Why predominantly PNP Germanium transistors?

[stevehertz]

Yes, why do germanium transistors mainly come as PNP types? I'm guessing that it was much easier to manufacture PNP types for some reason.

[jjl]

I think that it's most likely that N type germanium is more difficult to make than P type, but I don't know the details of why that should be. I know that arsenic and boron are used for doping silicon semiconductors and I think that the same elements are used for germanium too.

John

[Radio Wrangler]

It's the controllability of the rate of diffusion leads to the choice of which goes in as bulk into the substrate, and which is diffused in later. The aim is to control the thicknesses.

David

[stevehertz]

Originally Posted by Radio Wrangler

It's the controllability of the rate of diffusion leads to the choice of which goes in as bulk into the substrate, and which is diffused in later. The aim is to control the thicknesses.

David

Can you explain that bit further please David, making reference to PNP and NPN? Thanks.

I have wondered why, and as far as I can tell making an "N" substrate is easier and diffusing "P" is slower giving a more controlled result. Also Indium (P) is easier to handle than antimony (N). This is from memory and I am quite willing to be shot down.

[Keith956]

It's a long time since I was involved in semiconductor physics but here goes:

Semiconductors are typically group IV elements (Carbon, Silicon, Germanium or SiC) or more complex like III-V (GaAs, InSb, AlP, GaN) or even II-VI (CdTe), but in the early days germanium and silicon were the main candidates as they could be grown as single crystals in high purity fairly easily compared to the others.
It's very difficult doping carbon (Diamond), Silicon needs temperatures of 1000-1200C (it melts at about 1420C) , germanium much less so (melts at 940C), so germanium devices were easier to make at first.

In the early days it was reasonably straightforward to grow single crystal germanium by first removing impurities using a technique called zone refining, which swept a heat source along a bar of the material, melting it in a zone which concentrated the impurities. This was done several times and the end bit was discarded. It was then put into a 'puller' where it was meted and a single crystal dipped in and very slowly lifted out while rotating. This caused a long tubular crystal to grow. An impurity was deliberately added to the melt to dope the material. I can remember seeing a crystal grower still used in Philips (formerly Mullard) Southampton site, I think around 1981.

P type dopants are group III - Boron, Aluminium, Gallium, Indium. The solubility of these decreases in that order. N type dopants that can be used are Phosphorus, Arsenic, Antimony. In the case of germanium PNP devices, antimony was used as the N type dopant and Indium as P type; the P type regions for these devices like the OC71 were formed by alloying in a blob of indium onto the front and back of an N type germanium wafer.

As to why PNP germanium devices were easier to fabricate than NPN ones it is probably a combination of factors including the solubility of the dopants, their diffusion rates, the ease of making ohmic contacts and defect density of the resulting junctions.

Silicon largely took over as silicon dioxide is a much more stable (and better insulating) oxide than germanium oxide (which is water soluble), and silicon's higher bandgap (1,1eV vs 0.7eV) meant higher operating temperatures and less leakage.

[poppydog]

Nicely explained, even i understood that...

poppydog

[Keith956]

Originally Posted by Keith956

As to why PNP germanium devices were easier to fabricate than NPN ones it is probably a combination of factors including the solubility of the dopants, their diffusion rates, the ease of making ohmic contacts and defect density of the resulting junctions.

So I brushed the dust off my old textbooks... N type dopants in Ge diffuse an order or two of magnitude faster than P type dopants, so the base width of a NPN alloy diffused transistor would be harder to control than a PNP device. I can't see any other obvious reason.

[Radio Wrangler]

Diffusing P-type emitter and P type collector in from opposite faces of a disc of N-type base material was critically dependent on control of diffusion rates. You didn't want them to touch in the middle because that ruined the device, and the device gain was dependant on getting the base region thin. As you can imagine, errors and tolerances were exaggerated, leaving gain etc very poorly controlled.

People still had their valve brains fitted and were used to designing around much smaller variation of gm from device to device. Many of them didn't know what hit them when they changed to transistors. New circuit techniques took ages to develop.

David

[DMcMahon]

With later Silicon transistors and Silicon integrated circuits most of the old fashioned diffusion methods of doping Semi-conductors got superseded by Ion Implantation.

[stevehertz]

Originally Posted by Radio Wrangler

Diffusing P-type emitter and P type collector in from opposite faces of a disc of N-type base material was critically dependent on control of diffusion rates. You didn't want them to touch in the middle because that ruined the device, and the device gain was dependant on getting the base region thin. As you can imagine, errors and tolerances were exaggerated, leaving gain etc very poorly controlled.


David

I'm sure it's me not understanding what you're saying David, but to me it looks like you're describing a P-N-P 'sandwich' there and how critical and difficult a process it is? Yet PNP trannies are the more common?

[MotorBikeLes]

Most of what Kieth said rang a bell with me, but then at tech in the early '60's, we were given a Mullard booklet (still have it somewhere) with all the photos of furnace refining and info on "pulling".
One aspect of n and p type germanium and silicon is the difference in reliability, probably more so a few decades ago.Where together in a circuit, the n-type germanium was the likely failure, but the p- type in silicon pairs.
Personal observation.
Les.

