Sample Selection Criteria

We searched through the online RHESSI flare list (from 12-Feb-2002 through 02-May-2003) for appropriate limb flares, using critera similar to those introduced by Masuda (1994)and used by Petrosian et al. (2002). These are:

• Heliocentric longitude $\geq 70$ degrees . This provides sufficient angular separation between footpoint and looptop sources. The heliocentric (x, y) coordinates in the RHESSI flare list were converted into heliographic (longitude, latitude) coordinates. For flares whose locations are not available in the list, we obtained the locations by making full disk RHESSI images, usually in the 12-25 keV energy channel about the peak time, and use the position of the brightest pixel as the flare location.

• Peak count rate $\geq 30$ per second per detector in the 12-25 keV channel , allowing sufficiently good imaging quality.

We then carefully examined the sample flares satisfying these criteria and eliminated those with strong particle events, severe pileup or decimation, poor data quality, etc. 22 flares are included in our final sample. For each event, we performed imaging spectroscopy and lightcurve study as follows.