Device Development History
Optical Devices (2) - Process Technology Progress -
Basic optical devices have layered structure with a long-wavelength composition semiconductor as the light-emitting (active) layer sandwiched by a short-wavelength (cladding) layers. To improve performance and reliability, it is important to refine both the quality of the crystal material formed on the substrate as well as to precisely crystal lattice matching of the active layer with the substrate. The LD lasing wavelength is determined by the combination ratio of the active layer InGaAsP elements. The thickness of the layer is on the order of 0.1 µm. Consequently, an early key problem was how to grow a thin semiconductor layer of the right composition and crystal-lattice on a substrate without defects
Structure of Basic Optical Device
1. Transitions of Crystal Growth Equipment
Liquid Phase Epitaxy
Liquid Phase Epitaxy (LPE) is a precipitation growth method in which the substrate is in contact with molten semiconductor materials. First, the growth sequence is arranged using semiconductor growth solutions dissolved in solution reservoirs at a sufficiently high temperature. Next, the substrate mounted on a holder is slid along so that the substrate comes into contact with the molten solutions. As the temperature is gradually lowered under these conditions, the semiconductor materials that are not dissolved in the solution precipitate out on the semiconductor substrate. When the semiconductor layer reaches the required thickness, the substrate slides through the next semiconductor solution, and the same process is repeated over. The layer thickness is controlled by the temperature drop rate and holding time in the solution.
Schematic of Liquid Phase Epitaxial Growth Equipment
The composition of the semiconductor layer and the emitted light wavelength are determined by the proportion of the elements in solution. Consequently, the weights of each material in the solution reservoirs are measured so as to satisfy a target weight ratio. This requires precision adjustments in µg units using an electronic balance and if the measurements are not correct, they are repeated over until correct by cutting out materials using a surgical scalpel. This procedure is used for the materials of each layer on the substrate. Since the assembled materials are unevenly distributed in the solution immediately after dissolving, pre-work preparation time is required before use so each element is completely dissolved in a homogenous solution.
The LPE method uses relatively simple equipment and has the advantage of generating a high-quality layer matching the substrate crystal lattice. This equipment was introduced by our company in 1975 when development of optical devices started in earnest and was used for release of many products, such as LDs for Anritsu’ optical pulse testers. A disadvantage was that growth of a complex multi-layer structure was restricted by the number of reservoirs, it was difficult to grow layers thinner than 0.1 µm with good reproducibility, and limited on the size of growable substrates. Today we have transitioned to the vapor phase epitaxy method as explained next.
Vapor Phase Epitaxy
Using the LPE method, the raw materials for growing crystals are supplied in liquid form whereas with Vapor Phase Epitaxy (VPE), the raw materials are supplied as a vapor to grow crystals on the substrate. There are many variations of VPE, including chemical reaction methods using pyrolysis and physical methods like vacuum deposition to form the semiconductor crystal on the substrate. But whatever the method variation, they all feature the ability to grow uniform crystals with high accuracy. The former is typically called Metal Organic Chemical Vapor Deposition (MOCVD), while the latter is known as Molecular Beam Epitaxy (MBE). The active layer of recent LDs is not a single layer as described above but uses a quantum-well structure with multiple layered semiconductor consisting of mono-atomic layers, which greatly improves performance.
The MOCVD method is performed in a normal atmosphere or low-pressure environment with a balanced feed of gaseous raw materials containing organic metals; pyrolysis of the gaseous raw materials on the heated substrate results in growth of the required semiconductor layer. Various crystals can be formed by changing the proportions of gaseous feed raw materials and this method is used widely due to support for growth of multiple wafers. However, care must be exercised regarding safety measures, because many of the gas feedstocks have highly toxic. We started introduction of this equipment in 1990 and it is still in active use today.
MOCVD Equipment Schematic
The MBE is a method in which the raw materials is heated and evaporated into a beam in a high vacuum to grow crystals directly on the substrate. The film composition is controlled by the amount of vapor deposition of each raw material, and the quality can be evaluated as the film grows. As a result, this method is good for growing accurate mono-atomic films and is still used for R&D development. However, it is difficult to manage of an extremely high vacuum, and is not used for mass-production because of its slow growth rates and an unsuitable method for growing multiple wafers.
MBE Equipment Schematic
2. Post-processing Advances
After completing crystal growth and electrode forming so that current can flow, the wafer is divided into individual LD chips. There was no automated machine that you can set up and the device will automatically take care of it, and every work was done manually.
Optical devices have an active layer with a thickness of several µm in the upper layer the substrate. Since the substrate is usually about 300 to 400-µm thick, there is actually an area that is not used, but handling is difficult if this area is too thin at the start. Grown wafers are machine-polished to a thickness of about 120 µm and then hand-finished finally to about 100 µm. As an expert operators, thickness was known by touch of the finger. The wafer is likely to bend and break if over-polished to excessive thinness. New trainees learning the process sometimes polish the opposite side of the wafer by mistake to remove the active layer—causing embarrassed faces all round.
This process attaches gold-based metal and gold plating to the multiple layers of the semiconductor wafer. At the start, it was a simple metal vapor deposition process on the both side without an electrode pattern. Electrode forming is an important process that can cause increased resistance if there are problems with fabrication materials, coating method, and annealing (heating). Since the properties can change considerably depending on the equipment, when introducing new equipment, it is important to perform adequate testing of coating conditions before using it. At first glance, this electrode forming may seem inconspicuous but since it has a large impact on performance and reliability, the process continues to be improved.
After electrode forming, the wafer is divided into individual LD chips by a process called cleaving. Silicon wafer, etc., chips are separated by circular knife blade, but this method cannot be used for LD chips because the cut surface is used as a laser-reflection mirror. Generally, minerals and crystals have a property called cleavage where they can be split along a plane where the atomic bonds are weak, resulting in a flat smooth surface. Division of wafers of compound semiconductor crystals use this cleavage property. After positioning the wafer growth surface face down and covering with a film, gently inserting a surgical scalpel at an angle splits the wafer along the cleavage plane. Repeating this process over cuts the chip to the required size. Only skilled technicians working with a microscope could divide a wafer into its original chip size of 300-µm in length and 400 µm in width. Now these procedures are automated and the people who know about this work is gradually decreasing.
First, LD chips are sorted for faults using a pulse current test. In the early days, the number of good chips was extremely low and finding one was like a treasure hunt while praying. There was little know-how about antistatic measures and many LD chips died suddenly due to static discharge in dry winter air. Mounting good chips to a heatsink is called bonding in which the LD chip is soldered (below figure) on a diamond sub-mount attached on a copper heatsink. The product specifications required using AuSn (gold-tin alloy solder), but the chip was soldered to the heat sink occasionally using soft indium solder without a diamond sub-mount at the experimental stage. Diamond sub-mounts were purchased with pre-formed electrodes but were difficult to solder directly, which required time and effort to solder-coat materials beforehand. It was an impressive sight to see many skilled workers lined up opposite bonding machines all day.
Magnification of Initial LD Chip
Nowadays, bonding has switched to direct attachment to ceramic heat sinks and soldering uses automated processing which became possible due to purchase of pattern solder masks, so it feels like a different era. However, even now, small production lots for R&D use manual processes rather than automatic machines. The following figure shows an example of current semiconductor optical amplifier device on a chip carrier.
Please visit our sensing and devices website for a more detailed description of LD structure and manufacturing processes.
Current Optical Device