One of my interests is earth-mode (through the ground) communications at VLF frequencies. Most of my tests have been done around 8.97kHz but I have also tested at frequencies from 1-18kHz. Best DX so far is 6km transmitting with 5W into earth-electrodes 20m apart at home. One electrode was at the bottom of the garden and the other is a connection to the copper central heating pipes in the home. At the receiving end, a 80cm diameter 30 turn resonated loop feeds a high impedance FET amplifier. This fed a netbook PC running Spectran software. The beacon transmitter sends either 10wpm CW or QRSS3 or QRSS30. At 6km the QRSS3 signal is a good signal at certain locations. The propagation is by utilities assistance i.e. the signal is helped by buried pipes and cables, but it is not clear which ones or why the signal gets so far. I find it almost beggars belief than such low power can be detected so far away.
In recent months I have been very ill and I have been unable to do any field work. Since Aug 2013 I moved QTH, but I have yet to do proper VLF earth-mode tests from this new QTH.
This is a picture of the earth-mode VLF transmitter. A K1EL keyer IC feeds a TDA2003 5W audio amplifier which is matched to the earth-electrode antenna with a 3C90 toroid transformer.
The actual radiated power from this set-up (the main propagation is by conduction and induction) was calculated as just 37nW. About 20dB more is needed to get a signal that can be detected at around 100km or more using weak signal modes in extremely narrow bandwidths. This could be achieved by increasing the earth-electrode baseline and by increasing the power output.