arXiv subjects

Sergey Frolov

Publications and source records attributed to Sergey Frolov.

At least 19 recordsLinked to original sources

arXiv2026

Bridging short- and medium-range weather forecasting with machine learning

The National Oceanic and Atmospheric Administration (NOAA) employs independent prediction systems for distinct forecast products. While some separation is practical, we argue that combining short- and medium-range weather into a single prediction system would provide the public with a useful distillation of global weather and its impacts. To this end, we present Nested-EAGLE (Experimental Artificial intelligence Global and Limited-area Ensemble): a 0.25{\deg} global weather model with a 6 km refinement over the Contiguous United States (CONUS). The model achieves significantly lower mean-squared error in near-surface and low-level quantities over CONUS compared to NOAA's Global Forecast System and High-Resolution Rapid Refresh (HRRR), while remaining competitive throughout the rest of the global atmosphere. We show that the skill gains for near-surface fields stem from incorporating high-resolution regional analysis data into training through the nesting process. Forecasts of precipitation amounts are less skillful than those from HRRR, owing to deterministic training. However, we show that Nested-EAGLE provides the most accurate forecasts of storm locations at longer leads, despite blurred extrema. Our results motivate future work to extend the skill gains beyond CONUS and improve precipitation representation.

physics.ao-ph↗

arXiv2026

Long-window 4DVar for reanalysis using a differentiable weather model

Atmospheric reanalyses combine observations with model forecasts using complex data assimilation systems. We test whether a differentiable weather model permits a simpler and more accurate method based on a long-window four-dimensional variational data assimilation (4D-Var) formulation that omits the conventional background-error term. The method uses automatic differentiation to find optimal NeuralGCM initial conditions that minimize the misfit to real surface-pressure observations distributed across overlapping windows of two to seven days, assuming no model error. Cycling at 6-hour intervals for three months beginning 1 January 2015 yields a stable reanalysis with smaller error relative to ERA5 in 500-hPa geopotential height than the Twentieth Century Reanalysis version 3 (20CRv3), which uses an ensemble Kalman filter to assimilate the same observations. Every window produces smaller errors than 20CRv3, with analysis error for the four-day window approximately 55% smaller than for 20CRv3. At the end of the four-day window, which does not benefit from future observations, error remains approximately 38% smaller than 20CRv3. Analyses degrade slightly beyond four days, which we attribute to the increasing importance of model error.

physics.ao-ph↗

arXiv2026

Skillful Global Ocean Emulation and the Role of Correlation-Aware Loss

Machine learning emulators have shown extraordinary skill in forecasting atmospheric states, and their application to global ocean dynamics offers similar promise. Here, we adapt the GraphCast architecture into a dedicated ocean-only emulator, driven by prescribed atmospheric conditions, for medium-range predictions. The emulator is trained on NOAA's UFS-Replay dataset. Using a 24 hour time step, single initial condition, and without using autoregressive training, we produce an emulator that provides skillful forecasts for 10-15 day lead times. We further demonstrate the use of Mahalanobis distance as loss that improves the forecast skill compared to the Mean Squared Error loss by explicitly accounting for the correlations between tendencies of the target variables. Using spatial correlation analysis of the forecasted fields, we also show that the proposed correlation-aware loss acts as a statistical-dynamical regularizer for the slow, correlated dynamics of the global oceans, offering a better background forecast for downstream tasks like data assimilation.

physics.ao-ph↗

arXiv2026

Deep-Learned Observation Operators for Artificial Intelligence Weather Forecasting Models

Satellite observation operators play an essential role in atmospheric data assimilation by translating model state variables into observation space. Previous work has shown that deep-learned emulators can effectively predict the outputs of classic observation operators, like the Community Radiative Transfer Model (CRTM), with reduced inference time. This study expands previous work to show the potential for integrating observation operators into artificial intelligence (AI) weather forecasting models. Specifically, this study shows that (1) deep-learned models can effectively predict the innovations (or differences between the simulated and observed radiances) used by data assimilation models and (2) deep-learned observation models suffer only minor degradations in performance when the model state is represented with fewer vertical levels, as is commonly used by AI forecasting models. Experiments were performed using the Unified Forecast System (UFS) replay dataset, including Gridpoint Statistical Interpolation (GSI) observational data for the Advanced Technology Microwave Sounder (ATMS) sensor from 2022 and 2023. Code is available at https://github.com/mitre/deep-obs.

