About

Dirk Englund is a professor affiliated with the MIT Department of Electrical Engineering and Computer Science, with a focus on research areas including electronic, magnetic, optical, and quantum materials and devices, as well as nanoscale materials, devices, and systems. His work involves developing groundbreaking sensors, energy transducers, and physical substrates for computation, addressing shared human challenges through experimental, theoretical, and computational tools. His research also encompasses quantum computing, communication, and sensing, contributing to advancements in photonic devices, quantum devices, and high-speed optical communications. Englund's contributions include the development of new photonic devices such as light-emitting structures for advanced displays and quantum computers, as well as photonic processors capable of high-speed AI computations and real-time data analysis. His recent work has led to innovations like devices that perform deep learning at the speed of light and enable ultrafast AI computations with extreme energy efficiency. His research aims to enable the design of quantum devices, improve atomic-scale imaging, and advance wireless signal processing technologies, reflecting a broad impact across quantum information science and photonics.

Research topics

• Computer Science
• Physics
• Engineering
• Quantum mechanics
• Electronic engineering
• Electrical engineering
• Data science
• Materials science
• Telecommunications
• Optoelectronics
• Artificial Intelligence
• Optics
• Engineering physics
• Library science
• Computational physics
• Condensed matter physics
• Computer architecture
• History

Selected publications

• Single-shot matrix-matrix photonic processor based on spatial-spectral hypermultiplexed parallel diffraction

  Nature Communications · 2026-01-13 · 1 citations

  articleOpen access

  The ever-increasing data demand craves advancements in high-speed and energy-efficient computing hardware. Analog optical neural network (ONN) processors have emerged as a promising solution, offering benefits in bandwidth and energy consumption. However, existing ONN processors exhibit limited computational parallelism, and while certain architectures achieve high parallelism, they encounter serious scaling up roadblocks for large-scale implementation. Here, we introduce a spatial-wavelength-temporal hyper-multiplexed ONN processor, which is based on parallel diffractive beam routing. The architecture supports high three-dimensional data, high O(N3) computing parallelism, and is feasible for large-scale implementation. A 16 × 16 parallel diffractive beam routing is demonstrated, enabling a large-scale (16 × 16 − by − 16 × 16), high-parallelism (4096 multiply-and-accumulates/shot (MACs/shot)), high-speed (2 Gsa/s), single-shot matrix-matrix multiplication (MMM) optical tensor processor. It accelerates convolutional neural networks (CNNs) and deep neural networks (DNNs) through parallel matrix multiplication. We demonstrate benchmark image recognition using a CNN and a subsequently fully connected DNN in the optical domain. The network works with an ultra-low optical energy of ≈ 20 attojoules (aJ)/MAC at 96.4% classification accuracy. The ONN system supports broad spectral and spatial bandwidths and is capable for large-scale scaling up, paving the way for highly efficient large-scale optical computing for next-generation deep learning. Optical neural network processors offering benefits in bandwidth and energy consumption but problems in scaling and parallelism. The author presenting a novel optical tensor processor capable of optically performing large-scale, high-speed matrix-matrix multiplication in a single step.

  PublisherOA PDFDOI

• Thermal detection of single photons using Dirac fermions

  Nature Communications · 2026-03-12 · 1 citations

  articleOpen access

  Abstract Detecting single photons is a crucial process in quantum science, quantum networking, biology, and advanced imaging. To detect the small quantum of energy carried in a photon, conventional mechanisms rely on energy excitation across either a semiconductor bandgap or superconducting gap that hinders their applications to low-energy photons. Here, we detect single near-infrared photons using the thermal properties of Dirac fermions in graphene. By exploiting the extremely low heat capacity of Dirac electrons near its charge neutrality point, we observe a temperature rise up to ~ 2 K using a hybrid Josephson junction. In this proof-of-principle experiment, we achieve an intrinsic quantum efficiency of 87% (75%) with dark count &lt; 1 per second (per week), reaching an effective noise equivalent power of 2 × 10 −22 W/ $$\sqrt{{{{\rm{Hz}}}}}$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:msqrt> <mml:mrow> <mml:mi>Hz</mml:mi> </mml:mrow> </mml:msqrt> </mml:math> . The highest operation temperature is 1.2 K. Our results highlight the potential of graphene bolometers for detecting lower-energy photons from the mid-IR to microwave regimes, opening pathways to study space science in far-infrared regime, to potential applications in dark matter searches, and to advance quantum technologies across a broader electromagnetic spectrum.

