A research team at China’s Xidian University says it is developing a silicon-germanium (SiGe) approach for short-wave infrared (SWIR) detection that could, in theory, reduce sensor costs to as little as one percent of current prices. In an institutional release, Xidian University describes the work led by Prof. Hu Huiyong as targeting SWIR wavelengths used in areas ranging from fiber-optic communications to sensing applications often associated with defense and autonomy. If the cost projections hold up under real-world manufacturing conditions, the technology could broaden access to advanced infrared imaging, a capability long constrained by the expense of materials such as indium gallium arsenide.
What is verified so far
The core claim originates from Xidian University’s own innovation office. According to that institutional release, Prof. Hu Huiyong’s team has pursued a silicon-germanium route for SWIR detection that replaces conventional indium gallium arsenide (InGaAs) chips. The same report states that the theoretical cost of SiGe-based SWIR detection could fall to 1/100 to 1/10 of what InGaAs sensors cost today. That is the origin of the “99 percent” figure circulating in defense technology discussions. The framing matters: Xidian describes this as a theoretical price floor, not a demonstrated manufacturing outcome.
The technical approach centers on making infrared detection compatible with standard CMOS fabrication, the same mass-production process used for consumer electronics chips. Xidian’s Hangzhou Institute describes the project as targeting SWIR photon detection at 1310 and 1550 nanometers, wavelengths used in fiber-optic communications and military sensing alike. The detector uses an absorption-charge-multiplication separation avalanche structure, and the project spans an end-to-end chain from device modeling through process optimization, readout circuit design, and imaging system integration.
Why does CMOS compatibility matter so much for cost? InGaAs chips require specialized III-V semiconductor fabrication lines that handle small wafer volumes at high prices. Silicon-germanium, by contrast, can ride the economics of existing silicon foundries that already produce billions of chips per year. The gap between a niche III-V fab and a high-volume CMOS line is not incremental; it is the difference between artisan production and industrial scale. That structural cost advantage is what makes the 1/100 claim at least plausible in principle, even if no one has demonstrated it in practice yet.
Independent technical work supports the idea that SiGe-based single-photon avalanche diodes (SPADs) can achieve high performance. A paper hosted on arXiv describes a SPAD array fabricated in a 130nm SiGe BiCMOS process with a 1.8-micrometer pixel pitch and 47-picosecond timing jitter. Those numbers indicate that SiGe processes can produce detectors with fine spatial resolution and fast response times. A separate study reports a germanium-silicon device operating at room temperature with improved noise-equivalent power over earlier germanium-based designs, removing the need for expensive cryogenic cooling that adds weight and cost to fielded systems.
These arXiv papers do not come from Hu Huiyong’s group directly. They represent parallel research efforts that confirm the broader viability of SiGe for photon-counting infrared detection. ArXiv itself operates as an open-access preprint server maintained by Cornell University and supported by a global network of member institutions. The papers have not undergone traditional peer review, though arXiv applies moderation filters, and the technical claims include specific, reproducible measurements that other labs can attempt to replicate.
What remains uncertain
The distance between a theoretical cost model and a working production line is significant. Xidian’s report does not include yield data, wafer-level test results, or pricing from an actual fabrication run. No publicly available source confirms that a SiGe SWIR imaging array has been manufactured at scale, tested in field conditions, or evaluated by military procurement officials. The project description stops at the conceptual optimization stage and does not reference deployment timelines or defense contracts.
Performance comparisons also remain incomplete. InGaAs sensors have decades of field-proven reliability across temperature extremes, humidity, and vibration. SiGe detectors operating at room temperature represent a real advance, but no independent comparative trial has measured SiGe sensitivity against InGaAs under the same operational conditions. Lab metrics like noise-equivalent power, dark count rate, and timing jitter are necessary but not sufficient to predict battlefield performance or long-term reliability in harsh environments.
There is also a question of how the cost claim was constructed. The 1/100 to 1/10 range is wide. At the optimistic end, a sensor that currently costs the equivalent of tens of thousands of yuan could drop to a few hundred. At the conservative end, the savings would still be meaningful but far less dramatic than the headline figure suggests. Without a breakdown of which cost components were modeled, whether packaging, testing, and system integration are included, or what production volume is assumed, the projection is difficult to evaluate independently.
No non-Chinese institution has published a direct replication or independent audit of the Xidian team’s specific device performance. The arXiv papers cited above demonstrate that the underlying physics works, but they address different device architectures and were produced by different research groups. Treating them as validation of Xidian’s specific cost and performance claims would therefore overstate the evidence and blur the line between general feasibility and demonstrated implementation.
How to read the evidence
Three categories of evidence are in play, and they carry different weights. The strongest is the body of technical documentation on SiGe SPADs and related devices, which provides measured parameters from fabricated chips. These papers show that SiGe infrared detection at short-wave wavelengths is physically achievable with existing semiconductor processes and can, in some configurations, operate at room temperature with competitive noise levels.
The second category is the institutional reporting from Xidian University itself. As with many university innovation offices, the language in the release is aspirational and designed to highlight potential impact. It is reasonable to treat the stated device structure and targeted wavelengths as factual, while reading the most aggressive cost claims as projections. Until more detailed process data or prototype test results are released, those projections remain an internal estimate rather than an externally validated benchmark.
The third category is the broader ecosystem context. ArXiv is sustained by community support and institutional backing, which encourages early sharing of results but also means that not every posted paper has passed through peer review. For readers, this creates a hierarchy of confidence: preprints can reliably signal what is technically possible and what researchers are attempting, but they do not, on their own, confirm that a particular national lab or university has solved the engineering problems required for mass production.
When these strands are combined, a cautious picture emerges. The physics of SiGe-based SWIR detection is well supported by independent experiments. The economic argument for CMOS-based fabrication is structurally sound and consistent with decades of experience in the semiconductor industry. What is missing are the bridging details: wafer yields, defect densities, long-term stability data, and fully costed manufacturing runs that would translate theory into market reality.
Potential implications if the claims hold
If Xidian’s projections prove accurate, the implications would extend beyond any single military program. Lower-cost SWIR sensors could make high-resolution night vision, long-range laser detection, and all-weather imaging affordable for a far wider set of actors. Militaries with smaller budgets, paramilitary forces, and even non-state groups could gain access to capabilities that are currently limited by the price and export controls surrounding InGaAs technology.
Civilian sectors would also be affected. Autonomous vehicles, industrial inspection, agricultural monitoring, and medical imaging all stand to benefit from cheaper SWIR cameras. Today, many of these applications are constrained to research settings or premium products because sensor costs dominate the bill of materials. A shift to SiGe could unlock new commercial markets, but it would also complicate export-control regimes that rely on the scarcity and traceability of high-end infrared components.
For now, the prudent stance is to separate what is clearly demonstrated from what is merely promised. SiGe SPADs and related devices have been built, measured, and reported in the open literature. Xidian University has publicly committed to a CMOS-compatible SWIR detector program and articulated ambitious cost goals. Until more detailed performance data and manufacturing results emerge, however, the “99 percent cheaper” narrative should be treated as an informed aspiration rather than a settled fact. The next decisive evidence will not be another preprint or press release, but the appearance of commercially available SiGe SWIR cameras and independent tests of how they perform outside the lab.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
