TOPEM, a new temperature modulated DSC technique: A critical review

Transcription

1 FACULTY OF PHARMACEUTICAL SCIENCES INSTITUTE FOR PHARMACEUTICAL TECHNOLOGY AND BIOPHARMACY UNIVERSITY OF DUESSELDORF Academic year TOPEM, a new temperature modulated DSC technique: A critical review Laurine PARMENTIER First Master of Pharmacy, Pharmaceutical Care Promotor Dr. M. Thommes Co-promotor Prof. Dr. C. Vervaet Commissioners Prof. Dr. W. Baeyens Dr. K. Remaut 1

2 De auteur en de promotor geven de toelating deze masterproef voor consultatie beschikbaar te stellen en delen ervan te kopiëren voor persoonlijk gebruik. Elk ander gebruik valt onder de beperkingen van het auteursrecht, in het bijzonder met betrekking tot de verplichting uitdrukkelijk de bron te vermelden bij het aanhalen van de resultaten uit deze masterproef. 28/05/2009 Promotor Prof. Dr. C. Vervaet Auteur Laurine Parmentier 2

3 Word of Thanks I would like to thank Prof Dr. C. Vervaet and Prof. Dr. P. Kleinebudde for generally leading my master thesis. I would like to express my warmest thanks to Dr. M. Thommes for helping me throughout this study, especially for his patience and great help conducting me into the right direction. I would also like to thank the whole staff of the institute for the pleasant environment and especially Ms. K. Matthée who stood by me from the first day. I would like to thank all the friends, especially F. Koppers for always being ready to help. Further, I would like to thank my parents for their eternal support. Finally, I would like to thank the Erasmus programm, which made this whole experience possible. 3

4 LIST OF CONTENTS WORD OF THANKS LIST OF CONTENTS LIST OF ABBREVIATIONS 1 INTRODUCTION Thermal analysis In general Theory First law of thermodynamics Second law of thermodynamics Thermal analysis methods Differential thermal analysis Differential scanning calorimetry General principles Power compensation DSC Heat flux DSC Temperature modulated differential scanning calorimetry General principles Isothermal step method differential scanning calorimetry Alternating differential scanning calorimetry TOPEM MATERIALS AND METHODS Materials Methods Differential scanning calorimetry Balance

5 2.2.3 α-mannitol Water uptake of sugars OBJECTIVES RESULTS Preliminary tests Reversing and non-reversing heat flow In general Griseofulvin Sucrose-water mixture Moist substances Calculation window In general Mannitol Calculation window parameters Effect of the pulse width on the calculation window Heat capacity Frequency Resolution CONCLUSIONS LITERATURE LIST

6 LIST OF ABBREVIATIONS ADSC Cp Cp* DMA DSC DTA HR ISMDSC m p PET PH PVP TGA TMA TMDSC TOA USP Alternating Differential Scanning Calorimetry Heat capacity Complex heat capacity Dynamic Mechanical Analysis Differential Scanning Calorimetry Differential Thermal Analysis Heating rate Isothermal Step Method Differential Scanning Calorimetry Sample mass Period Polyethylene terephtalate Pulse height Polyvinylpyrolidone Thermogravimetric Analysis Thermal Mechanical Analysis Temperature Modulated Differential Scanning Calorimetry Thermooptical Analysis United States Pharmacopeia 6

7 1 INTRODUCTION 1.1 THERMAL ANALYSIS IN GENERAL In pharmaceutical research, the development of new medicines is the main goal. Besides that, already existing products need to be evaluated. There are several ways to determine the characteristics of drugs during and after the development. Information is needed about several characteristics such as purity, interactions, solubility, molecular weight, structure and melting point. All these physicochemical properties can be determined with a variety of methods and devices. A few examples are chromatographic methods, spectroscopy, titration and thermal analysis. In thermal analysis, the behavior of a substance is analyzed when a certain temperature program is applied. Thermal analysis is a group of techniques in which a physical property is measured as a function of temperature. Therefore the substance is subjected to a controlled temperature program. This can be a dynamic or an isothermal program. The obtained curve is the measured characteristic in function of the time or the applied temperature. (Schwarz & de Buhr, 1998) Thermal analysis includes several techniques of which the most important are: differential scanning calorimetry (DSC), differential thermal analysis (DTA) thermogravimetric analysis (TGA), thermal mechanical analysis (TMA), dynamic mechanical analysis (DMA) and thermooptical analysis (TOA). TOA is the visual perception or measurement of the permeability of light or light reflection of a sample. In DMA, a force is supplied to the sample whereby the sample deforms and information about the stiffness (viscosity and elasticity) can be determined. In TMA, the deformation (change in width or length) of the sample under a load is measured. In TGA, the mass of the sample is measured, under a defined atmosphere. DTA measures the temperature difference between a sample and a reference. DSC contains several techniques which will be discussed later. All the techniques for thermal analysis are measured with an underlying temperature program. (Schwarz & de Buhr, 1998) It is the latest new DSC method, TOPEM, which will be the subject of the research in this study. 1

8 With thermal analysis methods, several characteristics can be measured: melting point, crystallization behavior, glass transition, expansion, mass loss, thermal stability, decomposition temperature, oxidation time, network formation, purity, viscoelastic behavior and chemical reactions. (Schwarz & de Buhr, 1998) 1.2 THEORY FIRST LAW OF THERMODYNAMICS The first law of thermodynamics is an expression of the law of conservation of energy. No energy can be destroyed or created, only changes in energy can take place. The internal energy is the sum of heat and work done by the system. (Van der Plaats, 1992) U = Q + W (1) Where: U: internal energy (J) Q: heat (J) W: work (J) A calorimeter for example is a system in which mass and pressure are normally constant. This means that the energy is only depending on temperature and volume. U = U (T, V) (2) Where: U: internal energy (J) T: temperature ( C) V: volume (m 3 ) A new state function is introduced, namely the enthalpy. State functions are functions which always give the same result when going from point A to B, independent from the way which is taken to get there. The change of enthalpy equals the heat under constant pressure. Further is the work given by the pressure and volume. The minus sign means that the work is done by the system. This means that equation 1 can be rewritten using equations 3 and 4. (Van der Plaats, 1992) ( H) p = Q (3) W = -pv (4) (dh) p = du + pdv (5) 2

