On pg. 9-2 (in the chapter titled "Encapsulation", it says: "PNP and NNP germanium transistors are very sensitive to changes in water-vapor pressure. Reverse current and current amplification will improve or deteriorate, depending upon the previous surface conditions. Very generally, however, it seems that NPN transistors are more sensitive to moisture when being encapsulated than PNP transistors.
Thus, it once more must be stressed that the purpose of the encapsulation is to maintain the original surface condition the designer has created. Realizing however, that even if we seal our device hermetically without changing the surface structure, there must exist a temperature-dependent sorption-desorption equilibrium on the surface. The problem of maintaing the electrical parameters of our device constant under all conditions of storage or use at various ambient temperatures seems to be a very difficult one. For very critical applications three obvious solutions now exist to help us with this problem:
(1) Heat the device on a pump and backfill with very dry, inert atmosphere after pumping.
(2) Operate the device at as low a junction temperature as possible by the use of a lower rating for maximum collector dissipation and ambient temperature.
(3) Try to put a substance in the can which helps to maintain a temperature-independent water-vapor pressure. Several manufacturers have tried various inorganic compounds for this purpose, such as calcium sulfate, molecular sieve, and many others. None of these admixtures unilt now have solved the problem entirely, but some improvement certainly has been obtained. Much depends for each recipe upon the previous surface treatment, however."

Googling, I found a 1962 article that tested the efficacy of barium oxide mixed with calcium sulfate and other desiccants in germanium transistors (calcium sulfate + barium oxide being the best). It also posits an explanation as to why water vapor messes with Ge transistor behavior.^*

The powder I've seen in the IBM transistors indeed looks like calcium sulfate (gypsum); it's certainly not a "molecular sieve".
https://en.wikipedia.org/wiki/Calcium_sulfate
"Calcium sulfate (or calcium sulphate) is the inorganic compound with the formula CaSO4 and related hydrates. In the form of γ-anhydrite (the anhydrous form), it is used as a desiccant. One particular hydrate is better known as plaster of Paris, and another occurs naturally as the mineral gypsum. It has many uses in industry. All forms are white solids that are poorly soluble in water.[5] Calcium sulfate causes permanent hardness in water."

Calcium sulfate hemihydrate

If our IBM transistors also contain barium oxide, look out - it's a skin irritant. And don't eat it:
https://en.wikipedia.org/wiki/Barium_oxide
"Barium oxide is an irritant. If it contacts the skin or the eyes or is inhaled it causes pain and redness. However, it is more dangerous when ingested. It can cause nausea and diarrhea, muscle paralysis, cardiac arrhythmia, and can cause death. If ingested, medical attention should be sought immediately."

Best Regards,

- Robert

^* The article: "Selection of Germanium Transistor Parameters by Control of Moisture at Low Levels within the Device Encapsulation”, Robert J. Gnaedinger Jr., Steward S. Flaschen, Marie A. Hall and Edward J. Richez, Journal of The Electrochemical Society, Volume 109, Number 7
https://iopscience.iop.org/article/10.1149/1.2425502/meta

I went ahead and purchased it, ... excerpts below should be OK.

"These experiments have indicated that moisture vapor in air within the sealed, device encapsulations has a pronounced effect on the surface re- combination velocity of germanium transistors at extremely low relative humidities, relative humidities as low as 10-8% (0.2 ppm)."

"It is well-known that the chemical composition of the gas phase adjacent to a semiconductor surface plays an important role in determining the electrical properties of that surface (1-3). Water vapor is perhaps the most striking example of this fact. High relative humidities produce severe and erratic current leakage paths across the surface of silicon and germanium crystals; changes in the humidity at moderate levels produce pronounced changes in the surface conductivity and surface recombination velocity. In the manufacture of semiconductor devices it was this high sensitivity to changes in the moisture level which caused the introduction of hermetic sealing of devices. Yet, although this hermetic sealing generally removed the erratic behaviors associated with gross moisture, there remained day-to-day, seasonal, and aging variations in the parameters of production devices. In order to reduce these variations, it was found necessary to insert moisture adsorbing materials into the hermetically sealed device encapsulations. This technique yielded a more reproducible product, but it frequently accomplished this at the expense of large shifts in the levels of some of the device parameters.
It is the purpose of this paper to present the results of new experimental studies of the effects of moisture on the parameters of germanium transistors. These results indicate that the presence of moisture vapor at very low levels continues to exert a strong influence on semiconductor surface properties, a conclusion suggested by previous work (4-7), but not experimentally verified. In addition, these results show that the rather disruptive role commonly associated with moisture in the manufacture of semiconductor devices can be converted into a constructive one by selection and control of the moisture level. Lastly, these results emphasize the necessity of being able to make a quantitative, rather than qualitative, statement of the dryness level in the adjacent gaseous ambient in experimental studies of semiconductor surfaces and semiconductor devices."

“The data presented above indicate that certain parameters of germanium transistors depend very sensitively and reproducibly on moisture partial pressures at levels from about 80 ppm to below 1 ppm in air. The parameters thus affected are those which are strongly influenced by surface recombination velocity (or surface generation rate). The simplest interpretation to be made of these observations is that even at these very low partial pressures, sufficient water is adsorbed onto the germanium (or germanium oxide) surface to affect the surface properties appreciably. According to the work of Law (14), monolayer adsorption of water onto a germanium surface does not become complete until a relative humidity of about 10% is reached (this relative humidity in air at 1 atm pressure at 25~ corresponds to about 2000 ppm of H2O). Consequently, the present results suggest that water plays perhaps its dominant role in affecting surface recombination velocity in the region below one monolayer, a conclusion that is consistent with the work of Law and Meigs (4).
The detailed mechanism of how adsorbed water achieves this effect is not clear. However, since water is known to induce n-type channels (15) on p-type germanium, it has been postulated that absorbed water molecules behave like slow, hole-trapping states. This same postulate has been used to explain observations of the effect of Brattain-Bardeen ambient cycling on the surface recombination velocity of germanium.”

p.s. Inside the front cover of my copy of Hunter's tome reads: "Product Development Library, IBM Poughkeepsie, May 14, 1962."

