Q&A: James R. Biard, GaAs Infrared LED Inventor

Your biography mentions your employment at TI began in 1957, three days after you received your PhD from Texas A&M. This was one year prior to the hiring of Jack Kilby, the inventor of the integrated circuit. Do you recall what the atmosphere was like at the TI plant in the summer of 1958 when he developed his first germanium IC?

It was a very exciting time. A lot of interesting things were happening. Transistors up until Kilby’s time had been mainly grown junction, alloy junction, or mesa devices made from either germanium or silicon. All these were discrete transistor chips mounted in headers of various kinds, which had three leads (emitter, base, and collector). These transistors were used with other components like resistors and capacitors for constructing useful electronic circuits normally on printed circuit boards. This gave a dramatic size reduction compared to earlier vacuum tube versions of similar circuits. However, as the circuits became more complex, the transistor circuits also became quite large.

Since Kilby had been recently hired, he had no vacation time. He worked while everyone else took mass vacation, which was the practice at TI at that time. During that mass vacation he came up with a unique insight into a technique that would dramatically reduce the size of electronic circuits by making the transistors, resistors, and capacitors all from the semiconductor. His first integrated circuit was a single stage amplifier made from germanium. It included a transistor, resistors, and a P-N junction diode used as a capacitor all made in the same germanium chip.  These circuit elements were interconnected by wire bonds. Based on his insight, TI filed a patent for the integrated circuit. However, the germanium integrated circuit never caught on as a product. The semiconductor industry was just beginning to develop silicon planar technology and about six months later Robert Noyce at Fairchild Semiconductor came up with a technique for interconnecting transistors, resistors, and P-N junction capacitors into a working circuit using aluminum metallization over a silicon dioxide layer formed on the surface of the planar silicon chip. This was the development that made Kilby’s original idea for the integrated circuit practical. Kilby and Noyce are often thought of as the co-inventors of the integrated circuit. Kilby’s basic idea came first, but Noyce made it work. Robert Noyce died in 1990 at age 62. Kilby received the Nobel Prize in Physics for the invention of the integrated circuit in 2000 and died in 2005 at age 81.

Jack Kilby was my boss for several years prior to 1969 when I worked for TI and he was instrumental in getting me started as a part-time Adjunct Professor of Electrical Engineering at Texas A&M.

One of your first projects at TI was working on a low drift dc amplifier with your former Texas A&M professor, Walter T. Matzen. This work ultimately led to one of your first patents, the DC differential amplifier. Can you mention a few details about how this patent came about and its significance?

Walter Matzen was my major professor at A&M and he also served on my graduate committee. We both went to work at TI on June 3, 1957, and continued to work together for some time. My education in semiconductors began with my employment at TI. The main emphasis of our first project was on describing the characteristics of the bipolar transistor that affect offset and drift. There are two main parameters: 1) The way the emitter-base voltage depends on collector current and temperature, and 2) The dependence of the base current on collector current and temperature. After describing these dependencies, we went on to develop a low offset transistorized dc differential amplifier with low drift over temperature. The amplifier was implemented using carefully matched pairs of discrete grown junction silicon transistors.

We filed our patent in March of 1958. We then published an article in Electronics magazine in January of 1959. Honestly, I don't remember if our transistor dc amplifier was the first or just one of the first. Things were happening pretty fast back then. However, the drift equivalent circuit and method of analysis developed on this program formed the basis for much of the work that followed at TI on differential transistor amplifiers. With one of the electronics technicians, we designed and constructed one of the first completely automatic transistor testing facilities. It was called the Sequential Mechanism for Automatic Recording and Testing (SMART). SMART-1 could perform up to 18 ac and dc tests on a transistor and record the results of these tests directly on an IBM card without human intervention. This work resulted in a commercial line of SMART machines, which were manufactured and sold by the Industrial Instrumentation Division at TI. The commercial SMART machines were used primarily for incoming lot acceptance and life testing where it was necessary to maintain unit identification with recorded values of measured parameters.

