Greetings!

Profile picture
© Xin Long

I am Fabian Koehler, a quantum physicist with a passion for software development and artificial intelligence. Presently, my endeavours orbit around the development of satellite-based quantum key distribution systems aiming at securing communication links on a global scale against both classical and quantum threats.

My academic background is anchored in quantum physics with a focus on numerical methods, software development and high performance computing. This scientific voyage culminated in a PhD in theoretical physics where I honed my expertise in the development and application of algorithms to investigate the rich phenomenology of quantum many-body systems.

Beyond the realm of physics, I harbor an insatiable curiosity for the field of machine learning and artifical intelligence. The crafting of novel deep learning systems and the exploration of their capabilities to solve real-world problems is a source of endless fascination for me.

In moments of reprieve, I find solace in contributing to open-source projects, strumming melodies on my guitar, and enjoying the outdoors.

Publications

Exploring disordered quantum spin models with a multilayer multi­configurational approach

Publication · arXiv
Abstract

Numerical simulations of quantum spin models are crucial for a profound understanding of many-body phenomena in a variety of research areas in physics. An outstanding problem is the availability of methods to tackle systems that violate area laws of entanglement entropy. Such scenarios cover a wide range of compelling physical situations including disordered quantum spin systems among others. In this paper, we employ a numerical technique referred to as multilayer multiconfiguration time-dependent Hartree (ML-MCTDH) to evaluate the ground state of several disordered spin models. ML-MCTDH has previously been used to study problems of high-dimensional quantum dynamics in molecular and ultracold physics but is here applied to study spin systems. We exploit the inherent flexibility of the method to present results in one and two spatial dimensions and treat challenging setups that incorporate long-range interactions as well as disorder. Our results suggest that the hierarchical multilayering inherent to ML-MCTDH allows to tackle a wide range of quantum many-body problems such as spin dynamics of varying dimensionality.

Bosonic quantum dynamics following colliding potential wells

Publication · arXiv
Abstract

We employ the multiconfiguration time-dependent Hartree method for bosons in order to investigate the correlated nonequilibrium quantum dynamics of two bosons confined in two colliding and uniformly accelerated Gaussian wells. As the wells approach each other an effective, transient double-well structure is formed. This induces a transient and oscillatory over-barrier transport. We monitor both the amplitude of the intrawell dipole mode in the course of the dynamics as well as the final distribution of the particles between the two wells. For fast collisions we observe an emission process which we attribute to two distinct mechanisms. Energy transfer processes lead to an untrapped fraction of bosons and a resonant enhancement of the deconfinement for certain kinematic configurations can be observed. Despite the comparatively weak interaction strengths employed in this work, we identify strong interparticle correlations by analyzing the corresponding von Neumann entropy.

Dynamical pruning of the non-equilibrium quantum dynamics of trapped ultracold bosons

Publication · arXiv
Abstract

The investigation of the nonequilibrium quantum dynamics of bosonic many-body systems is very challenging due to the excessively growing Hilbert space and poses a major problem for their theoretical description and simulation. We present a novel dynamical pruning approach in the framework of the multiconfiguration time-dependent Hartree method for bosons (MCTDHB) to tackle this issue by dynamically detecting the most relevant number states of the underlying physical system and modifying the many-body Hamiltonian accordingly. We discuss two different number state selection criteria as well as two different ways to modify the Hamiltonian. Our scheme regularly re-evaluates the number state selection in order to dynamically adapt to the time evolution of the system. To benchmark our methodology, we study the nonequilibrium dynamics of bosonic particles confined either in an optical lattice or in a double-well potential. It is shown that our approach reproduces the unpruned MCTDHB results accurately while yielding a significant reduction of the simulation time. The speedup is particularly pronounced in the case of the optical lattice.

Projects

KTailctl

A GUI to monitor and manage Tailscale on your Linux desktop.

mpskit

C++17 toolkit to study the static properties of discrete quantum systems.

mlxtk

Toolkit for ML-MCTDH(X) simulations using QDTK

pelican-math-svg

Render mathematical expressions for your pelican site to SVG in a smart way.

pelican-references

Handle references and bibliographies for pelican posts.

pelican-syntax-highlighting

Syntax highlighting for your pelican site using various backends.

cpp-perf

Small header-only C++11 to measure code execution time and performance.