  PublisherOA PDFDOI

• Foundry-Enabled Patterning of Diamond Quantum Microchiplets for Scalable Quantum Photonics

  Nano Letters · 2026-05-07

  articleSenior authorCorresponding

  Quantum technologies promise secure communication and advanced information processing, but scaling these systems remains a challenge. Diamond is a promising platform because it hosts defects that emit single photons and store quantum information with high stability. However, the conventional fabrication of diamond optical structures is slow and difficult to scale. Here, we present a manufacturing approach that moves diamond quantum photonics closer to industrial production. Instead of patterning each device directly on diamond, we create high-precision silicon masks in commercial foundries and transfer them onto diamond by using microtransfer printing. This enables large arrays of nanoscale optical structures while improving uniformity, yield, and throughput. Using this method, we demonstrate hundreds of diamond quantum microchiplets with enhanced optical performance and controlled coupling to quantum emitters. The chiplet approach also allows faulty devices to be replaced and supports integration with existing photonic and electronic systems, offering a scalable path toward practical quantum technologies.

  PublisherDOI

• Ax-Prover: A Deep Reasoning Agentic Framework for Theorem Proving in Mathematics and Quantum Physics

  ArXiv.org · 2025-10-14

  preprintOpen accessSenior author

  We present Ax-Prover, a multi-agent system for automated theorem proving in Lean that can solve problems across diverse scientific domains and operate either autonomously or collaboratively with human experts. To achieve this, Ax-Prover approaches scientific problem solving through formal proof generation, a process that demands both creative reasoning and strict syntactic rigor. Ax-Prover meets this challenge by equipping Large Language Models (LLMs), which provide knowledge and reasoning, with Lean tools via the Model Context Protocol (MCP), which ensure formal correctness. To evaluate its performance as an autonomous prover, we benchmark our approach against frontier LLMs and specialized prover models on two public math benchmarks and on two Lean benchmarks we introduce in the fields of abstract algebra and quantum theory. On public datasets, Ax-Prover is competitive with state-of-the-art provers, while it largely outperforms them on the new benchmarks. This shows that, unlike specialized systems that struggle to generalize, our tool-based agentic theorem prover approach offers a generalizable methodology for formal verification across diverse scientific domains. Furthermore, we demonstrate Ax-Prover's assistant capabilities in a practical use case, showing how it enabled an expert mathematician to formalize the proof of a complex cryptography theorem.

  PublisherOA PDFDOI

• Programmable Quantum Matter: Heralding Large Cluster States in Driven Inhomogeneous Spin Ensembles

  ArXiv.org · 2025-09-03

  preprintOpen accessSenior author

  Atom-like emitters in solids are promising platforms for quantum sensing and information processing, but inhomogeneities in the emitter fine structure complicate quantum control. We present a framework that leverages this diversity to reduce the resources for generating optically heralded spin cluster states across $N_q$ emitters from the conventional order $O(N_q)$ to $O(1)$ in ensembles of $N_q \sim 10$-$100$. An optimized pulse sequence simultaneously corrects pulse-length and detuning errors, achieving single-qubit gate fidelities exceeding $99.99\%$ for errors (normalized relative to the Rabi drive strength) up to 0.3, while maintaining fidelities above $99\%$ for errors as large as 0.4. Applied as a Carr-Purcell-Meiboom-Gill (CPMG) dynamical decoupling protocol to the dominant noise spectrum of silicon-vacancy centers in diamond, it enhances ensemble coherence times by over $7\times$ compared to interleaved bang-bang based CPMG. For state-of-the-art dilution refrigerators, global resonant optimal decoupling across $N_q$ spins sharply reduces heating, addressing the trade-off between the spin coherence and scaling to $N_q \gg 1$. We further introduce a modified single-photon entanglement protocol with an efficient algorithm for deterministic entanglement compilation. Depending on the decoupling time window, our method yields order $O(10^2$-$10^4)$ more entanglement links than bang-bang sequences, with theoretical guarantees of order $Ω(N_q)$ unique links, improvable by control tuning. Together, these techniques provide scalable tools - including global control, phase denoising, remote entanglement, and compilation - for robust quantum computing architectures with heterogeneous spin ensembles.

