Interpretation of DSC curves

Part 1 : Dynamic measurements

The art of interpreting curves has yet t be integrated into commercially available computer programs. The interpretation of a DSC measurement curve is therefore still something you have to do yourself. It requires a considerable amount of experience in thermal analysis as well as a knowledge of the possible reactions that your particular sample can undergo.
This article presents tips and information that should help you with the systematic interpretation of DSC curves.

▣ Recognizing artifacts

The first thing to do is to examine the curve for any obvious artifacts that could lead to a possible misinterpretation of the results. Artifacts are effects that are not caused by the sample under investigation. Figure 1 shows examples of a number of such artifacts. They include:

A) An abrupt change of the heat transfer between the sample and the pan:
1) Samples of irregular form can topple over in the pan.
2) Polymer films that have not been pressed against the base of the pan first change shape(no longer lie flat) on initial warming. Afterward, on melting, they make good contact with the pan(Fig.2)

B) An abrupt change of the heat transfer between the pan and the DSC sensor:
1) Distortion of a hermetically sealed Al pan due to the vapor pressure of the sample.
2) Slight shift of the Al pan during a dynamic temperature program due to different coefficients of expansion(Al:~24ppm/K, DSC sensor~9ppm/K, see also Fig.2). This artifact does not occur with Pt pans(~8ppm/K).
3) The measuring cell suffers a mechanical shock: The pans jump around on the sensor and can move sideways if they do not have a central locating pin.

C) The entry of cool air into the measuring cell due to a poorly adjusted measuring cell lid leads to temperature fluctuations which cause a very noisy signal.

D) Electrical effects :
1) Discharge of static electricity in a metallic part of the system, or power supply disturbances(spikes)
2) Radio emitters, mobile(cellular)phones and other sources of high frequency interference.

E) A sudden change of room temperature, e.g. through sunshine.

F) The lid of the pan bursts as a result of increasing vapor pressure of the sample This produces an endothermic peak with a height of 0.1mW to 100mW depending on the quantity of gas or vapor evolved.

G) Intermittent(often periodic) closing of the hole in the lid of the pan due to droplets that condense or to samples that foam.

H) Contamination of the sensors caused by residues of a sample from previous experiments. The thermal effects characteristic for this substance always occur at the same temperature. This problem can often be overcome by heating the system in air or oxygen. This type of artifact is very dependent on the contaminant. Artifacts caused by pans that are not inert also look very similar. Figure 3 shows an example of this.

Artifacts can also interfere with automatic evaluations(with EvalMacro), especially those using automatic limits. Isolated artifacts that have been definitely identified as such can be eliminated from the measurement curve using TA/Baseline.

▣ Measurement conditions

You define the temperature range and the heating rate for the measurement based on your knowledge of the physical and chemical properties of the sample.
. Choose a temperature range that is on the large side. At a heating rate of 20K/min, you do not in fact lose too much time if the range measured is 100K too large.
. Use a sample weight of about 5 mg for the first measurement. Make a note of the total weight of the sample and pan so that you can detect a loss of weight by reweighing after the analysis. The first measurement is often performed using a pan with a pierced lid and nitrogen as a purge gas.

. The first heating curve is usually measured from room temperature to the desired final temperature at a heating rate of 20K/min.

. Interpretation is often facilitated by measuring a cooling curve directly afterward. The cooling rate that can be used depends on the cooling option installed in your system.

. It is a good idea to heat the sample a second time. differences between the first and the second heating curves can be very informative.

. Another helpful variation is to shock cool the sample after it has been heated for the first time to the final temperature. This freezes any possible metastable states. The  sample is then measured a second time. A very convenient way to shock cool the sample to room temperature is t use the automatic sample robot. It deposits the hot sample on the cold aluminum turntable, which cools it down to room temperature within a few seconds. If you do not have a sample robot, you can wait until the sample robot, you can wait until the sample has reached its final temperature and then remove the pan with tweezers and place it on a cold aluminum surface(with a 2 mm diameter hole for the pin) or immerse it for about 10 seconds in liquid nitrogen.