In September 1961, you and your LED co-inventor Gary Pittman were working on a project to develop GaAs varactor diodes when you first observed light emission. Can you explain what you two were investigating when you built the test structure on the GaAs substrate?

We initially thought we were seeing excess current in the valley region of the tunnel diodes we were building and felt this could have an impact on the varactor diodes. I came up with a theory to describe the excess current as a radiative recombination generation in which an emitted photon was stimulating other centers to discharge. Based on this theory, I thought it might be possible to use the stimulated emission to make a GaAs laser diode. We built the structure with the tunnel diode on a semi-insulating GaAs substrate to test the theory and found that the excess current was actually occurring in the forward bias region of the diode curve. When we forward biased the tunnel diode, it changed the conductivity of the semi-insulating GaAs substrate. This made us think that the GaAs diodes might be emitting photons. However, the light being emitted was infrared, which is not visible to the human eye. About that time a Japanese firm brought an infrared microscope to TI for the Quality Control Department to use for inspecting silicon wafers. When Gary and I found out it was in the building, we took some GaAs varactor diodes and tunnel diodes with clip leads and batteries to the Quality Control Department. We found that when the varactor and tunnel diodes were forward biased, they all emitted infrared light that could be seen on the infrared microscope. We filed our patent about a year later. The Central Research group at TI was of the opinion that semiconductors did not emit light, but we proved them wrong.

In October 1962, TI released the SNX-100 GaAs LED based on your design. This was the first commercial LED product. In October 1963, TI released the first hemispherical LED product, the SNX-110 GaAs LED. What were the reasons you and your team switched from a flat square crystal to a dome-shaped one?

When the LED chip is rectangular, only about 4% of the internally emitted light is at a small enough angle to escape the top flat surface of the chip. This is because the index of refraction of GaAs is so high (about 3.6), which gives an angle for total internal reflection of about 16.1⁰. If you could get all of the upward directed light out without absorption, it would increase the output by a factor of 25. When the chip shape is a hemispherical dome with the junction area limited to the central portion of the flat face of the dome, all of the unabsorbed photons emitted at the junction strike the exit surface of the dome at an angle less than the critical angle. That happens when the radius of the dome is greater than the radius of the junction by a ratio equal to or greater than the index of refraction of the GaAs. This increases the light output by a significant amount by eliminating total internal reflection, however, the increase is not by a factor of 25 due to the absorption that happens in the GaAs in the larger radius of the dome. The N-type GaAs between the junction and the exit surface is not completely transparent to the wavelength emitted at the PN junction. The greater path length for the photons in the dome LED structure produces more absorption than the short path length in the rectangular LED chip. However, the hemispherical dome LED resulted in a significant increase in total power output for a given forward bias current compared to the rectangular LED.

In the 1960s, one of the main contributors to integrated circuit research was the U.S. Air Force. Was it a similar situation for the LED?

Yes. Some of the development work we did on LEDs was funded by the U.S. Air Force at Wright Patterson AFB.

In December 1962, Nick Holonyak, Jr., of General Electric reported visible red light emission from an LED. Have you ever met him? If so, did you discuss with him your own work on the LED?

I know Nick Holonyak and have corresponded with him over the years. Apparently my engineering notebook entry for the GaAs tunnel diode on the semi-insulating substrate preceded his group’s earliest work at GE on visible red LEDs in gallium-arsenide-phosphide (GaAsP). When we heard about the work they were doing, we immediately started a program at TI on red LEDs. I once sent Nick a copy of my PowerPoint file that documents the invention of the LED. I think he generally accepts that my IR LED patent preceded his red LED.

Did you know Albert Einstein’s grandson, Bernhard Caesar Einstein, when he worked at TI in the 1960s?

