Colloquia

Scott A. Diddams

Optical Frequency Combs and Quantum Metrology

The optical frequency comb is one of the most significant advances in laser physics since the development of the laser itself. It has made routine the counting and synthesis of the oscillations of light on the femtosecond time scale, and it is an essential component of all present and future optical clocks and time-transfer systems. It further enables the most accurate measurement of any fundamental physical quantity—that of the quantized energy states of atoms and ions with 18 digits of precision. Despite this close connection to quantum systems, there are few demonstrations of how an optical frequency comb could yield a quantum advantage for metrology. The most important limitation remains in photodetection, where shot noise sets the fundamental signal-to-noise ratio. However, there are important and impactful differences in the detection of frequency comb light that yield surprising results, with time-stationary shot-noise limits being surpassed. We are exploring these limits with the goal of defining the standard quantum limit for metrology with optical frequency combs. Highlights will be provided for measurement scenarios that impact applications in clocks, climate and health diagnostics and exoplanet searches. Moreover, I will discuss frequency comb interferometry with non-classical states of light and electric-field-correlation spectroscopy of thermal light that could form a critical component of future long baseline interferometry in the mid and near-infrared.

 Scott Diddams holds the Robert H. Davis Chair at the University of Colorado Boulder, where he is also Professor of Electrical Engineering and Physics. He carries out experimental research in the fields of precision spectroscopy and quantum metrology, nonlinear optics, microwave photonics and ultrafast lasers. Diddams received the Ph.D. degree from the University of New Mexico in 1996.  From 1996 through 2000, he did postdoctroral work at JILA, NIST and the University of Colorado. Subsequently, Diddams was a Research Physicist, Group Leader, and Fellow at NIST (the National Institute of Standards and Technology). In 2022 he transitioned to his present position where he also assumed the role of Faculty Director of the Quantum Engineering Initiative in the College of Engineering and Applied Science. As a postdoc Diddams built the first optical frequency combs in the lab of Nobel laureate John Hall, and throughout his career, he has pioneered the use of these tools for optical clocks, tests of fundamental physics, novel spectroscopy, and astronomy.  His research has been documented in more than 750 peer-reviewed publications, conference papers, and invited talks. The work of Dr. Diddams and his research group has also been recognized by multiple awards. These include the Distinguished Presidential Rank Award, the Department of Commerce Gold and Silver Medals for “revolutionizing the way frequency is measured”, as well as the Presidential Early Career Award in Science and Engineering (PECASE), the OPTICA C.E.K. Mees Medal, the IEEE Photonics Society Laser Instrumentation Award, and the IEEE Rabi award. He is a Fellow of OPTICA (formerly OSA), the American Physical Society, IEEE and a member of the US National Academy of Engineering.

Scott A. Diddams

Optical Frequency Combs and Quantum Metrology

The optical frequency comb is one of the most significant advances in laser physics since the development of the laser itself. It has made routine the counting and synthesis of the oscillations of light on the femtosecond time scale, and it is an essential component of all present and future optical clocks and time-transfer systems. It further enables the most accurate measurement of any fundamental physical quantity—that of the quantized energy states of atoms and ions with 18 digits of precision. Despite this close connection to quantum systems, there are few demonstrations of how an optical frequency comb could yield a quantum advantage for metrology. The most important limitation remains in photodetection, where shot noise sets the fundamental signal-to-noise ratio. However, there are important and impactful differences in the detection of frequency comb light that yield surprising results, with time-stationary shot-noise limits being surpassed. We are exploring these limits with the goal of defining the standard quantum limit for metrology with optical frequency combs. Highlights will be provided for measurement scenarios that impact applications in clocks, climate and health diagnostics and exoplanet searches. Moreover, I will discuss frequency comb interferometry with non-classical states of light and electric-field-correlation spectroscopy of thermal light that could form a critical component of future long baseline interferometry in the mid and near-infrared.

