The technique at the heart of modern overlay control is diffraction-based overlay, or DBO. Instead of imaging the wafer, a DBO tool shines light onto a specially designed grating target and measures how the light diffracts into positive and negative first orders. In the ideal case, the difference between the two diffraction signals is strictly proportional to the overlay shift between layers. This approach has largely displaced older image-based metrology because it offers superior precision and better resilience to tool-induced shifts. But the clean proportionality assumption breaks down in the real world. Processes such as chemical mechanical polishing and plasma etching leave their fingerprints on the grating targets: sidewalls that lean at different angles on the left and right, rounded corners, and imbalanced critical dimensions. These geometric imperfections generate spurious asymmetry signals that can masquerade as genuine overlay shifts, producing what metrologists call process-induced shift errors that quietly corrupt the measurement.

Hsieh and Yang frame this problem in an illuminating way: process-induced asymmetry is fundamentally an optical leakage problem. Structural perturbations of the grating generate diffraction asymmetry that has nothing to do with overlay, and this nuisance signal leaks into the measurement channel that is supposed to carry only overlay information. Their proposed remedy is to stop reducing the diffraction data to a single scalar asymmetry value at one wavelength, and instead capture the full broadband spectral response of the target. By recording the positive and negative first-order diffraction efficiencies for both TE and TM polarizations at 61 wavelengths spanning 400 to 700 nanometers, each measurement becomes a 244-dimensional vector. The physical premise is that wavelength-dependent diffraction acts as a fingerprint of the grating’s state, one rich enough to distinguish true overlay from the spectral distortions introduced by imperfect processing.

To test this premise rigorously, the researchers built a simulation framework based on the finite-difference time-domain method, using a nonuniform mesh with steps as fine as 0.25 nanometers and, in the critical asymmetric target region, 0.10 nanometers. They designed a hierarchy of three evaluation phases of increasing structural complexity. Phase I isolates a single nuisance mechanism: asymmetric sidewall angles, with the left and right sidewall angles of the bottom grating independently randomized between 80 and 90 degrees. Phase II adds top-corner rounding and global film-thickness variation on top of the sidewall asymmetry. Phase III moves to an advanced PolyFin spacer architecture, mimicking the multilayer dielectric environment of modern gate structures, with overlay plus eight independently varied nuisance variables including left and right poly critical dimensions, sidewall angles, spacer corner radii, and two thickness variations. The datasets grew from 932 samples in Phase I to 2,394 in Phase III, each sample generated by a separate full electromagnetic simulation.

On top of these simulated spectra, the authors posed the inverse problem: given only the raw 244-feature spectrum, with no geometry metadata, estimate the overlay. Crucially, they did not bet everything on deep learning. They compared linear regression, Ridge regression, RBF support vector regression, XGBoost, and a four-hidden-layer deep neural network, all trained on identical raw inputs with identical data partitions and validation-only model selection. This design deliberately separates two questions that are often conflated: how much overlay information does the broadband optical response actually contain, and how much does the choice of estimator matter? The results delivered a surprising verdict. In the noise-free Phase I dataset, plain linear regression achieved a root-mean-square error of just 0.0015 nanometers, meaning overlay was almost perfectly linearly decodable from the broadband spectrum. In the most complex Phase III scenario, linear regression and Ridge still reached 0.229 and 0.230 nanometers respectively, beating the DNN’s 0.374 nanometers and far outperforming XGBoost at 1.564 nanometers.

The contrast with conventional narrowband approaches was stark. A fixed single-wavelength reference at 550 nanometers collapsed to errors of roughly 8 to 9 nanometers under compound process imperfections, and even the best single wavelength managed only 5.2 nanometers in the PolyFin case. Multi-wavelength inputs changed the picture entirely, and a Ridge regression on paired-target broadband features reached 0.299 nanometers in Phase III. The authors are careful about what this does and does not prove. The deep network is a competitive nonlinear estimator, but the study explicitly does not establish universal DNN superiority; rather, it shows that broadband spectral observability, rather than model complexity alone, drives much of the performance. No inverse model, however expressive, can recover overlay information that is absent or locally confounded in the measured channels.

The team backed the headline numbers with an unusually thorough battery of robustness analyses. Empirical spectral diagnostics showed that the wavelength carrying the strongest overlay association shifts with the target: 485 nanometers in Phase I, 590 in Phase II, and 690 in Phase III. At individual wavelengths, the overlay response can become nearly parallel to a nuisance response, creating local ambiguity, but across the full 244-feature space the largest global overlay-geometry alignment in Phase III was a moderate cosine of 0.339, and in Phase I the overlay and sidewall-asymmetry responses were nearly orthogonal. Singular-value analysis revealed that the spectra are highly redundant, with only 4, 5, and 14 components explaining 95 percent of the training variance across the three phases, though the very large condition numbers warn that unregularized inversion can amplify small perturbations dramatically.

