The "Open" compensation measurement measures, as you say yourself,  the stray capacitance of the fixture you built. All three compensation measurements are used to compensate for all the impedances you do not want to measure - stray inductances, stray capacitances and stray conductances. This three element compensation method is quite adequate for most purposes. Consult "Agilent Technologies Impedance Measurement Handbook" (July 2006). With regards  Petr  

Thank you Petr. I have another question, probably you know it. A sample has a resistance of 150 Giga ohms which I am getting from I-V measurement. Can Agillent 4294A analyzer measure the impedance (the  conductance) of the device? I can see that the maximum limiting load resistance (probably for the open circuit) that we can define in the device itself in load compensation is 100 Giga Ohms.

Dear Gyanendra,

If I remember correctly, the specified max. impedance of the 4294A Agilent Impedance Analyser lies around ~1GOhm (100 Hz - 100 kHz - see their maual impedance range table). Now, unguaranteed (with proper calibration and in 4TP cable configuration), one may be able to stretch it to some 10 GOhm (I recall ~50% error at 10 GOhm), but not really above this value.

I would therefore trust more your I-V measurements with well specified resistors. In order to find how high up we could trust Agilent instrument, we usually took known high value resistors ( "pure", calibrated resistors (1, 10, 100, 1000 GOhm), ) and use them as DUT in the Agilent 4294A instrument. The low frequency values of the real part of the measured impedance then started to deviate from the specified values at around 1 GOhm. The usual trend in these type of impedance measurements is that higher the impedance (actually the dc resistance->real part of the measured impedance), lower is the frequency, where you measure it. So my hope for you is that although your resistances are > 100 GOHm,  the measured real part of the impedance at the low frequency limit of 4294A is not saturated yet (at the limit w->0, the real  part of the impedance becomes identical to the I-V dc value resistance) and one can get an estimate of this dc level from the curvature of the real part of the impedance. This is done usually through a R,C,L fit. 

And a last comment. Agilent instruments are not good enough at low frequencies (

I am testing a very small GaN sample (4 sq mm area and .5 mm thickness). Even the capacitance is very low (I am getting 0.8 pF which is still not accurate), the impedance is purely capacitive. The susceptance (B) is 22 times larger than the conductance (G) at 40 Hz and you know the difference increases with increasing frequency. And as you said, the noise level increases highly below 1 kHz. So, I am thinking probably it is not possible to get some charge transfer properties of the  sample with my setup.

Thank you for the valuable information.

If I take the typical data for GaN, I get 0.72 pF for your bulk capacitance Cb which is what you get (I took an average of the dielectric constant of zincblend and wurtzite structure). However , you should be measuring the bulk resistance Rb ~ 125 kOhm if the data Tables can be trusted (conductivity at 300 K ~6-12 [1/Ohm.cm]). With these values I get the ratio of real part to imaginary part of the admittance Y1/Y2 ~ almost 10^5 !! That is clearly too low to be able to measure. The situation is different at  10^7 Hz. There you should be able to measure both Y1=1/Rb and Y2= w.Cb.

All this though by assuming that you have got Ohmic contacts and all you measure is the GaN bulk. If this is not case, then you have got an extra series impedance to your bulk impedance and that is the GaN-electrode interface (depletion region). At low w you measure interface and at high w GaN bulk ( very approximate first order approximation).

So, in conclusion, I still believe that you should be able to measure the dissipative parts of your sample (what you call "the charge transfer properties).

With greetings

Petr 

It is semi insulating GaN. The contacts are ohmic and the bulk resistance we got with IV is probably resonable. The additional capacitance I am getting might be because of fringing as the sample is thick and electrodes/fixture are not very good. The real part of the impedance I get is mostly noise  around 0 which I think is resonable because of noise in phase angle.  A slight plus in 90 degrees makes the real part go negative.

What I was thinking before is the noise might be because the electrodes are not making good contacts with the sample and the dissipative property is not displayed in spectra. But now as I think the resistance is very large, probably I might get only the dielectric properties (capacitance and loss factor) but not the loss because of ohmic conduction.

The motive of the work was to find if IS could be used to study defects in the sample.

Thank you. 

Dear Gyanendra,

I agree with almost everything you say, apart from the real part of the impedance (resistance) beyond 500 kHz.

If your contacts are ohmic, then you measure the bulk GaN. I would plot the data as real and imaginary part of the impedance and look how the real part changes with frequency. If it goes like Z1= Rb/(1+w^2*tau^2), where tau=Rb*Cb, then on the flat part (w*tau < 1), you are measuring bulk Rb and this Rb should be your dc resistance measured by I-V "Ohmic" measurement. If this is not the case and Z1 has different form, Rb is not measurable and any real part is most probably due to dielectric loss, as you correctly point out.

If Rb is too high to be measured, then you have a very suitable situation for studying defects (deep levels) in your system. Here I would suggest that you consult US patent 5 627 479, May 6, 1997, where it is shown how to study these defects using EIS (Electrical Impedance Spectroscopy). This might help you on the way.

Regards

Petr