Search

Kelvin probe force microscopy (KPFM) is a powerful tool for studying contact electrification, using an tiny tip to image voltages caused by transferred charge. It has been used in stationary studies focused on finding patterns (e.g. heterogeneity) in transferred charge, but also in dynamic studies aimed at understanding how charge escapes. Here, we show that the charge dynamics in all but the very best insulators are too fast for patterns found in stationary studies to be meaningful. Using a custom-built system, we are able to quickly (~30 s) transfer samples from our contact-charging apparatus to the atomic force microscopy (AFM). For materials at the lower end of `good insulators', we see potential decay that is shorter than the timescale of a typical KPFM scan (~10 minutes). We develop a minimal model to explain this decay based on bulk conductivity, and show that the measured timescale increases with the prediction from this model. To rule out surface conductivity, we perform additional experiments with the macroscopic method of scanning Kelvin probe microscopy. Our results highlight an important but widely overlooked consideration for dynamic vs. stationary studies in contact-electrified surfaces, and suggest that some patterns observed in the latter, e.g. charge heterogeneity, are not as widespread as previously thought.

Utilising JWST, ALMA and the VLA we present high angular resolution (0.06''- 0.42''), multi-wavelength (4 micron - 3cm) observations of the VLA 1623-2417 protostellar system to characterise the origin, morphology and, properties of the continuum emission. JWST observations at 4.4 micron reveal outflow cavities for VLA 1623 A and, for the first time, VLA 1623 B, as well as scattered light from the upper layers of the VLA 1623 W disk. We model the millimetre-centimetre spectral energy distributions to quantify the relative contributions of dust and ionised gas emission, calculate dust masses, and use spectral index maps to determine where optical depth hinders this analysis. In general, all objects appear to be optically thick down to ~90 GHz, show evidence for significant amounts (10's - 100's M_Earth) of large (>1 mm) dust grains, and are dominated by ionised gas emission for frequencies ~<15 GHz. In addition, we find evidence of unsettled millimetre dust in the inclined disk of VLA 1623 B possibly attributed to instabilities within the circumstellar disk, adding to the growing catalogue of unsettled Class 0/I disks. Our results represent some of the highest resolution observations possible with current instrumentation, particularly in the case of the VLA. However, our interpretation is still limited at low frequencies (~<22 GHz) and thus motivates the need for next-generation interferometers operating at centimetre wavelengths., Comment: Accepted for publication in ApJ. 35 pages, 10 figures

Benchmarks are important tools for tracking the rapid advancements in large language model (LLM) capabilities. However, benchmarks are not keeping pace in difficulty: LLMs now achieve over 90\% accuracy on popular benchmarks like MMLU, limiting informed measurement of state-of-the-art LLM capabilities. In response, we introduce Humanity's Last Exam (HLE), a multi-modal benchmark at the frontier of human knowledge, designed to be the final closed-ended academic benchmark of its kind with broad subject coverage. HLE consists of 2,700 questions across dozens of subjects, including mathematics, humanities, and the natural sciences. HLE is developed globally by subject-matter experts and consists of multiple-choice and short-answer questions suitable for automated grading. Each question has a known solution that is unambiguous and easily verifiable, but cannot be quickly answered via internet retrieval. State-of-the-art LLMs demonstrate low accuracy and calibration on HLE, highlighting a significant gap between current LLM capabilities and the expert human frontier on closed-ended academic questions. To inform research and policymaking upon a clear understanding of model capabilities, we publicly release HLE at https://lastexam.ai., Comment: 27 pages, 6 figures

The propagation of light that undergoes multiple-scattering by resonant atomic vapor can be described as a L\'evy flight. L\'evy flight is a random walk with heavy tailed step-size (r) distribution, decaying asymptotically as $P(r)\sim r^{-1-\alpha}$, with $\alpha<2$. The large steps, typical of L\'evy flights, have its origins in frequency redistribution of the light scattered by the vapor. We calculate the frequency redistribution function and the step-size distribution for light diffusion in atomic vapor. From the step-size distribution we extract a L\'evy parameter $\alpha$ that depends on the step's size. We investigate how the frequency redistribution function and step-size distribution are influenced by the finite size of the vapor and the many-level structure typical for alkali vapors. Finite size of the vapor introduces cutoff on the light scattered spectrum and thus in the size of steps. Multi-level structure introduces oscillations in $P(r)$ slope. Both effects might have an impact on measurables related to the L\'evy flight random walk., Comment: 37 pages, 12 figures

