Matter42 today is mostly software: spectral parsers, simulation pipelines, project memory, the research agent, the calibration library. The next step is closing the loop with samples we grow ourselves.

We're sketching a small cleanroom that pairs growth tools with the spectroscopic and simulation tooling we already build. Nothing here is final, but the floor plan is far enough along to share.

What goes where

Process tools

The center of the room is a small grid of process tools, sized to cover the deposition and contact-formation steps a 2D-materials process needs end to end. Exact tool selection is still in progress.

The grid layout exists for a reason. It keeps every tool reachable from one of the central aisles, lets the workbenches sit downstream of every tool, and leaves clean sightlines from the entrance for tool techs and visitors.

Workbench / sample prep

Two narrow sample-prep stations sit between the tools and the right wall. This is where FOUPs get loaded and unloaded, samples get transferred to carriers, and most of the day-to-day handling happens. Putting one workbench per tool row keeps short loops short — a sample never has to travel diagonally across the room to get from a tool to a workbench to a hood.

Wet chem and exhaust

The right wall and the top wall are everything that needs scrubbed exhaust:

• A solvent cabinet and precursor fridge sit on the top wall, close to the tools they feed.
• Two scrubbed fume hoods handle wet chem (etches, transfers, cleans).
• A sink rounds out the wet bench.

Every one of these, plus the process tools, has access to scrubbed exhaust (EXSC), house vacuum, and N₂. That constraint shaped the floor plan more than anything else.

Storage

The bottom strip is FOUP racks, shark cages, and general storage. Wafer carriers, gloves, sample boxes, the boring but necessary stuff. We wanted it on the wall furthest from the entrance so foot traffic doesn't move past open carriers.

Entrance

A single entrance on the left wall, between the two tool rows. One way in, one way out, with a clear sightline down both central aisles.

Cleanroom specs

• A small footprint, sized for a research-scale operation rather than a production fab.
• A cleanroom class chosen for the chemistry we run, not the smallest particle count we could pay for.
• Humidity control throughout the room.
• Scrubbed exhaust, house vacuum, and N₂ to every tool, every hood, and the solvent cabinet.

We're being deliberately conservative on cleanroom class. Most TMD growth recipes aren't particle-limited at the monolayer scale. They're limited by precursor purity, chamber chemistry, and contamination from prior runs. A tighter class buys cleanliness we can't use yet, at a cost we'd rather spend on tools.

Closing the loop

The whole point of Matter42's simulation-to-ML pipeline is that DFT and MLIP simulations and Raman maps describe the same physical defect from different angles. Today we get experimental data from academic collaborators, which has driven a lot of the analysis tooling, and we're grateful for it.

Having our own growth tools changes a few things. We can set growth conditions and measure the resulting defects with our own optical stack on the same day, on samples whose history we control end to end. That tightens the feedback between process knobs and measurable signatures in ways that aren't possible when there's a campus or a country between the two halves of the loop. It also lets us check the predictions of our growth simulations against films we actually grew, rather than against published spectra.

This is the research-side counterpart to the production loop we wrote about in AI-native manufacturing for 2D materials. It's hard to build the production version of that loop convincingly without first being able to run the research version of it under one roof.

What's next

Tool selection, vendor decisions, build-out timeline, and the first samples are all in motion. We'll post updates as the room lands.