[Goldieoldie]

Interesting early USA radios ( Regency tr1 ) etc and sone Japanese used NPN transistors
Don't know why this was ?
Cheers
Pete

[Radio Wrangler]

Originally Posted by stevehertz

I'm sure it's me not understanding what you're saying David, but to me it looks like you're describing a P-N-P 'sandwich' there and how critical and difficult a process it is? Yet PNP trannies are the more common?

It's the way early germanium transistors were made. P diffusion is relatively slow and therefore easier to control.

If they tried to make an NPN transistor the same way, starting with a disc of P doped germanium, and then applying some N dopant in the middle of each face, the stuff would diffuse in much faster and make it much more difficult to stop things at just the right spacing before the dopants from each side met in the middle and ruined the transistor.

Think of applying a drop of food dye to opposite faces of a sugar cube, hoping to get the migration of the dye to stop at a point leaving a thin zone of undyed sugar between the two dyed regions!

With semiconductors, the mobility of the dopant is controlled by the temperature. The collector and emitter need the mobile doping process to form them, so they benefit from the polarity of doping which is easiest to control, P. and that leaves only N for the base.

Years later, things moved to silicon to take advantage of its wider bandgap allowing higher temperatures and better thermal stability, but another big factor were the excellent insulating properties of silicon dioxide. Germanium's oxide is not so good and is water soluble. By this time, transistors were made by monolithic processes.

A wafer is made with N doping. A bathtub of P dope is diffused in from the top side. Then a smaller bathtub of N is diffused into part of the P bathtub... the N doping is strong enough to cancel the P in this smaller bathtub and leave it N We now have an NPN transistor. Electrons enter via a bond wire attached to the smallest bathtub - the emitter. They pass to the base - the thin film of P-ishness under the emitter bathtub, some leave via the base connection, the rest continue onwards to the N-type substrate, the collector.

Most heat is created in the collector, and that is the bottom layer which is easiest to plant on a heatsinking piece of copper. This is why metal canned transistors have the collector connected to the can (both NPN and PNP types)

PNP transistors can be made in the same way, just with the dopants all swapped.

BUT electrons are more agile than holes, so NPN transistors have a bit of a gain and frequency advantage over PNP (NPN work mostly with electrons, PNP with holes) So along came a preference for NPN.

In ICs, the processing is chosen to favour NPN transistors being made for their advantages. The polarity of the substrate makes PNP hard to make and they often were 'lateral' types where the electron flow was horizontal. This fitted in with the processing without great expense, but the transistors made were terribly slow.

Si ICs designers got even more impulsion to use NPN wherever possible.

Since then, fancy processes allow much better PNP to be made and this has enabled the design of some fancy fast opamps.

A simplified whistle-stop tour, but it ought to give the general route of development.

David

[stevehertz]

Thanks David, that's great info.

But NPN germanium transistors were made?

[ajgriff]

Originally Posted by stevehertz

But NPN germanium transistors were made?

I suppose the AC187 is just one well known example of an NPN germanium transistor. Often paired with the PNP AC188 as the audio output stage of sixties and early seventies portable radios. NPN germanium transistors were usually more expensive than PNP types for the reasons outlined in previous posts.

Alan

[Keith956]

Originally Posted by Radio Wrangler

In ICs, the processing is chosen to favour NPN transistors being made for their advantages. The polarity of the substrate makes PNP hard to make and they often were 'lateral' types where the electron flow was horizontal. This fitted in with the processing without great expense, but the transistors made were terribly slow.

Si ICs designers got even more impulsion to use NPN wherever possible.

David

One of the reasons why NPN devices are better for Si ICs is that you can use a N+ buried layer - typically arsenic doped - which has a very low resistance, compared to the collector which needs to be relatively lightly N doped, and hence higher resistance. And you don't want transistors with a high Rc. Arsenic also diffuses slowly compared to the only practical p type dopant for Si, boron, which was not much use for buried layers.

There are all sorts of tradeoffs in device design - in Si things like latchup was a problem in early cmos (and the subject of my thesis

Keith

[kalee20]

With the alloy-junction technique - described by Keith956 and Radio Wrangler - you need the blobs of dopant, which form the emitter and collector terminals, to be metallic so they conduct, and so that you can tack a leadout wire on.

For germanium PNP transistors, pellets of indium are used, so that's fine. The P dopant needs to be a Group III element; the N dopant a Group V element. Unfortunately, going across the periodic table from left to right, the elements get less metallic. Little dots of phosphorus as N dopants to make NPN transistors? Fine, until you come to try to weld wires to them! Arsenic similarly. Not a metal. Antimony is about the only option, and Sod's law says with fewer choices, there's more likely to be some near-showstopper.

[Diabolical Artificer]

This video - https://www.youtube.com/watch?v=ihkRwArnc1k goes into detail of early transistor making, it might throw light on the subject.

Andy.

[John10b]

In the excellent explanation given by David in post #14 it states “But Electrons are more agile than Holes..” if I remember correctly from college days, a very long time ago, a hole is created by a vacant Electron, if this is so how can it be more agile?
John