Gravitational Wave (GW) detectors routinely encounter transient noise bursts, known as glitches, which are caused by either instrumental or environmental factors. Due to their high occurrence rate, glitches can overlap with GW signals, as in the notable case of GW170817, the first detection of a binary neutron star merger. Accurate reconstruction and subtraction of these glitches is a challenging problem that must be addressed to ensure that scientific conclusions drawn from the data are reliable. This problem will intensify with third-generation observatories like the Einstein Telescope (ET) due to their higher detection rates of GWs and the longer duration of signals within the sensitivity band of the detectors. Robust glitch mitigation algorithms are, therefore, crucial for maximizing the scientific output of next-generation GW observatories. For the first time, we demonstrate how the null stream inherent in ET's unique triangular configuration can be leveraged by state-of-the-art glitch characterization methodology to essentially undo the effect of glitches for the purpose of estimating the parameters of the source. The null stream based approach enables characterization and subtraction of glitches that occur arbitrarily close to the peak of the signal without any significant effect on the quality of parameter measurements, and achieves an order of magnitude computational speed-up compared to when the null stream is not available. By contrast, without the null stream, significant biases can occur in the glitch reconstruction, which deteriorate the quality of subsequent measurements of the source parameters. This demonstrates a clear edge which the null stream can offer for precision GW science in the ET era.

Building upon the statistical formulation for parameter estimation in the presence of correlated noise proposed by Cireddu et al., we present an initial study to incorporate the effects of correlated noise into the analyses of various detector designs' performance. We consider a two L-shaped detector configuration located in the European Union, and compare the expectation of parameter estimation between the non-colocated and a hypothetical colocated configurations. In our study, we posit the existence of low-frequency correlated noise within the $5\text{ Hz}$ to $10\text{ Hz}$ range for the colocated detector configuration, with a varying degree of correlation. In this specific detector setup, our observations indicate an enhancement in the precision of intrinsic parameter measurements as the degree of correlation increases. This trend suggests that higher degrees of noise correlation may beneficially influence the accuracy of parameter estimation. In particular, when the noise is highly correlated, the uncertainty on chirp mass decreases by up to $30\%$. The absence of an inter-European baseline does hinder the estimation of the extrinsic parameters. However, given a realistic global network with the additional detector located in the United States, the uncertainty of extrinsic parameters is significantly reduced. This reduction is further amplified as the degree of noise correlation increases. When noise correlation exceeds a certain level, the colocated configuration outperforms the non-colocated one, reducing the $90\%$ credible area of sky location by up to $10\%$. We conclude that noise correlation significantly impacts detector performance, potentially altering both quantitative and qualitative outcomes. Thus, we recommend including noise correlation in comprehensive assessments of third-generation gravitational wave detector designs., Comment: 10 pages, 2 figures

Scanning Kelvin probe microscopy (SKPM) is a powerful technique for investigating the electrostatic properties of material surfaces, enabling the imaging of variations in work function, topology, surface charge density, or combinations thereof. Regardless of the underlying signal source, SKPM results in a voltage image which is spatially distorted due to the finite size of the probe, long-range electrostatic interactions, mechanical and electrical noise, and the finite response time of the electronics. In order to recover the underlying signal, it is necessary to deconvolve the measurement with an appropriate point spread function (PSF) that accounts the aforementioned distortions, but determining this PSF is difficult. Here we describe how such PSFs can be determined experimentally, and show how they can be used to recover the underlying information of interest. We first consider the physical principles that enable SKPM, and discuss how these affect the system PSF. We then show how one can experimentally measure PSFs by looking at well defined features, and that these compare well to simulated PSFs, provided scans are performed extremely slowly and carefully. Next, we work at realistic scan speeds, and show that the idealised PSFs fail to capture temporal distortions in the scan direction. While simulating PSFs for these situations would be quite challenging, we show that measuring PSFs with similar scan conditions works well. Our approach clarifies the basic principles of and inherent challenges to SKPM measurements, and gives practical methods to improve results, Comment: Updated with doi after publication

In this paper, we present a new computer-controlled weaving technology that enables the fabrication of woven structures in the shape of given 3D surfaces by using threads in non-traditional materials with high bending-stiffness, allowing for multiple applications with the resultant woven fabrics. A new weaving machine and a new manufacturing process are developed to realize the function of 3D surface weaving by the principle of short-row shaping. A computational solution is investigated to convert input 3D freeform surfaces into the corresponding weaving operations (indicated as W-code) to guide the operation of this system. A variety of examples using cotton threads, conductive threads and optical fibres are fabricated by our prototype system to demonstrate its functionality.

