff subsets N≤c for c ≥ 0, satisfying the following conditions:
(1) One has N≤0 = 0;
(2) Each N≤c is symmetric, i.e. −N≤c = N≤c;
(3) It is exhaustive: N =
S
c0 N≤c;
(4) It satisfies the triangle inequality: N≤c + N≤c0 ⊂ N≤c+c0 ;
(5) It is continuous: N≤c =
T
c0c N≤c0 .
In that situation, we let NB be the condensed abelian group which takes any profinite set S to the
completion of the normed abelian group of locally constant maps from S to N.
In other words, as discussed in the appendix, NB is the condensed abelian group associated to
the normed abstract abelian group N(∗).
Proposition 8.9. There is a unique map NB → N of condensed sets extending the identity
on underlying abelian groups. It is an injective map of condensed abelian groups.

Proof. Given a map f : S → NB, choose a sequence of locally constant maps f0, f1, . . . : S →
N with limit f (by definition NB(S) is the space of Cauchy sequences modulo the space of null
sequences). This defines a map S × N → N, and it is easy to see that it takes image in N≤c for
some c. Let Γ ⊂ (S ×N)×N≤c be its graph, and let Γ ⊂ (S ×(N∪{∞}))×N≤c be the closure. We
claim that Γ → S ×(N∪{∞}) is an isomorphism. As it is a map of compact Hausdorff spaces, it is
enough to check bijectivity, so pick some s ∈ S. We need to see that there is a unique preimage of
(s, ∞). Note that existence is clear, as Γ has dense image, and thus full image. On the other hand,
uniqueness is a simple consequence of the Cauchy property and the separatedness 0 =
T
c0 N≤c
(which follows from condition (1) and (5)).
In particular, restricting to ∞, we get a map S → N≤c, thus defining a map S → N. It is
clear that this is functorial, additive, and independent of the choice of the fi (as nullsequences will
converge to zero), and defines a map of condensed sets NB → N (and is the only possible choice).
The rest of the assertions are also easy to verify.
In our situation, the following proposition ensures that replacing Mr00,i by MB
r00,i is harmless.
Proposition 8.10. For any r0 ≥ r00 r000 r, the map Mr00,i → Mr000,i factors over MB
r000,i.
Proof. By Proposition 8.5, one has Mr00,i
∼
= Z((T))
n0
i
r00 for some n0
i, and then
Mr000,i = Mr00,i ⊗Z((T))r00 Z((T))r000 ∼
= Z((T))
n0
i
r000 .
Thus, it suffices to see that Z((T))r00 → Z((T))r000 factors over Z((T))B
r000 . It suffices to check this
on Z((T))r00,≤1, and then it follows from the observation that Z((T))r00,≤1 ∩ TmZ[[T]] maps into
Z((T))r000,≤( r000
r00 )m , so writing any S-valued section of Z((T))r00,≤1 as
P
n≥0 anTn with an ∈ C(S, Z),
this sum is actually convergent in Z((T))B
r000 .
This implies that
lim
−
→
r00r
[
c0
Y
i
Mr00,i,≤c = lim
−
→
r00r
[
c0
Y
i
(MB
r00,i,≤c).
After this change, we can actually prove a statement for an individual r00 (which gives the previous
statement by passage to the filtered colimit over all r00 r, using Proposition 8.3 again). We are
reduced to the following statement (where the r in the present statement is any of the r00 in the
previous statement).
Theorem 8.11. Fix radii 1 r0 r 0. Let S1 be a profinite set and let fi : S1 →
Z((T))ni
r be maps for some index set i ∈ I. Let Mi ⊂ Z((T))ni
r be the image of the induced maps
M(S1, Z((T))r) → Z((T))ni
r , with the subspace Mi,≤c ⊂ Mi defined as the image of M(S1, Z((T))r)≤c.
Then, letting
M =
[
c0
Y
i
(MB
i,≤c),
the map
][S], M)
is an isomorphism in D(Cond(Ab)) for all profinite sets S.

Note that we are doing something slightly funny here as we are using the procedure N 7→ NB
inside the product over i: While
S
c0
Q
i Mi,≤c is of Smith-type (countable union of compact
Hausdorff), the present construction is a strange mixture.
We will now get rid of the product over i by formulating a quantitative version of the statement
for a single i. To show that the internal Hom agrees, we have to see that for any profinite set S0,
one has
R HomZ[T−1](M(S, Z((T))r0 )[S0
]/Z[T−1
][S × S0
], M) = 0.
The source is a pseudocoherent Z[T−1]-module, so we may pick a projective resolution P• =
P•(r0, S, S0) of M(S, Z((T))r0 )[S0]/Z[T−1][S × S0], where each Pi = Z[T−1][Ei] is the free con-
densed Z[T−1]-module on some extremally disconnected set Ei. Then we have to see that the
corresponding complex
0 → M(E0) → M(E1) → . . .
is acyclic.
Now we observe that M is naturally the union of the subsets M≤c =
Q
i(MB
i,≤c) over all c 0.
Using this observation, it is enough to prove the following quantitative result.
Theorem 8.12. Fix profinite sets S and S0 and radii 1 r0 r 0, as well as a projective
resolution P• of M(S, Z((T))r0 )[S0]/Z[T−1][S×S0], where each Pi = Z[T−1][Ei] is the free condensed
Z[T−1]-module on some extremally disconnected set Ei.
Then for any m there is some Cm 0 with the following property. For any profinite set S1 with
a map f : S1 → Z((T))n
r, letting M be the image of the induced map M(S1, Z((T))r) → Z((T))n
r
with its subspaces M≤c, the complex
0 → MB
(E0) → MB
(E1) → . . .
computing R Hom(M(S, Z((T))r0 )[S0]/Z[T−1][S × S0], M) is exact, and if g ∈ ker(MB(Em) →
MB(Em+1)) with ||g|| ≤ c, then there is some h ∈ MB(Em−1) with ||h|| ≤ Cmc such that d(h) = g.
Now MB with its norm is an example of the following structure:
Definition 8.13. An r-normed Z[T±1]-module is a normed Z[T±1]-module V satisfying ||Tv|| =
r||v|| for all v ∈ V .
To an r-normed Z[T±1]-module V , one can associate a condensed Z[T±1]-module b
V , as in the
appendix. One main reason that we reduced to this “Banach space” setting is that one can use
resolutions by profinite sets, instead of extremally disconnected sets, by Proposition 8.19. We end
this lecture by reducing to the following statement.
Theorem 8.14. Fix radii 1 r0 r 0. Then for all r-normed Z[T±1]-modules V and all
profinite sets S, the map
R HomZ[T−1](M(S, Z((T))r0 ), b
V ) → b
V (S)
is a quasi-isomorphism.
Theorem 8.14 implies Theorem 8.12. Fix the profinite set S and take a projective resolu-
tion P• of M(S, Z((T))r0 )/Z[T−1][S], so that each Pi = Z[T−1][Ei] is the free condensed Z[T−1]-
module on an extremally disconnected set Ei.
As RΓ(S, b
V ) = b
V (S) is concentrated in degree 0, Theorem 8.14 implies that the complex
0 → b
V (E0) → b
V (E1) → . . .

computing R HomZ[T−1](M(S, Z((T))r0 )/Z[T−1][S], V ), is acyclic. We claim that this implies that
for all m ≥ 0, there is some constant Cm (independent of V ) such that for all g ∈ ker(b
V (Em) →
b
V (Em+1)) with ||g|| ≤ c, there is some h ∈ b
V (Em−1) with ||h|| ≤ Cmc such that d(h) = g.
Indeed, assume the contrary. Then we can find a sequence V1, V2, . . . of r-normed Z[T±1]-
modules, equipped with norms || · ||i on Vi, together with elements gi ∈ ker(b
Vi(Em) → b
Vi(Em+1))
with ||gi|| ≤ c for some fixed c, such that for each i = 1, 2, . . . there is no h ∈ Vi(Em−1) with
||hi|| ≤ r−2ic such that d(hi) = gi.
Let V be the direct sum V1 ⊕ V2 ⊕ . . . equipped with the supremum norm; this is again an
r-normed Z[T±1]-module. Moreover, g = (Tg1, T2g2, . . .) defines an element of b
V (Em). Using the
exactness of the complex for V , we get some h ∈ V with d(h) = g. Then there is some i such
that ||h|| ≤ r−ic, and then the image T−ihi ∈ Vi of h satisfies ||hi|| ≤ r−2ic and d(hi) = gi, which
contradicts our assumption.
Now, we also see that for all profinite sets S0, the complex
0 → b
V (E0 × S0
) → b
V (E1 × S0
) → . . .
is exact, and for all m ≥ 0, there is some constant Cm (independent of V ) such that for all
g ∈ ker(b
V (Em×S0) → b
V (Em+1×S0)) with ||g|| ≤ c, there is some h ∈ b
V (Em−1×S0) with ||h|| ≤ Cmc
such that d(h) = g. Indeed, this complex is obtained by completing the complex of locally constant
maps from S0 into the previous one, which gives the desired bounds by Proposition 8.17.
But this complex computes R HomZ[T−1](M(S, Z((T))r0 )[S0]/Z[T−1][S × S0], V ), via the resolu-
tion P•[S0]. This is not quite a resolution by extremally disconnected sets, but one can refine it by
one (resolving each term), and use Proposition 8.19. Applying this to V = MB with its norm, we
get the result.
Remark 8.15. Reductions very similar to the ones in this lecture could be done directly in
the setting of liquid R-vector spaces, reducing Theorem 6.6 to the following statement: For any
0 p0 p ≤ 1 and any p-Banach space V and profinite set S, the map
R HomR(Mp0 (S), V ) → V (S)
is a quasi-isomorphism. Equivalently, for all i 0,
Exti
R(Mp0 (S), V ) = 0.
It is in the proof of this statement that one runs into the issue pointed out after Theorem 6.7.

Appendix to Lecture VIII: Completions of normed abelian groups
Definition 8.16.
(1) A normed abelian group is an abelian group M equipped with a map
|| · || : M → R≥0
satisfying ||0|| = 0, || − m|| = ||m|| and ||m + n|| ≤ ||m|| + ||n|| for all m, n ∈ M.
(2) A normed abelian group is separated if ||m|| = 0 implies m = 0.
(3) A normed abelian group is complete if it is separated and for every Cauchy sequence
(m0, m1, . . .) (i.e., ||mi − mj|| → 0 as i, j → ∞) there is some m ∈ M with ||m − mi|| → 0
as i → ∞.
As is well-known, the inclusions of complete into separated into all normed abelian groups
admit left adjoints: Separation takes M to the quotient of M by {m ∈ M | ||m|| = 0}; and
completion takes M to the quotient of the space of Cauchy sequences (m0, m1, . . .) (||mi −mj|| → 0
as i, j → ∞) by the space of nullsequences (m0, m1, . . .) (||mi|| → 0 as i → ∞), with the norm
defined by declaring ||(m0, m1, . . .)|| to be the limit of ||mi|| as i → ∞.
We often use the following exactness property:
Proposition 8.17. Let M0
d0
−
→ M1
d1
−
→ M2
d2
−
→ M3 be a four-term complex of bounded maps
of normed abelian groups. Assume that, for some positive constants C and D, for all y ∈ ker(d1 :
M1 → M2) there is some x ∈ M0 with d0(x) = y and ||x|| ≤ C||y||, and similarly for all z ∈ ker(d2 :
M2 → M3), there is some y ∈ M1 with d1(y) = z and ||y|| ≤ D||z||.
Then c
M0
b
d0
−
→ c
M1
b
d1
−
→ c
M2
b
d2
−
→ c
M3 is a complex, and for all b
y ∈ c
M1 and all 0 there is some
b
x ∈ c
M0 with b
d1(b
x) = b
y and ||b
x|| ≤ (C + )||b
y||.
Proof. First, we claim that ker(d1 : M1 → M2) is dense in ker(b
d1 : c
M1 → c
M2). Pick any
b
y ∈ ker(b
d1) and δ 0 and take y ∈ M1 such that ||b
y − y|| ≤ δ. Let z = d1(y) ∈ M2, which has
norm ||z|| = ||d1(y)|| = ||d1(y − b
y)|| bounded by Cd1 δ, where Cd1 is the norm of d1. We can thus
find some y0 ∈ M1 with ||y0|| ≤ DCd1 δ and d1(y0) = z. Replacing y by y − y0, we can thus find
y ∈ ker(d1 : M1 → M2) such that still ||b
y − y|| ≤ (1 + DCd1 )δ; as δ was arbitrary, this gives the
desired density.
This implies that one can write b
y as a sum y0 +y1 +. . . with yi ∈ ker(d1) and ||yi|| ≤ i for i 0
for any given sequence of positive numbers 1 ≥ 2 ≥ . . .. Indeed, we can inductively choose the yi
so that ||b
y − y0 − . . . − yi|| ≤ 1
2i+1, in which case ||yi|| ≤ 1
2(i + i+1) ≤ i. Taking the sequence of
i’s sufficiently small so that
P
i0 i ≤ ||b
y||
2C , we can lift all yi to xi with ||xi|| ≤ C||yi||, and then
b
x = x0 + x1 + . . . maps to b
y and satisfies
||b
x|| ≤ ||x0|| + C
X
i0
i ≤ C||y0|| + C
X
i0
i ≤ C||b
y|| + 2C
X
i0
i ≤ (C + )||b
y||.
Definition 8.18. Let M be a normed abelian group. Let c
M be the condensed abelian group
taking any profinite S to the completion of the normed abelian group of locally constant maps from
S to M (equipped with the supremum norm).
We note that implicit here is that this actually is a condensed abelian group.

APPENDIX TO LECTURE VIII: COMPLETIONS OF NORMED ABELIAN GROUPS 55
Proposition 8.19. The condensed abelian group c
M is canonically identified with the condensed
abelian group associated to the topological abelian group c
Mtop given by the completion of M equipped
with the topology induced by the norm. The norm defines a natural map of condensed sets
|| · || : c
M → R≥0.
Moreover, for any hypercover S• → S of a profinite set S by profinite sets Si, the complex
0 → c
M(S) → c
M(S0) → c
M(S1) → . . .
is exact, and whenever f ∈ ker(c
M(Sm) → c
M(Sm+1)) with ||f|| ≤ c, then for any 0 there is
some g ∈ c
M(Sm−1) with ||g|| ≤ (1 + )c such that d(g) = f.
Proof. For the final assertion, follow the proof of [Sch19, Theorem 3.3]: When S and all Si
are finite, the hypercover splits, so a contracting homotopy gives the result with constant 1. In
general, write the hypercover as a cofiltered limit of hypercovers of finite sets by finite sets, pass to
the filtered colimit, and complete, using Proposition 8.17.
For the identification with the condensed abelian group associated to the topological abelian
group c
Mtop, note that in the supremum norm any continuous function from S to c
Mtop can be
approximated by locally constant functions arbitrarily well, and that the space of continuous func-
tions from S to c
Mtop is complete with respect to the supremum norm. That || · || defines a map of
condensed sets c
M → R≥0 follows for example from this identification with c
Mtop, as the norm is by
definition a continuous map c
Mtop → R≥0.

Appendix to Lecture VIII: Derived inverse limits
We recall the following construction of derived inverse limits. Let I be some index category,
and let F : I → A be a functor to some abelian category A that admits all colimits and compact
projective generators (in particular, infinite products are exact in A).
We build a cosimplicial object L•
F of A whose n-th term is
Ln
F =
Y
i0→...→in
F(in)
where the product is over all chains of n composable morphisms i0 → i1 → . . . → in. We note that
the index set of the product is equivalently the set of maps ∆n
cat → I, where ∆n
cat is the category
associated to the ordered set {0, . . . , n}. Moreover, F(in) can be understood as the colimit of F
restricted to ∆n
cat. With this description, it is clear that a map of simplices ∆n → ∆m induces a
map Ln
F → Lm
F .
Before going on, let us analyze the case that I has an initial object i ∈ I. In that case, L•
F first
of all extends to an augmented cosimplicial object F(i) → L•
F , and this augmented cosimplicial
object has terms Y
i=i−1→i0→...→in
F(in)
for n = −1, 0, . . ., which are even functorial in maps ∆n+1 → ∆m+1 preserving the initial vertex.
Formally (cf. e.g. (the dual of) [Lur09, Lemma 6.1.3.16]), this implies that the complex
0 → F(i) → L0
F → L1
F → . . .
associated to the augmented cosimplicial object is acyclic. More concretely, this is a direct verifi-
cation: the contracting homotopy is given by the map L0
F → F(i) taking the component at i, and
the maps Ln
F → Ln−1
F given by restricting to components with i = i0.
The following proposition is known as the “Bousfield–Kan formula”, cf. [BK72, Chapter XI].
Proposition 8.20. The complex associated to L•
F by Dold-Kan computes R limI F.
We will prove this only when I is cofiltered, which is the argument that is used in the main
text.
Proof. It is clear that limI F is the kernel of L0
F → L1
F . Picking an injective resolution of F
(in the category of functors I → A) and using exactness of infinite products, we get a natural map
from R limI F to L•
F . We claim that this is an equivalence.
If I is cofiltered, we can write I as a filtered union of finite subcategories Ij that contain an
initial object. Then both sides are equal to the derived cofiltered limit of their variants for the Ij.
Thus, we can assume that I is finite and has an initial object i. But in this case, we have proved
the exactness above.