Not that I remember; however, I’ve read we had the same boss, Dr. Richard Petritz. When Gary Pittman and I were in SRDL (Semiconductor Research and Development Laboratory), Dr. Petritz was the director. He supervised over 300 technical personnel. I think Dr. Petritz was in TI’s Central Research division before he transferred over to the Semiconductor Division to head up the SRDL. If Bernhard worked in the Central Research division, that might explain why I didn’t know him. Apparently he worked with Hans Jurg Stocker. I do remember that name. Stocker worked on infrared detectors and indium-antimonide (InSb) diodes. At the time, Gary and I worked almost exclusively on gallium-arsenide and gallium-arsenide-phosphide.

In 1964, you filed a patent for Schottky bipolar TTL logic circuits. In 1969, Intel released their first product, which used Schottky TTL. Can you explain how your patent came about? Was Intel was ever made aware of the work you did 5 years prior?

The first Schottky clamped transistors I developed were used in amplifiers I had designed to work with photodiodes in optical receivers. Without the Schottky diode clamp, a large optical input would cause the input bipolar transistor to saturate. The resulting minority carrier storage would kill the bandwidth of the optical receiver. The engineer in the office next to mine was having a similar saturation problem with bipolar transistor TTL logic circuits. Since the transistors of a standard TTL gate are saturated switches, minority carrier storage time in each junction limited the switching speed of the device. I used what I learned with my optical receiver amplifiers and applied it to the bipolar logic circuits. In general, the Schottky clamped transistor has utility in more than just logic circuits; however, the logic circuits far outstripped any other use in terms of volume.

When I filed the patent for the Schottky clamped logic circuit, it used aluminum for the metal contact in the Schottky diode. Aluminum did not make high-yield Schottky diodes on silicon. With aluminum you had to be very careful not to let it get too hot during the manufacturing process; otherwise, some of it would diffuse into the silicon and form a PN junction. When done properly, aluminum could make a good Schottky diode. The development of platinum silicide is what allowed for a reliable manufacturing process. Due to the way my patent was written, the use of different Schottky diodes was still covered.

In August 1969, my patent issued. That same year, Intel released its first product, the i3101 Schottky TTL bipolar 64-bit static random-access memory (SRAM). In 1971, TI introduced the 74S Series TTL logic family using Schottky diodes for high-speed applications. Low-power Schottky, designated 7400 LS, came along five years later and caused quite a revolution in logic circuits. Tim B. Smith was the one who led the effort to develop Low Power Schottky circuits for logic applications. TI used my patent to force Intel out of the Schottky TTL product line that they had started. As a result, Intel went into CMOS integrated circuit products. A colleague of mine, Fred Strieter, jokingly says this means I am partially responsible for Intel’s success.

You were later honored at Texas Instruments, correct?

Yes. I received the Patrick E. Haggerty Innovation award in 1985 for my invention of Schottky Clamped TTL Logic. Pat Haggerty was the president of TI when I worked there, but his award came about a number of years after I left TI in 1969 to join Spectronics. Apparently Schottky TTL made TI a lot of money.

Can you briefly explain your role at Spectronics?  Did you continue working on the improvement of LEDs?

At Spectronics I served as Vice President of Research. I continued to work on Optoelectronics including the design and development of silicon photodiodes, phototransistors, photodarlington devices, and GaAs light-emitting diodes. I also designed and developed most of the special test equipment used for component calibration and evaluation. The LED progressed into the Vertical Cavity Surface Emitting Laser (VCSEL) and we soon came up with the VCSEL diode. The VCSEL has become the work horse light-emitter for fiber optic interconnect in the large server farms that handle all of our e-mail and Internet traffic at places like Google. Fiber optic interconnect is what I continued to work on after leaving TI. Honeywell still makes and sells IR LED products that are used in optical isolators, remote control devices, and automotive anti-collision systems. IR LED products are still around in 2016, 54 years after we released the SNX-100 LED in 1962.

In the summer of 2015, you officially retired after 58 years in the industry. Have you been involved in any projects since that time?

