Differential Scanning Calorimetry
- Introduction to Differential Scanning Calorimetry (DSC)
- Important Aspects of DSC
- Principle of the DSC technique
- Heat Flow
- The Cell
- The Cell Controller
- The Cooling System and Purge Gas
- Experimental Procedure
- Sample Preparation
- Loading the Sample
- Creating the Protocol for Analysis
- Data Analysis
- Case Studied
- Melting Rates
- Crystallization Rates
- Storage Time
Calorimetry is a technique used to study the transfer of energy in the form of heat between a system and its environment. The transfer can be from the environment to the system or vice-versa. The melting of a fat is an endothermic event, wherein the fat sample absorbs energy from the environment, while crystallization is an exothermic process, one in which energy is given to the environment.
Differential Scanning Calorimetry (DSC) is a technique that measures difference in heat flow (hence differential) between the “system” and a “reference” as a function of a temperature or time (hence scanning). The “system” is the material under study, while the “reference” can be any known material like air. One instrument to carry out calorimetric measurements is called differential scanning calorimeter, also abbreviated DSC.
The main important aspect of a DSC instrument and the DSC technique are introduced in the following summary table.
DSC instruments are separated into two categories: heat-flux DSC and power compensation DSC. These calorimeters differ in design and measuring principles, but they have in common that the measured signal is proportional to the heat flow Φ.
A heat-flux DSC contains one furnace that consists of a base and two raised platforms. In this kind of calorimeter, a well-defined heat conduction path of known thermal resistances is used for the exchange of heat. The measured signal is the temperature difference between the sample’s pan and the reference’s pan. Two independent detectors (thermocouples) mounted on each of the raised platform are used to measure each temperature independently. A third detector (To) measures the temperature of the sensor base. A temperature difference between the detectors corresponding to the two pans will result in heat flow between them such that both pans are maintained at identical temperatures. These temperatures measurements are used to calculate the heat that flows into the sample with the help of the right calibrations, hence the importance of having correct calibration files.
A power-compensation DSC contains two separate furnaces: one furnace houses the sample pan and the second furnace houses the reference pan. Each furnace houses also a detector. The temperature difference between the two furnaces is the input of a second control circuit. The second control circuit compensates for the additional heat being absorbed or liberated by either increasing or reducing the power supplied to the sample furnace. The sample temperature is thus adjusted (by increasing or decreasing the heat supplied by the furnace to the sample pan) in such a way that its temperature is maintained identical to that of the reference.
Some of the physical phenomena studied with a DSC are glass transitions, melting profiles, heats of fusion, amount of crystallinity, oxidative stability, curing kinetics, crystallization kinetics, and other phase transitions. DSC data can also be used to study the reaction kinetics of phase transitions, but this is beyond the scope of this discussion.
The focus of this chapter is on the use of a heat-flux DSC to characterize the thermal behavior of edible fat samples by obtaining the temperature at which the edible fat melts and crystallizes, as well as the enthalpy of the transitions.
When the DSC is used in the scanning mode, the temperature is changed linearly. The heat flow rate Φ (Eq. 1) is proportional to the heating rate
with K as the proportionality factor. The differential heat flow rate depends on the differential heat capacity between a pan containing the sample and a pan used as reference usually filled with “air” and called empty, and heating rate. The measured heat flow in scanning mode is never zero and is made up of three parts (Eq. 2)
is due to the difference in temperature between the sample and reference positions, is due to the difference in heat capacity between sample and reference, and is the heat flow contribution from the latent heat of the transition in the sample. The first and second terms define the baseline and the third term defines the 'peak' of the measured curve.
This section uses the TA Q2000 DSC (TA Instruments, New Castle, Delaware, USA) to illustrate how a DSC works, but the principles apply to all heat-flux DSCs. The readers should refer to their own manuals for instrument/mode/maker specifications.
The TA Q2000 DSC consists of four components (Fig. 1): a furnace or cell, a cell controller, an external cooling system (RCS), and a purge system.
A DSC is a twin instrument, comprising one furnace or enclosure where the individual sample and reference platforms or calorimeters are positioned (Fig. 2). The term “calorimeter” refers to each raised platform and its respective detector. The enclosure is the cell seen in Fig.1.
