💡 In This Article
- The four numbers a single TGA curve gives you: Td5%, onset temperature, DTG peak, and residue
- Why the same sample gives different results in nitrogen and in air
- Why residue alone gives the wrong filler content, and how to correct it
- A diagnostic table for guessing the cause of weight loss: moisture, solvent, flame retardant, or polymer backbone
So far in this series, we have looked at thermal analyzers through one question: how do you measure Tg? DSC detects the change in heat capacity, TMA the change in expansion (CTE), and DMA and the rheometer the change in modulus. Each of them also tells you something beyond Tg. So can TGA, another thermal analyzer, measure Tg too?
No. At Tg the molecular chains begin to move differently, but the mass does not change. A sample can go from glassy to rubbery and the reading on the balance stays exactly where it was. On a TGA curve, Tg passes without leaving a trace.
Even so, TGA is often the first instrument to run before a Tg measurement. It sets the upper temperature limit for DSC and DMA runs. Heat a sample into its decomposition range and you contaminate the instrument and make the data meaningless.
What TGA does best is answer two other questions: “How hot can this adhesive go?” and “How much filler is in this sample?” Neither is about how the material changes. Both are about where it starts to disappear and what is left when it is done. This article is about the thermal analyzer that cannot measure Tg.
What Does TGA Measure?
A TGA puts a sample on a very precise balance, raises the temperature, and records how the weight drops. Typical samples are 5–20 mg, and the balance can detect changes down to the microgram level.
The principle is simple. As the temperature rises, the components leave the sample one by one. Moisture and residual solvent go first, then low-molecular-weight additives, and finally the crosslinked epoxy network itself breaks down into gas. What remains is whatever does not burn: inorganic filler and carbonaceous char.
DSC, TMA, and DMA watch a material change. TGA watches it disappear. That is why a TGA curve, unlike other thermal analysis plots, almost always steps downward from left to right.
Reading a TGA Curve in Three Regions
The x-axis is temperature, and the y-axis is the remaining weight (wt%), with the starting weight set to 100%. The curve of a cured epoxy adhesive is usually read in three regions. The exact temperatures depend on the resin, curing agent, and additives, so treat the ranges below as rough guides.
- Low-temperature region (below about 200 °C): Absorbed moisture, residual solvent, and unreacted low-molecular-weight species leave. In a well-cured sample, the loss here should be small, under about 1%.
- Main decomposition region (about 300–450 °C): The crosslinked network breaks and the weight falls sharply. This is the steepest cliff on the curve.
- High-temperature region (about 500 °C and above): The weight change flattens out again, and what remains at the end is the residue.
Schematic TGA and DTG curves of a typical cured epoxy, divided into three regions, with Td5%, onset, DTG peak, and residue marked.
A large weight loss in the low-temperature region is a signal in itself. It can mean incomplete cure, leftover solvent, or moisture picked up in storage, and in the actual process it can lead to bubbles, voids, and outgassing. It should not be dismissed as just another number.
Four Numbers to Read from the Curve
① Td5% (5% weight-loss temperature). This is the temperature at which the sample has lost 5% of its starting weight. The “heat resistance” figure on a datasheet is usually this number. The definition is simple, which makes it convenient for comparing different materials. But if a sample loses 2–3% moisture at low temperature, Td5% can come out low regardless of what the resin is doing. So when I look at Td5%, I also ask what is inside that first 5%.
② Onset temperature. This is where the extension of the flat baseline before decomposition meets the tangent to the decomposition cliff. It tells you more than Td5% about where serious decomposition begins. The value depends on where the tangent is drawn, though, so when comparing, check that the same method was used.
③ DTG peak (temperature of maximum decomposition rate). Differentiating the TGA curve with respect to temperature gives the DTG (derivative thermogravimetry) curve. Its peak marks the temperature at which weight is lost fastest. What looks like a single cliff on the TGA curve often splits into two or three peaks on DTG, which means different components are decomposing at their own temperatures. The more complex the formulation, with flame retardants, tougheners, or blended resins, the more DTG tells you compared with TGA.
④ Residue. This is the weight left at the final temperature, usually 700–800 °C. For a filled adhesive, you can use it to estimate the filler content. It is also where the most common mistake happens, and to explain it I first need to talk about atmosphere.
Same Sample, Different Results in Nitrogen and in Air
The gas flowing around the sample turns TGA into a different experiment.
| Nitrogen (N₂) | Air | |
|---|---|---|
| Reaction | Pyrolysis (heat alone breaks the chains, no oxygen) | Thermo-oxidative decomposition (oxygen takes part) |
| What it shows | Thermal stability of the polymer structure itself | Heat resistance closer to real service conditions |
| Curve shape | One main decomposition step, then char remains | After the main step, the char oxidizes and weight drops again in a second step |
| Residue | Inorganic filler + resin char | Almost only inorganic filler |
Even a neat epoxy resin measured in nitrogen can leave a few percent of char, and sometimes more than 10%. The more aromatic the resin, the more char. In air, that char burns off at high temperature. So if two datasheets for the same adhesive show noticeably different Td5% values, check whether the test atmosphere differs before concluding that the material does.
