Differential Scanning Calorimetry (DSC)

Two pans, one ramp, and the difference in between

Put a few milligrams of your material in one tiny aluminum pan, and seal an empty pan beside it as a reference. Drive both up a programmed temperature ramp — say 10 °C every minute — and keep them at exactly the same temperature the whole way. The trick of DSC is that it does not measure temperature. It measures the difference in heat flow needed to keep the two pans on the same ramp.

As long as nothing is happening inside the sample, both pans need almost the same power to warm up, and the recorder draws a flat baseline. But the moment the sample does something thermal — melts, crystallizes, decomposes, cures, dehydrates, or softens through a glass transition — it suddenly needs more or less heat than the empty reference. The instrument tracks that imbalance in milliwatts. Plot it against temperature and you get a thermogram: the material's thermal fingerprint.

Heat flow (mW)        endotherm ↓ (sample absorbs heat)
  │
  │   ___ Tg step ___
  │  /              \____
  │ /  glass→rubber       \         crystallization (exo, sample releases heat)
  │/   ΔCp ≈ 0.3 J/g·K      \      ╱‾╲
  │                          \    ╱   ╲___________
  │                           \__╱                \      melting (endo)
  │                          (cold cryst.)          \    ╱╲
  │                                                  \  ╱  ╲
  │                                                   \╱    ╲___
  └────────────────────────────────────────────────────────────→ Temperature
       Tg                Tc                              Tm

Each feature carries a number. The position of a peak tells you at what temperature something happens (melting point, glass transition). The area of a peak tells you how much energy the event involved (enthalpy of fusion, heat of crystallization, heat of reaction). The shape and sharpness tell you about purity and kinetics. From 2–10 mg of powder you can read out a startling amount of physical chemistry in under an hour.

Endotherms, exotherms, and what the y-axis means

A DSC peak points one of two ways. An endotherm is a process that absorbs heat — the sample falls behind the reference and the instrument feeds it extra power to catch up. An exotherm releases heat — the sample races ahead and the instrument throttles back. The split is exactly the sign of ΔH for the event:

Event	Direction	Typical ΔH	What it is physically
Melting (fusion)	Endotherm	+60 to +250 J/g	Crystalline order breaking; latent heat absorbed
Vaporization / drying	Endotherm	+2260 J/g (water)	Solvent or moisture leaving
Glass transition	Step (no peak)	0 latent (ΔCp 0.1–0.5 J/g·K)	Amorphous chains unfreezing — second order
Crystallization	Exotherm	−40 to −150 J/g	Chains ordering into crystals; latent heat released
Curing / polymerization	Exotherm	−300 to −600 J/g (epoxy)	Covalent network forming
Oxidation / decomposition	Exotherm	variable, often large	Bonds breaking and reforming exothermically

One persistent trap: whether an endotherm points up or down on the chart is purely a plotting convention. Power-compensation instruments historically plot "endo up"; many heat-flux instruments default to "exo up." The peak's identity never changes — melting always absorbs heat — but a misread axis sends students reporting crystallization as melting. Always find the little ↑ exo or ↓ endo label before you interpret a single feature.

Turning peak area into enthalpy

The signal the instrument records is heat-flow power, q, in milliwatts (mJ/s). Because the temperature is ramped at a constant rate β = dT/dt, time and temperature are interchangeable, and integrating power over time gives energy:

Q = ∫ q dt          [joules]
ΔH = Q / m          [J/g]   ← divide by sample mass m
ΔH_molar = ΔH · M   [J/mol]  ← multiply by molar mass M (÷1000 for kJ/mol)

Operationally, you draw a baseline under the peak (where the signal would have been with no event), integrate the area between peak and baseline, and divide by mass. The baseline matters: a sigmoidal or tangent baseline is chosen depending on whether the heat capacity changes across the transition.