9 Where: U: internal energy (J) H: enthalpy (J) p: pressure (Pa) V: volume (m 3 ) Q: heat (J) W: work (J) Under constant pressure, the heat given to the system is the same as the increase in enthalpy. Therefore, the heat capacity of a sample under constant pressure is the change of enthalpy in time. (Van der Plaats, 1992) c p = p (6) Where: c p : heat capacity (J/K) H: enthalpy (J) T: temperature (K) p: pressure (Pa) SECOND LAW OF THERMODYNAMICS There are different ways to determine the second law of thermodynamics: It s impossible to transfer heat from a cold to a warm place, without changing the environment It s impossible to transform heat completely without changing the environment A new state function is introduced, namely the entropy. The entropy is heat divided by temperature, in a reversible process. (Van der Plaats, 1992) S = (7) Where: S: entropy (J/K) Q: Heat (J) T: temperature (K) Using equation 4 and 7, equation 1 can be rewritten as follows: (Van der Plaats, 1992) du = TdS pdv (8) Where: U: internal energy (J) T: temperature (K) S: entropy (J/K) p: pressure (Pa) V: volume (m 3 ) 3

10 Further is also the Gibbs free energy defined as G = H TS. A system will always try to find the lowest free energy state. According to the equation, this means that the enthalpy has to be minimal and entropy has to be maximal to achieve this. The Gibbs free energy is minimal when the system is at equilibrium under constant pressure and temperature. However, pressure and temperature can be influenced in thermal analysis. Therefore, the Gibbs free energy cannot reach the mininmum. (Van der Plaats,1992) G = H TS (9) Where: G: Gibbs energy (J) H: enthalpy (J) T: temperature (K) S: entropy (J/K) 1.3 THERMAL ANALYSIS METHODS DIFFERENTIAL THERMAL ANALYSIS DTA is the oldest technique in thermal analysis. In DTA, a sample and reference undergo the same temperature program. The sample and reference are put in a pan. As a reference, mostly an empty pan is used. When a reaction occurs in the sample, energy is needed (for an endothermal reaction) or released (for an exothermal reaction) and therefore a temperature difference is measured. The temperature difference is measured between the sample and the reference by two thermal elements, one for each. When a temperature difference is measured, the oven will compensate this, so that both sample and reference have the same temperature. The temperature difference is used as a measured value. Any change in temperature is detected with DTA. (Widman & Riesen, 1984) DIFFERENTIAL SCANNING CALORIMETRY General principles As in DTA, in DSC a sample and reference, put into a pan, undergo a dynamic or isothermal temperature program. Both sample and reference are placed on identical measuring cells next to each other in an oven. The temperature difference is compensated by the measurement cells and not by the oven. Because the sample and reference have their own measuring cell, the detection of the temperature difference goes faster and with more sensitivity as in DTA. The output value is the 4

11 heat flow. This heat flow is actually measured as a heat power, with the unit Watt. By integration of the heat flow in time, the enthalpy of a sample is calculated, with the unit Joule. (Widman & Riesen, 1984) Power compensation DSC The sample and the reference both have their own measuring cell and oven. Both ovens separately follow a temperature program. When there is a thermal symmetry between both systems, the same amount of power will be needed to heat them. When a transition (endothermal or exothermal) occurs, this will lead to a temperature difference. A proportioning controller will be activated and puts as much heat in the sample or cools it down as needed for the transition. That s why the temperature difference is kept a constant. (Widman & Riesen, 1984) The compensated power is proportional to the measured temperature difference. P = -K 1 T (10) Where: P: power (W) T: temperature (K) K: calibration factor, fixed value of the controller When a good calibration is used, the temperature difference can be used to measure the heat flow in the sample. Φ S = -K 2 T (11) Where: Φ: heat flow (W) T: temperature (K) K: calibration factor FIGURE 1.1. CROSS-SECTION OF THE OVENS USED IN POWER COMPENSATION DSC. ( 5

12 Heat flux DSC In heat flux DSC, the sample and reference are placed in one oven. The heat flows from the oven, over a defined heat resistance to the sample and reference. The driving force for the heat flow is the temperature difference over the heat resistance. The heat flow to the sample is different to the one to the reference. (Widman & Riesen, 1984) When heat flows under steady state conditions, the temperature difference between the reference and sample is proportional to the difference in heat flow from the oven to the sample (OS) and to the reference (OR). Φ = Φ OS Φ OR = -K T = Φ S (12) Where: Φ: heat flow (W) T: temperature (K) O: oven R: reference S: sample K: calibration factor that includes the asymmetry of the system By measuring the temperature difference between a monster and reference, using this equation, the heat flow Φ S can be measured. A B FIGURE 1.2. A) CROSS-SECTION OF THE OVEN USED IN HEAT FLUX DSC. B) TOP-VIEW OF THE SENSOR USED IN HEAT FLUX DSC. S = SAMPLE, R = REFERENCE ( 6

