About this document
In heterogeneous silicon photonics integration, each added material exposes heat-sensitive optical devices to the bonding temperature again. This Technical Note sets out the principle, process and equipment of interface-selective bonding, which converts light into heat only at the designed bond interface. The full report is available in Korean. An English summary is provided on this page.
Download
An English edition is not available yet. The Korean document is below.
Summary
Optical devices degrade at bonding temperatures that electronic devices tolerate, yet conventional bonding heats the whole chip. Light at 1.55 μm passes through Si and InP and turns into heat only in an absorber a few tens of nanometers thick at the interface. Pulse width limits the heated depth, and irradiation position and sequence limit the heated volume, decoupling bonding temperature from device temperature. A list of related patent applications is included.
Conclusion
Unlike electronic devices, optical devices lose performance at the bonding temperature, so photonic integration depends on bonding with as little heat as possible. Conventional bonding heats the entire chip, so the heat needed for bonding reaches every optical device, and each additional material reheats the devices already in place. Interface-selective bonding does not lower the bonding temperature. It separates the bonding temperature from the temperature of the optical devices. At 1.55 μm the photon energy is 0.80 eV, below the bandgaps of Si (1.12 eV) and InP (1.34 eV), so the light passes through waveguides, cladding and handler to reach the interface. An absorber 20-60 nm thick at the interface converts it into heat, and the bonding-pad mask fixes both the position and the thickness of that heat source. Because thermal diffusion length scales with the square root of irradiation time, pulse width sets the heated depth. Controlling irradiation position, sequence and dose for each site makes the heated volume the sum of the bond sites rather than the whole wafer. The source document gives 375 °C at the bond interface and 52 °C in the optically active layer 30 μm away, under a 1 μs pulse. The 323 K difference between the two points at the same moment is obtained without lowering the bonding temperature. Since heating and cooling both take place within microseconds, oxide treatment and metal bonding finish before reoxidation can occur, which removes flux application and cleaning from the process flow. The same selective-absorption principle drives splitting, healing and bonding in layer transfer, allowing InP and InGaAsP active layers to be harvested repeatedly from a single donor wafer. In a double-sided package on a sapphire handler, adding bonding steps does not renew the thermal history of interfaces already bonded. The bonder separates the two functions: the pressing head handles only load and flatness, while heat arrives through an optical path from below, so heating and cooling the head drops out of the process time. The first application is co-packaged optics (CPO), where the electronic IC (EIC) and photonic IC (PIC) must be bonded together. Heating only the interface protects a PIC that would otherwise see the bonding temperature. As silicon photonics adds more steps that join electronic and optical parts, the number of interfaces this principle applies to grows with it. The document closes with a list of 107 patent applications, ranging from core principles such as selective absorption and optical paths to material-specific applications and equipment, control and inspection.
Topics
- Photonics
- CPO