The IBM 083, NPN, alloy junction (equiv 2N1302), used in most of the SMS logic cards. (Thanks to IBM Almaden machine shop for cleanly decapitating.) Words and music on image & below by Rick Dill ;-)) The round germanium die worked for the alloy transistor because it was uniformly doped and didn't matter which side was used for emitter and base. When the diffused base transistors were introduced, the same machine could turn them out with lots of adjustments, but no major re-engineering. The die for the diffused base transistors had a unique shape without mirror symmetry so that the parts could be fed proper side up by the Syntron bowl feeders. The parts were sonically driven up a long spiral track to the point where they were dispensed to the handler that put the die into the fixture. Notches and other features on the track would derail any die that was upside down of any other anomaly such as a couple of emitter or collector spheres which were stuck together and not free.
View of the emitter side. Note the alignment tab, the closest tab on the case base, to aid machine testing and also machine insertion into circuit cards. Very common for this "TO-5" very popular package. TO-5
The IBM 108, PNP, used to drive a 1403 print hammer. Rick Dill adds "This transistor has a "ring" shaped emitter with a base contact inside and outside. In power transistors, most of the current flows right at the edge of the emitter due to base resistance drop as one gets farther from the edge. The ring gives lots of edge with relatively less inactive area. "The collector is probably a circular alloy "dot" which is direct soldered to the can for heat sink purposes. "I am not sure whether this transistor was IBM design of whether it came from Delco or Motorola who were into power transistors for car radios and other potential automotive applications."

From: Rick Dill < rdill@cyburban.com > 
09/03/2008 10:12 AM 
To Robert B Garner/Almaden/IBM@IBMUS  
cc: ...  
Subject Alloy Transistors 

  

Robert,

The early 1950's were an interesting time.  In the summer of 1951, Bell 
Labs had a "summer school" for a fair sized group of university 
professors.  The professor who taught optics attended and had lectures 
from the likes of Shockley and Bardeen as well as some laboratory hands 
on time.  They published a paper with everyone as an author in which 
they confirmed the Einstein Relationship between drift of electrons and 
holes under electric field and thermal diffusion.  They returned to 
campus with a small kit of experiments which a fellow student, Jim 
Boyden and I latched onto and used both for recreation and every 
opportunity for a project.  We were sophomores when we got our hands on 
this. 

It included a germanium alloy diode for measuring IV curves, a chip of 
germanium with rhodium plating on the back (which we had to learn how to 
replace after etching it off) to be used for making point contact 
transistors, and a bar of germanium for the Shockley Hall measurement of 
drift and diffusion of electrons.  We fabricated our own point contacts 
and crude manipulators to put them down near where we wanted them. 

In the summer of 1954. fresh with a B.S. in physics I had a job at IBM 
in Poughkeepsie.  It was my first industrial research experience and a 
wonderful experience.  Joe Logue was the manager of the small group 
which included Hannon York, the engineer who invented the current switch 
circuit (non-saturating), Bob Henle who was a really good circuit guy 
and lead the company a few years later to abandon core memories and go 
to semiconductor memories. 

At the time, the group was just off of the CRT based electrostatic 
memories used in the 701 and the later mica target CRT-like memories 
used in the following computers up to the introduction of magnetic cores 
which were just coming visible in 1954.

My assignment was really a gift.  Joe wanted higher function circuits 
with the example being the neon ring counter.  As he has told you, with 
no real hands on experience I was able to postulate that we might be 
able to produce such a thing with a double-based diode (unijunction 
transistor) which had multiple emitters and with the help of the small 
group in the pickle factory actually get a four-state device 
prototyped.  Dick Rutz and John Marinace were the key people in that 
group and I eventually ended up working for in Rutz's group.

I also postulated an adder circuit based upon the Shockley Hall 
structure, but with deflectors to steer the electron cloud sideways to 
multiple collectors.  We didn't make this, but it did show up a decade 
later as an academic achievement.

While Bell Labs was stuck on point contact transistors (and even made a 
computer out of them), Joe Logue has recognized the superiority of 
junction transistors.  From a circuit standpoint, the ones available had 
too much base resistance to work well in digital circuits, although this 
didn't hamper them for communications.  He tried to encourage the 
vendors to make transistors with low base resistance to little avail 
since computers were not important to the electronics world at that time.

The task got handed to Research where on the fabrication side the team 
of Rutz and Marinace and their technicians made alloy transistors with a 
small alloy emitter surrounded by a circular base contact, and a larger 
collector junction on the back side of the die.  Contrary to what most 
people believed, alloying was a well determined metallurgical process 
when done right.

To do it right, you needed the crystal to have a <111> orientation.  
This is the slow dissolving direction when subject to dissolution by 
molten metal, so the sides of the dissolved region were angled along 
<111> planes and the bottom flat.  The depth depended on the volume of 
the alloy "dot" and the temperature.  On cooling, the germanium first 
cleanly regrew precipitated from the melt and was doped by the materials 
in the dot (indium gallium for PNP transistors and tin antimony for 
NPN).  The collector on the backside was simply larger and etched more 
deeply into the germanium chip. 

The design was IBM's, but we went to TI and contracted with them to 
manufacture germanium transistors for us.  We were allowed only to make 
10% of what we needed, which gave us room for special applications such 
as core drivers or advanced devices before releasing them to TI.

In spite of Joe Logue trying to get me to stay and do graduate work in 
the Syracuse MS program, I went back to Carnegie Tech and moved from 
physics to EE.  18 months after that Bob Henle visited campus following 
up on a summer job one of the young faculty had a year after me.  IBM 
wrote three separate contracts.  One supported my research with the 
requirement that I come to Poughkeepsie roughly monthly to report on my 
progress.  The other two supported Dale Critchlow, the young faculty 
member, and Bob Dennard, one of his students.  Both Dale and Bob were 
working in magnetics at the time.  I was the first to join IBM in 
February 1958 and Critchlow and Dennard joined the following summer.  We 
were all in the same group trying to do circuits with multi-hole 
magnetics and transistors.  I left that project in summer of 1958 to go 
to Poughkeepsie and work for Dick Rutz with my initial assignment being 
to duplicate Esaki's work in Japan on tunnel diodes. 