13 1.3.3 TEMPERATURE MODULATED DIFFERENTIAL SCANNING CALORIMETRY General principles In conventional DSC, the sample undergoes a linear temperature program. The output value is the difference in heat flow between the sample and reference. In TMDSC however, a temperature modulation is superimposed over the linear temperature program. This allows us to separate the heat flow into two components, the reversing and non-reversing heat flow. The separating of these can be useful when overlaying effects occur or thermal events cannot be clearly identified with classical DSC. (Cao, 2007) The reversing heat flow is the sensible heat flow component. This is associated with processes that can be reversed. Thermodynamically seen, these are processes that occur close to a local metastable state. When a process is reversible, the measured curve should be reproducible after cooling the sample down and reheating it. For such processes the heat input results in a temperature change. Some examples are: glass transitions, melting of polymers, temperature change when no thermal transition occurs. (Schawe, 2005a) The non-reversing heat flow is the latent heat flow component. It is associated with irreversible processes. Thermodynamically seen, such processes start in a nonequilibrium state and end in an equilibrium state. They are time dependent. The heat input doesn t change the temperature, because the heat is used for changes in the sample. Such changes take all the energy of the heat and therefore no heat is left for a temperature increase. Examples are: non-reversible chemical reactions, melting, and crystallization. (Schawe, 2005a) For measuring the change of heat capacity with time or temperature, temperature changes of the sample are measured. Due to that temperature difference, there s a change in enthalpy, given by H = m c p T (13) Where : H: enthalpy (J) m: sample mass (mg) c p : heat capacity (J/K) T: temperature (K) 7

14 Next to the heat flow as a result from the heat capacity, there are other energy changes due to chemical reaction or physical transitions. Therefore we should add the heat related to chemical reactions, related to physical transitions and the drift of the measuring system to equation 13. With respect to time, we can also split the heat flow in the sensible heat flow and latent heat flow. (Riesen 1994) Φ = m c p β Φ b + Φ r + Φ t (14) Φ s Φ l Where: Φ: heat flow (W) Φ r : heat flow of chemical reaction (W) m: sample mass (mg) Φ t : heat flow of physical transition (W) c p : heat capacity (J/K) Φ b : drift of the measurement (W) β: heating rate (K/min) Φ s :sensible or reversing heat flow (W) Φ l :latent or non-reversing heat flow(w) In conventional DSC the temperature program is given by T = T0 + ß t. In temperature modulated DSC, a temperature modulation is applied to a programmed underlying heating rate and therefore a time-dependent function is added to the last equation. This function is a saw tooth (ISM DSC), a sine wave (ADSC) or a stochastic modulation (TOPEM). T = T 0 + ß t + f(t) (15) Where: T: temperature ß: heating rate T 0 : start temperature t: time f(t): function that describes the temperature modulation Isothermal step method differential scanning calorimetry In ISM DSC, the temperature program consists of alternating heating and isothermal segments. The isothermal segment should last until the heat flow remains constant. If the temperature steps are too small, it is possible that the sample cannot follow. When the steps are big enough, quasi-static conditions can be achieved. This means that the sample can follow the set oven temperature and has time to equilibrate at the isothermal parts. When the sample can follow well, there is a good 8

15 separation of reversing and non-reversing heat flow. If the sample doesn t have the time to equilibrate, this separation cannot be completed. (Schawe, 2005b) FIGURE 1.3. TEMPERATURE PROFILE OF AN ISM DSC MEASUREMENT (METTLER-TOLEDO 2009) Alternating differential scanning calorimetry The temperature program in ADSC is a periodic sinusoidal change. Several frequencies can be chose, in a range from to 0.1 Hz. The disadvantage however is that for every new selected frequency, a new measurement has to be done. When low pulse heights and low underlying heating rates are used, there is an increase in resolution. (Schawe, 2005b; Jörimann et al. 1999) T(t) = T 0 + βt + A T sin(ωt) (16) ω = 2π/p (17) Where: T: temperature (K) t: time (s) ω: angular frequency (rad/s) p: period (s) T 0 : start temperature (K) β: heating rate (K/min) A T : amplitude of temperature modulation (K) FIGURE 1.4. SINUSOIDAL HEATING RATE OF AN ADSC MEASUREMENT (METTLER-TOLEDO 2009) 9

16 From the raw data, the total heat flow is calculated, by averaging the signal. This is the same approach as in conventional DSC measurements. The measured heat flow is also a sine wave, but has a phase shift or phase lag compared to the applied heating rate. (Jörimann, 1999) Φ = A Φ cos(ωt + φ) (18) Where: Φ: heat flow (W) A Φ : amplitude of the heat flow (K) t: time (s) ω: angular frequency (rad/s) φ: phase (rad/s) For a certain frequency f, the heat capacity c p (f) can be calculated from periodic changes in heat flow. Further on, the reversing heat flow and heat capacity can be measured. The non-reversing heat flow is the difference between the total heat flow and reversing heat flow. (Jörimann, 1999) Φ rev = β c p (19) c p = A Φ (20) Φ rev = β A Φ (21) Φ non = Φ tot - Φ rev (22) Where: β: heating rate (K/min) c p : heat capacity (J/K) p: period (s) A Φ : amplitude of heat flow (K) A T : amplitude of temperature modulation (K) Φ tot : total heat flow (W) Φ rev : reversing heat flow (W) Φ non : non-reversing heat flow (W) TOPEM In general In the TOPEM method, the temperature program is a stochastically changing pulse. This means that the pulse widths change at random between a chosen minimum and maximum. Also the pulse height has to be chosen. Figure 1.5. shows an example of how the pulse can look like. 10

