Thermal Analysis: DSC and TGA

Stop staring — start heating

The first three guides in this rung all gave you ways to freeze a material and look at it. Metallography under the light microscope shows grains on a polished, etched face; the SEM and TEM push that to fracture surfaces and single dislocations; X-ray diffraction reads out the crystal fingerprint. Every one of them is a snapshot — the material held perfectly still while you photograph it. Thermal analysis is different in spirit, and that difference is the whole point. Instead of stopping the material and staring, you push it: turn the temperature up at a steady, controlled rate and watch what the material does as it warms.

Transitions that are completely invisible in a still image announce themselves loudly the moment you heat through them. A polymer softens as its chains unfreeze; a metal precipitate forms; a solid melts; a thermoset cures; a plastic decomposes and gasses off. Each of these is a change of state or structure, and each one either swallows heat, gives out heat, or changes the sample's weight as it happens. Thermal analysis is simply the family of instruments that record those signals against temperature. Two of them do most of the work: differential scanning calorimetry (DSC), which watches heat, and thermogravimetric analysis (TGA), which watches weight.

DSC: counting heat, milligram by milligram

Picture two tiny metal pans side by side on a heater. One holds a few milligrams of your sample; the other is empty — the reference. A program ramps both up together, typically at 10 degrees C per minute. The instrument constantly measures the DIFFERENCE in heat flow needed to keep the two pans marching at exactly the same temperature. When nothing is happening inside the sample, both pans need almost the same trickle of heat, and the trace is a flat, boring baseline. That flat line is the whole trick: it means the machine is watching for the moment the sample suddenly needs more or less heat than its empty twin.

When a transition arrives, the baseline breaks. Melting is endothermic — it soaks up latent heat — so the sample pan suddenly demands a burst of extra heat, and the trace swings into an endothermic peak. Its position on the temperature axis is the melting temperature (Tm), and the AREA under it is the heat of fusion, in joules per gram. Crystallisation and the curing of a thermoset resin go the other way: they are exothermic, releasing heat as chains lock into order or crosslinks snap shut, so they show as a peak in the opposite direction. A single first-heat DSC of an uncured epoxy shows a big cure exotherm you will never see again once the resin is set — proof the part is fully cross-linked.

The subtlest and most important feature is the glass transition, and it does NOT look like a peak. At the glass transition temperature (Tg) an amorphous polymer goes from frozen-glassy to floppy-rubbery — but nothing melts, no latent heat is swallowed, no new phase forms. All that changes is the heat capacity: wobbling chains need a little more heat per degree than frozen ones. So Tg shows up not as a spike but as a small STEP in the baseline, a shift to a slightly higher level. That is why a beginner scanning a plastic sees an obvious melt peak and almost misses the gentle Tg step — yet for a rubbery or amorphous polymer that step is the single most important number on the whole trace.

Reading a DSC trace — with real numbers

Put the three features together on one scan of a semicrystalline plastic and you can read its whole thermal biography left to right. First a step (Tg), where the glassy chains loosen. Then, a bit hotter, a small exothermic dip called cold crystallisation — chains that were quenched into disorder finally have enough wiggle to snap into crystals, releasing heat. Finally, hottest of all, the big endothermic melt peak (Tm) as those crystals come apart. A fully amorphous plastic like polycarbonate shows only the Tg step and never a sharp melt; a highly crystalline one shows a strong melt; a thermoset shows a Tg but no melt at all, because its crosslinks refuse to let the chains flow.

  DSC of a semicrystalline polymer   (heating 10 deg C/min,  endo = UP)

  heat  |                                          .-.
  flow  |                                         /   \      MELT  (Tm)
        |          Tg = a STEP                    /     \     endotherm: it ABSORBS
  endo  |          (not a peak)                  /       '--  the heat of fusion
   up   |     __________                         /            (peak AREA = dHf)
  base .|.__./          \___          .---.      /
  line  |  (Cp jumps as      \       /     \    /   cold crystallisation
        |   glass->rubber)     \_____/       \__/    exotherm: RELEASES heat
        +--------------------------------------------------------> T (deg C)
              75                     130               255