The heat-flux signal itself relates to the measured temperature difference ΔT between sample and reference through a calibrated thermal resistance R:

q = ΔT / R          (heat-flux DSC, Tian's equation)

Calibration is non-negotiable. You run a high-purity metal standard — almost always indium (melting point 156.60 °C, enthalpy of fusion 28.45 J/g) — to lock both the temperature axis and the heat axis. Zinc (419.5 °C) and tin (231.9 °C) extend the temperature calibration; sapphire is the standard for absolute heat-capacity measurement.

Worked example: enthalpy of fusion and percent crystallinity

Suppose you scan 5.00 mg of a polyethylene sample and the melting endotherm integrates to 0.430 J (430 mJ) over the peak. The enthalpy of fusion of this sample is:

ΔHm = Q / m = 0.430 J / 0.00500 g = 86.0 J/g

Now compare it against the literature value for a 100 % crystalline polyethylene, ΔHm° = 293 J/g. The fraction of the sample that was crystalline is simply the ratio:

crystallinity = ΔHm / ΔHm° = 86.0 / 293 = 0.294 = 29.4 %

So this polyethylene is about 29 % crystalline and 71 % amorphous — a routine, quantitative result that no microscope or spectrometer hands you so directly. If the same scan also showed a cold-crystallization exotherm of, say, 12 J/g before melting, you would subtract it first (ΔHm,net = 86.0 − 12 = 74 J/g) because that crystallinity formed in the instrument, not in the original sample.

For a pure crystalline drug, the width of the melting peak feeds the van 't Hoff purity analysis: a 0.5 °C melting range implies >99.9 mol % purity, while a 3 °C smear can mean a 2–3 % impurity. The principle is the same colligative melting-point depression a chemist sees in a capillary, but quantified to two decimal places.

DSC vs DTA vs TGA

DSC sits in a family of thermal-analysis techniques. The distinctions matter because each answers a different question, and people routinely confuse them:

	DSC	DTA	TGA
Primary signal	Heat-flow difference (mW)	Temperature difference ΔT	Mass change (mg or %)
Gives enthalpy?	Yes — quantitative ΔH from area	No — qualitative only	No
Detects Tg?	Yes — baseline step	Weakly	No
Detects melting?	Yes (no mass change)	Yes	No (mass constant)
Detects decomposition / loss	Sees the heat	Sees the heat	Sees the weight loss directly
Typical temperature range	−180 to 700 °C	up to 1600 °C	up to 1000–1500 °C
Sample mass	2–10 mg	10–50 mg	5–20 mg
Best for	Tg, Tm, crystallinity, ΔH, cure, purity	High-T phase transitions, minerals	Volatiles, filler content, thermal stability

A practical rule: if the event changes the sample's mass (drying, decomposition, oxidation, off-gassing), reach for TGA — often coupled to DSC as simultaneous thermal analysis (STA) so a single run tells you both how much heat and how much mass. If the event is a pure phase change or transition with no mass loss (melting, Tg, crystallization), DSC is the instrument. DTA is the historical ancestor; modern heat-flux DSC is essentially a calibrated, quantitative DTA.

Heat-flux vs power-compensation hardware

Two engineering philosophies give the same kind of thermogram by different routes:

• Heat-flux DSC. Both pans sit on a single thermoelectric disk inside one furnace. Thermocouples read the small temperature difference between them, and a calibrated thermal resistance converts ΔT to heat flow (q = ΔT/R). Robust, inexpensive, and the most common design. The signal is reconstructed, not measured directly.
• Power-compensation DSC. Sample and reference sit in two separate microfurnaces, each with its own platinum heater and sensor. A feedback loop pumps whatever power is needed to hold them at identical temperature, and the power difference is read off directly. Faster response and a cleaner enthalpy, at higher cost. This is the Perkin-Elmer design lineage descending from Watson and O'Neill's 1962 invention.

The atmosphere matters too. A nitrogen purge (typically 20–50 mL/min) keeps things inert so you see only physical transitions; switching to an oxygen or air purge lets you measure oxidation induction time (OIT) — how long a polymer or oil resists oxidation at a fixed temperature, a key spec for cable insulation and lubricants. Hermetic and high-pressure pans contain volatile samples; an open pan lets moisture escape as a drying endotherm you can integrate to get water content.

