Talk:Diffusion pump

	This article has not yet been rated on Wikipedia's content assessment scale. It is of interest to the following WikiProjects:
Chemical and Bio Engineering (inactive) This article is within the scope of WikiProject Chemical and Bio Engineering, a project which is currently considered to be inactive.Chemical and Bio EngineeringWikipedia:WikiProject Chemical and Bio EngineeringTemplate:WikiProject Chemical and Bio EngineeringChemical and Bio Engineering

I seem to recall that cold traps enhance diffusion pump ability, not detract from it. Baffles probably do interfer though. -rmhermen

---

To be honest, my experience indicates that pumping ability is improved when the cold trap is in use. However, the book I used as a reference for writing this entry, says that cold traps interfere with pumping ability. My guess is that the additional inlet tube length required for the cold trap reduces conductance and thus pumping speed, so pumping speed may be higher without any cold trap. However, if a cold trap is already installed, using it reduces flow of oil into the chamber, resulting in an improvement relative to not using the already installed cold trap. Admittedly, I'm not an expert, so I'd welcome input from anyone with more knowledge on the subject.

--Matt Stoker

---

I think the trouble is that diff pumps are really bad at pumping water, which is mostly what you're pumping when you pump from atmosphere. Thus, employing the cold trap seems to improve pumping speed, since it pumps water very well.

--Brian Perkins

---

I have a few years of experience with a variety of pumps and traps, and I've found that cold traps can sometimes improve and sometimes hinder pumping efficiency. Here's why:

1) Cold traps place baffles between the pump and chamber, and this reduces conductance to the pump. Use of undersize cold traps can also create a throat that is smaller than the pump, also reducing efficiency. Cold traps are usually not long enough for the length to have an impact. The problem is that in a high vacuum environment, gas molecules have to diffuse into the pump; they're not pushed into it by a pressure gradient. So the molecules have to accidentally find the hole or path leading to the pump in the course of their random walk around the chamber. Smaller holes and larger baffles hinder pumping.

2) Cold traps cryopump the chamber. You wouldn't know that from reading the wikipedia article on cryopumps, but it's true. Vapor molecules, especially water and other outgassing products, will condense out of your chamber and onto the cold trap. In a general-purpose chamber that has not been baked out, your cold trap will typically reduce your pressure by a factor of 100, until it saturates after about 8 hours of operation. In a tighter chamber with bakeout, the cold trap willl not make as much of an improvement, (because the vacuum is already improved to begin with,) but it won't saturate as fast.

In summary, a cold trap acts as a poor man's cryopump in cheap systems, but act as a hairball in ultra-clean systems. Check out the vacuum page for some more discussion.

--Yannick 04:22, 10 Jun 2005 (UTC)

Has anybody had any gain in the use of these pumps in Vacuum Metallizing?? —The preceding unsigned comment was added by 196.209.22.59 (talk) 17:59, 11 December 2006 (UTC).[reply]

In my experience typical systems made for metallizing under vacuum use a diff pump. D6stringer 22:38, 9 July 2007 (UTC)[reply]

This article currently says that diffusion pumps cannot discharge directly to atmosphere, and also says that a steam ejector is a diffusion pump. These cannot both be right as steam ejectors can and do operate directly to atmosphere.

I would guess a diffusion pump is to an ejector as a turbomolecular pump is to a fan-superficially similar but designed to work at very low pressures where gases behave as free molecules rather than as a continuum.--QuantumEngineer 17:04, 30 June 2007 (UTC)[reply]

These are two different application of diffusion pumps. An injector can exhaust to atmosphere, but they can produce a very poor vaccum. If you use them as high-vacuum pumps, they need forepumping. It should be mentioned that mercury diffusion pumps were used for decades. Generally they are somewhat more efficient than oil diffusion pumps, but due to environmental problems they are very rarely used nowadays. (Valdez from Hungary 9 Oct 2007) —Preceding unsigned comment added by 84.0.211.232 (talk) 17:43, 9 October 2007 (UTC)[reply]

I suggest to split the article into two parts: oil diffusion pumps (which are a type of high vacuum pumps) and steam/pressurized air ejectors. For the oil diffusion pumps it is true that they cannot exhaust to atmosphere. (Peter.steier 14:43, 21 October 2007 (UTC))[reply]

Ok, I did clean up the contradiction myself (I think), Now the articel is mainly on oil diffusion pumps. (Peter.steier 16:31, 21 October 2007 (UTC))[reply]

I have browsed around a little more and think now the ejectors are no diffusion pumps and should be removed from this page. They are well covered in Aspirators and work by the Venturi effect observed in the laminar flow regime and not by momentum transfer in molecular flow. I will do this during the following week if no one protests. (Peter.steier 17:05, 21 October 2007 (UTC))[reply]

I wonder why this kind of pump is named "diffusion", because diffusion does not seem to play a roll in the pumping principle. I remember that when I encountered this kind of pump the first time - as a student some 20 years ago - another student explained me the principle roughly as:

"The oil is heated in the swamp and the entrapped gas diffuses out, and is pumped away by the forepump. The oil condenses on the cold surfaces and the residual gas diffuses into the oil, which flows back to the swamp."

For sure, this explanation is completely wrong, but maybe this is what also the inventor thought in 1913, spawning the strange name? Who has this paper describing the first diffusion pump?