Fig.1. DSC artifacts(details are given in the text) : An artifact can very often be identified by repeating the measurement with a new sample of the same substance and observing whether the effect occurs again either at the same place or at a different place on the curve. Exceptions to this are f and h, which can be very reproducible.


Fig.2. Above : Artifact due to a PE film that was not pressed down firmly in the pan(dotted line). The sample of film that was pressed down on the base of the pan with the lid of a light Al pan gave the "correct" melting curve.
Below : DSC heating curve of 1.92mg polystyrene showing a typical artifact at about 78oC caused by the thermal expansion of the Al pan. This artifact, which is of the order of 10+W, is only visible with large scale expansion(ordinate scale < 1mW).


Fig.3. Below : In an open pan, water evaporates before the boiling point is reached. Middle : In a self-generated atmosphere(50mw hole in the lid), the boiling point can be measured as the onset.
Above : In a hermetically sealed pan(at constant volume), there is no boiling point. The DSC curve is a straight line until the Al pan suddenly bursts at about 119˚C. If the ordinate scale is expanded 20 times an exothermic peak can be observed that is due to the reaction of aluminum with water(see the expanded section of the curve).

▣ If no thermal effect occurs

In this case your sample is inert in the temperature range used for the measurement and you have only measured the(temperature dependent) heat capacity. An inert sample does not undergo any loss of weight(except<30mg surface moisure). After opening the pan, it looks exactly the same as before the measurement. This can be confirmed with the aid of a microscope for reflected light.
If you are interested in cp values, you need a suitable blank curve. Check the plausibility of the results you obtain : values for cp are usually in the range 0.1 to 5Jg-1K-1. To make absolutely sure that no effects occur, extend the temperature range of the measurement and measure larger samples.

▣ If thermal effects are visible

Thermal effects are distinct deviations from the more or less straight line DSC curve. They are caused by the sample undergoing physical transitions or chemical reactions. If two effects overlap, try to separate them by using faster or slower heating rates, and smaller sample weights. Here, one should take into account that faster heating rates cause a marked shift in the peak maxima of chemical reactions to higher temperatures. To a lesser extent, this also applies to solid-solid transitions and glass transitions. The onset temperatures of the melting processes of nonpolymeric substances are, however, independent of the heating rate.

If several effects occur with significant loss of weight ( >39mg), you would of course like to assign the latter to a particular peak-weight loss is usually an endothermic effect due to the work of expansion resulting from the formation of gas. One method is to heat a new sample step by step through the individual peaks and determine the weight of the pan and contents at each stage(at Mettler Toledo, "off-line thermogravimetry"). The best way is to measure a new sample in a TGA, and use the same type of pan as for the DSC, measurement.
The shape of the DSC curve is usually very characteristic and helps to identify the nature of the effect.

▣ Physical transitions

Physical transitions can in principle be measured as many times as desired if

. on cooling, the sample reverts to the same state as before the transition. This, however, is not always the case and depends on the sample and the cooling rate. Manu substances in fact solidify from the melt at fast cooling rates to a glassy amorphous state. This is the reason why no melting peak occurs on heating the same sample a second time. Some metastable crystal modifications crystallize only in the presence of certain solvents.

. the sample does not escape from the pan through evaporation, sublimation, or(chemical) decomposition, or does not undergo transformation. Any sample lost by evaporation cannot of course condense in the sample pan on cooling because the purge gas has already removed it from the measuring cell.

▣ Melting, crystallization and mesophase transitions

The heat of fusion and the melting point can be determined from the melting curve. With pure substances, where the low temperature side of the melting peak is almost a straight line(Fig.4a), the melting point corresponds to the oneset. Impure and polymeric samples, whose melting curves are concave in shape, are characterized by the temperatures of their peak maxima(Fig.4b and c). Partially crystalline polymers give rise to very broad melting peaks because of the size distribution of the crystalites(Fig.4c).

