Subfactors in a nutshell

Subfactors in a nutshell
Or: ADE and all that
Daniel Tubbenhauer
June 2022
Daniel Tubbenhauer Subfactors June 2022 1 / 6

Jones’ revolution
Throughout
Please convince yourself that I haven’t messed up
while picking my pictures and text from my stolen material

Jones’ revolution
The history of subfactors can roughly be divided into three parts:
I Subfactors in operator theory Before Jones ∼1983--
I Invariants of subfactors Jones’ revolution ∼1983-1985
I Subfactors without subfactors!? Jones ∼1985++

Jones’ revolution
Along the way Jones
rediscovered the Temperley–Lieb (TL) algebra
and found a Markov trace on it

Jones’ revolution
Along the way Jones
rediscovered the Temperley–Lieb (TL) algebra
and found a Markov trace on it
Why is that important? Well, because of the (Birman–)Jones polynomial:

Jones’ revolution
Jones also implicitly coined the name “TL algebra” ∼1985:

Jones’ revolution
Today’s talk is based on:
my memory (horrible reference...)
The above are easy to google (it is worth it!)

Antediluvian ∼1983--
Why was everyone related to this part of math in Los Alamos?
I Factor = von Neumann algebra with trivial center
I Murray–von Neumann ∼1930++ classified factors by types: I, II1, II∞ and III
I II1 are the “most exciting” ones They have a unique trace!
We will stick with these momentarily (I drop the “of type II1”)
I A subfactor is an inclusion of factors N ⊂ M

Chemistry Group theory Operator theory
Matter Groups von Neumann algebras
Elements Simple groups Factors
Simpler substances Jordan–Hölder theorem The theorem below
Periodic table Classification of simple groups Classification of factors
Theorem (von Neumann ∼1949)
Every von Neumann algebra on a separable Hilbert space has an essentially unique
decomposition into direct integrals of factors

Example (factor)
Bounded operators B(H) on a Hilbert space H
If dim H < ∞, then these are just matrices

Example (factor)
Bounded operators B(H) on a Hilbert space H
If dim H < ∞, then these are just matrices
I need the following terminology
Projections E = EE = E∗
!
orthogonal projection onto some closed subspace
Subspaces and thus projection are ordered by inclusion
Finite projections No 0 < F < E equivalent to E
For example, in fin-dim-land everything is finite
Minimal projections No 0 < F < E

Type I = there is a minimal nonzero projection
Example (type I factor)
Every type I factor is isomorphic to
bounded operators B(H) on a Hilbert space H

Type I = there is a minimal nonzero projection
Example (type I factor)
Every type I factor is isomorphic to
bounded operators B(H) on a Hilbert space H
Type III = no nonzero finite projection
Non-example (type III factor)
Murray–von Neumann write ∼ 1936:
and do not solve Problem 3 since they did not find a type III factor
They exist! But giving an example gets me too far off track
They are also too hard for me anyway...

Type II = no minimal projections but there are nonzero finite projections
Type II1 = id operator is finite; Type II∞ = rest of type II
Example (type II factor)
Discrete group G with infinite nonidentity conjugacy classes, e.g. F2
von Neumann group algebra L(G) is a type II1 factor
(L(G) = the commutant of the left regular representation on `2
(G))
Reminder
`2
(G) = formal sequences
P
g∈G λg g with
P
g∈G |λg |2
< ∞
If |G| < ∞, then L(G) = C[G]

Type II = no minimal projections but there are nonzero finite projections
Type II1 = id operator is finite; Type II∞ = rest of type II
Example (type II factor)
Discrete group G with infinite nonidentity conjugacy classes, e.g. F2
von Neumann group algebra L(G) is a type II1 factor
(L(G) = the commutant of the left regular representation on `2
(G))
Reminder
`2
(G) = formal sequences
P
g∈G λg g with
P
g∈G |λg |2
< ∞
If |G| < ∞, then L(G) = C[G]
A famous open problem
L(F2) ∼
= L(Fk ) for k > 2

I did not find a
picture of Murray

I did not find a
picture of Murray
The word “algebra” is highlighted
because there will be representations!