Scott Diddams holds the Robert H. Davis Chair at the University of Colorado Boulder, where he is also Professor of Electrical Engineering and Physics. He carries out experimental research in the fields of precision spectroscopy and quantum metrology, nonlinear optics, microwave photonics and ultrafast lasers. Diddams received the Ph.D. degree from the University of New Mexico in 1996.  From 1996 through 2000, he did postdoctroral work at JILA, NIST and the University of Colorado. Subsequently, Diddams was a Research Physicist, Group Leader, and Fellow at NIST (the National Institute of Standards and Technology). In 2022 he transitioned to his present position where he also assumed the role of Faculty Director of the Quantum Engineering Initiative in the College of Engineering and Applied Science. As a postdoc Diddams built the first optical frequency combs in the lab of Nobel laureate John Hall, and throughout his career, he has pioneered the use of these tools for optical clocks, tests of fundamental physics, novel spectroscopy, and astronomy.  His research has been documented in more than 750 peer-reviewed publications, conference papers, and invited talks. The work of Dr. Diddams and his research group has also been recognized by multiple awards. These include the Distinguished Presidential Rank Award, the Department of Commerce Gold and Silver Medals for “revolutionizing the way frequency is measured”, as well as the Presidential Early Career Award in Science and Engineering (PECASE), the OPTICA C.E.K. Mees Medal, the IEEE Photonics Society Laser Instrumentation Award, and the IEEE Rabi award. He is a Fellow of OPTICA (formerly OSA), the American Physical Society, IEEE and a member of the US National Academy of Engineering.

Jun Ye

Quantum System Engineering for Clocks and Fundamental Physics

Laser and quantum sciences have fueled revolutionary developments in atomic, molecular, and fundamental physics. Quantum state engineering and many-body physics provide coherent quantum systems at increasingly large sizes, revolutionizing the performance of clocks and metrology and promising new discovery opportunities. Quantum technology has brought many thousands of atoms to minute-long coherence times, and it is now also knocking on the door of nuclear physics, heralded by the recent breakthrough of quantum-state-resolved laser spectroscopy of thorium-229 nuclear transition. These progresses in quantum metrology provide new tools for quantum sensing, and raise the prospect of using quantum sensors to search for new physics and probe the interface of gravity and quantum mechanics.

Jun Ye

Quantum System Engineering for Clocks and Fundamental Physics

Laser and quantum sciences have fueled revolutionary developments in atomic, molecular, and fundamental physics. Quantum state engineering and many-body physics provide coherent quantum systems at increasingly large sizes, revolutionizing the performance of clocks and metrology and promising new discovery opportunities. Quantum technology has brought many thousands of atoms to minute-long coherence times, and it is now also knocking on the door of nuclear physics, heralded by the recent breakthrough of quantum-state-resolved laser spectroscopy of thorium-229 nuclear transition. These progresses in quantum metrology provide new tools for quantum sensing, and raise the prospect of using quantum sensors to search for new physics and probe the interface of gravity and quantum mechanics.

Sonia Buckley

Emergent Learning in Physical Neural Networks

Abstract: Modern machine learning has become incredibly powerful, driving technologies like image recognition, translation, and voice assistants. However, the way these machine learning algorithms work is very different from how the brain learns. For example, backpropagation, the standard chain-rule algorithm used to calculate the gradient for “stochastic gradient descent”, is highly implausible in the brain [1]. This points to a fundamental gap in our understanding of how learning happens in the brain, and provides us with hints as to why developing analog hardware for AI has been so challenging. One promising direction is to develop physical neural networks—systems where the neural network is embodied directly in hardware rather than simulated on a digital computer—and to explore how such systems might learn on their own.

In this talk, we introduce multiplexed-gradient descent (MGD) [2,3], a bottom-up framework for designing physical neural networks that learn autonomously. In MGD, individual devices are engineered so that, when combined, the entire system exhibits stochastic gradient-descent learning as an emergent property. We demonstrate through simulation how the MGD approach can train standard image recognition benchmarks in a similar amount of time to a typical GPU, while requiring no knowledge of individual device properties. We also show an example of using MGD in analog photonic hardware to continually adapt the weights to solve a time-varying classification task. Finally, we share early results on building a mixed analog-digital testbed to explore how emergent learning could scale into future hardware systems.

