Ill-posedness formulation of the emission source localization in the radio- detection arrays of extensive air showers initiated by Ultra High Energy Cosmic Rays

RAPPORT D’ACTIVIT É
Ill-posedness formulation of the emission source localization in the radio-
detection arrays of extensive air showers initiated by Ultra High Energy Cos-
mic Rays
AHMED REBAI1 FOR THE CODALEMA COLLABORATION,
1 Subatech IN2P3-CNRS/Université de Nantes/ École des Mines de Nantes, Nantes, France.
ahmed.rebai@subatech.in2p3.fr
Abstract: In the field of radio detection in astroparticle physics, many studies have shown the strong dependence
of the solution of the radio-transient sources localization problem (the radio-shower time of arrival on antennas)
such solutions are purely numerical artifacts. Based on a detailed analysis of some already published results of
radio-detection experiments like : CODALEMA 3 in France, AERA in Argentina and TREND in China, we
demonstrate the ill-posed character of this problem in the sens of Hadamard. Two approaches have been used as
the existence of solutions degeneration and the bad conditioning of the mathematical formulation problem. A
comparison between experimental results and simulations have been made, to highlight the mathematical studies.
Many properties of the non-linear least square function are discussed such as the configuration of the set of
solutions and the bias.
Keywords: UHECR; radio-detection; antennas; non-convex analysis; optimization; ill-posed problem
1 Introduction
The determination of the ultra-high energy cosmic rays
(UHECR) nature is an old fundamental problem in cosmic
rays studies. New promising approaches could emerge from
the use of the radio-detection method which uses, through
antennas, the radio emission that accompanies the exten-
sive air shower emission (EAS). Several experiences like
CODALEMA [1] in France and LOPES [2] in Germany
shown the feasibility and the potential of the method. Many
parameters have been reconstructed as the arrival direction,
the shower core at ground, the electric field lateral distribu-
tion function and the primary particle energy. However, the
temporal radio wavefront characteristics remain still poorly
determined. In fact, the arrival time distribution defined by
the filtred radio signal maximum amplitude. On the oth-
er hand, the migration of small scale radio experiments to
large scale experiments with huge surfaces of several ten-
s of 1000km2 using autonomous antennas, is challenging.
This technique is subjected limitations in regard to UHE-
CR recognition, due to noises induced by human activities
(high voltage power lines, electric transformers, cars, trains
and planes) or by stormy weather conditions (lightning).
Figure 1 shows a typical reconstruction of sources obtained
with the CODALEMA experiment [?], by using a spheri-
cal wave minimization. Such results are also observed in
many others radio experiments [?, ?]. In general, one of
the striking results is that the radio emission sources are
reconstructed with great inaccuracy, although they are fixed
and although the huge measured events number.
The frequentist approach uses an objective function
minimization that depends the wavefront shape, employing
the arrival times and antennas positions.
2 Reconstruction with common algorithms
The performances of different algorithms has been tested
using the simplest test array of antennas. Within the con-
straints imposed by the number of free parameters used for
Fig. 1: Typical result of reconstruction of two entropic
emitters at ground, observed with the stand-alone stations of
CODALEMA, through standard minimization algorithms.
Despite the spreading of the reconstructed positions, these
two transmitters are, in reality, two stationary point sources.
reconstruction, we choose an array of 5 antennas for which
the antennas positions −→ri = (xi,yi,zi) are fixed (see Fig. 4)
(this corresponds to a multiplicity of antennas similar to
that sought at the detection threshold in current setups).
A source S with a spatial position −→rs = (xs,ys,zs) is set
at the desired value. Assuming ts the unknown instant of the
wave emission from S, c the wave velocity in the medium
considered constant during the propagation, and assuming
that the emitted wave is spherical, the reception time ti on
each antenna i ∈ {1,...,N} can written:
ti = ts +
(xi −xs)2
+(yi −ys)2
+(zi −zs)2
c
+G(0,σt)
where G(0,σt) is the Gaussian probability density function
centered to t = 0 and of standard deviation σt. This latter
parameter stand for the the global time resolution, which