I have had a consulting job for a start-up company, but I am not involved actively with anything else at the present time. I am enjoying my retirement.

In 2014, it was announced the Nobel Prize in Physics was being awarded to the inventors of the blue LED. What was your reaction when you heard this news?

I was happy for them. When Gary and I filed the GaAs IR LED patent, the policy at TI at the time was to pay $1 for each patent filed. Things have certainly changed a lot over the years.

There are many people who think you didn’t get the recognition you deserved for inventing the LED. Have you had many inquiries from people writing about the LED in a historical context?

In 1978, Tom Hyltin published a book titled The Digital Electronic Watch in which he cited the significance of our LED to the development of the digital watch. I worked pretty closely with him when I was at TI. He was an engineering project manager who had worked on the first solid state radar system. After leaving TI, he went on to form his own company manufacturing liquid crystal displays. In 1998, Carlene Stephens from the Smithsonian Lemelson Center used Tom's book as the basis for a web feature called The Quartz Watch. Within the last few years the Lemelson Center redesigned their website and took that web feature down to do some updates. As far as I know, they're still in the process of reinstating it.

In 2012, Mike Whelan from the Edison Tech Center created a timeline on their website about the LED. He lists Gary and me on the timeline and includes a few paragraphs about our LED. In November of last year, Mike shared a paper I co-wrote with my grandson, Thomas Okon, about the development of the LED at TI in the 1960s. Thomas became very familiar with the story over the years and used that knowledge to create my Wikipedia biography. In 2013, Southern Methodist University (SMU) honored me with the degree of Doctor of Science, honoris causa, for my contributions in the field of optoelectronics including the invention of the LED. Kay Bailey Hutchison, the former Texas senator, was also awarded a degree honoris causa at the ceremony. That same year, Paris High School inducted me as a Distinguished Graduate. Twenty years prior (in 1993), I was recognized as a Distinguished Graduate by Paris Junior College. Paris, Texas, is where I grew up and attended school prior to transferring to Texas A&M University. Texas A&M recognized me as a Distinguished Alumnus in 1986.

In 2014, a writer named Deana Totzke from Texas A&M University contacted me following the announcement of the inventors of the blue LED receiving the Nobel Prize in Physics. She put together a good article about my involvement with the LED. She was of the opinion that I’d be a logical candidate for receiving the Nobel Prize. I was flattered by the suggestion, however, I don’t know if the IR LED will ever get the credit the visible LED gets. I think a lot of people in the media audience are either not aware of how the IR LEDs work or they think because you cannot see the IR light with the naked eye that means it’s not a light-emitting diode. I’ve had a good, long career in the semiconductor industry and I’m happy with what I’ve been able to achieve. I don’t plan on holding my breath waiting on a Nobel Prize.

Do you consider yourself to be the inventor of the IR LED, the GaAs IR LED, or the LED?

I suppose my preference is inventor of the GaAs IR LED, however, answering that question is a bit difficult. We filed our patent (U.S. 3,293,513) in Aug. of 1962, however, it didn’t issue until Dec. 20, 1966. It took four years to issue because it was in interference in the U.S. patent office with submissions from GE, RCA, and Lincoln Labs at MIT. When the patent office checked everyone's engineering notebooks, they found that Gary Pittman and I had the earliest notebook entries, so we got the patent. Since then, the rules have changed in the patent office. Now it's the first person to file that gets the patent; not the earliest notebook entry. Things like that just seem to happen when the technology advances enough to support the result.   

Immediately after Gary and I filed our patent, we began a project to manufacture infrared diodes. In October 1962, we announced the first commercial LED product, the SNX-100 GaAs IR LED. That same month, Nick Holonyak of GE submitted his report on red light emission from a GaAsP LED. In October 1963, we released the first commercial hemispherical LED, the SNX-110. In May of 1965, Nick filed his patent (U.S. 3,249,473), which issued in May of 1966.