The Einstein Telescope faces a critical data analysis challenge with correlated noise, often overlooked in current parameter estimation analyses. We address this issue by presenting the statistical formulation of the likelihood that includes correlated noise for the Einstein Telescope or any detector network. By considering varying degrees of correlation, we probe the impact of noise correlations on the parameter estimation analysis of a GW150914-like event. We show that neglecting these correlations may significantly reduce the accuracy of the chirp mass reconstruction. This emphasizes how critical a proper treatment of correlated noise is, as presented in this work, to unlocking the wealth of results promised by the Einstein Telescope., Comment: Accepted for publication in Physical Review D; 11 pages, 3 figures

Be stars make up almost 20% of the B star population, and are rapidly rotating stars surrounded by a disc; however the origin of this rotation remains unclear. Mass transfer within close binaries provides the leading hypothesis, with previous detections of stripped companions to Be stars supporting this. Here, we exploit the exquisite astrometric precision of Gaia to carry out the largest to date comparative study into the binarity of matched samples of nearby B and Be stars from the Bright Star Catalogue. By utilising new "proper motion anomaly" values, derived from Gaia DR2 and DR3 astrometric data alongside previous values calculated using Hipparcos and Gaia data, and the Gaia provided RUWE, we demonstrate that we can identify unresolved binaries down to separations of 0.02". Using these measures, we find that the binary fractions of B and Be stars are similar between 0.04 - 10", but the Be binary fraction is significantly lower than that of the B stars for separations below 0.04". As the separation range of these "missing" binaries is too large for mass transfer, and stripped companions are not retrieved by these measures, we suggest the companions migrate inwards via binary hardening within a triple system. This confirms statistically for the first time the hypothesis that binary interaction causes the Be phenomenon, with migration causing the dearth of Be binaries between 0.02 - 0.04". Furthermore, we suggest that triplicity plays a vital role in this migration, and thus in the formation of Be stars as a whole., Comment: 11 pages, 4 figures; typos corrected

The current study examined predictors of outcomes across two school clinician-delivered treatments (i.e., treatment as usual [TAU] or modular cognitive-behavioral therapy [M-CBT]) for youth with anxiety disorders. Predictors reflected two broad domains -- therapist factors (i.e., education, years of experience, therapeutic orientation, work related stressors/barriers, self-efficacy, and attitudes towards evidence-based practices) and treatment-related factors (i.e., dosage, child compliance with treatment, therapeutic alliance, therapeutic nonspecifics, and proportion and quality of evidence-based structure elements). One hundred and ninety-five youth (mean age 10.98; 50.3% female, 53.5% non-Hispanic Caucasian) and 54 therapists (90.7% female, 72.2% non-Hispanic Caucasian) that were enrolled and randomized in a previous study (Ginsburg et al., 2020) participated. Results showed that incorporation of more evidence-based structure elements (e.g., agenda setting, assigning homework) and higher child compliance with treatment significantly increased odds of response to treatment. These findings provide important information about key ingredients to successful treatments, regardless of treatment condition/modality, and can be used to inform future training and implementation of school-based treatments for youth with anxiety disorders. [This is the in press version of an article published in "Journal of Behavioral and Cognitive Therapy" (ISSN 2589-9791).]

Through digital imaging, microscopy has evolved from primarily being a means for visual observation of life at the micro- and nano-scale, to a quantitative tool with ever-increasing resolution and throughput. Artificial intelligence, deep neural networks, and machine learning are all niche terms describing computational methods that have gained a pivotal role in microscopy-based research over the past decade. This Roadmap is written collectively by prominent researchers and encompasses selected aspects of how machine learning is applied to microscopy image data, with the aim of gaining scientific knowledge by improved image quality, automated detection, segmentation, classification and tracking of objects, and efficient merging of information from multiple imaging modalities. We aim to give the reader an overview of the key developments and an understanding of possibilities and limitations of machine learning for microscopy. It will be of interest to a wide cross-disciplinary audience in the physical sciences and life sciences.

