III-V Epi Wafers: Why They're so Important

The compound semiconductor industry has always played a secondary role compared with its more mature sibling: the silicon semiconductor industry. In fact, for a long time many were convinced that compound semiconductors would always be the material of the future.

Somehow all that changed in the late 1990s. During the past few years, an impressive amount of capital was invested in these exotic materials all around the world. Startups sprang up overnight from Silicon Valley to Taiwan to Europe. To understand what factors drove this change, it is important to compare and contrast the unique properties of III-V materials with silicon properties, and review the various useful applications of these novel materials.

III-V materials get their name from their placement in the Periodic Table. For example, gallium (Ga) comes from Group III and arsenide (As) comes from Group V. The carrier mobility is much higher in III-Vs vs. silicon, and they have a very useful light-emitting or -detecting property. Although silicon may continue to encroach on III-V territory for various high-speed and high-frequency devices, it is not the material of choice for such photonic devices as LEDs, lasers, detectors, etc. Some III-V strongholds like HBTs are under attack from SiGe, but one might make a case that it is also a compound semiconductor. Even so, many companies continue to jump into the InP HBT field.

From an application point of view, the III-V semiconductor industry can be broadly classified into the following segments:

• Lasers. For telecom and datacom applications, these include pump lasers, tunable lasers, VCSELs (vertical-cavity surface-emitting lasers), detectors, modulators, amplifiers, etc., which are used mainly for increasing bandwidth. Despite the recent industry slowdown, Nielsen/Netratings in July observed that there was a 100% increase in the number of broadband users over the previous 12 months. Even today, industry forecasts continue to point toward a worldwide increase in the number of broadband users. Many believe that the storage sector is poised for a huge increase in affordable storage capacity thanks to the ongoing developments in GaN-based blue lasers, which have a shorter wavelength than conventional red lasers. Yet another application for lasers is in medical instrumentation.
• LEDs. High-brightness LEDs have certainly reached a critical mass-market stage with booming applications in traffic lights, signage, automobiles, very large video screens, etc. This sector has enjoyed robust growth of more than 50% per year over the past several years.¹ The advantages of LEDs in traffic lights have been very well documented, and the momentum for switching to LED traffic lights is so great that there is a possibility of lawmakers making it illegal to use an incandescent bulb in a traffic signal.² In the case of very large video screens as those found in sports arenas, LED screens are replacing conventional CRTs in large part because of superior resolution. The past two years have seen enormous capacity expansion for LED plants all over the world, but particularly in Asia. As prices tumble, we are sure to see even more LEDs in our day-to-day lives. Recently, white LEDs have started appearing in consumer products and will continue to get significant attention in R&D. Of course, the Holy Grail is for LEDs to challenge the stranglehold of incandescent and fluorescent lamps in the general lighting area. The global market for general lighting applications is estimated to be a staggering $12B.¹
• Wireless/telecom/datacom devices. It is safe to say that DWDM, CATV and ultrahigh-bandwidth networks would not have been possible without III-V semiconductors. Products based on pHEMT, HBT and MESFET technologies go into making every cell phone and wireless network. III-V devices are routinely used for amplification, switching and frequency conversion. This segment is bound to have more exciting news going forward due in large part to efforts in the 2.5G and 3G cell phones. Although the mood for adopting these technologies has tempered this year, no one doubts that the demand for the wireless high-speed Internet will be huge. The emerging optical network will depend on III-V devices for intelligence. In data communications, III-V devices are required to make network processors and support the physical layer requirements of Fibre Channel and the Gigabit Ethernet. High-data-rate, high-capacity crosspoint switches are also expected to usher in the eagerly anticipated era of HDTV.
• Optoelectronic or photonic ICs. This is a field still in its infancy but one that has generated a lot of interest because of the anticipated need of bringing together lasers, modulators, amplifiers, waveguides and detectors onto a single chip.
• Solar cells. This has remained at best a niche market. Despite the environment-friendly nature of this renewable technology and regular improvements in photovoltaic efficiency, solar cells are not as cost-effective as conventional energy sources except in special applications.

All these applications have one requirement in common and that is a III-V semiconductor epiwafer of uncompromising quality. The epi deposition step is the key enabling process, which makes it possible to build these devices. Compared with a silicon CMOS epi structure, a III-V epi structure can be very complex and is made up of several layers of differing materials, concentrations and thicknesses. For example, these epi layers could be made of GaAs, AlGaAs, InP, InGaP, InGaAs or a variety of other III-V materials (Fig. 1). There are also some unique and novel efforts to grow GaN on silicon (e.g., Nitronex), and to grow GaAs on silicon (e.g., Motorola).

1. Cross section of a laser device called a NECSEL. (Source: Novalux)

While there are several epi deposition methods, there are mainly two popular techniques called MOCVD (metal organic chemical vapor deposition) and MBE (molecular beam epitaxy). In an MOCVD reactor, metalorganic compounds in liquid or gaseous form are used to deposit a layer of desired material composition on a substrate in a CVD reactor. In an MBE reactor, the substrate and several evaporation furnaces containing elemental species are placed in a UHV chamber. The operator uses various shutters and, through precise control of temperature, can manipulate the deposition of very complex single-crystal structures of virtually any material composition and doping. Generally, the MOCVD technique is used for high growth rates in volume production requirements of LEDs and other optoelectronic devices, and the MBE technique for superior epi layer control in wireless devices.