Many organic compounds melt with decomposition(exothermic or endothermic, Figs. 4d and 4e).
An endothermic peak in a DSC heating curve is a melting peak if
. the sample weight does not decrease significantly over the course of the peak. A number of substances exhibit a marked degree of sublimation around the melting temperature. If hermetically sealed pans are used, the DSC curve is not affected by sublimation and evaporation.
. the sample appears to have visibly melted after the measurement. Powdery organic substances, in particular, form a melt that on cooling either solidifies to a glass(with no exothermic crystallization peak) or crystallizes with an exothermic peak.
Comment : Many melts have a high melting point oxide layer on their surface. After melting, the oxide layer remains behind as a rigid envelope. This is the reason why, on opening the pan, the sample looks exactly the same as before melting-it would in fact require samples weighing several grams to deform the oxide layer under the force of gravity, so that the sample fits the shape of the pan. Precious metals have on oxide layer and form spherical droplets on melting.
. its surface area is between about 10Jg-1 and 400Jg-1. The heat of fusion on nonopolymeric organic substances is almost always between 120Jg-1 and 170Jg-1.
. its width at half height(half-width) is significantly less than 10K (partially crystalline polymers can melt over a wider range). The melting peak is increasingly sharper, the purer the substance and the smaller the size of the sample. Very small quantities of pure substances give peaks with half-widths of less than 1 K.

Impure samples and mixtures often show several peaks. Substances with eutectic impurities exhibit two peaks(Fig.4b): first the eutectic peak, whose size is proportional to the amount of impurity, and then the main melting peak. Sometimes the eutectic is amorphous so the first peak is missing. Liquid crystals remain anisotropic even after the melting peak. The melt does not become isotropic until one or mor small sharp peaks of mesophase transitions have ocurred(Fig.4f).
. the peak are is about the same as the melting peak-since the heat of fusion is temperature dependent, a difference of up to 20% can arise depending on the degree of supercooling.
. the degree of supercooling(the difference between the onset temperatures of melting and crystallization) is between 1 K and about 50K. Substances that crystallize rapidly show an almost vertical line after nucleation until(if the sample is large enouth) the melting temperature is reached(Figs. 5a, 5g).

If the liquid phase consists of a number of individual droplets, the degree of supercooling of each droplet is different so that several peaks are observed(Fig. 5b). Organic and other "poorly crystallizing" compounds form a solid glass on cooling(Fig.5c). Such amorphous samples can then crystallize on heating to temperatures above the glass transition temperature(devitrification, cold crystallization). Cold crystallization can often occur in two steps. On further heating, polymorphic transitions can occur before the solid phase finally melts(Fig.5e).

When the melt of a sample containing eutectic impurities is cooled, the main component often crystallizes out(Fig. 5d). It can, hoever, solidify to a glass(Fig.5c). Very often the eutectic remains amorphous so that the eutectic peak is missing.
A polymer melt crystallizes after supercooing by about 30K9Fig.5f). many polymers solidify to glasses on rapid cooling(Fig. 5c).

When the melt of a liquid crystal is cooled, the mesophase transitions occur first(often without any supercooling). The subsequent crystallization exhibits the usual supercooling(Fig.5g).


Fig.4. Melting processes : a: a nonpolymeric pure substance; b: a sample wit a eutectic impurity; c: a partially crystalline polymer; d and e : melting with decomposition; f: a liquid crystal.


Fig.5. Crystallization: a: a pure substance(Tf is the melting point); b:separate droplets solidify with individual degrees of supercooling; c: a melt that solidifies amorphously; d: a sample with a eutectic impurity; e: a shock-cooled melt crystallizes on warming above the glass transition temperature(cold crystallization)l f: a partially crystalline polymer; g: a liquid crystal


Fig.6. Monotropic transition : a: the arrow marks the solid-solid transition, afterward the a-modification just formed melts; b: in this case the solid-solid transition is so slow that a crystallizesk c: the pure a'-form melts low; d: the pure a-form melts high.

▣ Solid-solid transitions polymorphism

Solid-solid transitions can be identified by the fact that a sample in powder form is still a powder even after the transition. The monotropic solid-solid transition of metastable crystals(marked a' in Fig.6) to the stable a-form, which is frequently observed in organic compounds, is exothermic(Fig. 6a). As the name implies, monotropoic transitions go in one direction only(they are irreversible).