Rebai A. Ill-posedness formulation of the emission source localization in the radio-detection arrays
Fig. 2: ypical result of the reconstruction of a source ob-
served with AERA (Auger Engineering Radio Array) exper-
iment for an entropic emitter at ground, the minimization
has been implemented usign Minuit with the Simplex and
Migrad algorithm and shown in [10].
Fig. 3: Reconstruction results in the TREND experiment
with the same algorithm in the case of an radio emitter
located on ground inside the antennas array. The algorithm
provide a good estimation of the emitter position.
Fig. 4: Scheme of the antenna array used for the simulations.
The antenna location is took from a uniform distribution of
1m width.
depends as well on technological specifications of the
apparatus than on analysis methods.
The theoretical predictions are compared to the recon-
structions given by the different algorithms. The latter are
setup in two steps. First, a planar adjustment is made, in
order to pres-tress the region of the zenith angle θ and az-
imuth angle φ of the source arrival direction. It specifies
a target region in this subset of the phase space, reducing
the computing time of the search of the minimum of the
objective function of the spherical emission. Reconstruction
of the source location is achieved, choosing an objective-
function that measures the agreement between the data and
the model of the form, by calculating the difference be-
tween data and a theoretical model (in frequentist statistics,
the objective-function is conventionally arranged so that
small values represent close agreement):
f(rs,t∗
s ) =
1
2
N
∑
i=1
−→rs −−→ri
2
−(t∗
s −t∗
i )2
2
(1)
The partial terms −→rs −−→ri
2
−(t∗
s −t∗
i )2
represents the
difference between the square of the radius calculated using
coordinates and the square of the radius calculated using
wave propagation time for each of the N antennas. The
functional f can be interpreted as the sum of squared errors.
Intuitively the source positions −→rs at the instant ts is one
that minimizes this error.
In the context of this paper, we did not use genetic al-
gorithms or multivariate analysis methods but we focused
on three minimization algorithms, used extensively in sta-
tistical data analysis software of high energy physics[?, ?]:
Simplex, Line-Search and Levenberg-Marquardt (see table
1). They can be found in many scientific libraries as the
Optimization Toolbox in Matlab, the MPFIT in IDL and the
library Minuit in Root that uses 2 algorithms Migrad and
Simplex which are based respectively on a variable-metric
linear search method with calculation of the objective func-
tion first derivative and a simple search method. For the
present study, we have used with their default parameters.
We tested three time resolutions with times values took
within 3σt.
• σt = 0ns plays the role of the perfect theoretical
detection and serves as reference;
• σt = 3ns reflects the optimum performances expected
in the current state of the art;
• σt = 10ns stands for the timing resolution estimate
of an experiment like CODALEMA [?].
For every source distance and temporal resolution, one
million events were generated. Antenna location was taken
in a uniform distribution of 1m width. A blind search was
simulated using uniform distribution of the initial rs values
from 0.1km to 20km. Typical results obtained with our
simulations are presented in Figures 5 and 6. The summary
of the reconstructed parameters is given in table ??.
Whatever the simulations samples (versus any source dis-
tances, arrival directions, time resolutions), (also with sever-
al detector configurations) and the three minimization algo-
rithms, large spreads were generally observed for the source
locations reconstructed. This suggests that the objective-
function presents local minima. Moreover, the results de-
pend strongly on initial conditions. All these phenomena
may indicate that we are facing an ill-posed problem. In-
deed, condition number calculations [?] (see Fig. 2), which
measures the sensitivity of the solution to errors in the data
(as the distance of the source, the timing resolution, etc.),
indicate large values (> 104), when a well-posed problem
should induce values close to 1.
To understand the observed source reconstruction pat-
terns, we have undertaken to study the main features of this
objective-function.

