Thermogravimetric analysis (TGA) measures the mass of a sample continuously as it is heated (or held isothermally) under a controlled gas atmosphere, producing a curve of mass versus temperature or time. Because mass only changes when material actually leaves the sample — as vapor, gas, or combustion products — a TGA curve is a direct, quantitative record of what a material loses and at what temperature it loses it. This guide covers how to set up and run a scan (crucible, purge gas, heating rate), how to read the resulting curve and its derivative, and how to convert a series of mass-loss steps into a compositional breakdown, worked through with real numbers.
What a Thermobalance Measures
A TGA instrument (a thermobalance) suspends a small sample — typically 5–20 mg — in a crucible from a microbalance arm, inside a furnace purged with a controlled gas flow. As the furnace temperature is ramped according to a programmed method, the balance continuously records mass. The raw output is a TGA curve: mass (as a percentage of the starting mass) plotted against temperature or time. Because the balance itself does not identify what is being lost — only that mass left the sample — TGA is normally interpreted alongside knowledge of the sample’s likely composition, and often alongside a complementary technique such as evolved-gas analysis or FTIR/mass spectrometry coupled to the furnace exhaust when the identity of a given mass-loss step needs independent confirmation rather than inference from temperature alone.
Method: Setting Up and Running a Scan
- Choose the crucible. Crucible material affects both chemical compatibility and thermal behavior; see the comparison table below. Tare the empty crucible on the instrument’s own balance immediately before loading — do not rely on a separate analytical balance, since even small handling differences between instruments will show up as an offset in percentage mass-loss values. (If a separate balance is used to pre-weigh material, its calibration matters just as much here as it does for any other quantitative weighing step — see this site’s guide to analytical balance calibration and weighing technique.)
- Load and record the sample mass. Spread solids in a thin, even layer rather than a mounded pile — a thick or uneven bed slows heat and gas transfer through the sample, broadening and shifting mass-loss steps to higher apparent temperatures. Record the exact starting mass; every later percentage in the analysis is calculated against this number, so an imprecise starting weight propagates through the entire curve.
- Select the purge gas and flow rate. Inert gas (nitrogen or argon) suppresses oxidation and isolates thermal decomposition; an oxidizing gas (air or oxygen) is used deliberately when the goal is to quantify combustible material such as carbon black or residual carbon. Typical purge flow rates run roughly 20–100 mL/min — consult the instrument manufacturer’s guidance, since flow rate affects both how quickly evolved gases are swept away from the sample and how much the buoyancy artifact (below) shifts the baseline.
- Set the temperature program. Define the starting temperature, the final temperature, the heating rate (or a segmented program with different rates or isothermal holds at specific points), and any planned gas-switch point. See the heating-rate trade-off section below before defaulting to the fastest rate the instrument allows.
- Run a blank first if precision matters. An empty-crucible blank run, using the identical method and gas flow, captures instrument-specific baseline drift (buoyancy, furnace expansion) that can then be subtracted from the sample run — this matters most when a mass-loss step is small relative to total baseline drift.
- Run the scan and inspect the raw curve before analyzing it. Check for a smooth, monotonic mass loss within each step and a flat, stable baseline before and after — noise, spikes, or a non-returning baseline usually indicate a physical problem (sample splattering, crucible contact issues, gas-flow interruption) rather than a real compositional feature, and the run should be repeated rather than interpreted as-is.
