User Com. Thermal Analysis. Choosing the right baseline. Information for Users. Dear Customer, Contents 1/2007. Introduction

Transcription

1 Thermal Analysis Information for Users 25 User Com Dear Customer, We are very pleased to receive more and more articles from you for publication in UserCom. Thanks to new techniques and better performance, thermal analysis is being used in an ever-increasing number of scientific fields. Hyphenated techniques such as evolved gas analysis, microscopy and chemiluminescence yield much more information about samples and very often greatly simplify the interpretation of measurement results. We think this issue of UserCom will once again give you ideas for applications in new and interesting areas using the multitude of techniques now available. Choosing the right baseline Dr. Rudolf Riesen Choosing the correct baseline is crucial for the determination of the enthalpy of a transition or a reaction. The baseline represents the DSC curve that would be measured if no transition or reaction occurred. The examples described in this article illustrate how to choose the right baseline for a particular evaluation. Contents 1/2007 TA Tip - Choosing the right baseline 1 Applications - Determination of the Noack evaporation loss of lubricants by TGA 7 - The characterization of polymorphs by thermal analysis 9 - Analysis of melting processes using TOPEM 13 - Characterization of delivery systems by thermogravimetry 18 Tips and hints - Detection and evaluation of weak sample effects in DSC 21 Dates - Exhibitions 23 - Courses and Seminars 23 Introduction In thermal analysis, baselines are mostly used in connection with the integration of peaks. The peak area is determined by integrating the area between the measurement curve and a virtual or true baseline. In the same way, the peak temperature is defined as the point on the curve where the distance to the baseline is greatest. Extrapolated baselines are important for the determination of glass transition temperatures

2 TA Tip Figure 1. Drawing interpolated DSC baselines (the endothermic direction is upward). and the onset temperatures of effects. In the literature and in standards, the term baseline is sometimes defined differently, or different terms are used for the same thing. The terms most frequently encountered have therefore been summarized together with some brief comments. A number of application examples are then discussed to illustrate the rules governing the choice of baselines and that show which type of baseline should be used for the optimum evaluation of a particular DSC curve. Terminology The terms used in thermal analysis are summarized and explained in various standards. However, since the definitions are not always the same, the terms used have been summarized below for the discussion of baselines that follows. Further definitions can be found in the book by Höhne [1] as well as in the standards mentioned (ISO [2], DIN [3], ASTM [4, 5]). The preferred terms are highlighted, but other terms are also included. Blank, blank curve, zero line [3], instrument baseline [2]: A thermal analysis curve measured under the same conditions as the sample but without the sample; the mass of the crucibles used must be the same. Blank curves are essential for specific heat capacity determinations. Comment: In some cases, the zero line [1] is also understood as a curve measured without the sample or crucibles. Sample blank: A curve that is obtained from a fully converted sample. This is usually the second heating run of the same sample under the same conditions. The effect measured in the first heating run no longer appears. Baseline (also sample baseline [2]): Part of the curve that does not exhibit any transitions or reactions. This is an isothermal baseline if the temperature is held constant. A dynamic baseline is obtained when the temperature is changed through heating or cooling. The baseline depends on the heat capacity of the sample (with an empty reference crucible) and the blank curve. Comment: In practice, the term is also used to mean the virtual baseline used for integration. Virtual baseline [2]: An imaginary line in the region of a reaction or transition that the DSC curve would show if no reaction or transition enthalpy were produced. Interpolated baseline [1]: This is a line that joins the measured curve before and after the peak. Extrapolated baseline: This is a line that extends the measured curve before or after the thermal effect. The types of virtual baselines normally used are explained in the applications. True baseline: In the region of the transition or reaction, the baseline can be calculated according to physical data or even measured. Factors influencing the baseline The influence of measurement conditions on the DSC curve and the baseline should always be taken into account when interpreting curves and evaluating numerical data. Furthermore, the course of the blank curve and its reproducibility should be known. Possible important parameters that can change during a transition are [1]: 1. Mass, shape and structure of the sample, e.g. powder or film; 2. Thermal conductivity and contact of the sample with the bottom of the crucible, e.g. a powder liquefies during melting; 3. Heat transfer from the crucible to the sensor, e.g. deformation of the crucible due to an increase in the internal pressure or through products escaping from the crucible; 4. Heating rate, e.g. when it changes from dynamic to isothermal; 5. Thermal history of the sample and measuring system. If it is difficult to choose the baseline, it often helps to examine the sample and crucible after measurement with regard to the above points. Principles for constructing virtual baselines The basic principle for constructing a virtual baseline can be summarized as follows: The interpolated baseline for the determination of the transition enthalpy or the reaction enthalpy leaves the DSC curve tangentially before the thermal effect and joins the curve again tangentially after the effect. A good example to illustrate this is the take-off and landing of an aircraft. In special cases there are some exceptions to this that will be described in the examples. Figure 1 shows how these principles are applied. a) 1 nonsensical; 2 good (Line), b) 1 unsatisfactory (horizontal straight line); 2 good (integral baseline, possibly Spline), 2