Fig. 2 shows a schematic diagram of the inside of the cell as well as the location of both pans. Sample and reference pans are placed on top of the two platforms projecting from the sensor base. The cell is heated at a linear heating rate. The two calorimeters are part of a larger unit, the “sensor” (Danley, 2001). Each calorimeter will measure heat flow as a function of temperature or time. The two calorimeters are assumed to be identical. The sensor body is made out of constantan (an alloy made of 55% copper and 45% nickel), consisting of a thick flat base and a pair of raised platforms where the sample pan and reference pan are positioned for the analysis (Fig. 2 insert). The thin wall of the sensor creates the thermal resistance. This constantan material provides good thermal conduction to the sample and the reference. As the temperature of the furnace is changed, heat is transferred from the silver base of the enclosure to the sensor body and thus to the pans. The temperature of the furnace is controlled by the refrigerated cooling system (RCS). Thermocouples on the underside of each platform are used to measure the temperature of the sample and the reference. A third sensor is used to measure the temperature of the sensor base.
Major sources of error in the measurements are the effects of differences in the sample and reference sensor heating rates as well as the sample and reference pan heating rates during a phase change (Cassel, TA Instrument). The TA Q2000 uses a Tzero™ technology that accounts for this and includes these effects in the heat flow measurement. The result is sharper onsets, higher peaks and faster baseline returns after transitions.
The Cell Controller
The cell controller houses all the electronics and is the heart of the DSC. It consists of a touch display window and an automatic cell that can be opened or closed from the commands on the touch screen window. The controller connects to the computer from which all commands are given to the instrument. The TA Q2000 controller houses a computer inside it.
The Cooling System and Purge Gas
The refrigerated cooling system (RCS) provides cooling or heating to the cell in order to precisely control the temperature of the sample and the reference pans, allowing an operating temperature range from -90 to 550°C. The RCS is turned on/off directly from its switch and also from the touch screen, or it can also be controlled directly from the software in the computer selecting the option events. The RCS system is refrigerated by two different materials, the first stage uses propylene while the second stage uses a blend of ethylene and propane.
A purge gas is used to continuously purge the cell. The purge gas helps with the heat transfer efficiency inside the cell by eliminating hot spots and removing any off-gasses and moisture emanating from the pan or sample. Nitrogen is most commonly used as a purge gas as it is inexpensive, inert and easily available. In addition, it does not interfere with heat measurements due to its low thermal conductivity. This purge gas gets heated before entering the cell in order to equilibrate its temperature with that of the cell.
The RCS also requires a purge gas to purge the interior of the RCS cooling head when the cell is open during loading/unloading. At the back of the DSC it should be seen that the two lines of the purge gas are connected to the instrument.
A heat-flux instrument like the TA Q2000 computes the heat flow from calibrated files, hence these files are very important.
The calibration files for the TA Q2000 are obtained from three runs that can be carried out by following the calibration wizard in the instrument’s operating software. Typically, a calibration is performed once a year or after a major change in the system takes place. Calibration results are saved and reused each time an experiment is run. The calibration is re-done when key experimental conditions are changed (e.g., a different purge gas is used and/or heating/cooling rates are altered) or when the sample cell is burnt. The cell is burnt when there are too many contaminants inside the cell. Burning the cell comprises a step where cell is taken to a high temperature, like 350°C or even 550°C if the system allows for it.
As a heat-flux instrument, the TA Q2000 calorimeter was been designed so that the heat flowing into one of the platforms does not affect the temperature of the other. The reference platform is then independent of the sample platform. A constantan and a chromel wire are welded to the centre of the sensor base structure and constitute the cell sensor that reports To. The instrument reports the temperatures, Ts and Tr, of the chromel disks attached to the sample platform and the reference platform, respectively, with respect to the body detector (Danley, 2003) in a way that the differential temperatures are defined as:
where To = Tchromel wire – Tconstantan is the temperature at the base of the sensor body (see Fig. 2), and subscripts r and s refer to reference and sample, respectively.
The algorithm followed by TA instruments is called the “the lumped heat capacity” method (Danley, 2003) where the signal is converted to a heat flow rate, Φ, using a temperature dependent proportionality factor so that:
Each cell within the calorimeter comprises a thermal resistance, R, and a heat capacity, C. The heat flow rate for the sample and reference is written as:
Calibration for the TA Q2000 instrument involves the determination of the values for Rs, Rr, Cs and Cr.
