Analysis

What Aeluma gains by growing photonics on larger wafers

Aeluma’s new reactor can handle 150 mm and 200 mm wafers. The extra area is useful only to the extent that it produces detectors and lasers that still meet their specifications after fabrication and packaging.
A close-up of a silicon wafer with bands of red, gold and green reflected across the patterned surface.

Light catches the patterned surface of a silicon wafer. The photograph illustrates the wafer format, rather than showing an Aeluma device or growth run.

Image: .RGB. / Wikimedia Commons · CC BY 2.0

Aeluma has bought an AIXTRON G10-AsP reactor to grow the semiconductor layers for the high-speed photodetectors and quantum dot lasers it’s developing for data links. The tool accepts wafers of 150 mm and 200 mm, so going from the smaller size to the larger one gives it about 78% more area before any allowance for the wafer edge or rejected dies. That’s a geometric comparison, not a forecast of production: a bigger wafer makes room for more devices, but the value of that space depends on what comes out of the reactor and how well it holds up through the rest of the process.

The wafer itself is part of that calculation. Aeluma plans to grow the active material on gallium arsenide or silicon instead of depending on indium phosphide substrates, which gives it more freedom to choose a material for the optical job and a starting wafer for the fabrication route. It also means that increasing wafer diameter and changing the underlying material are two different decisions. A reactor that accepts both formats gives Aeluma room to work on either, without establishing that a process which works on one will transfer intact to the other.

Where the extra area can disappear

A wafer can meet its average layer thickness and still produce detectors with quite different leakage currents or lasers that need different amounts of current before they begin to lase. Those differences start with what a defect does to the device: it may give current a route through a detector in the dark, or let carriers in a laser recombine without giving off light. In the latter case the active layer has a harder time producing enough gain to overcome losses in the cavity, so a defect that begins during growth can change a measurement made much later on a finished die.

Researchers who demonstrated continuous-wave operation in III–V quantum dot lasers grown on silicon linked their result to reducing threading dislocations in the semiconductor layers. Their process doesn’t establish what Aeluma can achieve, but it shows why two wafers with the same diameter and nominal layer thickness needn’t deliver the same number of working lasers. Even the dies that pass screening still have to meet their optical requirements after they’ve been packaged and coupled to the rest of the system, which is why a bare count of good die can overstate what a customer will get.

Where the work moves to a partner

Aeluma already has a route into partner manufacturing through its September agreement with Sumitomo Chemical Advanced Technologies, which initially uses deposition equipment in Phoenix. It has also named Tower Semiconductor as a manufacturing partner and says it works with established gallium arsenide and silicon foundries. Bringing more growth in-house gives it more control over experiments on the active layers, while the results still have to travel to another line that can handle the wafers and process them within its own tolerances.

Suppose a batch of lasers comes back needing more current to reach threshold. The cause could be in the layers grown in the reactor, but losses introduced later during fabrication can also move the threshold. Looking at how the material behaved before handover alongside the finished devices helps narrow down where that change began. There’s more value in the reactor if it helps Aeluma and its partners separate those effects as they work on larger wafers, especially if the next run changes both the diameter and the substrate.

Aeluma’s quarterly update puts the detectors and lasers in advanced development and mentions discussions with customers, without disclosing qualification dates or yield. Until those figures are available, that 78% is a measure of the extra room in the reactor rather than of the devices a customer can expect to buy.

Sources: Aeluma’s G10-AsP purchase agreement; AIXTRON on the tool and wafer formats; Aeluma’s wafer production agreement with Sumitomo Chemical Advanced Technologies; Aeluma’s quarterly development update; Aeluma’s September update on manufacturing partners; Chen et al., quantum dot lasers grown on silicon.

Continue reading