3 c) good (Integral tangential, or possibly Spline), d) melting with exothermic decomposition, 1 good (straight line to the point of intersection with the DSC curve); 2 rather arbitrary because the DSC curve is the sum of simultaneously occurring processes, e) two overlapping peaks, e.g. eutectic and melting peak of the main component, 1 good for the total integral, 2 good for the integration of the first peak (peak interpreted as sitting on the main peak, Spline). The transition line from one tangent to another can have different shapes and be displayed as a straight line or as a sigmoidal curve (S-shaped function). The type of interpolated baseline chosen depends mostly on the physical conditions or chemical changes involved, for example: The specific heat capacity of the sample, c p, hardly changes during the transition or it changes linearly with temperature. The transition is accompanied by a significant change in the heat capacity The heat transfer to the sample changes during the transition. The mass of the sample changes during the transition. The STAR e software provides several different types of baseline to accommodate the changes shown by the DSC curve during a transition. Table 1 describes the baselines and their typical applications. The extrapolated virtual baselines are the tangents to the measured curve at the evaluation limits, just as they are used for interpolation with the baselines. Typi- Baseline type Description Typical DSC application Line Tangential left Tangential right Horizontal left Horizontal right Spline Integral tangential Integral horizontal Zero line Polygon line This is a straight line that joins two evaluation limits on the measured curve. This is the extension of the tangent to the measured curve at the left evaluation limit. This is the extension of the tangent to the measured curve at the right evaluation limit. This is the horizontal line through the point of intersection of the measured curve with the left limit. This is the horizontal line through the point of intersection of the measured curve with the right limit. The Spline baseline is the curve obtained using a flexible ruler to manually interpolate between two given points (known as a Bezier curve). It is determined as a 2 nd order polynomial through the tangents at the evaluation limits. This bow-shaped or S-shaped baseline is based on the tangents left and right. Starting with a trial baseline, the integral baseline is calculated using an iterative process. The conversion calculated from the integration between the evaluation limits on the measured curve is normalized. Like the Spline curve, this bow-shaped or S-shaped baseline is based on the tangents left and right. This baseline is calculated using an iterative process like the Integral tangential baseline. This S-shaped baseline always begins and ends horizontally. This is the horizontal line that intersects the ordinate at the zero point. It requires a blank curve subtraction. The baseline can be determined through a curved line or a straight line from individually chosen points. The polygon line is then first subtracted from the measured curve and the resulting peak integrated using a straight baseline. Reactions, without abrupt c p changes, that exhibit a constant c p increase or a constant c p. This baseline is the default setting. Integration of a melting peak on a measured curve with subsequent decomposition of a substance. Melting of semicrystalline plastics with significant c p temperature function below the melting range. Peak integration when substances decompose. Isothermal reactions, DSC purity determination. With overlapping effects. Samples with different c p temperature functions before and after the effect. The Line baseline would possibly cross the DSC curve and lead to large integration errors depending on the limits chosen. Samples whose heat capacity changes markedly, e.g. through vaporization and decomposition. The Line baseline would possibly cross the DSC curve and lead to large integration errors depending on the limits chosen. Determination of transition enthalpies including sensible heat. In special cases. Table 1. List of virtual baseline types for integration. 3

4 TA Tip Figure 2. Examples of frequently used types of baseline. cal applications of extrapolated baselines are for the determination of the: glass transition temperature extrapolated onset temperature (also as first deviation from the measured curve) step height. In all evaluations in which extrapolated tangents are used, one must make sure that artifacts on the measured curve or signal noise do not interfere with the determination of the slope of the tangent. This would result in the tangents being in the wrong place. Application examples The different types of baseline presented in Figure 1 are illustrated in the follow- ing practical examples. Figure 2 shows the most frequently used virtual baselines: a) Spline: This is used to determine the reaction enthalpy of a postcuring reaction that is overlapped by the beginning of decomposition. b) Horizontal right: Isothermal curing of an epoxy resin at 140 C. When the reaction has finished the DSC curve is horizontal. The baseline can be drawn horizontally through the last measured points. c) Integral horizontal: The DSC curve of mg water, which on heating evaporates through a 50-µm hole in the crucible lid. The loss of mass causes a change in the sample heat capacity, which is reduced proportionally to the amount vaporized. At the end of the measurement the crucible is empty and the DSC signal is practically 0 mw. d) Line: The DSC curve shows a glass transition of the amorphous part of the polyethylene terephthalate (PET) followed by cold crystallization and melting of the crystallites. The straight baseline is the virtual extension of the DSC curve after the glass transition to the curve after the melting and shows the trend of the curve without crystallization and melting. The integral of the two effects yields 22.8 J/g as the difference between the exothermic and endothermic processes. This means that crystallites were already present at the beginning of the measurement. In relation to the melting enthalpy of 100% crystalline PET, this shows that the degree of crystallization of the sample was initially about 16% and was therefore not fully amorphous. Figure 3. Example for the change of the heat capacity during the transition ice/water. Figure 3 discusses how to draw the interpolated baseline if the baselines before and after the peak are at different levels, for example because the specific heat capacities of ice (2.1 J/gK) and water (4.2 J/gK) are very different. The figure shows four identical curves of the same part of the DSC melting peak of 1.87 mg water measured at 5 K/min (melting enthalpy 333 J/g). Each curve has a different baseline type. 1) The Horizontal left baseline does not take the change in heat capacity in the evaluation range into account and yields a peak area that is too large. 2) The Line baseline is clearly unfavorable and contradicts the basic principles (no tangents, it crosses the DSC curve). In reality, the change in heat capacity is of course not linear with temperature between the evaluation limits as the dotted line wrongly shows. 3) The Spline baseline is somewhat better, but also crosses the DSC curve. 4) In this case, the Integral horizontal baseline is optimal. It draws the baseline proportional to the peak area from the level before the peak to the level after the peak and so takes into account the change in heat capacity. 4