The monotropic transition is slow and is most rapid a few degrees K below the melting point of the metastable phase. In spite of this, the peak height is usually less than 0.5mW and can therefore easily be overlooked alongside the following melting peak of about 10mW(gray arrow in Fig.6b). It is often best to measure the monotropic transition isothermally.
At heating rates greater than 5K/min, it is easy to "run over" the slow transition(Fig.6b) and so reach the melting temperature of the metastable form. The monotropic solid-solid transition is either not visible or it could be falsely interpreted as a slightly exothermic "baseline shift" before the melting peak. If some stable crystals are present that can serve as nuclei for the crystallization of the liquid phase formed, the melting peak merges directly into the exothermic crystallization via the liquid phase-on immediate cooling to room temperature, the sample would have visibly melted. Finally the melting temperature of the stable midification is reached.

If no a-nuclei are present, there is no a-crystallization peak and of course no a-melting peak(Fig. 6c). If the sample consists entirely of the stable form, then only the a-melting peak appears and the polymorphic effect is not observed(Fig.6d).  Depending on the substance, the-form melts at temperatures that are 1 K to 40K lower than the stable modification. The enanitiotropic solid-solid transition, which occurs less often, is reversible. The a->b transition, starting from the low temperature a-form to the high temperature b-form is endothermic. The enantiotropic transition gives rise to peaks of different shape depending on the particle size of the sample because the nucleation rate of each crystal is different. For statistical reasons, samples that are finely crystalline give rise to bell-shaped(gaussian) peaks(Figs.7a and 7c). A small number of larger crystals can give rise to peaks with very bizarre shapes. This is especially the case for the reverse b->a transition(Figs. 7b and 7d). The peaks of enantiotropic transitions typically have a half-width of 10K.


Fig.7. Reversible enantiotropic transition : a: a fine powder; b: coarse crystals; c: reverse transition of the fine powder; d: reverse transition of the coarse crystals; at Tt, a and b are in thermodynamic equilibrium.

▣ Transitions with a distinct loss of weight

These types of transitions can of course only be observed in open pans, i.e. either a pan with no lid, or a pan with a lid and a 1 mm hole to protect the measuring cell from substances that creep out or that splitter.

Examples are :
. the  evaporation of liquid samples(Fig.3, below and Fig. 8a),
. drying (desorbtion of adsorbed moisture or solvents, Fig.8b)
. the sublimation of solid samples(Fig.8b) and the
. decomposition of hydrates9or solvates) with the elimination of the water of crystallization. In an open crucible, the shape of the curve corresponds that shown in Fig.8b, and in a self-generated atmosphere to that in Fig.8c.

These peaks have a half-width of > 20K(except in a self-generated atmosphere) and have a shape similar to that exhibited by chemical reactions. The decomposition of solvates is known as pseudo-polymorphism(probably because in a hermitically sealed pan, a new melting point occurs when the sample melts in its own water of crystallization) and can also be regarded as a chemical reaction. In a self-generated atmosphere(with a 50mm hole in the lid of the pan), the evaporation of liquids is severely hindered. The usual very sharp boiling peak(Fig.3, middle and Fig.8d) does not occur until the boiling point is reached.

Apart from the appreciable loss of weight, these reactions have another feature in common, namely that the baseline shifts in the exothermic direction due to the decreasing heat capacity of the sample.

▣ The glass transition

At the glass transition of amorphous substances, the specific heat increases by about 0.1 to 0.5Jg-1K-1. This is the reason shy the DSC curve shows a characteristic shift in the endothermic direction(Fig.2, below and Fig.9a).

Typically
. the radius of curvature at the onset is significantly greater than at the endset and
. before the transition, the slope is clearly endothermic, and after the transition the curve is (almost)horizpntal.

The first measurement of a sample that has been stored for a long time below the glass transition temperature, Tg, often exhibits an endothermic relaxation peak with an are of 1Jg-1 to a maximum of about 10Jg-1(Fig.9b). This peak can no longer be observed on cooling(Fig.9c), or on heating a second time. The glass transition covers a temperature range of 10K to about 30K.
You can identify an effect that resembles a glass transition by checking whether the sample is visibly soft, almost liquid or rubbery-like above the Tg. If you do not have access to a TMA or DMA instrument, you can check this by heating a sample up to a temperature of Tg + 20K in a pan without a lid. After several minutes at this temperature, you open the lid of the measuring cell and press the sample with a spatula or a needle. It is, however, difficult to detect softening in this way especially with polymers containing large amounts of fillers.