Residue Is Not Filler Content
This is where most mistakes happen. Say a filled adhesive measured in nitrogen leaves 72% residue. Calling that 72% filler would be wrong, because the residue also contains char from the resin. The residue of a composite is the sum of the filler’s residue and the resin’s residue. Let w be the filler weight fraction and R the residue fraction of each component measured alone:
Rc = ω · Rf + (1 – ω) · Rr
ω = (Rc – Rr) / (Rf – Rr)
- Rc : residue of the composite (the formulation)
- Rr : residue of the resin alone, measured under the same conditions
- Rf : residue of the filler alone (close to 100% for stable fillers such as silica or alumina)
If the neat resin in the example leaves 10%, the actual filler content is (72 − 10) ÷ (100 − 10) ≈ 69%. A 3-point difference may look small, but in a highly filled formulation it is enough to change viscosity and CTE significantly.
Rf is not 100% for every filler. Hydroxide flame-retardant fillers are the typical case. On heating, they release their bound water and convert to oxides, losing weight along the way. That dehydration reaction is also the basis of their flame-retardant effect. If you take Rf = 100% for such a filler, you will badly underestimate the filler content. Silane-treated fillers also lose a small amount of weight from the organic layer on their surface.
Weight percent and volume percent are different things as well. What actually governs viscosity and thermal conductivity is the volume fraction, which you get by converting through density (ρ):
Seventy wt% silica (about 2.2 g/cm³) in epoxy (about 1.2 g/cm³) is roughly 56 vol%. The same 70 wt% of alumina (about 3.9 g/cm³) comes to only about 42 vol%. This is why, in the article [Why Does Adding More Filler Increase Viscosity So Dramatically?], the basis is volume rather than weight.
Cautions When Running TGA
- Heating rate. Usually 10 or 20 °C/min. The thermal lag discussed in the TMA article occurs in TGA too, so a faster rate gives a higher apparent decomposition temperature. If you compare data taken at two different rates, you are looking at the difference in conditions, not in materials.
- Sample amount and sampling position. With only a few milligrams, sedimented or unevenly dispersed filler means the residue depends on where the piece was taken. If you are judging filler content from residue, take samples from different positions and measure at least 2–3 times. Conversely, you can use that scatter to gauge dispersion or sedimentation.
- Degree of cure and thermal history. An uncured sample goes through further cure and releases low-molecular-weight species during the run, so the low-temperature region looks messy. As with TMA, you need to know how the sample was cured before you read the data.
- Isothermal TGA. Instead of ramping the temperature, you can hold the sample at a set temperature and watch the weight change over time. This shows how much leaves at a real process temperature, such as reflow or post-cure, and it is useful for tracing outgassing and void problems.
A Development Note: The Day I Read Residue as Filler Weight
While analyzing a competitor’s formulation, I once read the TGA residue directly as the filler content. Whatever survives to the end is inorganic filler, I thought, so the remaining weight must be the filler weight. It did not occur to me that a filler can turn into something else at high temperature.
Hydroxide flame-retardant fillers release water and become oxides when heated. For magnesium hydroxide:
Mg(OH)2 → MgO + H2O
By molecular weight, 18.0 of the 58.3 g/mol is water, so about 31% of the weight leaves and 69% remains as MgO. Suppose the test is run in air, so the resin burns away almost completely. A formulation with 50 wt% Mg(OH)₂ would then leave 50% × 0.69 ≈ 35%. Taken at face value, that number makes you wonder whether there is really 15% less filler than expected. Put Rf = 69% into the equation and you get the right answer: 35 ÷ 69 ≈ 50%.
Since then, I check one thing before looking at any TGA result: what does this filler turn into at 800 °C? The remaining weight is not the weight of what you put in. It is the weight of what survived.
Guessing the Cause from Where the Weight Is Lost
When you get a TGA curve, looking at where the weight was lost, and how much, points you toward the cause. Confirmation needs FT-IR or other analysis, but this is enough for a first judgment.
| Where the loss occurs | First suspect | Related article |
|---|---|---|
| –200 °C, above 1% | Absorbed moisture, Residual solvent, incomplete cure | [Why Does Epoxy Look Fine When Dry but Fail After Absorbing Moisture?] |
| 200–350 °C, a step before the main cliff | Dehydration of hydroxide flame retardants (temperature depends on the type), low-molecular-weight additives | [How Are Flame-Retardant Additives Used in Epoxy Formulations?] |
| Main cliff splits in two on DTG | Tougheners, blended resins, phase separation | [What Is the Difference Between Carboxylated NBR and Core-Shell Rubber?] |
| Residue varies between runs | Filler sedimentation, uneven dispersion | [Why Does Particle Size Distribution (PSD) Matter? — Why Fillers with the Same D50 Can Behave Differently] |
Final Thoughts
DSC, TMA, and DMA show how a material changes. TGA shows when it starts to fall apart and what it leaves behind. The weight that remains is a mixture of resin char and filler, so before trusting the number, check the atmosphere and how the measurement was made.
One thing slipped by earlier. A silane-treated filler loses just a little more weight in TGA. How can an organic layer making up less than 1% of the weight lower viscosity, raise adhesion to metal, and even change 85/85 reliability? In the next article, we will look at silane coupling agents, which link the filler, the resin, and the substrate.