Since the first detection of gravitational waves in 2015, gravitational-wave astronomy has emerged as a rapidly advancing field that holds great potential for studying the cosmos, from probing the properties of black holes to testing the limits of our current understanding of gravity. One important aspect of gravitational-wave astronomy is the phenomenon of gravitational lensing, where massive intervening objects can bend and magnify gravitational waves, providing a unique way to probe the distribution of matter in the universe, as well as finding applications to fundamental physics, astrophysics, and cosmology. However, current models for gravitational-wave millilensing - a specific form of lensing where small-scale astrophysical objects can split a gravitational wave signal into multiple copies - are often limited to simple isolated lenses, which is not realistic for complex lensing scenarios. In this paper, we present a novel phenomenological approach to incorporate millilensing in data analysis in a model-independent fashion. Our approach enables the recovery of arbitrary lens configurations without the need for extensive computational lens modeling, making it a more accurate and computationally efficient tool for studying the distribution of matter in the universe using gravitational-wave signals. When gravitational-wave lensing observations become possible, our method can provide a powerful tool for studying complex lens configurations, including dark matter subhalos and MACHOs., Comment: 12 pages, 8 figures

The ALPHA Collaboration, based at the CERN Antiproton Decelerator, has recently implemented a novel beamline for low-energy ($\lesssim$ 100 eV) positron and antiproton transport between cylindrical Penning traps that have strong axial magnetic fields. Here, we describe how a combination of semianalytical and numerical calculations were used to optimise the layout and design of this beamline. Using experimental measurements taken during the initial commissioning of the instrument, we evaluate its performance and validate the models used for its development. By combining data from a range of sources, we show that the beamline has a high transfer efficiency, and estimate that the percentage of particles captured in the experiments from each bunch is (78 $\pm$ 3)% for up to $10^{5}$ antiprotons, and (71 $\pm$ 5)% for bunches of up to $10^{7}$ positrons., Comment: 15 pages, 15 figures

This paper presents a novel Active Blacklist (ABL) based on a modification of the Simple Mail Transfer Protocol (SMTP). ABL was implemented in the Mail Transfer Agent (MTA) Postfix of the e-mail server Zimbra and assessed exhaustively in a series of experiments. The modified server Zimbra showed computational performance and costs similar to those of the original server Zimbra when receiving legitimate e-mails. When receiving spam, however, it showed better computing performance and costs than the original Zimbra. Moreover, there was a considerable computational cost on the spammer's server when it sent spam e-mails. ABL was assessed at the Federal University of Itajub\'{a}, Brazil, during a period of sixty-one days. It was responsible for rejecting a percentage of 20.94% of the spam e-mails received by the university during this period. After this period, it was deployed and remained in use, from July-2015 to July-2019, at the university. ABL is part of the new Open Machine-Learning-Based Anti-Spam (Open-MaLBAS). Both ABL and Open-MaLBAS are freely available on GitHub., Comment: 18 pages, 6 figures, 5 tables

We present the first systematic search for exotic compact mergers in Advanced LIGO and Virgo events. We compare the short gravitational-wave signals GW190521, GW190426$\_$190642, GW200220$\_$061928 and the trigger 200114$\_$020818 (or S200114f) to a new catalogue of 759 numerical simulations of head-on mergers of horizonless exotic compact objects known as Proca stars, interpreted as self-gravitating lumps of (fuzzy) dark matter sourced by an ultralight (vector) bosonic particle. The Proca-star merger hypothesis is strongly rejected with respect to the black hole merger one by GW190426, weakly rejected by GW200220 and weakly favoured by GW190521 and S200114f. GW190521 and GW200220 yield highly consistent boson masses of $\mu_{\rm B} = 8.69^{+0.61}_{-0.75}\times10^{-13}$ eV and $\mu_{\rm B} = 9.13^{+1.18}_{-1.30}\times10^{-13}$ eV at the $90\%$ credible level. We conduct a preliminary population study of the compact binaries behind these events. Excluding (including) S200114f as a real event, and ignoring boson-mass consistencies across events, we estimate a fraction of Proca-star mergers of $\zeta = 0.27^{+0.43}_{-0.25} \ (0.39^{+0.38}_{-0.33})$. We discuss the impact of boson-mass consistency across events in such estimates. Our results maintain GW190521 as a Proca-star merger candidate and pave the way towards population studies considering exotic compact objects., Comment: Version accepted in Phys.Rev.D. 27 pages, 16 Figures, 4 Appendixes