[1] Lillicrap, T.P., Santoro, A., Marris, L. et al. “Backpropagation and the brain”. Nat Rev Neurosci 21, 335–346 (2020).
[2] A. N. McCaughan, B. G. Oripov, N. Ganesh; S. W. Nam, A. Dienstfrey, S. M. Buckley “Multiplexed gradient descent: Fast online training of modern datasets on hardware neural networks without backpropagation” APL Mach. Learn. 1, 026118 (2023)
[3] Oripov, B.G., Dienstfrey, A., McCaughan, A.N. and Buckley, S.M., “Scaling of hardware-compatible perturbative training algorithms”. APL Machine Learning, 3(2) (2025)
[4] Guo, Z., Aadhi, A., McCaughan, A.N., Tait, A.N., Youngblood, N., Buckley, S.M. and Shastri, B.J., “Fully analog end-to-end online training with real-time adaptability on integrated photonic platform”. arXiv:2506.18041 (2025)

Bio: Dr. Sonia Buckley is a physicist at the National Institute of Standards and Technology (NIST), where she leads research in critical and emerging technologies, including quantum photonics metrology and standards, and the development of novel hardware and benchmarks for artificial intelligence. Her work aims to bridge the gaps between academic research and new commercial technologies. Dr. Buckley obtained her PhD in Applied Physics and MS in Electrical engineering from Stanford University in 2015.

Sonia Buckley

Emergent Learning in Physical Neural Networks

Abstract: Modern machine learning has become incredibly powerful, driving technologies like image recognition, translation, and voice assistants. However, the way these machine learning algorithms work is very different from how the brain learns. For example, backpropagation, the standard chain-rule algorithm used to calculate the gradient for “stochastic gradient descent”, is highly implausible in the brain [1]. This points to a fundamental gap in our understanding of how learning happens in the brain, and provides us with hints as to why developing analog hardware for AI has been so challenging. One promising direction is to develop physical neural networks—systems where the neural network is embodied directly in hardware rather than simulated on a digital computer—and to explore how such systems might learn on their own.

In this talk, we introduce multiplexed-gradient descent (MGD) [2,3], a bottom-up framework for designing physical neural networks that learn autonomously. In MGD, individual devices are engineered so that, when combined, the entire system exhibits stochastic gradient-descent learning as an emergent property. We demonstrate through simulation how the MGD approach can train standard image recognition benchmarks in a similar amount of time to a typical GPU, while requiring no knowledge of individual device properties. We also show an example of using MGD in analog photonic hardware to continually adapt the weights to solve a time-varying classification task. Finally, we share early results on building a mixed analog-digital testbed to explore how emergent learning could scale into future hardware systems.

[1] Lillicrap, T.P., Santoro, A., Marris, L. et al. “Backpropagation and the brain”. Nat Rev Neurosci 21, 335–346 (2020).
[2] A. N. McCaughan, B. G. Oripov, N. Ganesh; S. W. Nam, A. Dienstfrey, S. M. Buckley “Multiplexed gradient descent: Fast online training of modern datasets on hardware neural networks without backpropagation” APL Mach. Learn. 1, 026118 (2023)
[3] Oripov, B.G., Dienstfrey, A., McCaughan, A.N. and Buckley, S.M., “Scaling of hardware-compatible perturbative training algorithms”. APL Machine Learning, 3(2) (2025)
[4] Guo, Z., Aadhi, A., McCaughan, A.N., Tait, A.N., Youngblood, N., Buckley, S.M. and Shastri, B.J., “Fully analog end-to-end online training with real-time adaptability on integrated photonic platform”. arXiv:2506.18041 (2025)

Bio: Dr. Sonia Buckley is a physicist at the National Institute of Standards and Technology (NIST), where she leads research in critical and emerging technologies, including quantum photonics metrology and standards, and the development of novel hardware and benchmarks for artificial intelligence. Her work aims to bridge the gaps between academic research and new commercial technologies. Dr. Buckley obtained her PhD in Applied Physics and MS in Electrical engineering from Stanford University in 2015.

The graduate students won the showdown despite the heroic efforts of the faculty!

The cup will now be displayed at a location of the graduate students choosing for the whole year with associated bragging rights.

The graduate students won the showdown despite the heroic efforts of the faculty!

The cup will now be displayed at a location of the graduate students choosing for the whole year with associated bragging rights.