Where DSC earns its keep

• Pharmaceutical polymorphism. A drug that crystallizes in two forms can melt at, say, 165 °C and 172 °C with different enthalpies. The wrong polymorph can be less soluble and less bioavailable — the infamous ritonavir Form II crisis cost Abbott a reformulation. DSC is the front-line screen for which crystal form is in the tablet, and for detecting amorphous content that risks recrystallizing on the shelf.
• Polymer identification and processing. Tg sets the use temperature of a plastic: polystyrene softens at ~100 °C, polycarbonate at ~147 °C, PEEK at ~143 °C with melting near 343 °C. The heat–cool–heat protocol reports Tg, Tm, percent crystallinity, and recrystallization behavior — everything an injection molder needs.
• Thermoset cure. Epoxy and composite resins release 300–600 J/g as they cross-link. The cure exotherm's area tells you total reaction heat; running a partially cured sample and integrating the residual exotherm tells you the degree of cure already achieved — quality control for aerospace prepregs.
• Food and lipids. Cocoa butter has six polymorphs (I–VI); good chocolate is tempered to set Form V, which melts just below body temperature at ~34 °C for the right snap and mouthfeel. DSC reads the melting profile that distinguishes a well-tempered bar from a bloomed one.
• Protein stability. Differential scanning calorimetry of dilute protein solutions (a specialized micro-DSC) measures the unfolding endotherm; the peak temperature is the melting temperature Tm of the fold, and the area is the enthalpy of denaturation — central to formulating stable biologic drugs.

Modulated DSC: separating reversing from non-reversing heat

Standard DSC reports only the total heat flow. When two events overlap — a glass transition buried under an enthalpic relaxation peak, or a cure exotherm sitting on top of a Tg — you cannot cleanly read either. Modulated DSC (MDSC) solves this by superimposing a small sinusoidal temperature oscillation (typically ±0.5–1 °C every 40–60 s) on a slow underlying ramp of 2–5 °C/min.

A Fourier deconvolution then splits the total signal into two parts:

total heat flow = reversing + non-reversing
  reversing      → heat-capacity-driven, follows the modulation  (Tg, melting)
  non-reversing  → kinetic, time-dependent                      (cure, relaxation,
                                                                  cold crystallization,
                                                                  evaporation)

The reversing signal isolates the Cp step of the glass transition even when an enthalpic-relaxation endotherm overlaps it; the non-reversing signal isolates the kinetic event. MDSC is what lets you measure a 1 % amorphous fraction in a crystalline drug or pin down Tg in a sample that relaxes strongly — distinctions invisible to conventional scanning.

Common misconceptions and pitfalls

• Reporting Tg from the first heat. The first scan is contaminated by thermal history, residual solvent, processing stress, and enthalpic relaxation. Always quote Tg from the second heat after a controlled cool, unless the as-received history is exactly what you are studying.
• Confusing Tg with a melting point. Tg is a second-order transition — a step in heat capacity with no latent heat — that occurs only in the amorphous fraction. Melting is first-order, with a real peak and latent heat, in the crystalline fraction. A semicrystalline polymer shows both, at different temperatures.
• Ignoring heating-rate effects. Kinetic events (Tg, decomposition, cold crystallization) shift with ramp rate; faster scans push them to higher apparent temperatures and inflate peak height. A reported transition temperature is meaningless without the heating rate beside it.
• Bad baseline, wrong ΔH. The integrated area depends entirely on where you draw the baseline. A sloping or curved baseline under a broad peak, or a heat-capacity change across the transition, can change the reported enthalpy by 10–30 %. Use the correct sigmoidal/tangent baseline and always run an empty-pan blank.
• Sample contact and mass. Poor thermal contact (a curled film, a loose powder) broadens and shifts peaks. Too much sample saturates the detector and smears resolution; the 2–10 mg window exists for a reason. Crimp the pan flat against the sensor.
• Over-reading purity. The van 't Hoff melting-point analysis assumes a eutectic-forming impurity and breaks down above ~2–3 mol % impurity or for solid solutions. It is a screen, not a substitute for chromatography.