Fig. 8. Transitions with weight loss : a: evaporation in an open pan; b: desorption, sublimation; c: dehydration; d: boiling in a pan with a small hole in the lid, Tb is the boiling point.


Fig.9. Step transitions: a: a glass transition; b: a glass transition with enthalpy relaxation; c: the reverse transition; d: a Curie transition


Fig.10. Curve shapes of chemical reactions: a: an ideal exothermic reaction; b: reaction with "Interfering" physical transitions and the beginning of decomposition; c: chemical reaction with a secondary reaction; d: partial oxidation of organic samples with the residual oxygen in a hermetically sealed pan.

▣ Lambda transitions

These types of solid-solid transitions exhibit  L-shaped cp temperature functions. The most important is the ferromagnetic Curie transition, which was previously used to calibrate the temperature scale of TGA instruments. The DSC effect is however extremely weak(Fig.9d). To make sure, you can check that the sample is no longer magnetic above the Curie temperature with a small magnet.

▣ Chemical reactions

Chemical reactions can in general only be measured in the first heating run. On cooling to the starting temperature, the reaction product remains chemically stable, so that on heating a second time no further reaction takes place1. In some cases, however, the reaction does not go to completion during the first heating run, so that on heating a second time, a weak postreaction can be observed(e.g. the curing of epoxy resins).

The half-width of chemical reaction peaks is about 10K to 70K(usually about 50K at a heating rate of 10K/min to 20K/min). Reactions which show no significant loss of weight are usually exothermic(about 1Jg-1 to 20,000Jg-1, Figs. 10a and 10b). The others tend to be endothermic because the work of expansion predominates.

Ideally, DSC curves of a chemical reaction show a single smooth peak (Fig.10a). In practice, however, other effects and reactions often overlap and distort the peak shape, e.g. the melting of additives(Fig.10b), or secondary or decomposition reactions (Fig.10c).

Examples of reactions with significant loss of weight are:
. thermal decomposition(pyrolysis under an inert gas), with CO, short-chain alkanes, H₂O and N₂ as the most frequently occurring gaseous pyrolysis products,
. depolymerization with more or less quantitative formation of the monomer and
. polycondensation, for example the curing of phenol and melamine resins.2

Reactions with a significant increase of weight nearly always involve oxygen and are strongly exothermic. Examples are:
. the corrosion of metals such as iron and
. the initial uptake of oxygen at the beginning of the oxidation of organic compounds. During the course of the reaction, volatile oxidation products such as carbonic acids, CO₂ and H₂O are formed, so that finally a weight loss occurs(the initial increase in weight can be seen best in a TGA curve).

Examples of reactions with no significant change in weight are3.
. addition and polyaddition reactions, curing of epoxy resins,
. polymerizations, dimerizations, rearrangements and
. the oxidation of organic samples(e.g. polyethylene) with the residual atmospheric oxygen(about 10mg) in a hermetically sealed pan(Fig.10d).

▣ Final comments
The article should help you to interpret DSC curves. You will, however, often have to use additional methods for confirmation. Some important techniques are :
. thermogravimetric analysis, ideally in combination with DTA or SDTA.. The interpretation of DTA and SDTA? curves is analogous to DSC with limitations due to reduced sensitivity,
. thermomechanical and dynamic mechanical analysis,
. the analysis of the gaseous substances evolved(EGA, Evolved Gas Analysis) with MS or FTIR and
. the observation of the sample on a hot stage microscope

In addition, various other chemical or physical methods are available. These depend on the type of sample, and can be applied after each thermal effect has taken place.

¹ There are very few exceptions to this rule; one example is the polymerization of sulfur, which begins on heating at about 150˚C and which is then reverted on cooling at about 130˚C.

² These slightly exothermic reactions are often measured in high pressure crucibles in order to suppress the endothermic vaporization peak of the volatile side-products.

³ These reactions are often performed in hermetically sealed Al pans in order to prevent the release of small amounts of volatile components.