Crucible Selection
| Crucible material | Typical use | Trade-offs |
|---|---|---|
| Platinum | High-temperature work, samples requiring an inert, non-reactive surface | Excellent thermal conductivity and chemical inertness for most materials, but reusable platinum can be attacked or catalytically affected by certain samples (some metals, halogens, phosphorus compounds, molten salts) — check compatibility before use, and dedicate crucibles to specific sample classes where contamination is a concern. |
| Alumina (ceramic) | General-purpose routine analysis, the most common default | Inert to most samples, tolerates high temperatures, considerably cheaper than platinum, available open or with a pierced lid to slow vapor escape from volatile samples — but more brittle and less thermally conductive than platinum. |
| Quartz or borosilicate glass | Lower-temperature work only | Inexpensive and inert, but limited by the glass’s own softening point — not suitable for scans that extend into the high-temperature decomposition or ash-determination range. |
Purge Gas Selection
| Atmosphere | What it isolates | When to use it |
|---|---|---|
| Inert (nitrogen, argon) | Pyrolytic/thermal decomposition without combustion | Default for determining moisture, volatiles, and the true thermal decomposition profile of organic or polymeric material, without combustion reactions confounding the mass-loss steps. |
| Oxidizing (air, oxygen) | Combustion of carbon-containing residue | Used either for the entire scan, or switched in partway through a method (after the inert segment) specifically to burn off carbon black or residual char and isolate a true inorganic ash value at the end of the run. |
Heating Rate: The Resolution-versus-Throughput Trade-off
Heating rate is the single parameter with the largest effect on how easy a curve is to interpret. Slower rates (on the order of 1–5°C/min) give the sample more time to reach thermal equilibrium at each temperature, which sharpens mass-loss steps and better separates events that occur close together — at the cost of a much longer run. Faster rates (commonly 10–20°C/min for routine work, and considerably higher for rapid screening) shorten the run but introduce thermal lag between the programmed furnace temperature and the sample’s actual temperature, which broadens each step and shifts its apparent onset and peak to a higher recorded temperature than the same event would show at a slower rate. This is a kinetic effect, not an instrument error, and it is the reason published decomposition temperatures for the same material can differ meaningfully between methods run at different rates — always report the heating rate alongside any temperature value taken from a TGA curve, and use the same rate when comparing curves between samples or against a reference method.
Reading the Output: The TGA Curve and Its Derivative (DTG)
The primary curve plots mass, usually as a percentage of the starting mass, against temperature. Most analysis software also generates the derivative of that curve with respect to temperature or time — the DTG (derivative thermogravimetry) curve — which converts each mass-loss step on the primary curve into a peak. DTG peaks make overlapping or subtle steps far easier to see and to bound than the primary curve alone, and the temperature at a DTG peak is commonly reported as the characteristic decomposition temperature for that step. Two reference points are used to bound each step for the percentage calculation:
- Onset temperature — typically taken as the intersection of the extrapolated baseline before the step and the tangent to the curve’s steepest point of descent (the tangent method), giving a reproducible starting point that is less subjective than reading directly off the point where the curve visibly begins to bend.
- Endset/completion temperature — the point at which the curve returns to a stable, near-horizontal baseline, marking the end of that mass-loss event and the start of the next isothermal or ramping segment.
A well-behaved multi-step curve on a filled polymer or elastomer typically shows: an early, low-temperature step (moisture and low-boiling volatiles), a mid-range step under inert gas (decomposition of the organic/polymer backbone), and — if the atmosphere is switched to oxygen or air partway through — a further step corresponding to combustion of carbon black or residual char, leaving a flat final residue that represents inorganic filler or ash. Not every sample shows all four; a purely inorganic mineral sample may show only a moisture step and a decomposition step (e.g., a carbonate losing CO₂), while a solvent-free thermoset may show no distinct low-temperature volatiles step at all.
Worked Calculation: Assigning Mass-Loss Steps to Composition
The arithmetic itself is simple — percentage mass loss for a given step is (mass at start of step − mass at end of step) ÷ starting sample mass × 100 — but getting a clean set of numbers requires bounding each step correctly first, using the onset/endset points described above. The illustrative example below (representative numbers, not a specific reported dataset) walks through a four-step scan typical of a filled elastomer — the same general step pattern that a standard method such as ASTM D6370 (rubber compositional analysis by TGA) or ASTM E1131 (general compositional analysis by thermogravimetry) is built around: an inert-atmosphere segment for volatiles and polymer decomposition, then a switch to an oxidizing atmosphere to combust carbon black and isolate ash.