Physically, the operator needs to perform three different experiments: (1) empty cell or no pans, (2) using two discs of sapphire that are placed on the sample and reference platforms according to the instructions provided by TA Instrument and (3) using a sealed sample pan with a high purity material, usually indium. Experiments 1 and 2 constitute the baseline calibration, while experiment 3 is the temperature and heat of fusion calibration.
The baseline calibration compensates for subtle differences between the reference and sample thermocouples.The output of this calibration is the calculation of a baseline slope and offset values, which are used to flatten the baseline and zero the heat flow signal. Most instruments perform only one run for this calibration, which consists of heating the empty cell through the instrument’s temperature range. The TA Q2000 DSC calibration also includes a second run with two sapphire disks, without being sealed inside a pan. These two disks are used without the pans because this material gives equal heating rate as it has no discontinuous (first order) phase transitions in this temperature range. It is recommended to perform these two calibrations for the maximum temperature range of interest for fats (-70°C to 350°C). The manual recommends a heating ramp of 20°C/min.
The temperature and heat of fusion calibrations ensure that the sample thermocouple reading is correct under the experimental conditions chosen. It involves the use of standard materials, such as high-purity metals, indium, and gallium, which are melted at the same heating rate used in the analysis of future samples. High-purity materials are used because they give narrow peaks. This calibration gives three different outputs: the cell constant, the onset slope and the temperature calibration (Thermal Advantage Manual, 2000). The cell constant is the ratio between the theoretical heat of fusion and the measured experimental heat of fusion of the standard. The onset slope or thermal resistance is a measure of the temperature drop that occurs in a melting sample in relation to the thermocouple. Theoretically, a standard high-purity sample should melt at a unique temperature. However, as it melts and draws more heat, a temperature difference develops between the sample and the sample thermocouple. The thermal resistance between these two points is calculated as the onset slope of the heat flow versus temperature curve on the low-temperature side of the melting peak. The temperature calibration makes a correction based upon the difference between the observed and the theoretical melting temperature of the high-purity metal used. If only one standard is used, then the calibration shifts the sample temperature by a constant amount. A two-point calibration shifts the temperature with a linear correction (straight line) and projects this correction to temperatures above and below the two calibration points.
Pans are made from gold, copper, aluminium, graphite and platinum or alodine-aluminium (Guide for choosing DSC pans, TA Instrument). The lids can either provide a complete enclosure or can contain a pin hole to release the pressure that builds up inside, due to evaporation of water as the temperature is increased. Most samples can be run in non-hermetically sealed pans either uncovered or crimped with an aligned or inverted cover. Atmospheric interaction is optimised by using an open (uncovered) pan. Crimped pans improve the thermal contact between the sample, pan and disc, reduce thermal gradients in the sample, minimize spillage, and enable retention of the sample for further study. In our laboratory, fat samples are placed in alodine-aluminium pans containing no more than 10 mg of sample that are hermetically sealed before the study is carried out. These pans are coated with an inert fluoro-phosphate layer which gives the pans a slightly yellow or gold colour rather than the typical silver colour of aluminium. This coating renders the pans inert to many chemicals.
Calorimetric measurements require that the instrument be set to its experimental mode. The operator can use the instrument’s operation software wizard to define and edit an experimental protocol. The software from TA is self-explanatory and employs one-line pre-determined instructions or segments that aid in the creation of the protocol.
The experiment procedure requires: 1) booth of the system and opening of purge gas 2) preparation of sample pan 3) loading of samples and reference pan 4) set up of the protocol that the instrument needs to follows, 5) analysis of the data.
Sample preparation involves the encapsulation of the material in pans designed for calorimetric analysis. The pan requires between 1 and 10 mg of sample. The amount of sample depends on the material used. Care should be taken not to overfill the pan to the point that the sample spills when the pan is being sealed. Overfilled and leaking pans will lead to erroneous readings and most critically, contamination of the cell. The software uses the weight of the empty pan as well as the weight of the sample. It is important to accurately measure these values so that the transition enthalpy is correctly reported in units of J/g. The sample pan is composed of two pieces, a bottom, and a cover. Once the sample is deposited in the indent at the bottom of the pan, the pan is sealed using a press specially designed for this purpose. It is recommended that gloves or tweezers be used all the time to prevent any contamination from oily residues on the surfaces of the hands. The manipulation and transportation of the pans from the weighing and/or encapsulation station to the DSC cell should be done without introducing any contamination.
