Why are both deep-ultraviolet LEDs and GaN power devices competing fiercely for the same AlN wafer?

AlN single-crystal wafers represent a rather niche market segment; however, whenever they are mentioned, two primary applications invariably come to mind: deep-ultraviolet LEDs and GaN power devices. Interestingly, the operating principles and application scenarios of these two types of devices are worlds apart—one involves light emission, while the other involves switching current—but both have stringent requirements for AlN wafers. Today, we’ll break down these two applications separately to explore how the same substrate material has become an essential component in two entirely different fields.

Deep-ultraviolet LEDs: Defects are their primary adversary.

Deep-ultraviolet LEDs (commonly referred to as UVC LEDs, with a light emission wavelength ranging from 200–280 nm) represent the most mainstream application domain for disinfection and sterilization; they are widely used in water treatment, air purification, and surface disinfection. Their core material system consists of AlGaN with a high aluminum content; this material is fabricated by epitaxially growing an AlGaN multi-quantum well structure on a substrate.

The issue lies in the choice of substrate. The most mature substrate used in the industry is sapphire—cost-effective, available in large sizes, and supported by a well-established supply chain. However, there is a significant lattice mismatch between sapphire and AlGaN; as a result, the dislocation density in the epitaxial layer typically ranges from 10⁹ to 10¹⁰ cm⁻². Deep-ultraviolet LEDs are particularly sensitive to dislocations—these dislocations act as non-radiative recombination centers, where electrons and holes “silently” recombine without emitting light, thereby consuming a large portion of the total luminous efficiency. As a result, the external quantum efficiency of deep-ultraviolet LEDs in the industry is generally low, and the dislocation density is one of the primary contributing factors.

Using AlN as the substrate yields entirely different results. Since AlN and AlGaN belong to the same material system and their lattice constants are naturally close, the dislocation density in the epitaxial layer can be reduced to below 10⁶ cm⁻² – three to four orders of magnitude lower than that achieved with sapphire substrates. A research team has fabricated deep-ultraviolet LEDs on AlN substrates using the HVPE method; with the dislocation density in the p-GaN layer controlled below 10⁶ cm⁻², the estimated continuous operating lifetime exceeds 5,000 hours (at a current of 150 mA).

In addition to its dislocation density, AlN offers another advantage that sapphire cannot match: deep-ultraviolet transparency. Deep-ultraviolet LEDs typically employ a vertical light-emission structure in which light must travel through the substrate before being emitted; if the substrate itself absorbs deep-ultraviolet light, the light extraction efficiency is reduced. AlN exhibits excellent transmittance in typical deep-ultraviolet wavelength ranges—such as 265 nm and 230 nm—making this an additional benefit compared to sapphire.

GaN power devices: Insulation and heat dissipation are essential.

GaN power devices—particularly high-electron-mobility transistors (HEMTs)—involve entirely different considerations compared to deep-ultraviolet LEDs: they do not emit light; instead, their design focuses on high voltage withstandability, low on-resistance, and high-frequency performance. The substrate requirements for such devices can be summarized into three key points:

First is the semi-insulating characteristic.

During operation, there must be no significant leakage path between the substrate and the device; otherwise, it will increase parasitic capacitance, degrade the device’s high-frequency performance, or even directly affect its withstand voltage capability. AlN exhibits a resistivity on the order of 10¹³ Ω·cm, making it a naturally excellent insulator that effectively isolates the device from the substrate and reduces leakage current.

Second, heat dissipation.

GaN power devices operate under high power densities, with heat concentrated in a very small area; however, when the substrate’s heat dissipation capability cannot keep up, the device may quickly degrade or even fail due to excessive temperature. AlN has a thermal conductivity ranging from 170 to 340 W/(m·K), which is comparable to that of SiC, and its heat dissipation performance is significantly superior to that of sapphire.

The third issue is the age-old problem of lattice matching.

Traditional GaN HEMTs are predominantly grown on silicon or sapphire substrates; however, the high dislocation density and thermal mismatch issues arising from lattice mismatch also limit the thickness of the GaN buffer layer and the reliability of the device. Using an AlN substrate enables the functional layers (such as GaN and AlGaN) to be grown under conditions that approximate “homogeneous epitaxy,” thereby reducing both strain and defect density.

Over the past two years, the academic community has continued to explore more radical device architectures—such as fabricating XHEMTs by sandwiching a GaN channel layer between two AlN layers and epitaxially growing them directly on an AlN single-crystal substrate. Due to the nearly perfect thermal matching between the substrate and the epitaxial layer, this structure offers superior thermal management performance compared to traditional architectures grown on 4H-SiC substrates, despite SiC itself having a higher thermal conductivity; this is because the interfacial thermal resistance between the AlN substrate and the AlN buffer layer is virtually negligible, whereas the interfacial thermal resistance of a heterogeneous substrate remains an unavoidable limitation. At a frequency of 10 GHz, such devices can achieve an output power of nearly 6 W/mm and a power-added efficiency of 65%, which is on par with the current state-of-the-art GaN HEMT technology.

Two distinct development paths with different focuses on wafer parameters.

Although both applications require AlN wafers, they place entirely different emphasis on wafer quality parameters:

Simply put, deep-ultraviolet LEDs prioritize crystal quality, whereas power devices prioritize electrical and thermal performance – although both fundamentally rely on the same underlying principle: lattice matching between AlN and the target epitaxial material.

Applying this to real-world products: when selecting a substrate, first clearly determine which side you need.

Products such as Semicera’s 10×10 mm monocrystalline AlN substrates are listed on the official website as two distinct primary application areas – deep-ultraviolet optoelectronics and high-voltage, high-temperature materials research. These products are categorized into three grades—P, R, and D—thereby providing customers with varying budgets and project phases with a range of options to choose from.

For epitaxial research and development focused on deep-ultraviolet LEDs, key metrics for assessing crystal quality—such as the half-width at half-maximum (HWHM) of the XRD rocking curve—should be closely monitored; a higher quality grade (e.g., within 100 arcseconds for P-grade) indicates better control over dislocation density.

When developing GaN power devices or HEMTs, in addition to crystal quality, it is also necessary to verify whether parameters such as resistivity and thermal conductivity meet the required specifications; these often need to be confirmed separately with the supplier, as not all product grades are automatically labeled with such information.

Final Note

When the same material attracts interest from two entirely distinct application fields simultaneously, it usually indicates that the material addresses a sufficiently fundamental challenge—namely, “lattice matching.” Deep-ultraviolet LEDs rely on this principle to enhance their luminous efficiency, while GaN power devices depend on it to ensure robust insulation and heat dissipation performance. Although their application pathways differ, their starting point is the same: finding a substrate that shares a compatible structural similarity with their own, thereby addressing the inherent defects of the epitaxial layer at their root.