| Step | Temperature range | Atmosphere | Mass at start | Mass at end | Mass loss | Mass loss (%) | Assignment |
|---|---|---|---|---|---|---|---|
| — | Start | N₂ | — | 10.000 mg | — | — | Starting sample mass |
| 1 | 30–250°C | N₂ | 10.000 mg | 9.620 mg | 0.380 mg | 3.80% | Moisture and low-boiling volatiles (residual water, processing oils, plasticizer) |
| 2 | 250–550°C | N₂ | 9.620 mg | 6.050 mg | 3.570 mg | 35.70% | Polymer/organic backbone decomposition |
| 3 | 550–750°C | Switched to air/O₂ at 550°C | 6.050 mg | 2.850 mg | 3.200 mg | 32.00% | Combustion of carbon black / residual carbon |
| 4 (residue) | Held at 750°C | Air/O₂ | 2.850 mg | 2.850 mg | — | 28.50% | Inorganic ash / mineral filler (residue remaining) |
Checking the arithmetic: 3.80 + 35.70 + 32.00 + 28.50 = 100.00%, which should always be true (within rounding and any small buoyancy-driven baseline offset) since every percentage in the table is calculated against the same starting mass. If a set of step percentages from a real run does not sum close to 100%, the most common causes are a miscalibrated or unsubtracted baseline, a step boundary drawn incorrectly (onset or endset placed at the wrong point), or an incomplete final residue reading taken before the curve had actually flattened.
The real ASTM D6370 method for filled rubber specifies its own fixed program rather than leaving temperature breakpoints to judgment — heating from roughly 50°C to 560°C at 10°C/min under nitrogen, holding, then continuing from 300°C to 800°C at 10°C/min after switching to an oxidizing atmosphere, on a sample of roughly 10–12 mg — specifically so that results are comparable between laboratories running the same standard method. When comparability across labs or against a specification matters, run the published standard method rather than an ad hoc temperature program.
TGA vs. DSC and DTA
TGA measures only mass change — it cannot on its own distinguish an endothermic transition (such as melting, which involves no mass loss at all) from an exothermic one, and it says nothing about events that involve no mass change, such as most crystalline phase transitions or glass transitions. Differential scanning calorimetry (DSC) and differential thermal analysis (DTA) measure heat flow rather than mass, and are the complementary techniques used to characterize those events. Many mass-loss steps on a TGA curve are worth checking against a simultaneous or parallel DSC/DTA run (sometimes run on the same instrument as simultaneous thermal analysis, STA) specifically to confirm whether a given step is endothermic (consistent with simple evaporation or desorption) or exothermic (consistent with combustion or an exothermic decomposition reaction) — the mass-loss curve alone cannot tell you which.
Common Applications
- Moisture and volatile content — pharmaceutical excipients, food powders, and other hygroscopic materials, as a faster alternative or cross-check to loss-on-drying methods.
- Compositional analysis of filled polymers and rubber — separating polymer, plasticizer/oil, carbon black or other reinforcing filler, and inorganic ash, per methods such as ASTM D6370 and ISO 11358.
- Thermal stability screening — comparing decomposition onset temperatures between formulations or batches, for quality control or shelf-life-related work.
- Filler and ash content — mineral fillers in plastics, inorganic content in coatings, and residual catalyst or inorganic contamination in a formulated product.
- Kinetic studies — running the same material at multiple heating rates to estimate decomposition kinetics (the basis of methods such as ISO 11358-2’s activation-energy determination), an application that goes well beyond a single routine scan.
