There can be trivial and non-tirvial reasons. The trivial reason may be that the thermocouple never measures exactly the sample temperature, but its own temperature. If you use a heating rate above 2 degC/min, you can easily get different readings for the same sample temperature on heating and cooling. (Depending on the size of the sample holder it can be fairly high). The non-trival reason is that in the case of polymers there is significant difference between "melting point" and "freezing point" - better to say, crystaliization temperature. If you measure by DSC you can see that crystallization occurs at lower temperature and the peak is more symmetric than on heating. This is due to crystallite size dsitribution and on the nucleation conditions. (In the prsence of nucleating agents th crystallization begins earlier). Post-crystalization after cooling may be initiated by annealing below, ut close the nominal melting point. I would suggest to measure the same amtarial by DSC (ith helps only with crystallization. It also detects Tg as a hump in th cp curve, but sometimes that is smeared).

Is this effect is reproducible?

What kind of material (ferroelectrics?) and in what form - crystal, ceramics?

dear Gyorgy banhegyi sir,

your explanation is good. pls can u attach the reference related to DSC.

Dear Venkat

see e.g. the following intorduction:

http://www.masontechnology.ie/x/Usercom_11.pdf