That warning proved prophetic in the noise stress tests. When trained on noise-free data and then confronted with test spectra perturbed by independent relative noise, the models diverged sharply. Unregularized linear regression, which had appeared almost magical on clean data, exploded to errors of tens of nanometers at just 1 percent noise, while Ridge and the DNN degraded far more gracefully; the Phase III DNN moved only from 0.374 to 0.406 nanometers at 1 percent noise. A blocked holdout test, in which the 20 percent of samples closest to the boundary of the simulated process space were withheld entirely, showed all models degrading by 35 to 68 percent relative to in-domain performance, a reminder that interpolation within a known process window does not automatically transfer to unseen conditions. Bootstrap resampling and repeated data partitions confirmed that the Ridge-versus-DNN difference was statistically robust for the fixed test set, while wavelength ablations showed that five evenly spaced combined-polarization wavelengths already deliver sub-nanometer accuracy, though matching the full-spectrum result to within 10 percent required all 61 labels.

The authors are candid that this is a simulation-based study with defined limits: the two-dimensional model excludes finite mark size, line-edge roughness, wafer-scale nonuniformity, and tool drift, and the noise model is a generic sensitivity test rather than a calibrated instrument specification. The sub-nanometer errors therefore establish feasibility within the evaluated simulations, not validated production-line accuracy. Yet the practical pathway they sketch is concrete. They propose a hybrid digital-twin workflow in which a convergence-verified FDTD library pretrains an inverse estimator, a deliberately selected set of measured calibration data adapts it to the real tool, active learning targets high-uncertainty regions, and continuous monitoring of residuals and spectral distance triggers fallback to reference metrology when drift or out-of-distribution conditions appear. They also point out that target pitch shapes which wavelengths carry overlay information, opening the door to co-designing grating targets and measurement bandwidths.

For an industry where every fraction of a nanometer of edge-placement error translates directly into yield and performance, the message of this work is quietly radical: the information needed to defeat process-induced overlay errors may already be present in the full spectrum of light bouncing off the wafer, waiting to be read properly. The finding that simple, well-regularized linear models can rival deep networks when given rich broadband inputs is a useful corrective to the assumption that harder problems always demand bigger models. As chipmakers push toward gate-all-around transistors and beyond, the combination of full-spectrum diffraction fingerprints, physics-guided simulation, and disciplined statistical validation may prove to be the microscope that keeps the nanoscale world in focus.

Subject of Research: Broadband diffraction-based overlay metrology for semiconductor lithography under process-induced structural asymmetry

Article Title: Broadband spectral metrology for robust diffraction-based overlay estimation under process-induced asymmetry

Article References: Hsieh, H.-C., & Yang, C.-S. (2026). Broadband spectral metrology for robust diffraction-based overlay estimation under process-induced asymmetry. Results in Optics, 25, Article 101156. https://doi.org/10.1016/j.rio.2026.101156

Image Credits: AI Generated

DOI: 10.1016/j.rio.2026.101156

Keywords: overlay metrology, diffraction-based overlay, semiconductor manufacturing, broadband spectroscopy, FDTD simulation, process-induced shift, deep learning, ridge regression, lithography, nanometrology, digital twin, PolyFin spacer

The experiment is, at heart, a modern descendant of Henry Cavendish’s 1798 torsion balance, the apparatus he used to weigh the Earth. A torsion pendulum responds to a horizontal force by twisting a thin suspension fibre, and because the restoring torque of such a fibre can be made extraordinarily small, even minuscule forces produce measurable deflections. Cavendish-type balances remain among the most sensitive instruments for measuring gravity from small source masses, and they have been used to pin down the Newtonian gravitational constant and to test whether gravity behaves as expected at centimetre and sub-centimetre scales. The challenge the Shanghai-led team set out to solve is one of readout: how do you detect the nanometre-scale motion of a pendulum twisted by a gravitational pull of order 10^-16 newtons without adding noise, bulk or complexity?

Their answer is an optical microcavity coupled directly to the pendulum. The sensor uses a whispering-gallery-mode microresonator, a tiny optical resonator in which light circulates around the rim of a microsphere by continuous total internal reflection. Light is coupled into and out of the cavity through a tapered optical fibre whose tip is held roughly 100 nanometres from the microsphere surface. In this evanescent coupling regime, the amount of light that crosses the nanoscale gap depends exquisitely on its exact width. When the torsion pendulum moves, it changes the cavity-fibre separation, which in turn changes the optical transmission. The exponentially decaying evanescent field thus converts nanoscale displacements into optical signals across that roughly 100-nanometre coupling window, acting as an extremely steep mechanical-to-optical transducer.

The real innovation, however, lies in how the team manipulated the light inside the cavity. By exploiting coupled photon-phonon effects, in which the optical field interacts coherently with a mechanical acoustic mode of the resonator, they induced a regime known as Brillouin-scattering-induced transparency, an analogue of electromagnetically induced transparency familiar from atomic physics. Within this narrow spectral window, the optical dispersion becomes extremely steep, and the group velocity of light, the speed at which information-carrying pulses propagate, collapses to just 2,000 metres per second. That is roughly 150,000 times slower than light in vacuum, and it places the device firmly in the ‘stopped light’ regime that the same group has explored previously for light storage near exceptional points.

