The paper Measurement of Impulsive Thrust from a Closed Radio-Frequency Cavity in Vacuum (H. White et al, J. Propulsion & Power, November, 2016, http://dx.doi.org/10.2514/1.B36120) refers to an unusually shaped copper cavity with a resonance at about 1.94 GHz. This is described in the quoted section below. (further reading: https://space.stackexchange.com/questions/tagged/emdrive)

Fig. 4 suggests that the Q of this cavity is over 7,000 (7E+03). As far as I can tell there is no suggestion of an unusually conductive coating inside the copper.

My question is about the extremely high Q. I think among those with experience with ~GHz resonant copper cavities should be able to answer this based on experience, without it being too opinion based. Could a copper RF cavity like this be reasonably expected to have a Q > 7000?

I'm curious - with a drive of 50W, what would be the order of magnitude electric fields inside? kV/m? MV/m? I can break this off as a separate question if necessary.

An example of anything close in configuration and Q could be the basis of a "yes" and an example of anything close in configuration, highly optimized, and not even close in Q could be the basis of a "no" answer.

B. Test Article

The RF resonance test article is a copper frustum with an inner diameter of 27.9 cm on the big end, an inner diameter of 15.9 cm on the small end, and an axial length of 22.9 cm. The test article contains a 5.4-cm-thick disk of polyethylene with an outer diameter of 15.6 cm that is mounted to the inside face of the smaller diameter end of the frustum. A 13.5-mm-diam loop antenna drives the system in the TM212 mode at 1937 MHz. Because there are no analytical solutions for the resonant modes of a truncated cone, the use of the term TM212 describes a mode with two nodes in the axial direction and four nodes in the azimuthal direction. A small whip antenna provides feedback to the phase-locked loop (PLL) system. Figure 3 provides a block diagram of the test article’s major elements.

enter image description here

above: Figure 4 from here. Right click to open in a separate window to view clearly as full size, or view at the original link.

enter image description here

above: "Fig. 14 Forward thrust mounting configuration (heat sink is black finned item between test article and amplifier)." from here

enter image description here

above: "Fig. 17 Null thrust mounting configuration, b) view from side" from here

  • 2
    \$\begingroup\$ If the Q is so high and the output thrust (and presumably power) so low then why is there a bloody great big heatsink on the end of the copper bucket? Where is all the power going? \$\endgroup\$
    – Andy aka
    Jan 24, 2017 at 13:59
  • 3
    \$\begingroup\$ @Andyaka That looks like a lovely heatsink to use where there's convection going on. Shame they're using it in a vacuum. \$\endgroup\$ Jan 24, 2017 at 20:03
  • \$\begingroup\$ @Andyaka I think the heatsink is on the drive electronics, not the resonator. Quite what it does in a vacuum is another matter! \$\endgroup\$ Jan 25, 2017 at 10:53

2 Answers 2


The trick to getting good microwave resonant-cavity Q is to have a good conductor, a smooth finish, precise alignment, light coupling of the input signal, and limited microphonic pickup.

The design in the picture looks like it might have been limited by microphonics, and then reworked to eliminate them. For example, it uses a large heatsink instead of a fan. It also looks like alignment would be a real chore!

The loaded Q specification for the Keysight Split Cylinder Resonator is >20,000 at 10 GHz. If you look into one of the resonator halves, you will see yourself in the mirror surface finish. The resonator is gold plated and precision diamond turned. The parts look so good that they used clear plastic for the instrument covers! Very unusual for Keysight gear.

Here is more background information about the Split Cylinder Resonator, in case anyone is interested:

The alignment is done with a kinematic mount, similar to how a telescope mirror is adjusted. The resonator halves can then be adjusted back and forth, while maintaining the alignment. A measurement sample is placed in the gap. The sample changes the Q and resonant frequency of the resonator. This, along with a Network Analyzer, enables measurement of the sample dielectric constant and loss. The accuracy of the dielectric measurement relies on having a high-Q resonator.

Here are the specifics on the surface finish from the datasheet: "Cylinders are precision diamond turned Al 6061-T6 plated with 0.5 μm Cu, 0.25 μm PdNi, and 2.0 μm Au."

Full disclosure: I am speaking for myself, not Keysight, even though I work there.

  • \$\begingroup\$ This answer is very helpful, since you've given a lot of practical background information. Considering that the question says "An example of anything close in configuration and Q could be the basis of a 'yes'..." and that's just what you're showing here, I can assume it is a reasonable expectation, provided one knows what one is doing. Thanks! \$\endgroup\$
    – uhoh
    Jan 25, 2017 at 10:49
  • \$\begingroup\$ Sounds like Keysight ought to offer to make Nasa a better one to see if improved Q improves thrust... \$\endgroup\$ Jan 25, 2017 at 10:51
  • \$\begingroup\$ note: in this application the cavity and RF source are sitting on top of a very sensitive balance, and micro-Newton forces are being inferred, so I think a fan would be excluded from the beginning. Consider also the title of the paper: "Measurement of Impulsive Thrust from a Closed Radio-Frequency Cavity in Vacuum" \$\endgroup\$
    – uhoh
    Jan 25, 2017 at 10:52
  • \$\begingroup\$ The Keysight resonators are based on the research done by NIST, see nvlpubs.nist.gov/nistpubs/Legacy/TN/nbstechnicalnote1354.pdf . This uncertainty analysis from NIST was useful for product development. One of the challenges is to design with shapes that can be mechanically measured to great precision, so that the mechanical measurements can be related through the uncertainty model into a prediction of the microwave performance. This is the basis for microwave calibration and verification standards. \$\endgroup\$ Jan 25, 2017 at 18:52

