Advancing the frontiers of quantum science and engineering

Abstract: Pseudorandomness has played an indispensable role in variousfields of computer science, and in the quantum world things are even moreamazing. While in the classical world, one-way functions (OWFs), pseudorandomgenerators (PRGs), and pseudorandom functions (PRFs) are all equivalent, in thequantum world, even if BQP=QMA (relative to a quantum oracle) or if P=NP(relative to a classical oracle), pseudorandom quantum states (PRSs) andpseudorandom function-like quantum state (PRFS) generators can still exist.In our recent work, we construct an (oracle model) isometryalgorithm, whose output can be arbitrarily close to a Haar-random state, giventhe quantum access to a random function. The goal is to prepare a quantum statewhose coefficients follow the normalized standard complex Gaussian randomvector (which behaves as a Haar-random state), and our key step is similar toGrover-Rudolph algorithm: We run a Beta sampling algorithm in superposition andapply controlled rotations qubit by qubit. Using this technique, we obtain:1. A new scalable PRS construction.2. The first scalable and adaptive PRFS construction.Our PRFS construction implies various prior primitives, includinglong-input PRFS, short-input PRFS, short-output PRFS, non-adaptive PRFS, andclassical-accessible adaptive PRFS. This new construction may be helpful insome simplification of the microcrypt zoo (https://sattath.github.io/microcrypt-zoo/).

Abstract: Two topics will be covered in this talk.1. Exploring fundamental quantum cryptographicprimitives: One-way functions (OWF) are the weakest primitive of classicalcryptography—no classical cryptographic scheme can exist without assuming theexistence of OWFs. Interestingly, it has been shown that pseudorandom quantumstates (PRS) can exist even when BQP = QMA relative to a quantum oracle[Kre21], or when P = NP relative to a classical oracle [KQST23]. In theseoracle worlds, OWFs secure against quantum (or classical) polynomial-timeadversaries do not exist. Motivated by the possibility of exploiting quantumadvantages, researchers have proposed numerous cryptographic primitives thatmay be weaker than OWFs in the quantum setting, leading to a "zoo" ofprimitives. Among these, EFI has emerged as the weakest known primitive, withsome other primitives shown to be equivalent to it. To better understand andsimplify this zoo, we investigate potential primitives that might be equivalentto EFI.2. Toward unified definitions of quantumcommitments: Bit commitment is a fundamental cryptographic primitive in whichAlice commits to a bit while keeping it hidden from Bob, only to reveal itlater. However, naive quantum analogues of classical security definitions havebeen shown to lead to insecure quantum bit commitment schemes [ABU14, SZ25]. Toaddress this, a revised security definition was proposed in [Unr16]. Morerecently, definitions for quantum state commitments—a generalization of bitcommitments to quantum states—have been introduced [GJMZ23]. These definitions,however, are formulated in different frameworks. In the second part of thetalk, we explore approaches to unify the definitions of quantum bit commitmentsand quantum state commitments under a common framework.

Integrated photonics offers a scalable solution for realizing complex photonic quantum systems, which are essential for practical computing and simulation applications. Recently, thin-film lithium niobate (TFLN) has emerged as a promising material platform for integrated nonlinear and quantum photonics. It offers many attractive properties that are critically missing in traditional silicon photonics, such as large electro-optic and piezoelectric coefficients, strong second-order nonlinearity, and engineerable ferroelectric domains. In this talk, we will discuss our recent efforts in developing nonlinear-based quantum light sources on TFLN, including photon-pair generation using new ferrielectric engineering strategies, squeezed-light generation in high-quality periodically poled TFLN waveguides, bi-photon frequency comb generation through cavity integration, and other related topics.

Abstract: We implement a range of experiments—from quantum foundations through to quantum algorithms—using programmable photonic integrated circuits [1] driven by a semiconductor quantum-dot single-photon source [2].

Grover’s algorithm [3] interrogates an N-element unstructured database in √N searches as opposed to the best classical algorithm, which requires N/2. The original version is non-deterministic, with a significant probability of failing that varies with the size of the database and the number of marked elements. A recent version is deterministic, in principle achieving 100% success probability for any database size, as long as the number of marked elements are ¼ or less of the database size. We implement both versions of Grover’s on programmable quantum photonic processors in silicon nitride.