I didn’t find out until recently that Rubin Braunstein and Egon Loebner of RCA had filed a patent for a germanium LED in 1958. Their patent (U.S. 3,102,201) issued in August 1963, a year after we had filed ours. I didn’t know Rubin Braunstein. I did know Egon Loebner when I worked at TI, however, I had never heard of anyone making an LED from germanium. Germanium has a direct band gap (~0.85eV), which means it should have infrared radiative recombination at about 1.4um wavelength. As I understand it, they were driving the germanium diode with a constant dc current and it had radiative recombination current in the bulk (area) of the junction and non-radiative recombination at the back surface of the chip opposite the junction. As far as I know, RCA never tried to make a commercial component out of the patent. I imagine it would have been very slow due to the long carrier lifetime in the bulk of the germanium chip, since they had to diffuse electrons a distance of 7.0 mils (0.007 inches) across the thickness of the germanium chip to get them to the non-radiative recombination on the back side of the chip.

I’ve been told that we were all preceded by the Russian scientist Oleg Losev who died in 1942. Apparently he saw light emission from point contact diodes in carborundum. I think he really got overlooked.

In July 1966, you filed a patent with Bob Crawford for a ‘Binary Decoder.’ Apparently this device was the first time a Read Only Memory had been made using MOS transistors. Can you explain how this patent came about and its importance to early calculators?

In 1964, I was placed in charge of both the Optoelectronic branch and the MOS branch in SRDL at TI. The Opto branch had developed a monolithic visible LED element consisting of a 3x5 array of red LEDs capable of displaying the numbers 0-9. The device was lacking a means of driving the array, so Bob Crawford (from the MOS branch) and I developed a P-channel MOS decoder circuit for converting binary numbers into the decimal number display. The MOS circuit worked on the first pass and was implemented into a simulated cockpit altimeter. In March of that year, TI displayed the altimeter in a booth at the New York IEEE show and convention. When we filed our patent in July 1966, we referred to it as a ‘Binary Decoder.’ We didn’t know at the time that the device was an MOS ROM.

The first handheld digital calculator that Jack Kilby co-invented with Jerry Merryman and James Van Tassel at TI in December 1966 used bipolar transistors and integrated circuits. I suspect the arithmetic operations in that first calculator were hardwired in the circuit and not stored in memory. It wasn’t until later that MOS ROM was used for the storage of calculators’ fixed programs that do the various arithmetic operations. The MOS transistors were preferred to the bipolar transistors because they consumed less power and were smaller in size.

It was in the early 1970s that TI released their first line of calculators. By the early 1980s, most portable calculators used MOS technology; however, they were made by an N-channel silicon gate process. In 1986, TI filed a complaint with the International Trade Commission against multiple Japanese firms who had manufactured chips which potentially infringed our MOS ROM patent. I was working for Honeywell at the time and TI flew me out to Washington D.C. in order to testify. It was determined that the Japanese firms had not violated our patent because there were enough differences between the implementation of our MOS ROM circuits and theirs. By 1987, our patent had expired. Patents are only given protection for 17 years.

Can you briefly discuss your involvement (if any) at TI with Jerry Merryman and/or James Van Tassel?

I worked beside both of them every day; however, I did not work directly with them on the digital handheld calculator. Jerry Merryman is one of the smartest, most inventive engineers I ever worked with. He went to school at Texas A&M, but he never graduated. He was hired at TI on the recommendation of Walter Matzen. Walt had Jerry in some of his classes. I share two patents with Jerry. One of them is for an opto thermal audio amplifier, which we filed in December 1966. That was the same month his team developed their first working model of the digital calculator.

Is there any advice you’d like to give to future generations of researchers in the semiconductor industry?

Keep your eyes open and don’t be afraid to adopt what seems to be a new approach. You never know when you are going to discover something new and interesting. Keeping my eyes open has resulted in my 72 U.S. patents and numerous published papers during my career.

Looking for parts? Go to SourceESB.