ultra small current amplification, please help

Dear Peng,
I have some advice on this, although you must check really what is best in your application. The best performance option by far, in all respects excepting "speed", is to use a reed delay. There is no leakage current, no noise, no charge injection. I think if you can tolerate a long delay in the switching, a reed relay may be ok. The switching action itself is fast, at least for mercury wetted contacts. The delay may be long but jitter in the delay probably much less, if that matters. Switching at 0.5 Hz should be ok as far as relay lifetime, at least for some applications. If you can somehow tolerate the timing, and are not prohibited by vibration or external magnetic fields or anything like that, it may be a good solution.
Another way I suppose will be to use a diode bridge with a floating drive circuit, e.g., use a photovoltaic isolator to provide static reverse bias, and transformer coupling for the forward bias. Duty cycle for forward bias will be low, but maybe that is ok for your application?
Well, I'm just offering these ideas in case they are helpful. I happened to see your post and the question is an interesting one, and both reed relays and diode bridges have been very useful in somewhat similar applications in my past experience.
Good luck with your design challenge,

Gerard Visser

p.s. Perhaps you should also consider a third way: If you will digitize the output continuously and do some processing to make your measurment of interest, maybe you don't have to reset to zero particularly? If you instead only need to keep the integrator out of saturation, you could pump some charge in to offset the accumulated charge (either with a diode/capacitor arrangement, or a voltage pulse through a high volaue resistor)?