5 In the first three cases, the result of the integration can be improved by choosing better limits, but even so, the virtual baselines do not correspond to the physical facts. Overlapping thermal effects are usually the most difficult with regard to choosing a realistic virtual baseline. Figure 4 shows how a second heating run of the reacted sample helps to locate the exact position of the baseline. An epoxy resin was partially cured at 100 C for 80 min, causing the material to vitrify [6]. The DSC curves shown in Figure 4 were then measured at 5 K/min. The postcuring reaction begins at the glass transition (curve 1). Curve 2 shows the DSC second measurement of the same, fully cured, sample. The straight line 3 (dotted) describes the course of the DSC curve after complete curing above the glass transition. It therefore represents the baseline for the integration and serves as a tangent for the determination of the glass transition temperature. It can be assumed that the behavior of the heat capacity above the glass transition during postcuring is about the same as that of the fully cured sample. Figure 4. Example showing a special baseline (Polygon using just two points, X) to determine the postcuring reaction. The Zero line baseline was used to integrate the peaks in curve 4. Figure 5. Curve interpretation and the choice of integration limits and baselines. The postcuring enthalpy is determined as follows: The dotted line (curve 3) is subtracted from curve 1 yielding curve 4. The peak in this curve is integrated using the Zero line baseline type within the limits shown. Separation of the partially overlapping peaks could also be achieved using temperature-modulated DSC. The purpose of the example in Figure 5 is to show how important correct interpretation of the DSC curve is. The choice of the integration limits and the baseline type should be good enough to obtain results that provide consistent information for further investigations. Figure 5 shows the DSC curve of a 40% solution of sucrose in water measured at 5 K/min after slow cooling. The glass transition occurs at about 45 C and the ice that had crystallized out melts in the sucrose solution in the range 37 C to 0 C. Integration from 29 C (dotted line) would assume that the specific heat capacity decreases, which is not the case here. The Line and Spline baselines yield enthalpy values that are 5% too low. Only the Integral tangential baseline from 37 C gives the correct value that can be used for a consistent quantitative evaluation. Although one thinks mostly about interpolated baselines, extrapolated baselines are in fact just as important. This is shown in the examples in Figure 6: a) Oxidation induction time (OIT) of a mineral oil, measured at 180 C at an oxygen pressure of 3.5 MPa. b) Melting point of benzoic acid determined as the extrapolated onset. Since jumps in heat capacity rarely occur, a virtual baseline should be conc) Glass transition temperature of polystyrene determined as the midpoint according to how the tangents are constructed (depends on the particular standard). d) At the glass transition, the specific heat capacity, c p, increases leading to a step in the c p curve. The step height is characteristic of the amorphous content of the sample. Conclusions Whenever possible, physical changes must be taken into account when choosing the optimum baseline for an integration or onset determination. 5

6 TA Tip Figure 6. Examples of extrapolated baselines. structed that is smooth and free from any irregularities or discontinuities. The correct choice of baseline assumes that the curve has been properly and consistently interpreted [7]. Furthermore, the integration limits must be carefully chosen depending on the information required. The rules and types of baseline discussed here using DSC measurements as examples can be applied to other TA measurement techniques, e.g. for the integration of peaks from SDTA, DTG analyses and other mathematically derived measurement curves. Literature [1] G. W. H. Höhne, W. Hemminger and H.-J. Flammersheim: Differential Scanning Calorimetry, Springer Verlag, Chapter The DSC Curve [2] ISO (1997) Plastics DSC. General principles [3] DIN Thermal analysis; Terms [4] ASTM E473 Standard Terminology Relating to Thermal Analysis and Rheology [5] ASTM E 2161 Standard Terminology Relating to Performance Validation in Thermal Analysis [6] J. Schawe, METTLER TOLEDO UserCom 14, 17 [7] METTLER TOLEDO UserCom with articles on curve interpretation: DSC, UserComs 11 and 12; TGA, UserCom 13 6