17 FIGURE 1.5. EXAMPLE OF POSSIBLE PULSES IN TOPEM. THE VERTICAL ARROW IS THE PULSE HEIGHT. THE HORIZONTAL ARROWS ARE TWO POSSIBLE PULSE WIDTHS. (METTLER-TOLEDO 2009) The stochastic pulses contain many different frequencies. After the measurement, the heat capacity is calculated. This curve is then used for the frequency evaluation. For a certain frequency, out of the heat capacity curve a new heat capacity curve is calculated, which is called the complex heat capacity. Several frequencies can be chosen, ideal for analyzing frequency dependent effects. The heat capacity is determined under quasi-static conditions. The calculated heat flow is the sum of the sensible and latent heat flow. (Schawe, 2005b) The chosen temperature program defines the input signal. The output signal is the measured heat flow. A mathematical procedure calculates the correlation between the input (heating rate) and output signal (heat flow). After this procedure, a component is obtained related to the input signal and one that doesn t. The correlated one is the reversing heat flow, the non-correlated one is the non-reversing heat flow. (Schawe et al., 2005e) The temperature program is defined by the underlying heating rate, switching time and pulse height. The heating rate cannot be too high so that enough data points are obtained. The switching time of the pulse width is set between 15 and 30 seconds. For low frequency measurements, a high switching time is needed and vice versa. Too low or too high switching times result in noisy results. The pulse height is determined by the behavior of the sample. The sample has to be able to follow the oven. A sample is able to follow a little pulse height better than a high pulse height 11

18 because it needs less time to equilibrate. On the other hand, more information is gained with higher pulse heights because the sample is heated for a longer period. This means that a compromise has to be found between higher and lower pulse heights. FIGURE 1.6. TYPICAL TEMPERATURE PROFILE (LEFT) AND CORRESPONDING HEAT FLOW (RIGHT) IN A TOPEM MEASUREMENT (SCHAWE, 2005B) Advantages and disadvantages of TOPEM The new TMDSC method has many claimed benefits: - In one measurement, there is a simultaneous determination of sample properties as a function of time and temperature over a wide frequency range. - Very accurate determination of the quasi-static specific heat capacity by calculation of this heat capacity from the pulse response. - High sensitivity and high resolution allows the measurement of low energy transitions or effect close to another. - Separation of sensible and latent heat flow. Heat capacities can be determined even if the effects overlap. - Simple interpretation of the curves. Effects depending on frequency can easily be distinguished from those independent of frequency. - The dynamics of the system can be analyzed over a broad range of frequencies in one single measurement. (Schawe 2005b) 12

19 The disadvantages of TOPEM are: - Melting of pure substances cannot be measured. During the melting, the sample temperature doesn t change and the sample can therefore not follow a temperature modulation. (Schawe 2005b) - For the evaluation, a calculation window has to be set. The recommended width of that calculation window by Mettler-Toledo is one third of the transition interval. This is however not strict enough. It should be less than about one tenth of the transition interval. (Fraga et al. 2007) - The limit frequency for TOPEM is 4 mhz, independent on the experimental or calculation parameters. This means it is intrinsic to the TOPEM method. (Fraga et al. 2007) - The measured complex specific heat capacities at a selected frequency shift to higher temperatures for higher sample mass. (Fraga et al. 2007) - In comparison to ADSC, the measured temperature of a transition is significantly larger. This is because both use a different temperature modulation. (Fraga et al. 2007) Theory For the measured heat flow both the reversing and non-reversing heat flow can be calculated. The total heat flow is the sum of the reversing and non-reversing heat flow according to the equation Φ tot = Φ rev + Φ non (23) Φ rev = m ß (24) Where: Φ tot : total heat flow (W) Φ rev : reversing heat flow (W) Φ non : non-reversing heat flow (W) m: sample mass (mg) : quasi-static heat capacity ß: heating rate (K/min) If the temperature modulation is sufficiently small, it is assumed that the current state of the sample is almost unaffected and that it is in equilibrium. Therefore, in a limited temperature range the sample can be described as a linear system. (Schawe, 2005d) 13

20 u(t) g(t) y(t) u(t) = T(t) linear system y(t) = Φ measured (t) Temperature as a function of time The system is defined by the sample and instrument Heat flow as a function of time. FIGURE 1.7. GENERAL PRINCIPLE OF THE TOPEM TECHNIQUE For a linear time dependent system with an input signal u(t) and output signal y(t), the correlation is given as: (TOPEM The new advanced multi-frequency TMDSC technique) y(t) = g(t) u(t) (25) Where: y(t): output signal g(t): pulse response of the system u(t): input signal t: time (s) The pulse response is calculated with a z-transformation. Generally, a transformation is the description of the signal with another set of basic units as was the case before. The z-transformation is the discrete analog of the Laplace transformation. It is used to convert differential equations which come up for time discrete processes. The time-domain signal is converted by the z-transformation into a complex frequency-domain representation. Equation 25 can be rewritten in the z- plane as y(z) = H(z)u(z) (26) Equations of the time-domain can be described and solved more easily in the z-plane. Several functions can be used to solve H(z), but often it can be exactly described with a rational function. (27) Where B(z) and A(z) are polynomials of degree q or p in the variable z. 14

21 By using the last equation, y(z) can be rewritten as y(z) = u(z) (28) or A(z)y(z) = B(z)u(z) (29) When the last equation is converted to the time domain, the unknown parameters for A(z) and B(z) have to be defined. These parameters are determined using the measured input and output quantities. If these parameters are known, the values of the pulse response g(t) for a certain frequency can easily be calculated. Once we get the pulse response, the quasi-static heat capacity is calculated from this pulse response. (TOPEM The new advanced multi-frequency TMDSC technique) Where: m: sample mass (mg) : quasi-static heat capacity (30) g(t): pulse response t: time (s) 15