(Peter.steier 16:40, 21 October 2007 (UTC))[reply]

In low vacuums, pumping creates a pressure gradient that helps push gas towards the inlet of the pump. At high vacuums, a pressure gradient still exists, but the absolute pressures are so low as to have negligeable effect on the movement of the gas. In this regime, gas movement can be better modelled by diffusion. This is the transition from continuum to molecular flow. The practical consequence is that the seal leakage of your pump is now not nearly as important as the throat diameter, because you have to wait for gas molecules to diffuse into the throat, and the pressures are so low that they can be retained by a very fragile seal like an oil curtain. The diffusion pump was the first of these large open-throat devices meant to accept gas by diffusion. After the turbomolecular pump was invented, they were grouped under the umbrella term of momentum transfer pumps, even though they are both diffusion pumps in the old sense.--Yannick 17:11, 21 October 2007 (UTC)[reply]

Thanx. This should be probably noted in the article. I try it. However, I lack reeferences. (Peter.steier 17:28, 21 October 2007 (UTC))[reply]

Aw, be bold and put it in. I have references, and I can sprinkle them in when I have the time, or when anyone challenges us. I'm too busy with other articles to do elegant writing on this particular page right now, so your help would be appreciated. Most people are better at putting things into lay terms than I am anyway. By the way, note I made a minor correction to my post above.--Yannick 17:41, 21 October 2007 (UTC)[reply]

When I was looking for the year of invention and the inventor, I finally found a paper from which Gaede's reasons for using the name "diffusion" are mentioned. (Peter.steier 20:36, 21 October 2007 (UTC))[reply]

Good work. If we're going to talk about their widespread use, we may want to mention that they are being replaced by turbomolecular pumps in most modern designs. That's original research on my part, but I suspect it would be possible to find a source. Modern turbomoleculars use magnetic bearings, so the diffusion pump's reliability advantage has disappeared. Its only advantage left is low cost, and the savings can easily be wiped out by your first pump stall.--Yannick 03:05, 22 October 2007 (UTC)[reply]

Old comment, but even in 2011 the oil diffusion pump isn't going anywhere. The turbo pump still cannot compete price and size wise with the diffusion pump. Many large scale industrial vacuum chambers are still diff pumped as they are very low maintenance, low cost and can scale up easily. The largest diff pumps are over a meter in diameter and can pump upward of 50,000 liters a second of air. Turbo pumps top out around 3000-4000 l/s air. Also, magnetic bearings do nothing to make them more reliable. Sudden pressure changes can destroy the fragile blades in a turbo pump causing the lower blades to ingest the upper broken blades leading to a large pile of scrap metal in the pump. If the diff pump oil burns, just pull the pump and clean it out with some solvent. A large diff pump is a few thousand, a large turbo pump is tens of thousands. A diff pump only needs a power source for a heater plate, a turbo pump needs an expensive and elaborate power supply to supply a high frequency AC current to its motor to get the blades spinning upward of 70,000+ RPM. For sensitive lab experiments and equipment, the turbo pump has indeed replaced the diffusion pump. But for large scale systems and even processes where a turbo pump could be damaged by harmful process vapors, the diffusion pump is the first and only choice. 24.186.130.27 (talk) 13:38, 13 June 2011 (UTC)[reply]

"Within the nozzles, the flow changes from laminar, to supersonic and molecular."

Shouldn't this be the other way? i.e. "Within the nozzles, the flow changes from molecular, to supersonic and laminar."?

I am not 100% certain, so thats why I rather ask than change. —Preceding unsigned comment added by Davor (talk • contribs) 00:47, 11 June 2008 (UTC)[reply]

The figure seems to show that pumping speed at low pressures is constant when the pump is baffled, while it drops off with decreasing pressure with no baffle. I am guessing this refers to baffling the input throat of the diffusion pump. I would expect just the opposite. Not an expert here, so I just note it, that it may be unclearly or incorrectly labeled....

The figure is adapted from 3.12 in "Building Scientific Apparatus" (ISBN 0813340063). The pumping speed drops as the pressure above the pump approaches the vapor pressure of the pump oil and backstreaming becomes an issue. This is less an issue with a cold trap or baffle. I will try to clarify this point in the article when I get the chance. Thanks for pointing this out. --Kkmurray (talk) 03:46, 12 September 2009 (UTC)[reply]

I'm after a section near the beginning along the lines of "Vacuum pumps for Dummies". I am looking for more general information on how vacuum pumps work and what the issues and problems are or were. I have heard about them in my early science days, but after reading the current article I still don't know exactly whats going on.07:00, 15 June 2014 (UTC) — Preceding unsigned comment added by 118.211.175.2 (talk)

I came to this article because I'm reading about Irving Langmuir and his contributions to the development of electronic vacuum tubes; I wanted to see if I could get a general idea of his contributions to the development of vacuum pumps. I have a college education but not in physics or mechanics, and just as a "data point" I can hardly understand any of the writing in this article. If someone who understands the subject matter has the time, the communication skills and the mercy, please consider rewriting at least the introductory part of this article so that an intelligent layperson with an average, general education can understand it. Thank you! DSatz (talk) 00:29, 9 August 2025 (UTC)[reply]

The image chosen to lead the illustrations for the article—an installation of a rusty unit in someone's laboratory—is a pedagogically awful choice, communicating absolutely nothing about the subject or its workings. Once a professor. 76.136.83.142 (talk) 17:14, 29 September 2025 (UTC)[reply]