7 Determination of the Noack evaporation loss of lubricants by TGA Applications Dr. Rudolf Riesen Abstract For quality and environmental reasons, lubricants for engines and other applications must only exhibit a low evaporation rate. The loss of volatile components from an oil increases its viscosity and leads to increased oil consumption, coking and wear. The Noack method is a widely used standard test method for measuring the evaporation loss from lubricating oils. According to the ILSAC GF-3 and API-SL specifications the evaporation loss must not be greater than 15%. The ASTM standard test method D6375 for the determination of the evaporation loss of lubricating oils by the Noack method [1] uses thermogravimetric analysis, TGA. This method yields the same results as other standard test methods (e.g. ASTM D5800 [2], DIN [3], JPI-5S [4]). This article describes how the Noack evaporation loss is determined in comparison to a reference oil sample using TGA. Introduction The increase in the usable lifetime of lubricants coupled with faster oil circulation rates, longer oil change intervals and lower lubricant consumption means that lubricants are subjected to greater stress. Higher temperatures coupled with smaller oil volumes and higher performance lead to a constant increase in the demands placed on the performance and quality of the lubricants. To ensure that the lubricants are properly used, they must be properly specified and classified. The specifications describe the physical properties of engine oils such as the viscosity, evaporation loss and shear stability. Performance behavior is also tested in engine tests. This includes wear protection and cleanliness as well as the influence on fuel consumption and the changes in the engine oil during operation due to viscosity changes (thickening). The classification is provided by organizations such as ILSAC, API or SAE (see the table of acronyms). One of the commonly used specifications is the evaporation loss. The low molecular mass constituents of an engine oil, which consists of fractions of different hydrocarbons with different chain lengths and molecular masses, can evaporate under increased thermal stress. This usually leads to an increase in the viscosity of the lubricant. At the same time, the solubility of the additives in the base oil is affected. The evaporation is important for all lubricant groups (e.g. also for synthetic oils) if they are used at higher temperatures. For example with engine oils, evaporation losses can occur through high temperatures at the piston rings and elsewhere. These losses lead to undesirable oil thickening and increased oil consumption. The Noack evaporation loss test according to ASTM D6375 The Noack test to quantitatively determine the evaporation loss of oils under standard conditions was introduced many years ago. For example, the DIN [3] test method measures the evaporation loss over a period of one hour at 250 C under vacuum (2 mbar). The ASTM D6375 standard thermogravimetric test method was developed [5] to combine the advantages of the gas chromatographic method [6] with the realistic conditions of the traditional Noack test. The method is quicker and more reliable than both and can be performed with less sample material. According to the ASTM D6375 method, a sample is heated rapidly in a crucible to 249 C and held isothermally for 30 min at this temperature during which time the TGA curve is recorded. The Noack evaporation loss is the loss in mass up to the Noack reference time. This time is determined beforehand under the same experimental conditions with a Noack reference oil. In this method, it is important that the sample temperature quickly rises to a value between 247 and 249 C but does not overshoot. To simulate the traditional Noack method, a sample is usually heated to 220 C at 100 K/min and then to 249 C at 10 K/min. The sample mass (m s ) to be used is determined from the internal diameter of the crucible (d) using the following equation: m s = 350d 3 where d is in cm and m s in mg. Performing a Noack test The Noack evaporation test (also referred to as Noack volatility) was performed under the following conditions using a METTLER TOLEDO TGA instrument: Crucible: 100 µl aluminum without lid (internal diameter 0.56 cm) Sample mass: 61 ± 3 mg Purge and protective gas: total 80 ml/min air Noack reference oil: W with 10.93% mass loss up to the Noack time; supplier: Walter Herzog GmbH Temperature program: 50 to 220 C at 100 K/min, followed by heating to 249 C at 10 K/min and isothermal at 249 C. To achieve the above condition, the t lag parameter for this crucible was adjusted to zero. The sample (test oil) used was a synthetic 5W40 engine oil. Figure 1 shows the TGA curve (black line) of the reference oil. The Noack reference time is min; this is the time at which the certified loss of 10.93% is reached (see the black arrows). The TGA 7

8 Applications curve of the sample is shown as the red curve. After min the loss is read off as 7.80% (see red arrows): the synthetic 5W40 engine oil thus has a Noack evaporation loss of 7.8%. This is often given in the oil specification simply as NOACK 7.8%. According to ASTM D6375, the repeatability with 8% loss is about 1% for two determinations by the same laboratory, and the reproducibility is about 1.4% for two determinations by different laboratories. According to ASTM D6375, the TGA furnace must be regularly heated out. The method recommends that this is done after about ten determinations. The furnace should be heated to 1000 C without a crucible and held isothermally at this temperature for about 5 min. The air gas flow is left at about 80 ml/min. Conclusions Over the past years, the demands placed on lubricants in many application areas have changed significantly. Thermal oxidation stability, low tendency to evaporate and their influence on our natural and working environments have become very important. The innovative development of modern lubricants and their proper application have far-reaching economical consequences. Lubricants (base liquid and additives) that have been optimized for the different tasks, for example with low evaporation losses, save energy reduce service intervals minimize wear increase engine service life increase oil change intervals (lifetime) and result in considerable economical savings. The determination of the evaporation loss by thermogravimetry is therefore an important step in the qualification of lubricant. The METTLER TOLEDO TGA system with sample robot and automated evaluation provides high sample throughput and rapid pass/fail assessment of the oil in question. Acronyms API: American Petroleum Institute ASTM: ASTM International, originally known as the American Society for Testing and Materials DIN: Deutsches Institut für Normung (German Institute for Standardization) ILSAC: International Lubricant Standardization and Approval Committee JPI: Petroleum Association of Japan SAE: SAE International, Society of Automotive Engineers Literature [1] ASTM D6375 Standard Test Method for Evaporation Loss of Lubricating Oils by Thermogravimetric Analyzer (TGA) Noack Method. [2] ASTM D-5800 Test Method for Evaporation Loss of Lubricating Oils by the Noack Method. [3] DIN , Determination of the evaporative loss of petroleum products by the Noack method Part 1 (original in German). [4] JPI-5S Determination of Evaporation Loss of Engine Oils (Noack Method). [5] E. F. de Paz, C. B. Sneyd The Thermogravimetric Noack Test: a Precise, Safe and Fast Method for Measuring Lubricant Volatility, Subjects in Engine Oil Rheology and Tribology, SP1209, International Fall Fuel and Lubricants Meeting, San Antonio 1996, available also as SAE Technical Papers, Document Number: [6] DIN , Determination of the evaporation loss of petroleum products by gas chromatography Part 2 (original in German). Figure 1. Determination of the evaporation loss (Noack volatility) from a synthetic engine oil using the TGA Noack test method. Black continuous curve: reference oil; red curve: synthetic 5W40 engine oil; black dashed curve: sample temperature. 8