I did not find a
picture of Murray
Example
M is a factor, G a nice group, then MG
⊂ M is a subfactor
Example
G a nice group with infinite conjugacy classes
H ⊂ G a nice subgroup with infinite conjugacy classes
L(H) ⊂ L(G) is a subfactor

I did not find a
picture of Murray
Example
M is a factor, G a nice group, then MG
⊂ M is a subfactor
Example
G a nice group with infinite conjugacy classes
H ⊂ G a nice subgroup with infinite conjugacy classes
L(H) ⊂ L(G) is a subfactor
Vague slogan Subfactors ! fixed points of a “quantum group” G action
The is also a version of Galois correspondence
and one can recover G from MG
⊂ M
In this sense subfactor theory generalizes group theory

I did not find a
picture of Murray
Philosophy
(hard to track back – happened around Jones’ revolution)
Focus on the abstract symmetry defined by a subfactor
+
de-emphasize the factors themselves
Really good invariants which we see today
and that shift the focus to combinatorics/categories
Index (number), principal graph, standard invariant (“tensor category”)

Jones’ construction ∼1983-1985
[M : N] ! eM ∈ R≥0
I Subfactor N ⊂ M, M is a N-module by left multiplication
I Assume that M is finitely generated projective N-module
The index [M : N] ∈ R≥0 ∪ {∞} is the trace of the idempotent eM for M
I Jones’ index theorem ∼1983 The index is an invariant of N ⊂ M and
[M : N] ∈ {4 cos2
( π
k+2 )|k ∈ N} ∪ [4, ∞]

[M : N] ! eM ∈ R≥0
[M : N] ∈ {4 cos2
( π
k+2 )|k ∈ N} ∪ [4, ∞]
Note the “quantization” below 4:
This was a weird/exciting result!
Jones: The most challenging part is constructing these subfactors

N ⊂ N o S3 : ,
I Mk = M ⊗N M ⊗N ... (k copies of M), Mk is a N-N, N-M, M-N and M-M bimodule
I Enter, the principal graph(s) Γ!

N ⊂ N o S3 : ,
Example
Principal graph for N ⊂ N o G:
as an N-N bimodule N o G is the direct sum, over G, of g-twisted trivial bimodules
Even vertices are the group elements
Example
Dual principal graph for N ⊂ N o G:
Even vertices are the simple group reps
with as many edges as dimension of the reps

N ⊂ N o S3 : ,
Perron–Frobenius theorem (Perron ∼1907, Frobenius ∼1912)
Irreducible matrices with entries from R≥0
have an eigenvalue pf ∈ R≥0 and
an associated eigenvector ~
pf ∈ Rn
≥0
The growth rate of RN
is roughly given by pf N

N ⊂ N o S3 : ,
Sketch of Jones’ quantization argument
[M : N] = pf (Γ)2
(unless [M : N] = ∞)
Then use Kronecker’s theorem ∼1857
Upshot of Jones’ construction
“Subfactors = graphs + combinatorics”
with huge quotation marks

N ⊂ N o S3 : ,
Jones’ type invariants were later also
introduced in other field, for example the study of tensor categories ∼2005
Example (principal graph of Rep(S3))
The graph for the action of the standard rep of S3 is:
PF miracle
Take T = M(triv) + M(std)2
+ M(triv)
pf (T) = 6 = |S3|
pf eigenvector ! regular rep C[S3]

Jones’ projectors satisfy the scaled TL relations for δ = [M : N] = 4 cos2
( π
k+2 )
ek ! 1
[M:N]
k
I ek ek = ek ! =
I ek ek±1ek = 1
[M:N] ek ! = 1
[M:N]
I ek el = el ek for |k − l| > 1 ! far commutativity

( π
k+2 )
ek ! 1
[M:N]
k
I ek ek = ek ! =
I ek ek±1ek = 1
[M:N] ek ! = 1
[M:N]
I will be more precise later
(using a more modern language)
but for now, Jones shows that
{ei |i ≥ n} generate a factor Rn
R2 ⊂ R1 is a subfactor of index 4 cos2
( π
k+2
)

