# Why do PNP silicon transistors cost more to produce? NPN is cheaper?

I got a few suggestions/answers but nothing definite after searching and digging around awhile .. I am not so sure about any of these so correct any of them that you see

Thanks

Alex Van den Bossche · Ghent University Yes, speed is one thing, but a PNP transistor requires 3 times more SI chip surface, so it gets uneconomical when the chip costs are a big part of the component.

Silane diborane and phosphine gases all react differently .. it is more difficult to control borane p-type doping of silicon also the impurities in polysilicon (it can never be completely pure) act as n-type dopants .. so for quality p-type silicon you have to get much purer polysilicon so n-type silicon is easier -- On the Enhancement of Silicon Chemical Vapor Deposition Rates at Low Temperatures - chang1976.pdf

My last guess was that it is just economics of scale .. NPN are simply far more mass produced so supply and demand .. but it had to start out that way for a reason

It depends on which one is easier to produce, using the materials that are to hand. Given that most circuits can be reconfigured to accept either type, except the obvious complementary totem pole which needs both, production tends to be concentrated on the cheaper, better form.

In the bad old days of germanium, PNP was the easiest transistor to make out of the available doped forms.

With silicon, NPN is easier.

Why is one or the other easier? It has to do with the relative carrier mobilities and conductivities of the doped semiconductor. One type will turn out to have smaller parasitic elements than the other. For the same size die, the better one will carry more current. For the same properties, the better one is smaller and so cheaper.

In an ideal world, we would like to have dopant atoms with properties that would give us excellent semiconductors. In the real world, we are stuck with the properties that boron, nitrogen and the other group 3 and group 5 elements happen to have, which just give us good ones. When we go beyond germanium and silicon, and create other substrates, the so-called 3-5 alloys, we get 'better' semiconductors, which don't necessarily have the same guidelines as for silicon and germanium.

• So Alex was right? the PNP transistors are larger in surface area due to the slower transport speed of P-type silicon? ecee.colorado.edu/~bart/book/book/chapter2/ch2_7.htm table 2.7.2 – Jimmyboy Jun 13 '17 at 13:06
• I thought this was helpful: so +1. – jonk Jun 13 '17 at 17:30

Mobility of electrons are about 2-3 times the mobility of holes. In PNP BJTs, holes are the majority carrier.

Mobility relates to a kind of "effective inertial mass" of a charge and so mobility also relates directly to mean velocity under the same accelerating electric field strength. (Analogous to the idea that a garden hose sprayer will push along small gravel faster than larger gravel.) In fact, the units of mobility are exactly what is needed so that all you have to do is multiply by the electric field strength in order to get the mean velocity of the charges: $\overline{v}=\mu\cdot \mathscr{E}$.

The magnitude of a current (in amps) is the number of charges (in Coulombs) passing through a cross-sectional area per second. Charges of the same polarity tend to uniformly distribute themselves throughout the volume (for obvious reasons) and therefore throughout a cross-section of that volume. If their average velocity is lower, and if the charge density in the material is otherwise similar, you simply then need more cross-sectional area in order to achieve the same current.

It's a pretty simple idea. (Well, it's designed on the simplifying "charge cloud" model and the idea that accelerating charges come to an abrupt stop upon impact of "something" [atom?] before starting back up again. Which, of course, isn't right at the quantum level. But arises as a quite good model when taken at the large-scale statistical level we usually encounter in circuits.)

So the page you mentioned in a comment is actually pretty good to read to get the part of the view the author is discussing. I like that page.

I've also been involved, peripherally to be honest, with some of the gases used in FABs. The ones I remember more are for vapor dep, epitaxy, and doping: arsine (TLV of 50ppb -- highly toxic, flammable), phosphane (pyrophoric, toxic), and silane (explosive, corrosive, and toxic.) All "very bad" stuff.

However, on the "pureness" comment you quoted I have to say that this problem has been largely solved. Silicon is available readily at remarkable purities for IC processing and I don't think that part is a problem, of late.

I don't know much about the business end. It would seem reasonable to my limited views to suggest NPN BJTs are cheaper also because there is more production experience and longer production ramp-up times and knowledge, as well as greater demand volumes.

But I think the mobility is, at best, only one of many contributing factors.

For discrete parts, I honestly don't know if the increased area needed due to the lower mobility means that the dies they cut from a wafer are smaller for NPN vs PNP. If it's possible and reasonable for the overall goals, then I'm sure they are doing it to get a higher yield from a wafer. But there are other considerations that would come to my mind, anyway. Part of that is how difficult it might be to take advantage of a supposedly smaller die when cutting them, and how the size of the die they cut might relate to its ability to dissipates power into the resulting package, and how a smaller die might complicate wire-bonding to the leads, and so on. There are many more considerations than just the mobility that manufacturers need to worry over. So I don't really know if they get a higher number of dies out of a wafer for discrete BJTs. If not, then this doesn't impact "cost" as much as we might imagine from only the mobility question.