22 2 MATERIALS AND METHODS 2.1 MATERIALS TABLE 3.1. LIST OF ALL USED CHEMICALS Products Company City Country Griseofulvin (USP micronized) Hawkins Pharmaceutical Group Minneapolis USA Mannitol (Pearlitol 60) Roquette Lestrem France Indium Mettler-Toledo Gießen Germany PET BASF Ludwigshafen Germany Sucrose Pfeifer & Langen Cologne Germany Glucose Roquette Lestrem Germany Dextrose (monohydrate ST) Roquette Lestrem France Lactose (granulac 200) Meggleburg Wasserburg Germany Sodium nitrate Grüssing Filsum Germany PVP CL BASF Ludwigshafen Germany PVP CL-M BASF Ludwigshafen Germany PVP 17PF BASF Ludwigshafen Germany PVP 25 BASF Ludwigshafen Germany PVP 90F BASF Ludwigshafen Germany PVP IR BASF Ludwigshafen Germany PEG PVA copolymer VA64 BASF Ludwigshafen Germany 16

23 2.2 METHODS DIFFERENTIAL SCANNING CALORIMETRY All experiments are done with the DSC1 from Mettler-Toledo (Gießen, Germany) equipped with sample robot and a Haake Intra-Cooler EK/MT. Sample parameters and data-analysis are realized with the STAR e software. This software also determines if the DSC is used as a conventional DSC, ADSC or as TOPEM. All experiments were done using Aluminium pans without pin of 40μl (Mettler-Toledo, Gießen, Germany). Al of the pans had pierced lids unless told else. All measurements are done under N 2 with a flow of 50ml/min. The used method will be shown in the diagrams, when provided or else in the text. The used abbreviations are HR for heating rate, PH for pulse height and p for period BALANCE All samples are weighed on the balance MC210P from Sartorius AG (Goethingen, Germany). First, sample pan holder, pan and lid are set to zero before weighing the sample. After sealing the pan with the lid, the sample is weighed again and this value is used α-mannitol In order to make α-mannitol, ß-mannitol is heated until it is completely molten. Then it s poured into a mortar. When it s completely crystallized, it is pulverized into powder and stored at room temperature WATER UPTAKE OF SUGARS For some measurements, moist substances were needed. Glucose, fructose, dextrose and sucrose were heated until completely molten and poured into a petri dish. The glassy substances were equilibrated at 23 C at 60% relative humidity for at least 10 hours. 17

24 3 OBJECTIVES In thermal analysis, differential scanning calorimetry (DSC) is a commonly used technique. Later, temperature modulated DSC techniques made their entrance, namely isothermal step DSC and alternating DSC. Recently, TOPEM, a new temperature modulated DSC technique has been developed. However, not much research has been done on this technique. In addition, most available information is linked to the company which developed TOPEM. The goal of this study is to make a critical review about this new technique. TOPEM will be compared to regular DSC and ADSC in order to evaluate its possible advantages and drawbacks. The main question is: does TOPEM has any added value for differential scanning calorimetry? In the literature, some advantages and disadvantages can be found (see also introduction p.13) and based on the described advantages and disadvantages, the following points should be investigated: - separation of reversing and non-reversing heat flow: measurements of several examples - calculation window: determination of a width of calculation window and the influence of the pulse width on it - heat capacity: comparison of several methods to calculate heat capacity and evaluation of the shift in complex heat capacity - frequency: comparison with ADSC, determination of its limits - resolution 18

25 4 RESULTS 4.1 PRELIMINARY TESTS To start with, there was a search for a substance to do the investigations on TOPEM with, because no standard substances are described in the literature. As a first substance, Indium was measured because it is commonly used as a standard for regular DSC. The melting peak with high intensity was found at the same temperature as DSC. No conclusions could be drawn from the reversing and nonreversing heat flow curves because of the big intensity of the melting over small temperature range. In addition, pure substances such as Indium cannot follow a temperature modulation. (Schawe 2005b) Next step was to take substances described in the literature and do measurements with the same parameters. Own measurements were compared with measurements described in literature. The goal was to work on with a certain substance when comparable results were found. A lactose-water mixture containing 4,5% lactose was measured as described by Schawe (2005a). The article shows a peak at -30 C of which the origin is not clear. There was no transition found, even after several measurements. However, the described peak in the literature is small. It is possible that the DSC used by Schawe (2005a) is able to detect smaller effects because it is more sensitive. Afterwards, polyethelene terephtalate was measured as described by Schawe (2005c). Pet was crystallized for 10 min at 170 C. Compared to the article, smaller curves are obtained and there is noise even after repeating the measurement. The reason for this is that polymers often have a batch-to-batch variability. Therefore, the next measured substance was a crystalline substance, namely sodium nitrate as described by Schawe (2005a). The measurement showed comparable curves to the literature with a step in the curves at the same temperature (see figure 4.1.). From these tests was concluded to do no measuring of melting of pure substances. Pure substances can be used when effects overly or are close to each other. Because polymers often have differences between batches, it is better to take crystalline substances. Also mixtures will be used for investigation. 19

26 A B FIGURE 4.1. A. TOTAL (UPPER), NON-REVERSING (MIDDLE) AND REVERSING (LOWER) HEAT FLOW CURVES OF A TOPEM MEASUREMENT OF SODIUM NITRATE (SCHAWE, 2005C) B. TOTAL (BLACK), REVERSING (GREEN) AND NON-REVERSING (BLUE) HEAT FLOW CURVE OF A TOPEM MEASUREMENT OF SODIUM NITRATE 4.2 REVERSING AND NON-REVERSING HEAT FLOW IN GENERAL Separation of reversing and non-reversing heat flow is the main goal of using temperature modulated DSC. To evaluate this, first griseofulvin was measured because it is expected that TOPEM can give more information about the results obtained from regular DSC. Also a 40:60 mixture of sucrose and water was measured and compared to the literature. There were also several moist sugar samples measured. Moist substances give in regular DSC only a peak between 30 C and 120 C because of the evaporation of water. This peak overlies glass transitions occurring in this range. TOPEM should be able to detect this glass transition GRISEOFULVIN The conventional DSC curve of griseofulvin shows a small peak before the melting peak which could be due to the sublimation of griseofulvin. The clue for that is the description of a weight loss due to sublimation at about 200 C, measured with thermogravimetry. (Analytical profiles of drug substances and excipient) 20

