For getting high quality data there is little alternative to doing measurements on empty substrates, preferably of the same batch as those used to produce the films, and preferably of the same dimensions, and measured in the same geometries as your film lateron.

Diamagnetic contributions are essentially temperature independent, while paramagnetic contributions probably are not. Paramagnetic contributions may originate from paramagnetic impurities in the substrate and as such vary in their (chemical) nature and abundance (concentration) between different production batches and producers.

Which level of accuracy is good enough for your purposes depends a lot on the kind of information you wish to obtain from your samples. If, for example, you wish to retrieve information on anisotropy contributions based on how magnetization approaches saturation, then your "calibration" or reference measurements need to be very carefully done. Simply assuming that the slope of m(H) at high fields corresponds exactly to the background you wish to get rid of is then a risky assumption.

Substrate contribution will scale with its mass (volume) and therefore precise weighing of the substrates in use is mandatory. In terms of magnetization (of films, I suppose), the accuracy you can achieve will never exceed the accuracy of you knowledge of film composition and thickness (mass, volume).

Finally, using the wrong tools in sample handling (e.g. iron based tweezers or the like) may severely compromise the reliability of your data.

Dear colleague

You need to measure the sample holder without the sample. Then measure the sample with a sample holder. Thereafter, signal of the sample holder to subtract the signal sample with a sample holder. 

In my opinion this is the best way.

I agree with Dr. Koshkid'ko, however if your curve shows a sharp increase a very low temperatures (caracteristic of paramagnetic) you can fit this part of the curve with a Langevin function. The diamagnetic contribution must be very low in comparison with the ferromagnetic magnetization. See paper attached.

Dear Prof. Gallardo and Dr. Yu,

Thank you very much for your comments.

I can see only diamagnetic contribution at low temperature. As Dr. Yu suggested that measure the sample holder separately and subtract the contribution. But I think contribution from sample holder is very small however mm thick substrate gives large diamagnetic contribution. how to subtract this diamagnetic contribution from raw data?

If you have a saturated hysteresis loop, to remove diamagnetic contribución trom the substrate you simply substrate the slope obtained by linear fitting of the raw data at high field, where sample FM signal should be saturated= constant.

You can measure independently sample+substrate+sample holder, substrate+sample holder, sample holder. If you find substrate+sample holder is diamagnetic (linear), you can do linear fitting and obtain slope m and background offset c. Then use these m and c to subtract diamagnetic contribution in your  sample+substrate+sample holder data

Dear Shalu Pathak,

You may also take into account the material that you use to mount your sample onto the sample holder. Some of the tapes/glues sometimes used for this purpose are diamagnetic. This contribution into your overall signal can be negligible but it is worth checking (for example, several layers of PTFE tape generate quite a noticeable diamagnetic signal in a magnetic field).