The realised unitaries vary considerably from the ideal due to intrinsic circuit noise: a major issue is thermal cross-talk between the heaters used to program the circuit. We explored three regimes: 1) using the native device settings; then further correcting for circuit imperfections by 2) fast sequential vary and measure characterisation; and 3) scalable machine learning-assisted clear-box characterisation [5]. Regime 1 saw poor performance, regimes 2 and 3 improved considerably upon this with regime 3 achieving near-ideal performance. In every regime we find that the deterministic version is considerably more robust against device imperfections. Our work confirms the operational robustness of the deterministic Grover’s algorithm and the pathway for quantum photonic processors to move towards relevant quantum algorithmic implementations.

[1]. C. Taballione, M. Correa Anguita, M. de Goede, P. Venderbosch, B. Kassenberg, H. Snijders, N. Kannan, W. L. Vleeshouwers, D. Smith, J. P. Epping, R. van der Meer, P. W. H. Pinkse, H. van den Vlekkert, and J. J. Renema, Quantum 7, 1071 (2023). [2]. P. Senellart, G. Solomon, and A. G. White, Nature Nanotechnology 12, 1026–1039 (2017). [3]. L. K. Grover, Physical Review Letters 79, 325 (1997). [4]. T. Roy, L. Jiang, and D. I. Schuster, Physical Review Research 4, L022013 (2022). [5]. A. Fyrillas, O. Faure, N. Maring, J. Senellart, and N. Belabas, Optica 11, 427-436 (2024). [6]. F. Mohit, A. Roy, J. Guanzon, J. McKinlay, T. J. Weinhold, M. P. Almeida, M. Rambach, and A. G. White, arXiv:2506.06435 (2025).

Abstract: Artificial intelligence aims to model agents that learn and adapt through interaction. Among its frameworks, reinforcement learning stands out as the most natural paradigm, centered around the fundamental dilemma between exploration and exploitation. At the core of this paradigm lies the multi-armed bandit, a minimal yet powerful model capturing this trade-off. While there have been various efforts to extend AI into the quantum realm, we ask: have we truly studied simple reinforcement learning scenarios with quantum data? And more broadly, can we rethink reinforcement learning through the lens of quantum information? This thesis explores reinforcement learning in quantum environments through the multi-armed quantum bandit framework. Given streaming access to an unknown quantum state, we sequentially choose observables to maximize expected value or minimize regret—the cumulative gap to the best possible outcome. We provide information-theoretic lower bounds and matching upper bounds, showing that regret typically grows with the square root of the number of rounds. We apply this framework to quantum state tomography, aiming to learn the state efficiently while minimizing measurement disturbance. For pure states, we achieve polylogarithmic regret using a sample-efficient algorithm based on a weighted online least squares estimator. Finally, we demonstrate the framework’s flexibility through applications in thermodynamic work extraction and quantum data recommendation.

Abstract: Neutral atom arrays have emerged as one of the leading architectures for implementing quantum processors and quantum simulators. In order to facilitate interatomic interactions required for entangling operations, the atoms must be driven to high-lying Rydberg states, where their large polarizability leads to a second-order dipole-dipole interaction that can blockade the doubly-excited state. However, this excitation process is highly susceptible to noise, which limits the fidelity of gate operations as well as the useful timescales over which many-body simulations remain reliable. I will discuss our approach towards realising a high-fidelity controlled-Z gate on a Caesium quantum processor, with a comprehensive error budgeting analysis over various noise channels. Building upon this, I will go into our approach for designing a low noise Rydberg laser system and our current progress on our experiment. Finally, I will briefly describe our future plans for a blue-detuned tweezer array capable of trapping atoms in both the Rydberg and ground state, and its implications for quantum processors and simulators.

Abstract: Rare earth ion dopants in cryogenic crystalline hosts offer a unique platform for quantum technologies, combining long coherence times with the potential for integration into scalable architectures. Their shielded 4f electrons enable narrow optical and spin transitions even in solid-state environments, making them promising candidates for quantum memories, transducers, and processors. After a general introduction I will focus on microwave to optical transduction and our recent work highlight the potential of rare earths in magnetic host crystals. In particular, we have shown that despite the very high densities of spins extremely coherence times are potentially possible. We also have observed strong coherent coupling between rare earths and collective magnetic resonance modes (magnons) of the host crystal. This has exciting implications for microwave optical transduction of quantum signals.