Table 1: Summary of the different algorithms and methods used to minimize the objective-function. The second row
indicates framework functions corresponding to each algorithm; third recalls the framework names. The key information
used for optimization are recalled down, noting that a differentiable optimization algorithm (ie. non-probabilistic and non-
heuristic) consists of building a sequence of points in the phase space as follows: xk+1 = xk +tk.dk, and that it is ranked
based on its calculation method of tk and dk parameters (see [11, 12, 13]).
Minimization algorithms Levenberg-Marquardt Simplex Line-Search
Libraries lsqnonlin - MPFIT fminsearch - SIMPLEX MIGRAD - lsqcurveﬁt
Software Optimization Toolbox Mat-
lab - IDL
Optimization Toolbox Mat-
lab - MINUIT-ROOT
Optimization Toolbox Mat-
lab - MINUIT-ROOT
Method Principles Gauss-Newton method
combined with trust region
method
Direct search method Compute the step-size by op-
timizing the merit function
f(x+t.d)
Used information Compute gradient (∇f)k
and an approximate hessian
(∇2 f)k
No use of numerical or ana-
lytical gradients
f(x + t.d, d) where d is a
direction descent computed
with gradient/hessian
Advantages / Disadvantages Stabilize ill-conditioned Hes-
sian matrix / time consuming
and local minimum trap
No reliable information
about parameter errors and
correlations
Need initialization with an-
other method, give the op-
timal step size for the opti-
mization algorithm then re-
duce the complexity
Fig. 5: Results of the reconstruction of a source with a
radius of curvature equal to 1 and 10 km with the LVM
algorithm. For Rtrue = 1km, the effect of the blind search
leads to non-convergence of the LVM algorithm, when
initialization values are greater than Rtrue = 1km.
Several properties of the objective-function f were stud-
ied: the coercive property to indicate the existence of at least
one minima, the non-convexity to indicate the existence of
several local minima, and the jacobian to locate the critical
Fig. 6: Results of the reconstruction of a source with a
radius of curvature equal to 1 and 10 km with the Simplex
algorithm.
points. (Bias study, which corresponds to a systematic shift
of the estimator, is postponed to another contribution). In
mathematical terms, this analysis amounts to:
• Estimate the limits of f to make evidence of critical
points; obviously, the objective function f is positive,

Fig. 7: Condition numbers obtained using the formula
Cond(Q) = Q . Q−1 with Q the Hessian matrix (see nex-
t section) as a function of the source distance and for dif-
ferent timing resolutions. The large values of conditioning
suggest that we face an ill-posed problem.
regular and coercive. Indeed, f tends to +∞ when
X → ±∞, because it is a polynomial and contains
positive square terms. So, f admits at least a mini-
mum.
• Verify the second optimality condition: the convexity
property of a function on a domain for a sufficiently
regular function is equivalent to positive-definiteness
character of its Hessian matrix.
• Solve the first optimality condition: ∇f(Xs) = 0
(jacobian) to find the critical points.
2.1 Convexity property
Using fi(Xs) = (Xs −Xi)T .M.(Xs −Xi) where M designates
the Minkowski matrix and given ∇fi(Xs) = 2.M(Xs − Xi),
the f gradient function can written:
1
2
∇f(Xs) = (∑ fi(Xs))M.Xs −M.(∑ fi(Xs)Xi)
The Hessian matrix, which is the f second derivative can
written:
∇2
f(Xs) = ∑∇ fi(Xs).∇ fT
i +∑ fi.∇2
fi
that becomes, replacing ∇fi by its expression:
∇2
f(Xs) = (∑ fi(Xs)).M +2M.[N.Xs.XT
s +
∑XiXT
i −Xs(∑Xi)T
−(∑Xi)XT
s ].M
Using a Taylor series expansion to order 2, an expanded
form of the Hessian matrix, equivalent to the previous
formula of the f second derivative, is:
This latter allowed us to study the convexity of f. Indeed,
because its mathematical form is not appropriate for a direct
use of the convexity definition, we have preferred to use
the property of semi-positive-definiteness of the Hessian
matrix. Our calculus lead to the conclusion that:
• Using the criterion of Sylvester [?] and the analysis
of the principal minors of the Hessian matrix , we find
that f is not convex on small domains, and thus is
likely to exhibit several local minima, according to Xs
and Xi. It is these minima, which induce convergence
problems to the correct solution for the common
minimization algorithms.
2.2 Critical points
The study of the first optimality condition (Jacobian =
0) gives the following system ∇ f(Xs) = 0 and allows
finding the critical points and their phase-space distributions.
Taking into account the following expression:
1
2 ∇ f( ¯Xs) = (∑ fi( ¯Xs))M. ¯Xs −M.(∑ fi( ¯Xs)Xi)
we get the relation:
Xs =
N
∑
i=1
fi(Xs)
∑j fj(Xs)
Xi (2)
This formula looks like the traditional relationship of a
barycenter. Thus, we interpret it in terms of the antennas
positions barycenter and its weights. The weight function
fi expressing the space-time distance error between the
position exact and calculated, the predominant direction will
be the one presenting the greatest error between its exact
and calculated position. The antennas of greatest weight
will be those the closest to the source.
In practice, because the analytical development of this
optimality condition in a three-dimensional formulation is
not practical, especially considering the nonlinear terms,
we chose to study particular cases. We considered the case
of a linear antennas array (1D) for which the optimality
condition is easier to express with an emission source
located in the same plane. This approach allows us to
understand the origin of the observed degeneration which
appears from the wave equation invariance by translation
and by time reversal (known reversibility of the wave
equation in theory of partial differential equations) and
provides us a intuition of the overall solution. It also
enlightens the importance of the position of the actual
source relative to the antennas array (the latter point is
linked to the convex hull of the antenna array and is the
object of the next section). Our study led to the following
interpretations:
• The iso-barycenter of the antenna array (of the lit an-
tennas for a given event) plays an important role in
explaining the observed numerical degeneration. The
nature of the critical points set determines the conver-
gence of algorithms and therefore the reconstruction
result.
• There are strong indications, in agreement with the
experimental results and our calculations (for 1D
geometry), that the critical points are distributed on a
line connecting the barycenter of the lit antennas and
the actual source location. We used this observation
to construct an alternative method of locating the
source (section 4).
• According to the source position relative to the anten-
na array, the reconstruction can lead to an ill-posed
or well-posed problem, in the sense of J. Hadamard.

