Geisel Web
Geisel Web
Geisel Web
The interplay of nonlinear dynamics and wave propagation is The phenomena we address in our own research reach from
a rich area of complex phenomena. Reaching back to the very the chaotic motion of ballistic electrons in graphene, to the
beginnings of nonlinear science and computational physics dynamics of Bose-Einstein condensates in optical lattices
in the celebrated Fermi-Pasta-Ulam computer experiment and to the occurrence of extreme events in wave propagati-
on the statistical physics of heat transport, this area of phy- on through complex media.
sics became prominent in the field of quantum chaos. Its im- Complex media due to their internal structure can often be
plications are now subject of research in many fundamental described as random fields with spatial correlations. When
and applied sciences, from atom-optics to oceanography. waves are weakly scattered by such a medium, the weak but
1500 km
Fig. 1: The simulation of a tsunami wave in a region of the Indian Ocean where the ocean depth variations have a standard deviation of only less
than 7% reveals the strong impact of branching on tsunami propagation [1].
1
Nonlinear Dynamics, Chaos, and Theoretical Brain Research
Fig. 2: The collective dynamics of the network of interacting nerve cells in our brains forms the basis of brain function and provides formidable
challenges for innovative theoretical research. When recording activity from a brain area (left), we are constrained to sampling a fraction of the
neurons (right). We are developing approaches for inferring the collective dynamics despite this subsampling constraint (figure generated using
TREES toolbox by H. Cuntz).
correlated random forces conspire with nonlinear dynamics We investigate these questions building on approaches from
and diffraction to cause waves of extreme height or intensity statistical physics and information theory. In collaboration
in characteristic branch-like spatial structures [2]. This phen- with experimentalists from the Hôpital Salpetrière in Paris
omenon can cause unexpected conductance features in se- we recently showed that the human brain can exhibit col-
miconductors as well as giant “rogue waves” at sea. And even lective dynamics close to a “critical state”, as in a 2nd order
tiny fluctuations in the height-profile of the ocean floor can phase transtition [3,5]. This state appears functionally parti-
scatter tsunami waves and focus their energy by an order of cularly versatile for brain function, because in models it has
magnitude (see Fig.1), with severe consequences for tsunami been found to maximize information processing capacity.
predictability. This line of our research is complemented by To precisely infer the properties of collective dynamics from
our studies on the intrinsic localization of nonlinear waves, neural recordings, we are deriving mathematical approaches
e.g. of Bose-Einstein condensates in leaky optical lattices, that allow to infer the collective dynamics from only a small
which also leads to extreme events albeit by very different subset of all neurons (subsampling). This approach for the
mechanisms. first time formalizes the inference of system dynamics from
observations that are limited to a small fraction of units, and
Neural Dynamics and Information Processing is thus relevant not only for brain research, but also for in-
fering system properties from sparsely sampled networks in
Studying the human brain, one is faced with a complex net- general.
work of 80 billion neurons, each of them interacting with
thousands of other neurons by means of action potentials, Human Dynamics and
short electrical pulses. In our research at the Institute for Neural Mechanisms of Timing
Nonlinear Dynamics and the Max Planck Institute for Dy-
namics and Self-Organization we address three major ques-
tions: What is the collective dynamics of the interacting Human dynamics is a new branch of statistical physics,
neural network? How does this network give rise to infor- which aims to understand statistical properties of human
mation processing and thus to our cognitive abilities? And behavior as expressed e.g. in inter-event times and waiting
how can we infer these properties from neural recordings, if time distributions. This is particularly important for the mo-
we can assess the activity of only a fraction of all neurons? deling and forecast of the spreading of epidemics, where re-
2
Institute for Nonlinear Dynamics
liable statistics of human traveling behavior is required. In a [1] H. Degueldre, J. J. Metzger, T. Geisel, and R. Fleischmann, „Random
focusing of tsunami waves,“ Nature Phys. 12, 259–262 (2016).
seminal study [6], we have used banknotes (dollar bills) as a
proxy for human travel and determined e.g. an inverse po- [2] J. J. Metzger, R. Fleischmann, and T. Geisel, “Statistics of Extreme Waves in
wer law for the traveling distances. More recently we have Random Media,” Phys. Rev. Lett. 112(20), 203903 (2014).
investigated the nature of temporal fluctuations in perfor- [3] V Priesemann, M Valderrama, M Wibral, M Quyen, “Neuronal Avalanches
med musical rhythms (Fig. 3). We found that these fluctu- Differ from Wakefulness to Deep Sleep–Evidence from Intracranial Depth
Recordings in Humans”, PLOS Comp Biol 9, e1002985 (2013)
ations exhibit long-range correlations, i.e., a small rhythmic
fluctuation at some point in time still influences rhythmic [4] H. Hennig, R. Fleischmann, and T. Geisel, “Musical rhythms: The science of
being slightly off,” Physics Today 65(7), 64–65 (2012).
fluctuations after tens of seconds [4]. Our findings have led
to patents for the so-called humanizing of computer-genera- [5] A. Levina, J.M. Herrmann, and T. Geisel, “Dynamical synapses causing self-
ted musical sequences. organized criticality in neural networks”, Nature Phys. 3:857-860 (2007).
www.nld.physik.uni-goettingen.de
www.nld.ds.mpg.de
Fig. 3: We aim to characterize the nature of temporal fluctuations Fig. 4: Some of the problems we address require extensive com-
in musical performances and the neuronal mechanisms underlying puting power. For that purpose our group operates dedicated
musical timing. high-performance computing platforms with approximately 10,000
CPU cores.