| |
By Michael L. Simpson and M. Nance Ericson
ORNL has developed optical application-specific integrated circuits to make improved photodetector imaging devices. ORNL also has built the first photospectrometer realized in a standard integrated circuit process.
Since very early in the semiconductor revolution, silicon has been the workhorse in our electronic devices. Here are three reasons. Silicon readily grows a stable oxide, which protects silicon from environmental degradation and creates an insulating layer, allowing the development of three-dimensional structures needed for electronic devices. In addition, the electrical properties of silicon enable the creation of two types of complementary transistors, while other semiconductor materials (e.g., gallium arsenide) allow the development of only one type of transistor. Finally, silicon is an inexpensive and readily available material.
Possibly more important, however, is the tremendous financial investment in the silicon processing infrastructure used by computer, electronics, and telecommunications industries worldwide. A challenger to silicon would have to offer overwhelming advantages for this investment to be discarded in its favor. So, to make a major impact in the integrated circuit arena, it is essential to find a way to add functionality to silicon devices manufactured using standard processes. Such is the aim of the optical application-specific integrated circuit (OASIC) research being jointly performed by ORNL's Instrumentation and Controls, Solid State, and Life Sciences divisions.
|
|
|
Fig. 1. Mike Simpson examines the bioluminescent bioreporter chip (magnified on the monitor in the background), an example of optical application-specific integrated circuit (OASIC) technology. The chip emits a visible blue-green light when its living sensors (bioluminescent bacteria) detect certain pollutants and explosives.
|
It has long been known that integrated circuits (ICs) composed of silicon devices are sensitive to light, but this photosensitivity has been thought of more often as a nuisance than a desirable feature. However, some researchers are beginning to use this feature to overcome recently observed disadvantages of a standard imaging camera called the charge-coupled device. They are trying to solve these problems and advance imaging technology through so-called "smart pixel" sensors that can be fabricated using standard IC processes. The strength of standard IC processes is that they allow analog and digital processing circuits (i.e., "smarts") to be integrated with the light-sensitive pixel on the same chip. Combining functions and distributing signal processing in advanced IC photodetectors provide advantages such as the ability to compress data into a smaller space, increase readout speed, and extract more information from images. However, it is widely believed that color information cannot be extracted from light signals forming images that are picked up by these IC photodetectors without the addition of nonstandard filters, dyes, or defractive devices. Unfortunately, once steps are added that move away from standard IC processes, the usual advantage of low-cost mass production is lost. We at ORNL believe that a more capable photodetector technology can be realized using integrated OASIC and standard IC processes. Development of such a technology could lead to complete laboratory instruments-on-a-chip that aid in the detection of environmental pollutants (such as the "critters on a chip" device shown in Fig. 1 and described in the article, "Critters on a Chip"), the discovery of new therapeutic drugs, rapid and inexpensive medical diagnoses, and solutions to many other detection problems.
OASIC research at ORNL is centered on achieving application-specific spectral responses using only the masks, materials, and processes inherent in a standard, readily available IC manufacturing method known as the complementary metal oxide semiconductor (CMOS) process. CMOS processes dominate existing IC markets that focus on microprocessors and many telecommunications applications. It is important to note that the technical drivers for these processes are large-volume computing markets. Because IC manufacturers will not make expensive process changes to accommodate OASICs, to achieve our goals we must exploit the wavelength-sensitive mechanisms native to the CMOS process.
The response of any photodetector is composed of three elements: how efficiently the light is transmitted to the detector (transmission properties); how efficiently light is absorbed by the detector (absorption properties), where the intensity of each light signal is converted to an electrical charge; and how efficiently each light-created charge is collected by the detector (charge transport properties). Each of these properties can be manipulated to form spectrally independent photodetector responses.
Transmission Properties
How efficiently light is transmitted to a photodetector chip depends heavily on the material covering the photodetector. In a standard CMOS process, three materials can be placed above the silicon base of the detector chip: aluminum, silicon dioxide (glass), and polycrystalline silicon (poly-Si). Aluminum is used for interconnects between circuits on a silicon IC chip, but for optical applications it is useful only as a light shield or mask. The glass layers serve as insulators on standard IC chips, but for OASICs they serve as spacers in stacks of thin-film materials used to form optical filters. The material that forms the heart of the OASIC chip is poly-Si because it is a wavelength-sensitive reflector, absorber, and transmitter of light, and its proper use can lead to very useful spectral functions (e.g., determining the intensity of each color present in the light signal). Indirectly, aluminum, glass, and poly-Si also play a role in manipulating charge transport properties, as will be described later in this report.