In 1955, I worked at RCA Labs for the summer on silicon diodes and the 
observation that they did not behave according to Shockley's theory.  
There I met Herb Kroemer who is credited (among other things) as the 
father of the "drift transistor".  Kroemer postulated (as a theorist) 
that a graded impurity doping in the base region would provide a field 
that greatly speed up transistors.  It was only after I got to IBM that 
I read a patent of Lloyd Hunter which is the patent on the structure, so 
IBM has at least as much claim as RCA.  By diffusing (probably 
phosphorus) into the germanium blanks, the IBM design became much faster.

In 1958 on returning to IBM, I found that the development group had 
built a fully automated factory for producing alloy transistors.  It 
used syntron sonic driven bowls to feed to germanium die, alloy spheres 
for emitter and collector, stub leads soldered to the spheres during 
alloying, and the dished base contact washer.  These were fed into high 
purity carbon fixtures, one for each transistor.  Once the assembly was 
together it went through a hydrogen furnace, after which the transistor 
was extracted, etched, washed, dried, tested, and then assembled onto a 
header.  The carbon fixture was sent back to be re-used.  This 
room-sized automated factory could product 40 million transistors a 
year, which was more than IBM needed.  We shipped the factory to TI with 
the stipulation that they could use it only for IBM production for a 
stated number of years.

When the diffused base transistor came in, it was only a small 
modification to the line to get the die right-side-up so that the 
diffused surface was facing the emitter.

About 1964 I visited TI and saw multiples of this production facility 
running full tilt to make germanium transistors for the world.

Silicon really came in with the system 360.  The group that met to work 
out what would be done was the Compact Committee.  I was a research rep 
on that team.  We should talk about that sometime.  Bill Harding was one 
of the key people and I believe that he is still around in Southern CA.  
What became solid logic technology (SLT) came from his declaration that 
he could produce individual transistors very cheaply and solder them 
directly to substrates.  This was all an act of faith, but it came to 
fruition and significantly delayed IBM's serious entry into integrated 
circuits for logic.

When we did get into integrated circuits for logic, electron beam 
lithography played an immense role in giving IBM a competitive 
advantage.  Hans Pfeiffer was the leader in this technology and is now 
in Carmel Valley.

This is just a little background for ou[r] Thursday discussion.

Rick

 

I met with Rick for 3 hours today.
He was a PhD summer intern at IBM in 1954, 
    working on multi-emitter Germanium devices.
Fascinating conversation on all what was going on a IBM at the time (and later).
He thinks he can  explain our loopy I-V curves 
    (water contamination causing the usual culprit - surface states.)


Finally I have someone to talk to about the physics and manufacturing 
    of our alloy junction transistors. (although I have some old books on topic now too.)
He also talked about how hard it was to make the 
    high-voltage core/hammer driver power Ge transistors...


- Robert
 

>> Have you seen early Philco transistors -- "look a bit like bullets
>>   - about 1/2 inch long, and maybe 3/16 inch diameter, 
>>      metal with a rounded top" -- on any SMS cards?

> A Google image search of 2N501 will show somw Philcos.
>
>> (I can send higher-res versions of these pics if you need.)
> Yes, please. On early transistors, sometimes the distinguishing marks
> for each manufacturer are very subtle.
>
> Thanks!
>
> --
> Will

The prototypes for the 1401 were undoubtedly made by two technicians who worked for Dick Rutz. Rutz did most of the exploratory device fabrication. The technicians were Ralph McGibbon and Earle Harden. They were masters of the build-by-hand exploratory germanium transistors.

Rutz pretty well demonstrated that the "forming" of the collector on a point contact produced an np junction at the collector which was why point could have an alpha greater than one. He also made a two collector point contact transistor which could perform the full adder function. That was in the summer of 1954. He also got some prototypes together for me to demonstrate the concept for a decimal counter based upon a multi-emitter unijunction transistor. That was my first integrated circuit.

In 1959 I joined Rutz's group where I learned to build transistors by hand, but without the amazing steady skill or McGibbon and Harden. My project was to reproduce the Esaki Diode (later tunnel) simply by reading Esaki's publication. Before long Earle Harden worked for me and we got into 3-5 tunnel diodes, high efficiency GaAs LEDs and IBM's first injection lasers.

Earle, in addition to being an incredible technician was an amazing master mechanic. He was key to a mask camera I was making to get no placement errors in overlay. The David Mann steppers were open loop lead screws and .... I will come back to this in another post later about mask making for SLT and beyond. Earle also defined the head lapping process for the first thin film magnetic recording heads. That same process is still used by every manufacturer or disk drives today.

Yes, in development and manufacturing, many jobs went to women. That is still true today.

Threading magnetic cores was commonly done by women who handled the care and dealt with frustration better than men, although in the beginning when the process wasn't stable, it usually was done with men who were brought to a much better understanding of why and how things were done.

IBM built machines for automatic threading of core arrays. These got more complicated as cores shrunk to smaller sizes to get faster switching, lower currents, and more cores per unit area of the array. My understanding was that the automatic core wiring machines vs humans doing the wiring was pretty much a wash and that both process options were used.

On the 650, that was the machine I first used. A close friend, Karl Konnerth, introduced me to programming and the 650 by handing me something like a 20 page instruction manual for Wolontis emulator that ran on the 650. It took half the memory (1000 words) for itself and gave the user the other half. What the user got was a 3 address floating point image and yes, it did run slowly. Still I was trying to model unijunction transistors for my PhD thesis and I needed numerical help with the non-linear differential equations.

