The high-frequency (HF) oscillations, which are driven by the electron
beam on the high-potential side of an electric double layer, are inve
stigated in a laboratory experiment. A new HF probe design has made it
possible to achieve a combination of absolute amplitude calibration a
nd spatial resolution which, for such high frequencies, has not been a
chieved before. The HF waves convert about 20% of the beam energy to o
scillations within a region extending from 100 to 400 Debye lengths on
the high-potential side of the double layer. This makes the HF waves
the dominant mechanism of local tapping of the beam energy, and the mo
dified energy and particle balance is investigated. The HF waves propa
gate with an approximately constant phase velocity which is slightly s
maller than the beam velocity. Time-averaged measurements locate them
to a HF region, extending approximately from 5 to 15 cm from the doubl
e layer, with a typical half width of 10 cm for the time-averaged elec
tric field amplitude. Time-resolved measurements show, however, a much
more narrow structure. The envelope of the electric field amplitude h
as a single maximum within the HF region and a typical half width of 1
-2 cm (about one wavelength) along the beam. We calf this envelope 'th
e HF spike'. The position of the HF spike changes in an apparently irr
egular fashion within the HF region and it moves with velocities in th
e range 0-15 km s(-1). The spatial increase and decrease in amplitude
of the wave is exponential over nearly two orders of magnitudes, with
electron-folding distances of only about 5 mm, so the maximum of the e
nvelope is very sharp. The amplitude increase agrees approximately wit
h the growth rate from linear beam-plasma theory and the maximum ampli
tude observed agrees with saturation by beam trapping. The strong spat
ial decrease in the wave amplitude is not understood. It is proposed t
hat the motion of the Hf spike is caused by fluctuations, on the ion a
coustic time scale, both in the growth length acid in the starting poi
nt for wave growth at the double layer, which moves back and forth. Th
is motion is shown to be strongly correlated to the motion of the HF s
pike in a particular case. Amplitude modulations, as observed by a sta
tionary probe, are found on two time scales. On a slower time scale of
typically 2-5 mu s, the motion of the HF spike, together with its lim
ited spatial extent, gives rise to a temporal burst. Within these burs
ts the waves are also modulated on a much faster time scale of 10-30 n
s (the wave period is 3 ns). The HF spike cannot be interpreted as a l
inear superposition of waves with the constant phase Velocity measured
, because this wave packet would have a spatial extention at least ten
times larger than the width of the HF spike.