TGA for Nanomaterials: Purity, Loading and Thermal Stability

The cheapest purity check in the laboratory

Thermogravimetric analysis measures mass as a function of temperature in a controlled atmosphere. That is all. From that one signal you can get the residual inorganic content of a carbon material, the organic loading on a functionalised particle, the moisture and residual solvent content of a powder, and a usable measure of thermal stability — and you can do it on a few milligrams in a couple of hours on an instrument most materials laboratories already own.

For carbon nanomaterials in particular, TGA is the standard answer to the question buyers distrust most: how much of what I paid for is catalyst. It is also, for exactly that reason, the measurement most often reported without the arithmetic that makes it correct.

What the instrument does

The measurement

A microbalance holds a small crucible inside a furnace under a controlled gas flow. The furnace follows a temperature programme, usually a linear ramp, and the balance records mass continuously. The output is a thermogram: mass, or mass percentage, against temperature.

Atmosphere is half the experiment

The same sample gives entirely different thermograms in oxidising and inert atmospheres, and the two answer different questions.

• Air or oxygen. Carbon burns off. What is left is the inorganic residue, oxidised. This is the purity experiment.
• Nitrogen or argon. Carbon does not burn. Organic material pyrolyses, volatiles leave, and a carbonaceous char often remains. This is the experiment for organic loading and for thermal stability in a non-oxidising service environment.

A thermogram reported without its atmosphere is uninterpretable. A supplier quoting a "decomposition temperature" without saying in what gas has given you half a specification.

Heating rate moves the answer

Mass loss is kinetic, not thermodynamic. Faster ramps shift apparent onset temperatures upward and blur closely spaced events; slower ramps resolve them and take longer. Comparing an onset temperature from a 20 K/min run against one from a 5 K/min run is not a valid comparison. Where the kinetics themselves are of interest, running several rates and applying an isoconversional analysis is the proper route, but for routine comparison the requirement is simply that everyone uses the same rate and says what it was.

Reading a thermogram

The steps

A well-separated thermogram is a staircase, and each step is a process. Typical assignments, in ascending temperature:

• Below about 150 °C — physisorbed water and residual solvent. Often ignored; should not be, because it sets the mass basis for everything above it.
• Roughly 150-400 °C — decomposition of organic ligands, surfactants and functional groups; loss of oxygen-containing groups in graphene oxide.
• Roughly 400-700 °C in air — oxidation of carbon. Where in that window depends strongly on how ordered the carbon is.
• Above — the residue, which for carbon materials is metal oxide from the growth catalyst.

The derivative curve is where the resolution is

Plot the first derivative of mass with respect to temperature — the DTG curve — and overlapping steps that look like one sloping shoulder on the mass curve resolve into distinct peaks. For carbon materials this is what separates amorphous carbon from nanotube, and it is the curve to ask for. A supplier who sends a mass curve without a derivative has sent the less informative half of the same data.

The residue

The final mass, expressed as a percentage of the starting mass. For a carbon material burned in air this is the inorganic content — and it is where the arithmetic below applies.

Carbon nanotube purity: the conversion everyone forgets

Burn a carbon nanotube sample in air. The carbon leaves as carbon dioxide. The iron, cobalt or nickel catalyst does not leave; it oxidises and stays. The residue is therefore metal oxide, not metal, and reporting the residue percentage as the catalyst content overstates the metal by the mass of the oxygen attached to it.

The correction is straightforward. For iron ending as the sesquioxide, the formula mass is about 159.7 and the two iron atoms contribute about 111.7, so iron is close to 70% of the residue mass by weight. A 5.0 wt% residue therefore corresponds to roughly 3.5 wt% iron — a difference of one and a half percentage points, on a specification where a percentage point matters and where two suppliers quoting "5% catalyst" may mean different things.

Two caveats that keep this honest. First, the oxide stoichiometry has to be known or assumed; iron can end as a mixed oxide, and cobalt and nickel oxides have their own conversion factors. Second, catalyst encapsulated inside a graphitic shell may not fully oxidise, in which case the residue underestimates the metal. Where the number really matters, the definitive measurement is elemental analysis by ICP-OES or ICP-MS after digestion, and TGA is the cheap routine check that flags when to spend that money.