When the temperature of crystallization or storage is different from room temperature, all tools and pans must be pre-tempered. Depending on the experiment, it may also be advisable to prepare (and seal) pans in a temperature-controlled environment. In this case, the DSC cell should also be pre-tempered. The user must keep in mind that any temperature variations will affect the thermal behavior of the sample (be it crystallization or melting) and hence affect results.
Loading the Sample
The sample can be loaded any time that the measuring cell is not in use. Care should be taken not to introduce any thermal changes in the sample when it is deposited inside the cell. It is recommended to pre-heat or pre-cool the cell to the desired starting temperature before opening the cell to deposit the pan. Both reference and sample pans should be well centered into each corresponding platform (Fig. 2). The reference pan is placed in the rear position of the cell. Tweezers and/or gloves might be necessary when handling the pans so as to not introduce any contamination.
Creating the Protocol for Analysis
Two different types of phenomena/phase changes can be studied with a DSC: crystallization and melting. The melting properties of a crystallized fat are investigated by heating the sample at a controlled rate until the fat is completely melted. The thermal history of the sample will determine the melting profile. The thermal history includes the conditions of crystallization prior to the DSC melt, the storage time and temperature. All of these variables, together with the heating rate will determine the outcome of the melting process. In the crystallization process, care is taken to erase all crystal memory from the sample (at the recommended 80°C for 15 minutes), thus results will not be affected by thermal history.
Depending on what one is studying, the test program will go either from a low temperature to a high one, or vice versa, including some holding time at particular temperatures.
An example is given here for a method used to study fully hydrogenated soybean oil (FHSO) (Bunge, Toronto, Ontario, Canada), which has a melting point of approximately 65°C. The protocol followed is given in Table 1. The objective of this example is to find the temperatures at which the fat undergoes phase changes from the liquid to solid and solid to liquid state. Most researches use 5°C/min for both melting and cooling. This rate seems to be good compromise between the run time and the precision of the results.
The following is a step-by-step explanation of the method outlined in Table 1:
- Step 1 is performed to set the starting temperature.
- Step 2 brings the temperature of the cell from 24°C to 80°C at a rate of 5°C/min. The sample will melt during this step.
- Step 3 holds the temperature at 80°C for 15 minutes. This is done in order to erase all memory of the crystal structure.
- Step 4 reduces the temperature from 80°C to 24°C at a rate of 5°C/min. The sample will crystallize during this step.
- Step 5 maintains the temperature at 24°C for 15 minutes. This is to ensure complete crystallization of the sample as well as to equilibrate the material at this final temperature. If the user wants to study storage time, then the pan is removed after this step and moved to an incubator at 24°C to be melted at a later time. For the purpose of this example, it was enough to keep it 15 minutes; this equilibrates the temperature of the material, without any concerns about polymorphic transformations.
- Step 6 completes the thermal study of the sample. During this ramp, the material will melt and the onset can be obtained from the observation. The thermal history needs to be reported as the conditions provided in steps 4 and 5.
Before going into details on how to analyse the data, one needs to understand the kind of information that it is obtained from the DSC.
Thermograms associated with phase changes of fat samples are created as the temperature is increased or decreased at a controlled rate. The cooling or heating rates are set by the operator. Thermograms can also be created as heat flows as a function of time at a constant temperature. Typical thermograms display the difference in heat flow on the Y axis, while the temperature or time is plotted along the X axis. Parameters commonly analyzed are the temperatures of the summit of the peak of crystallization, TC, the summit of the peak of melting, TM, the onset temperatures for crystallization, TOC and for melting, TOM and the enthalpy of crystallization, ΔHC, and melting ΔHM (Fig. 3).
The temperature at which the phase change peak (be it melting or crystallization) displays the lowest or highest heat flow value is the summit of the peak and is associated with the temperature at which at least half of the lipid species have gone through the phase transition. Associated with this peak is the onset temperature (TOM or TOC), defined as the temperature at which the first crystallites melt or form, which is observed as a deviation from the baseline. The area under the peak corresponds to the enthalpy of the phase transition ΔH. The enthalpy of a system is equal to the energy added through heat, only if the system is under constant pressure and when the only work done on the system is due to an expansion caused by heating alone. Under these circumstances, the enthalpy is equal to the heat, Q, supplied by the system so that ΔH = Q. The units are typically Joules (J) or J/g when specific enthalpy is reported.
