WST, the Wide-field Spectroscopic Telescope: dispersing elements

The Wide-field Spectroscopic Telescope (WST) is a proposed 12-m class facility entirely dedicated to spectroscopic surveys, combining a high-multiplex multi-object spectrograph operating at low (MOS-LR) and high (MOS-HR) spectral resolution with a giant panoramic integral-field spectrograph (IFS), all three operating in parallel. Diffraction gratings are the key dispersing elements of all three instruments and, given the very large number and size of the units required, drive critical trade-offs in throughput, feasibility and production cost. This paper reviews the two grating technologies under consideration for WST, Volume Phase Holographic Gratings (VPHG) and binary (lithographic, surface-relief) gratings, summarizing their working principles and the parameters that control their diffraction efficiency. We then present the current baseline grating parameters and vendor results for each instrument: low-dispersion, VPHGs for the IFS; a four-arm GRISM layout for MOS-LR; and two competing high-resolution disperser architectures (‘8M16D’ and ‘16M4D’) for MOS-HR, where binary gratings show a promising path to diffraction efficiencies beyond what is achievable with VPHGs. We conclude with the main open challenges, chiefly the mass production of hundreds of grating units within cost and schedule and the next steps foreseen to consolidate the disperser baseline for WST.

Improvements to AI Systems:

1. AI-Driven Grating Efficiency Optimization: Train a generative AI model on the physical parameters (groove density, blaze angle, modulation depth, refractive index contrast) and measured efficiency curves from the paper’s vendor results. The AI can then propose novel grating designs that maximize diffraction efficiency across target wavelength bands, automatically balancing trade-offs between VPHG and binary technologies.

2. Automated Cost-Schedule Risk Predictor: Build a machine learning system that ingests grating specifications (size, number of units, technology type) and historical production data from the paper’s “mass production” challenge. The AI predicts cost overruns and schedule delays, then recommends optimal batch sizes, material substitutions, or hybrid VPHG-binary layouts to meet WST’s budget.

3. Multi-Instrument Disperser Selection Agent: Develop a reinforcement learning agent that simulates the full WST instrument suite (IFS, MOS-LR, MOS-HR). The agent evaluates thousands of grating configurations (e.g., ‘8M16D’ vs ‘16M4D’) against scientific throughput, resolution, and parallel observation requirements, then outputs the Pareto-optimal disperser architecture for each arm.

4. Physics-Informed Surrogate Model for Diffraction Simulation: Replace slow electromagnetic solvers (e.g., RCWA) with a neural network trained on the paper’s efficiency data and first-principles physics. This AI can instantly predict efficiency for any grating geometry, enabling rapid iteration on design trade-offs and real-time optimization during fabrication.

5. Automated Quality Control via Anomaly Detection: Use a convolutional neural network on interferometric images of fabricated gratings to detect microscopic defects (e.g., groove irregularities, coating thickness variations) that degrade efficiency. The AI flags units that fail WST’s throughput thresholds, reducing waste and ensuring consistent performance across hundreds of mass-produced gratings.

What the Improved AI System Can Do:

• Design and validate optimal gratings in hours instead of months, directly reducing WST’s cost and schedule risk.

• Predict and mitigate production bottlenecks before they occur, ensuring the 100+ grating units are delivered on time.

• Automatically select the best disperser for each instrument based on real-time scientific priorities and manufacturing constraints.

• Provide a digital twin of the entire WST spectrograph system, allowing engineers to test “what-if” scenarios (e.g., switching from VPHG to binary for IFS) without physical prototypes.

• Ensure consistent, high-efficiency performance across all units, maximizing the telescope’s survey speed and spectral fidelity.