It is important to have a representative sample size when conducting TGA analysis of cementitious or other multi-phase materials. At NIST, we generally tried to use our large volume TGA pans and get a sample size close to or exceeding 100 mg.

Also, results often depend on from which temperature you are measuring your final weight loss. In your figure 7, one can observe that already by 100 C, when the sample should be losing mostly free water, samples 5 and 6 have much less mass loss than those of 1-4. Perhaps these samples were allowed to "dry out" more prior to being loaded into the DSC. You can try renormalizing your mass loss results using the 100 C masses of each specimen as the new basis. This should definitely get rid of some of the variability between specimens.

Regarding the heat flow plot, it would be better to plot the y-axis as heat flow per unit mass as otherwise it is difficult to compare the various curves.

From the mass loss curves, the mass loss between 400 C and 500 C is usually attributed to water lost from calcium hydroxide and while it is difficult to tell exactly from viewing the curves, the six specimens appear to have fairly similar absolute mass losses between 400 and 500 C. This would suggest that the free water loss (prior to 100 C) may indeed be biasing the subsequent cumulative mass loss curves and normalizing based on the mass measured at 100 C may be a good way to go.

Good luck,

Dale

D. P. Bentz thanks for the suggestions.

I have plotted the TGA graph as suggested, renormalizing the graph reduces the variability between the samples. I am not sure how to plot heat flow per unit mass. Is it same as DTG?

I have tried DTG curve for all samples and seems there is an additional peak (could be gypsum or monosulphate) with ettringite for samples 1 to 4. It might be the reason for less weight loss for sample 5 and 6. But the XRD shows gypsum in all samples.

With the avilable TGA equipment, I cannot go more than 35 mg, Could you please recommend some methods to minimize error due to small sample size.

Regards

Saurabh

In addition to the great advice from D. P. Bentz , I reccommend you to be really careful on the previous sample treatment before you introduce in TGA. Focus on what are you doing to stop the hydration, if maybe it was different for samples 5 and 6 and maybe, as D. P. Bentz says, you took some combined water out during the drying of the samples appart from the free water removal.

Susana G. Sanfélix , D. P. Bentz

thank you for the suggestions, I will repeat my analysis with more attention to restrict the hydration and keep the conditions identical before the TGA measurements.

regards

saurabh

Usually, if you cannot run a larger sample, the solution is to run multiple samples and report the average result.

For the heat flow, I was just suggesting that the results be normalized simply by the initial sample mass (35 or 50 mg, respectively). That should reduce the variability amongst the curves for your six specimens.

Dale

Also, in the way of perhaps useful information, in 2018 (so relatively recently), ASTM published a standard test method for performing thermogravimetric analysis of hydraulic cement that may have some good insights:

ASTM C1872 - 18e2 📷

Standard Test Method for Thermogravimetric Analysis of Hydraulic Cement

Dale

Some devices need pre-test of stabilization before running the program you set in tga and dsc. This can result in the fifference of initial wt loss and so on significantly. Multiple sampling tests are one of the best practices to obtain the representative data.