  PublisherOA PDFDOI

• On-chip high-order parametric downconversion in the excitonic Mott insulator Nb$_3$Cl$_8$ for programmable multiphoton entangled states

  ArXiv.org · 2025-12-01

  preprintOpen access

  Spontaneous parametric downconversion (SPDC) and four-wave mixing in $χ^{(2)}$ and $χ^{(3)}$ media underpin most entangled-photon sources, but direct generation of higher-order entangled multiphoton states by $n$-th order parametric downconversion remains extremely challenging because conventional materials exhibit tiny high-order nonlinearities. Here we show that single-layer Nb$_3$Cl$_8$, an excitonic Mott insulator on a breathing Kagome lattice, supports exceptionally large nonlinear susceptibilities up to seventh order. Many-body GW--Bethe--Salpeter and time-dependent BSE / Kadanoff--Baym simulations yield resonant $χ^{(2)}$--$χ^{(7)}$ for monolayer Nb$_3$Cl$_8$, with $|χ^{(4)}|$ and $|χ^{(5)}|$ surpassing values in prototypical transition metal dichalcogenides by 5--9 orders of magnitude. We trace this enhancement to flat bands and strongly bound Frenkel excitons with ferroelectrically aligned out-of-plane dipoles. Building on experimentally demonstrated 1$\times N$ integrated beam splitters with arbitrary power ratios, we propose an on-chip architecture where each output arm hosts an Nb$_3$Cl$_8$ patch, optionally gated by graphene to tune the complex $n$-photon amplitudes. Using the ab-initio $χ^{(3)}$ and $χ^{(4)}$ values, we predict that three-photon GHZ$_3$ and four-photon cluster-state sources in this platform can achieve $n$-photon generation rates up to $\sim 10^8$ and $\sim 10^6$ times larger, respectively, than silica-fiber- and MoS$_2$-based implementations with comparable geometry. We derive the quantum Hamiltonian and explicit $n$-photon generation rates for this platform, and show how suitable interferometric networks enable electrically and spectrally tunable GHZ, $W$, and cluster states based on genuine high-order nonlinear processes in a 2D excitonic Mott insulator.

  PublisherOA PDFDOI

• LightCode: Compiling LLM Inference for Photonic-Electronic Systems

  ArXiv.org · 2025-09-19

  preprintOpen accessSenior author

  The growing demand for low-latency, energy-efficient inference in large language models (LLMs) has catalyzed interest in heterogeneous architectures. While GPUs remain dominant, they are poorly suited for integration with emerging domain-specific accelerators like the Photonic Tensor Units (PTUs), which offer low-power, high-throughput linear computation. This motivates hybrid compilation strategies that combine photonic and electronic resources. We present LightCode, a compiler framework and simulator for mapping LLM inference workloads across hybrid photonic-electronic systems. LightCode introduces the Stacked Graph, an intermediate representation that encodes multiple hardware-specific realizations of each tensor operation. Hardware assignment is formulated as a constrained subgraph selection problem optimized for latency or energy under parametric cost models. We evaluate LightCode on the prefill stage of GPT-2 and Llama-7B showing that under our workload and hardware assumptions, (i) Photonic hardware reduced energy by up to 50% in our simulated workloads at maximum sequence length; (ii) multiplexing and assignment strategy yielded latency improvements exceeding 10x; and (iii) Optimizing for latency or energy resulted in distinct hardware mappings in our simulations. LightCode offers a module, foundational framework and simulator for compiling LLMs to emerging photonic accelerators.

