Your transistor selection looks WAY off!
The cmt70n03 is a 1-3v threshold. You need significantly MORE voltage than the threshold to turn it on all the way, and turn it on quickly. Failure to do either will result in lots and lots of heat.
Your schematic shows a diode protecting the PIC, so even assuming it's a Schottky, that's 3.9v-2.2v Vgs. You need a FAR lower Vth than this! This is at high risk of current going into saturation region, which will probably let the magic smoke leak out.
Also the capacitance is a huge issue. The heat generated during switching will typically be much greater than that generated due to I^2*Rds-on heating.
In fact, bigger current ratings typically mean proportionally higher capacitance (given similar construction), and thus the part actually can handle LESS current during switching with a modest gate drive like a PIC than one with a lower rating.
Another thing, the current ratings across different packages is funny. See the rating is based on how much heat it'll take to damage the junction. Bigger packages get higher ratings not because they've got lower resistance, but because of greater dissipation. Consequently, they're often higher rds-on and less efficient than the same current rating in a much smaller package.
Another significant factor, the ripple rating of the input cap. Tantalums typically have "high" ripple current rating for their size, but in these tiny caps it's still pretty small. I've had tantalums several times larger than these in flashlight drivers that can only handle like 200mA PWM'ed off a lithium cell before overheating. I found better ones which can handle more, but the tech's pushing it. Ceramic MLCC typically has MUCH lower ESR, and there are some darn high capacitance ones in tiny packages now.
. By the way i am building a 5x SSC P7 in an Ultrafire wf-500 host, direkt driven by 5x 18650, sunday i´ll push the button- if i am still alive, i´ll post a picture...