27 Since sublimation is a process related to latent heat flow, it should be seen in the non-reversing heat flow curve of a TOPEM measurement. As shown in figure 4.2.B the total heat flow curve also shows this second peak but the non-reversing curve doesn t. The reason for this could be that the gas phase escaped through the hole in the pierced lid of the pan and therefore no effect can be measured. Another measurement was done with a closed pan. As can be seen in figure 4.2.C the curves are comparable to the curves of the measurement with a pierced pan. This means that no extra information was gained using TOPEM. The peak at about 18 C is due to melting. Griseofulvin is a pure substance so its melting cannot be evaluated using TOPEM. Griseofulvin was investigated for the small peak under the first part of melting peak. It was not used for the investigation of the melting peak. A B C FIGURE 4.2. TOTAL HEAT FLOW CURVE OF A DSC MEASUREMENT OF GRISEOFULVIN (A) AND TOTAL (BLACK), REVERSING (GREEN) AND NON-REVERSING (BLUE) HEAT FLOW CURVES OF A TOPEM MEASUREMENT OF GRISEOFULVIN WITH PIERCED (B) AND SEALED (C) AL PAN 21

28 4.2.3 SUCROSE-WATER MIXTURE A TOPEM measurement of a mixture of 40% sucrose and 60% water was measured as described by Schawe (2005c). The total heat flow curves are comparable to our measurement. However, the curve for the non-reversing heat flow is not entirely published. In our measurement, a peak in the non-reversing heat flow curve at about -5 C can clearly be seen. The little peak around -35 C is explained in the article to be the melting of non-equilibrated crystals. The peak at -5 C is explained by Schawe (2005c) to be an artifact. It is interesting that the peak at -5 C, which is more than ten times larger than the peak at -35 C is not reported in the article. The peak at -5 is also present at all the measurements which were taken. Therefore we assume that it is not an artifact but that the peak is due to crystallization. FIGURE 4.3. TOTAL (BLACK), REVERSING (GREEN) AND NON-REVERSING (BLUE) HEAT FLOW CURVES OF A TOPEM MEASUREMENT OF A 40:60 SUCROSE-WATER MIXTURE 22

29 A B FIGURE 4.4. A. TOTAL HEAT FLOW (BLACK) AND MEASURED HEAT FLOW (RED) OF A TOPEM MEASUREMENT OF A 40:60 SUCROSE-WATER MIXTURE (SCHAWE, 2005C) B. TOTAL (BLACK), REVERSING (GREEN) AND NON-REVERSING (BLUE) HEAT FLOW CURVES OF A TOPEM MEASUREMENT OF A 40:60 SUCROSE-WATER MIXTURE (SCHAWE, 2005C) MOIST SUBSTANCES Moist substances show in regular DSC a peak between 30 C and 120 C because of the evaporation of water. When some underlying effects occur at the same range, such as a glass transition, this cannot be seen in the heat flow curve. TOPEM however can separate the latent and sensible heat flow. The water peak will still be seen in the total heat, but as a glass transition is a reversible process, it should be seen in the reversing heat flow. To examine the effects underlying the evaporation of water, several samples of polyvinylpyrolidone (PVP) were evaluated. All the selected samples showed a large water peak in measurements obtained by regular DSC in the past. All the PVP samples were measured between 0 C and 140 C with a heating rate of 1 C/min, a pulse height of 0.5 C and had a sample weight of about 3.00 mg. There was no underlying effect found for none of the PVP samples. Only the reversing heat flow curve of the TOPEM measurement of PVP VA64 does show a 23

30 glass transition at about the glass transition temperature of the DSC measurement. At that point however, it is not overwhelmed by a peak of evaporation of water. This means that for all the PVP samples, TOPEM cannot separate between reversing and non-reversing heat flow. The next measured substances were glucose, fructose, dextrose and sucrose. All have a defined glass transition temperature and take up water fast and easily. All samples were measured from 50 C below to 50 C above the glass transition temperature, with a heating rate of 1 C/min, pulse height of 0.5 C and a sample mass between 2.00 and 6.00 mg. In all curves, the water peak is visible, but no underlying effects were detected. For all the substances there is an extra peak at about the middle of the peak due to the desorption of water. It occurs at about 70 C for dextrose and 90 C for glucose (see figure 4.6.). Desorption means that, in this case, water molecules leave the surface to join the gas phase. It can be concluded that also for the sugars, TOPEM cannot separate between reversing and nonreversing heat flow. FIGURE 4.5. TOTAL (BLACK), REVERSING (GREEN) AND NON-REVERSING (BLUE) HEAT FLOW CURVES OF A TOPEM MEASUREMENT OF PVP VA64 24

31 FIGURE 4.6. TOTAL (BLACK), REVERSING (GREEN) AND NON-REVERSING (BLUE) HEAT FLOW CURVES OF A TOPEM MEASUREMENT OF DEXTROSE (UPPER) AND GLUCOSE (LOWER) 4.3 CALCULATION WINDOW IN GENERAL For the evaluation of the data, several parameters have to be defined. The width of calculation window, sample and pan weight, parameters for fitting the points in a curve and the width of smoothing window have to be determined. The calculation window is a set of data points which are considered for the calculation of the heat flow curves out of the raw data. The chosen width is a time interval which is considered for calculation for each data point. The shift of calculation window is how far the calculation window moves each time. The width of the calculation window is recommended by Mettler-Toledo to be one third of the transition whereas it is said to be about one tenth by Fraga (2007). This window width and its effect on the eventual curves will be investigated. When a narrow calculation window is used, the calculation goes faster, but less information is obtained. Using bigger calculation windows more information is obtained, but the calculation requires more time. Large window widths need low heating rates. A small shift in calculation window gives a good analysis but takes a long time. 25