9 The characterization of polymorphs by thermal analysis Saikat Roy, Bipul Sarma und Ashwini Nangia, School of Chemistry, University of Hyderabad, India Dr. Matthias Wagner, Dr. Rudolf Riesen TGA-FTIR was used to characterize and distinguish between different polymorphs. If two polymorphic forms of a solid are present, whereby one form melts and the other sublimes or vaporizes at about the same temperature, then evolved gas analysis can be used to obtain quantitative (mass loss) and qualitative (spectral) data to analyze such solids. The two pharmaceutically important compounds shown in Figure 1, the active pharmaceutical ingredient (API) venlafaxine hydrochloride (Structure 1) and the well-known host material 1,1-bis(4- hydroxyphenyl)cyclohexane (Structure 2), were analyzed by DSC, TGA, hot-stage microscopy (HSM) and TGA-FTIR to study the phase transitions that occur on heating. A substance is said to exhibit polymorphism if it can exist in two or more crystal lattice forms. These are called polymorphs and have different physical properties [1]. Figure 1. Venlafaxine ((±)-1-[2(dimethylamino)-1- (4-methoxyphenyl)ethyl]cyclohexanol hydrochloride (Structure 1)) and 1,1-bis(4-hydroxyphenyl)cyclohexane (Structure 2). 204 C, phase from the melt) and Form 4 ( C, hydrate/alcoholate). A new amorphous, transient, glassy (semisolid) phase (Form 5) was isolated by sublimation under vacuum during the course of our thermal studies on this API [2]. TGA, TGA-FTIR and DSC measurements The TGA curves of the marketed drug Forms 1 and 2 showed complete loss of mass between 220 and 260 C (Figure 2a). We interpreted the mass loss as being due to decomposition or vaporization of the sample after melting. gaseous products formed in the TGA were passed through a heated transfer line to the FTIR spectrometer and FTIR spectra continuously recorded. The TGA-FTIR spectra of the vaporized VenHCl Forms 1 and 2 were identical and the main peaks matched the peaks in the solid state FTIR spectrum of VenHCl. This meant that VenHCl vapor is evolved from both forms after the phase change between 214 and Figure 2 (a). TGA (above) and DSC (below) of VenHCl Form 1 and Form 2. Note the mass loss that accompanies sublimation. Figure 2 (b). FTIR spectra of the vapor from Forms 1 (black) and 2 (red). Venlafaxine hydrochloride (VenHCl) VenHCl is a widely sold anti-depressant. The hydrochloride salt of venlafaxine, (±)-1-[2(dimethylamino)-1-(4-methoxy-phenyl)ethyl]cyclohexanol, exists in several different polymorphic modifications. The polymorphs of VenHCl are classified according to their main melting temperatures in the DSC: Form 1 ( C), Form 2 ( C), Form 3 (202- The vaporization products were analyzed by simultaneous FTIR spectroscopy. The Temperature C Wavenumber cm -1 Absorbance Units Temperature C Wavenumber cm -1 Absorbance Units Figure 2 (c and d). FTIR spectra as a function of temperature ( C) in (c) Form 1 and (d) Form 2. 9

10 Applications 216 C, a phenomenon that accompanies sublimation of the solid during the broad endothermic effect between 220 and 260 C. The TGA curves show that Form 2 sublimes more rapidly than Form 1. When VenHCl was sublimed/vaporized at reduced pressure (0.2 Torr, ~160 C), amorphous, semi-solid droplets formed on the cold finger (Figure 3a). The glassy mass was immediately transferred to a glass plate as liquid-like droplets (Figure 3b). The DSC curve of the sublimed semi-solid material showed crystallization between 95 and 100 C (exothermic) followed by melting between 216 and 218 C (endothermic), and finally a broad endothermic effect between 220 and 260 C due to sublimation/vaporization (Figure 3d, left). The exothermic effect at 100 C is due to the solidification of the glassy mass, the endothermic effect between 217 and 218 C from melting. Vapor loss occurs between 220 an 260 C. Figure 3d, right, shows the DSC of Ven- HCl hydrate obtained from Form 5 after one day exposure in open air (above) and the hydrate Form 4 prepared by crystallization from methanol (below). The endothermic effect at 80 C is due to the loss of solvent/water. Figure 3 (d). DSC curves of the freshly formed Forms 4 and 5. Figure 4 (a). Heating-coolingheating experiments performed on Form 1 with different maximum temperatures resulting in the formation of different end products. Figure 3 (a, b and c). The transient semi-solid, glassy phase on the cold finger of the sublimation apparatus (a) and droplets immediately placed on a glass plate (b). The sublimed material, Form 5, transforms to the hydrate, Form 4, after exposure to the Hyderabad climate (25-30 C, RH 40 50%) for one day. DSC heating-cooling-heating experiments The presence of two endotherms between 210 and 220 C in the DSC curves of Forms 1 and 2 (Figure 2 effects marked 1 and 2) raised the following questions. 1) Is the first endothermic effect due to a phase transition and the second due to melting, or vice versa? 2) Is the endo-exo peak in Form 2 a melting crystallization phenomenon? 3) Which polymorph is more stable? Do they interconvert or transform to a new, different phase? To clarify this, a number of DSC heating-cooling-heating experiments were performed. A) Form 1 was heated at 2 K/min to 212 C, a temperature that is just after the larger endothermic peak marked 1 (in Figure 2) but before the small peak marked 2. The sample was then cooled to room temperature at 5 K/min in the DSC cell (Figure 4a). Reheating at 2 K/min showed a broad endothermic peak at 212 C. This meant that the solid is still in Form 1 and not a transformed product. The exothermic behavior at 195 C in the cooling segment is due to solidification/crystallization of the melted Form 1. We believe that the peak between 210 and 212 C in Figure 4a is due to melting and is not a phase transition. B) When the same procedure was repeated but heating was continued to 219 C, just past the second small peak, the DSC curve of the reheated Form 1 is very different. There is now 10

