Do a parametric search on low input current opamps. Some examples :
LPC661, LMC662, LMC6001A, ...

You should look for a amplifier intended for pH measurements. I think, 
NS has one: the E+ is well separated from all other pins and together 
with a guard track connected to E- this amplifier yelds a input 
impedance of (as far as I remember..) some 1E12 ohms.

W.S.

Do you remember the thype of this offset?
How constant is the offset voltage of these opamps? is it suficciant to 
mesure offseton open circuit and subtract, or does it change much over 
time? I thought of a 10kOhm shunt? Mesure Voltage before and the voltage 
drop over shunt. So effective Voltage can be calculated.
How would you design this?

The offset of these OPs is highly temperature dependent, similar to the 
leakage current of a diode. So one has to keep the temperature low and 
constant.

Just a Shunt is usually not a good option. The voltage Drop over the 
shunt will be to large and change the voltage.
The typical circuit is a transimpedance amplifier with one of the extra 
low bias OPs. Choose the feedback resistor as large as resonabe possible 
(e.g. in the 10 M range). For larger currents (e.g. more than 0,1 µA) a 
second range is probably needed.

Hi ulrich
many thanks

I thougt about a solution with a shunt and measuring the voltage before, 
but since I know the current and the shunt value i can calculate the 
voltage drop and suptract it from the mesured Voltage.

with the transimpedance amplifier I would need a resistor of about 1GOhm 
to get a mV Voltage at 1pA so i could even mesure below 1pA. How big are 
leakage currents on pcb etc? Would you recommend such a high resistor? 
And what opamp type has enough input resistance?

Leakage currents on the PCB are a problem. You may want to use extra 
isolation standoffs and have the critical pins free in the air, not on 
the PCB. If one wants the PCB, keep it really clean and use active 
shilds.

The input impedance of MOSFET input OPs is large enough (e.g. 10^12 Ohms 
range) - leakage from internal protection diodes may be a problem at 
higher temperature. You probably won't need 1 G Ohm as feedbach, because 
one can measure less than 1 mV. It should be OK to resolve something 
like 10-50 µV. So something like 10 M to 100 M is enough. This also 
gives more range with one resistor. With the OP its also important to 
look for low drift and low frequency noise, not only for extremly low 
bias currents.

So even something like the old ICL7650 may be an option - not the lowest 
bias, but low drift and LF noise. Here even less than 10 M is OK as a 
resistor.

I had the Idea to use a Fluke 189 Multimeter, unfortunately I cant 
change the Integration time of this device. (around 1/4 sec).
However since this device is a Multimeter and not a precision 
Voltagemeter i think I should have at leaset 1mV in order to mesure 
precisely. Do you see any Problems using a 1G resistor? (for the range I 
can also switch to 100M). I will probably build the circuit without PCB.

Ive seen, that the ICL has a relatively high offest Voltage of 1uV. 
Which affects the Voltage measurement.