Carnegie Tech had just gotten their first computer, an IBM 650. It was for education use and not administration. Karl got into the machine from assembler on up. SOAP gave the assembler control of assigning memory locations so that when an operation completed, it was stored in a location coming up soon. Karl and I worked together over the years. I got him to come to IBM for a summer to see how quickly we could turn a semiconductor laser on or off. I felt that this device was likely good for optical communication and he concurred. The answer was simply "faster than we can measure"

Over the years Karl and I spent a lot of time working together. He was involved with a Fishkill engineer in an optical communication demo at Expo in 1966. He and I created measurement tools for semiconductor manufacturing in the 1970's that also resulted in IBM trying new product areas. One in technical areas where our thin film measurement tool was offered to the semiconductor industry and to the fabric industry for color matching and dye usage. This was IBM instruments, which didn't survive. Our Fishkill partners wanted to be like HP or other standard measurement tool things. Our tools were also way ahead of their time. The other area a few years later was the formation of an IBM product group in medicine with the key product a hospital cart sized unit that acquired ECG data and performed an amazing level of analysis of even very difficult traces. That also failed because once you have a trace, the physician attending can tell if it is "normal". If not normal it goes to a cardiologist. At the very highest level, it treads on the egos of the very top guys who wanted to be paid for the answer rather than be told it. Karl came with me to Fishkill when we were setting up the last two bipolar factories. He took a small team and created the first fully paperless semiconductor manufacturing line ever. Now in his mid 80's he is still playing with single board systems. He might be a person of interest for CHM.

enough for tonight. This will go out without proof reading. HELP.

Rick

Jack Ward, principal of the online Transistor Museum, has generously donated many important historic transistors to the CHM collection. He is currently working on a new collection that will include early IBM devices, including types used in the 1401. Attached you will find a draft copy of the text that will accompany the donation. Please forward the document to any and all members of the 1401 team who may be interested, If you have any comments or corrections that you wish to send to Jack please contact him at the address above.

David Laws

TransistorMuseumIBM-GermaniumComputerTransistorHistoryDraftFeb2015-2.pdf

From: Robert Garner
To: Jack Ward transistormuseum Feb 22, 2015 12:14 am

Subject: Re: IBM Germanium Transistor History and Historic Devices Jack, Thanks for your follow-up, especially on your 077 PNPN germanium thyratron transistor. It would be interesting to characterize. And I wasn't aware of Hunter's patent.* I haven't taken a part one of the tall case types, but would be willing to donate one of these if you have any interest in case dis-assembly and photos, as you did with the TO-5 case. Yes, it would be a pleasure to open one up (if there's a "surplus"). Thanks very much for the 1960 Electronics magazine article on the IBM automated assembly system. I have looked for that for a while, and this is very helpful. If you need a better quality scan, I could secure one at the San Jose (State Univ) library which has the period Electronics magazines in its reference stacks. (I was able to scan its paper on the notched disk storage prototype developed at the NBS in the mid 1950s, studied by the IBM RAMAC developers in San Jose.) Speaking of characterization, are you aware of (or have observed) the loopy I-V germanium transistor curves I mentioned/showed below? Here an interesting movie of its tricks (taken by Bill Newman several years ago), with variable trace rates (including low frequencies): http://www.youtube.com/watch?v=9lAvLXQEv2c Cheers, - Robert * I have a copy of both the 1st (1956) and 2nd edition (1962) of the Handbook of Semiconductor Electronics, edited by Hunter. The 1st edition contains a "point-contact transistors" section that Joe Logue, in the oral history session (that I recorded at his home), told me that Hunter forced him to include. Joe had strongly felt that, with the development of the higher-yielding alloy-junction transistor, point-contact transistors were becoming obsolete, but some folks, including Hunter, especially given the initial high cost of transistor manufacture, were taken that "only one point-contact transistor was required in order to construct a trigger circuit." Rick Dill first mentioned this (impossible sounding proposition) to me and I didn't believe it until I read the section and saw its I-V curve with a negative slope (akin to a tunnel diode.)

transistormuseum@aol.com wrote: at 5:32 PM

Hello again Robert, If you'll supply an address, I'll mail you some of the early 1950s/1960s IBM germanium transistors for your research, so you can characterize performance of a variety of these, as well disassemble a "tall-case" unit. Should be very interesting indeed. I haven't done much in terms of electrical measurement and performance characteristics, although I do have some experience with point contact transistors and other negative resistance devices such as tunnel diodes, unijunction transistors and Shockley diodes. I can provide some of those 1950s devices as well if you are interested. Let me know. Did you record or transcribe your interview with Joe Logue? I'd be very interested in reviewing this. best regards, Jack PS - I'm traveling on business so it will be couple of weeks before I can mail the transistors. PPS - I've attached another pdf, this one decribing some historical germanium transistors used in the first U.S. satellites (Explorer I and Vanguard). Not sure if you are interested, but thought the description of germanium transistors in space would be useful.

Robert Garner to Jack transistormuseum@aol.com, at 8:16 PM

Jack, Thanks in advance for sending some of the earliest IBM germanium transistors (btw, I also have some) and the "tall case" unit to disassemble/photograph. I'll pass on the menagerie of negative resistance devices (for now anyway ;-) Yes, I video recorded my interview of Joe Logue. He was a strong outspoken leader. I should create a DVD or upload it... Thanks for the article on the Vanguard transistors. Looking at the page 4 transistor proto-family tree, it looks like I should properly say that the germanium transistors used in our 1401 are "3rd generation", after the 1st point-contact and after the 2nd "early junction transistor"? (Up to now, I've said that the 1401 used the 2nd type of transistor ever developed.) On this note, I've wondered why the 1401 didn't use the faster diffused base transistor. Were they that much more expensive, or perhaps more finicky? Or why didn't IBM switch to a more reliable silicon transistor? It's interesting that the transistors used in the 1401 were manufactured late enough to supply 1401s manufactured up until 1972 (and likely beyond for SMS subsystem(s), such as the 2821 used to attach 1403-N1 chain printers to S/360s.) btw, I have a fairly thick reference book/catalog of all the extant transistor manufacturing firms in the late 1950s (w/ transistor types, specs, etc). Most of the firms I'd never heard of. If you don't have it, I'd be happy to loan it out or get a copy made . (I can supply its title tomorrow, as it's at work.) Rick Dill, cc'd, also gave me a copy of IBM's course materials for their first transistor educational classes, and early Bell Labs instructional manuals. Cheers, - Robert