Why does slowing light help? The steep dispersion that accompanies stopped light means that a tiny change in the cavity resonance frequency, caused by a tiny displacement of the pendulum, produces a disproportionately large change in the phase and transmission of the probe light. In effect, the slow-light resonance amplifies the optical signature of small mechanical motion. This is conceptually similar to how optomechanically induced transparency and other cavity-enhanced readout schemes have boosted force sensitivity in micro- and nanomechanical resonators, but the stopped-light approach pushes the dispersion slope to an extreme, sharpening the transduction without requiring cryogenic temperatures or exotic materials. The whole readout chain operates at room temperature in a table-top apparatus.

The measured performance is impressive. The team reports a displacement sensitivity of 7.85 picometres per root hertz, meaning they can resolve pendulum motions of a few trillionths of a metre in a one-hertz measurement bandwidth. Converted through the mechanical response of the torsion pendulum, that corresponds to an acceleration sensitivity of 3.06 x 10^-16 g, where g is the standard acceleration due to Earth’s gravity. To put that in perspective, the sensor can detect accelerations roughly sixteen orders of magnitude smaller than the pull you feel standing on the ground. By periodically modulating the position of the source mass, a modulation technique that shifts the gravitational signal to a known frequency where technical noise is lower, the researchers resolved a minimum detectable change in gravitational force of 3.02 x 10^-16 newtons.

The source mass in these experiments was millimetre-scale, comparable to the masses used in the landmark 2021 Vienna experiment that first measured gravitational coupling between millimetre-sized gold spheres. That experiment demonstrated that gravity between small objects can be isolated; the new work shows how a photonic readout can be integrated with such a mechanical sensor to push sensitivity further while keeping the instrument compact. Integration has long been the sticking point for Cavendish-type balances: their optical levers and interferometric readouts are delicate, alignment-heavy and difficult to miniaturise. A microcavity readout, with light delivered through a fibre taper and the transduction happening in a micrometre-scale resonator, points toward on-chip gravitational sensing platforms.

The implications extend beyond metrology. Detecting the gravitational fields of microscopic objects is a prerequisite for microscale gravitational imaging, for constraining short-range deviations from Newtonian gravity, and for proposed tests of gravity’s quantum nature. Several theoretical proposals, including the influential 2017 Bose and Marletto-Vedral schemes, suggest that if two masses become entangled purely through their mutual gravitational interaction, gravity must be quantum-mechanical. Realising such experiments demands sensors that can resolve gravitational forces from ever-smaller masses held in ever-more-controlled quantum states. The acceleration sensitivity demonstrated here, combined with proposals for detecting single gravitons with quantum sensing and for gravitationally induced decoherence tests, suggests that stopped-light-enhanced readout could become a key enabling technology on that road.

There are, of course, practical hurdles between a table-top demonstration and a quantum-gravity laboratory. Torsion pendulums are inherently susceptible to thermal noise in the suspension fibre, seismic disturbances and Newtonian background forces from nearby objects, and the evanescent coupling gap of 100 nanometres demands nanometre-precision positioning stability. The researchers also note that the coupled photon-phonon physics that produces stopped light is the same physics that can introduce optomechanical backaction, so careful operating-point selection is needed to ensure that the dispersion enhancement is not eroded by the light’s own mechanical influence. Nonetheless, the demonstration that a room-temperature microcavity can transduce gravitational forces at the 10^-16-newton level marks a genuine advance in nanometrology.

What makes the result especially compelling is its elegance: rather than adding complexity, the team exploited a fundamental optical effect to make the sensor simultaneously simpler and more sensitive. Slow and stopped light has been celebrated for applications ranging from optical buffering to nonlinear enhancement, but using it as the readout engine of a gravitational sensor is a genuinely unexpected twist. If the approach can be scaled to smaller source masses, integrated with levitated or cryogenic mechanical systems, or combined with quantum state preparation of the pendulum itself, the humble Cavendish experiment may yet become a window into the deepest open question in physics: whether the force that holds the cosmos together plays by quantum rules at the smallest scales.

Subject of Research: Stopped-light-enhanced torsion-pendulum sensing of gravitational forces from millimetre-scale masses

Article Title: Stopped-light-enhanced gravitational force sensing

Article References: Zhu, Y., Geng, Q., Xue, B., Sun, Y., Zhang, R., Chen, X., Jiang, X., Azeem, F., Tierz, M., & Wan, W. (2026). Stopped-light-enhanced gravitational force sensing. Nature Nanotechnology. https://doi.org/10.1038/s41565-026-02298-8

Image Credits: AI Generated

DOI: 10.1038/s41565-026-02298-8

Keywords: gravity sensing, torsion pendulum, stopped light, optical microcavity, whispering gallery mode, cavity optomechanics, nanometrology, slow light, force sensitivity, quantum gravity, photon-phonon coupling, Nature Nanotechnology