TA Instrument Universal Analysis software for DSC is used to obtain these three parameters from the raw data. This TA software offers four different profiles (fits) to define the area of the peak. The four options are found under peak analysis-fitting and are named: lineal, sigmoidal horizontal, sigmoidal tangential and extrapolated. The choice of fitting requires that the user select the region of interest (start and finish temperature). When no difference in the flow rate is observed before and after the melting/crystallizing peak, then the linear fitting is advised. In this case, a straight line will be traced between the two selected points. The sigmoidal horizontal fitting automatically creates a curved line with one inflection between the two points, while the sigmoidal tangential option allows the user to manually choose the inflection point. Any of these two fittings is recommended when there is a marked difference between the baseline heat flow before and after the peak. The extrapolated fitting uses a horizontal line regardless of the shape of the peak between the two temperature points of interest.
Crystallization is an exothermic event, where energy from the sample is released during the process. Exothermal and endothermic events are defined by the user. Fig. 3 displays a positive peak for the crystallization peak or exothermic event, but the opposite can also be set. The melting of a fat is an endothermic event, where the sample absorbs energy from the environment. To keep consistency with what was said above, the melting peak has been chosen as a negative one as seen in Fig. 3.
Three different studies on FHSO (Bunge, Toronto, ON, Canada) were carried out and the results shown here. The aim is to illustrate the challenges that one can encounter when studying fats with a DSC. The idea is to highlight the importance of keeping track of the thermal history and the experimental conditions used to “manufacture” the fat.
Samples of FHSO were received from the supplier and stored for 1 year at room temperature (~24°C) before any study was carried out. Specimens of FHSO were hermetically sealed into alodine-aluminium pans. Three samples were analyzed following a melting ramp that started at 24°C and finished at 80°C (steps 1 and 2 in Table 1). Three different heating rates were used: 1, 5 and 10°C/min. Thermograms from each sample were analyzed using a linear fit. The results of the three different melts are shown in Fig. 4.
It can be observed from Fig. 4 that the greater the heating rate, the higher the temperature that correspond to the summit or the melting peak. The summit values obtained were 68.2°C, 69.5°C and 71.6°C for 1, 5 and 10°C/min respectively. However, the onset value for the three melting profiles was similar: 66.3°C. This is showing that the parameter that stays constant with the different cooling rates is the onset temperature. Simulations of melting behaviour had shown that the extrapolated peak onset temperature is relatively independent of experimental parameters (Höhne et al., 2003).
FHSO samples were introduced into hermetically sealed alodine-aluminium pans and heated to 80°C inside the DSC cell (steps 1 and 2 in Table 1). They were kept at this temperature for 15 minutes (step 3 in Table 1) and subsequently cooled to 24°C (step 4 in Table 1) for crystallization at the following cooling rates: 0.5, 1.0, 5.0, and 10.0°C/min. Each cooling rate was carried out using its corresponding calibration file. Thermogram data were analysed using a linear fit in each instance. The results are shown in Fig. 5.
The parameter values obtained from Fig. 5 are reported in Table 2. The results show that cooling rates affect the position of the crystallization peak maximum (summit), while the onset crystallization temperature is similar for the four cases.
The crystallization summit values are not the same as the summit melting values observed under the same cooling and heating rates, respectively (Fig. 4). Each run in this experiment was calibrated using temperature data for heating experiments. TA Instruments reports (Cassel, TA Reference) that unless a complex multiple rate calibration is carried out, there will be an error of about 1 to 2°C in the summit for the crystallization peak compared with the summit for the melting peak (Menczel and Leslie, 1990; Menczel, 1997).
This example shows one of the problems associated with working with fats: re-crystallization. In this case, a sample of FHSO was subjected to the protocol displayed in Table 1. The goal was to melt the sample after it was crystallized under controlled conditions. The melting profile is shown in Fig. 6.
Fig. 6 shows a melting peak with a summit at 52.1°C and a second melting peak with a summit at 62.3°C. Between these two peaks, an exothermic peak is observed. This can be explained by realizing that a polymorphic transformation took place while heating. The first melting peak might be due to the melting of one of the metastable polymorphic forms of a fat, either α or β′. The subsequent increase in temperature seems to trigger a re-ordering of the sample, with an exothermic peak that indicates crystallization. This might cause the rearrangement of the molecules to form a new crystalline form that melts at a higher temperature, as shown by the second melting peak. This is a good example of a sample that should be further studied using other techniques, such as XRD, in order to discover the true polymorphic form present in the original sample.