9. LECTURE IX: END OF PROOF 57
9. Lecture IX: End of proof
Recall that in the last lecture, we reduced Theorem 6.11 to the following result.
Theorem 9.1. Fix radii 1 r0 r 0. Then for all r-normed Z[T±1]-modules V and all
profinite sets S, the map
R HomZ[T−1](M(S, Z((T))r0 ), b
V ) → b
V (S)
is a quasi-isomorphism.
We note that actually the case of b
V = Z((T))B
r itself seems to carry all essential difficulty: We
believe that any argument one can give in that case will work in general.
In that sense, we have simplified the target of the R Hom as far as possible. It is time to
understand the source. In the last lecture, we observed that the theorem automatically implies a
more precise version bounding the norms of preimages of differentials. Our proof will actually go
back to such explicit bounds, but for explicit resolutions of the source, which we will now construct.
Write
Mr0 (S) := M(S, Z((T))r0 )/Z[T−1
][S].
This is, as a condensed abelian group (but not as Z[T−1]-module), a direct summand M(S, TZ[[T]]r0 )
of M(S, Z((T))r0 ), allowing only positive powers of T.16
To see this, use Proposition 2.1 to see that
the contribution to M(S, Z((T))r0 ) from nonpositive powers of T is exactly Z[T−1][S]. In particular,
we can write
Mr0 (S) =
[
c0
Mr0 (S)≤c,
where
Mr0 (S)≤c = lim
←
−
i
Mr0 (Si)≤c
when writing S as an inverse limit of finite sets Si, and for finite S
Mr0 (S)≤c = {(
X
n≥1
an,sTn
)s |
X
n≥1,s∈S
|an,s|rn
≤ c}.
We need to resolve this explicitly as a condensed Z[T−1]-module. The Breen-Deligne resolution
gives us the resolution
. . . → Z[Mr0 (S)2
] → Z[Mr0 (S)] → Mr0 (S) → 0
as condensed Z[T−1]-modules; as in the appendix, we assume (for notational convenience) that
each term is of the form Z[Mr0 (S)a] (instead of some finite direct sum of such). Moreover, each
term Z[Mr0 (S)a] admits a two-term resolution
0 → Z[T−1
][Mr0 (S)a
]
T−1−[T−1]
−
−
−
−
−
−
−
→ Z[T−1
][Mr0 (S)a
] → Z[Mr0 (S)a
] → 0
16At this point, we critically use that we chose Z[T−1
] as our base ring; already taking Z[T±1
] would destroy
this argument.

that is functorial in the Z[T−1]-module Z[Mr0 (S)a]. This gives a resolution of Mr0 (S) by the
double complex
. . . // Z[T−1][Mr0 (S)2]
T−1−[T−1]

// Z[T−1][Mr0 (S)]
T−1−[T−1]

. . . // Z[T−1][Mr0 (S)2] // Z[T−1][Mr0 (S)].
Unfortunately, the terms are not of the form Z[T−1][Ei] with profinite Ei. To remedy this, we write
the double complex as the filtered colimit of the double complexes
(9.1) . . . // Z[T−1][Mr0 (S)2
≤r0κ1c]
T−1−[T−1]

// Z[T−1][Mr0 (S)≤r0c]
T−1−[T−1]

. . . // Z[T−1][Mr0 (S)2
≤κ1c] // Z[T−1][Mr0 (S)≤c]
for varying c 0, where the constants κ1, . . . 0 are fixed and chosen so that the transition maps
are well-defined, cf. Lemma 9.11 in the appendix to this lecture.
It is easy to understand what happens to the vertical part under mapping to V :
Lemma 9.2. For any r-normed Z[T±1]-module V , any c 0 and any a, the map
b
V (Mr0 (S)a
≤c)
T−1−[T−1]∗
−
−
−
−
−
−
−
−
→ b
V (Mr0 (S)a
≤r0c)
is surjective, has norm bounded by r−1 +1, and for any f ∈ b
V (Mr0 (S)a
≤r0c) and 0 there is some
g ∈ b
V (Mr0 (S)a
≤c) with f(x) = T−1g(x) − g(T−1x) and ||g|| ≤ r
1−r (1 + )||f||.
Proof. Given f : Mr0 (S)a
≤r0c → b
V , choose an extension to a map e
f : Mr0 (S)a → b
V with
|| e
f|| ≤ (1 + )||f||. Such an extension exists: By induction (and using a sequence of n’s with
Q
n(1 + n) ≤ 1 + ), it suffices to see that for any closed immersion A ⊂ B of profinite sets and
a map fA : A → b
V , there is an extension fB : B → b
V of fA with ||fB|| ≤ (1 + )||fA||. To see
this, write fA as a (fast) convergent sum of maps that factor over a finite quotient of A; for maps
factoring over a finite quotient of A, the extension is clear (and can be done in a norm-preserving
way), as any map from A to a finite set can be extended to a map from B to the same finite set.
Given e
f, we can now define g : Mr0 (S)a
≤c by
g(x) = T e
f(x) + T2 e
f(T−1
x) + . . . + Tn+1 e
f(T−n
x) + . . . ∈ b
V ;
then ||g|| ≤ r
1−r || e
f|| ≤ r
1−r (1 + )||f||.
Let b
V (M(S)≤c)T−1
⊂ b
V (M(S)≤c) be the kernel of this map, with the subspace norm. For
varying c, we now get varying normed complexes. We will need to use the following qualitative
notion of exactness.
Definition 9.3. For each sufficiently large c (i.e. all c ≥ c0 for some c0 0), let
C•
c : C0
c → C1
c → . . .

be a complex of complete normed abelian groups, and for c0 c, let resi
c0,c : C•
c0 → C•
c be a
map of complexes, satisfying the obvious associativity condition. This datum is admissible if all
differentials and maps resi
c0,c are norm-nonincreasing.
For integers m ≥ 0 and constants k ≥ 1, c0
0 0, the datum (C•
c )c is ≤ k-exact in degrees ≤ m
and for c ≥ c0
0 if the following condition is satisfied. For all c ≥ c0
0 and all x ∈ Ci
kc with i ≤ m there
is some y ∈ Ci−1
c (which is defined to be 0 when i = 0) such that
||resi
kc,c(x) − di−1
c (y)||Ci
c
≤ k||di
kc(x)||Ci+1
kc
.
We apply this to
C•
c : b
V (Mr0 (S)≤c)T−1
→ b
V (Mr0 (S)2
≤κ1c)T−1
→ . . .
given by mapping (9.1) into b
V and using Lemma 9.2. Now we state the following result.
Theorem 9.4. Fix radii 1 r0 r 0. For any m there is some k and c0 such that for all
profinite sets S and all r-normed Z[T±1]-modules V , the system of complexes
C•
c : b
V (Mr0 (S)≤c)T−1
→ b
V (Mr0 (S)2
≤κ1c)T−1
→ . . .
is ≤ k-exact in degrees ≤ m for c ≥ c0.
Let us first check that this implies Theorem 9.1.
Theorem 9.4 implies Theorem 9.1. By the preceding discussion, one can compute
R HomZ[T−1](Mr0 (S), b
V )
as the derived inverse limit of C•
c over all c 0; equivalently, all c ≥ c0. Theorem 9.4 implies that
for any m ≥ 0 the pro-system of cohomology groups Hm(C•
c ) is pro-zero (as Hm(C•
kc) → Hm(C•
c )
is zero). Thus, the derived inverse limit vanishes, as desired.
We will prove Theorem 9.4 by induction on m. Unfortunately, the induction requires us to
prove a stronger statement. This goes as follows. Consider any polyhedral lattice Λ, by which
we mean a finite free abelian group equipped with a norm || · ||Λ : Λ ⊗ R → R (so Λ ⊗ R is a
Banach space) that is given by the supremum of finitely many linear functions on Λ with rational
coefficients; equivalently, the “unit ball” {λ ∈ Λ ⊗ R | ||λ||Λ ≤ 1} is a rational polyhedron.
Endow Hom(Λ, Mr0 (S)) with the subspaces
Hom(Λ, Mr0 (S))≤c = {f : Λ → Mr0 (S) | ∀x ∈ Λ, f(x) ∈ Mr0 (S)≤c||x||}.
As Λ is polyhedral, it is enough to check the given condition for finitely many x.
We can then define double complexes like (9.1). Lemma 9.2 stays true with the same constants.
Now we claim the following generalization of Theorem 9.4.
Theorem 9.5. Fix radii 1 r0 r 0. For any m there is some k such that for all polyhedral
lattices Λ there is a constant c0(Λ) 0 such that for all profinite sets S and all r-normed Z[T±1]-
modules V , the system of complexes
C•
Λ,c : b
V (Hom(Λ, Mr0 (S))≤c)T−1
→ b
V (Hom(Λ, Mr0 (S))2
≤κ1c)T−1
→ . . .
is ≤ k-exact in degrees ≤ m for c ≥ c0(Λ).

We note that the constants κ1, κ2, . . . implicit in the choice of the complex are chosen once and
for all (after fixing r and r0), and it can be ensured that the transition maps in the complex are
norm-nonincreasing. Indeed, with the κi chosen as in Lemma 9.11, the maps
b
V (Hom(Λ, Mr0 (S))ai
≤κic) → b
V (Hom(Λ, Mr0 (S))
ai+1
≤κi+1c)
will have bounded norm, independently of V (as they are a certain universal finite sum of maps
induced by maps between the profinite sets in paranthesis, each of which induces a map of norm
bounded by 1), so on the subspace of T−1-invariants, one can shrink the norm down to 1 by
shrinking κi+1. We make and fix this choice of the κi for the statement of Theorem 9.5, and the
rest of the proof.
Proposition 9.6. Fix an integer m ≥ 0 and a constant k. Then there exists an 0 and a
constant k0, depending (only) on k and m, with the following property.
Consider an admissible system of double complexes Mp,q
c , p, q ≥ 0, c ≥ c0, of complete normed
abelian groups as well as some k0 ≥ k0 and some H 0, such that
M0,0
c
d00,0
c
//
d0,0
c

M0,1
c
d00,1
c
//
d0,1
c

M0,2
c
d00,2
c
//
d0,2
c

. . .
M1,0
c
d01,0
c
//
d1,0
c

M1,1
c
d01,1
c
//
d1,1
c

M1,2
c
d01,2
c
//
d1,2
c

. . .
M2,0
c
d02,0
c
//
d2,0
c

M2,1
c
d02,1
c
//
d2,1
c

...
.
.
.
.
.
.
(1) for i = 0, . . . , m + 1, the rows Mi,q
c are ≤ k-exact in degrees ≤ m − 1 for c ≥ c0;
(2) for j = 0, . . . , m, the columns Mp,j
c are ≤ k-exact in degrees ≤ m for c ≥ c0;
(3) for q = 0, . . . , m and c ≥ c0, there is a map hq
k0c : M0,q+1
k0c → M1,q
c with
||hq
k0c(x)||M1,q
c
≤ H||x||M0,q+1
k0c
for all x ∈ M0,q+1
k0c , and such that for all c ≥ c0 and q = 0, . . . , m the “homotopic” map
res1,q
k02c,k0c
◦ d0,q
+ hq
k02c
◦ d00,q
k02c
+ d01,q−1
k0c ◦ hq−1
k02c
: M0,q
k02c
→ M1,q
k0c
factors as a composite of the restriction res0,q
k02c,c
and a map
δ0,q
c : M0,q
c → M1,q
k0c
that is a map of complexes (in degrees ≤ m), and satisfies the estimate
(9.2) ||δ0,q
c (x)||M1,q
k0c
≤ ||x||M0,q
c
for all x ∈ M0,q
c .
Then the first row is ≤ max(k02, 2k0H)-exact in degrees ≤ m for c ≥ c0.

We note that the homotopy in (3) is bounded in two ways: On the one hand, the homotopic
map factors over a deep restriction, and on the other hand its norm is bounded by .
Proof. First, we treat the case m = 0. If m = 0, we claim that one can take = 1
2k and k0 = k.
We have to prove exactness at the first step. Let xk02c ∈ M0,0
k02c
and denote xk0c = res0,0
k02c,k0c
(x) and
xc = res0,0
k02c,c
(x). Then by assumption (2) (and k0 ≥ k), we have
||xc||M0,0
c
≤ k||d0,0
k0c(xk0c)||M1,0
k0c
.
On the other hand, by (3),
||res1,0
k02c,k0c
(d0,0
k02c
(x)) + h0
k02c(d00,0
k02c
(x))||M1,0
k0c
≤ ||xc||M0,0
c
,
noting that the left-hand side agrees with δ0,0
c (xc) by assumption. In particular, noting that
res1,0
k02c,k0c
(d0,0
k02c
(x)) = d0,0
k0c(xk0c), we get
||xc||M0,0
c
≤ k||d0,0
k0c(xk0c)||M1,0
k0c
≤ k||xc||M0,0
c
+ kH||d00,0
k02c
(x)||M0,1
k02c
.
Thus, taking = 1
2k as promised, and bringing 1
2||xc||M0,0
c
to the left-hand side, this implies
||xc||M0,0
c
≤ 2kH||d00,0
k02c
(x)||M0,1
k02c
.
This gives the desired ≤ max(k02, 2k0H)-exactness in degrees ≤ m for c ≥ c0.
Now we argue by induction on m. Consider the complex Np,q given by Mp,q+1 for q ≥ 1
and Np,0 = Mp,1/Mp,0 (the quotient by the closure of the image, which is also the completion
of Mp,1/Mp,0), equipped with the quotient norm. Using the normed version of the snake lemma,
Proposition 9.10 in the appendix to this lecture, one checks that this satisfies the assumptions for
m − 1, with k replaced by max(k4, k3 + k + 1). To verify condition (3), note that the maps δ0,q
c
induce similar maps after passing to this quotient complex. To verify the estimate (9.2), note that
it is nontrivial only for N0,0 = M0,1/M0,0. In that case, for any given a 0 one can lift x ∈ N0,0
c
to e
x ∈ M0,1
c with ||e
x||N0,0
c
≤ ||x||M0,1
c
+ a. This implies
||δ0,q
c (x)||N1,0
k0c
≤ ||δ0,q
c (e
x)||M1,1
k0c
≤ ||e
x||M0,1
c
≤ ||x||M0,1
c
+ a
for all a 0, and hence the desired inequality by taking the infimum over all a.
Finally, we can prove the key combinatorial lemma, ensuring that any element of Hom(Λ, Mr0 (S))
can be decomposed into N elements whose norm is roughly 1
N of the original element. As prepa-
ration, we have the following simple result.
Lemma 9.7. Let Λ be a finite free abelian group, let N be a positive integer, and let λ1, . . . , λm ∈
Λ be elements. Then there is a finite subset A ⊂ Λ∨ such that for all x ∈ Λ∨ = Hom(Λ, Z) there is
some x0 ∈ A such that x − x0 ∈ NΛ∨ and for all i = 1, . . . , m, the numbers x0(λi) and (x − x0)(λi)
have the same sign, i.e. are both nonnegative or both nonpositive.
Proof. It suffices to prove the statement for all x such that λi(x) ≥ 0 for all i; indeed, applying
this variant to all ±λi, one gets the full statement.
Thus, consider the submonoid Λ∨
+ ⊂ Λ∨ of all x that pair nonnegatively with all λi. This is a
finitely generated monoid by standard results; let y1, . . . , yM be a set of generators. Then we can
take for A all sums n1y1 + . . . + nM yM where all nj ∈ {0, . . . , N − 1}.

Now we have the key lemma:
Lemma 9.8. Let Λ be a polyhedral lattice. Then for all positive integers N there is a constant
d such that for all c 0 and all profinite sets S one can write any x ∈ Hom(Λ, Mr0 (S))≤c as
x = x1 + . . . + xN
where all xi ∈ Hom(Λ, Mr0 (S))≤c/N+d.
Proof. The desired statement is equivalent to the surjectivity of the map of profinite sets
Hom(Λ, Mr0 (S))N
≤c/N+d ×Hom(Λ,Mr0 (S))≤c+Nd
Hom(Λ, Mr0 (S))≤c → Hom(Λ, Mr0 (S))≤c.
Note that, as a functor of S, both sides commute with cofiltered limits, so it is enough to handle
finite S, by Tychonoff.
Pick λ1, . . . , λm ∈ Λ generating the norm. We fix a finite subset A ⊂ Λ∨ satisfying the conclusion
of the previous lemma. Write, for finite S,
x =
X
n≥1,s∈S
xn,sTn
[s]
with xn,s ∈ Λ∨. Then we can decompose
xn,s = Nx0
n,s + x1
n,s
where x1
n,s ∈ A and we have the same-sign property of the last lemma. Letting x0 =
P
n≥1,s∈S x0
n,sTn[s],
we get a decomposition
x = Nx0
+
X
a∈A
axa
with xa ∈ Mr0 (S) (with the property that in the basis given by the Tn[s], all coefficients are 0 or
1). Crucially, we know that for all i = 1, . . . , m, we have
||x(λi)|| = N||x0
(λi)|| +
X
a∈A
|a(λi)|||xa||
by using the same sign property of the decomposition.
Using this decomposition of x, we decompose each term into N summands. This is trivial for
the first term Nx0, and each summand of the second term reduces to the similar problem for Λ = Z.
In that case, one can take d = 1, as follows by decomposing any sum with terms of size at most 1
into N such partial sums whose sums differ by at most 1. (It follows that in general one can take
for d the supremum over all i of
P
a∈A |a(λi)|.)
Proof of Theorem 9.5. We argue by induction on m, so assume the result for m − 1 (this
is no assumption for m = 0, so we do not need an induction start). This gives us some k 1 for
which the statement of Theorem 9.5 holds true for m − 1; if m = 0, simply take any k 1. In the
proof below, we will increase k further in a way that depends only on m and r. After this modified
choice of k, we fix and k0 as provided by Proposition 9.6. Moreover, we let k0 be the supremum
of k0 and the κ0
i from Lemma 9.12 (and 9.13) for i = 0, . . . , m. Finally, choose a positive integer b
so that 2k0( r
r0 )b ≤ , and let N be the minimal power of 2 that satisfies
k0
/N ≤ (r0
)b
.
Then in particular rbN ≤ 2k0( r
r0 )b ≤ .

We consider the diagonal embedding
Λ ,→ Λ0
= ΛN
,
where we endow Λ0 with the norm
||(λ1, . . . , λN )||Λ0 = 1
N (||λ1||Λ + . . . + ||λN ||Λ).
For any m ≥ 1, let Λ0(m) be given by Λ0m/Λ ⊗ (Zm)P
=0; then Λ0(•) is a cosimplicial polyhedral
lattice, the Čech conerve of Λ → Λ0. For m = 0, we set Λ0(0) = Λ. It is clear that all of these are
polyhedral lattices.
In particular, for any c 0, we have
Hom(Λ0(m)
, Mr0 (S))≤c = Hom(Λ0
, Mr0 (S))
m/ Hom(Λ,Mr0 (S))≤c
≤c ,
the m-fold fibre product of Hom(Λ0, Mr0 (S))≤c over Hom(Λ, Mr0 (S))≤c; and
Hom(Λ0
, Mr0 (S))≤c = Hom(Λ, Mr0 (S))N
≤c/N ,
with the map to Hom(Λ, Mr0 (S))≤c given by the sum map.
Consider the collection of double complexes C•
Λ0(•),c
associated to this cosimplicial polyhedral
lattice by Dold-Kan. Up to rescaling the norms in the complex for Λ0(m) by a universal constant
(something like (m + 2)!), the differentials are strictly compatible with norms (as they are an
alternating sum of m + 1 face maps, all of which are of norm ≤ 1), so this collection of normed
double complexes is admissible. By induction, the first condition of Proposition 9.6 is satisfied for
all c ≥ c0 with c0 large enough (depending on Λ but not V or S). By Lemma 9.8, and noting that
Hom(Λ0(•), Mr0 (S))≤c is the Čech nerve of
Hom(Λ, Mr0 (S))N
≤c/N
P
−
→ Hom(Λ, Mr0 (S))≤c,
also the second condition is satisfied, with k the maximum of the previous k and some constant
depending only on m and r, provided we take c0 large enough so that (k − 1)r0κic0/N is at least
the d of Lemma 9.8 for all i = 0, . . . , m (so this choice of c0 again depends on Λ). Indeed, then one
can splice a surjection of profinite sets between the maps
Hom(Λ, Mr0 (S))Na
≤κic/N → Hom(Λ, Mr0 (S))a
≤κic
and
Hom(Λ, Mr0 (S))Na
≤kκic/N → Hom(Λ, Mr0 (S))a
≤kκic,
and so the transition map between the columns of that double complex factors over a similar
complex arising from a simplicial hypercover of profinite sets, so the constants are bounded by
Proposition 8.19, Lemma 9.2, and (the version for kernels instead of cokernels of) Proposition 9.10.
At this point, we have finalized our choice of k (and, as promised, this choice depended only on m
and r), and so we also finalized the constants , k0 and N from the first paragraph of the proof.
Finally, to check the third condition, we use Lemma 9.13 to find, in degrees ≤ m, a homotopy
between the two maps from the first row
b
V (Hom(Λ, Mr0 (S))≤c)T−1
→ b
V (Hom(Λ, Mr0 (S))2
≤κ1c)T−1
→ . . .
to the second row
b
V (Hom(Λ, Mr0 (S))N
≤c/N )T−1
→ b
V (Hom(Λ, Mr0 (S))2N
≤κ1c/N )T−1
→ . . .

respectively induced by the addition Hom(Λ, Mr0 (S))N
≤c/N → Hom(Λ, Mr0 (S))≤c (which is the map
that forms part of the double complex), and the map that is the sum of the N maps induced by
the N projection maps
Hom(Λ, Mr0 (S))N
≤c/N → Hom(Λ, Mr0 (S))≤c/N ⊂ Hom(Λ, Mr0 (S))≤c.
By Lemma 9.13, we can find this homotopy between the complex for k0c and the complex for c, by
our choice of k0 ≥ κ0
i for i = 0, . . . , m. As N is fixed, the homotopy is the universal homotopy from
Lemma 9.13, and in particular its norm is bounded by some universal constant H.
Finally, it remains to establish the estimate (9.2) on the homotopic map. We note that this
takes x ∈ b
≤k02κic
)T−1
(with i = q in the notation of (9.2)) to the element
y ∈ b
V (Hom(Λ, Mr0 (S))Nai
≤k0κic/N )T−1
that is the sum of the N pullbacks along the N projection maps Hom(Λ, Mr0 (S))Nai
≤k0κic/N →
Hom(Λ, Mr0 (S))ai
≤k02κic
. We note that these actually take image in Hom(Λ, Mr0 (S))ai
≤κic as N ≥ k0,
so this actually gives a well-defined map
b
≤κic)T−1
→ b
≤k0κic/N )T−1
.
We need to see that this map is of norm ≤ . Now note that by our choice of N, we actually have
k0κic/N ≤ (r0)bκic, so this can be written as the composite of the restriction map
b
≤κic)T−1
→ b
≤(r0)bκic
)T−1
and
b
≤(r0)bκic
)T−1
→ b
≤k0κic/N )T−1
.
The first map has norm exactly rb, by T−1-invariance, and as multiplication by T scales the norm
with a factor of r on b
V .17
The second map has norm at most N (as it is a sum of N maps of norm
≤ 1). Thus, the total map has norm ≤ rbN. But by our choice of N, we have rbN ≤ , giving the
result.
Thus, we can apply Proposition 9.6, and get the desired ≤ max(k02, 2k0H)-exactness in degrees
≤ m for c ≥ c0, where k0, k0 and H were defined only in terms of k, m, r0 and r, while c0 depends
on Λ (but not on V or S). This proves the inductive step.
Question 9.9. Can one make the constants explicit, and how large are they?18
Modulo the
Breen-Deligne resolution, all the arguments give in principle explicit constants; and actually the
proof of the existence of the Breen-Deligne resolution should be explicit enough to ensure the
existence of bounds on the κi and κ0
i.
This completes the proof of all results announced so far.
17Here is where we use r0
r, ensuring different scaling behaviour of the norm on source and target.
18A back of the envelope calculation seems to suggest that k is roughly doubly exponential in m, and that N
has to be taken of roughly the same magnitude.