from Rick Dill to Robert Garner, Feb 25, 2015

Hi all, I was there at IBM when this all was happening. As a summer student working for Joe Logue in 1954, I saw a lot of his operation, which was much more a circuit based engineering effort looking for what would really work for computer circuits. There I saw the Philco transistors with very thin controlled basewidths first in surface barrier electrodes, then the same with micro-alloy (much better emitters and collectors) and a few years with micro-alloy and diffused base. Logue had two key engineers, Bob Henle and Jim Walsh. All three became early IBM Fellows. The Philco transistors were perhaps a good model with a ring base, but they were mechanically very weak and a little vibration would fracture them. At the time they were the best for switching speed. The research group in 1954 included a few basic physicists brought to Poughkeepsie from Watson Lab on Columbia campus and a very small and practical group of engineers and chemists including engineer Dick Rutz and chemist John Marinace. In my assignment from Logue to make a functional electronic device like a counter, I got to know both of them in trying to prototype my unijunction counter device that summer and later worked closely with them. It was Rutz who built the prototype devices for what went into the 1401 and beyond. Both Rutz and Marinace were very productive inventors with Rutz concentrating on devices and Marinace on process. Using point contact-like transistor technology (pnpn .. hook collectors), Rutz invented a point conact-like full adder transistor, which I believe that I have given to the CHM. In 1954, there wasn't really any development activity. In spite of Logue's trying to convince me to not go to grad school, but stay with him and the Syracuse MS program, I went back to CMU and set up the first lab there that could make transistors with a capital investment of only $1000. That served to build unijunction devices that I used to write the first semiconductor device related thesis in the EE Dept. I was basically without an advisor, but was greatly helped when Henle came to CMU and brought 3 research contracts for myself, Dennard (DRAM) who was then working on magnetics, and Dale Critchlow who had a great career at IBM culminating with deep trench dram chips. My contract required me to go to Poughkeepsie monthly to report on my progress which got me some guidance and lots of access to 650 computers, often sitting with their covers off. My work, while not prominently published was one of the first uses of carrier flow in a semiconductor device. Yes, solving Poisson's equation, numerically. Yes it was 1-D, but it was the first real model for the unijunction transistor. I came back to IBM after the 3.5 years it took to get my PhD. For about six months I worked on circuits using transistors and multi-aperture cores for logic to get a change and then went back to devices working for Dick Rutz in1958 where I was assigned to build tunnel diodes following up on Esaki's paper. That was great fun starting with germanium and soon moving along with the field to 3/5 compounds like GaAs or GaSb which had direct band-gaps and much higher on to off ratios (and remarkably short life for GaAs which came back a few years later with lasers). Over those few years, IBM's research had moved to the 701 building which had just opened the summer of 1954 in the heat of summer without air conditioning. The 702 and 703 buildings had been erected and staffed with development engineers and scientists and a number of people from other companies with experience in early semiconductor manufacturing. I got to know many of these well because they were only a building away and, like myself, interested in making things more than basic physics which was the general tenor in the research lab. In the 703 building (if I remember the number right) were a whole bunch of key people. Bob Slade has been discussed. Mel Klein did a rather amazing job designing a germanium transistor capable of driving core memories. This had to be capable of one amp pulses with a hundred volt against the non-linear inductive swing of the cores. Those were a key component in IBM's early core memory, perhaps more decisive than the cores themselves. Paul Low, later VP of the components division was there and went off to Stanford for a PhD and back. Ed Davis, an underrgrad friend from college came after his PhD at Stanford. He engineered the SLT module technology including the technology for making precision resistors our of screenprinted palladium oxide resistors trimmed to precision resistance with a narrow beam sandblasting. He was also CD president, a little before Paul Low. There was Bill Harding, but I'll come to him later. Rutz did a good job with the transistor prototyping, more from just building things and paying attention than from design. The 1401 transistor was basically out of his lab. While Herb Kroemer who ended up at RCA is credited with the diffused base (drift) transistor, he is the physicist who first showed it mathematically. The patent for the structure was actually held by Lloyd Hunter who was Logue's boss in the summer of 54 and eventually moved to Stanford and University of Rochester. Years later Hunter taught me and my new bride how to fly gliders. I was a research rep on the "Compact Committee", a broad range of people looking at the technology for next generation computers. We had system designers logic designers, semiconductor development people, and only a couple of us from research. This went on endlessly with little progress until Bill Harding stood up and declared that he could make a transistor for a nickel in silicon. That turned the whole discussion and became IBM's SLT technology that evolved though discrete transistors to small integrated circuits to modular design of circuits from building blocks. Bipolar was the thrust for performance reasons .. either small fast systems for small computers or full ECL for top end speed like the model 90. Both saturating and ECL were available to the designer. I wrestled with Bill Harding some years later on process and manufacturing automation. Bill was very senior to me and what came out was an automated factory concept that got implemented in Fishkill for quick turnaround of IC's using electron beam writing for wiring logic chips to get unique chips out quickly. It also ended up somewhat in Burlington for memory chips. My end was developing automated process metrology for in-line measurement. Today's thin film measurement systems, both sprectral and ellipsometric tools derive from what we invented then. It was in that period that I put together a team to understand model and characterize lithography, which was probably the biggest contribution of my career. Answers to some questions. Why didn't the 1401 get diffused base transistors? It was simply a matter of timing. The 1401 was designed when the only transistors available were alloy junctions. Even with them, it took the invention of the current switch circuit (ECL) by Hannon Yourke in 1956 to come of age in the Stretch system later. The diffused base transistors had great advantage in communication initially where the transistor didn't saturate. Once the transistor saturates and charge builds in the base, it has little advantage over an alloy transistor. Why didn't IBM use silicon transistors for better reliability? Again it was a matter of timing. At my first visit to TI sometime around 1960, I was shown the making of silicon transistors. These were grown NPN or PNP junctions by changing doping as a crystal was pulled. One then cut rods out of the crystal and probed to find the base. A wire was alloyed to contact the base and contact made at the ends. Expensive and only where you need other charcteristics. Silicon Reliability? Until some time after the FET came into use for dynamic logic, and particular for early CMOS, silicon reliability was poorly understood. It was relatively easy to make stable pnp devices, but NPN was really hard. IBM stumbled upon the solution in their SLT which deliberately chose to hand with single devices with precision resistors than IC's. The use on NPN transistors of a phosphorus emitter diffused from a phosphorus doped glass layer deposited over the oxide insulator trapped the mobile sodium that was the bane of reliability in silicon. It wasn't done for reliability but rather to build the device that was wanted. Shortly after the announcement of the 360 a technician was etching the oxide before metallization and caused a shut-down of the line because of leaky transistors, which is how the role of the phosphorus was discovered at IBM. Why did IBM got to Texas Instruments for Transistors? This didn't get asked. Everyone in the early transistor business saw analog and communications circuits as the most important thing. Early on that didn't require low base resistance as did logic. This was a prime item for Logue's team. They saw that the IBM circuits were designed for low base resistance. IBM did not really want to get into semiconductor manufacturing early on and the start-up TI was the only company willing to build transistors designed for IBM's needs. The contract built IBM's designs or TI's improvements on them. When the automated transistor making machine was built, it was capable of far more than the 4 million transistors a year IBM had allotted to itself for memory drivers, etc., so the machine, once fully demonstrated, was shipped to Dallas. TI used it only for IBM for the first few years and then replicated it for the vastly growing need for germanium transistors. That machine easily adapted to the diffused base which was a small change from the alloy transistor in terms handling of the parts. The diffused die for the diffused base transistor did have to have a special shape so that it could be rejected from the syntron sonic feeder bowl if it was upside down. Me, I was always restless and moving on. Still worked more across the lines to development than most in research. After my work on tunnel diodes and some diversion about the von Neuman scheme for making transistors using majority logic of parametrons (L/C circuits with non-linear diode capacitors). I moved back to GaAs light emitting diodes which physicists generally were excited because of their high efficiency. I got tied up with Nathan et al building devices which the theoreticians thought might be induced to lase. They were right and we published the semiconductor laser the same day as the team at GE. No Bell Labs didn't invent the semiconductor laser even though they claimed so in their 100th anniversry of the telephone publicity. That was great fun because I figured out how to make laser mirrors by crystal cleaving so we quickly had hundreds for any experiment anyone wanted to do. GE was still hand polishing up to just before I published my fabrication technique. Bored with that and with silicon well established in the development lab, I went back to germanium one more time to see if, as a semiconductor with high mobility for both electrons and holes, we could beat silicon in speed. That was great fun because we ran a silicon project under Dave Dewitt and a germanium under my control, helping each other. Germanium won the switching speed race in 1967, but the physics of germanium suggested a much smaller advantage than my back of the pad start and we exited germanium (to our good credit).

