Latest
What Atlas can do with WS₂
A 130 × 130 µm co-registered Raman and PL map of a PFIB-patterned WS₂ flake: segmentation, two independent defect-density estimators, strain–doping decomposition, exciton/trion fitting, edge profiles, and cross-modal correlation — with every calibration caveat Atlas raised.
Read the noteMore posts
What Atlas can do with MoS₂
ALD versus sputtered MoS₂ films across four process conditions: what Raman linewidth, peak separation, strain–doping spread, and defect-density estimators say about disordered large-area films, and how they track FET on/off ratio.
What Atlas can do with GaN and InGaN
Cathodoluminescence maps of InGaN/GaN quantum wells across a five-sample point-defect series: how spatial intensity uniformity from one Block orders the samples the same way independent lifetime measurements do, and what dedicated GaN Blocks add.
What Atlas can do with GaAs heterostructures
A 111,556-pixel micro-PL map of GaAs/GaAsP quantum-well-tube nanowires: confined-state energy, per-wire linewidth and energy spread, emitting fraction across flakes, and a deep-level ratio Atlas reports as uninformative for the right reason.
Where does physics enter the AI system?
Six levels of integrating physics into AI systems, from a constraint layer that forbids impossible answers to an autonomous scientist that runs its own lab - the level where Physics AI hands off to Physical AI. Lets dive in with our friends Rick and Marvin
AFM beyond roughness: what Atlas can do with a surface map
One 1 µm scan of CVD MoS₂ triangles, pushed as far as it goes: artifact-matched leveling, roughness moments, counting 40 flakes at a 0.56 nm monolayer step, line profiles, and the QNM channels hiding in the same file.
Beyond peak picking: what Atlas can do with XRD
A practical tour from broad-peak GIXRD and multi-scan BRML files to phase evidence, strain, texture, rocking curves, and rotational twins.
Stress-testing a Nature Communications result with Atlas
We gave Atlas nearly 3 GB of raw Raman and PL maps behind a published MoS₂ nanoribbon result. A naive PL clustering produced a convincing wrong answer; a geometry-aware edge analysis recovered the physical trend and added independent depletion, strain, and defect checks.
What MOCVD actually buys you
How metered precursor delivery turns TMD growth into a process that can be logged, compared, and tuned across runs.
How Raman spectroscopy works
An interactive walkthrough: photon scattering, why the shift is a chemical fingerprint, and what peak positions, widths, and the anti-Stokes ratio tell you.
Why WS₂ crystals grow as triangles
An interactive kinetic Wulff construction: how alternating S- and W-zigzag edges turn a hexagonal WS₂ nucleus into a triangle under CVD.
How a Raman linewidth becomes a defect density
Inside the simulation-to-calibration pipeline: MLIP phonons, defect supercells, and an inversion that states its own limits.
Beyond Moore: why characterization has to move faster
Advanced semiconductor programs need faster ways to connect materials evidence, simulation context, and process learning.
Designing repeatable characterization workflows
How we think about turning mixed spectroscopy, microscopy, notes, and simulation outputs into a reproducible project workflow.
AI-native characterization for 2D materials
How calibrated tools, surrogate models, and project memory can make spectroscopy-driven materials decisions easier to review.
Toward a research agent for 2D materials
Why a calibrated research agent — not a generic chatbot — is the right interface to spectroscopy, simulation, documents, and project memory for 2D-material teams.