APPENDIX TO LECTURE IX: SOME NORMED HOMOLOGICAL ALGEBRA 65
Appendix to Lecture IX: Some normed homological algebra
In this appendix, we gather a few results about homological algebra with normed abelian groups,
the proofs of which are just obtained by keeping track of constants in the standard proofs.
Proposition 9.10. Let M•
c and M0•
c be two admissible collections of complexes of complete
normed abelian groups, where c ≥ c0. Let f•
c : M•
c → M0•
c be a collection of maps between these
collections of complexes that is strictly compatible with the norm and commutes with restriction
maps, and assume that it satisfies
||resi
kc,c(x)||Mi
c
≤ k||fi
kc(x)||M0i
kc
for all i = 0, . . . , m + 1 and all x ∈ Mi
kc. Let N•
c = M0•
c /M•
c (which equals the completion of
M0•
c /M•
c ) be the collection of quotient complexes, with the quotient norm; this is again an admissible
collection of complexes.
Assume that M•
c and M0•
c are ≤ k-exact in degrees ≤ m for c ≥ c0. Then N•
c is ≤ max(k4, k3 +
k + 1)-exact in degrees ≤ m − 1 for c ≥ c0.
Proof. We make the following preliminary observation. Take any i = 0, . . . , m + 1 and m0
kc ∈
M0i
kc with image nkc ∈ Ni
kc. By the definition of the quotient norm, for any 0 we can find some
mkc ∈ Mi
kc such that ||m0
kc − fi
kc(mkc)|| ≤ ||nkc|| + . We would like to replace this by the stronger
assertion that we can find mkc ∈ Mi
kc such that
||m0
kc − fi
kc(mkc)|| ≤ (1 + )||nkc||.
This is obviously possible as long as ||nkc|| 0, but in case ||nkc|| = 0, it may not be possible,
because M•
c → M0•
c may not have closed image.
However, we claim that, letting m0
c ∈ M0i
c be the restriction of m0
kc ∈ M0i
kc, with image nc ∈ Ni
c,
we can always find some mc ∈ Mi
c such that
||m0
c − fi
c(mc)|| ≤ (1 + )||nkc||.
By the above, we only need to prove this when ||nkc|| = 0. Choose a sequence mkc,0, mkc,1, . . . in Mi
kc
such that ||m0
kc −fi
kc(mkc,j)|| → 0 for j → ∞. In particular, ||fi
kc(mkc,j −mkc,j0 )|| → 0 for j, j0 → ∞.
By the displayed bound in the statement of the proposition, this ensure that ||mc,j − mc,j0 || → 0
where mc,j ∈ Mi
c is the image of mkc,j. Thus, we get a Cauchy sequence in Mi
c whose limit mc ∈ Mi
c
will satisfy ||m0
c − fi
c(mc)|| = 0 (i.e. m0
c = fi
c(mc)).
Now we start the proof of the proposition. Let ni
k4c ∈ Ni
k4c for i ≤ m − 1, with image ni+1
k4c
∈
Ni+1
k4c
, and let C := ||ni+1
k4c
||Ni+1
k4c
. We need to find an element ni
c ∈ Ni−1
c such that
||ni
c − di−1
N,c(ni−1
c )||Ni
c
≤ (k3
+ k + 1)C,
where we change the subscript when applying restriction maps.
Pick any preimage m0i
k4c ∈ M0i
k4c of ni
k4c, and let m0i+1
k4c
∈ M0i+1
k4c
be its image. By the preliminary
observation, we can find mi+1
k3c
∈ Mi+1
k3c
such that
m0i+1
k3c
= fi+1
k3c
(mi+1
k3c
) + m00i+1
k3c
with ||m00i+1
k3c
||M0i+1
k3c
≤ (1 + )C, where we choose so that (k3 + k)(1 + ) ≤ k3 + k + 1.

Let mi+2
k3c
∈ Mi+2
k3c
be the image of mi+1
k3c
. Applying the differential to the last displayed equation,
and using that this kills m0i+1
k3c
, and that f•
k3c is a map of complexes, we see that
fi+2
k3c
(mi+2
k3c
) = −m00i+2
k3c
,
where similarly m00i+2
k3c
is the differential of m00i+1
k3c
. We get
||mi+2
k2c
||Mi+2
k2c
≤ k||fi+2
k3c
(mi+2
k3c
)||M0i+2
k3c
= k||m00i+2
k3c
||M0i+2
k3c
≤ k||m00i+1
k3c
||M0i+1
k3c
≤ k(1 + )C.
On the other hand, we can find some mi
kc ∈ Mi
kc such that
||mi+1
kc − di
kc(mi
kc)|| ≤ k||mi+2
k2c
||Mi+2
k2c
≤ k2
(1 + )C.
Now let m0i
kc,new = m0i
kc − fi
kc(mi
kc) ∈ M0i
kc; this is a lift of ni
kc. Then the image m0i+1
kc,new in M0i+1
kc
satisfies
m0i+1
kc,new = m0i+1
kc − fi+1
kc (mi+1
kc ) + fi+1
kc (mi+1
kc − di
kc(mi
kc)) = m00i+1
kc + fi+1
kc (mi+1
kc − di
kc(mi
kc)).
In particular,
||m0i+1
kc,new||M0i+1
kc
≤ (1 + )C + k2
(1 + )C.
Now we can find m0i−1
c ∈ M0i−1
c such that
||m0i
c,new − d0i−1
c (m0i−1
c )||M0i
c
≤ k||m0i+1
kc,new||M0i+1
kc
≤ (k3
+ k)(1 + )C.
In particular, letting ni−1
c ∈ Ni−1
c be the image of m0i−1
c , we get
||ni
c − di−1
N,c(ni−1
c )||Ni
c
≤ (k3
+ k)(1 + )C,
so by our choice of this gives the desired result.
We need the following results about the Breen-Deligne resolution for normed abelian groups. Let
us consider here abelian groups M (in any topos) equipped with an increasing filtration M≤c ⊂ M
by subobjects indexed by the positive real numbers, such that 0 ∈ M≤c, −M≤c = M≤c and
M≤c + M≤c0 ⊂ M≤c+c0 ; we need no further conditions. Let us call these pseudo-normed abelian
groups.
Fix a choice of a functorial Breen-Deligne resolution
C(M) : . . . → Z[Mai
] → . . . → Z[Ma1
] → Z[Ma0
] → M → 0
of an abelian group M; purely for notational convenience, we can and do assume that each term is
of the form Z[Mai ] (as opposed to a finite direct sum of such). The possibility of doing this follows
from the proof of [Sch19, Theorem 4.10], noting that a functor of the form A 7→ Z[An] ⊕ Z[Am]
admits a surjection from the functor A 7→ Z[An+m] ⊕ Z; this gives a resolution where all terms are
of the form Z[Aai ] ⊕ Zm. Now pass to the quotient of these complexes corresponding to the map
0 → A; this gives a complex all of whose terms are of the form Z[Aai ]/Z. Noting that Z[Aai ] is
functorially isomorphic to Z[Aai ]/Z ⊕ Z (via splitting 0 → Aai → 0), we can then add an acyclic
complex of Z’s in each degree to get a resolution all of whose terms are of the form Z[Aai ].

APPENDIX TO LECTURE IX: SOME NORMED HOMOLOGICAL ALGEBRA 67
Lemma 9.11. There are universal constants κ0 = 1, κ1, κ2, . . . so that the Breen-Deligne reso-
lution admits the subcomplex
C(M)≤c : . . . → Z[Mai
≤κic] → . . . → Z[Ma1
≤κ1c] → Z[Ma0
≤c]
for all pseudo-normed abelian group objects in any topos as above, and all c 0.
Proof. Each differential in the Breen-Deligne resolution is a finite sum of maps induced by
maps Mai+1 → Mai given by some ai × ai+1-matrix of integers. Given κi, one can thus find
some κi+1 so that M
ai+1
≤κi+1c maps into Mai
≤κic for each of those finitely many maps, which gives the
claim.
We also need some homotopies. More precisely, we start with the following homotopy.
Lemma 9.12. For an abelian group M, the maps σ1, σ2 from
C(M2
) : . . . → Z[M2ai
] → . . . → Z[M2a1
] → Z[M2a0
]
to
C(M) : . . . → Z[Mai
] → . . . → Z[Ma1
] → Z[Ma0
],
induced by addition M2 → M, respectively the sum of the two maps induced by two projections
M2 → M, are homotopic, via some functorial homotopy
hi : Z[M2ai
] → Z[Mai+1
].
If M is a pseudo-normed abelian group object in any topos, then σ1 and σ2 are well-defined as
maps of complexes from
C(M2
)≤c/2 : . . . → Z[M2ai
≤κic/2] → . . . → Z[M2a1
≤κ1c/2] → Z[M2a0
≤c/2]
to
C(M)≤c : . . . → Z[Mai
≤κic] → . . . → Z[Ma1
≤κ1c] → Z[Ma0
≤c]
for all c 0. In that case, for all i ≥ 0 there are universal constants κ0
i such that hi defines
well-defined maps
Z[M2ai
≤κic/2] → Z[M
ai+1
≤κ0
iκi+1c
]
for all c 0.
Proof. This is a consequence of the proof of the existence of the Breen-Deligne resolution,
proved in the same way as [Sch19, Proposition 4.17]. The existence of the constants κ0
i is again
formal, as in the last lemma.
Now we need the following generalization to adding N elements.
Lemma 9.13. Let N be a power of 2. The maps of complexes σ1, σ2 from
C(MN
) : . . . → Z[MNai
] → . . . → Z[MNa1
] → Z[MNa0
]
to
C(M) : . . . → Z[Mai
] → . . . → Z[Ma1
] → Z[Ma0
],
induced by addition MN → M, respectively the sum of the N maps induced by the N projections
MN → M, are homotopic, via some functorial homotopy
hN
i : Z[MNai
] → Z[Mai+1
]

which moreover satisfies the following bound, with the same constants κ0
0, κ0
1, . . . as in the previous
lemma:
If M is a pseudo-normed abelian group object in any topos, then σ1 and σ2 are well-defined as
maps of complexes from
C(MN
)≤c/N : . . . → Z[MNai
≤κic/N ] → . . . → Z[MNa1
≤κ1c/N ] → Z[MNa0
≤c/N ]
to
C(M)≤c : . . . → Z[Mai
≤κic] → . . . → Z[Ma1
≤κ1c] → Z[Ma0
≤c]
for all c 0. In that case, hN
i defines well-defined maps
Z[MNai
≤κic/N ] → Z[M
ai+1
≤κ0
iκi+1c
]
for all c 0.
Proof. Let N = 2m. For each j = 0, . . . , m − 1, the two maps from C(M2j+1
) to C(M2j
)
from the previous lemma are homotopic, and we use the homotopy from that lemma. Composing
homotopies (which amounts concretely to a certain sum) we get the desired homotopy from C(M2m
)
to C(M). It follows directly from this construction that the constants κ0
i are unchanged.

10. LECTURE X: SOME COMPUTATIONS WITH LIQUID MODULES 69
10. Lecture X: Some computations with liquid modules
Now we have done the hard work to prove that there is a well-behaved category of liquid real
vector spaces. In the rest of the course, we want to define analytic spaces, and show that various
classical objects, like complex-analytic spaces, or rigid-analytic varieties, give examples.
However, we will leave the discussion of analytic spaces for after Christmas, and in this last
lecture before Christmas, we simply want to get a feeling for the category of liquid real vector spaces,
or liquid Z((T))r-modules; some of these computations will also be necessary for the discussion
of complex-analytic spaces.
Let us do some basic computations of liquidifications.
Proposition 10.1. Fix any radius 0 r 1.
(1) For any compact Hausdorff space S, let
M(S, Z((T))r) ⊂
Y
n∈Z
Z[S] · Tn
be the union over all c 0 of the (compact Hausdorff) subspace M(S, Z((T))r)≤c whose
S0-valued points are all maps S0 →
Q
n∈Z Z[S] · Tn such that for all s0 ∈ S0, the induced
element (µn · Tn)n, µn ∈ Z[S] satisfies µn ∈ Z[S]≤an for some nonnegative integers an
such that
P
anrn ≤ c.
This definition agrees with the previous one for profinite sets S, and in general if
S• → S is a hypercover of a compact Hausdorff S by profinite sets Si, then
M(S•, Z((T))r) → M(S, Z((T))r)
is a resolution. The derived r-liquidification of Z[T−1][S] is given by
M(S, Z((T))r) = lim
−
→
r0r
M(S, Z((T))r0 ).
(2) For any compact abelian group A, the derived r-liquidification of A[T−1] is given by A((T)).
By the first part, we can also define Mp(S) via base change to R for general compact Hausdorff
S; equivalently, via resolving S by profinite sets Si and forming the complex Mp(S•), which will
be quasi-isomorphic to Mp(S) in degree 0. Then the derived p-liquidification of R[S] is equal to
Mp(S), and in particular sits in degree 0. We note that it seems slightly tricky to do these things
directly over R.
Proof. For the first part, note that for profinite S, this is essentially the definition, with
Proposition 2.1. For general compact Hausdorff S, consider a simplicial resolution of S by profinite
sets Si, and the corresponding resolution
. . . → Z[S1] → Z[S0] → Z[S] → 0.
We consider this as a complex of pseudonormed abelian groups, with the subgroups Z[S]≤n. We
claim that whenever x ∈ Z[Si]≤n with d(x) = 0 then there is some y ∈ Z[Si+1]≤n with x = d(y).
Note that this statement depends only on the underlying sets of S and the Si, and the hypercover
S• → S of sets splits; this produces a contracting homotopy which easily gives the statement (the
contracting homotopy is like in the second appendix to Lecture VIII). From here it is easy to see
that also
M(S•, Z((T))r) → M(S, Z((T))r)

is a resolution. Passing to the colimit over r0 r gives the result (as M(S•, Z((T))r) is the
derived r-liquidification of Z[T−1][Si], and thus the complex M(S•, Z((T))r) computes the derived
r-liquidification of Z[T−1][S]).
For the second part, we need to see that A((T)) is r-liquid and that derived r-liquidification
of B = A((T))/A[T−1] vanishes. It is easy to see that any map S → A((T)) extends to a map
e
f : M(S, Z((T))r) → A((T)); as A((T)) is quasiseparated, uniqueness is automatic, so A((T)) is
r-liquid.
Now note that B = A((T))/A[T−1] is a compact abelian group with Z[T−1]-module structure. It
can be resolved via the Breen-Deligne resolution, so it suffices to see that for any compact Hausdorff
space S with an endomorphism T−1 : S → S, the derived r-liquidification of Z[S] (considered as
Z[T−1]-module) vanishes. This can be described as the cone of T−1 − [T−1] on the derived r-
liquidification of Z[T−1][S], the latter of which is described by the first part. Now T−1 − [T−1] has
the inverse T + [T−1]T2 + [T−1]2T3 + . . ., giving the result.
The following corollary generalizes [Sch19, Theorem 4.3 (ii)].
Corollary 10.2. Let V be a p-liquid R-vector space and let A be a compact abelian group.
Then RHom(A, V ) = 0. Equivalently, the derived p-liquidification of A vanishes.
Proof. Write R = Z((T))r/(2 − T−1) where r = 2−p. Then the derived p-liquidification of
A is the cone of 2 − T−1 on the derived r-liquidification of A[T−1]. The latter is A((T)) by part
(2) of the proposition, and 2 − T−1 is invertible on this (with inverse T + 2T2 + 4T3 + . . .).
Remark 10.3. In the proof of [Sch19, Theorem 4.3 (ii)], we also used the action of 2 − [2] on
a Breen-Deligne resolution; this seems closely related to writing R = Z((T))r/(2 − T−1).
The corollary implies formally that if V is a p-liquid R-vector space considered as a trivial
representation of a compact abelian group A, then the group cohomology Hi(A, V ) vanishes for
i 0. (Indeed, by general nonsense there is a spectral sequence starting with Exti
(
Vj
A, V )
converging to Hi+j(A, V ).) The proof uses commutativity critically.
Question 10.4. Let G be a compact nonabelian group, say G = SU(2), and let V be a p-liquid
R-vector space, e.g. a p-Banach. Does Hi(G, V ) vanish for i 0?
Here is a related question.
Question 10.5. Consider C as a condensed ring, and correspondingly the algebraic K-theory
K(C) as a condensed spectrum (sending an extremally disconnected S to K(C(S, C))). What is
the p-liquidification of K(C)? What is the r-liquidification of K(C)[T−1]?19
We note that the solidification can be computed:20
Proposition 10.6. The solidification of K(C) is ku, connective topological K-theory.
Proof. The essential point is that S[GLn(C)] ∼
= S[|GLn(C)|] where |GLn(C)| is the corre-
sponding anima (cf. [Sch19, Example 6.5]).
19For the latter, we remark that one can define liquidification also over S((T))r where S is the sphere spectrum.
20
Again, solidification can be defined over S.