Yes, a quality factor of 7000 is not even near the upper limit for cavity resonators made of copper at that frequency. Microwave copper cavities with quality factors of \$10^6\$ are not uncommon. Exotic superconducting cavities can reach Q factors of \$10^{12}\$ (!!!).

Calculating the energy stored in a truncated conical cavity is nontrivial, and requires integrating the transverse magnetic and transverse electric fields, computed for a given geometry using Maxwell's equations. How to do so is beyond the scope of this question, but there is an excellent walkthrough and solution set of differential equations for a truncated spherical cone (not quite the same as this, but close enough) here. In fact, that entire page is just a wonderful write up on this topic and I heartily recommend it to anyone interested in getting dirty with the math.

Let's just do an easy one, a resonant cavity that is a simple cylinder. It's not a completely terrible substitute for a truncated cone, I think you'd agree.

The Q factor for such a cavity is:

$$Q = \frac{2\pi f\cdot \frac{\mu}{2}\int_{v}H^2dv}{\frac{R}{2}\oint_{s}H_{t}^{2}ds}$$

and I already have heartburn so I'm going to do what any engineer would do and use the much simpler approximation instead! One can show that a resonant cavity will have a Q that is in the order of magnitude of:

$$ Q\cong \frac{2}{\delta}\cdot \frac{V}{A} $$

where \$\delta\$ is the skin depth at the resonant frequency in question, and V and A is the volume and surface area of the cavity. In other words, the ratio of a cavity's volume to surface area is going to set a fairly narrow range of Q factors that a cavity, regardless of the exact geometry, can possibly have.

It should be apparent by now that creating a simple cylindrical cavity out of copper with a Q well above 7000, more like between 10,000 and 100,000. 7000 actually seems unusually low for a cavity shaped like the one in the photos. At the skin depth they're at, surface smoothness and imperfections become a concern, so if the surface quality inside is crappy, this could cause the Q to drop significantly.

Anyway, to answer the unasked question here, which is how does this thing produce thrust.... well, it's not at all anamolous. It seems to be exactly the right magnitude for the expected thrust due to uneven radiation of heat, as can be seen by the write up I linked earlier. This does produce thrust, and it will work in a vacuum. Unfortunately, relativity enforces a rather depressing limit on the thrust per power.

This drive will never produce more than micronewtons per killowatt. This makes it the most inefficient and impractical means of space propulsion available, reaction mass or no. And it will not get better. At least, that's the conclusion I've drawn, but I would love to be proven wrong.

  • \$\begingroup\$ Nice analysis on the Q, and the supporting link justifies the expected thrust due to uneven radiation - or photon emission from a flashlight in vacuum - as 3.3uN/kw - as you suggest. But NASA's reported measurements in vacuum are orders of magnitude higher - around 1 uN/watt. \$\endgroup\$ Jan 24, 2017 at 20:29
  • \$\begingroup\$ This is a very interesting answer, and I'll take some time to look at the link. For the equation for Q of a cylindrical cavity, could you add an additional, separate link (within the answer) to a non-spacecraft-related site as well? I don't have any microwave texts handy. You are right - order of magnitude estimates are just fine for the purposes of this question. Thanks! \$\endgroup\$
    – uhoh
    Jan 25, 2017 at 0:48
  • \$\begingroup\$ side note: now that you've introduced me to the Greg Egan site(s), my productivity for the rest of the week is probably shot. cf gregegan.net/SCIENCE/Bearings/Bearings.html \$\endgroup\$
    – uhoh
    Jan 25, 2017 at 1:03
  • \$\begingroup\$ @Brian Drummond hmmm, an old, old controversy where measured thrust was far greater than the expected radiation reaction force was ...Crookes' radiometer. Eliminating the "radiometer force" artifact caused by trace gas or surface contaminants is non-trivial, especially if surface temp diff. is far higher than that of paddles in a light-mill. Even extremely hard vacuum may not be good enough. An expert in ultra-clean, UHV chambers perhaps could do it, but better would be to just put the damn thing in high earth-orbit environment, well cleaned, plus let it outgas for weeks before testing. \$\endgroup\$
    – wbeaty
    Jan 25, 2017 at 2:39
  • \$\begingroup\$ @wbeaty ... yes, if the observed phenomenon is outgassing, thrust can be expected to decline as the outgassing reaction mass is consumed. We shall see ... though I would find it hard to believe the NASA experimenters haven't already examined that hypothesis. \$\endgroup\$ Jan 25, 2017 at 10:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.