Detection and parameter inference of gravitational-wave signals \ncor{from compact mergers} rely on the comparison of the incoming detector strain data $d(t)$ to waveform templates for the gravitational-wave strain $h(t)$ that ultimately rely on the resolution of Einstein's equations via numerical relativity simulations. These, however, commonly output a quantity known as the Newman-Penrose scalar $\psi_4(t)$ which, under the Bondi gauge, is related to the gravitational-wave strain by $\psi_4(t)=\mathrm{d}^2h(t) / \mathrm{d}t^2$. Therefore, obtaining strain templates involves an integration process that introduces artefacts that need to be treated in a rather manual way. By taking second-order finite differences on the detector data and inferring the corresponding background noise distribution, we develop a framework to perform gravitational-wave data analysis directly using $\psi_4(t)$ templates. We first demonstrate this formalism, and the impact of integration artefacts in strain templates, through the recovery of numerically simulated signals from head-on collisions of Proca stars injected in Advanced LIGO noise. Next, we re-analyse the event GW190521 under the hypothesis of a Proca-star merger, obtaining results equivalent to those in Ref.[1], where we used the classical strain framework. We find, however, that integration errors would strongly impact our analysis if GW190521 was four times louder. Finally, we show that our framework fixes significant biases in the interpretation of the high-mass GW trigger S200114f arising from the usage of strain templates. We remove the need to obtain strain waveforms from numerical relativity simulations, avoiding the associated systematic errors., Comment: Version accepted for publication in Physical Review X. 27 pages, 11 Figures, 3 Appendixes

When a gravitational wave passes by a massive object, its trajectory gets curved which can result in multiple images separated by a time delay, a phenomenon known as strong lensing. Although gravitational wave lensing has not been observed yet, it is currently predicted that it will be possible to detect with third-generation detectors. Moreover, the future detectors are expected to significantly increase the gravitational-wave detection rate up to $O(10^5 - 10^6)$ $\mathrm{yr}^{-1}$ which could also increase the chance of observing gravitational wave lensing. Therefore, it is necessary to develop accurate and efficient tools to filter potentially lensed gravitational-wave signal pairs. In the study, we propose three statistics to analyze the overlap of two sky localization probability maps, which could provide a quick preliminary lensing analysis of two gravitational-wave signals within 3 seconds. We simulate 200 lensed pairs of gravitational-wave signals for 5 signal-to-noise ratios to investigate the performance of the search methodology. By setting up a threshold with a specific false positive rate $\mathrm{FPR} = 10^{-2}$ for the three overlap statistics, we conclude that it is possible to filter out more than $99\%$ of the non-lensed event pairs while keeping all lensed event pairs., Comment: 8 pages, 3 figures, 2 tables

The proposed third-generation gravitational-wave detectors Einstein Telescope will have a triangular design that consists of three colocated interferometers. Summing the strain outputs from the three interferometers will cancel any gravitational-wave signal and the resultant signal-free stream is known as null stream. The null stream is in a fixed subspace of the observation space of Einstein telescope where no gravitational-wave signal can exist. In this paper, we establish the decomposition of the observation space of Einstein Telescope into the signal space that contains all possible gravitational-wave signals and the null space that contains the null stream. We show that the results of Bayesian parameter estimation and model selection using the strain data in the signal space are identical to that using the full set of strain data. This implies that one could use a fraction of the strain data to extract all information of the source which reduces the memory usage and speeds up the likelihood evaluation. We also discuss the existence of a fixed null space in Einstein Telescope allows the unbiased estimation of the noise properties in the signal-free subspace., Comment: 9 pages

The ALPHA collaboration has successfully demonstrated the production and the confinement of cold antihydrogen, $\overline{\mathrm{H}}$. An analysis of trapping data allowed a stringent limit to be placed on the electric charge of the simplest antiatom. Charge neutrality of matter is known to a very high precision, hence a neutrality limit of $\overline{\mathrm{H}}$ provides a test of CPT invariance. The experimental technique is based on the measurement of the deflection of putatively charged $\overline{\mathrm{H}}$ in an electric field. The tendency for trapped $\overline{\mathrm{H}}$ atoms to be displaced by electrostatic fields is measured and compared to the results of a detailed simulation of $\overline{\mathrm{H}}$ dynamics in the trap. An extensive survey of the systematic errors is performed, with particular attention to those due to the silicon vertex detector, which is the device used to determine the $\overline{\mathrm{H}}$ annihilation position. The limit obtained on the charge of the $\overline{\mathrm{H}}$ atom is \mbox{$ Q = (-1.3\pm1.8\pm0.4)\times10^{-8}$}, representing the first precision measurement with $\overline{\mathrm{H}}$., Comment: 5 pages, 3 figures