Troubleshooting Table
| Symptom | Likely cause | Fix |
|---|---|---|
| Gradual mass drift with no real step (curve never returns to a flat 0% baseline drift line) | Buoyancy effect — gas density around the sample and crucible changes as the furnace heats, changing the apparent (not real) mass reading; can be worsened by furnace or purge-gas turbulence | Run and subtract an empty-crucible blank using the identical method; check that purge flow rate is stable and within the manufacturer’s recommended range. |
| Noisy or spiking signal during a mass-loss step | Sample splattering or spitting as trapped volatiles release rapidly, especially in samples that melt then boil, or that are loaded as a thick/mounded bed | Spread the sample thinner, reduce heating rate through that temperature range, or use a crucible with a pierced lid to moderate the rate of vapor escape. |
| Step onset shifted to a noticeably higher temperature than expected or than a reference value | Heating rate too fast, causing thermal lag between programmed and actual sample temperature | Reduce heating rate, especially through the region of interest, or run a matched heating-rate reference standard alongside the sample for comparison. |
| Sample appears fused, discolored, or reacted with the crucible after the run | Chemical incompatibility between sample and crucible material (common with platinum and certain metals, halogens, or molten salts) | Switch to an alumina or other inert crucible appropriate to the sample chemistry; dedicate crucibles to specific sample types. |
| Two mass-loss events overlap into a single broad, poorly defined step | Heating rate too fast to resolve events that occur close together in temperature, or a genuinely overlapping decomposition mechanism | Repeat at a slower heating rate to test whether the events separate; if they still overlap, consider evolved-gas analysis to confirm what is actually being lost at each point rather than relying on temperature alone. |
| Final residue percentage does not match an expected or specification value | Incomplete combustion of carbon black (oxidizing segment ended too early, or held at too low a temperature) or a baseline/calibration issue | Confirm the oxidizing segment ran long enough for the curve to flatten completely before recording the residue; verify balance calibration and run a blank if the discrepancy is small. |
Relevant Standards
ASTM E1131 (“Standard Test Method for Compositional Analysis by Thermogravimetry”) gives a general technique for determining volatile matter, combustible/organic material, and ash content across a broad range of solid and liquid materials, typically from room temperature up to 1000°C. ASTM D6370 applies the same underlying approach specifically to rubber compounds, with a defined temperature program for separating oil/volatiles, polymer, carbon black, and ash. ISO 11358 (in several parts) covers thermogravimetry of polymers specifically, including general principles (Part 1) and activation-energy/kinetics determination from multi-rate scans (Parts 2 and 3). Where a result needs to be defensible against a specification or comparable across laboratories, run the applicable standard method as written rather than an ad hoc program, and check whether the relevant laboratory holds ISO/IEC 17025 accreditation for the specific test method if that matters for the intended use of the result.
Frequently Asked Questions
What does TGA actually measure?
Mass, and only mass — as a function of temperature and/or time, under a controlled gas atmosphere. It does not directly measure heat flow, composition, or chemical identity; those are inferred from the temperature at which mass is lost, from the magnitude of each step, and often from a complementary technique.
How much sample do I need?
Typical TGA sample sizes run roughly 5–20 mg, though the right amount depends on the instrument’s furnace and balance design, the sample’s density and homogeneity, and how small a mass-loss step needs to be resolved — check the specific instrument’s guidance, and note that standard methods such as ASTM D6370 specify their own fixed sample-mass range for comparability between laboratories.
What’s the difference between TGA and DSC?
TGA measures mass change; DSC measures heat flow. A transition that involves no mass change (melting, most glass transitions, many crystalline phase changes) is invisible to TGA but visible to DSC, while DSC alone cannot confirm that a heat-flow event corresponds to material actually leaving the sample. The two are complementary, and are often run together as simultaneous thermal analysis (STA) specifically so that mass-loss and heat-flow data can be read against the same temperature axis.
Why does my decomposition temperature not match a published value for the same material?
Almost always heating rate, purge gas, or crucible/atmosphere differences between the two runs — decomposition onset and peak temperatures measured by TGA are method-dependent, not fixed physical constants, which is exactly why standard methods specify a fixed heating rate and atmosphere rather than leaving them open.
Can TGA identify an unknown material?
Not on its own, and not reliably. TGA tells you how much mass is lost and at what temperature, which can be diagnostic when compared against a known reference or library of curves, but it does not directly identify chemical composition. Evolved-gas analysis (TGA coupled to FTIR or mass spectrometry on the furnace exhaust) closes that gap when identification of a specific mass-loss event is required.
For related analytical-instrument methods on this site, see the guides to gas chromatography, HPLC, LC-MS, thin-layer chromatography, and flow cytometry, as well as the guides to spectrophotometer calibration and analytical balance calibration and weighing technique.
