I heard that microprocessors are usually made using silicon, yet germanium has a better conductive state.
Why isn't germanium not chosen instead, taking to account that better conductors provide better efficiency?
\$\begingroup\$
\$\endgroup\$
1
-
9\$\begingroup\$ I think you must know that there is a good reason for this so, where have you researched in order to form your proposition? I mean, would you choose a car based on its colour scheme alone or, would you consider a lot of other factors? \$\endgroup\$– Andy akaCommented Mar 16 at 14:28
Add a comment
|
6 Answers
\$\begingroup\$
\$\endgroup\$
The most important special property of Si is that its oxide is an unusually useful material. Mechanically robust, chemically inert, electrically insulating, easily deposited, etchable, ...
\$\begingroup\$
\$\endgroup\$
2
If a material is well conductive, that makes it a terrible choice for logic. Otherwise we'd see logic made of Copper. What you actually need is material that can be switched very well between conductive and insulating states with high contrast and efficiency. This is why we use materials with a bandgap (making the conductivity atrocious) in the first place.
Wider bandgap materials generally allow for higher such contrast (hence they are "better"), if they can be suitably doped to establish highly conductive p and n regions. Si can be suitably doped, which is currently not the case for even wider bandgap materials. So Si sits an a sweet spot for logic fabrication.
And it has a ton of "nice to have" advantages as mentioned in other answers.
-
1\$\begingroup\$ IBM has been using copper in processors for a time. \$\endgroup\$ Commented Mar 18 at 12:09
-
3\$\begingroup\$ @SimonRichter for logic ?? or interconnects? \$\endgroup\$– tobaltCommented Mar 18 at 21:16
\$\begingroup\$
\$\endgroup\$
5
Modern microprocessors produce a lot of heat. The maximum temperature a germanium semiconductor will work at is 70 degrees C. Can you imagine the cooling problem for a modern computer if you had to keep the temperature well below that.
There are also significant materials challenges in the fabrication of Ge MOSFETs. If these are solved you may see mixed Si-Ge logic but it is likely to be much more expensive than Si.
-
2\$\begingroup\$ High power Ge transistors were common before Si took over, so the poorer thermal conductivity was manageable in practice. \$\endgroup\$ Commented Mar 16 at 14:10
-
4\$\begingroup\$ Incidentally, Ge is commonly used in many modern fabs -- not usually pure, but as a mixture with Si to make heterojunction devices. If an advantageous structure were found, they would most likely be able to implement it pretty quickly :) \$\endgroup\$ Commented Mar 16 at 14:40
-
\$\begingroup\$ We were doing mixed signal SiGe chips (RF, digital, power switching) 3 years ago before I retired. \$\endgroup\$– SteveShCommented Mar 17 at 0:14
-
\$\begingroup\$ @SteveSh High density high speed logic? \$\endgroup\$– RoyCCommented Mar 18 at 9:41
-
\$\begingroup\$ @RoyC I guess that depends on your definition. The chip was mostly RF, with the digital only taking up 1/8 of the die area. Speed was moderate by today's standards, 25 MHz - 50 MHz IIRC. \$\endgroup\$– SteveShCommented Mar 18 at 11:00
\$\begingroup\$
\$\endgroup\$
5
The Germanium vs. Silicon Wafers – Why Silicon Is Preferred page from Wafer World Inc contains the following three advantaged for Silicon:
- More Abundant Supply:
Because of its abundant supply, silicon wafers are also cheaper to manufacture en masse than germanium-made ones.
- Greater Heat Resistance:
Another area where silicon is superior to germanium is heat resistance. Silicon can withstand up to 150 degrees Celsius or roughly 300 degrees Fahrenheit. On the other hand, germanium wafers are prone to breakage or melting at 70 degrees Celsius or 158 degrees Fahrenheit.
- Fewer Free Electrons at Room Temperature:
Elements are only good conductors if their free electrons are less than 3. At room temperature, silicon has fewer free electrons than germanium.
The above point about silicon having fewer Free Electrons, and therefore contributing to silicon being a better conductor at room temperature is a counter-argument to the following from the question:
I heard that microprocessors are usually made using silicon, yet germanium has a better conductive state.