We present a null-stream-based Bayesian unmodeled framework to probe generic gravitational-wave polarizations. Generic metric theories allow six gravitational-wave polarization states, but general relativity only permits the existence of two of them namely the tensorial polarizations. The strain signal measured by an interferometer is a linear combination of the polarization modes and such a linear combination depends on the geometry of the detector and the source location. The detector network of Advanced LIGO and Advanced Virgo allows us to measure different linear combinations of the polarization modes and therefore we can constrain the polarization content by analyzing how the polarization modes are linearly combined. We propose the basis formulation to construct a null stream along the polarization basis modes without requiring modeling the basis explicitly. We conduct a mock data study and we show that the framework is capable of probing pure and mixed polarizations in the Advanced LIGO-Advanced Virgo 3-detector network without knowing the sky location of the source from electromagnetic counterparts. We also discuss the effect of the presence of the uncaptured orthogonal polarization component in the framework, and we propose using the plug-in method to test the existence of the orthogonal polarizations., Comment: 26 pages, 20 figures

The organic terahertz (THz) generation crystal BNA has recently gained traction as a valuable source to produce broadband THz pulses. Even when pumped with 800-nm light, thin BNA crystals can produce relatively high electric fields with frequency components out to 5 THz. However, the THz output when pumped with 800-nm light is limited by the damage threshold of the organic crystal. Here we report that the damage threshold of BNA can be significantly improved by physically bonding BNA to a high-thermal conductivity sapphire window. When pumped with 800-nm light from an amplified Ti:sapphire laser system, our bonded BNA (BNA-sapphire) generates 2.5 times higher electric field strengths compared to bare BNA crystals. We characterize the average damage threshold for bare BNA and BNA-sapphire, measure peak-to-peak electric field strengths and THz waveforms, and determine the nonlinear transmission in BNA. Pumping BNA-sapphire with 800-nm light results in peak-to-peak electric fields exceeding 1 MV/cm, with strong broadband frequency components from 0.5-5 THz. Our BNA-sapphire THz source is a promising alternative to tilted pulse front LiNbO3 THz sources, which will enable many research groups without optical parametric amplifiers to perform high-field, broadband THz spectroscopy., Comment: 9 pages, 5 figures

We characterize the terahertz (THz) generation of N-benzyl-2-methyl-4-nitroaniline (BNA), with crystals ranging in thickness from 123-700 {\mu}m. We compare excitation using 800-nm and 1250 to 1500-nm wavelengths. Pumping BNA with 800-nm wavelengths and longer near-infrared wavelengths results in a broad spectrum, producing out to 6 THz using a 100-fs pump, provided the BNA crystal is thin enough. ~200 {\mu}m or thinner crystals are required to produce a broad spectrum with an 800-nm pump, whereas ~300 {\mu}m thick crystals are optimal for broadband THz generation using the longer wavelengths. We report the favorable THz generation and optical characteristics of our BNA crystals that make them attractive for broadband, high-field THz generation, and we also find significant differences to BNA results reported in other works., Comment: 7 pages, 5 figures

Since their invention in the 1980s [1], optical tweezers have found a wide range of applications, from biophotonics and mechanobiology to microscopy and optomechanics [2, 3, 4, 5]. Simulations of the motion of microscopic particles held by optical tweezers are often required to explore complex phenomena and to interpret experimental data [6, 7, 8, 9]. For the sake of computational efficiency, these simulations usually model the optical tweezers as an harmonic potential [10, 11]. However, more physically-accurate optical-scattering models [12, 13, 14, 15] are required to accurately model more onerous systems; this is especially true for optical traps generated with complex fields [16, 17, 18, 19]. Although accurate, these models tend to be prohibitively slow for problems with more than one or two degrees of freedom (DoF) [20], which has limited their broad adoption. Here, we demonstrate that machine learning permits one to combine the speed of the harmonic model with the accuracy of optical-scattering models. Specifically, we show that a neural network can be trained to rapidly and accurately predict the optical forces acting on a microscopic particle. We demonstrate the utility of this approach on two phenomena that are prohibitively slow to accurately simulate otherwise: the escape dynamics of swelling microparticles in an optical trap, and the rotation rates of particles in a superposition of beams with opposite orbital angular momenta. Thanks to its high speed and accuracy, this method can greatly enhance the range of phenomena that can be efficiently simulated and studied., Comment: 4 figures