10. LECTURE X: SOME COMPUTATIONS WITH LIQUID MODULES 71
Of course, solidification forgets all information about real vector spaces, and we want to re-
cover those via p-liquidification. This has the slightly annoying feature that, at least a priori, it
depends on p. The r-liquidification of K(C)[T−1] has the virtue that it specializes to all possible
p-liquidifications via specializing T to different values; and it remembers the integral information.
We note that for K1(C) = C×, the r-liquidification of C×[T−1] can be computed: Using the de-
composition C× ∼
= R × R/Z, it is
C×
((T))r = R((T))r × R/Z((T)).
On the other hand, let us try to understand the structure of M(S, Z((T))r) and Mp(S) better.
Proposition 10.7. Let S be a compact Hausdorff space and 0 p 1. Then for any µ ∈
Mp(S) (element of the underlying set) there is a sequence s0, s1, . . . in S and real numbers x0, x1, . . .
with
P
|xn|p ∞, such that
µ =
∞
X
n=0
xn[sn].
Similarly, for 0 r 1 and any µ ∈ M(S, Z((T))r), there is a sequence s0, s1, . . . ∈ S and
elements x0, x1, . . . ∈ Z((T))r such that xn ∈ Z((T))r,≤cn with
P
cn ∞ and
µ =
∞
X
n=0
xn[sn].
Thus, all p-measures are just countable sums of Dirac measures. In particular, Haar measures
which are “equidistributed” over the whole space, will never lie in Mp(S) for p 1 (except for
finite groups). This phenomenon, of measures being countable sums of Dirac measures, actually
happens more generally for measures “of bounded entropy”, and this is the minimal condition that
the Ribe extension forces on us.
We note that this is only a description of the underlying set: As a condensed set, Mp(S) is not
a filtered colimit over its subspaces for (closures of) countable subspaces of S. Indeed, the map
S → Mp(S) does not factor over such a subspace in general.
Proof. It suffices to handle the assertion over Z((T))r; the case of Mp(S) follows via base
change. But then there is some c ∞ such that
µ ∈ M(S, Z((T))r)≤c ⊂
Y
n
Z[S] · Tn
,
and each coefficients of Tn involves only finitely many elements of S. It follows that one can
write µ =
P∞
n=0 xn[sn] for some countable sequence of elements sn ∈ S, which we assume has
no repetitions. The condition µ ∈ M(S, Z((T))r) then amounts to the condition that if xn ∈
Z((T))r,≤cn with cn chosen minimal, then
P
cn ≤ c. This gives the result.
Finally, we use the formalism in the simplest situation of relevance to complex-analytic geom-
etry.
Proposition 10.8. Let V be the condensed C-vector space of overconvergent holomorphic func-
tions on the closed unit disc {|z| ≤ 1}. Equivalently,
V = lim
−
→
r1
{
X
anzn
∈ C[[z]] |
X
|an|r−n
∞}

where each term in the filtered colimit is regarded as a Banach space. Then for any 0 p 1, the
tensor product
V ⊗Lliq
C,p V
is given by
lim
−
→
r1
{
X
anbmzn
wm
∈ C[[z, w]] |
X
|anbm|r−n−m
∞}
of overconvergent holomorphic functions on the 2-dimensional disc {|z|, |w| ≤ 1}.
Proof. Decompose Mp(N ∪ {∞}) = W ⊕ R · [∞]. Then W is the union of the spaces Wp0
for p0 p whose compact subspaces are given by the space of sequences (x0, x1, . . .) ∈ R with
P
|xi|p0
≤ c. The key observation is now that
V = lim
−
→
r1
WC
along the maps fr : WC → V sending (x0, x1, . . .) to
P
xnrnzn; this is a simple verification, based on
the observation that multiplication by the sequence (1, t, t2, . . .) with t 1 takes bounded sequences
to `p-summable sequences for all 0 p 1.
But now the liquid tensor product of Mp(N ∪ {∞}) with itself is
Mp((N ∪ {∞}) × (N ∪ {∞}))
of which W ⊗Lliq
p W is the direct factor whose compact subspaces are the space of sequences
(xnm)n,m≥0 ∈ R with
P
|xn,m|p0
≤ c for some p0 p and c 0. This gives the result by passing to
the colimit over r.
Corollary 10.9. Let D1, D2 ⊂ C be two closed discs with empty intersection. Let O(D1)
and O(D2) be the spaces of overconvergent holomorphic functions, considered as condensed C[z]-
modules. Then
O(D1) ⊗Lliq
C[z],p O(D2) = 0.
Proof. Forming the tensor product over C, one gets O(D1 × D2) by the previous proposition.
On this space, z1 − z2 is invertible, so the tensor product over C[z] vanishes.
We see that for these questions, the choice of p is largely irrelevant, and that one gets the
expected tensor products.

11. LECTURE XI: TOWARDS LOCALIZATION 73
11. Lecture XI: Towards localization
In the remaining lectures, we will define the category of analytic spaces, and give various
examples. Let us at this point remark that the resulting theory of analytis spaces will, notably,
be an absolute theory: Usually, one defines complex-analytic spaces or rigid-analytic varieties over
some nonarchimedean field K, so works over a fixed base field (and works with K-algebras satisfying
some topological finiteness condition). Here, just like with schemes, it will not be necessary to
specify any base in advance.21
There are two ways to define schemes: As locally ringed topological spaces, or as certain functors
on the category of rings. In the latter approach, the essential non-formal input is just the notion
of a localization A → A[f−1] of rings; from the category of rings with this class of morphisms, one
can recover the category of schemes. Indeed, schemes will be a full subcategory of the category of
functors from rings to sets; in particular, we ask that they are Zariski sheaves (where a finite family
of localizations A → A[f−1
i ] is a cover if the base change functor from A-modules to
Q
i A[f−1
i ]-
modules is faithful). An affine scheme is a functor of the form FA : B 7→ Hom(A, B) for some ring
A. A map F → G of Zariski sheaves is an open immersion if it is injective and for every ring B
with a section s ∈ G(B), corresponding to a map FB → G, the fibre product F ×G FB is the colimit
of FA over all localizations B → A for which FA → FB factors over F ×G FB. Finally, a scheme is
a functor F that admits an open cover by FA’s.
We will follow this route for the definition of analytic spaces. Thus, the next goal is to define
the correct notion of localization A → B of analytic rings. The notion is supposed to have the
following properties:
(1) It is stable under base change.
(2) It is stable under composition.
(3) It is stable under filtered colimits.
The first condition actually requires some explanation; we will come to this later. Regarding
the third condition, note that this means that in the usual algebraic context we allow general
localizations A → A[S−1] at multiplicative subsets. It turns out that in the analytic category, there
is no reasonable “finiteness” one can a priori impose on the localizations, so we will allow general
ones.
Exercise 11.1. Verify that the above formal definition of schemes also gives a well-defined
notion when one allows general localizations A → A[S−1] of rings.
Let us first give several examples.
Example 11.2.
(1) If A → B = A[S−1] is a usual localization of (discrete) rings, with trivial analytic ring
structure, then A → B will be allowed.
(2) The map (Z[T], Z) → (Z[T], Z[T]), corresponding to localization to the subset {|T| ≤ 1}
of the adic space Spa(Z[T], Z).
21Let us mention that Ben-Bassat, Kremnitzer, and coauthors, have also recently undertaken a general study of
analytic spaces, including some notion of quasicoherent sheaves, cf. e.g. [BBK17]. Their theory is based on working
in the symmetric monoidal category of ind-Banach spaces over a fixed base field, roughly in the manner paraphrased
in the first lecture (that is the “relative algebraic geometry” of Toën–Vaquié–Vezzosi). In particular, it is a relative
theory; but also in the relative situation, our theory is quite different, making use of varying analytic ring structures.

(3) More generally, if X = Spa(A, A+) is a (reasonable) adic space and U ⊂ X is a rational
subset, the map
(A, A+
) → (OX(U), O+
X(U)).
(4) In the context of the last lecture, if D1 ⊂ D2 is an inclusion of closed discs in C, then
O(D2) → O(D1) (endowed with the analytic ring structure induced from the p-liquid
structure on C for some chosen 0 p ≤ 1) is a localization.
The parenthetical word “reasonable” occurs here as general adic spaces are not well-behaved:
For example, the structure presheaf may fail to be a sheaf. This failure will actually be corrected
by the formalism of analytic spaces, and the correction will be by slightly changing the localizations
in general.
For example, consider the case U = {|f| ≤ 1} for some f ∈ A. Then OX(U) is by definition the
quotient of the convergent power series algebra AhTi by the closure of the ideal generated by T −f.
On the other hand, in practice this ideal is already closed and moreover T − f is a non-zerodivisor
(in fact, by [KL19, Lemma 2.4.10], this is automatic if (A, A+) is sheafy and A has a topologically
nilpotent unit, for example is an algebra over a nonarchimedean field). Thus, in this case
OX(U) = AhTi/(T − f) = AhTi/L
(T − f) = [AhTi
T−f
−
−
−
→ AhTi].
On the other hand, in general both of these properties can fail. Note that algebraically, the
most natural object is the complex on the right, and indeed this will be the localization in the
world of analytic rings. Working with topological rings, it is impossible to work with the quotient
AhTi/(T − f) if the ideal (T − f) is not closed. However, passing to the condensed world, the
quotient AhTi/(T − f) is perfectly well-behaved. If T − f is a zerodivisor, it however becomes
necessary to work even with the derived reduction AhTi/L(T − f): A condensed animated ring.
We had seen last semester that localizations will in general not be flat, and turn modules
concentrated in degree 0 into complexes (in nonnegative homological degrees). The same will
happen in general for localizations of analytic rings, so we need to also derive our rings. This is the
subject of derived algebraic geometry, which replaces the usual category of rings by its animation,
the ∞-category of animated rings.22
11.1. Animation. Let us briefly recall the notion of animation.23
Let C be a category that
admits all small colimits. Recall that an object X ∈ C is compact (also called finitely presented) if
Hom(X, −) commutes with filtered colimits. An object X ∈ C is projective if Hom(X, −) commutes
with reflexive coequalizers, i.e. colimits along ∆op
≤1 (coequalizers of parallel arrows Y ⇒ Z with
a simultaneous section Z → Y of both maps). Taken together, an object X ∈ C is compact
projective if Hom(X, −) commutes with all filtered colimits and reflexive coequalizers; equivalently,
it commutes with all (so-called) 1-sifted colimits.
Let Ccp ⊂ C be the full subcategory of compact projective objects. There is a fully faithful em-
bedding sInd(Ccp) → C from the 1-sifted Ind-category of Ccp (the full subcategory of Fun((Ccp)op, Set)
22We refer to the work of Toën–Vezzosi, [TV08], and Lurie, [Lur18], for developments of derived algebraic
geometry. Note that these sources say “simplicial ring” when they mean “animated ring”.
23In classical language, this is a non-abelian derived category in the sense of Quillen. It has also been studied
by Rosicky, [Ros07], and we follow Lurie, [Lur09, Section 5.5.8], with language that has been coined by Clausen
inspired by Beilinson’s [Bei07], and used first in [ČS19].

generated under small 1-sifted colimits by the Yoneda image; equivalently, the category freely gen-
erated under small 1-sifted colimits by Ccp). If C is generated under small colimits by Ccp, then
this functor is an equivalence
sInd(Ccp
) ∼
= C.
If Ccp is small, then
sInd(Ccp
) ⊂ Fun((Ccp
)op
, Set)
is exactly the full subcategory of functors that take finite coproducts in Ccp to products in Set.
Example 11.3.
(1) If C = Set is the category of sets, then Ccp is the category of finite sets, which generates C
under small colimits.
(2) If C = Ab is the category of abelian groups, then Ccp is the category of finite free abelian
groups, which generates C under small colimits.
(3) If C = Ring is the category of commutative rings, then Ccp is the category of retracts of
polynomial rings Z[X1, . . . , Xn], which generates C under small colimits.
(4) If C = Cond(Set) is the category of condensed sets, then Ccp is the category of extremally
disconnected profinite sets, which generates C under small colimits.24
(5) If C = Cond(Ab) is the category of condensed abelian groups, then Ccp is the category of
direct summands of Z[S] for extremally disconnected S, which generates C under small
colimits.
(6) If C = Cond(Ring) is the category of condensed rings, then Ccp is the category of retracts
of Z[N[S]] for extremally disconnected S, where N[S] is free condensed abelian monoid on
S and thus Z[N[S]] is the free condensed ring on S. Again, Ccp generates C under small
colimits.
Definition 11.4. Let C be a category that admits all small colimits and is generated under
small colimits by Ccp. The animation of C is the ∞-category Ani(C) freely generated under sifted
colimits by Ccp.
Example 11.5. If C = Set, then Ani(C) = Ani(Set) =: Ani is the ∞-category of animated sets,
or anima for brevity. In standard language, this is the ∞-category of “spaces”.
Let us describe the nature of the ∞-category of anima. Any anima has a set of connected
components, giving a functor π0 : Ani → Set (in fact, π0 is simply given by the universal property
of Ani, as Set is a category with all sifted colimits with a functor from finite sets), which has a fully
faithful right adjoint Set ,→ Ani. Given an anima A with a point a ∈ A (meaning a map a : ∗ → A),
one can define groups πi(A, a) for i ≥ 1, which are abelian for i ≥ 2. The map a : ∗ → A is an
equivalence if and only if π0A is a point and πi(A, a) = 0 for all i ≥ 1. An anima is defined to be
i-truncated if πj(A, a) = 0 for all a ∈ A and j i. Then A is 0-truncated if and only if it is in the
essential image of Set ,→ Ani. The inclusion of i-truncated anima into all anima has a left adjoint
τ≤i. For all anima A, the natural map
A → lim
i
τ≤iA
is an equivalence; this is the “convergence of the Postnikov tower”. Picking any a ∈ A and i ≥ 1,
the fibre of τ≤iA → τ≤i−1A over the image of a is an Eilenberg-MacLane anima K(πi(A, a), i).
24In this example (and the ones to follow), Ccp
is itself a large category.

Here, an Eilenberg-MacLane anima K(π, i), with i ≥ 1 an integer and π a group that is abelian if
i 1, is a pointed connected anima with πj = 0 for j 6= i and πi = π. It is unique up to unique
isomorphism. In fact, the ∞-category of pointed connected anima (A, a) with πj(A, a) = 0 for j 6= i
is equivalent to the category of groups when i = 1, and to the category of abelian groups when
i ≥ 2.
There are several ways to construct Ani(C). It can be defined as the full sub-∞-category of
Fun((Ccp
)op
, Ani)
generated under sifted colimits by the Yoneda image. If Ccp is small, this agrees with the full
sub-∞-category of all contravariant functors F : Ccp → Ani that take finite coproducts to products.
On the other hand, recall that all sifted colimits are generated by filtered colimits and geometric
realizations of simplicial objects. This makes simplicial objects central. In fact, Ani can be defined
as the ∞-category obtained from the category of simplicial sets sSet by inverting weak equivalences.
Similarly, for any ∞-category C generated under small colimits by Ccp, one can describe Ani(C) as
the ∞-category obtained from simplicial objects in C by inverting weak equivalences.25
Example 11.6.
(1) As indicated above, for C = Set, one gets the ∞-category of anima (a.k.a. spaces).
(2) For C = Ab, the ∞-category Ani(Ab) of animated abelian groups is, by the Dold-Kan
equivalence, equivalent to the ∞-derived category of abelian groups D≥0(Ab) in nonneg-
ative homological degrees. This motivates the term “nonabelian deried category” for the
general construction of animation.
(3) For C = Ring, one gets the ∞-category of animated rings.
(4) For C = Cond(Set), one gets the ∞-category Ani(Cond(Set)) of animated condensed sets.
The two operations Ani(−) and Cond(−) actually commute by Lemma 11.8 below, so this
is equivalent the ∞-category Cond(Ani(Set)) = Cond(Ani) of condensed anima.
(5) For C = Cond(Ab), one gets, by Dold-Kan again, the ∞-derived category D≥0(Cond(Ab))
of condensed abelian groups in nonnegative homological degrees.
(6) For C = Cond(Ring), one gets the ∞-category of animated condensed rings, or equivalently
of condensed animated rings.
The operations of animation and passage to condensed objects commute:
Definition 11.7. Let C be an ∞-category that admits all small colimits. For any uncountable
strong limit cardinal κ, the ∞-category Condκ(C) of κ-condensed objects of C is the category of
contravariant functors from κ-small extremally disconnected profinite sets S to C that take finite
coproducts to products.
Moreover, using the fully faithful left adjoints to the forgetful functors,
Cond(C) := lim
−
→
κ
Condκ(C).
Lemma 11.8. Let C be a category that is generated under small colimits by Ccp. Then Cond(C)
is still generated under small colimits by its compact projective objects, and there is a natural
equivalence of ∞-categories
Cond(Ani(C)) ∼
= Ani(Cond(C)).
25We regard it thus as a misnomer to call objects of Ani(Ring) simplicial rings: One has changed the morphisms
drastically.

Proof. A family of compact projective generators of Cond(C) is parametrized by pairs (S, X)
of an extremally disconnected S and a compact projective object X ∈ Ccp, given by the sheafification
of sending T to the colimit over Hom(T, S) many copies of X. These objects still define compact
projective objects of the ∞-category Cond(Ani(C)), and generate the latter under small colimits.
This implies the desired result.
We will be interested in condensed animated rings. However, let us make some remarks about
the nature of the ∞-category of condensed anima, as this combines two different flavours of topology.
The traditional approach to homotopy theory is to define the ∞-category of anima, or “spaces”,
by taking the category of CW complexes, and inverting weak equivalences. This becomes potentially
very confusing when one is also considering condensed sets. Namely, CW complexes also embed
fully faithfully into condensed sets.
Thus, there are two natural functors from CW complexes to condensed anima: A fully faithful
functor, via the full subcategory of condensed sets; and another non-faithful functor, factoring over
the full-∞-subcategory of anima. Let us (abusively) denote the first functor by X 7→ X, and the
second by X 7→ |X|. Then S1 is a physical circle, while |S1| is some ghostly appearance of a point
with an internal automorphism (the anima BZ).
Lemma 11.9. Let X be a CW complex (more generally, any condensed set that is a filtered
colimit of condensed sets that are built out of pushouts of Sn−1 = ∂Dn ,→ Dn). There is a
universal anima Y with a map X → Y of condensed anima. There is a functorial identification
Y ∼
= |X|.
In particular, there is a natural map X → |X| of condensed anima. For S1, we get a cartesian
square (of condensed anima)
R //

∗

S1 // |S1|
involving the universal cover R → S1. If X is a connected CW complex, then X ×|X| τ≥2|X| → X
is the universal cover of X. When X has higher homotopy groups, then X ×|X| ∗ is an actual
condensed anima in that it is nontrivial both in the condensed and in the animated direction. We
invite the reader to get a mental image of S2 ×|S2| ∗.
Proof. As X is an iterated pushout of Sn−1 = ∂Dn ,→ Dn (and these pushouts are compatible
with the fully faithful functors from compactly generated topological spaces to condensed sets to
condensed anima), it suffices to prove the existence of Y in those cases; and this reduces, moreover,
inductively to the case X = Dn. In that case, the natural map Y → Hom(X, Y ) is an equivalence
for any anima Y . To check this, it suffices to see that it induces a bijection on π0: Indeed,
applying the statement on the level of π0 to Hom(Z, Y ) for any other anima Z will then give
π0 Hom(Z, Y ) ∼
= π0 Hom(Z, Hom(X, Y )) for all Z (as Hom(Z, Hom(X, Y )) ∼
= Hom(X, Hom(Z, Y ))),
which means that Y → Hom(X, Y ) is an equivalence.
To check that π0Y → π0 Hom(Dn, Y ) is an isomorphism, we pass up the Postnikov tower. If
Y is discrete, this follows from connectedness of X. If Y = BG for some discrete group G, then
Hom(X, Y ) classifies G-torsors over X (as a condensed set). It is not hard to see that these are
locally on X trivial, as any trivialization at a point spreads uniquely to small closed neighborhoods

(after a resolution by extremally disconnected sets, and thus by descent on X). Thus, they are
classified by usual G-torsors on the topological space X = Dn, and these are all split. If Y = K(π, n)
for n 1, then Hom(X, Y ) is given by Hn(X, π) (cohomology of X as a condensed set). By last
semester, this agrees with singular cohomology, and so it vanishes for X = Dn.
We see that for X = Dn, the map X → |X| = ∗ is the universal map to an anima. As the
functor sending X to this initial anima commutes with colimits, as does X → |X|, this gives the
identification in general.