The transmission properties of the photodetector are manipulated by specifying the use and placement of two layers of poly-Si and five layers of glass to form a stack of thin-film filters. Fortunately, this use of poly-Si and glass layers is allowed in a CMOS process, although these materials are not used in this way in standard IC designs. The use of existing layers in standard CMOS designs allows spectrally selective thin film filters to be implemented using widely available, low-cost, standard CMOS IC processes.
Absorption Properties
When a photon of light is absorbed by a photodetector chip, an immobile electron in the valence band (outside electron orbit of a silicon atom) becomes a mobile carrier of electric charge in the conduction band of silicon. This process also leaves behind a mobile hole in the valence band (i.e., a missing electron leaves a positively charged hole that can attract another electron).
Because silicon is an indirect semiconductor, the probability of a photon being absorbed a given depth in a silicon chip depends largely on the photon's energy. Short-wavelength, high-energy light is absorbed near the silicon surface, and longer-wavelength, lower-energy light is absorbed deeper in the material. At long enough wavelengths, the photons lack sufficient energy to move an electron from the valence band to the conduction band, so the silicon chip effectively becomes transparent to this light. Silicon stops absorbing light in the near-infrared region at a wavelength of about 1100 nanometers.
This depth-dependent absorption of light can be used to determine spectral content (i.e., what colors are in an image and how bright is each color in each pixel). A double-junction photodiode that can be constructed using CMOS processes has one diode junction near the surface and another junction deeper in the material. The shallow-junction device in a silicon chip will preferentially collect the charges created by short-wavelength light, while the deeper junction collects the charges created by longer-wavelength light. Some standard IC processes may have a third useful junction depth, but unfortunately the number of useful junction depths is limited to just a few in any process. Use of multiple junction depths is another method that can be used to tailor the spectral response of photodetectors in standard processes.
Charge Transport
|
|
|
Fig. 2: (a) P-channel MOSFET cross section showing multiple junctions and (b) detection of photogenerated electron-hole pairs using a single pn junction.
|
Once a mobile charge is created by absorption of light, one of two possible events will occur. Either the charge recombines-that is, the electron drops back into the valence band and eliminates the hole-or it crosses a diode junction, producing an electrical signal. Because the charge that crosses the diode junction creates a useful signal while the charge that recombines with the hole is lost, charge transport properties significantly affect the spectral response of the photodetector. Because of the physical properties of silicon, recombination usually takes place in "trapping centers" created by impurities or disruptions in the lattice structure. One very significant location for trapping centers is the silicon-glass interface at the surface of the photodetector where the silicon lattice is abruptly terminated. Because shorter-wavelength light is absorbed nearer the silicon surface than is longer-wavelength light, shorter-wavelength light is more strongly affected by surface trapping. All silicon-based IC processes exhibit surface traps though special processing steps can be taken to minimize surface trap density.
How an OASIC Photodetector Works
Standard CMOS fabrication processes form a special type of transistor called a metal oxide semiconductor field effect transistor (MOSFET). A MOSFET can be either an n-channel or p-channel transistor. (Figure 2a shows the cross section of a p-channel MOSFET transistor.) These devices function in a complementary fashion to form larger functional circuit blocks such as digital gates and amplifiers. For this example, two junctions exist that can be used for photodetection-the p-diffusion (p-diff) to n-well junction, and the n-well to p-substrate junction. Most standard CMOS fabrication processes have only two junctions available, as this is all that is needed to fabricate MOSFETs.
Careful connection of the standard materials used to make MOSFETs results in the formation of a photodetector. When light strikes the photodiode structure, an electron-hole pair is generated within the semiconductor material, provided that the photon energy is greater than the bandgap energy for silicon (see Fig. 2b). A depletion region is formed at the interface between oppositely doped materials, resulting in an electric field that accelerates the holes to the p-doped region, and the electrons to the n-doped region. The result is an electric current that is proportional to the light incident on the photodetector.
|
The density of surface states is controlled by the "quality" of the oxide layer that terminates the silicon lattice. In CMOS processes this initial oxide layer is supposed to be either a carefully grown thin oxide (gate oxide), or a less controlled, relatively thick grown oxide (field oxide). However, by manipulating the process in a nontraditional manner, the terminating layer can be a deposited oxide or a spun-on oxide. Each of these four oxides provides a different density of surface states, strongly influencing the response of the photodetector in the near-ultraviolet, blue, and green regions.
Some nontraditional oxide manipulations cause an aluminum layer to be deposited on the photodetector surface and then etched away leaving behind additional surface trapping states. However, if a poly-Si layer is placed over the detector, the aluminum is deposited on the poly-Si. Although the poly-Si does not survive the aluminum-etch step, the silicon surface is protected and the excess surface trapping states are not formed. Surface trap density can be manipulated in standard silicon processes to influence short-wavelength sensitivity by selective placement of layers that will be etched away during processing steps, resulting in the addition of surface traps.
