A loop flare that occurred on 22 April 1993 near the disk center is examine
d using the Yohkoh Hard X-ray Telescope (HXT). We specifically looked into
the faint early phase of the flare prior to the start of the strong impulsi
ve phase. The pre-impulsive phase, though weak in intensity, is expected to
contain essential clues to the mechanism of loop flares according to the c
ausality principle, but it has not received attention previously, probably
due to the insufficient dynamic range and cadence of observations by the in
struments on earlier satellites. Observations with Yohkoh/HXT can clarify w
hat occurs in this phase. This flare, like many other flares of this type,
shows a relatively weak emission with a smooth and gradual increase during
this pre-impulsive phase, followed by impulsive bursts, and then turns into
a smooth decay phase without impulsive bursts. First, we found that the sp
ectrum for the initial smooth rise part is consistent with a thin-thermal s
ource at a temperature around 80 MK. Imaging of this phase in the HXT/L and
M bands shows a single source between the footpoint sources that will come
up in the impulsive phase following this phase, suggesting that this hyper
hot source is located at a high part of the loop between the footpoints, si
nce this flare takes a form of a loop. Furthermore, as we go up to the earl
iest times of the flare before this 'hyperhot' source phase, two fainter so
urces are found near the footpoint sources that will appear later in the im
pulsive phase. The spectra of these sources at this earliest time of the fl
are, in contrast to the 'hyperhot' source, cannot be determined from the HX
T because the instrument was not in flare mode, and HXT/M1, M2, and H-band
data are, unfortunately, not available at this very initial time. We can gu
ess, however, that they are also of thermal character because the time prof
ile is smooth without any spikes just as in the following `hyperhot' therma
l phase, and in the post-impulsive `superhot' thermal phase coming up much
later. These findings suggest that there is an important, and probably dyna
mic, early phase in loop flares that has been unnoticed in the still dark p
re-impulsive phase, because the very early footpoint sources change into th
e loop top source in a matter of 20-30 s, comparable to the dynamic Alfven
time scale. Some implications of our new findings are discussed.