Thermal Analysis: DSC, TGA & DMA

What thermal analysis tells you

Three techniques answer three different questions about a polymer, and together they cover most of what you need to characterize a material. DSC measures heat flow, so it sees every transition that absorbs or releases energy: the glass transition, crystallization, and melting, plus the enthalpies behind them. TGA measures mass, so it sees thermal stability and, by watching what is lost and when, the composition: moisture, plasticizer, polymer, carbon, and inorganic filler. DMA measures mechanical stiffness under an oscillating load, so it sees the same transitions DSC does but through their effect on modulus and damping, and it is far and away the most sensitive way to find a weak glass transition.

Every chart below is live: change the sample, the heating rate, the atmosphere, or the test frequency and watch the curve and the read-out values respond the way a real instrument would. The numbers behind the demos are drawn from instrument-maker application notes and the polymer literature; they are teaching approximations, not a substitute for running your own standard.

Which technique answers your question?

Pick what you want to measure, and the technique (or techniques) that give it light up.

DSC — Differential Scanning Calorimetry

DSC records the difference in heat flow between your sample and an empty reference pan as both follow the same temperature program. A step in the baseline is the glass transition (T_g), where the amorphous fraction gains segmental mobility and its heat capacity jumps. A peak is a first-order event: an endotherm (melting) absorbs heat, an exotherm (crystallization) releases it. A quenched semicrystalline polymer like PET shows a cold-crystallization exotherm on heating, between T_g and melting, as frozen chains finally crystallize. Because a crystallization exotherm and a melting endotherm point in opposite directions, the plot is meaningless without an endo-up or endo-down arrow.

The first heat carries the sample's processing history, including a small enthalpy-relaxation overshoot on the T_g step if the glass has aged. A controlled cool followed by a second heat erases that history, which is why intrinsic values are quoted from the second heat. Percent crystallinity comes from the melting enthalpy: X_c = (ΔH_m − ΔH_cc) / ΔH_m° × 100, subtracting the cold-crystallization enthalpy so you count only crystals that were there before the scan.

On the numbers: tabulated T_g and T_m vary with sample and method, and because the glass transition is at least partly kinetic, the T_g also depends on measurement rate (a faster scan reads a higher T_g). The transition is not a sharp point but the cessation of long-range segmental motion, spread over a modest temperature range for chain segments of different lengths (about 5–20 chain atoms), so which temperature you read depends on how fast you probe it. The cooling rate that formed the glass matters as well: a faster quench freezes in a higher T_g, so a reheat scan also carries the history of how the sample was cooled. Standard reference tables (Brandrup's Polymer Handbook, as tabulated in Odian's Principles of Polymerization) list PET at T_g 61 °C with a crystalline T_m of 270 °C, while a 10 K/min DSC scan of quenched PET shows an apparent T_g around 75–80 °C and a melting peak near 250 °C. The heating-rate slider below shows why the T_g figure is method-dependent.

TGA — Thermogravimetric Analysis

TGA weighs a few milligrams of sample continuously as it heats, plotting mass as a percent of the start. Each downward step is something leaving: adsorbed moisture below ~150 °C, plasticizer or process oil in the 150–350 °C range, then the polymer backbone decomposing in the largest step. The first derivative, DTG, turns each step into a peak whose apex is T_max, the temperature of fastest mass loss, and resolves overlapping steps a plain curve would blur.

The purge gas is the trick that turns TGA into a compositional tool. Under inert N₂, carbon (polymer char and carbon black) survives, so the residue is carbon plus inorganic filler. Switch to air and that carbon burns off as CO₂, leaving only mineral ash. The gap between the two residues is the carbon content. A carbonate filler gives itself away too: CaCO₃ decomposes to CaO near 700 °C, losing exactly 44% of its mass as CO₂, so its loading is back-calculated from that step.

DMA — Dynamic Mechanical Analysis

DMA applies a small oscillating deformation and measures the stiffness, split into an in-phase elastic part, the storage modulus E′, and an out-of-phase viscous part, the loss modulus E″. Their ratio, tan δ = E″/E′, is the damping. On heating, E′ sits on a stiff glassy plateau near 10⁹ Pa, falls three to four decades through the glass transition, and settles on a lower rubbery plateau. That drop is the most sensitive T_g measurement there is.

DMA reports T_g three ways, and they do not agree: the onset of the E′ drop (lowest, the conservative engineering limit), the E″ peak (middle), and the tan δ peak (highest, most often quoted), typically spread 10–25 °C apart, so a DMA T_g is meaningless without saying which marker and what frequency. Because T_g is kinetic, raising the frequency shifts the whole transition to higher temperature, about 6 °C per decade. And for a crosslinked network the rubbery plateau is set by crosslink density through E′_rubbery = 3νRT, so DMA doubles as a way to measure how tightly cured a thermoset is.

Reading the three side by side

Traps worth remembering. A first-heat DSC T_g includes processing history and an aging overshoot; quote the second heat. Forgetting to subtract ΔH_cc overstates crystallinity. A TGA “decomposition temperature” is ambiguous (extrapolated onset, T_5%, and first deviation all differ), and every temperature shifts with heating rate and sample mass. A melting point is convention-dependent in the same way: the classic T_m is the extrapolated onset of melting, while DSC software usually quotes the endotherm peak, so a reported T_m depends on which one is meant. A DMA T_g is not comparable between labs unless the marker and frequency match, and the tan δ peak already overstates the usable service temperature.

Worked examples

Each of these reproduces, by hand, a number the interactive read-outs above give you.

1. Crystallinity from a DSC scan. A quenched PET first heat shows a melting enthalpy ΔH_m = 35 J/g and a cold-crystallization exotherm ΔH_cc = 30 J/g; the perfect-crystal reference for PET is ΔH_m° = 140 J/g. Subtract the cold crystallization so you count only crystals present before the scan:

A clean HDPE melt with ΔH_m = 200 J/g and no cold crystallization (ΔH_m° = 293 J/g) instead gives X_c = 200/293 = 68%.

2. Composition from a TGA compositional run. A filled compound leaves 16.1% residue after a full inert (N₂) run and 10.1% after a full air run, and shows a distinct CO₂-loss step near 720 °C that removes 7.9% of the original mass. The inert residue keeps carbon plus mineral; the air residue is mineral only; the carbonate is fingerprinted by its 44% CO₂ loss:

3. Crosslink density from a DMA rubbery plateau. The default network on the DMA plot reads a rubbery storage modulus E′ ≈ 1.2 MPa on the plateau ~40 °C above T_g (T ≈ 373 K). Ideal rubber elasticity, E′ = 3νRT, inverts to the crosslink density ν and the molar mass between crosslinks M_c (with ρ ≈ 1 g/cm³):

Tighten the cure and E′, ν, and T_g all rise while M_c falls; loosen it and they move the other way. Drag the crosslink-density slider to watch it.

Related tools

Predict a copolymer or blend T_g before you measure it with the Tg Predictor (Fox equation). Turn a molecular-weight distribution question into numbers with the GPC Calibration Converter and the GPC Peak Interpretation guide. Every term here (T_g, T_m, crystallinity, storage modulus, crosslink density) is defined in the Glossary.