11 C) D) a broad exothermic effect at 110 C and an endothermic effect at 200 C. The exothermic effect corresponds to the crystallization of the transformed Form 3, which melts at 200 C. Thus Form 1 undergoes a phase change to Form 3 on heating to C and cooling. In a similar procedure, Form 2 was heated in the DSC at 2 K/min to the endo-exo peak at 213 C and then cooled at 5 K/min to room temperature (Figure 4b). The cooling curve is flat and shows no crystallization, which means that the crystallization of Form 2 (exothermic effect at 213 C) was correctly assigned. Reheating at 2 K/min shows a sharp endothermic peak between 218 and 220 C corresponding to the melting of Form 5, the phase obtained by sublimation. On heating Form 2 to beyond the second endothermic effect up to 220 C, cooling to room temperature and then reheating, different peaks occur. Now the DSC cooling curve shows crystallization at 150 C and endothermic peaks that resemble Form 3. Form 1 30 C C C C Form 2 30 C C C C Figure 4 (b). Heating-coolingheating experiments performed on Form 2 to two different end temperatures again yield different results. Figure 5 (a and b). HSM of Form 1 and HSM of Form 2. The heating-cooling-heating curves show that Forms 1 and 2 first melt and then phase transform to different solidstate forms (3 and 5) in the range 210 to 220 C C Microcrystals under polarized light after cooling C Microcrystals under polarized light after cooling Hot-stage microscopy (HSM) Morphological and phase transitions in Forms 1 and 2 and the thermal events leading to sublimation of Form 5 were studied by HSM. The photomicrographs in Figure 5 show snapshots of the transformation of both solids to Form 5. Whereas the extent of vaporization was almost complete when the starting form was Form 2, it was only partial in the case of Form 1. Figure 6. BHPC (Structure 2) p 12. DSC of 2m and 2s polymorphs. The metastable phase 2m has a lower melting temperature and shows phase transition to the higher melting thermodynamically stable Form 2s. HSM measurements confirm the interpretation of the DSC curves in Figure 4 and the existence of the new transient, glassy phase Form 5. 11

12 Applications Figure 7 (a, b and c). TGA (a) and FTIR spectra of the evolved vapor of polymorphs 2s (b) and 2m (c). The measured FTIR peaks match with those expected for BHPC. 1,1-Bis(4-hydroxyphenyl)cyclohexane (BHPC) 1,1-bis(4-hydroxyphenyl)cyclohexane molecules are highly prone to forming inclusion complexes in over 30 host-guest crystal structures. We employed two solvent-free conditions, melt crystallization and sublimation under vacuum, to crystallize guestfree forms. Single crystals of BHPC were solved and refined in triclinic space group P1 (2s, Z = 1, sublimation phase) and in orthorhombic space group Pbca (2m, Z = 2, phase from the melt). Z is the number of symmetry-independent molecules in the crystallographic unit cell. TGA, TGA-FTIR and DSC measurements Phase relationships of 2s and 2m and possible mechanisms for their interconversion were studied by DSC and HSM [3]. The DSC curve of 2s showed a single broad endothermic peak at ~184 ºC (T peak, peak 1), while 2m shows two sharp endothermic peaks at 183 ºC and 188 ºC (Figure 6, peaks 2 and 3). These two peaks were assigned to the melting of 2m (peak 2), then crystallization (exothermic, peak 4) to 2s and finally fusion of the sublimed form (peak 3). On heating a second time in the DSC, both forms showed a single endothermic peak (peaks 5 and 6), implying transformation to the stable 2s polymorph. Polymorph 2m is a metastable phase, which shows a phase transition to the thermodynamic, sublimation polymorph 2s on heating to 200 C. Under the same conditions, polymorph 2s does not show phase changes except the vaporization endotherm. In general, the endothermic peak obtained on reheating is shifted by about 5 K to lower temperature compared to the first heating run due to better thermal contact of the sample with the crucible after melting. TGA-FTIR measurements of the evolved vapor were performed in order to confirm the sublimation process in BHPC. This was done by heating 8 12 mg of substance at 10 K/min in a dry nitrogen gas flow of 50 ml/min. The FTIR spectra showed that vaporization of both forms occurs after the phase change between 183 and 185 ºC (Figure 7). Sublimation below melting is two to three times more pronounced with polymorph 2s compared to the melt phase 2m, even though only a marginal loss of mass is observed on the TGA curve (Figure 7a). Hot-stage microscopy (HSM) Hot-stage microscopy shows blocks of melt crystals beginning to melt between 178 and 182 ºC with complete melting between 183 and 185 ºC. With both forms, cooling resulted in sublimed crystals with fine needle morphology (Figure 8). On the other hand, sublimed crystals 2s did not show any apparent crystal form change in a similar heating-cooling cycle on the hot stage. A combination of thermoanalytical methods such as TGA, DSC and HSM indicate that 2m is the metastable polymorph and 2s the thermodynamically stable phase (T 2m = 183 ± 1 ºC; T 2s = 188 ± 1 ºC). The single endothermic peak after reheating is the thermodynamically stable, Form m Form s (a) (d) (b) (e) (c) (f) Time in min Wavenumber cm -1 Transmitance in % Time in min Wavenumber cm -1 Transmitance in % Figure 8 (a to f). HSM images. 2m: (a) at 25 ºC, (b) ºC, (c) after cooling to room temperature 2s: (d) at 25 ºC, (e) ºC, (f) after cooling to room temperature. Phase transition of a block crystal of 2m to needle fibers of 2s (a c). (b) (c) 12