12. LECTURE XII: LOCALIZATIONS 79
12. Lecture XII: Localizations
We need to generalize the theory of analytic rings from condensed rings to condensed animated
rings. As this works in the context of (still unital) associative rings, let us temporarily work in
that setting. It has the advantage that the ∞-category of animated condensed associative rings
can also be described as the ∞-category of E1-algebras in the symmetric monoidal ∞-category
D≥0(Cond(Ab)) = Ani(Cond(Ab)) of animated condensed abelian groups. Any animated con-
densed ring A thus has an ∞-category D≥0(A) of modules in animated condensed abelian groups.
This is a prestable ∞-category in the sense of Lurie (cf. [Lur18, Appendix C]); in particular, it em-
beds fully faithfully into a stable ∞-category D(A), as the positive part of a t-structure. The heart
is equivalent to the category of condensed modules over the condensed ring π0A. (The preceding
discussion applies to sheaves on any site.)
The following is a direct generalization of [Sch19, Definition 7.1, 7.4], but we decided to change
notation.
Definition 12.1. An analytic animated associative ring is a pair (A, M) consisting of a con-
densed animated associative ring A together with a covariant functor M : S 7→ M[S] from ex-
tremally disconnected profinite sets S to condensed animated A-modules, taking finite coproducts
to finite direct sums, together with a natural transformation S → M[S] of condensed anima, with
the following property. For any object C ∈ D≥0(A) that is a sifted colimit of objects of the form
M[S], the natural map
HomD≥0(A)(M[S0
], C) → HomD≥0(A)(A[S0
], C)
of condensed anima is an equivalence for all extremally disconnected profinite sets S0.
Remark 12.2. The letter M can be thought of as denoting spaces of measures, or as denoting
extra data on modules, singling out the correct category of modules inside all of D≥0(A).
We note that in the condition, the map is actually a map of condensed animated abelian
groups, and by embedding into D(A), we could also use the RHom and ask for an isomorphism in
D(Cond(Ab)), as in [Sch19, Definition 7.4]. (A priori, the condition on RHom is stronger, but it
follows from the given condition applied to all shifts C[i], i ≥ 0.)
Apart from language, the only difference with the situation of [Sch19, Definition 7.1, 7.4] is
that we allow both A and M[S] to be complexes, living in nonnegative homological degrees.
Definition 12.3. Let (A, M) be an analytic animated associative ring. The ∞-category
D≥0(A, M) is defined to be the full ∞-subcategory
D≥0(A, M) ⊂ D≥0(A)
spanned by all C ∈ D≥0(A) such that for all extremally disconnected profinite sets S, the map
HomD≥0(A)(M[S], C) → HomD≥0(A)(A[S], C)
of condensed anima is an equivalence.
Proposition 12.4. Let (A, M) be an analytic animated associative ring. The ∞-category
D≥0(A, M) is generated under sifted colimits by the objects M[S] for varying extremally discon-
nected profinite sets S, which are compact projective objects of D≥0(A, M). The full ∞-subcategory
D≥0(A, M) ⊂ D≥0(A)

is stable under all limits and colimits and admits a left adjoint
− ⊗A (A, M) : D≥0(A) → D≥0(A, M)
sending A[S] to M[S].
The ∞-category D≥0(A, M) is prestable. Its heart is the abelian category D♥(A, M) that is the
full subcategory of condensed π0A-modules generated under colimits by π0M[S] for varying S. An
object C ∈ D≥0(A) lies in D≥0(A, M) if and only if all Hi(C) lie in D♥(A, M).
If A has the structure of a condensed animated commutative ring so that D≥0(A) is naturally
a symmetric monoidal ∞-category, there is a unique symmetric monoidal structure on D≥0(A, M)
making − ⊗A (A, M) symmetric monoidal.
Remark 12.5. Passing to “spectrum objects” (a formal procedure that in the current situation
recovers D(−) from D≥0(−)), similar statements hold true on the level of D(A, M) ⊂ D(A).
A consequence of the proposition is that M[S] is determined by the full ∞-subcategory D≥0(A, M) ⊂
D≥0(A) as the image of A[S] under the left adjoint to the full inclusion. Moreover, this ∞-
subcategory is determined already by the collection {π0M[S]}S of condensed π0A-modules. In
other words, analytic ring structures on A are completely determined by data on the level of usual
abelian categories. In the first appendix to this lecture, we will characterize analytic ring structures
from this perspective.
Proof. By the definition of analytic animated rings, all sifted colimits of objects M[S] are in
D≥0(A, M). On the other hand, by the definition of D≥0(A, M), the objects M[S] are compact
projective generators.
By definition of D≥0(A, M), it is stable under all limits. For the left adjoint, it is clear that it
exists on A[S] with value M[S] by definition of D≥0(A, M), and then it also exists for all of their
sifted colimits by the assumption that (A, M) is an analytic ring; but these sifted colimits exhaust
D≥0(A). Now the existence of general colimits follows, by forming them first in D≥0(A), and then
applying the left adjoint.
It is now formal that D≥0(A, M) is prestable, cf. [Lur18, Corollary C.1.2.3]. If C ∈ D≥0(A, M),
then also τ≥1C ∈ D≥0(A, M) as this is the suspension of the loops of C and D≥0(A, M) is stable
under all limits and colimits. Thus, H0(C)[0] ∈ D≥0(A, M). The statement about the heart is
then formal.
Regarding the symmetric monoidal structure, one possible reference is [NS18, Theorem I.3.6],
noting that the condition of being an analytic ring ensures that the kernel of − ⊗A (A, M) is a
⊗-ideal (as the kernel is generated under colimits by the cones of A[S] → M[S], and tensoring them
over Z with Z[T] for varying T still lies in the kernel, by evaluating the internal Hom at T).
The definition of analytic animated associative rings suggests to define a map (A, M) → (B, N)
of such to be a map of condensed animated associative rings A → B together with a natural
transformation M[S] → N[S] linear over this map and commuting with the maps from S.
Following the (updated) discussion in [Sch19, Lecture 7], we instead define a map (A, M) →
(B, N) to be a map A → B of condensed animated associative rings such that the forgetful functor
D≥0(B, N) → D≥0(A) takes image in D≥0(A, M). It is enough to check this condition for N[S],
and in fact for π0N[S], as these generate the abelian heart, and the categories are determined by
the heart.

Given a map (A, M) → (B, N) of analytic animated associative rings, the maps S → N[S]
extend uniquely to maps M[S] → N[S] linear over A → B, so one gets a necessarily unique map in
the naive sense. The converse is also close to being true: If one has functorial maps M[S] → N[S]
linear over A → B and commuting with the map from S, then this defines a map (A, M) → (B, N)
under a mild condition. In fact, it is enough to give functorial maps π0M[S] → π0N[S] linear over
π0A → π0B such that for all maps S → π0A from some extremally disconnected profinite set S,
the induced diagram
π0M[S] //

π0M[∗]

π0N[S] // π0N[∗]
commutes, cf. [Sch19, Proposition 7.14] (whose proof immediately passes to the present situation).
Proposition 12.6. Let f : (A, M) → (B, N) be a map of analytic animated associative rings.
The forgetful functor D≥0(B, N) → D≥0(A, M) admits a left adjoint
− ⊗(A,M) (B, N) : D≥0(A, M) → D≥0(B, N)
sending M[S] to N[S]. This functor is naturally symmetric monoidal when A and B and f : A → B
have the structure of condensed animated commutative rings (resp. of a map between such).
Remark 12.7. We should really encode more functoriality, especially regarding composition.
This is best done by defining the relevant (co)Cartesian fibrations and straightforward in the present
situation.
Proof. The existence and description of the left adjoint is clear. The symmetric monoidal
structure follows from [NS18, Theorem I.3.6] again.
One can base change analytic ring structures along maps of condensed rings, as follows. We
emphasize that this proposition works only in the animated setting – the relevant base changes will
in general be derived, and we have to keep track of the derived structure.
Proposition 12.8. Let (A, M) be an analytic animated associative ring and let g : A → B be
a map of condensed animated associative rings. Then the functor
S 7→ N[S] := B[S] ⊗A (A, M)
defines an analytic animated associative ring (B, N).
Proof. Note that all N[S] are, as A-modules, by definition in D≥0(A, M); thus, for any sifted
colimit N ∈ D≥0(B) of these, forgetting to D≥0(A) defines an object of D≥0(A, M).
We want to prove
HomD≥0(B)(N[S], N) = HomD≥0(B)(B[S], N).
From N ∈ D≥0(A, M) and the definition of N[S], we know that
HomD≥0(A)(N[S], N) = HomD≥0(A)(B[S], N).
We claim that this implies the desired result formally. Indeed, we first rewrite this as
HomD≥0(B)(N[S] ⊗A B, N) = HomD≥0(B)(B[S] ⊗A B, N)

which moreover implies, by taking HomD≥0(B)(M, −) for varying M ∈ D≥0(B), that
HomD≥0(B)(N[S] ⊗A M, N) = HomD≥0(B)(B[S] ⊗A M, N)
for all such M. This can be rewritten as
HomD≥0(B)(N[S] ⊗B (B ⊗A M), N) = HomD≥0(B)(B[S] ⊗B (B ⊗A M), N).
But the objects of the form B ⊗A M ∈ D≥0(B) for M ∈ D≥0(B) generate all of D≥0(B) under
colimits; for example, there is a resolution of B of the form
. . . → B ⊗A B ⊗A B → B ⊗A B → B
(as this forms a split simplicial diagram). Thus, also
HomD≥0(B)(N[S], N) = HomD≥0(B)(B[S], N).

In the third appendix to this lecture, we show that if (A, M) is an analytic animated associative
ring, then the map A → M[∗] is naturally a map of condensed animated associative rings; in
particular, M[∗] acquires the structure of a condensed animated associative ring. In fact, M[∗]
is initial in the ∞-category of condensed animated associative rings B under A with the property
that B ∈ D≥0(A, M) ⊂ D≥0(A).
Definition 12.9. An analytic animated associative ring (A, M) is normalized if the map A →
M[∗] is an isomorphism.
The preceding discussion shows that for any analytic animated associative ring (A, M), there
is an initial analytic animated associative ring under it, given by applying the previous proposition
to (A, M) and the map A → M[∗]. Normalization does not affect D≥0(A, M).
In the commutative case, passing to normalized rings is actually more subtle. The problem is
that we have to endow M[∗] with the structure of a condensed animated commutative ring, and,
as analyzed in the third appendix, this requires that the Symn
-operations on D≥0(A) descend to
D≥0(A, M). As proved there, this is implied by insisting on the following property in the definition
of condensed animated commutative rings (A, M):
Definition 12.10. An analytic animated commutative ring is a condensed animated commuta-
tive ring A together with a functor M : S 7→ M[S] as above so that (A, M) is an analytic animated
associative ring, and such that for all primes p, the Frobenius map φp : A → A/Lp induces a map
of analytic animated associative rings φp : (A, M) → (A/Lp, M/Lp). A map of analytic animated
commutative rings (A, MA) → (B, MB) is a map of condensed animated commutative rings A → B
that induces a map of analytic animated associative rings.
This condition on Frobenius is analyzed in the second appendix, and shown there to be auto-
matic in a number of situations; it may well be the case that it is automatic in general. By the
results of the third appendix, one can normalize analytic animated commutative rings.26
In the following, we will only work with normalized analytic rings.
26If one works in the context of connective E∞-rings, this subtlety about Frobenius does not come up, as then
it is clear that normalizations exist (as the tensor unit in a symmetric monoidal ∞-category is an E∞-ring); also all
the rest of the discussion of these lectures immediately adapts to that context.

Definition 12.11. Let AnRing be the ∞-category of normalized analytic animated commuta-
tive rings.
Let us analyze this ∞-category. For the following proposition, it is again critical to work in the
animated context.
Proposition 12.12. The ∞-category AnRing admits all small colimits. The initial object is
Z with S 7→ Z[S]. Sifted colimits commute with the functor (A, M) 7→ M[S] to D≥0(Cond(Ab)).
Pushouts are computed as follows:
Let (B, MB) ← (A, MA) → (C, MC) be a diagram in AnRing. The pushout (B, MB) ⊗(A,MA)
(C, MC) = (E, ME) of this diagram in AnRing exists. It can be defined as the normalization of an
analytic ring structure on
B ⊗A C.
The corresponding functor
D≥0(E, ME) → D≥0(B ⊗A C)
is fully faithful, with essential image given by all C ∈ D≥0(B ⊗A C) whose images in D≥0(B) and
D≥0(C) lie in D≥0(B, MB) and D≥0(C, MC).
The left adjoint
− ⊗B⊗AC (E, ME) : D≥0(B ⊗A C) → D≥0(E, ME)
is given by the sequential colimit
− → −⊗B(B, MB) → (−⊗B(B, MB))⊗C (C, MC) → ((−⊗B(B, MB))⊗C (C, MC))⊗B(B, MB) → . . .
(where each functor is considered as an endofunctor of D≥0(B ⊗A C)).
In particular, pushouts in AnRing are in general subtle, as the operations of “MB-completion”
− ⊗B (B, MB) and “C-completion” − ⊗C (C, MC) do not in general commute.
Proof. The statement about the initial object is clear. For the statement about sifted colimits,
let (A•, M•) be any sifted diagram in AnRing.27
Then for any i and S, the sifted colimit of M•[S]
is an object of D≥0(colim A•) whose restriction to D≥0(Ai) lies in D≥0(Ai, Mi), and thus for any
sifted colimit C of A•[S]’s,
HomD≥0(Ai)(Mi[S], C) = HomD≥0(Ai)(Ai[S], C)
or equivalently
HomD≥0(colim A•)(Mi[S] ⊗Ai colim A•, C) = HomD≥0(colim A•)(colim A•[S], C).
Taking the limit over i of this statement and using that the colimit of Mi[S] ⊗Ai colim A• is just
colim M•[S] as the colimit is sifted, we get the desired result.
For pushouts, we note that any object in (E0, ME0 ) ∈ AnRing with compatible maps from
(B, MB) ← (A, MA) → (C, MC) will admit a natural map D≥0(E0, ME0 ) → D≥0(B ⊗A C) whose
essential image lies in the full ∞-subcategory of all objects whose restrictions to the two factors
lie in D≥0(B, MB) and D≥0(C, MC). Thus, we see that it is enough to prove that the recipee in
the statement defines an (a priori unnormalized) analytic ring structure on B ⊗A C; but this is
easy to see (using, for example, Proposition 12.20). Passing to the normalization gives the desired
pushout.
27On first reading, assume that it is a filtered colimit.

One instance where the base change in AnRing is simple is for induced analytic ring structures
as in Proposition 12.8.
Another such instance is the following class of maps.
Definition 12.13. A map f : (A, MA) → (B, MB) in AnRing is steady if for all maps g :
(A, MA) → (C, MC) in AnRing, the functor M 7→ M ⊗B (B, MB) preserves the full ∞-subcategory
of D≥0(C ⊗A B) of all objects whose restriction to C lies in D≥0(C, MC).
Equivalently, for all M ∈ D≥0(C, MC), the object
M ⊗A (B, MB) = M ⊗(A,MA) (B, MB)
which is a priori an object of D≥0(C⊗AB), lies in D≥0(C, MC) when restricted to C. As its restriction
to B lies in D≥0(B, MB) by construction, this will then actually define an object of
D≥0((B, MB) ⊗(A,MA) (C, MC)).
In fact, the condition that f is steady means precisely that for all g the colimit
− → −⊗C (C, MC) → (−⊗C (C, MC))⊗B (B, MB) → ((−⊗C (C, MC))⊗B (B, MB))⊗C (C, MC) → . . .
stabilizes at (−⊗C (C, MC))⊗B (B, MB). The pushout (E, ME) of (C, MC) ← (A, MA) → (B, MB)
is then given by the functor
ME : S 7→ MC[S] ⊗(A,MA) (B, MB).
Yet another equivalent characterization is the following.
Proposition 12.14. A map f : (A, MA) → (B, MB) in AnRing is steady if and only if for all
pushout diagrams
(E, ME) (C, MC)
e
f
oo
(B, MB)
e
g
OO
(A, MA)
f
oo
g
OO
and all C ∈ D(C, MC), the base change map
(C|A) ⊗(A,MA) (B, MB) → (C ⊗(C,MC) (E, ME))|B
is an isomorphism.
In the geometric language of the next lecture, this means that in the cartesian diagram
AnSpec(E, ME)
e
f
//
e
g

AnSpec(C, MC)
g

AnSpec(B, MB)
f
// AnSpec(A, MA),
the map
f∗
g∗C → e
g∗
e
f∗
C
is an isomorphism (using the usual notation for pullbacks and pushforwards of quasicoherent
sheaves). Thus, pushforward commutes with base change along steady maps, and this property
characterizes steady maps.

Proof. As both sides commute with filtered colimits in C, it suffices to check on C = MC[S],
which amounts to the condition above.
The class of steady maps has good properties.
Proposition 12.15. The class of steady maps is stable under base change and composition.
If a steady map f : (A, MA) → (B, MB) in AnRing factors over a map f0 : (A0, MA0 ) →
(B, MB), then f0 is also steady. Moreover, the class of steady maps is closed under all colim-
its (in Fun(∆1, AnRing)).
Proof. Exercise.
Definition 12.16. A map f : (A, MA) → (B, MB) in AnRing is a localization if the forgetful
functor D≥0(B, MB) → D≥0(A, MA) is fully faithful. The map f is a steady localization if it is a
localization and steady.
Exercise 12.17. Show that if a map f : (A, MA) → (B, MB) in AnRing is a localization,
then the induced map (B, MB) → (B, MB) ⊗(A,MA) (B, MB) is an isomorphism. (Hint: First
establish that any base change of a localization is again a localization, by noting that (B, MB) can
be regarded as the normalization of a non-normalized analytic ring structure on A.) If f is steady,
prove the converse. (Hint: The condition of fully faithfulness is equivalent to the statement that
for all M ∈ D≥0(B, MB), the map M → M ⊗(A,MA) (B, MB) is an equivalence. Rewrite the latter
as M ⊗(B,MB) ((B, MB) ⊗(A,MA) (B, MB)).)
With this definition, we have the following descent of modules.
Proposition 12.18. Let fi : (A, M) → (Ai, Mi), i ∈ I, be a finite family of steady localizations
in AnRing such that the functor D≥0(A, M) →
Q
i D≥0(Ai, Mi) is conservative. Let CI be the
category of nonempty subsets of I; we get a functor CI → AnRing sending any J ⊂ I to (AJ , MJ ) :=
N
i∈J,/(A,M)(Ai, Mi).
Then for any M ∈ D(A, M) the natural map
M → lim
J∈CI
(M ⊗(A,M) (AJ , MJ ))
is an isomorphism, and
D(A, M) → lim
J∈CI
D(AJ , MJ )
is an equivalence.
Proof. This is formal, cf. (proof of) [Sch19, Proposition 10.5]. The key point is that steady
localizations commute with any base change. We also use that localizations commute with finite
limits, for which we have to pass from the prestable ∞-category D≥0 to the stable ∞-category D
(where they can be reinterpreted as finite colimits).
Next week, we will discuss some examples. In particular, we will see that for the analytic
rings corresponding to Huber pairs (A, A+), the condition of (A, A+) → (B, B+) being steady is
closely related to the condition that A → B is adic in Huber’s sense, i.e. given compatible rings of
definition A0 ⊂ A, B0 ⊂ B, A0 → B0, if I ⊂ A0 is an ideal of definition, then IB0 ⊂ B0 is an ideal
of definition. Thus the present discussion mirrors a standard discussion on Huber pairs: In that
situation, general pushouts do not exist, but they do exist when one of the maps is adic.