Raphael Pestourie

Input-Space Machine Learning to Accelerate Physics-Certified Inverse Design in Nanophotonics

Designing nanophotonic devices (metamaterials and metasurfaces) requires large electromagnetic simulations that resolve sub-wavelength features across wide domains; making iterative design loops prohibitively expensive. I will show how machine learning may address this challenge in two ways: (1) training fast, accurate surrogate-based approximate solvers from limited full-wave simulations on subdomains, and (2) learning structure in the input design space to either guide searches toward promising configurations or build more trustworthy surrogate models. This input-space learning accelerates optimization while preserving physics-based fidelity, enabling large-scale, physics-certified inverse design.

Raphael Pestourie

Input-Space Machine Learning to Accelerate Physics-Certified Inverse Design in Nanophotonics

Designing nanophotonic devices (metamaterials and metasurfaces) requires large electromagnetic simulations that resolve sub-wavelength features across wide domains; making iterative design loops prohibitively expensive. I will show how machine learning may address this challenge in two ways: (1) training fast, accurate surrogate-based approximate solvers from limited full-wave simulations on subdomains, and (2) learning structure in the input design space to either guide searches toward promising configurations or build more trustworthy surrogate models. This input-space learning accelerates optimization while preserving physics-based fidelity, enabling large-scale, physics-certified inverse design.

David Schmidt

From Ignition to Energy: Laser Fusion’s Next Chapter

The National Ignition Facility’s (NIF) achievement of fusion ignition marked a historic milestone for inertial fusion energy (IFE), proving that lasers can ignite fusion reactions. The next challenge is turning this scientific success into a practical and economically viable energy source. In this talk, I will discuss how Xcimer Energy is advancing laser-driven IFE by rethinking laser architecture, target design, and system scale. Our approach leverages excimer laser technology and nonlinear gas–laser phenomena to achieve dramatically higher energy, efficiency, and repetition rates than traditional solid-state systems—enabling the scale required for fusion power plants.

I will highlight key technical innovations, upcoming projects, and the remaining challenges in target physics, beam conditioning, and large-scale integration.

David Schmidt

From Ignition to Energy: Laser Fusion’s Next Chapter

The National Ignition Facility’s (NIF) achievement of fusion ignition marked a historic milestone for inertial fusion energy (IFE), proving that lasers can ignite fusion reactions. The next challenge is turning this scientific success into a practical and economically viable energy source. In this talk, I will discuss how Xcimer Energy is advancing laser-driven IFE by rethinking laser architecture, target design, and system scale. Our approach leverages excimer laser technology and nonlinear gas–laser phenomena to achieve dramatically higher energy, efficiency, and repetition rates than traditional solid-state systems—enabling the scale required for fusion power plants.

I will highlight key technical innovations, upcoming projects, and the remaining challenges in target physics, beam conditioning, and large-scale integration.

Stephanie Wissel

Tuning into Cosmic Neutrinos at High Elevation 

Neutrinos are the ideal messenger for high-energy astrophysics. Weakly interacting and uncharged, they propagate undeterred and unabsorbed through the universe. In the last decade, the IceCube experiment has brought us the discovery of a flux of high-energy, TeV-scale neutrinos, and through a multi-messenger lens — the combined observations of neutrinos and other messengers like photons —  we are starting to see hints of energetic neutrino sources for the first time. At higher energies still, beyond the PeV scale, we can probe the most energetic sources of both neutrinos and cosmic rays, but current neutrino experiments become too small to observe a sizable flux. Radio experiments can achieve the large exposures necessary by taking advantage of the coherent broadband radio emission resulting from ultra-high-energy (E>10^17 eV) neutrino interactions as well as the large volumes visible from high elevations. In this talk, I will review results from current and future high-elevation radio experiments, both from balloon-borne instruments like PUEO and from mountaintop experiments like BEACON and HERON.