From: From: transistormuseum@aol.com - Date: Thu, Feb 26, 2015 4:34 pm

... The attached "Semiconductor Ancestry Tree" might shed some light on this - this is found in a early 1960s WECO handbook "Solid State Devices".

On Thu, Feb 26, 2015 at 7:05 PM, < transistormuseum@aol.com > wrote:

Hi Rick, I very much enjoyed your detailed discussion about early IBM transistor development. I have read Bashe et al IBM's Early Computers, which also contains much information about the early IBM transistor work, and your email discussion really adds much to this. I have a couple of questions which I'd like to pose: Did you ever come across or test any of the point contact transistors hand-made at IBM in 1950 by Dickinson's group, as mentioned in Chapter 10 of the Bashe book? These things must be exceedingly rare, with likely few made and fewer still around. As the text reads in the Bashe book, these devices were hand-made using germanium crystals extracted by cracking open 1N48 GE diodes. I'm not sure, but I think this is how most companies got their first experience with making point contact transistors. 1950 is very early in the transistor timeline and I've searched for years for any further info on these IBM devices, so any info or recollections you might have would be much appreciated. I'm also very interested in your comments about TI. The relationship with IBM must have taken off very rapidly, and IBM must have been using really large quantities of TI transistors - I'm thinking this because I came across a TI award object (see the attached photo) states that 10,000,000 transistors had been made by TI for IBM by 1960. In my opinion, that's an amazing number! Any thoughts on this? Finally, the PNPN or hook transistor you describe must be one of the most unique early semiconductors produced by IBM. I say this because to my knowledge all other commercial PNPN devices were silicon, from such companies as GE, Westinghouse, Shockley and others. It seems that IBM was the lone producer of germanium PNPN transistors. I'm also interested in the early X-4 made by Rutz combined point contact and alloy junctions in a proto thyratron. I've attached a brief writeup on this, and hoping you can add some clarity and historical context. (See attached X-4 writeup). Was this the type of device used in the binary adder you donated to CHM? Thanks in advance for your comments. Regards, Jack [Ward]