Superconducting nanowire single-photon detectors (SNSPD) systems have progressed rapidly over the past decade. Recent devices exhibit leading performance figures including sub-10 ps system timing jitter, 1.5 GHz detection rates, arrays up to 400,000 pixels, sub-hertz dark count rates, high efficiency, as well as intrinsic and array-based photon number resolution. Quantum networks must be built upon protocols that support high data throughput in bits/s and ebit/s to achieve broad utility and maturity. If built with high-bandwidth nonlinear optics or emerging ultra-fast electro-optical modulators, the photonic sources in these networks no longer limit their maximum clock rate or throughput. Instead, the timing resolution and count rate of single-photon detectors define the maximum throughput of these systems. In this talk, we will review progress towards next-generation SNSPDs which combine many state-of-the-art metrics into a single system. These detectors will be showcased in a series of record-setting communication protocols: NASA’s deep space optical communication demonstration, a wavelength-multiplexed 12.5 GHz entanglement distribution system, and a 20 GHz single-photon laser communication system. We will also discuss on-going efforts to develop arrays of SNSPDs for next generation wavelength-division multiplexed quantum networks.

Abstract: TBA

Abstract: Non-local quantum computation (NLQC) replaces a localinteraction between two systems with a single round of communication and sharedentanglement. It arises naturally in the analysis of attacks on quantumposition verification (QPV). Despite many partial results, it is known that acharacterization of entanglement cost in at least certain NLQC tasks wouldimply significant breakthroughs in complexity theory. Here, we avoid theseobstructions and take an indirect approach to understanding resource requirementsin NLQC, which mimics the approach used by complexity theorists: we study therelative hardness of different NLQC tasks by identifying resource efficientreductions between them. Most significantly, we prove that f-measure (the taskof measuring in a certain BB84 basis according to a classical function f) andf-route (the task of routing a qubit according to a classical function f), thetwo best studied NLQC tasks, are in fact equivalent under O(1) overheadreductions. This result simplifies many existing proofs in the literature asresults were often proved separately for the two, it also extends several newproperties to f-measure that were only known for f-routing. For instance, weobtain sub-exponential upper bounds on f-measure for all functions.This talk is based on joint work with the same title withAndreas Bluhm, Simon Höfer, Alex May, Mikka Stasiuk and Henry Yuen https://arxiv.org/abs/2505.23893

Abstract: Coherent states conveniently represent the classically meaningful wavelike states of light, as opposed to the corpuscular number states, but superposingcoherent states is mind-bogglingly unclassical, including as examples "cat states", "comb states" and "compass states". I present a history of superposing coherent states, especially for oscillators (including electromagnetic field states) and for spin, and extend to entangled coherent states. I then show that superposed coherent states are a viable path for photonic quantum computing and for reaching towards the ultimate limits of sensing. Finally, I discuss our experimental realisations (at University of New South Wales) of superposed coherent states in a spin-7/2 Antimony nucleus in a silicon substrate.

Abstract: This talk traces the development of quantum physics throughout the 20th century, emphasizing its intricate entanglement with economic, political, and cultural contexts. Unexpected associations—linking streetlights to Planck’s quantum of action, lasers to ballet dancers, and timekeeping to ion traps—serve as narrative threads illustrating the multifaceted nature of quantum discoveries. These examples, drawn from the “Light and Matter” exhibition at the Deutsches Museum in Munich, Germany, offer historical vantage points from which to reflect on the evolving landscape of quantum technologies today.

The workshop concerns both the theoretical and experimental aspects of this subject. This workshop aims to build connections between young scientists and promotes open discussions between participants. As such, 30-minute talks by all participants will be held at the beginning of the week, leaving two full days open for discussions on the presented work (and beyond).

Please note that participation in this workshop is by invitation only.

Quantum Techniques in Machine Learning (QTML) is a leading international conference at the forefront of quantum science and machine learning. Held annually, it brings together researchers and industry experts to explore how quantum computing can transform learning, optimisation, and data-driven discovery. CQT is proud to support planning for the ninth edition of this international conference, QTML 2025, to take place in Singapore 16-21 November. Registration will open from 11 August.