greypilgrim said:
I have no intuition about a reasonable scale. Maybe for startes let's take something lab-scale that could in principle be tested, like ##A=1\ \text{m}^2, d=1\ \text{m}, V=10000\ \text{V}, M=1\ \text{kg}##?
OK, let's consider an idealization of your device, consisting of:
- An antenna engineered to direct electromagnetic radiation in a single direction.
- A parallel-plate capacitor of area ##A## and separation ##d##, charged to voltage ##V##, that energizes the antenna.
- Conductors with no resistive losses.
Then assume, upon capacitor discharge, that the antenna emits a unidirectional burst of radiation carrying energy ##E_{\text{rad}}## and momentum ##E_{\text{rad}}/c##, and the whole device of mass ##M## recoils in the opposite direction with speed ##v##.
Ignoring the capacitor's fringing fields, its capacitance is ##C=\epsilon_0\frac{A}{d}## and it stores an energy ##E_{\text{cap}}=\frac{1}{2}C V^2##. By conservation, the total energy and linear-momentum is the same before and after capacitor discharge, so:$$E_{\text{cap}}=E_{\text{rad}}+\frac{1}{2}Mv^{2},\quad0=Mv-\frac{E_{\text{rad}}}{c}\tag{1,2}$$
(Here I've used the non-relativistic expressions for the energy and momentum of the device because I anticipate that ##v\ll c##.) Solving the conservation eqs.(1,2) gives:$$\frac{v}{c}=\frac{E_{\text{rad}}}{Mc^{2}}=\sqrt{1+2\left(\frac{E_{\text{cap}}}{Mc^{2}}\right)}-1\approx\frac{E_{\text{cap}}}{Mc^{2}}\text{ for }\frac{E_{\text{cap}}}{Mc^{2}}\ll1\tag{3}$$Now it's time to put in your numbers (slightly adjusted): ##A=1 \text{ m}^2,d=1\text{ }\mu\text{m},V=10\text{ kV},M=1\text{ gm}##. (Note that I've reduced your distance ##d## by ##10^{-6}## to boost the capacitance ##C## and lightened your mass ##M## by ##10^{-3}## to help the device "fly"!) Plugging these into (3), we find:$$C=8.85\:\mu\text{F},\:E_{\text{rad}}\approx E_{\text{cap}}=443\text{ J},\:Mc^{2}=9\times10^{13}\text{ J},\:v=1.48\text{ mm/s},\:\frac{1}{2}Mv^{2}=1.09\times10^{-9}\text{ J}$$So this idealized device achieves a speed on the order of just a millimeter per second. And any real test device will almost certainly be slower due to a much poorer directivity of the radiated momentum, as well as conductor losses. But maybe you could detect the effect by suspending the test device in a vacuum and looking for minute movements?