From: Rick Dill < rickdill@gmail.com > - Date: Thu, Feb 26, 2015 9:18 pm

Hi Jack, Life is too busy now for me to go search for any artifacts I have, but I appreciate your collecting. Now to answer your questions on point contact vs pnpn. I haven't read any books,but am looking back in my foggy mind. Hand made point contact transistors are a good point to start. What got me into the field was a summer course Bell Labs had for university faculty in the summer of 1951. The professor who taught optics, Leito, but was better known for his ability as an estimator for costs of brick work, went. Among other things was a joint paper by all of the attendees confirming the correctness of a long ago relationship between drift and diffusion of charged particles by Einstein. The important thing to me and to Jim Boyden, my friend, was the kit he brought home with experiments to use in teaching. Jim and I latched onto it and did one experiment after another whenever we needed a project. One project was a germanium chip with a rhodium plated back onto which we were supposed to construct "microprobes" to place two point contacts onto the top surface. We found phosphor bronze wire and crudely pointed (I think with angled cutters) and I built some really crude things to get them close together on the surface. [In the process we etched the plating off the back and had to figure out how to replace it] So we made point contact transistor(s) and "formed" the collector with a capacitive discharge and they worked. That would have been about 1952. There were other experiments, like the I/V measurements of a grown p/n junction and a bar of germanium on which we could contact the two ends and inject carriers with one point and detect them with another to show the drift velocity and diffusion. It was part of the independent education of individuals! Jim was last in my view as the technical guy with Paul Allenon the X-Prize. He stayed in physics and I strayed into engineering and devices. When I got to IBM in 1954 and then back a few years later, there were various views on the point contact. I never heard of the the work of Dickerson, but wouldn't be surprised. IBM was very deeply into germanium, even in the vacuum tube era. While some built computers using vacuum tubes for the logic, IBM early focused on using diodes for logic and vacuum tubes for amplification to the next stage. That meant that the speed depended more on the recovery time of the diodes, which got excess carriers stored if they were on very long, much like saturated transistor circuits. IBM worked on fast recovery diodes with Hughes, who made them for microwave receivers for the military mostly. To get fast recovery, the crystals of germanium were heated up to near melting and mechanically twisted to introduce mechanical defects. While the on/off ratio was reduced, so was the recovery time. These diodes were probably in the IBM 650 series that enabled my phd thesis. They would have been a good source of germanium surfaces which one could put point contacts on, particularly if one were more mechanically sophisticated than we were as undergrads. When I got to IBM the lore was that the current gain of the point contacts was really that the collector was a np junction from the phosphor in the phosphor bronze contact and the forming process. I can relate to that when we built diffused-alloy transistors later in germanium with an an arsenic doped aluminum alloy emitter into a diffused-base type substrate. With the right crystallographic orientation the alloy penetrated as a v-shaped trench and the arsenic diffused forward in the brief time at about 700C to form the base. The surface diffusion provided the connection to the base contact. With that we achieved raw switching times of 100 picoseconds, but that was in the 1960's with cooperation with TI who was making similar devices for low noise communication apps. Joe Logue's team simply rejected point contact. The group had been formed after Joe had led the work on the memories for the 701 using surplus radar scope tubes. They had a hard engineering approach. Bell Labs did build a computer with point contacts, but it hardly made a peep in the growing technology. Dick Rutz' full adder transistor was essentially point contact (i.e. deliberate np collector), and I don't remember exactly how he produced the collector. There were cigar box sized demo units which would add and display .. at very primitive levels. .................................. IBM went to everyone in the industry to get someone who would build low base resistance transistors. That was a characteristic that Logue's circuit guys demanded. No one was interested. The washer shaped base contact was specifically for that purpose. TI, which was just moving up from instrumentation for oil drilling into semiconductors was the only one who stepped up. Everyone else thought that transistors were for communication and that base resistance didn't matter. At that point the group under Rutz had prototyped the transistors, as they did for several generations until IBM made the big change to silicon and SLT. I'm sure that the intitial orders were more modest, but I think that IBM was only allowed to build 4 million transistors a year. 1960 was early in the delivery cycle and the big production had hardly started. IBM also got power devices from Motorola and Delco as well as building some special devices like core memory drivers which needed higher voltage capability than was generally expected from germanium. There were also experiments with transistors made with deposited oxide on germanium .. something we came back to in the mid-1960's as we pushed for speed. We did lots of things. I tried GaAs transistors after my experience with that material for tunnel diodes. I found it really hard to work with given the primitive technology I had. Still I was ready when the GaAs LED rush came in with the laser following. Dick Rutz had assembled a small group of key technicians. Ralph McGibbon was ex-navy and steady building devices. Earle Harden was a master toolmaker, put together many of the leading transistors with his fingers, and was a master of precision machining. When I showed up with a new PhD, there was no office available, but a bench in the lab. They taught me as much of the trade as I was capable of learning. How to brace your hands so you could move the tweezers precisely. How to cut a shard of a tiny sphere of indium/gallium and heat it into a smaller sphere on a graphite hot stage for alloying. I learned the hard way to drink less coffee. The day they discovered that I wasn't just another technician was when the site tour guide brought the King of Denmark .. or something like that behind my stool as I was alloying on a hot stage and then introduced me as Dr..... I could see them in the background almost falling off their stools. Earl Harden went on later ... still a mechanical tech .. to define the lapping process still used in making thin film recording heads, but that is a far different story.

Let's look at the state of the art when the 1401 transistors were conceived. There were no silicon alloy transistors worth talking about. The diffused base transistors were brand new in the early 1950's and most of the first of them were implemented in germanium by RCA, following Kroemer's drift transistor theory. Interestingly, I believe that the patent to the diffused base transistor (the structure, not the theory) was held by Lloyd Hunter, who was Logue's boss.

The only silicon transistors were grown junction, which made narrow base transistors, but not in any volume manufacturing manner.

The IBM transistors were sealed in hermetic cans and had their surfaces etched and cleaned so that they were stable in a dry environment. The transistors you see are typical of transistors in open air since they have long lost their hermetic seal.

While the process for making them (either by hand or by IBM's automation) was fairly straightforward and if the thickness of the germanium was controlled, the sizes of the emitter and collector shot spheres correct, and the temperature cycle of alloying to spec, the yield of good transistors was reasonably high. In the automated manufacturing, they were tested and etched before being mounted on a header. The alloy process has messy edge margins that need general clean-up. With the right etch, the yield was high and the surface stable until mounted on the header and sealed in. There may have been a burn-in to eliminate flaky devices, but I don't recall whether there was. This certainly was required for early silicon transistors and ICs.

Semiconductor surfaces are not particularly stable, whether they have an oxide coat or not. Early work on silicon (until well into the MOS era) had most oxides contaminated with sodium which was charged and changed device characteristics and leakage, just with the application of voltages across the device. By the time we really understood how to make really stable silicon devices, the germanium era was past. It didn't have an oxide, but it was perfectly possible to treat the surface and then encapsulate the device to last for many years.

