The previous article mentioned that indium phosphide (InP) plays a key role in the upgrade of optical communication driven by AI computing power. But to understand why it" It must be it. It is necessary to start from the most fundamental physical principles.
The essence of semiconductor luminescence is that electrons transition from high energy levels to low ones and release photons after recombining with holes. However, the energy band structures of different materials vary, and the release efficiency of light differs greatly. Silicon (Si) is an indirect bandgap semiconductor. The momenta of electrons and holes are inconsistent, and additional phonons are needed during recombination. The probability of light emission is extremely low and the efficiency is extremely poor - this is the fundamental reason why silicon almost cannot emit light and cannot be used as a light source for lasers. InP belongs to direct bandgap semiconductors. The momentum of electrons and holes is the same. When recombining, photons are directly released, and the internal quantum efficiency can reach nearly 100%.
This physical difference has existed for more than a century since the discovery of the material. In 1910, Thiel first synthesized InP. Over the following decades, scientists gradually recognized its unique value in the field of optoelectronics. In the 1970s, lasers based on InP were successfully developed. In the 1980s, InP lasers had already been practically applied in optical fiber communication engineering.
The bandgap width of InP is 1.35eV, and the corresponding emission wavelength is approximately 0.92μm. More importantly, by epitaxially growing quaternary alloy materials such as InGaAsP and InGaAlAs on its surface, the emission wavelength can be precisely regulated, covering the two lowest loss Windows in optical fiber communication: 1310nm (with low dispersion) and 1550nm (with low loss). The emission wavelength of silicon is within 1100nm, which is completely unable to cover these two communication Windows.
The core physical parameters of InP make it superior to traditional silicon and gallium arsenide materials in multiple dimensions. According to publicly available material data, the electron mobility of InP can reach up to 4600 cm²/(V·s), and the hole mobility can reach 150 cm²/(V·s), which means that the electrical signal switching is faster and it supports high-frequency modulation above 50GHz. In contrast, the electron mobility of silicon materials is typically only one fifth to one tenth of that of InP. The thermal conductivity of InP is 0.7 W/(cm·K), which is superior to that of GaAs material. It has better heat dissipation performance and is conducive to the stability of high-power devices. Its microhardness is (435±20) mm⁻¹. The single crystal is brittle and soft, dark gray in color, with a metallic luster. It is stable in air at room temperature and begins to decompose at 360°C.
Overall, each of the three mainstream semiconductor materials has its own applicable boundaries. However, in the field of optical communication, the unique physical properties of InP make it the only choice: silicon is an indirect bandgap semiconductor that cannot emit lasers by itself, has a low electron mobility, and weak radio frequency performance, making it only suitable for passive optical integrated devices. The luminescence range of gallium arsenide is concentrated around 850nm, which is only suitable for short-distance low-speed interconnection. In the long-distance high-speed transmission scenario of data centers, the loss is too high. InP can directly generate lasers, has extremely low transmission loss in the communication band, and its electron mobility is much higher than that of silicon and gallium arsenide. Its radio frequency performance is also outstanding. It is currently the only material system that can simultaneously provide light source, high-speed modulation and detection functions.
The global optical module market data also confirms InP's dominant position. According to industry research data, in the demand applications of indium phosphide, the telecommunications market accounts for approximately 73%, data centers about 14%, and CATV about 4.9%. EML electro-absorption modulated lasers are currently the product type with the largest consumption of indium phosphide.
| "Comparison dimension" | Silicon (Si | Gallium arsenide (GaAs | Indium phosphide (InP |
|---|---|---|---|
| Semiconductor generational | The first generation | The second generation | The second generation |
| Bandgap type | Indirect bandgap | Direct bandgap | Direct bandgap |
| Bandgap width (eV | 1.12 | 1.42 | 1.35 |
| Electron mobility (cm²/V·s) | ~ 1400 | ~ 8500 | ~ 4600 |
| Hole mobility (cm²/V·s) | ~ 450 | ~ 400 | ~ 150 |
| Emission wavelength range | Non-luminescent (indirect band gap | ~850nm | Tunable to 1310nm/1550nm |
| Can lasers be prepared? | ❌ cannot | ✅ can (short distance, low speed) | ✅ energy (high-speed long-range, core of communication band) |
| Thermal conductivity (W/cm·K) | 1.5 | 0.46 | 0.7 |
| Dielectric constant | 11.7 | 12.9 | 10.8 |
| Melting point (°C | 1414 | 1238 | 1070 |
| Density (g/cm³ | 2.33 | 5.32 | 4.787 |
| Main application fields | Integrated circuits, discrete devices, photovoltaics | Radio frequency devices, VCSEL, short-range optical communication | High-speed lasers, photodetectors, radio frequency devices, satellite communications |
| The role in optical communication | Passive optical waveguide/modulator | VCSEL chip (Short range | EML/DFB/CW lasers, PIN/APD detectors (long-range high-speed) |
| Relative manufacturing cost | low | In the | high |
| Scarcity of raw materials | Rich | Ga is relatively rare. | Indium is an associated metal and is scarce |
| Mass production of 8 inches and above | ✅ mature mass production | ✅ can be achieved | ⚠️ 6 inches is the current watershed |
In the manufacturing and application of InP substrates, size specifications are a key variable that directly affects industrial efficiency. In the industry, wafers are classified into three specifications: 2 inches, 4 inches, and 6 inches. The larger the wafer size, the more chips can be cut from a single substrate, and the lower the manufacturing cost per wafer. Data from Coherent's 6-inch production line in Sherman, Texas, shows that the chip output of 6-inch wafers is more than four times that of 3-inch ones, while the cost is less than half.
However, the manufacturing threshold for large-sized InP substrates is very high. The growth of InP crystals must be carried out under conditions of high temperature (approximately 1070°C) and high pressure (exceeding 27 atmospheres). When the crystal size expands from 2 inches to 6 inches, the original micro-sized crystal growth technology cannot be simply replicated. Enterprises need to independently develop or upgrade large autoclaves, high-power heaters, and higher-precision temperature control systems. Therefore, 6 inches is regarded by the industry as "..." "Technological watershed" . At present, the mainstream single crystal growth methods include VGF (Vertical Gradient Solidification Method), VB (Vertical Bridgerman method), LEC (Liquid-sealed Krausski method), etc. Among them, the VGF method performs best in terms of crystal uniformity and dislocation density control, and is the mainstream route for 6-inch mass production.
After the crystal ingot is grown, it still needs to go through precise processes such as cutting, edge grinding, grinding, polishing and cleaning. The surface roughness of the finished substrate product must reach Ra<. 0.2nm, warpage Bow<" 10μm, these indicators directly affect the downstream epitaxial quality and device yield.
In the next article, we will start from" Where is InP used? From this perspective, let's take a look at how indium phosphide plays a core role in 800G optical modules, automotive lidars, and 6G communications.