Deconvolution of directly precipitating and trap-precipitating electrons in solar flare hard X-rays. II. Compton Gamma Ray Observatory data analysis

Based on the deconvolution method developed in the first paper of this seri es, we present here the data analysis of 20-200 keV hard X-ray (HXR) data f rom the Burst and Transient Source Experiment (BATSE) on board the Compton Gamma Ray Observatory (CGRO) recorded during 103 solar flares in 1991-1995. These are all of the flares simultaneously observed by CGRO with high time resolution (64 ms) and by Yohkoh in hare mode. The deconvolution method ta kes the measured HXR count rates as function of energy and time, I(epsilon, t), and computes the following self-consistently: the electron injection f unction n(E, t), the directly precipitating electron flux n(prec)(E, t), th e trapped-precipitating flux n(trap)(E, t), the fraction of directly precip itating electrons (q(prec)), the electron time-of-flight distance (l(TOF)), and the electron density at the loss cone site of the trap (n(e)). We find that the electron time-of-flight; distances (l(TOF) = 20.0 +/- 7.3 Mm) inf erred with the deconvolution method are fully consistent with those obtaine d earlier using a Fourier filter method. The trap electron densities (n(e) = 10(10.96+/-0.57) cm(-3)) obtained from deconvolving the e-folding decay t imes of HXR pulses (according to the trap model of Melrose & Brown) are fou nd to be statistically a factor of 1.5 lower than those inferred from cross -correlation delays. The fraction q(prec) of directly precipitating electro ns, measured for the first time here, is found to have a mean (and standard deviation) of q(prec) = 0.42 +/- 0.16. Based on this precipitation fractio n, we infer loss cone angles of alpha(0) approximate to 20 degrees-70 degre es and magnetic mirror ratios of R = B-loss/B-inj approximate to 1.2 - 3 (w ith a median value of R-median = 1.6) between the loss cone site and inject ion/acceleration site, assuming an isotropic pitch angle distribution at th e injection site. The TOF distances and mirror ratios constrain magnetic sc ale heights in hare loops to lambda(B) = 10-150 Mm. The fact that this two- component model (with free-streaming and trapped electrons) satisfactorily fits the energy-dependent time delays in virtually all flares provides stro ng evidence that electron time-of-flight differences and collisional scatte ring times dominate the observed HXR timing, while the injection of electro ns appears to be synchronized (independent of energy) and does not reveal t he timing of the acceleration process.