  PublisherOA PDFDOI

• Vector magnetometry using cavity-enhanced microwave readout in nitrogen-vacancy diamond

  ArXiv.org · 2025-11-14

  preprintOpen access

  We demonstrate $4π$-steradian vector magnetic field sensing using an ensemble of nitrogen-vacancy (NV) centers in a single-crystal diamond coupled to a microwave (MW) cavity. The MW cavity enhances the spin-photon coupling which enables efficient, high-contrast spin-state readout via MW interrogation and removes the need for bulky optical collection components. An applied AC bias magnetic field lifts the zero-field degeneracy of the four crystallographic NV orientations, allowing each orientation to be individually addressed and used for vector reconstruction of the magnetic field. The resulting magnetometer has a 40\% contrast (20x higher than typical for optical spin-ensemble readout) and achieves a single-axis sensitivity of 250 pT/$\sqrt{\mathrm{Hz}}$ which is flat from DC to 1 kHz. Noise models of the composite spin-cavity system establish MW amplitude noise as the dominant noise source and predict a thermal noise limit of 2 pT/$\sqrt{\mathrm{Hz}}$.

  PublisherOA PDFDOI

• Supplemental document

  AIP Publishing · 2025-12-30

  articleOpen access

  Data and additional information for the main manuscript.

  PublisherDOI

• Breaking On/Off-coupling Loss Degeneracies via Bidirectional Nonlinear Optics

  arXiv (Cornell University) · 2025-10-15

  preprintOpen access

  Accurate evaluation of nonlinear photonic integrated circuits requires separating input and output coupling efficiencies (i.e., $η_1$ and $η_2$), yet the conventional linear-transmission calibration method recovers only their product (i.e., $η_1\,η_2$) and therefore introduces systematic bias when inferring on-chip performance from off-chip data. We present bidirectional nonlinear optical tomography (BNOT), a direction-aware metrology that uses forward and backward pumping of complementary nonlinear probes, with process-appropriate detection, to break the ``degeneracy'' of $η_1\,η_2$ and estimate individual interface efficiencies with tight confidence intervals. The method links off-chip measurements to on-chip quantities through a compact observation model that explicitly incorporates pump fluctuations and detector noise, and it frames efficiency extraction as a joint constrained optimization. Monte Carlo studies show unbiased convergence of the estimated efficiencies to ground truth with low error across realistic operating regimes. Using these efficiency estimates to reconstruct on-chip nonlinear figures of merit yields distributions centered on the true values with reduced variance, whereas conventional ``degenerate'' calibration is biased and can substantially misestimate on-chip performance. BNOT is hardware-compatible and platform-agnostic, and provides unbiased characterization of off- and on-chip coupling efficiencies across nonlinear processes, enabling reproducible, coupling-resolved benchmarking for scalable systems in quantum optics, frequency conversion, and precision metrology.

  PublisherOA PDFDOI

Recent grants

• NIH Grant R43MH102942

  NIH · $289k · 2016

• Collaborative research: Quantum Communication with Loss-Protected Photonic Encoding

  NSF · $263k · 2019–2022

• EFRI ACQUIRE: Scalable Quantum Communications with Error-Corrected Semiconductor Qubits

  NSF · $2.0M · 2016–2021

• EAGER:Scalable Photonic AI Accelerators Based on Photoelectric Multiplication

  NSF · $200k · 2019–2022

• RAISE TAQS: Very Large Scale Integrated Electronics and Phontonics Platform for Scaleable Quantum Information Processing

  NSF · $999k · 2018–2022

Frequent coauthors

• Matthew E. Trusheim

  DEVCOM Army Research Laboratory

  172 shared

• Jelena Vučković

  112 shared

• Ryan Hamerly

  103 shared

• Mark Dong

  82 shared

• Matt Eichenfield

  81 shared

• Kevin C. Chen

  80 shared

• Tim Schröder

  Humboldt-Universität zu Berlin

  78 shared

• Hyeongrak Choi

  Massachusetts Institute of Technology

  77 shared

Education

• BS, Physics

  California Institute of Technology

  2012

• PhD, Applied Physics

  Stanford University

  2008

• MS, Electrical Engineering

  Stanford University

  2008