There is a well-known gap in technology development called the valley of death: the space between a laboratory demonstration and a commercial product. In excitonic photonics, the valley is particularly wide. A device that works once in a laboratory, under careful conditions, with a hand-selected sample, is far from a product that works thousands of times, in varied environments, at a competitive price.
This article explains what happens after a device is demonstrated: yield, packaging, reliability, qualification, cost, and the business decisions that determine whether a technology crosses the valley or remains a research result.
Why the laboratory is not the factory
A laboratory device is optimized for a single outcome: demonstrating a physics effect or a device concept. The researcher can select the best flake, tune the measurement conditions, and report the best result. The device may require cryogenic cooling, optical pumping, or manual alignment. It may operate for minutes, not years. It may be one of a handful made, with no data on how many failed.
A product must meet specifications across a population of devices, under specified operating conditions, for a specified lifetime, at a specified cost. It must be manufactured with a process that is repeatable, controllable, and documented. It must be tested, packaged, and qualified. It must have a supply chain for materials and components. And it must have a market that will pay for it.
The transition from laboratory to factory therefore requires a shift in thinking: from best-device performance to statistical performance, from demonstration to manufacturing, and from physics to engineering and business. Many promising technologies fail in this transition not because the physics is wrong but because the engineering, manufacturing, or economics do not close.
Yield: the statistic that matters most
Yield is the fraction of devices on a wafer or in a batch that meet specifications. It is the single most important manufacturing metric, because it determines cost. A process with 10 percent yield must discard 90 percent of its output, which makes each good device ten times more expensive than the process cost would suggest. A process with 90 percent yield is far more economical, even if its per-wafer cost is higher.
For excitonic devices, yield is challenged at every step. Growth may produce nonuniform films. Transfer may introduce defects. Lithography may damage atomically thin layers. Contacts may be inconsistent. Encapsulation may fail. Each step has its own yield, and the overall yield is the product of all step yields. A process with ten steps each at 90 percent yield has an overall yield of about 35 percent. Improving yield requires improving every step, not just the most visible one.
Yield also depends on specifications. A device with relaxed specifications — for example, a sensor that tolerates a wide range of responsivity — will have higher yield than a device with tight specifications, such as a laser that requires a narrow wavelength tolerance. Tightening specifications to meet a product requirement can dramatically reduce yield, which is why product specifications and manufacturing capability must be co-designed.
Reporting yield is essential for evaluating a technology’s maturity. A paper that reports results on a single device, or on selected devices, without reporting how many devices were made and how many met specifications, is not providing yield information. A manufacturing report that shows distributions of key metrics across a wafer, and yield as a function of specifications, is providing the information needed to assess commercial prospects.
Packaging: protecting the device from the world
Packaging is often treated as an afterthought in device research, but it is a major determinant of product success. A photonic device must be protected from moisture, oxygen, mechanical stress, and thermal cycling, while allowing light to enter and exit, electrical connections to be made, and heat to be dissipated. For excitonic devices based on air-sensitive materials such as perovskites or monolayer TMDs, packaging is even more critical.
Optical packaging is particularly challenging. A fiber must be aligned to a waveguide or emitter with sub-micrometer precision and held stable over temperature and vibration. A lens must be aligned to a detector array. An encapsulant must be transparent at the operating wavelength, stable under illumination, and compatible with the device materials. Any of these can fail, and a packaging failure is a device failure from the customer’s perspective.
Thermal packaging is equally important. A laser or high-brightness LED generates heat that must be removed. A detector’s dark current increases with temperature, so thermal management affects sensitivity. A modulator’s resonance shifts with temperature, so thermal stability affects performance. Packaging must provide a thermal path that keeps the device at its operating temperature without adding excessive cost or size.
Hermetic packaging — sealing the device in a package that excludes moisture and oxygen — is standard for many photonic products, but it adds cost and size. Thin-film encapsulation, where a barrier layer is deposited directly on the device, can be thinner and cheaper but must achieve very low water vapor transmission rates to protect sensitive materials. The choice depends on the material’s sensitivity, the product’s lifetime requirement, and the cost target.
Reliability and qualification
Reliability is the ability of a device to operate for its specified lifetime under specified conditions without failing. Qualification is the process of demonstrating reliability through accelerated testing. For photonic products, qualification typically includes high-temperature operating life tests, temperature cycling, humidity exposure, mechanical shock and vibration, and electrostatic discharge testing, following standards from organizations such as JEDEC, Telcordia, or automotive standards bodies.
Excitonic devices face specific reliability challenges. Perovskites can degrade under current, heat, and moisture. Organic devices can degrade under exciton and charge-induced chemical reactions. TMD devices can degrade through oxidation, defect migration, or contact degradation. Each mechanism must be understood, mitigated, and tested.
Accelerated testing is used to predict lifetime without waiting years. A device is operated at elevated temperature or current, and its degradation rate is extrapolated to normal operating conditions using models such as the Arrhenius equation for temperature acceleration. The validity of the extrapolation depends on whether the degradation mechanism is the same at accelerated and normal conditions. If a new mechanism appears at high temperature that does not occur at normal temperature, the extrapolation will be wrong. This is why qualification requires understanding degradation mechanisms, not just running accelerated tests.