Fig. 8: Scheme of the reconstruction problem of spherical
waves for our testing array of antennas (2D), with a source
located at ground. For this configuration, the convex hull
becomes the surface depicted in red. The result is the same
for a source in the sky.
2.3 Convex hull
In the previous section we pointed that to face a well-
posed problem (no degeneration in solution set), it was
necessary to add constraints reflecting the propagation law
in the medium, the causality constraints, and a condition
linking the source location and the antenna array, the latter
inducing the concept of convex hull of the array of antennas.
We also saw that analytically the critical points evidence
could become very complex from the mathematical point
of view. Therefore, we chose again an intuitive approach to
characterize the convex hull, by exploring mathematically
the case of a linear array with an emission source located in
the same plane.
The results extend to a 2D antenna array, illuminated
by a source located anywhere at ground, arguing that it
is possible to separate the array into sub-arrays arranged
linearly. The superposition of all the convex segments of
the sub-arrays leads then to conceptualize a final convex
surface, built by all the peripheral antennas illuminated (see
Figure 8).
The generalization of these results to real practical expe-
rience (with a source located anywhere in the sky) was guid-
ed by our experimental observations (performed through
minimization algorithms) that provide a first idea of what
happens. For this, we chose to directly calculate numeri-
cally the objective function for both general topologies: a
source inside the antenna array (ie. and at ground level) and
an external source to the antenna array (in the sky ). As
can be deduced from the results (see Figs. 9 and 10), for a
surface antenna array, the convex hull is the surface defined
by the antennas illuminated. (An extrapolation of reasoning
to a 3D network (such as Ice Cube, ANTARES,...) should
lead, this time, to the convex volume of the setup).
Our results suggest the following interpretations:
• If the source is in the convex hull of the detector, the
solution is unique. In contrast, the location of the
source outside the convex hull of the detector, causes
degeneration of solutions (multiple local minimums)
regarding to the constrained optimization problem.
The source position, outside or inside the array, af-
fects the convergence of reconstruction algorithms.
Fig. 9: Plots of the objective-function versus R and versus
the phase space (R, t), in the case of our testing array (2D),
for a source on the ground and located inside the convex
surface of the antenna array. This configuration leads to a
single solution. In this case the problem is well-posed.
Fig. 10: Plots of the objective-function versus R and versus
the phase space (R, t), in the case of our testing array (2D)
for a source outside the convex hull. This configuration
leads to multiple local minima. All minima are located on
the line joining the antenna barycenter to the true source. In
this case the problem is ill-posed.
3 Conclusion
Experimental results indicated that the common methods of
minimization of spherical wavefronts could induce a mis-
localisation of the emission sources. In the current form
of our objective function, a first elementary mathematical
study indicates that the source localization method may
lead to ill-posed problems, according to the actual source
position.
References
[1] D. Ardouin, et al., Radio-detection signature of high-
energy cosmic rays by the CODALEMA experiment,

