Diploma Thesis
This site can also be reached via thesis.jonas-stadelmann.com.
All ressources on this site are for autorized use only. To ask for permission you may write an email to: stadelmannjonas9@gmail.com.
Embedded Files
Analyzing and Overcoming Optimization Obstacles in Adaptive Variational Quantum Algorithms
Analyzing and Overcoming Optimization Obstacles in Adaptive Variational Quantum Algorithms
Quantum computers promise to solve complex chemical and physical problems that today’s supercomputers cannot handle. One particularly promising tool for this is the “ADAPT-VQE” algorithm. What makes it unique is that it does not assemble its computational program in advance, but rather adapts it dynamically during the calculation. But this smart approach has a weakness—it often gets stuck in so-called “gradient troughs.” These are deceptive mathematical dead ends where the guidance signals for the algorithm become so weak that it gets stuck and sometimes even mistakenly considers the computation process to be over. We addressed this problem in our thesis.
Project Overview: More Efficient Computations for Quantum Computers
What is it about? Simulating complex systems, such as the behavior of molecules in chemistry, is one of the most important tasks for future quantum computers. A very promising algorithm for this is called ADAPT-VQE. What makes it unique is that it does not rigidly define the required quantum program (the so-called “circuit”) in advance, but rather builds it up dynamically, step by step, during the calculation.
The problem: “Gradient Troughs” Although this method offers many advantages, the algorithm is often blocked by a phenomenon called “gradient troughs". In such a phase, the mathematical guides that the program uses for orientation suddenly drop drastically. The algorithm “thinks” it has reached the final goal, even though it is actually still far from the optimal solution. This leads to the process being stopped prematurely or requiring an extremely large number of time-consuming measurements just to make any progress at all.
The Solution Until now, the standard algorithm has always simply appended new computational steps to the end of the existing circuit. In this project, a completely new approach was developed to identify and overcome these dead ends. Together with our project partners at Virginia Tech, we discovered that we can show the program new paths by inserting the computational steps at different positions within the circuit—that is, not just at the end, but sometimes at the very beginning or in the middle.
The Benefit Through this clever placement, the algorithm receives clear signals again regarding the direction in which it must continue computing. It can thus successfully escape the “dead ends” while requiring significantly less measurement effort. As a result, this method leads to faster and more reliable calculations without making the quantum circuit unnecessarily large. This brings us a big step closer to the practical and efficient use of ADAPT-VQE in quantum computers.
Our Publication
Our paper on the results of this work is already available as a preprint at the following link: www.arxiv.org/abs/2512.25004. Updates regarding peer review and journal publication will be shared here as soon as they become available.
The Thesis
Since we have not yet defended the thesis, we are currently unable to make it publicly available. We will publish it here as soon as possible. To be notified when it becomes available, please register here.
Updates
You can follow the latest news and updates about the thesis and future projects on this topic on our Instagram channel.
Important dates in the coming weeks:
HTL Vorarlberg Award 2026 (May 21, 2026)
Jugend Innovativ 2026 (May 27–29, 2026)
Page updated
Google Sites
Report abuse