For the green and the red curve, it is extremely difficult. For the blue you may connect the more or less straight parts at low and high temperature with a straight or slightly curved line (polynomial).

The chosen baseline for the exothermic peak in the red curve is meaningless. You may use a straight line extrapolating the low-temperature region and integrate the curve against this line.

It is always dangerous, to use separate baselines for the different peaks. In general, the baseline should represent the heat capacity without contributions from latent heat. However, this is not known in many cases.

See, e.g., Article Heat capacity, enthalpy and crystallinity of polymers from D...

Article Differential scanning calorimetry (DSC) of semicrystalline polymers

Article The melting of polymers-a three-phase approach

Thank you for this very enlightening answer, Professor Christoph Schick

I will take note of that principle.

I have requested for the articles you shared. I look forward to reading them and will come back with more questions. Thanks again Prof!

Hi, Caleb Ojo,

this is rather general rule than the answer to your specific problem, but you may use the irreversibility of some phase transformations that may occur upon heating. As Christoph Schick mentioned, in a perfect world you would need a curve of a similar material without contributions of latent heat. This is not easy to obtain, if possible at all. But in some cases you can make a trick: perform the 1st heating, then perform the cooling with the same rate, and perform the 2nd heating. If your sample undergoes irreversible transformations during heating, it will not be observed on the cooling curve (good news), and it will not be observed on the 2nd heating curve, neither (very good news). You can use the 2nd curve as the baseline for the 1st curve.

As I mentioned, this is not always possible, but worth to give a try, if you are not sure of what kind of transformation you are dealing with. Of course, irreversible transformations may mean decomposition of the sample, so the 2nd curve will not be the same as the initial state, but it should be closer to the non-latent-heat curve than your attempts to draw a baseline shown in the plots that you presented.

Good luck with your study.

Jarosław Ferenc

Before you perform any calculations, you should understand the thermal events and their correlated materials. Hence, you would be able to identify and assign each peak respectively, regrdless separated and merged.

In your case, some peaks/or thermal events apparently overlapped, which could lead to possible misleading results or conclusions. This type of calculations were greatly prine to errors, subject to the operators' knowledge, skills, experiences and mindsets.

Thanks. If I am to understand correctly, what you mean is that understanding the typical shape of certain thermal events would increase the likelihood of correct establishment of baseline?

Is this what you mean? Jp Wu

Jarosław Ferenc thank you very much for such an insightful suggestion. I understand what you mean and I believe it would be of great help.

I would put it to the test.

You should establish the mechanism of the thermal events to their peak profile. Even so, it is still impossible to build up a relationship between the materials quantity and the calculated values of thermal events due to overlapping. Baseline correction can minimise the error of selection of, e.g., starting/ending points, but not eliminate the impact of the overlapped events. You can, however, isolate the sharp peak from the broad one by using 2nd derivatives. The results regarding the sharp peaks analysis by using derivatives would be reliable and acceptable.

Be aware that the DSC curve represents the superposition of endo- and exothermic events in the sample. The occurrence of a peak only indicates the dominance of one of the processes but not the absence of the other. A good example are the melting-recrystallizatio-remelting processes in semicrystalline polymers. By varying the heating rate, you can change the number, size, position, and sharpness of peaks in the DSC curve. Therefore peak separation is often difficult and sometimes misleading in identifying the underlying processes.

Dr Jp Wu I once used the Fityk software to deconvolute an overlapping peak into 3 different peaks.

Thanks Professor Christoph Schick

Great suggestion.

I tried these different heating rates for one of my samples (2.5, 5, 10, 20 K/min). I have attached the results. The baseline is wrong based on the conversation we have been having, but please, what can you say about the progression as the heating rate increases? Your comments will be greatly valued.

Deconvolution is a good method to dissolve and separate submerged events. However, in order to do that, you have to fully understand what's happening and how many events are involved and their respective mechanisms.

For instance, why three events (corresponging to three peaks), not 2, 4 or even more? Math can get much more peaks deconvolved with minimum error, but are they right from a scientific point of view?

Jp Wu you make an exceptionally good point! Noted! Thanks much!