Now a tiny worked example, because the area under the melt peak is not just decoration. To find percent crystallinity you divide the sample's net heat of fusion by the heat of fusion a 100-percent-crystalline version of the same polymer would give. Say a piece of PET shows a melt endotherm of 45 J/g but also a cold-crystallisation exotherm of 5 J/g; the net is 45 - 5 = 40 J/g. A perfectly crystalline PET would release about 140 J/g. So crystallinity is roughly 40 / 140 = 0.29, about 29 percent — and that single number governs the plastic's stiffness, clarity, and barrier to gases. One honest caution: the very first heat carries the part's processing and ageing history baked in. Labs routinely heat once to erase that memory, cool at a known rate, then read Tg and Tm off the cleaner SECOND heat.

TGA: weighing what boils and burns away

TGA swaps the calorimeter for a scale. The sample — again just milligrams — hangs from an ultra-sensitive microbalance inside a furnace, and as the temperature ramps the balance records the mass continuously. Anything that leaves the sample shows as a downward step in the mass curve: first loosely held moisture near 100 degrees C, then any plasticiser or solvent, then at higher temperatures the polymer chains themselves cracking into volatile fragments — decomposition. The temperature where that big drop begins is the material's thermal-stability limit, a genuinely important safety and processing number. Because each step's height is a straight weight percent, TGA reads out composition directly, without ever taking a picture.

1. Room temperature to about 150 deg C: the mass slips by roughly 1-2 percent as adsorbed moisture is driven off. On a hygroscopic nylon this alone can matter for moulding.
2. About 350 to 470 deg C under inert nitrogen: a big ~68 percent step as the nylon chains decompose into volatile fragments. Where this step BEGINS is the practical upper service and processing temperature.
3. Switch the purge gas to air near 600 deg C: any leftover carbon char now burns off, separating combustible residue from the truly inorganic part.
4. Residue at 700 deg C: about 30 percent remains and refuses to leave — the glass fibre, which does not decompose. That final plateau IS the filler content, confirming the 30-percent-glass spec.

Which tool — and the honest limits

So the two split the world cleanly. DSC measures energy, so it hands you the temperatures and heats of transitions: Tg, Tm, crystallisation, cure — and it is not only for polymers. On a metal, DSC catches precipitation as an alloy age-hardens, the release of stored energy on recrystallisation, and the peaks of precipitation strengthening or tempering. TGA measures mass, so it hands you composition, moisture, filler and ash content, and thermal-stability onset. Because they ask different questions, they are often built into one instrument that runs both at once on a single sample — simultaneous thermal analysis, or STA.

Now the honesty, because thermal analysis is easy to over-read. Both tools tell you WHEN a thing happens (at what temperature) and HOW MUCH (how much heat or mass), but never WHAT it is chemically. A weight-loss step does not name the gas coming off; to identify it you couple the TGA outlet to an infrared or mass spectrometer — evolved-gas analysis. A DSC peak marks a transition, not a molecule. Worse, Tg itself is rate-dependent: heat faster and Tg reads a few degrees higher, because the glass transition is kinetic, not a true equilibrium phase change. So you ALWAYS report the heating rate, and you never compare a Tg measured at 20 deg C/min against one at 2.

Two last cautions and a link forward. The sample is a speck of a few milligrams, so if the part is inhomogeneous — a welded joint, a composite laminate, a filled moulding with settled fibre — one tiny pinch may not represent the whole; sampling is part of the answer. And the numbers mean nothing without calibration: you check the temperature and heat scales against a standard whose melting point and heat of fusion are known exactly (indium melts at 156.6 deg C with about 28.5 J/g), or your peaks are just wiggles. Neither DSC nor TGA ever shows you a picture, so pair them with the imaging of guides 1 and 2 and the crystal fingerprint of guide 3. In the QC lab a DSC Tg/Tm fingerprint verifies an incoming plastic is the right grade, a cure exotherm proves a composite is fully set, and a TGA reads filler content and decomposition onset straight onto a safety datasheet. Guide 5 steps all the way back to non-destructive testing and the real question underneath this whole rung: which tool for which question.