This example emphasizes the changes that can come about due to storage. Storing a sample for a long time might allow it to equilibrate and undergo a transition to the stable polymorphic form. This example shows that even when the thermal history is the same, the storage time affects the final results.
Samples were melted in the DSC at 80°C and kept at that temperature for 15 minutes, after which they were cooled to 24°C at a rate of 5°C/min. Two different crystallization/storage times were evaluated. One sample was kept isothermally at 24°C in the DSC cell for 15 minutes prior to the melt (solid line in Fig. 7). The other sample was transferred to a temperature-controlled incubator set at 24°C for 72 hours and subsequently melted in the DSC (dashed line in Fig. 7). A rate of 5°C/min was used to melt the samples from the incubator.
This experiment shows the importance of the history of the sample as well as understanding the dynamics of polymorphic changes in the sample. The main issue always is whether a specific polymorphic form was originally present in the sample or if it was created upon heating while carrying out the DSC measurement. Results for the sample stored for 15 minutes shows two peaks that were identified as α and β’ (Small, 1966). The sample stored for 72 hours also showed these two peaks together with a third one that was identified as the β polymorphic form (Small, 1966). It appears that having kept the sample at a constant temperature for 72 hours induced a solid-state polymorphic transformation to the β phase.
Every operator is encouraged to spend some time familiarizing with the equipment by thoroughly reading the manuals before performing any experiments. Investing time to understand the basic operation of the equipment, as well as the theory behind the technique, will guarantee the success of the measurements. Avoiding this step may lead to erroneous results that might put the validity of the results in jeopardy. Training sessions with experts in the area, specifically with the ones that built the equipment, might be a reasonable way to answer some of the inevitable questions. A good result is not equipment-dependent and its reproducibility should be universal, assuming that the correct technique has been used.
The thermal behaviour of fats and oil can be characterized easily and systematically using a differential scanning calorimeter. In a heat-flux DSC, the calibration of the DSC is of outmost important as it is from those values that the heat flow is computed. The onset value is a parameter that can be used to report both the onset temperature of melting and the onset of crystallization, as it is not heating/cooling-rate dependent. The drawback is that one does not know the temperature at which all the sample was melted/crystallised. The enthalpy is the energy that a system absorbs or releases during a phase transition. DSC results show that the enthalpy values obtained when using different cooling or heating rates are not the same.
One has to have in mind that small variations in the instrumentation or in the handling of the sample, the pans, and the cell might lead to discrepancies in the values obtained at different heating/cooling rates for this parameter. It is then imperative to carry out multiple repetitions, using new pans and specimens.
Regarding calibration, whenever possible, the operator should use a calibration routine that calibrates simultaneously for heating and cooling.
- Danley, R. New heat flux DSC measurement technique. Thermochim. Acta, 295, 201-208 (2003).
- Danley, R.L and Caulfield, P.A. DSC Baseline Improvements Obtained by a New Heat Flow Measurement Technique. TA Instruments, New Castle, Delaware 19720 USA (2001).
- Cassel, R.B. Communication from TA: How Tzero™ Technology Improves DSC Performance Part VI: Simplifying Temperature Calibration for Cooling Experiments. TA Instruments, New Castle Delaware 19720, USA.
- Guide for Choosing DSC pans. TA Instruments. Thermal applications note. TN-12.
- Höhne, G.W.H., Hemmiger, W.F. and Flammersheim, H.J. Differential Scanning Calorimetry. (Springer-Verlag Berlin, Heidelberg) (2003).
- Menczel, J.D. Temperature calibration of heat flux DSC’s on cooling. J. Thermal Anal., 49, 193-199 (1997).
- Menczel, J.D., and Leslie, T.M. Temperature calibration of a power compensation DSC on cooling. Thermochim. Acta, 166, 309-317 (1990).
- Small, D.M. Handbook of Lipid Research. (Plenum Press, New York, and London) (1966).
- Thermal Advantage Manual. DSC User Reference Guide (TA Instruments, New Castle, Delaware, US) (2000).
Update: February 2, 2017