Albeit I'm not a designer of integrated circuits so may have misunderstood the relevance of silicon having fewer free electrons than germanium.
answered Mar 16 at 13:39
-
\$\begingroup\$ I'm not sure where they got "if their free electrons are less than 3" from. The more free electrons, the better for conductivity. \$\endgroup\$– HearthCommented Mar 16 at 15:39
-
5\$\begingroup\$ I think the "fewer electrons" thing is an awkward way of saying that one wants a certain minimum bandgap to keep room temperature leakage below certain levels for VLSI. Germanium VLSI might show too much leakage in the "off" state. At the time, Si was the "wide bandgap semiconductor", with all the advantages that we attribute to WBG semiconductors of our time 😉 \$\endgroup\$– tobaltCommented Mar 16 at 16:17
-
2\$\begingroup\$ The properties of silicon and germanium are found here. The resistivity is markedly different between the two. The linked page has a poor choice of words, writing "good conductors" without defining the meaning of good and talking about semiconductors when using the word conductors. A "good semiconductor" doesn't prioritize properties in the same way as a "good conductor" does. \$\endgroup\$ Commented Mar 16 at 19:25
-
\$\begingroup\$ This is an excellent answer. I want to add that Germanium is water soluble. \$\endgroup\$ Commented Mar 18 at 9:21
-
1\$\begingroup\$ That source looks like really obnoxious SEO bait. I'm inherently skeptical. \$\endgroup\$ Commented Mar 18 at 14:31
\$\begingroup\$
\$\endgroup\$
2
None of these answers are correct:
A native oxide is not necessary, since 28nm, HKMG transistors have used HfO as the main dielectric.
GAA-FET’s use epitaxial growth, so the wafer material itself isn’t that important
Heat is irrelevant for reason 2
The real answer is that the advantage in carrier mobility is not worth the integration difficulties and effort
To an extent it is used, since most PMOS devices use SiGe
-
2\$\begingroup\$ I think part of this question is historical. High quality SiOx was so convenient, it historically was a dominant reason Si was preferred. That's part of the reason why Si integration is easier than Ge/SiGe, even today when HfO is even higher quality and more common. \$\endgroup\$– KD9PDPCommented Mar 17 at 3:26
-
\$\begingroup\$ @KD9PDP yes, that's a fair point, and GeOx still has some issues in the modern day (though granted no one really has put a lot of effort into it) \$\endgroup\$ Commented Mar 18 at 6:51
\$\begingroup\$
\$\endgroup\$
3
Silicon vs germanium in IGBT's.
Follow the money:
Which has a lower cost of goods sold - silicon.
Silicon wins.
Mechanically:
Which has a greater Strong's modulus of plasticity - silicon.
Mechanically:
Which has a greater Strong's modulus of elasticity - silicon.
That is be enough for me to say that silicon is a better choice.
Even if I was wrong about the mechanical properties, still the cost of goods sold is an overriding constraint.
Silicon wins.
Manufacturing/Processing:
If the chips lines are cut with laser then the heat resistivity of the Silicon being higher can lead to less melt-bleed-off.
Silicon wins.
Manufacturing/Processing:
If the chip lines are cut with laser then the heat transmissibility of Silicon, I think would be less (but I am not certain about that) could also lead to less melt-bleed-off.
Silicon might win if this is correct.
-
\$\begingroup\$ There are cheaper and stronger materials than high purity Si wafers. And what is it about laser cutting? \$\endgroup\$– tobaltCommented Mar 17 at 5:36
-
\$\begingroup\$ @tobalt, The question in title was about two specific materials. I was not deviating from the question title. I know there are more than two options, but I supply two for the following: Acid etching = known to be not very accurate and having a tendency to bleed into and away from the desired line. Laser cutting, if properly polarized lasers are used then the cut of the desired line can be more uniform with less chance of bleed off from or into the line. But with laser there is the heat factor, thus some of my further detail in my answer. I intended to help not make an entire dissertation. \$\endgroup\$ Commented Mar 18 at 16:59
-
\$\begingroup\$ @tobalt, Furthermore, if the option of printing the lines is compared to laser cutting, then due to the printing processes (as far as I know currently) are not only using laser, but are melting the line for it's placement which might allow for bleed off from the line. OK, so now I come to the option of laser incorporated printing with laser trimming of the line edges, which seems to be currently the best. Thus, when I made my comparisons of the two materials, I chose laser interactions as part of a basis. I guessed that this would be obvious, but maybe not. Thank you for asking. \$\endgroup\$ Commented Mar 18 at 17:11