As the market for lasers, detectors and LEDs continues to grow, the volume of compound semiconductor wafers has been steadily increasing. This has resulted in several merchant III-V epi wafer suppliers entering the market. Naturally, this is followed by price erosion and, therefore, wafer yields become extremely important. Increasingly, epi growth managers demand the ability to carry out rapid non-destructive assessment on whole wafers to map the morphological, structural, electrical and optical properties of wafers. However, there are some destructive techniques, which are indispensable and continue to be popular.

The non-contact characterization of epi wafers is traditionally done by a few different methods:

• Mapping the photoluminescence (PL) properties of the epi wafer. When a laser of a wavelength shorter than that of the material under investigation strikes the epi layer it emits a PL signal. The characteristics of this emission are important to understand the properties of the epi layer. Traditionally, a diffraction-grating-based spectrometer is used with a single PIN diode coupled with an automated stage to measure this signal. More recently, innovative systems have been introduced that use PIN diode arrays for detection, resulting in ultrafast mapping of whole wafers (Fig. 2).³ These tools can also be used to obtain epi thickness or reflectivity maps. One particularly popular process control application is mapping the Fabry-Perot dip or the stop band center for a VCSEL wafer. Some PL measurements are also done at cryogenic temperatures or with an ultrafine spatial resolution. However, these applications tend to be primarily for specialized research interests.
  2. PL image maps of a 2 in. diameter, strained multi-quantum well VCSEL wafer. (Source: Nortel Networks’ CoreTek Division)
• Mapping the X-ray properties of the wafer. X-ray diffraction provides a unique and indispensable tool to directly probe the electron density variations within semiconductor materials. In III-V semiconductors, this technique focuses on the periodic arrangement of the unit cell, physical boundaries of the unit cell periodicities and the orientation of unit cell periodicities.⁴ Typically, the analysis is simplified when the diffraction information from the substrate material is used as a fixed reference point from which the various epi layers are characterized. Also, with the robustness and reliability of the genetic algorithms coupled with pattern recognition, automated analysis of high-resolution X-ray diffraction patterns is now a reality within a production control environment.⁵

• 	3. A typical GaN wafer sheet resistance map. (Source: Lehighton Electronics)

  Mapping the resistivity of the epi layer. The measurement of sheet resistance requires inducing eddy currents in a semiconductor layer. Under certain conditions, the power absorbed by the test material can be detected and this is inversely proportional to the sheet resistance (Fig. 3).
• Mapping the thickness and carrier concentration of epi layers by FTIR spectrometers. Recent advances in model-based algorithms are employed in measuring the properties of individual films in a complex stack of epi films as in a PIN detector structure.

There are also a few destructive tests that are still used routinely:

• Electrochemical C-V profiling uses what is essentially a C-V profiler that can obtain carrier concentration as a function of depth to unlimited depths. This is not possible in a conventional C-V system because, as you raise the voltage beyond a certain point, the sample is in avalanche breakdown condition and Schottky equations are no longer valid. In an electrochemical C-V profiler, a capillary column of an electrochemical solution is used as a Schottky contact rather than a metal contact. The advantage of such a contact is that one can make measurements with negligible bias, etch the material a tiny amount and continue to repeat the measurement cycle to unlimited depth. This is important for III-V epi structures, which can have very complex layers of various epi films. Taken together, this stack can measure up to several microns in thickness. Some have worked in conjunction with a photovoltage spectrometer, making it possible to measure the bandgap as a function of depth.
• Hall measurement is a classic test run on a Van der Pauw patterned or clover-shaped sample in a magnetic field per ASTM F-76 standard. The objective of this test is to measure the resistivity, carrier concentration and mobility of the epi layer. Depending on the material being investigated, the measurement may have to be tested at cryogenic or elevated temperature and some tests may require a variable magnetic field.

One recent change in the III-V industry is that several tool manufacturers have introduced new cassette-to-cassette tools in response to customer demand. The GaAs foundries have used such equipment for the past several years, and today even the optoelectronics companies have embraced these tools of mass production. What is taken for granted and accepted as the standard in the silicon industry is now bringing improved productivity to the III-V sector.

A 2 in. wafer can carry 25,000 VCSELs and can cost hundreds of thousands of dollars. One cost-of-ownership calculation of mapping PL tools suggests a cost of only 16 cents per wafer. Contrast this with the typical cost of depositing epi, which is in excess of $20 per wafer. Such demonstrations of cost-effectiveness have resulted in epi growth managers and epi customers mandating maps of every wafer rather than just random sampling. Mapping wafers with superior resolution also allows one to catch flaws in epi wafers that would otherwise have been missed, increasing productivity, yield and profits. Thus, as we wait for the current downturn to end, only those companies dedicated to cutting costs and improving yields will embrace new technology and survive. Novalux is a good example of this.

Novalux characterizes its NECSEL devices all the way through the manufacturing process from wafer level to module. "We can build a 3-D wafer map at any stage to show us power, wavelength and other important characteristics. This data is key to known good die and the NECSEL's promise of high powers in a cost-effective package," said Gary Oppedahl, Novalux's vice president of operations. This is a far cry from just three or four years ago when III-V semiconductor manufacturing was accomplished using platoons of PhDs with broken wafer pieces because they were simply too valuable to be discarded.