Derek S. Wang

Derek S. Wang

Research Scientist

IBM Quantum

I explore the limits of today’s quantum computers. My Research interests include quantum simulation of molecules & materials, learning & mitigating quantum noise, and dynamic quantum circuits for near-to-far-term quantum computations. I’m open to working with researchers of all experience levels, including those who are new to my field or to research in general. For a broad overview of my current lines of inquiry, see below.

I get to do what I do only because of a pantheon of teachers and mentors that have guided me through many hoops. I am interested in democratizing these jumps, including the chemistry olympiad, university, summer research programs and jobs, graduate school, scholarships, and fellowships. To this end, I’ve compiled some informal Resources. I’m also happy to share my application materials and tips.

I am recreationally interested in Metascience, or the science of science. Feel free to reach out to me to discuss this, or anything else mentioned here.


Quantum simulation of molecules and materials

I’m ultimately interested in designing molecules and materials that unlock technological leaps. In particular, during my PhD, I used classical first-principles methods to probe complex physics at the intersection of quantum chemistry, quantum optics, and many-body physics. Results include design rules for the generation of entangled photons from quantum optical matter, engineered point defects in solid-state materials as a qubit platform, transduction of quantum information in nanomagnonic cavities and the classical origin of cavity-modified chemical reactivity. (For an overview of these past efforts, see my PhD dissertation.) Despite theoretical advances and access to powerful classical computational clusters, many flavors of practically relevant physics remain out of reach of existing classical methods.

Now, I am interested in exploring the potential of today’s quantum computers to answer classically hard, but quantumly easy questions in quantum simulation of molecules and materials. As a start, I have explored large-scale quantum computational studies of quantum many-body dynamics, quantum chromodynamics, topological excitations, and polariton chemistry. Moving forward, I am interested in the following questions:

  • Akin to Jacob’s ladder of increasingly accurate and complex density functionals, I suggest we define ``Feynman’s Ladder," where rungs are quantum simulations of increasing complexity eventually suitable for large-scale & fault-tolerant quantum computers. What should these rungs, or checkpoints, be?
  • Is it possible to achieve a practical quantum advantage before fault tolerance?
  • Is there a simple way to quantify the overhead of simulating non-native many-body systems (one of our examples here) on an analogue quantum computer?
  • What is the most efficient way to encode mixed fermion-boson systems on a qubit-based quantum computer?
  • Non-equilibrium quantum dynamics is often cited as the first field where we expect a quantum advantage. What is the path from learning something fundamental about non-equilibrium quantum dynamics to practical utility?
  • Historically, how useful has it been in industry to simulate materials and molecules from first principles? Do we expect improved simulation capabilities to make a substantial difference? (Some light research on this topic here.)

Learning and mitigating quantum noise

The main challenge to quantum computers is a complex set of quantum noise processes that corrupt calculations. Learning faithful representations of quantum noise can reveal the underlying physical processes, thus allowing one to suppress, mitigate, or even correct the noise. Naively learning the full set of noise processes is challenging in a large-scale quantum device due to the exponential number of learned parameters. Furthermore, protocols that attempt to learn the noise are themselves affected by noise. We presented a method that enhances this learning process, even in the presence of imperfections in the learning process itself.

Once the noise is learned, quantum error mitigation strategies can harness imperfect quantum computers to yield near noise-free and meaningful results despite the presence of unmonitored errors. We have developed new strategies, such as a physics-based composite QEM workflow, a practical machine learning-based method that outperforms the physics-based composite QEM workflow, and a Bayesian inference-based one for measuring properties of quantum systems with shallow shadow tomography. Follow-on questions include:

  • What is the most efficient way to mitigate quantum noise at scale, across many different classes of circuits, and over long times?
  • What methods interpolate between error mitigation and correction?
  • Can we embed error mitigation methods directly within the circuit compilation step?
  • Certain error mitigation methods work surprisingly well. Why?

Dynamic quantum circuits toward fault-tolerant quantum computation

Quantum systems present two distinct modes of evolution: deterministic unitary evolution, and stochastic evolution as the consequence of quantum measurements. To date, quantum computations predominantly use unitary evolution to generate complex quantum states for information processing and simulation. However, especially for calculations that do not require quantum effects, such as the addition of two bits, classical computations are likely to be faster and more reliable. Exploiting these trade-offs has resulted in the development of dynamic quantum circuits, namely quantum circuits that collect classical information from mid-circuit measurements, classically process the results, and feed-forward operations within a single shot of the circuit. A flagship application of dynamic circuits is active quantum error correction. But there exist applications of dynamic circuits in the near term. For instance, with dynamic circuits, we have demonstrated the generation of long-range entanglement, generalized measurements, and the quantum Fourier transform algorithmic primitive with record-breaking fidelity. In the future, I’m interested in the following:

  • Can dynamic quantum circuits accelerate quantum simulations of molecules and materials?
  • How can error suppression and mitigation methods be generalized for dynamic quantum circuits?
  • How can we systematically discover efficient compilations of quantum circuits with dynamic instructions?
  • How should software tools for quantum computing, such as circuit simulation, be adapted for dynamic quantum circuits?


Here are some questions on my mind, some of which are accompanied by light notes.

  • What made some research labs, small or large, so successful, and can these same principles be applied to a quantum computing-specific lab? Woefully incomplete reading list.
  • What is the best way for training someone to ask and answer new & important questions? Similarly, can we formalize what it takes to have “good taste in research”, or good meta-decision-making abilities? Is there a correlation between being able to choose good research problems vs. solving them?
  • Which best practices for conducting research should be formalized and taught to new researchers?
  • What would the equivalent of a Hippocratic oath but for scientists look like? (My take is here.) Should scientists push for one, and how would they advocate for its adoption amidst the forces of industry research labs, publishers, and academia?
  • What is, will be, and should be the role of artificial intelligence in science? How can a machine learning model be used to discover new physical phenomena, other than serving as a null hypothesis?
  • How can the sum total of knowledge be quantified, so as to quantify research progress? How much new knowledge is produced per unit of research output, and how much previous knowledge is used to generate it, now vs. before? This article is a nice start.
  • What are examples of deep technologies where interest & investment (from the government, investors, and the public) came at the right time and in the right quantities?
  • How should pedalogical practices be adapted to account for student access to nearly limitless, freely available, and now, flexibly synthesized information? My teaching philosophy attempts to address this challenge, among others.
  • What cultural differences in research are there, and how can these gaps be bridged?