For a new material, qualification is a long process. It requires building many devices, testing them under multiple conditions, and analyzing failures. It can take months to years, and it must be repeated if the process or materials change significantly. This is one reason why established materials have an advantage: they have years of reliability data, while new materials must build that data from scratch.
Cost: materials, process, and scale
Cost is the final gate. A device that performs well and is reliable will not be adopted if it costs more than the market will pay. Cost has three components: materials cost, process cost, and yield-adjusted cost. A process that uses expensive precursors, high-temperature steps, or slow methods will have high process cost. A process with low yield will have high yield-adjusted cost, even if its materials and process costs are low.
For excitonic devices, materials cost is often low — a monolayer of TMD uses a negligible amount of material, and solution processing uses inexpensive precursors. Process cost is the larger factor: growth, transfer, encapsulation, and packaging can be expensive if they require specialized equipment, manual steps, or low-throughput methods. Scaling to high volume can reduce process cost through automation, larger wafers, and improved yield, but it requires capital investment.
The economics of scale create a chicken-and-egg problem. High volume reduces cost, but high volume requires demand, and demand requires a product that is cost-competitive. Breaking this cycle typically requires a niche market where the device’s unique capability justifies a higher price, or a strategic investment that funds scale-up before demand is proven. Government funding, venture capital, and corporate R&D all play roles in bridging this gap.
Cost targets vary by market. A display pixel must cost a fraction of a cent. A datacenter transceiver must cost tens to hundreds of dollars. A scientific instrument component can cost thousands. A new technology is more likely to enter through a market with a higher cost tolerance, where its performance advantage justifies the price, and then move to cost-sensitive markets as volume grows and costs fall.
The valley of death: why good physics fails
The valley of death is not a single barrier but a series of them: manufacturing, packaging, reliability, qualification, cost, and market adoption. Each must be crossed. A technology can fail at any stage, even if it succeeded at earlier stages.
Common failure modes include: a process that works at small scale but cannot be scaled; a device that is efficient but unstable; a product that is reliable but too expensive; and a product that is technically excellent but has no market because it does not solve a problem customers will pay to solve. The last failure mode is particularly common in photonics: a device that is a solution in search of a problem.
Avoiding these failures requires early attention to manufacturing, packaging, and market needs, not just device physics. It requires engaging with manufacturing partners, packaging vendors, and potential customers early, before the device design is locked. It requires reporting manufacturing and reliability data, not just device performance. And it requires a realistic assessment of what the market needs and what incumbents offer.
How technologies cross the valley
Technologies that successfully cross the valley typically share several characteristics. They solve a problem that incumbents cannot solve, or they solve it significantly better or cheaper. They have a clear path to manufacturing that leverages existing processes where possible. They have a niche market that will pay for the initial product, providing revenue and learning. And they have a team that combines device physics, manufacturing engineering, and business understanding.
Organic LEDs are an example. They succeeded because they offered a form factor — thin, flexible, high-contrast displays — that incumbents could not match, and because the display industry invested heavily in manufacturing, encapsulation, and qualification over many years. Quantum dot displays succeeded by integrating quantum dots into existing LCD architectures, leveraging existing manufacturing rather than replacing it. Silicon photonics succeeded by leveraging CMOS manufacturing and by targeting datacenter interconnects, where the need for bandwidth and energy efficiency was acute.
For excitonic photonics, the path is likely similar: find a niche where excitonic physics provides a unique advantage, leverage existing manufacturing where possible, and build reliability and yield data in that niche before expanding to broader markets. The niche could be a specialized sensor, a tunable filter, a compact modulator, or a multispectral imager. Success in a niche builds the manufacturing capability and credibility needed for larger markets.
What to watch
For readers evaluating whether an excitonic technology is crossing the valley, the signals are manufacturing and business signals, not just physics signals. Has the company or research group reported wafer-scale yields? Has it demonstrated packaging and reliability data? Has it qualified a product to an industry standard? Has it announced a customer, a pilot production line, or a manufacturing partnership? Has it raised funding or secured a strategic investment for scale-up?
Physics milestones — higher efficiency, lower threshold, faster switching — remain important, but they are not sufficient. A technology that reports only physics milestones, without manufacturing, packaging, or reliability data, is still in the laboratory. A technology that reports yields, qualification results, and customer engagements is moving toward a product.
The valley of death is not a reason for pessimism. It is a reminder that commercialization is a different kind of work from research, with different metrics, different skills, and different timelines. The excitonic field has made remarkable progress on the physics. The next decade will test whether it can make equal progress on manufacturing, packaging, and business — the work that turns a demonstration into a product.
Sources and further reading
This article draws on literature on semiconductor manufacturing, photonic packaging, reliability engineering, and technology commercialization. Useful sources include textbooks on semiconductor manufacturing and yield, reliability engineering, and photonic packaging, as well as review articles in Nature Photonics, Nature Reviews Materials, and IEEE journals on manufacturing and packaging of photonic devices. Standards from JEDEC, Telcordia, and automotive bodies provide the qualification frameworks referenced here. Specific yield, cost, and lifetime numbers vary widely by device type and market, so this explainer avoids citing single-product values as general facts and describes the commercialization challenges qualitatively.