13 higher melting crystallized phase. TGA- FTIR confirms the vaporization/sublimation of BHPC. Figure 9. Phase transformations in VenHCl polymorphs 1 to 5. Conclusions The phase relationships between VenHCl polymorphs are summarized in Figure 9. In addition to recording quantitative and reproducible data on four previously reported Forms 1 to 4 of VenHCl, a new form, Form 5, was obtained by sublimation. Form 5 is short-lived under inert conditions (stable for a few hours up to one day). It transforms to the hydrate, Form 4, in the open air and to Form 1 under dry conditions. TGA, DSC and hot-stage microscopy show that the 5 K lower melting solid 2m form of BHPC is the metastable modification and 2s is the thermodynamically stable phase. The single endothermic peak after reheating both forms is ascribed to the thermodynamically stable, higher melting phase, which can also be obtained by vaporization/sublimation. In addition to the above application, we have also used TGA-FTIR to differentiate between aniline and phenol inclusion in a guest-selective host lattice [4]. Literature [1] W. C. McCrone, in Physics and Chemistry of the Organic Solid State, Vol. 2, D. Fox, M. M. Labes and A. Weissberger (Eds.), Wiley Interscience: New York, 1965, pp [2] S. Roy, S. Aitipamula and A. Nangia, Cryst. Growth Des. 2005, 5, [3] B. Sarma, S. Roy, and A. Nangia, Chem. Commun. 2006, [4] S. Aitipamula and A. Nangia, Chem. Eur. J. 2005, 11, Analysis of melting processes using TOPEM Dr. Jürgen Schawe This article discusses the conditions required for analyzing melting processes using TOPEM. If these conditions are fulfilled, the reversing heat flow measures processes that occur under equilibrium conditions and the non-reversing heat flow processes that involve supercooling or superheating. This separation allows a classification of melting processes and the differention of crystal structures of different stability. Introduction The measurement and interpretation of melting processes using temperature modulated DSC (TMDSC) is one of the more demanding tasks in thermal analysis. This is possibly the reason why a number of ideas and proposals can be found in the scientific literature that do not stand up to a critical analysis. Despite this, TMDSC can provide interesting and important information about melting behavior that would otherwise be difficult to obtain. Starting out from the basic principles of melting behavior discussed in reference [1], we want to show with the aid of suitable examples how melting behavior can be investigated using TOPEM. Basic principles of temperature modulated DSC Measurement principles and requirements In TMDSC, a conventional temperature program (heating or cooling at a constant rate, or isothermal conditions) is overlaid with a small temperature perturbation (modulation). In the evaluation algorithm, it is assumed that the reaction of the sample to the conventional temperature program and the modulation do 13

14 Applications Figure 1. Schematic diagram of free enthalpy as a function of temperature. Processes are shown that occur under local stable, metastable and unstable conditions. not influence each other. The underlying part (from the conventional temperature program) and the part from the modulation can then be separated. While just as in conventional DSC the underlying part of the heat flow (total heat flow) contains the entire information, the part that is generated by the modulation, only contains information about processes that can more or less follow the modulation. In all modulation techniques, the measurement conditions must be chosen so that measurement and evaluation take place under linear and almost stationary conditions. This means that the result is independent of the intensity (amplitude) of the modulation and that the total heat flow during a relevant evaluation window (period) does not change much. The quality of the measurement improves as the underlying heating rate is reduced. Especially in the analysis of melting processes, small modulations must be used because otherwise artifacts are measured that lead to the misinterpretation of results. TOPEM is a modern TMDSC technique that differs from conventional TMDSC with regard to the type of modulation function and evaluation. In TOPEM, a stochastic function is used for modulation. The intensity of the modulation function is the height of the pulse. The evaluation consists of a correlation analysis of the measured heat flow and heating rate in a selectable evaluation window. [2, 3]. Total, reversing and non-reversing heat flow In all TMDSC techniques, three heat flow components are determined from the measured heat flow. These are the total heat flow, Φ tot, the reversing heat flow, Φ rev, and the non-reversing heat flow, Φ non. In conventional TMDSC, the total heat flow is obtained from the measured heat flow by averaging over at least one period. The reversing heat flow is determined from the modulated component. The non-reversing heat flow is given by the difference: In TOPEM the evaluation is carried out by means of a correlation analysis of the heat flow and the heating rate. This yields the component of the measured heat flow that correlates with the heating rate and another component that does not correlate with the heating rate. The non-correlating component is the non-reversing heat flow, Φ non. The reversing heat flow is determined from the correlating heat flow part [3]. The total heat flow is calculated from the sum of the two quantities: At first sight, this difference in approach seems relatively unimportant. In TOPEM, however, it allows a consistency test of the measurement to be performed as described below. Sensible and latent heat flow In principle, the heat flow consists of two components namely the sensible heat flow, Φ s, and the latent heat flow, Φ l, [3, 4]. The latent heat flow does not explicitly depend on the temperature but on the kinetics of the thermal event. An example is the curing reaction of an adhesive. A temperature change during the reaction cannot cause the sample to return to its initial state. It will only change the reaction rate. The sensible heat flow depends explicitly on the heating rate. An example is the heat flow into an inert sample, which is directly proportional to the heating rate. Here, the proportionality factor is the heat capacity. Basic principles The starting point is the description of melting and crystallization processes by means of free enthalpy given in reference [1]. A diagram summarizing this is shown in Figure 1. The red, black and green curves are the free enthalpies of the melt, the crystal and the glass. The dashed curves represent intermediate states. The curve with the smallest free enthalpy characterizes the stable state. All other states are metastable. The system tries to achieve the stable state but is hindered by kinetic processes (e.g. nucleation). Processes occurring in a TOPEM measurement are marked by blue ellipses or arrows in Figure 1. These are processes that occur under quasi-stable, metastable and unstable conditions: Quasi-stable processes in (local) equilibrium are for example the measurement of heat capacity without another thermal event occurring. In processes under metastable conditions the system departs only slightly from local equilibrium. Examples of this are glass transitions or melting and crystallization processes close to local equilibrium conditions such as those occurring in the melting region of impure substances (see [1]). These processes can be practically reversed through a small change in temperature. 14