Appendix to Lecture XII: Topological invariance of analytic ring structures
In this appendix, we characterize analytic ring structures in terms of the full sub-∞-category
D≥0(A, M) ⊂ D≥0(A). As an application, we prove that analytic ring structures are invariant
under nilpotent thickenings. First, we have the following result.
Proposition 12.19. Let A be a condensed animated associative ring. The collection of full
sub-∞-categories D ⊂ D≥0(A) stable under all limits and colimits is in natural bijection with the
collection of all full subcategories C ⊂ Modcond
π0A stable under all limits, colimits, and extensions, via
sending D to the intersection with Modcond
π0A , and C to the full sub-∞-category D of all C ∈ D≥0(A)
such that all Hi(C) ∈ C for i ≥ 0.
Proof. If D ⊂ D≥0(A) is stable under all limits and colimits, then in particular for any C ∈ D,
also τ≥1C ∈ D as the suspension of the loops of C, and thus H0(C)[0] ∈ D. It follows that defining
C to be the intersection of D with Modcond
π0A , one has C ∈ D if and only if all Hi(C) ∈ C. Indeed,
the forward direction follows by inductively applying this argument, and the converse follows as D
is stable under extensions (as these can be written as cofibers) and (Postnikov) limits. Moreover,
stability of D under limits and colimits implies stability of C under limits, colimits, and extensions.
Conversely, if C ⊂ Modcond
π0A is stable under limits, colimits, and extensions, then defining D
as in the statement of the proposition, one sees that D is stable under fibres and cofibres, as well
as infinite direct sums or infinite direct products, and thus under all limits and colimits. The
intersection of D with Modcond
π0A is by definition C, giving the result.
Now we can characterize analytic ring structures.
Proposition 12.20. Let A be a condensed animated associative ring. A full sub-∞-category
D ⊂ D≥0(A) is of the form D≥0(A, M) for a necessarily unique analytic ring structure (A, M) on
A if and only if satisfies the following conditions:
(1) The sub-∞-category D ⊂ D≥0(A) is stable under all limits and colimits.
(2) The sub-∞-category D ⊂ D≥0(A) is stable under HomD≥0(Cond(Ab))(Z[S], −) for any ex-
tremally disconnected profinite set S.
(3) The inclusion D ⊂ D≥0(A) admits a left adjoint.
We remark that the existence of the left adjoint is automatic modulo set-theoretic issues.
Proof. This is clear in the forward direction. Conversely, we can define M[S] as the image of
A[S] under the left adjoint (guaranteed by (3)). Then we formally know that for all C ∈ D,
HomD≥0(A)(M[S], C) → HomD≥0(A)(A[S], C)
is an isomorphism of anima. By condition (2), we actually see that it is an isomorphism of condensed
anima. As by (1), D contains all sifted colimits of M[S]’s, we see that this defines an analytic ring
structure.
As an application, one can show that analytic ring structures are independent of the animated
structure.
Proposition 12.21. Let f : A → B be map of condensed animated associative rings such that
π0A → π0B is an isomorphism.

APPENDIX TO LECTURE XII: TOPOLOGICAL INVARIANCE OF ANALYTIC RING STRUCTURES 87
Then there is a bijective correspondence between analytic ring structures (A, M) on A and
analytic ring structures (B, N) on B. In the forward direction, this takes (A, M) to an induced
analytic ring structure
N[S] := B[S] ⊗A (A, M).
Remark 12.22. Everything applies, with identical proofs, to general E1-algebras in condensed
connective spectra. For example one can define a solid analytic ring structure on S, and for 0 r 1
and r-liquid analytic ring structure on S[T−1] (whose normalization is some condensed E∞-ring
S((T))r that seems subtle to define directly).
Proof. By Proposition 12.8, the functor is well-defined. To prove that it this is a bijection,
we can reduce to the case where f is the projection A → π0A.
The hearts of D≥0(A) and D≥0(π0A) agree, and the functor identifies the corresponding hearts
of D≥0(A, M) and D≥0(π0A ⊗A (A, M)). To see that this defines the desired bijection, we need to
see that for any analytic ring structure (π0A, M0) on π0A with corresponding heart C ⊂ Modcond
π0A ,
one can define an analytic ring structure (A, M) such that C ∈ D≥0(A, M) if and only if all
Hi(C) ∈ C. By Proposition 12.19, this defines some full sub-∞-category D ⊂ D≥0(A) stable under
all limits and colimits. We need to check conditions (2) and (3) of Proposition 12.20. For condition
(2), we can assume (by Postnikov limits and filtrations) that C = X[j] for some X ∈ C and j ≥ 0,
in which case C comes from a π0A-module, and the result follows from that case.
For condition (3), we need to see that the functor M 7→ M(S) on D is representable. We prove
by induction on i that there is some i-truncated Mi[S] ∈ D with a map A[S] → Mi[S] such that
HomD(Mi[S], M) → M(S)
is an isomorphism of anima for all i-truncated M ∈ D. For i = 0, we can take M0[S] = π0M0[S].
Given Mi[S], let Ni be the cofiber of A[S] → Mi[S] in D≥0(A) and let
N0
i = Ni ⊗A (π0A, M0
).
As the forgetful functor D≥0(π0A, M0) → D≥0(A) lands in D, we can regard N0
i as an object of D.
Then we let Mi+1[S] be the fibre of Mi[S] → Ni → N0
i, and we note that A[S] → Mi[S] factors
naturally over Mi+1[S].
To see that
HomD(Mi+1[S], M) → M(S)
is an isomorphism of anima for all i + 1-truncated M ∈ D, it suffices to check for M = X[i + 1]
with X ∈ C. Note that for such M, we have the fibre sequence
HomD≥0(A)(Ni, M) → HomD(Mi[S], M) → M(S)
where the right-most term is concentrated in degree i + 1, while the middle term is isomorphic to
M(S) after applying τ≥1, by induction. It follows that HomD≥0(A)(Ni, M) sits in degree 0. This
can also be written as
HomD≥0(π0A)(Ni ⊗A π0A, M)
which again for all M = X[i+1] with X ∈ C sits in degree 0. It follows that N0
i = Ni ⊗A (π0A, M0)
lies in D≥i+1(π0A, M0), and we still have
HomD≥0(π0A)(N0
i, M) = HomD≥0(π0A)(Ni ⊗A π0A, M).

On the other hand, the map
HomD≥0(π0A)(N0
i, M) → HomD≥0(A)(N0
i, M)
is also an isomorphism, as N0
i ∈ D≥i+1 and M is i + 1-truncated.
In summary, we see that the natural map
HomD≥0(A)(Ni, M) ← HomD(N0
i, M)
is an isomorphism, where both sides are concentrated in degree 0. We defined Mi+1[S] as the fiber
of Mi[S] → N0
i. Recall that N0
i was in D≥i+1, in particular its π0 vanishes. This implies that we
get a cofiber sequence
HomD(N0
i, M) → HomD(Mi[S], M) → HomD(Mi+1[S], M)
and comparing with
HomD≥0(A)(Ni, M) → HomD(Mi[S], M) → M(S)
gives the result.
One can moreover prove invariance under nilpotent thickenings.
Proposition 12.23. Let f : A → B be a map of condensed animated associative rings such
that π0A → π0B is surjective with nilpotent kernel I ⊂ A. Then analytic ring structures (A, M)
on A are in bijection with analytic ring structures (B, N) on B, via taking induced analytic ring
structures.
Proof. By Proposition 12.21, we can assume that A and B are 0-truncated. We may also
assume that I2 = 0 by induction. Let us define the inverse functor, so take an analytic ring structure
(B, N) on B, corresponding to some abelian category CB ⊂ Modcond
B . Let C ⊂ Modcond
A be the
subcategory of all M ∈ Modcond
A such that M/IM and IM, which lie naturally in Modcond
B , are in CB.
This contains CB and is stable under all limits, colimits, and extensions; it can also be characterized
as the category generated by CB under extensions. By Proposition 12.19, it corresponds to some
full sub-∞-category D ⊂ D≥0(A). We need to check the criteria of Proposition 12.20. Condition
(2) can again be checked for C = X[j] for some X ∈ C and j ≥ 0, where it follows by filtering in
terms of two objects from CB.
It remains to check condition (3). This follows from the argument in the proof of Proposi-
tion 12.21 with a slightly refined induction. Namely, besides the class of i-truncated objects of
D, we also consider the class of i-truncated objects C ∈ D such that Hi(C) is killed by I. Then
we start the induction with this restricted class of 0-truncated objects, then go to all 0-truncated
objects, then to the restricted class of 1-truncated objects, to all 1-truncated objects, etc. .

APPENDIX TO LECTURE XII: FROBENIUS 89
Appendix to Lecture XII: Frobenius
Fix a prime p. Consider a condensed animated commutative Fp-algebra A, equipped as an
associative with an analytic ring structure (A, M). One may wonder whether the Frobenius φ :
A → A induces a map of analytic rings (A, M) → (A, M). This condition is stable under all
colimits. The goal of this appendix is to show that it is true in some important cases. We denote
by Cp the cyclic group of order p.
Proposition 12.24. Extend the functor S 7→ M[S] to all profinite sets by using simplicial
resolutions by extremally disconnected sets. Assume the following condition:
Assumption 12.25. For all profinite sets S with Cp-action and fixed points S0 = SCp , the
natural map
M[S0]tCp
→ M[S]tCp
is an isomorphism.
Then φ induces a map of analytic rings (A, M) → (A, M).
Moreover, the assumption is satisfied in the following two situations:
(1) if (A, M) is an algebra over Fp,;
(2) if for all profinite sets S, M[S] is m-truncated for some m.
By (the version for animated rings of) [Sch19, Proposition 7.14], it is enough to construct
functorial φ-linear maps
φS : π0M[S] → π0M[S]
for all extremally disconnected sets S, subject to the condition that for any map f : T → A from
an extremally disconnected profinite set, letting e
f : M[T] → M[∗] be the induced extension, the
diagram
π0M[T]
φT
//
π0
e
f

π0M[T]
π0
e
f

π0M[∗]
φ
// π0M[∗]
commutes.
To do this, recall that for any abelian group M, there is a natural linear map
M → Ȟ0
(Cp, M ⊗ . . . ⊗ M) = cofib(Nm : (M ⊗ . . . ⊗ M)Cp → (M ⊗ . . . ⊗)Cp
),
where M ⊗ . . . ⊗ M (with p tensor factors) is equipped with the Cp-action permuting the tensor
factors cyclically, and we take the 0-th Tate cohomology of the Cp-action, i.e. the cofiber of the
norm map from the coinvariants to the invariants. This map is induced by the (non-additive) map
M → (M ⊗ . . . ⊗ M)Cp
: m 7→ m ⊗ . . . ⊗ m,
noting that it becomes additive after passing to the cofiber of the norm map.
Applying this to the condensed abelian group π0M[S], we get a natural map
π0M[S] → Ȟ0
(Cp, π0M[S] ⊗ . . . ⊗ π0M[S])
that we can compose with the Cp-equivariant multiplication map
π0M[S] ⊗ . . . ⊗ π0M[S] → π0M[Sp
]

to get a functorial map
π0M[S] → Ȟ0
(Cp, π0M[Sp
]).
Now note that there is a map π0M[S] → π0M[Sp] induced by the diagonal inclusion S ,→ Sp.
We would like to use that the map Ȟ0(Cp, π0M[S]) → Ȟ0(Cp, π0M[Sp]) is an isomorphism. As
Cp acts trivially on the p-torsion module π0M[S], this would give the desired map
π0M[S] → Ȟ0
(Cp, π0M[S]) = π0M[S].
But we have passed to π0 too early; we need to stay derived for a little longer to get the passage
from M[Sp] to M[S]. For this, recall that one can define Tate cohomology in general for a spectrum
X with Cp-action, as
XtCp
= cofib(Nm : XhCp → XhCp
)
the cofibre of the norm map from homotopy orbits to homotopy fixed points, cf. e.g. [NS18, Chapter
I]. The map M → Ȟ0(Cp, M ⊗ . . . ⊗ M) generalizes to the following. Let X be any spectrum in
the sense of stable homotopy theory, and consider X ⊗ . . . ⊗ X, the p-fold tensor product, with the
cyclic Cp-action. Then there is a unique functorial lax symmetric monoidal map
X → (X ⊗ . . . ⊗ X)tCp
,
the so-called Tate diagonal. (This map exists only for spectra, not in D(Ab).)
Now following the above, we get a functorial map
M[S] → (M[S] ⊗ . . . ⊗ M[S])tCp
→ M[Sp
]tCp
.
Again, we have the map M[S] → M[Sp] induced by the diagonal S → Sp, and now the assumption
guarantees that this induces an isomorphism after applying −tCp .
In total, we have produced a functorial map M[S] → M[S]tCp . Now is the time to pass to π0:
First, on the target, where we get a natural map to (π0M[S])tCp . In the resulting map
M[S] → (π0M[S])tCp
we take π0, which gives the desired functorial map
π0M[S] → π0M[S].
We need to see that this is linear over φ : π0A → π0A. But M[S] → M[Sp]tCp ∼
= M[S]tCp is linear
over the Tate-valued Frobenius A → AtCp by construction, and the Tate-valued Frobenius induces
the usual Frobenius on π0.
It remains to see that for any map f : T → A from some extremally disconnected T, inducing
a map e
f : M[T] → M[∗], the diagram
π0M[T]
φT
//
π0
e
f

π0M[T]
π0
e
f

π0M[∗]
φ
// π0M[∗]
commutes. For this, we note that φT has the property that the composite
π0M[T]
φT
−
−
→ π0M[T] → Ȟ0
(Cp, π0M[Tp
])

APPENDIX TO LECTURE XII: FROBENIUS 91
is the naive map φ0
T constructed in the beginning. Note that π0M[∗] → Ȟ0(Cp, π0M[∗]) is an
isomorphism. We see that it suffices to see that the diagram
π0M[T]
φ0
T
//
π0
e
f

Ȟ0(Cp, π0M[Tp])
π0
e
fp

π0M[∗]
φ
// Ȟ0(Cp, π0M[∗])
commutes, where e
fp : M[Tp] → M[∗] is the p-fold tensor product of e
f in D≥0(A, M). This is the
composite of the commutative diagram
π0M[T]
φ0
T
//
π0
e
f

Ȟ0(Cp, π0M[T] ⊗π0A . . . ⊗π0A π0M[T])
π0
e
f⊗p

π0M[∗]
φ
// Ȟ0(Cp, π0M[∗] ⊗π0A . . . ⊗π0A M[∗])
(by functoriality of M → Ȟ0(Cp, M ⊗A . . . ⊗A M)) and the diagram
Ȟ0(Cp, π0M[T] ⊗π0A . . . ⊗π0A π0M[T]) //
π0
e
f⊗p

Ȟ0(Cp, π0M[Tp])
π0
e
fp

Ȟ0(Cp, π0M[∗] ⊗π0A . . . ⊗π0A M[∗]) // Ȟ0(Cp, π0M[∗]).
For commutativity of this diagram, it suffices to prove that
π0M[T] ⊗π0A . . . ⊗π0A π0M[T] //
π0
e
f⊗p

π0M[Tp]
π0
e
fp

π0M[∗] ⊗π0A . . . ⊗π0A M[∗] // π0M[∗].
commutes. But this is a diagram in condensed π0A-modules, and the target π0M[∗] lies in
D≥0(A, M), so the result follows from e
fp being, by definition, the image of e
f⊗p in D≥0(A, M).
It remains to show that the assumption is satisfied if A is an algebra over Fp,, or if all M[S]
are truncated. For this, we note that we can write the cofiber of
M[S0]tCp
→ M[S]tCp
as the limit of
. . . → M[S][−2n]
β
−
→ . . .
β
−
→ M[S][−2]
β
−
→ M[S]
along a certain map β : M[S][−2] → M[S], where S is the quotient of the pointed profinite set
S/S0 by the free (in pointed profinite sets) Cp-action, and M is the extension of M to pointed
profinite sets (taking (T, ∗) to M[T]/M[∗]). This comes from a general formula for XtCp as a limit
of XhCp [−2n] when X ∈ D(Ab).
Moreover, by functoriality of the whole situation, the map β base changes. In the case of Fp,,
it is equal to 0 as then M[S] is projective for all profinite sets S; thus, it is also 0 for any (A, M)
over Fp,, and hence the limit above vanishes.

On the other hand, if M[S] is truncated, then the limit vanishes as the terms become more
and more coconnective.

APPENDIX TO LECTURE XII: NORMALIZATIONS OF ANALYTIC ANIMATED RINGS 93
Appendix to Lecture XII: Normalizations of analytic animated rings
The goal of this appendix is to indicate the proof of the following result.
Proposition 12.26. Let (A, M) be an analytic animated associative (resp. commutative) ring.
Consider the ∞-category of condensed animated associative (resp. commutative) rings B with a
map from A such that B ∈ D≥0(A, M) ⊂ D≥0(A). This has an initial object, whose underlying
object in D≥0(A) is M[∗].
First, we prove this in the associative case. We use that the ∞-category of condensed animated
associative rings B/A is monadic over the D≥0(A), as follows from the ∞-categorical version of the
Barr–Beck theorem [?, Theorem 4.7.0.3]. Here, the relevant left adjoint functor can be defined by
animation. More precisely, consider the category of pairs of maps of associative rings A → B, with
its forgetful functor to the category of pairs (A, M) of an associative ring A and an A-module M.
This admits a left adjoint, sending M to the free associative algebra
TA(M) = A ⊕ M ⊗Z A ⊕ M ⊗Z M ⊗Z A ⊕ . . . .
Animating this adjunction defines an adjunction between maps of animated associative rings A → B
and the ∞-category of pairs of an animated ring A with an animated A-module M. We can
then also pass to condensed objects, and to the fibre over the given condensed animated ring
A. This shows that the relevant monad T is given by a monad structure on the functor M 7→
A ⊕ M ⊗Z A ⊕ M ⊗Z M ⊗Z A ⊕ . . . on D≥0(A).
Now we observe that if M → N is a map in D≥0(A) that is sent to an isomorphism in D≥0(A, M)
under − ⊗A (A, M), then also T(M) → T(N) is sent to an isomorphism in D≥0(A, M), by the
formula for T, and using that the kernel of the localization is stable under − ⊗Z C for all C ∈
D≥0(Cond(Ab)). This implies formally that T descends to a monad TM on D≥0(A, M) such that
T-algebras whose underlying object lies in D≥0(A, M) are equivalent to TM-algebras. Moreover,
− ⊗A (A, M) defines a left adjoint to the fully faithful forgetful functor from TM-algebras to
T-algebras. This proves the desired result.
In the commutative case, we argue in the same way, but we observe that the relevant monad
structure is now induced from the functor
M 7→ A ⊕ M ⊕ Sym2
M ⊕ . . . ⊕ Symn
M ⊕ . . .
in the case of usual rings and modules. (See [Lur18, Section 25.2.2, Construction 25.2.2.6] for a
more careful discussion.) In particular, we note that if A is a condensed animated commutative
ring, then by animating Symn
: (A, M) 7→ (A, Symn
M), passing to condensed objects, and the
fibre over the condensed animated ring A, there is a natural functor Symn
on D≥0(A). In order to
argue as above, we need the following lemma.
Lemma 12.27. Let (A, M) be an analytic animated commutative ring. Let M → N be a map
in D≥0(A) that becomes an isomorphism after −⊗A (A, M). Then Symn
M → Symn
N becomes an
isomorphism after − ⊗A (A, M) for all n ≥ 0.
Proof. Let Q be the cofibre of M → N. By [Lur18, Construction 25.2.5.4], Symn
N admits
a filtration whose graded pieces are
Symn
M, Symn−1
M ⊗ Q, . . . , Symn−i
M ⊗ Symi
Q, . . . , Symn
Q.

(This can be proved by animating the corresponding statement for usual (finite projective) mod-
ules.) Thus, it suffices to see that for all Q in the kernel K of − ⊗A (A, M), also Symi
Q lies in K
for all i 0.
We argue by induction on i. Then we see by induction, using the above filtration, that the
composite functor F : Q 7→ Symi
Q ⊗A (A, M) from K to D≥0(A, M) commutes with cofibres. In
particular, F(Q[n]) ∼
= F(Q)[n] for all n, or
Symi
Q ⊗A (A, M) ∼
= Symi
(Q[n])[−n] ⊗A (A, M)
for all n. Passing to the colimit over n, we note that the functor Q 7→ colimn Symi
(Q[n])[−n] is
an exact endofunctor of D≥0(A). One can show that this functor vanishes unless i is power pm of
a prime p (the first “Goodwillie derivative of Symi
”, in this situation going back to Dold–Puppe,
[DP61]), and in that case it is filtered by copies of the functor Q 7→ Q ⊗A A/p where the map
A → A/p is the Frobenius a 7→ apm
. Now the result follows from the definition of analytic animated
commutative rings.