( π
k+2 )
ek ! 1
[M:N]
k
I ek ek = ek ! =
I ek ek±1ek = 1
[M:N] ek ! = 1
[M:N]
The Markov property!
In braid pictures (crossing is given by Kauffman skein formula )
In hindsight the crucial result
Jones’ proved about TLC(δ)
is the existence of a Markov trace ⇒ Jones polynomial

Postdeluvian ∼1985++
NBim MBim
N MM
M MN
I We have collections of bimodules of the four flavors
I One flavor of bimodule forms a tensor category
I C∗
-2-category with two objects (corresponding to M and N) together with a
choice of generating 1-morphism (corresponding to N MM )+axioms =
Standard invariant
I The N-N bimodules and the M-M bimodules are the even parts of the
standard invariant, and the others the odd part

NBim MBim
N MM
M MN
I C∗
Standard invariant
Theorem (Popa ∼1994)
(Amenable = nice for the purpose of this talk)
ΦΨ = id and ΨΦ = id
An amenable subfactor can be reconstructed from its standard invariant

NBim MBim
N MM
M MN
I C∗
Standard invariant
This is also called quantum Tannaka–Krein duality (TKD)
Roughly, TKD = reconstruction of compact groups from their representations
I didn’t find a pdf copy of
M. G. Krein, A principle of duality for a bicompact group and a square block algebra
Dokl. Akad. Nauk. SSSR, N. S. 69, 725–728 (1949)

NBim MBim
N MM
M MN
I C∗
Standard invariant
Example
Here the N-N tensor categories are Vect(S3) and Rep(S3)

Type E is not displayed
A reformulation of Jones’ result is:
I Let L1 be a generator of a fusion category with pf (L1) < 2
I Then pf (L1) = 2 cos( π
k+2 ) and the fusion graph of L1 is of ADE type
I In type A the fusion category is Rep(Uq(sl2))ss
for q = exp( 2πi
k+2 )

for q = exp( 2πi
k+2 )
Jones constructed Rep(Uq(sl2))ss
via the TL calculus
Main toolbox are planar algebras
This theorem is due to Ocneanu and others ∼1988

for q = exp( 2πi
k+2 )
“Quantum McKay correspondence” Theorem (Popa ∼1994)
The pf 2
= 4 case is given by affine Dynkin diagrams:
A∞ is again the TL calculus

for q = exp( 2πi
k+2 )
Jones’s ADE subfactors are known to be related to
CFTs, knot theory, representation theory,
category theory, TQFT, integrable models, more...
Big question (open as far as I can tell):
Do other subfactors give “exotic” (non quantum group) versions of these?
(Supertransitivity=longest line in the graph)

Jones’ revolution
I did not find a
picture of Murray
Philosophy
+
[M : N] ! eM ∈ R≥0
[M : N] ∈ {4 cos2
( π
k+2 )|k ∈ N} ∪ [4, ∞]
N ⊂ N o S3 : ,
N ⊂ N o S3 : ,
pf ∈ Rn
≥0
N ⊂ N o S3 : ,
[M : N] = pf (Γ)2
NBim MBim
N MM
M MN
I C∗
Standard invariant
for q = exp( 2πi
k+2 )
There is still much to do...

Jones’ revolution
I did not find a
picture of Murray
Philosophy
+
[M : N] ! eM ∈ R≥0
[M : N] ∈ {4 cos2
( π
k+2 )|k ∈ N} ∪ [4, ∞]
N ⊂ N o S3 : ,
N ⊂ N o S3 : ,
pf ∈ Rn
≥0
N ⊂ N o S3 : ,
[M : N] = pf (Γ)2
NBim MBim
N MM
M MN
I C∗
Standard invariant
for q = exp( 2πi
k+2 )
Thanks for your attention!

Subfactors in a nutshell

More Related Content

Similar to Subfactors in a nutshell (20)

More from Daniel Tubbenhauer (13)

Recently uploaded (20)

Subfactors in a nutshell