When the iron is oxidized or corroded out of the invar leads on germanium devices, they were no longer encapsulated. The GOOD NEWS is that enough devices still are reasonably encapsulated so that the 1401s work.

Even poorly encapsulated devices worked pretty well. The Raytheon hobby transistors were in plastic, but it was permeable to water and other contaminants. That didn't keep people from using them.

The big thing is that we did not the MOS technology, particularly MOS capacitors, to characterize surfaces. That really didn't become commonly available until the 1970's when I had some great fun with a summer student playing with it to look at both surfaces and point defects in silicon surfaces. I left germanium in the mid 1960's and easy surface metrology wasn't there. It was a decade until I had a lab with home built computer controlled measurement stuff. Since few had anything like we had at the time, lots of things were fun to look at, but there was too much going on to dig down and write papers.

Rick

"David Laws" < laws@computerhistory.org > 02/26/2009 05:03 PM
To: Robert B Garner / Almaden / IBM@IBMUS
cc: "Michael Riordan" < mriordan@ucsc.edu >, "Rosemary Remacle" < rremacle@computerhistory.org >
Subject RE: Pics of early Kilby TI IC's made for IBM's Joe Logue in 1966 (and Ge Alloy-junction picture and loopy I-V curves and scan of 1955 "transistorized 604")

Robert,

Thanks for the interesting items of IBM semiconductor history. I'll talk to our oral history committee about the opportunity.

In the meantime I thought you might enjoy the comments of Morry Tanenbaum, the man who built the first silicon transistor at Bell Labs, regarding the vagaries of germanium devices. This is taken from an oral history panel on Silicon Research and Development at Bell Telephone Laboratories recorded late last year by Michael Riordan.

Tanenbaum: Germanium had some problems. First of all, its energy gap wasn't as large as one would have liked it to be, and so it couldn't operate very much above room temperature without cooling. The other problem was that it was highly surface sensitive. There was something called the ”friendly effect” that if you had a germanium transistor in an ordinary room, and waved your hand at it, and watched on the oscilloscope, the curves would wave back to you. So germanium transistors had to be carefully encapsulated, either in metal or ceramic or something, to control the atmosphere.

The photos of the two devices are examples of the discretionary wiring approach to building custom ICs developed by TI for the Air Force in the mid 1960s. We included a refence to this technique on the Silicon Engine at: http://www.computerhistory.org/semiconductor/timeline/1967-ASICs.html

Your comments about the Bell Labs summer course for university professors back in the early 1950s sparked my interest and I have been doing some digging through the early transistor literature and think I came across some useful info on this course.

Attached is a two page scan from the August 1952 Bell Labs Record, page scan 1, page scan 2, providing a couple of photos and description of a course held at Bell Labs in June 1952. According to the article, a total of 63 students (university professors) representing 33 schools attended. Only a few names are mentioned, and I'm still looking for a complete list of attendees.

It would be terrific to find Dr. Leito mentioned. Notice there is a discussion about some to the professors building experimental kits to bring back to their schools. Possibly this is the origin of the kit that you and your friend Jim Boyden worked with.

I can find no specifics about the kit material that was developed for this course. However, Western Electric continued this tradition throughout the 1950s and 1960s. Attached is a photo of a 1953 set of cartridge type point contact transistors which were performance rejects but were packaged for use as study kits for universities. Also attached is a photo of an early 1960s College Relations Transistor Kit.

One last note. The Bell Labs Record article also includes a discussion of the transistor licenses that Western Electric had sold to twenty-six domestic and nine foreign firms. I'm sure that IBM was one of the domestic licenses.

Thanks for your earlier comments that provided the inspiration for this followup.

best regards,

Jack

I probably have the name of the optics professor wrong. There is one good way to access to a technical library (real or on-line).

The Shockley-Hall experiment (one of those in the kit) consisted of a bar of p-type germanium a couple of cm long. It had (I believe rhodium plated) contacts on the ends. The experimenters needed to connect electrically to these to get a known electric field and hole current. They were to place two point contacts on the bar and measure the distance between them (call the spacing between the point contacts 1.5 cm and the space between the end contacts 2 cm)

One point contact was the "emitter" which injected a pulse of electrons. The other was reverse biased and served as "collector" and showed increased current as the pulse of holes drifting down the bar passed it. The electrical pulse for the emitter triggered an oscilloscope and you could see how long it took for the electrons to arrive at the collector.

This was a great experiment. Jim Boyden and I set it up as one of the "projects" in a junior year experimental physics lab course. It was dazzling because you could actually measure the velocity of electrons under an electric field (electron mobility), which was usually done with the magnetic Hall Effect which was then being used to study carrier motion in metals and other conductors. In semiconductors, Hall effect measured the majority carriers, but not the minority ones.

Jim and I were asked to demo it for a department open house. After we'd graduated and he had gone on to Cal Tech and I departing for the local EE department, I was called back for Physics Open House evenings to set up the experiment and show it again.

This clearly was a one of the highlights of the summer course, The group of faculty observed that you could see the electron mobility, but you could also see the diffusion of electrons from the spreading of the pulse. You could play games and double the field applied to the bar and see the change in pulse shape. This resulted in a paper by everyone in on that demonstration and discovery.

The paper, published in late 1952 or 1953 (probably Physical Review) was the first verification of Einstein's theoretical relationship between drift and diffusion of electrons. If that could be found, I'd love to see it. It had a record number of authors since everyone in attendance at that demo was included.

Anyone have access to technical libraries. That is one thing that I think is wrong. Publication of technical papers is a large income generator for non-profits who get the material free, or even with payment (page charges). They then use the profit from selling journals at modest prices to individuals and extreme prices to libraries and companies. The reason for publication should be availability of the papers for advancement of the field by all. Having been in IBM, this was less obvious to me, but as I took part in IEEE at management levels (from society president to board of directors), it became clear that the model is wrong. Information should be FREE. It is in some fields, at least to many of the people who need it.

Any takers on looking for this article on Verifying Einstein's relation between drift and diffusion?

Rick