Amorphous carbon versus nanotube

Disordered carbon oxidises at a lower temperature than well-graphitised carbon, because there are more reactive edge and defect sites to start from. In a good thermogram of a mixed sample the DTG curve shows two peaks, and the ratio of their areas estimates the amorphous fraction. In a poor one they merge into a single broad feature and the split is a matter of opinion.

Metal catalyst complicates this further by catalysing carbon oxidation, which shifts the onset downward — so a low oxidation onset can indicate poor graphitisation or high catalyst content, and TGA alone does not separate the two. Pair it with Raman spectroscopy, which reports on structural order directly.

Organic loading on functionalised particles

For a nanoparticle with an organic coating, TGA under inert gas gives the mass fraction of organic material directly: the step between the moisture loss and the plateau. That is already useful. Combined with a surface area it becomes considerably more useful, because it yields a grafting density — molecules per square nanometre — which is the number that actually describes a functionalised surface.

The calculation is: take the organic mass fraction divided by the inorganic mass fraction, divide by the molar mass of the grafted molecule to get moles of ligand per gram of inorganic core, multiply by the Avogadro constant to get molecules, and divide by the specific surface area of the core expressed in square nanometres per gram. Every quantity in that chain is measurable, and the weak links are worth naming: the molar mass assumes the whole ligand is lost and nothing else is, and the surface area must be that of the bare core, measured on an uncoated sample, not on the coated product. The BET guide covers where that area number goes wrong.

Cross-check the identity of what was lost. TGA tells you mass left; it does not tell you what. Coupling the instrument to a mass spectrometer or an infrared spectrometer — evolved gas analysis — identifies the products and turns a mass loss into a chemical statement.

Graphene oxide, where TGA is unusually diagnostic

Graphene oxide has a characteristic thermogram: water loss below about 150 °C, then a pronounced mass loss in the region of 200 °C as the labile oxygen-containing groups decompose, releasing carbon dioxide, carbon monoxide and water. That step is essentially the oxidation level of the material made visible, and it is a good quick discriminator between graphene oxide, reduced graphene oxide and graphite. After reduction the low-temperature step shrinks markedly. The graphene, GO and rGO comparison puts this alongside the XRD, Raman and XPS evidence.

Where it goes wrong

• No blank subtraction. Gas density changes with temperature, so the apparent buoyancy on the crucible changes as the furnace heats. Without a blank run under identical conditions there is a systematic drift in the baseline, and for small mass losses it is not negligible.
• Too much sample. A deep bed develops thermal and gas-diffusion gradients: the outside burns while the inside has not started. Steps smear and onsets shift. Small samples, spread thin.
• The wrong crucible. Platinum catalyses some oxidations and alloys with some metals; alumina is the usual safe default. A crucible that reacts with the sample is a mass change you will attribute to the sample.
• Contaminated or leaking gas. Trace oxygen in a nominally inert atmosphere will oxidise carbon at high temperature and produce a mass loss that is an artefact of the plumbing.
• Reading onset temperature as a material constant. It depends on heating rate, sample mass, particle size, gas flow and crucible geometry. It is a comparative figure under stated conditions, not a property.
• Ignoring mass gain. Some samples gain mass — metal nanoparticles oxidising in air, for example — and a step upward is as informative as a step down.

What to ask a supplier for

• The thermogram and the DTG curve, not only a residue percentage.
• Atmosphere, gas flow rate, heating rate and temperature range.
• Sample mass and crucible material.
• Whether the quoted catalyst content is residue as measured or metal after oxide conversion, and which oxide was assumed.
• Whether the residue has been cross-checked by elemental analysis.

That is the same discipline the rest of the document deserves — see how to read a certificate of analysis.

Standards

The ISO 11358 series covers the general principles of thermogravimetry for polymers, and its treatment of procedure and reporting transfers directly to other materials. ASTM E1131 describes compositional analysis by thermogravimetry — the standard framework for splitting a thermogram into moisture, volatiles, combustible content and ash. Neither is nanomaterial-specific, and both are more rigorous about reporting conditions than most nanomaterial certificates are.

Related guides

• Where thermal analysis fits: the characterisation workflow.
• Structural order, which TGA infers and Raman measures: Raman spectroscopy of carbon nanomaterials.
• The surface-chemistry half of a functionalisation claim: XPS for nanomaterials.