physics.ao-ph↗

arXiv2026

HealDA: Highlighting the importance of initial errors in end-to-end AI weather forecasts

AI weather models now rival leading numerical weather prediction (NWP) systems in medium-range skill. However, almost all still rely on NWP data assimilation (DA) to provide initial conditions, tying them to expensive infrastructure and limiting the practical speed and accuracy gains of ML. More recently, ML-based DA systems have been proposed, which are often trained and evaluated end-to-end with a forecast model, making it difficult to assess the quality of their analysis fields. We introduce HealDA, a global ML-based DA system that maps a short window of satellite and conventional observations directly to a 1{\deg} atmospheric state on the HEALPix grid, using a smaller sensor suite than operational NWP. We treat HealDA strictly as a DA module: its analyses are used to initialize off-the-shelf ML forecast models without any fine-tuning of either. For a variety of off-the-shelf ML forecast models, including FourCastNet3 (FCN3), Aurora, and FengWu, HealDA-initialized forecasts lose less than one day of effective lead time when scored against ERA5. HealDA-initialized FCN3 ensembles similarly trail those of the ECMWF IFS ENS system by < 24 h. We find that forecast error growth in these models is unchanged from HealDA initialization, and the skill gap primarily arises from the larger initial error of the HealDA analysis. Spectral analysis reveals that this stems from overfitting to the large scales and upper-tropospheric fields. We also demonstrate that small changes in the verification setup can shift apparent skill by 12--24h, underscoring the need for consistent scoring. Taken together, these results clarify the current performance of ML-based DA systems and show that a relatively simple, direct observation-to-state network can already provide initial conditions that are usable by state-of-the-art ML forecast models with only modest loss in medium-range skill.

physics.ao-ph↗

arXiv2025

On the $AdS_3\times S^3\times S^3\times S^1$ dressing factors

We propose dressing factors for massive excitations of the worldsheet S matrix of $AdS_3\times S^3\times S^3\times S^1$ supported by mixed Ramond--Ramond and Neveu-Schwarz--Neveu-Schwarz flux, in the "string" and "mirror" kinematics. Our proposal is compatible with crossing and unitarity, and it reproduces the available perturbative results for any ratio of the two three-spheres' radii.

hep-th↗

arXiv2025

Cryogenic growth of aluminum: structural morphology, optical properties, superconductivity and microwave dielectric loss

We explore the molecular beam epitaxy synthesis of superconducting aluminum thin films grown on c-plane sapphire substrates at cryogenic temperatures of 6 K and compare their behavior with films synthesized at room temperature. We demonstrate that cryogenic growth increases structural disorder, producing crystalline grains that modify the optical, electrical, and superconducting properties of aluminum. We observe that cryogenic deposition changes the color of aluminum from fully reflective to yellow and correlate the pseudo-dielectric function and reflectance with structural changes in the film. We find that smaller grain sizes enhance the superconductivity of aluminum, increasing its critical temperature and critical field. We then estimate the superconducting gap and coherence length of Cooper pairs in aluminum in the presence of disorder. Finally, we fabricate superconducting microwave resonators on these films and find that, independently of the growth temperature, the system is dominated by two-level system loss with similar quality factors in the high and low power regimes. We further measure a higher kinetic inductance in the cryogenically grown films.

cond-mat.supr-con↗

arXiv2025

Dressing Factors and Mirror Thermodynamic Bethe Ansatz for mixed-flux AdS3/CFT2

We complete the derivation of the dressing factors for the $AdS_3\times S^3\times T^4$ S matrix with mixed Ramond--Ramond and Neveu-Schwarz-Neveu-Schwarz flux, in the "string" and "mirror" kinematics. Using these, we propose the mirror Thermodynamic Bethe Ansatz equations which describe the spectrum of the model at any string tension.

hep-th↗

arXiv2025

HRRRCast: a data-driven emulator for regional weather forecasting at convection allowing scales