13. LECTURE XIII: ANALYTIC SPACES 95
13. Lecture XIII: Analytic spaces
Finally, we define analytic spaces. In this lecture, we use the term analytic ring to refer to
objects of AnRing, i.e. analytic animated commutative rings (A, M) with A = M[∗].28
Consider the ∞-category of functors from AnRing to anima, i.e. the ∞-category of presheaves
of anima on AnRingop
. By the Yoneda lemma, this admits AnRingop
as a full subcategory, via the
functor
(A, M) 7→ AnSpec(A, M) : (B, N) 7→ HomAnRing((A, M), (B, N)).
We endow AnRingop
with the Grothendieck topology generated by the families of maps
{AnSpec(Ai, Mi) → AnSpec(A, M)}
whenever (A, M) → (Ai, Mi) is a finite family of steady localizations such that
D≥0(A, M) →
Y
i
D≥0(Ai, Mi)
is conservative. We note that pullbacks of such finite families of maps are of the same form.
Proposition 13.1. For any analytic ring (A, M), the functor AnSpec(A, M) is a sheaf on
AnRingop
.
Proof. This follows directly from Proposition 12.18.
If f : F → G is a map of functors from AnRing to anima, we write F ⊂ G if for all analytic
rings (A, M), all fibres of
F(A, M) → G(A, M)
are either empty or contractible.
Definition 13.2. Let (A, M) be an analytic ring. A steady subspace of AnSpec(A, M) is a
subfunctor U ⊂ AnSpec(A, M) such that the natural map
colimAnSpec(A0,M0)⊂U AnSpec(A0
, M0
) → U
is an isomorphism, where the colimit runs over all steady localizations (A, M) → (A0, M0) for
which AnSpec(A0, M0) ⊂ AnSpec(A, M) factors over U.
Remark 13.3. As the colimit is taken over a non-filtered index category, one might worry
that it may produce undesired outcomes. This is not so: Compute the colimit first as presheaves.
Then for all analytic rings (B, N) with a fixed map to U, the fibre product computes the colimit
of a point over all AnSpec(A0, M0) contained in U and such that AnSpec(B, N) → U factors over
AnSpec(A0, M0). If nonempty, index category is cofiltered (as with any two steady localizations,
it contains their product), so the geometric realization of the index category is contractible, and
hence this colimit is itself a point or empty. Now note that if G is a sheaf of anima and F ⊂ G
is a presheaf of anima (the inclusion meaning that all fibres are empty or contractible), then the
sheafification F0 of F still has the property that F0 → G has empty or contractible fibres; it is the
subsheaf of G of all sections that locally lift to F. To see this, note that sheafification commutes
with taking fibres, so it is enough to recall that the sheafification of a presheaf whose values are
empty or a point has values which still are empty or a point.
28One can consider a variant using analytic connective E∞-rings instead.

This argument shows that U ⊂ AnSpec(A, M) is a steady subspace if and only if there is some
collection (A, M) → (Ai, Mi)i of steady localizations such that U is the image of
G
i
AnSpec(Ai, Mi) → AnSpec(A, M)(B, N).
One verifies immediately that pullbacks of steady subspaces are steady, and conversely U ⊂
AnSpec(A, M) is a steady subspace if and only the pullbacks Ui ⊂ AnSpec(Ai, Mi) are steady
subspaces for all terms in a finite cover of (A, M) by steady localizations (Ai, Mi). Thus, the
following definition has reasonable properties.
Definition 13.4. An inclusion F ⊂ G of sheaves of anima on AnRing is a steady subspace if
for all analytic rings (A, M) the fibre product
F ×G AnSpec(A, M) → AnSpec(A, M)
is a steady subspace.
Finally, we can define analytic spaces.
Definition 13.5. An analytic space is a sheaf of anima X on AnRing such that the map
colimAnSpec(A,M)⊂X AnSpec(A, M) → X
is an isomorphism, where the colimit runs over all affine steady subspaces AnSpec(A, M) ⊂ X.
Again, one might worry that the colimit does not behave well as the index category is not well-
behaved. However, we have the following result, saying that we could equivalently define analytic
spaces X as those sheaves that can be covered by steady subspaces of the form AnSpec(A, M) ⊂ X.
Proposition 13.6. If X is a sheaf of anima on AnRing such that the map
colimAnSpec(A,M)⊂X AnSpec(A, M) → X
is surjective, where as above the colimit runs over all affine steady subspaces AnSpec(A, M) ⊂ X,
then X is an analytic space, i.e. the map is an isomorphism.
In particular, any steady subspace of an analytic space is again an analytic space.
Proof. Assume first that X admits an injection into some affine e
X = AnSpec( e
A, f
M); let us
call such X quasi-affine for the rest of this proof. In that case, fibres products over X can also be
computes as fibre products over e
X. For any (B, N) mapping to X, the fibre of the displayed map
is computed by the colimit over a point over the index category of all AnSpec(A, M) ⊂ X over
which AnSpec(B, N) → X, which will be cofiltered in this case. Thus, this colimit is either empty
or a point, which gives the result in this case.
At this point, we already see that any steady subspace of some quasi-affine X is again an
analytic space (and clearly quasi-affine). Now let X be general and consider the colimit
colimY ⊂X Y → X
where the colimit runs over all quasi-affine analytic spaces Y that are steady subspaces of X. By
the same argument as in the first paragraph, this map is an isomorphism. Finally, the map
colimAnSpec(A,M)⊂X AnSpec(A, M) → colimY ⊂X Y

is an isomorphism, by comparing both with
colimAnSpec(A,M)⊂Y ⊂X AnSpec(A, M) :
This is equal to the first, as the set of possible Y is cofiltered, and it is equal to the second as each
Y can be written as colimAnSpec(A,M)⊂Y AnSpec(A, M).
By Proposition 12.18, we can define an ∞-category of quasicoherent sheaves on X.
Definition 13.7. Let X be an analytic space. The ∞-category of quasicoherent sheaves on X
is
D(X) := lim
AnSpec(A,M)⊂X
D(A, M).
Equivalently, the association (A, M) 7→ D(A, M) defines a sheaf of ∞-categories on AnRingop
,
and D(X) is the value of this sheaf on X. In particular:
Proposition 13.8. Let X be an analytic space and let Ui ⊂ X, i ∈ I, be steady subspaces that
cover X (i.e.
F
i Ui → X is a surjection). Let CI be the category of finite nonempty subsets of I,
and consider the functor taking J ∈ CI to UJ =
T
i∈J Ui. Then
D(X) → lim
J∈CI
D(UJ )
is an equivalence of ∞-categories.
If X = AnSpec(A, M) is affine, then D(X) ∼
= D(A, M).
We see that there is a well-defined ∞-category of analytic spaces, defined very analogously to
schemes, and it comes with an ∞-category of quasicoherent sheaves, satisfying descent.
It is high time for some examples. First, we need some examples of steady localizations. All of
them will be produced by the following criterion.
Proposition 13.9. Let f : (A, M) → (B, N) be a map of analytic rings. Assume that for any
extremally disconnected profinite set S and any M ∈ D≥0(A, M), the natural map
(HomD(A)(A[S], M) ⊗(A,M) (B, N))(∗) → (M ⊗(A,M) (B, N))(S)
in D(Ab) is an equivalence. Then f is steady.
Proof. Consider any analytic ring (C, MC) over (A, M) and let M ∈ D≥0(C, MC). Then
(M⊗(A,M)(B, N))(S) = (HomD(A)(A[S], M)⊗(A,M)(B, N))(∗) = (HomD(C,MC)(MC[S], M)⊗(A,M)(B, N))(∗).
Varying S, the right-hand side of this formula defines a functor from the ∞-category of C-modules
isomorphic to some MC[S] towards anima, commuting with finite products, and thus an object of
D≥0(C, MC), that necessarily agrees with M ⊗(A,M) (B, N).
An interesting question is whether an induced analytic ring structure is steady. This leads to
the following class of modules. Here and in the following, we write
P∨
= HomD(A)(P, A) ∈ D(A)
for P ∈ D(A).

Definition 13.10. Let (A, M) be an analytic ring. An object C ∈ D(A, M) is nuclear if for
all extremally disconnected profinite sets S, the natural map
(A[S]∨
⊗(A,M) C)(∗) → C(S)
in D(Ab) is an isomorphism.
Let us characterize this class of modules. For the characterization, the following class of maps
is critical.
Definition 13.11. Let (A, M) be an analytic ring. A map f : P → Q between compact objects
of D(A, M) is trace-class if there is some map
g : A → P∨
⊗(A,M) Q
such that f is the composite
P
1⊗g
−
−
→ P ⊗(A,M) P∨
⊗(A,M) Q → Q
where the second map contracts the first two factors.
Definition 13.12. Let (A, M) be an analytic ring. An object C ∈ D(A, M) is basic nuclear
if it can be written as a sequential colimit
C = colim(P0 → P1 → . . .)
of compact Pi ∈ D(A, M) along trace-class maps.
Proposition 13.13. Let (A, M) be an analytic ring and let C ∈ D(A, M). Then C is nuclear
if and only if it can be written as a filtered colimit of basic nuclear objects. The class of basic
nuclear objects is stable under all countable colimits, and the class of nuclear objects is stable under
all colimits.
Proof. Assume first that C is basic nuclear, so C is a colimit of compact Pi along trace-
class maps. Choose maps A → P∨
i ⊗A Pi+1 giving rise to these trace-class maps Pi → Pi+1. Let
Q = A[S]. Then nuclearity means that (Q∨ ⊗(A,M) C)(∗) → HomD(A)(Q, C) is an isomorphism (in
D(Ab)). Both sides commute with colimits in C, and there are natural backwards maps
HomD(A)(Q, Pi) → (HomD(A)(Q, Pi) ⊗(A,M) P∨
i ⊗(A,M) Pi+1)(∗) → (Q∨
⊗(A,M) Pi+1)(∗),
showing that C is nuclear.
Next, we show that the class of basic nuclear C is stable under finite colimits; equivalently,
under passage to cones. Let f : C → C0 be a map of basic nuclear objects, and write C resp. C0 as
a filtered colimit of Pi resp. P0
i as in the definition. As all Pi are compact, the map C → C0 can be
realized via compatible maps Pi → P0
i , up to reindexing the P0
i . Let Qi be the cone of Pi → P0
i , so
the cone of f is the sequential colimit of the Qi. We claim that Qi → Qi+2 is trace-class for all i. To
see this, we need to find a map A → Q∨
i ⊗(A,M) Qi+2 defining this map. Note that Q∨
i ⊗(A,M) Qi+2
is the fibre of
(P0
i )∨
⊗(A,M) Qi+2 → P∨
i ⊗(A,M) Qi+2
and to define a map there, it suffices to exhibit a section of (P0
i )∨ ⊗(A,M) P0
i+2 and a section of
P∨
i ⊗(A,M) Pi+2 whose images in P∨
i ⊗(A,M) P0
i+2 agree. To define this, pick sections
α : A → P∨
i ⊗(A,M) Pi+1 , β : A → (P0
i+1)∨
⊗(A,M) P0
i+2

witnessing that Pi → Pi+1 and P0
i+1 → P0
i+2 are trace-class, and take the induced sections of
P∨
i ⊗(A,M) Pi+2 and P∨
i ⊗(A,M) P0
i+2.
To see stability of basic nuclear C under all countable colimits, it remains to observe that it is
stable under countable direct sums, which is clear from the definition.
It is clear from the definition that the class of all nuclear objects is stable under all colimits.
Now let C be a general nuclear object. We can look at the ω1-filtered diagram of all basic nuclear
C0 mapping to C (here we use that basic nuclear objects admit countable colimits), and replacing
C by the cone of the colimit of this diagram mapping to C, we can assume that there are no maps
from basic nuclear C0’s to C. We need to see that then C = 0. If not, there is some compact P0
with a nonzero map P0 → C. As by nuclearity,
(P∨
0 ⊗A C)(∗) = Hom(P0, C),
and both sides commute with filtered colimits as functors of C, we can find some compact P1 → C
and a section of (P∨
0 ⊗A P1)(∗) so that P0 → C is the composite of P1 → C with the corresponding
trace-class map P0 → P1. Continuing, we find a nonzero map from some basic nuclear colim(P0 →
P1 → . . .) to C, contradiction.
The proposition implies that nuclear modules have stronger properties than a priori guaranteed
by the definition:
Proposition 13.14. Let (A, M) be an analytic ring and let C ∈ D(A, M) be nuclear. Then
for all M ∈ D(A, M) and extremally disconnected profinite sets S, the map
(HomD(A))(A[S], M) ⊗(A,M) C)(∗) → (M ⊗(A,M) C)(S)
in D(Ab) is an isomorphism.
In particular, if A → B is a map of condensed animated rings such that B ∈ D(A, M) is
nuclear, and (B, N) is the induced analytic ring structure on B, then (A, M) → (B, N) satisfies
the criterion from Proposition 13.9, and hence is steady.
Proof. As both sides commute with colimits in C, we can assume that C is basic nuclear, and
write it as the sequential colimit of compact Pi along trace-class maps Pi → Pi+1. In that case,
there are backwards maps
(M ⊗(A,M) Pi)(S) = HomD(A)(A[S], M ⊗(A,M) Pi)(∗)
→ (HomD(A)(A[S], M ⊗(A,M) Pi) ⊗(A,M) P∨
i ⊗(A,M) Pi+1)(∗)
→ (HomD(A)(A[S], M ⊗(A,M) Pi ⊗(A,M) P∨
i ) ⊗(A,M) Pi+1)(∗)
→ (HomD(A))(A[S], M) ⊗(A,M) Pi+1)(∗)
and so passing to the filtered colimit over the Pi gives the desired equivalence.
The final sentence follows directly.
We will now verify this condition in the examples listed two lectures ago.
Example 13.15. The following are steady localizations of analytic rings.
(1) If A → B = A[S−1] is a usual localization of (discrete) rings, with trivial analytic ring
structure, then A → B is a steady localization. Indeed, B, like any discrete module, is
nuclear over A.

(2) The map (Z[T], Z) → (Z[T], Z[T]), corresponding to localization to the subset {|T| ≤ 1}
of the adic space Spa(Z[T], Z). Recall that in this case (cf. [Sch19, Lecture VII]),
M ⊗(Z[T],Z)
(Z[T], Z[T]) = HomD((Z[T],Z))((Z((T−1
))/Z[T])[−1], M).
Then
(HomD(Z[T])(Z[T][S], M)⊗(Z[T],Z)
(Z[T], Z[T]))(∗) = HomD(Z[T])((Z((T−1
))/Z[T])[−1]⊗ZZ[S], M)
while also
(M ⊗(Z[T],Z)
(Z[T], Z[T]))(S) = HomD((Z[T],Z))((Z((T−1
))/Z[T])[−1] ⊗Z Z[S], M),
giving the desired result.
(3) Any rational subset of an adic space is of the form U = {|fi| ≤ |g| 6= 0} for finitely many
elements f1, . . . , fn, g ∈ A (satisfying some condition). In the present situation of analytic
rings, we can realize this localization by first inverting g, and then asking the condition
|fi
g | ≤ 1. Inverting g is handled by a base change of (1), while asking |fi
g | ≤ 1 is handled
by a base change of (2).
(4) If D1 ⊂ D2 is an inclusion of closed discs in C, then O(D2) → O(D1) (endowed with the
analytic ring structure induced from the p-liquid structure on C for some chosen 0 p ≤ 1)
is a steady localization. This follows from nuclearity of O(D2) → O(D1).
Regarding (3), let us at least recall how general Huber pairs fit into the present framework.
Proposition 13.16. There is a fully faithful functor (A, A+) 7→ (A, A+) from the category
of Huber pairs (A, A+) to the ∞-category AnRing, defined as follows. For a finitely generated
Z-algebra R, let R be the analytic ring structure on R given by
R[S] = lim
←
−
i
R[Si]
for a profinite set S = lim
←
−i
Si. For a general discrete Z-algebra R, let R = lim
−
→R0→R
R0
, where the
colimit is over all finitely generated Z-algebras R0 mapping to R. Finally,
(A, A+
) = A ⊗A+
disc
(A+
disc)
is an induced analytic ring structure.
Remark 13.17. A priori, this induced analytic ring structure might be animated. However,
Andreychev has proved that (A, A+)[S] is always 0-truncated.
The analytic ring (A, A+) is always normalized; this follows from A being a module over
(A+
disc), which can be deduced by writing it as the limit of A/In where I ⊂ A0 ⊂ A is an ideal of
definition that is also an R-module, where R ⊂ A+ is any chosen finitely generated subalgebra.
Remark 13.18. The second part of a Huber pair is an open and integrally closed subring
A+ ⊂ A consisting of powerbounded elements. One may wonder how these precise conditions
arise. One answer is the following: Consider all possible normalized analytic ring structures on
A admitting a map from Z and that can be obtained via filtered colimits from pushouts of the
localization (Z[T], Z) → (Z[T], Z[T]). Then this class is canonically in bijection with the open
and integrally closed subrings A+ ⊂ A contained in the powerbounded elements. Given A+ ⊂ A,
one allows the localization via the pushout of (Z[T], Z) → (Z[T], Z[T]) along any map Z[T] → A
sending T to an element of A+.

Proof. That the functor is well-defined follows from the results of [Sch19] and the result on
induced analytic ring structures. For fully faithfulness, let (A, A+) and (B, B+) be Huber pairs.
By definition of maps of analytic rings, maps from (A, A+) to (B, B+) embed into maps from A
to B. As A and B are metrizable, these agree with maps from A to B. We need to see that a map
A → B sends A+ into B+ if and only if it induces a map of analytic rings (A, A+) → (B, B+).
This is clear in the forward direction.
Thus, assume that f : A → B is a map of Huber rings that induces a map of analytic rings
(A, A+) → (B, B+), but does not send A+ into B+. Take any g ∈ A+ such that f(g) 6∈ B+. We
get a map (Z[T], Z[T]) → (A, A+) sending T to g; precomposing with this map, we may assume
that (A, A+) = (Z[T], Z[T]). Note that the localization from (Z[T], Z) to (Z[T], Z[T]) is obtained
by killing the algebra object Z((T))−1. Thus, our assumption is that
Z((T−1
)) ⊗(Z[T],Z)
(B, B+
) = 0.
As asking whether an algebra is just asking whether 1 = 0 in this algebra, and if this happens in
a filtered colimit, it happens at a finite stage, we can then find some finitely generated subalgebra
R ⊂ B+ such that
Z((T−1
)) ⊗(Z[T],Z)
(B, R) = 0.
Using the resolution
0 →
Y
N
Z ⊗Z Z[T]
1−shift⊗T
−
−
−
−
−
−
→
Y
N
Z ⊗Z Z[T] → Z((T−1
)) → 0,
this means that the map
Y
N
R ⊗R
B
1−shift⊗g
−
−
−
−
−
−
→
Y
N
R ⊗R
B
is an isomorphism. We need to see that this implies that g is integral over R.
In other words, we now start with a finitely generated Z-algebra R, some Huber R-algebra B,
and an element g ∈ B. Choose a ring of definition B0 ⊂ B containing R and some ideal of definition
I ⊂ B0. Then in particular B = limn B/In as condensed R-modules. We want to show that if
Y
N
R ⊗R
B
1−shift⊗g
−
−
−
−
−
−
→
Y
N
R ⊗R
B
is an isomorphism, then g is integral over R + I (which is a subring of B and contained in B+).
For any n, Y
N
R ⊗R B/In
= lim
−
→
M⊂B/In
Y
N
R ⊗R M = lim
−
→
M⊂B/In
Y
N
M ⊂
Y
N
R/In
where the colimit runs over finitely generated R-submodules M of B/In. We see that the inverse
limit over all n maps to
Q
N B, and the image is contained in sequences (b0, b1, . . .) such that for
all n, the submodule of B/In generated by all bi is finitely generated.
If the map is surjective, then in particular 1 is in the image, which means that we can find such
a sequence (b0, b1, . . .) in B satisfying b0 = 1, b1 − gb0 = 0, b2 − gb1 = 0, etc., i.e. bi = gi. This
sequence needs to have the property that the R-submodule of B/I generated by all gi is finitely
generated. But this means that there is some relation
gn
+ rn−1gn−1
+ . . . + r0 = 0 ∈ B/I,
which as desired shows that g is integral over R + I.

