My colleagues at JPL and I have discussed this at length, being hams, and building radios for Mars as our “day job”.

I think something like 10 GHz band would currently be optimum. Given a reflector size there’s a tradeoff on pointing accuracy. A 3 meter aperture would give you a beam width around 1 degree, which is practical for pointing. It’s about 45 dBi gain.

Free space path loss between isotropes at 10 GHz with a 150 million km path (1AU) is 300 dB (32 + 100 + 163)

Assuming I want to communicate at 10 bits/second, and I can get a decent noise figure, I need a received signal around -160 to -150 dBm. Let’s assume -150 dBm.

If both ends of the link have 3 meter apertures, the antenna gain is 45 dB per antenna, so we are now at 210 dB path loss, so I’d need a transmitter power of 60 dBm to get my desired -150 at the other end. 60 dBm is 1 kW, which is a practical (but not cheap) power at 10 GHz.

If you go to 5 meter apertures (pretty big, but there are people who have them) on both ends, you could pick up another 8–9 dB.

Sure, you could go up to the next ham band at 24 GHz, but parts and test equipment is starting to get expensive.

There are ambitious amateurs who have received the X-band (8.45 GHz) carrier of the downlink from Juno, at Jupiter - Juno has a 25W transmitter and a 3 meter dish. Mars is a lot easier.