The High-Resolution Rapid Refresh (HRRR) model is a convection-allowing model used in operational weather forecasting across the contiguous United States (CONUS). To provide a computationally efficient alternative, we introduce HRRRCast, a data-driven emulator built with advanced machine learning techniques. HRRRCast includes two architectures: a ResNet-based model (ResHRRR) and a Graph Neural Network-based model (GraphHRRR). ResHRRR uses convolutional neural networks enhanced with squeeze-and-excitation blocks and Feature-wise Linear Modulation, and supports probabilistic forecasting via the Denoising Diffusion Implicit Model (DDIM). To better handle longer lead times, we train a single model to predict multiple lead times (1h, 3h, and 6h), then use a greedy rollout strategy during inference. When evaluated on composite reflectivity over the full CONUS domain using ensembles of 3 to 10 members, ResHRRR outperforms HRRR forecast at light rainfall threshold (20 dBZ) and achieves competitive performance at moderate thresholds (30 dBZ). Our work advances the StormCast model of Pathak et al. [21] by: a) training on the full CONUS domain, b) using multiple lead times to improve long-range skill, c) training on analysis data instead of the +1h post-analysis data inadvertently used in StormCast, and d) incorporating future GFS states as inputs, enabling downscaling that improves long-lead accuracy. Grid-, neighborhood-, and object-based metrics confirm better storm placement, lower frequency bias, and higher success ratios than HRRR. HRRRCast ensemble forecasts also maintain sharper spatial detail, with power spectra more closely matching HRRR analysis. While GraphHRRR underperforms in its current form, it lays groundwork for future graph-based forecasting. HRRRCast represents a step toward efficient, data-driven regional weather prediction with competitive accuracy and ensemble capability.

physics.ao-ph↗

arXiv2025

Exchange relations and crossing

We discuss the scattering matrix of two-dimensional integrable QFTs whose fields obey non-trivial exchange relations. We show that crossing equations for such models have to be modified, and propose their consistent modification. This modification opens the way to constructing new integrable S~matrices. As a check, we consider the crossing equations for the $SU(N)$ chiral Gross-Neveu model, and for the $\Phi_{21}$ deformation of the tricritical Ising model, finding an agreement with the existing proposals. Finally, we reconsider the crossing equations for massless excitations of the mixed-flux $AdS_3\times S^3\times T^4$ light-cone gauge superstring sigma model, and conjecture that the massless excitations satisfy non-trivial exchange relations. This changes the crossing equations and leads to a simpler massless dressing factor.

hep-th↗

arXiv2025

Massive dressing factors for mixed-flux AdS$_3$/CFT$_2$

We follow up on our proposal for dressing factors for the mixed-flux $AdS_3\times S^3\times T^4$ background presented in arXiv:2402.11732. We discuss in detail the analytic properties of the dressing factors in the string and mirror kinematics for fundamental massive particles and bound states. We prove that the dressing factors are unitary and CP-invariant in the string kinematics, parity invariant in the mirror, and solve the crossing equations in both kinematics. In the limit of pure Ramond-Ramond flux they reduce to the known ones. Finally, we present their expansion at strong tension, as well as in the (small-RR-flux) relativistic limit, finding agreement with the literature.

hep-th↗

arXiv2024

Assimilating Observed Surface Pressure into ML Weather Prediction Models

There has been a recent surge in development of accurate machine learning (ML) weather prediction models, but evaluation of these models has mainly been focused on medium-range forecasts, not their performance in cycling data assimilation (DA) systems. Cycling DA provides a statistically optimal estimate of model initial conditions, given observations and previous model forecasts. Here, real surface pressure observations are assimilated into several popular ML models using an ensemble Kalman filter, where accurate ensemble covariance estimation is essential to constrain unobserved state variables from sparse observations. In this cycling DA system, deterministic ML models accumulate small-scale noise until they diverge. Mitigating this noise with a spectral filter can stabilize the system, but with larger errors than traditional models. Perturbation experiments illustrate that these models do not accurately represent short-term error growth, leading to poor estimation of cross-variable covariances.

physics.ao-ph↗

arXiv2024

Comments on Integrability in the Symmetric Orbifold

We present a map between the excitation of the symmetric-product orbifold CFT of $T^4$, and of the worldsheet-integrability description of $AdS_3\times S^3\times T^4$ of Lloyd, Ohlsson Sax, Sfondrini, and Stefański at $k=1$. We discuss the map in the absence of RR fluxes, when the theory is free, and at small RR flux, $h\ll 1$, where the symmetric-orbifold CFT is deformed by a marginal operator from the twist-two sector. We discuss the recent results of Gaberdiel, Gopakumar, and Nairz, who computed from the perturbed symmetric-product orbifold the central extension to the symmetry algebra of the theory and its coproduct. We show that it coincides with the $h\ll 1$ expansion of the lightcone symmetry algebra known from worldsheet integrability, and that hence the S matrix found by Gaberdiel, Gopakumar, and Nairz maps to the one bootstrapped by the worldsheet integrability approach.