In the context of adic spaces, one can show that nuclearity is closely related to adic maps;
for example, if A → B is a map of Tate-Huber rings (i.e. Huber rings admitting a topologically
nilpotent unit), then B is always a nuclear A-module, for any choice of A+. On the other hand, if
A → B is not adic, then B is never nuclear as A-module, and in fact one can show that for any
choices of A+ mapping to B+, the map (A, A+) → (B, B+) will not be steady.

14. LECTURE XIV: VARIA 103
14. Lecture XIV: Varia
In this final lecture, we briefly outline a few more ideas. We do not give proofs here; however,
they are not especially difficult. (The hard work was in obtaining the examples of the solid analytic
ring structure on Z last semester, and the liquid analytic ring structures on R this semester.)
First, one may wonder whether one can define analytic spaces more in the spirit of locally ringed
topological spaces. This is possible. If X is an analytic space, then we can consider the site |X| of
steady subspaces U ⊂ X of X. This forms a locale. If X = AnSpec(A, M) is affine, then a basis is
given by the affine steady subspaces, which are quasicompact and quasiseparated. Thus, Deligne’s
theorem on the existence of points implies that, if there is only a set (instead of a proper class) of
affine steady subspaces of X, then |X| can be identified with the site of open subsets of a spectral
space. For general (non-affine) X, it would then be the case that |X| is a locally spectral space.
Moreover, |X| carries a sheaf OX of condensed animated commutative rings (obtained by sheafi-
fying its value on affine subspaces), as well as a sheaf MX[S] of condensed animated OX-modules,
for every extremally disconnected profinite set S (functorial in S, taking finite disjoint unions to
finite direct sums). We want to define what “locally ringed” means, i.e. we want an analogue of
the condition that the stalks of OX are local.
Definition 14.1. A locally analytically ringed locale is a triple (X, OX, MX) consisting of a
locale X, a sheaf of condensed animated commutative rings OX on X, and a functor MX : S 7→
MX[S] from extremally disconnected profinite sets S to condensed animated OX-modules, taking
finite disjoint unions to finite direct sums, subject to the following conditions:
(1) (“analytically ringed”) there is a basis of subsets U ⊂ X for which (OX(U), MX(U)) is
an analytic ring; on this basis, the restriction maps are maps of analytic rings;
(2) (“stalks are local”) for every U ⊂ X with a map (A, M) → (OX(U), MX(U)) from some
analytic ring (A, M), and every steady cover
G
i
AnSpec(Ai, Mi) → AnSpec(A, M)
there is a cover of U by Ui ⊂ U such that (A, M) → (OX(Ui), MX(Ui)) factors over
(Ai, Mi).
A map f : (Y, OY , MY ) → (X, OX, MX) of locally analytically ringed locales is a map f : Y →
X of locales together with a map f∗OX → OY of sheaves of condensed animated commutative
rings, with the following properties.
(1) (“analytically ringed map”) If V ⊂ Y maps into U ⊂ X, then (OX(U), MX(U)) →
(OY (V ), MY (V )) is a map of analytic rings whenever both are.
(2) (“maps on local rings local”) Moreover, in the same situation, if there is some steady
localization
(A, M) = (OX(U), MX(U)) → (A0
, M0
)
over which (A, M) → (OY (V ), MY (V )) factors, then one can find U0 ⊂ U containing the
image of V such that (A, M) → (OX(U0), MX(U0)) factors over (A0, M0).
It is now not hard to prove the following proposition.
Proposition 14.2. The functor X 7→ (|X|, OX, MX) from analytic spaces to locally analytically
ringed locales is well-defined and fully faithful. An object lies in the essential image if and only

if it is locally isomorphic to the locally analytically ringed locale associated to some analytic ring
(A, M).
Using this language, we can explain how to embed complex-analytic spaces into analytic spaces
over C equipped with its p-liquid analytic ring structure (C, Mp) (for any fixed 0 p ≤ 1).
Proposition 14.3. Let (X, OX) be a complex-analytic space; so X is a topological space and
OX is a sheaf of condensed C-algebras, locally Zariski closed in an open n-dimensional complex ball.
Consider the functor taking any analytic space Y over (C, Mp) to maps of locales f : |Y | → X
together with a map of sheaves of condensed animated C-algebras f∗OX → OY . This functor is
representable by an analytic space Xan
p over AnSpec(C, Mp).
Let us give an example: Let X = {z | |z| 1} be the open unit disc in C. Let Dr = {z |
|z| ≤ r} ⊂ X be the closed discs of radius r 1. Let O(Dr) be the condensed C-algebra of
overconvergent holomorphic functions on Dr. This is a nuclear algebra over (C, Mp). Endow
O(Dr) with the analytic ring structure induced from (C, Mp). Then Xan
p is the increasing union
of AnSpec O(Dr) over all r 1.
In fact, similar results hold true for real-analytic, smooth, or topological manifolds: If (X, OX) is
one such, then the functor taking an analytic space Y over (R, Mp) to maps of locales f : |Y | → X
together with a map of sheaves of condensed animated R-algebras f∗OX → OY is representable
by an analytic space Xan
p over AnSpec(R, Mp). The explicit description of the functor is similar,
using the affine analytic spaces associated with overconvergent real-analytic, smooth, or continuous
functions on a closed ball.
Warning 14.4. In the topological case, the functor does not land in steady analytic spaces
over AnSpec(R, Mp), and the functor does not commute with products (it does in the other
cases). The issue is that for example C(S1, R) ⊗L
(R,Mp) C(S1, R) is not given by C((S1)2, R) (and
non-steadyness comes from non-nuclearity of C(S1, R)).
We see that analytic spaces over AnSpec(R, Mp) are able to handle all flavours of real or
complex manifolds simultaneously. In fact, nothing is holding one back from even mixing them,
taking products of real-analytic manifolds with topological manifolds, etc. . Even more, one can
also mix with algebraic varieties:
Proposition 14.5. For any analytic ring (A, M), there is a fully faithful functor X 7→ Xan
from schemes X over the (animated) ring A(∗) to analytic spaces over AnSpec(A, M). In fact,
Xan represents the functor taking an analytic space Y over AnSpec(A, M) to maps f : |Y | → X
together with a map f∗OX → OY of sheaves of animated A(∗)-algebras that induces local maps on
local rings. Moreover, there is a fully faithful functor Dqc(X) ,→ D(Xan).
Remark 14.6. The analytification of A1
C in this sense is AnSpec(C[T], Mp[T]), taking the
usual ring of polynomials over C with the condensed structure and the analytic structure induced
from (C, Mp). This is not what is usually regarded as the analytification of A1
C, which would be
the increasing union of open balls of radius r around the origin, letting r → ∞. In the present
situation, this is still a steady subspace of AnSpec(C[T], Mp[T]). The latter is however larger,
and contains points “infinitesimally close to infinity”.
We see that the present framework of analytic spaces makes it possible to have analytic
spaces that are globally algebraic but locally analytic – in the sense that the global functions

of AnSpec(C[T], Mp[T]) are just the polynomials C[T], but locally one gets certain algebras of
convergent functions.
Let us now turn to adic spaces.
Proposition 14.7. Let (A, A+) be a Huber pair and X = AnSpec((A, A+)). The underlying
locale |X| maps to the topological space Spa(A, A+) defined by Huber. Moreover, for any rational
subspace U ⊂ Spa(A, A+) with preimage U0 ⊂ X, the map
(A, A+
) → (OX(U0
), MX(U0
))
extends uniquely to an isomorphism
(OSpa(A,A+)(U), O+
Spa(A,A+)
(U))
∼
= (OX(U0
), MX(U0
))
in the following cases:
(1) if A is discrete;
(2) if A admits a noetherian ring of definition A0 over which it is finitely generated;
(3) if A is Tate and sheafy.
Remark 14.8. Let us quickly indicate the construction of the map |X| → Spa(A, A+). An ele-
ment a ∈ A is defined to be ≤ 1 on some affine subspace U ⊂ X if and only if the (OX(U), MX(U))-
module Z((T−1)) ⊗Z[T] (OX(U), MX(U)) vanishes, where T is mapped to a. Extending this defi-
nition to local fractions a
b , this defines a valuation on A locally on X. This valuation is however
a priori not continuous. Let Spv(A, A+) be the spectral space of all (not necessarily continu-
ous) valuations | · | on A with |A+| ≤ 1 and |A◦◦| 1. Then Spa(A, A+) ⊂ Spv(A, A+) and
there is a retraction, cf. [Hub93, Proposition 2.6, Theorem 3.1]. Now the construction above
gives a natural map |X| → Spv(A, A+), functorial in (A, A+). Composing with the retraction
Spv(A, A+) → Spa(A, A+) gives the desired map |X| → Spa(A, A+).
We warn the reader that the retraction Spv(A, A+) → Spa(A, A+) is not natural in (A, A+) for
non-adic maps (A, A+) → (B, B+) (and hence neither is |X| → Spa(A, A+)). For example, pulling
back
Spa(Z[T±1
], Z) ⊂ Spa(Z[T], Z)
to Spa(QphTi, ZphTi), one gets a punctured unit disc, which is not quasicompact anymore. In
the context of analytic spaces, the pullback would be given by an analytic ring structure on
QphTi[T−1], which is not a Huber ring anymore. Relatedly, there are points of the analytic
space AnSpec((QphTi, ZphTi)) that are “infinitesimally close to the origin”. These will map in
Spa(QphTi, ZphTi) to the origin, while their image in AnSpec((Z[T], Z)) will not map to the origin
of Spa(Z[T], Z).
A different way to understand the situation is to regard only Spv(A, A+) and the map |X| →
Spv(A, A+) as fundamental, and then observe that Huber was only able to define the structure
on those rational subsets of Spv(A, A+) that arise via pullback from Spa(A, A+), as only on those
rational subsets one (usually) gets analytic rings corresponding to Huber pairs again.
The proposition says that in cases (1) – (3), Huber has defined the “correct” structure sheaf.
From the proposition, one sees that there is a functor from adic spaces glued out of Spa(A, A+)’s
with A satisfying one of the given conditions, to analytic spaces.

One can also use the results of these lectures to recover (and slightly generalize) the results on
gluing of vector bundles (or coherent sheaves), cf. e.g. [KL19]. For this, note that from Proposi-
tion 12.18 one formally gets descent for dualizable objects. Now we have the following result.
Theorem 14.9. Let (A, A+) be a Huber pair. Then the dualizable objects in D((A, A+)) are
equivalent to the perfect complexes of A-modules. A similar result holds true for the analytic spaces
associated to topological, smooth or real-analytic manifolds, or complex-analytic spaces: taking an
algebra A of overconvergent continuous, smooth, real-analytic, or complex-analytic functions on
some ball, the dualizable objects of D((R, Mp) ⊗R A) are equivalent to the perfect complexes of
A(∗)-modules.
Generally, for any analytic ring (A, M), the dualizable objects in D(A, M) are generated (under
cones, shifts, and retracts) by the cones of 1 − f on M[S], ranging over extremally disconnected
profinite sets S and trace-class maps f : M[S] → M[S].
Remark 14.10. One can show (for any analytic ring (A, M)) that for perfect complexes the
perfect amplitude descends, giving in particular descent for vector bundles.
Let us end these lectures by giving one somewhat funny example of an analytic ring; it is
essentially the Novikov ring appearing in symplectic geometry.
For any 0 r 1, consider the ring
Z((TR
))r = {
X
x∈R
axTx
|
X
x
|ax|rx
∞}, 29
equipped with the following condensed ring structure. Write
Z((TR
))r =
[
c0
Z((TR
))r,≤c
where
Z((TR
))r,≤c = {
X
x∈R
axTx
|
X
x
|ax|rx
≤ c},
which we give the structure of a compact Hausdorff space by writing it as a quotient of the compact
Hausdorff space
{(x+
1 , x−
1 , x+
2 , x−
2 , . . .) | x+
i , x−
i ∈ [(log(c) − log(i))/ log r, ∞],
X
i
(rx+
i + rx−
i ) ≤ c}
via
(x+
1 , x−
1 , x+
2 , x−
2 , . . .) 7→
X
i
(Tx+
i − Tx−
i ).
In particular, the map R ∪ {∞} → Z((TR))r : x 7→ Tx defines a map of condensed sets, and hence
0 = T∞ is connected to 1 = T0 in Z((TR))r. We note that Z((TR))r has a rather peculiar condensed
structure, in that a null-sequence x0, x1, . . . need not go to 0 in the ≤ c-sense: For example, the
sequence
1 − T, 1 − T1/2
, . . . , 1 − T1/n
, . . .
converges to 1 − T0 = 1 − 1 = 0, so is a null-sequence, but none of the terms lies in Z((TR))r,≤1.
29In particular, at most countably many ax are nonzero.

Again, we can define similarly for any finite set S a subset
Z((TR
))[S]r,≤c ⊂ Z((TR
))[S]r,
and then define for a profinite set S = lim
←
−i
Si
M(S, Z((TR
))r)≤c = lim
←
−
i
Z((TR
))[Si]r,≤c,
and
M(S, Z((TR
))r) =
[
c0
M(S, Z((TR
))r)≤c.
Finally, we also define
Z((TR
))r =
[
r0r
Z((TR
))r0
and the same for a profinite set S.
Note that there is an action of the condensed semigroup R≥1 on Z((TR))r via T 7→ Tt for
t ∈ R≥1. Restricted to a prime p ∈ R≥1, this action defines a Frobenius lift, so Z((TR))r is a novel
kind of λ-ring where the individual Frobenius lifts combine into an action of R≥1. Note that the
R≥1-action actually also induces canonical isomorphisms between Z((TR))r for different values of
r, so this ring is actually independent of the choice of r (up to canonical isomorphism).
Theorem 14.11. This construction defines an analytic ring structure on Z((TR))r. In fact,
it is induced from the analytic ring structure on Z((T))r via base change along Z[T] → Z[TR].
Analytic geometry over Z((TR))r is then some funny crossover between arithmetic geometry
and complex or real-analytic geometry, allowing actions of connected groups like R≥1 in character-
istic p; we believe it is unlike anything that has been studied.
What does the corresponding analytic space look like? It maps at least to the Berkovich
spectrum of Z:
Proposition 14.12. There is a natural map from | AnSpec(Z((T))r)| to the Berkovich spec-
trum M(Z) of Z, of all multiplicative norms | · | : Z → R≥0. Restricting to | AnSpec(Z((TR))r)|,
the map
| AnSpec(Z((TR
))r)| → M(Z)
is equivariant for the action of R≥1 (acting via rescaling the norm on M(Z)).
In this picture, the various theories (R, Mp) map to the point p ∈ (0, 1] on the half-line
corresponding to real valuations. In particular, taking the limit for p → 0 makes one enter the
arithmetic part of the Berkovich space M(Z), so we see that the analytic ring structures on R are
inextricably linked to arithmetic, which arises in the limit p → 0.

Q
F2
F3
F5
Fp
Q2
Q3
Q5
Qp
(R, M0.5)
(R, M1)
(R, M0.2)
Figure 1. The Berkovich space M(Z)

Bibliography
[ASC+
16] A. Avilés, F. C. Sánchez, J. M. F. Castillo, M. González, and Y. Moreno, Separably injective Banach
spaces, Lecture Notes in Mathematics, vol. 2132, Springer, [Cham], 2016.
[BBK17] O. Ben-Bassat and K. Kremnizer, Non-Archimedean analytic geometry as relative algebraic geometry, Ann.
Fac. Sci. Toulouse Math. (6) 26 (2017), no. 1, 49–126.
[Bei07] A. Beilinson, Topological -factors, Pure Appl. Math. Q. 3 (2007), no. 1, Special Issue: In honor of Robert
D. MacPherson. Part 3, 357–391.
[BK72] A. K. Bousfield and D. M. Kan, Homotopy limits, completions and localizations, Lecture Notes in Math-
ematics, Vol. 304, Springer-Verlag, Berlin-New York, 1972.
[BS15] B. Bhatt and P. Scholze, The pro-étale topology for schemes, Astérisque (2015), no. 369, 99–201.
[CC15] Alain Connes and Caterina Consani, Universal thickening of the field of real numbers, Advances in the
theory of numbers, Fields Inst. Commun., vol. 77, Fields Inst. Res. Math. Sci., Toronto, ON, 2015, pp. 11–
74.
[CCR+
82] R. R. Coifman, M. Cwikel, R. Rochberg, Y. Sagher, and G. Weiss, A theory of complex interpolation for
families of Banach spaces, Adv. in Math. 43 (1982), no. 3, 203–229.
[Cem84] P. Cembranos, C(K, E) contains a complemented copy of c0, Proc. Amer. Math. Soc. 91 (1984), no. 4,
556–558.
[Con11] Alain Connes, The Witt construction in characteristic one and quantization, Noncommutative geometry
and global analysis, Contemp. Math., vol. 546, Amer. Math. Soc., Providence, RI, 2011, pp. 83–113.
MR 2815131
[ČS19] K. Česnavičius and P. Scholze, Purity for flat cohomology, arXiv:1912.10932, 2019.
[CSCK15] F. Cabello Sánchez, J. M. F. Castillo, and N. J. Kalton, Complex interpolation and twisted twisted Hilbert
spaces, Pacific J. Math. 276 (2015), no. 2, 287–307.
[DFS08] J. Diestel, J. H. Fourie, and J. Swart, The metric theory of tensor products, American Mathematical
Society, Providence, RI, 2008, Grothendieck’s résumé revisited.
[DP61] A. Dold and D. Puppe, Homologie nicht-additiver Funktoren. Anwendungen., Ann. Inst. Fourier 11 (1961),
201–312 (German).
[Enf73] P. Enflo, A counterexample to the approximation problem in Banach spaces, Acta Math. 130 (1973),
309–317.
[Gro55] A. Grothendieck, Produits tensoriels topologiques et espaces nucléaires, Mem. Amer. Math. Soc. No. 16
(1955), Chapter 1: 196 pp.; Chapter 2: 140.
[Har84] D. Harbater, Convergent arithmetic power series, Amer. J. Math. 106 (1984), no. 4, 801–846.
[Hub93] R. Huber, Continuous valuations, Math. Z. 212 (1993), no. 3, 455–477.
[Kal81] N. J. Kalton, Convexity, type and the three space problem, Studia Math. 69 (1980/81), no. 3, 247–287.
[KL19] K. S. Kedlaya and R. Liu, Relative p-adic Hodge theory, II: Imperfect period rings, arXiv:1602.06899v3,
2019.
[KP79] N. J. Kalton and N. T. Peck, Twisted sums of sequence spaces and the three space problem, Trans. Amer.
Math. Soc. 255 (1979), 1–30.
[Lur09] J. Lurie, Higher topos theory, Annals of Mathematics Studies, vol. 170, Princeton University Press, Prince-
ton, NJ, 2009.
[Lur18] , Spectral algebraic geometry, 2018.
[NS18] T. Nikolaus and P. Scholze, On topological cyclic homology, Acta Math. 221 (2018), no. 2, 203–409.
109

110 BIBLIOGRAPHY
[Rib79] M. Ribe, Examples for the nonlocally convex three space problem, Proc. Amer. Math. Soc. 73 (1979), no. 3,
351–355.
[Ros07] J. Rosický, On homotopy varieties, Adv. Math. 214 (2007), no. 2, 525–550.
[Sch19] P. Scholze, Lectures on Condensed Mathematics, 2019.
[Smi52] M. F. Smith, The Pontrjagin duality theorem in linear spaces, Ann. of Math. (2) 56 (1952), 248–253.
[TV08] B. Toën and G. Vezzosi, Homotopical algebraic geometry. II. Geometric stacks and applications, Mem.
Amer. Math. Soc. 193 (2008), no. 902, x+224. MR 2394633

Lectures on Analytic Geometry

More Related Content

What's hot (20)

Similar to Lectures on Analytic Geometry (20)

More from Anatol Alizar (20)

Recently uploaded (20)

Lectures on Analytic Geometry