Nucl. Instr. and Meth. A 555 (2005) 148.
[2] H. Falcke, et al., Detection and imaging of atmospheric
radio flashes from cosmic ray air showers, 2005,
Nature 435, 313-316.
[3] W. D. Apel et al. , Lateral distribution of the radio
signal in extensive air showers measured with LOPES,
Astropart. Phys., (32) 2010, 294.
[4] D. Ardouin et al., Geomagnetic origin of the radio
emission from cosmic ray induced air showers
observed by CODALEMA, Astropart. Phys., 31
(2009), 294.
[5] A. Horneffer, et al., Primary particle energy calibration
of the EAS radio pulse height, Proc. of the 30th ICRC,
Mrida 2007, Vol 4. (HE part 1), 83-86.
[6] A. REBAI, et al. Ill-posed formulation of the emission
source localization in the radio-detection experiments
of extensive air showers. arxiv 1208.3539.
[7] S. Lafebre, et al., Prospects for determining air shower
characteristics through geo-synchrotron emission
arrival times, Astropart. Phys. 34 (2010) 12-17
[8] D. Ardouin et al. (Trend collaboration): First detection
of extensive air showers by the TREND self-triggering
radio experiment, Astropart., (34) 2011, 717.
[9] K. Weidenhaupt for the Pierre Auger Collaboration,
The Auger Engeneering Radio Array, XIV Vulcano
Workshop, May 28 - June 2, 2012, Sicily, Italy, to be
published in Acta Polytechnica 2013 Vol. 53 N. 1. (ed.
by F. Giovannelli and G. Mannocchi).
[10] L. Mohrmann, Measurement of Radio Emission from
Cosmic Ray induced Air Showers at the Pierre Auger
Observatory with a Spherical Wave Reconstruction,
Masterarbeit in Physik, Univ. of Aachen 2011.
[11] F. James and M. Roos, MINUIT - A system for
function minimization and analysis of the parameter
errors and correlations, Computer Physics
Communications 10 (1975), 343-367.
[12] C. B. Markwardt, Non-linear Least Squares Fitting in
IDL with MPFIT, 2009,
arXiv.org/astro-ph/arXiv:0902.2850.
[13] J. F. Bonnans et al. , Numerical Optimization:
Theoretical and Practical Aspects, Springer-Verlag,
Universitext, (second ed.) 2003.
[14] R. A. Horn, C. R. Johnson, Matrix Analysis,
Cambridge University Press, (first ed. 1985) 1999.
[15] F. Delprat-Jannaud and P. Lailly, Ill-Posed and Well-
Posed Formulation of the Reflection Travel Time
Tomography Problem, J. of Geophysical Research, vol.
98, No. B4, p. 6589, April 10, 1993.

Ill-posedness formulation of the emission source localization in the radio- detection arrays of extensive air showers initiated by Ultra High Energy Cosmic Rays

More Related Content

What's hot (14)

Similar to Ill-posedness formulation of the emission source localization in the radio- detection arrays of extensive air showers initiated by Ultra High Energy Cosmic Rays (20)

More from Ahmed Ammar Rebai PhD (20)

Recently uploaded (20)

Ill-posedness formulation of the emission source localization in the radio- detection arrays of extensive air showers initiated by Ultra High Energy Cosmic Rays