hep-th↗

arXiv2024

Dressing Factors for Mixed-Flux $AdS_3\times S^3\times T^4$ Superstrings

We propose the dressing factors for the scattering of massive particles on the worldsheet of mixed-flux $AdS_3\times S^3\times T^4$ superstrings, in the string and mirror kinematics. The proposal passes all self-consistency checks in the both kinematics, including for bound states. It matches with perturbative and semiclassical computations from the string sigma model, and with its relativistic limit.

hep-th↗

arXiv2024

Ubiquitous non-Majorana Zero-Bias Conductance Peaks in Nanowire Devices

We perform tunneling measurements on indium antimonide nanowire/superconductor hybrid devices fabricated for the studies of Majorana bound states. At finite magnetic field, resonances that strongly resemble Majorana bound states, including zero-bias pinning, become common to the point of ubiquity. Since Majorana bound states are predicted in only a limited parameter range in nanowire devices, we seek an alternative explanation for the observed zero-bias peaks. With the help of a self-consistent Poission-Schrödinger multiband model developed in parallel, we identify several families of trivial subgap states which overlap and interact, giving rise to a crowded spectrum near zero energy and zero-bias conductance peaks in experiments. These findings advance the search for Majorana bound states through improved understanding of broader phenomena found in superconductor-semiconductor systems.

cond-mat.mes-hall↗

arXiv2023

On mixed-flux worldsheet scattering in AdS3/CFT2

Strings on AdS3xS3xT4 with mixed Ramond-Ramond and Neveu-Schwarz-Neveu-Schwarz flux are known to be classically integrable. This is a crucial property of this model, which cannot be studied by conventional worldsheet-CFT techniques. Integrability should carry over to the quantum level, and the worldsheet S matrix in the lightcone gauge is known up to the so-called dressing factors. In this work we study the kinematics of mixed-flux theories and consider a relativistic limit of the S matrix whereby we can complete the bootstrap program, including the dressing factors for fundamental particles and bound states. This provides an important test for the dressing factors of the full worldsheet model, and offers new insights on the features of the model when the amount of NSNS flux is low.

hep-th↗

arXiv2023

Ground state energy of twisted $AdS_{3}\times S^{3}\times T^{4}$ superstring and the TBA

We use the lightcone $AdS_{3}\times S^{3}\times T^{4}$ superstring sigma model with fermions and bosons subject to twisted boundary conditions to find the ground state energy in the semi-classical approximation where effective string tension $h$ and the light-cone momentum $L$ are sent to infinity in such a way that ${\cal J}\equiv L/h$ is kept fixed. We then analyse the ground state energy of the model by means of the mirror TBA equations for the $AdS_{3}\times S^{3}\times T^{4}$ superstring in the pure RR background. The calculation is performed for small twist $μ$ with $L$ and $h$ fixed, for large $L$ with $μ$ and $h$ fixed, and for small $h$ with $μ$ and $L$ fixed. In these limits the contribution of the gapless worldsheet modes coming from the $T^4$ bosons and fermions can be computed exactly, and is shown to be proportional to $hL/(4L^2-1)$. Comparison with the semi-classical result shows that the TBA equations involve only one $Y_0$-function for massless excitations but not two as was conjectured before. Some of the results obtained are generalised to the mixed-flux $AdS_{3}\times S^{3}\times T^{4}$ superstring.

hep-th↗

arXiv2022

We cannot believe we overlooked these Majorana discoveries

In 2011-2012 we performed experiments on hybrid superconductor-semiconductor nanowire devices which yielded signatures of Majorana fermions based on zero-bias peaks in tunneling measurements. The research field that grew out of those findings and other contemporary works has advanced significantly, and a lot of new knowledge and insights were gained. However, key smoking gun evidence of Majorana is still lacking. In this paper, we report that while reviewing our old data recently, armed with a decade of knowledge, we realized that back in 2012 our results contained two breakthrough Majorana discoveries. Specifically, we have observed quantized zero-bias peaks, the hallmark of ideal Majorana states. Furthermore, we have observed the closing and re-opening of the induced gap perfectly correlated with the emergence of the zero-bias peak - clear evidence of the topological quantum phase superconducting transition. These insights should pave the way to topological Majorana qubits, and you should also check supplementary information for important disclosures.

cond-mat.mes-hall↗