At the very start of the impulsive phase of two solar flares the temperature derived from medium-energy (~ 16 keV) X-ray countrates was observed to rise abruptly, by several times 10 7 K above the temperature derived from low-energy X-ray (~ 7 keV) countrates. The difference between the two temperatures relaxed to zero thereafter, quasi-exponentially, with a characteristic time of ~ 1.5 min. This differential temperature variation appears to mimique the differences between the ionic kinetic and the electron temperatures derived from spectral observations (Figures 1 and2).
These observations are explained in a quantitatively supported model of the flare kernel (Figure 4) in which the kernel is heated by electron beams from above. The low-energy electrons are stopped above the kernel and only the medium and high energy electrons penetrate down to the top of the chromosphere, causing heating of the chromospheric gas to about 50 MK, and ablation ('evaporation'), leading to the abrupt formation of a superhot flare kernel and a likely superhot 'dome' above it (Figure 4), through which gas rises up and spreads out convectively, while cooling down in approximately the same time (45 s). The heating process lasts only for a few minutes. The difference between the Doppler temperature and the electron temperature derived from line intensity ratios or from low energy countrate ratios is ascribed to truncation of the tail of the electron energy distribution in the kernel. The kernel is about 2500 km deep; H~ emission is radiated by a thin layer at its basis.