Photospectrometer Developed at ORNL
The above discussion describes how to fabricate photodetectors that have independent spectral responses, using a standard CMOS IC process. Unfortunately, the responses of the devices taken individually are not very useful. However, by taking advantage of the true strength of OASICs (signal processing integrated with the photodetector on the same chip), an ORNL team has developed what is believed to be the first photospectrometer realized in a standard IC process. The team includes researchers in the Instrumentation and Controls Division (Michael L. Simpson, M. Nance Ericson, Alan L. Wintenberg, William B. Dress, and David N. Sitter) and the Solid State Division (G. E. Jellison).
|
|
|
Fig. 3. A single-chip photospectrometer
|
In this new single-chip photospectrometer, the input spectrum is measured by computing weighted sums of the detector responses to form a set of outputs proportional to the signal power in discrete optical bands. This device, which measures 2.2 millimeters (mm) ´ 2.2 mm ´ 0.5 mm, is shown in Fig. 3.
This photospectrometer will be an integral part of a new generation of microinstrumentation that will include spectrally sensitive electro-optical detection, analog and digital signal processing, microelectromechanical systems (MEMS), and wireless communications circuits. Planned developments include a photospectrometer-on-a-fiber for in-vivo and in-vitro bioanalytical applications (with Tuan Vo-Dinh in ORNL's Life Sciences Division), a four-color imager for DNA sequencing instrumentation (with William B. Whitten in ORNL's Chemical and Analytical Sciences Division), and a color silicon retina.
Conclusions
While silicon will continue to be the dominant material used in IC processes for the foreseeable future, significant advances in the state of the art of microinstrumentation can be made by exploiting many features of these standard processes that are usually neglected. OASICs provide one example where a great deal of measurement capability has been underutilized by traditional circuit techniques. The continuing development of OASICs at ORNL will contribute to a new generation of microinstrumentation that will take measurements out of the laboratory and into the environment to advance scientific understanding.
BIOGRAPHICAL SKETCH
MICHAEL L. SIMPSON received a Ph.D. degree in electrical engineering from the University of Tennessee at Knoxville (UTK) in 1991 and shortly thereafter joined the Monolithic Systems Development Goup in ORNL's Instrumentation and Controls (I&C) Division. He has been a member of the international collaboration that developed the WA-98 electromagnetic calorimeter now in operation at the CERN laboratory near Geneva, Switzerland, and of the PHENIX collaboration that is developing a detector system for the Relativistic Heavy Ion Collider under construction at DOE's Brookhaven National Laboratory. Simpson's recent research interests have been in the emerging field of monolithic sensors. He was the principal investigator for a two-year research project to create an optical application-specific integrated circuit (OASIC) capability at ORNL. This work has resulted in the first photospectrometer realized in a standard integrated circuit process. Simpson also has co-developed a novel living sensor concept combining an OASIC with genetically engineered bioluminescent bacteria. A patent based on this technology, which has generated a great deal of commercial interest, is now being written. Simpson is an adjunct faculty member of UTK's Electrical Engineering Department and the ORNL coordinator of the UT/ORNL Joint Program in Mixed-Signal Very Large-Scale Integrated Circuits and Monolithic Sensors. He is a senior member of the Institute for Electrical and Electronic Engineers (IEEE) and an active member of the Nuclear Science and Solid-State Circuit Societies. He holds three patents and has three patents pending. He is a member of the Eta Kappa Nu Electrical Engineering Honor Society and will be listed in the fourth edition of Who's Who in Science and Engineering.
M. NANCE ERICSON is a research and development engineer in the I&C Division's Monolithic Systems Development Group. He received a B.S. degree in electrical engineering from Christian Brothers University in Memphis, Tennessee, in 1987, and an M.S. degree from UTK in 1993. In 1987 he joined ORNL. He has been involved in a variety of research and development projects focusing on mixed-signal CMOS integrated circuits and integrated sensors. He is currently involved with sensors and systems for in vivo physiological monitoring, silicon-integrated optical sensors for low-level light detection, and high-temperature CMOS integrated circuits for well-logging applications. His research interests include integrated sensor and readout methods for in vivo measurement, wireless data communications, and integrated photo-spectrometers. Ericson is a member of Tau Beta Pi, Eta Kappa Nu, and IEEE, and has authored or co-authored more than 25 publications. He holds one patent for wide-temperature logarithmic current measurement and has three patents pending.
| |
|