32 FIGURE 4.7. WIDTH AND SHIFT OF CALCULATION WINDOW (METTLER-TOLEDO 2008) The sample and pan weight can be filled in to quantify measurements or to define a heat capacity. These weights are without sample and are a blank. The sample and instrument response parameters are the degrees of the polynomials used to fit the data points in a curve. After the curves are calculated with the calculation window, they still have to be smoothened. Therefore the width of smoothing window has to be set. It has to be smaller than the calculation window because it smoothens within this window MANNITOL For the investigation of the calculation window, a 50:50 mixture of α-ß mannitol was measured. Mannitol can crystallize in several forms of which two polymorphic forms are widely described in the literature: α-mannitol and ß-mannitol. ß-mannitol is the most stable one and the metastable α-mannitol transforms to the ß-form over time. The orthorombic α and ß-forms have many similarities. Their physical properties such as melting temperature, enthalpy of fusion, density and specific heat, only differ little. They can be separated using X-ray diffraction and infrared spectroscopy. It s not possible to differentiate them with conventional DSC measurements because their melting points are too close to each other: α-mannitol melts at 166 C and ß-mannitol at C. (Burger et al., 1999) 26

33 When a mixture of α and ß-mannitol is being melted, the metastable α-form will melt first. Because of the presence of ß-mannitol, the molten α-mannitol will recrystallize into ß-mannitol. This is a non-reversible effect because ß-mannitol will never crystallize into the less stable α-form when the more stable ß-form is present. This means that this effect should be seen in the non-reversing heat flow CALCULATION WINDOW PARAMETERS During the evaluation of mannitol, it was found that the width of the calculation window influenced the curves. The correct width of calculation window is the one which gives a good fit between the total and the measured heat flow. That is always the case when the total heat flow is calculated from the measured heat flow, as in ADSC for example. But in TOPEM, it is the reversing and non-reversing heat flows which are calculated from the measured heat flow and the total heat flow is the sum of both. Afterwards, it has to be checked if this total heat flow and the measured heat flow do fit. If not, no conclusions can be drawn. To make these curves fit, the evaluation parameters have to be set. With every new calculation window a new curve is obtained. The window that gives the best fitted total heat flow curve should be taken for interpretation of the data. To see if there is a good fit, the measured heat flow and the total heat flow were subtracted. The obtained curve should not have large fluctuations. These curves were compared for several calculation windows for several mixtures of α-ß mannitol. To analyze the fluctuations in the subtraction curves, the amplitudes at each point are summed up. If there are large fluctuations, the sum of the amplitudes is high. The lowest sum of amplitudes accords to the best fit for total and measured heat flow and the corresponding width of calculation window is the best to analyze the sample. 27

34 FIGURE 4.8. TOTAL HEAT FLOW (BLACK), MEASURED HEAT FLOW (RED) AND SUBTRACTION OF BOTH (PURPLE) OF A TOPEM MEASUREMENT OF Α-ß MANNITOL; CALCULATION WINDOW IS 120S (UPPER), 300S (MIDDLE) AND 550S (LOWER) The calculated window should correspond to the recommended calculation window, namely one third of the transition interval. The transition interval for every mixture is divided by three to obtain the recommended width of the calculation window. As can be seen in table 4.1. the recommended and optimal calculation windows are different. Only for the 80:20 mixture, one third of the transition interval is close to the optimal width of the calculation window. For all the others, the window should be bigger. In figure 4.9. can be seen that for all the mixtures, except the 60:40 mixture, several calculation window widths will yield acceptable results, because their sums of amplitudes are in the same range. From this data was concluded that it is important to determine the correct calculation window before any conclusions are drawn. However, it is difficult to determine the optimal calculation window width since similar results were obtained. When the average of these was made, the calculation window width should be around one fifth of the transition interval, but this should be evaluated for each sample separately. 28

35 A B C D E FIGURE 4.9. SUM OF THE AMPLITUDES OF THE SUBTRACTION CURVE BETWEEN THE TOTAL HEAT FLOW AND MEASURED HEAT FLOW FOR SEVERAL α-ß MANNITOL MIXTURES: 20:80 (A), 40:60 (B), 50:50 (C), 60:40 (D) AND 80:20 (E) α-ß MANNITOL; THE LOWEST SUM IS MARKED (RED). 29

36 TABLE 4.1. COMPARISON OF CALCULATED BEST AND RECOMMENDED CALCULATION WINDOW (1/3 OF TRANSITION INTERVAL) Mixture 1/3 of Calculated best Part of Transition transition calculation transition interval (s) interval (s) window (s) interval 20: /5 40: /6 50: /5 60: /4 80: / EFFECT OF THE PULSE WIDTH ON THE CALCULATION WINDOW As already pointed out, the width of the calculation window has a big influence on the outcome of the final heat flow curves. The goal was to find a defined width of calculation window. The problem is that TOPEM uses a stochastically changing pulse width. In addition, TOPEM does not provide the raw data, there has already been a mathematical modulation executed. In order to overcome the problem with the pulse width and the data, a TOPEM simulation was carried out with regular DSC. The modulation was done with a periodic step function of 60s. The measured substance was a 50:50 mixture of α and ß-mannitol. The following parameters were used: heating rate 0.05 /min, pulse height 0.02 C, period 60s and sample mass between 2.00 and 5.50 mg. We are especially interested in this substance because the ADSC measurement shows no reversing effect while the TOPEM measurement does show a peak in the reversing heat flow. This is shown in figure Figure shows the raw data of the sample, an empty pan and the reference pan. Not the heat flow but the temperature is taken for analysis because the temperature is the only true measured value and the heat flow is correlated to it. 30