Stephanie Wissel

Tuning into Cosmic Neutrinos at High Elevation 

Neutrinos are the ideal messenger for high-energy astrophysics. Weakly interacting and uncharged, they propagate undeterred and unabsorbed through the universe. In the last decade, the IceCube experiment has brought us the discovery of a flux of high-energy, TeV-scale neutrinos, and through a multi-messenger lens — the combined observations of neutrinos and other messengers like photons —  we are starting to see hints of energetic neutrino sources for the first time. At higher energies still, beyond the PeV scale, we can probe the most energetic sources of both neutrinos and cosmic rays, but current neutrino experiments become too small to observe a sizable flux. Radio experiments can achieve the large exposures necessary by taking advantage of the coherent broadband radio emission resulting from ultra-high-energy (E>10^17 eV) neutrino interactions as well as the large volumes visible from high elevations. In this talk, I will review results from current and future high-elevation radio experiments, both from balloon-borne instruments like PUEO and from mountaintop experiments like BEACON and HERON.

Michael Litos

Miniaturizing Particle Accelerators with the Power of Plasma

Plasma-based particle accelerators offer an opportunity to significantly reduce the size and cost of high-energy particle beams for applications ranging from ultrafast electron diffraction, to X-ray free electron lasers, to high-energy particle colliders. These applications in turn serve users in a variety of research fields by permitting access to ultrafast dynamics at atomic scales, or even fundamental particle interactions. Plasma wakefield accelerators (PWFAs) can sustain accelerating electric fields that are orders of magnitude greater than conventional metallic accelerating structures due in part to the fact that the plasma medium cannot itself be destroyed by the fields, in contrast to metallic structures. Researchers have shown that PWFAs can provide the promised large rates of acceleration to electron bunches, and the next great challenge for the field is to is to preserve the quality (i.e. emittance) of the accelerated bunches. This will be achieved by utilizing the plasma source itself to precisely focus the electron bunches into the PWFA, matching the natural divergence of the electron beam to the strong focusing force experienced in the plasma. Experiments planned at SLAC National Accelerator Laboratory’s FACET-II facility aim to accomplish this alongside other tangential research goals utilizing relativistic particle beams and plasmas.

Bio: Dr. Litos received his PhD in 2010 from Boston University where his research focused on neutrino oscillations and proton decay. He changed his research direction to plasma wakefield acceleration when he began his postdoc at SLAC National Accelerator Laboratory’s Facility for Advanced Accelerator Experimental Tests (FACET). He was promoted to staff scientist at SLAC in 2014, and then became a faculty member at the University of Colorado Boulder in 2016, where he continues to focus on plasma wakefield acceleration and closely related topics, carrying out research in his university laser lab, as well as at SLAC’s FACET-II facility.

Michael Litos

Miniaturizing Particle Accelerators with the Power of Plasma

Plasma-based particle accelerators offer an opportunity to significantly reduce the size and cost of high-energy particle beams for applications ranging from ultrafast electron diffraction, to X-ray free electron lasers, to high-energy particle colliders. These applications in turn serve users in a variety of research fields by permitting access to ultrafast dynamics at atomic scales, or even fundamental particle interactions. Plasma wakefield accelerators (PWFAs) can sustain accelerating electric fields that are orders of magnitude greater than conventional metallic accelerating structures due in part to the fact that the plasma medium cannot itself be destroyed by the fields, in contrast to metallic structures. Researchers have shown that PWFAs can provide the promised large rates of acceleration to electron bunches, and the next great challenge for the field is to is to preserve the quality (i.e. emittance) of the accelerated bunches. This will be achieved by utilizing the plasma source itself to precisely focus the electron bunches into the PWFA, matching the natural divergence of the electron beam to the strong focusing force experienced in the plasma. Experiments planned at SLAC National Accelerator Laboratory’s FACET-II facility aim to accomplish this alongside other tangential research goals utilizing relativistic particle beams and plasmas.

Bio: Dr. Litos received his PhD in 2010 from Boston University where his research focused on neutrino oscillations and proton decay. He changed his research direction to plasma wakefield acceleration when he began his postdoc at SLAC National Accelerator Laboratory’s Facility for Advanced Accelerator Experimental Tests (FACET). He was promoted to staff scientist at SLAC in 2014, and then became a faculty member at the University of Colorado Boulder in 2016, where he continues to focus on plasma wakefield acceleration and closely related topics, carrying out research in his university laser lab, as well as at SLAC’s FACET-II facility.