37 A B FIGURE MEASURED (RED), TOTAL (BLACK), REVERSING (GREEN) AND NON- REVERSING (BLUE) HEAT FLOW CURVE OF A TOPEM MEASUREMENT (A) AND AN ADSC MEASUREMENT (B) OF A 50:50 MIXTURE OF α-ß MANNITOL FIGURE RAW DATA OF THE SAMPLE, EMPTY PAN AND REFERENCE PAN OF A PERIODIC TMDSC MEASUREMENT. CURVES ARE A FUNCTION OF TIME AND TEMPERATURE. 31

38 The three curves are still a function of time and of temperature. The aim is to eliminate the factor temperature so that the slope of the curve becomes zero. In order to do that, the average of the pan curve is calculated and subtracted of all the curves. This average is actually the heating rate of the measurement. After this is done, the peak of melting can be seen in the sample curve (pink curve in figure 4.12.). This curve is considered for further analysis of the calculation window. Since this is a periodic modulation, the period should be used as a calculation window width. For comparison, two other widths are considered: 200s and 330s. As transition interval the segment between 2600 and 3600 seconds is considered. The width of 200 corresponds to about one fifth of the transition interval, which was found to be the correct width. The width of 330s corresponds to about one third of the transition interval, which is the recommended calculation window width. As can be seen in figure the width of the period gives the best fit. The width of 200s and 330s are not flat anymore. FIGURE RAW DATA OF THE SAMPLE OF A PERIODIC TMDSC MEASUREMENT. AVERAGE CURVES OF THE SAMPLE ARE GIVEN FOR A PERIOD, 200S AND 330S. CURVES ARE ONLY A FUNCTION OF TIME. 32

39 To take a closer look at how the sample responses to the set value and in order to compare each segment to another, the melting peak had to be eliminated. For further calculations, figure is considered. The sample curve (pink) and its averages were subtracted for all three average curves. This means that a subtraction was done for the width of the period (pink minus blue curve), the width of 200s (pink minus green curve) and for the width of 330s (pink minus brown curve). These subtraction curves were compared to the reference. The resulting curves are shown in figure The same colors are used for the same widths of calculation window. In figure can again be seen that different curves are created for the different calculation windows and that the period gives the best width. To compare the three window widths better, every period was overlapped to see any changes in the curves. This means that all the curves in figure 4.13 were cut every 60s and all these pieced were laid over each other. The result is shown in figure Only the first half of period is shown, but again the curves are different for the three window widths. For the width of 330s, the curves even go below zero. The sample is able to follow the oven temperature quickly. FIGURE BEHAVIOR OF THE SAMPLE; COMPARISON OF DIFFERENT CALCULATION WINDOWS 33

40 FIGURE BEHAVIOUR OF THE SAMPLE; COMPARISON OF DIFFERENT CALCULATION WINDOWS; BLUE CURVES ARE BEFORE THE TRANSITION, PINK ARE DURING AND YELLOW ARE AFTER In order to understand the shape of the curves in figures and 4.14., the behavior of the sample should be investigated more. This is explained in figure There are two effects which have to be considered: the effect of melting (green) and the effect of the diffusion of the heat through the sample (brown). Both curves result in the sample curve (blue). In the mathematical evaluation of the data (see introduction under ), these curves are part of the equation 27. The equation y(t) = g(t)u(t) was rewritten in the z-plane as y(z) = H(z)u(z) in which H(z) can be described as a rational function. y(t) = g(t) u(t) Where: y(t): output signal g(t): pulse response of the system u(t): input signal t: time (s) Where B(z) and A(z) are polynomials of degree q in the variable z. 34

41 The last equation holds the functions which describe the melting part and the heat diffusion part of the sample. B(z) is the melting part and A(z) is the heat diffusion of the sample. Using these equations, in the time domain, y(t) becomes the following: y(t) = locally y(t) = u(t) in general The output value or sample curve y(t), the input value or set value u(t) and the heat diffusion A(t) are known variables, so the melting B(t) can be determined. FIGURE BEHAVIOR OF THE SAMPLE, THE EFFECT OF MELTING AND EFFECT OF HEAT DIFFUSION A last step in comparison of the different calculation windows was to evaluate the melting behavior. For the three different sample curves (blue, green and brown curve) as shown in figure the average is calculated. Since there was no reversing effect found in the ADSC measurement, it is also expected not to see a reversing effect in this measurement. This means that the calculated average curve 35

42 should be flat. In figure can be seen that this is almost the case for the periodcurve, but the bigger the calculation window, the more the curves deviate. FIGURE AVERAGE OF THE SAMPLE CURVE FOR DIFFERENT CALCULATION WINDOWS It was already clear that several widths of calculation window give different heat flow curves. The curves shown in figure 4.13, and also prove this. In addition, several ways proved that the period is the best width for the calculation window. The problem with TOPEM is that, because of the stochastic pulses, there is no fixed period. It is because of this that the calculation of the heat flow curves becomes difficult. Therefore we would recommend using TOPEM in a periodic way to overcome the problem with the calculation window. This means however that no frequency evaluation can be done since a periodic modulation holds only one frequency. The difference with ADSC is that TOPEM uses a step function instead of a sine wave as modulation. A sample cannot follow a sine wave well because there is no time for equilibration. Because of the many fluctuations, a blank and calibration curve need to be executed for every new method in ADSC. Using a step function as in TOPEM, the sample has the time to equilibrate every step upwards and downwards. That s why there are no blank and calibration curves needed for TOPEM as there is for ADSC. When TOPEM is used in a periodic way, more measurements 36