15 In processes with a large change in the sensible heat flow, which includes the free enthalpy, the system starts in processes that take place close to equilibrium. The latent heat flow is described by metastable equilibrium and drops down into the new more stable state. the last term. It includes processes that The process is hardly influenced by start far way from equilibrium. small temperature changes. Examples are crystallization processes after a Heat flow separation by TOPEM sufficiently large degree of supercooling or melting processes of crystals flows measured by TOPEM can be as- The reversing and non-reversing heat with superheating. signed to the sensible and latent heat flow if the linearity and stationarity requirements are adhered to within the bounds of measurement accuracy: Description of sensible and latent heat flows Let us assume that two different processes take place in a sample that can each be described with an order parameter, x. On melting, x, describes the degree of disorder and changes from x = 0 (ideal crystal) to x = 1 (equilibrated melt). The process with the order parameter, x me, takes place close to local equilibrium. The other process begins far way from equilibrium and has the order parameter x i. The measured heat is given by: where c p is the specific heat capacity, dt/dt is the heating rate, h me and h i are the specific transition enthalpies assigned to the corresponding processes. Since the process (me) takes place close to local equilibrium, x me can follow the small temperature modulation. For this case, we can write: Testing measurement conditions Linearity Since the evaluation methodology in TMDSC is based on the analysis of linear systems, the measurement program must be chosen so that the measured heat flow satisfies linear conditions. This is the case if the reversing heat flow is independent of the intensity of the temperature modulation (the pulse height, i.e. the amplitude). The maximum intensity depends on the material and the events to be investigated. With melting, the linearity limit is usually less than 0.1 K. Figure 2 shows an example of a linearity test using the melting of polyethylene terephthalate (PET). Two samples of similar mass were measured with different pulse heights and an underlying heating rate of 0.3 K/min. The pulse heights were ±5 mk and ±50 mk. The reversing heat flow curves in Figure 2 are independent of the pulse height. The noise is however larger with the small modulation intensity. The blue curve is the difference between the two Φ rev curves. This material can be measured with a pulse height of ±50 mk. Stationarity The total heat flow should at the most change only slightly in an evaluation window. This condition cannot always be fulfilled particularly at higher heating rates during relatively rapid thermal events. In contrast to conventional TMDSC, TOPEM offers the possibility to detect regions of curves in which an interpretation is critical. This is done by comparing the measured heat flow and the total heat flow. Figure 3 shows these curves in the melting region of a 40:60 mass % sucrose-water mixture. Over a wide range, the total heat flow corresponds to the mean value of the measured heat flow. In this range, the curves obtained can be evaluated both quantitatively and qualitatively. The total heat flow is however too small in the region of the peak maximum and the In the non-equilibrium process (i), the order parameter does not follow the small temperature change, T, determined by the modulation function. So that in this case: Substitution of eq (4) in eq (3) gives the measured heat flow: Figure 2. Test of the linearity condition using PET that had been crystallized at 170 C. Above: measurement curves. Below: reversing heat flow and the difference between the two curves (blue). The first term in eq (6) is an explicit function of the heating rate. Here it is 15

16 Applications Figure 3. TOPEM measurement in the melting region of a sucrosewater mixture. high temperature side of the peak. The measurement curve is primarily determined by the heat transport and less by the melting process. To obtain quantitative results with TOPEM in this region, the heating rate must be reduced. Melting processes When TOPEM is applied to melting processes, several different cases have to be considered. Melting of pure materials Pure materials (e.g. indium) melt at the equilibrium melting temperature. During the melting process the sample temperature does not change [5] and can not therefore follow a temperature modulation. The measured TMDSC curves originate mainly from the change in the heat transport conditions. TMDSC methods are thus unsuitable for the measurement of sharp transitions of this type. Reversible melting close to local equilibrium Melting processes during which crystals and melt exist in local equilibrium occur for example with mixtures that have a broad melting range. Such processes should according to eqs (8) and (9) supply a contribution to the reversing heat flow, while the non-reversing heat flow is small. The diagram on the left side of Figure 4 shows the simplified phase diagram of the sucrose-water system [10]. The path taken on heating is marked by arrows. The melting process begins at about 36 C with the melting of small nonequilibrated crystals. This gives rise to a melt with a critical concentration of about 80 mass% sucrose. The non-reversing heat flow shows the corresponding peak. Afterward the melting process follows the liquidus line whereby crystals and melt are in local equilibrium. This part of the melting process supplies a contribution to the reversing heat flow. The melting processes that give rise to a reversing heat flow are called reversible melting and the others non-reversible melting. Non-equilibrium melting behavior: superheating of polymers With many semicrystalline polymers, relatively stable crystallites superheat and melt above their thermodynamic melting temperature. In this situation, the melt and crystallites are not in thermodynamic equilibrium. The melting process is non-reversible. A sample of polyethylene terephthalate (PET) was first crystallized for 10 min at 170 C. The sample was then measured using a pulse height of ±50 mk and an underlying heating rate von 0.3 K/min. The measurement curves in the region of the main melting peak are shown in Figure 5. As expected, the peak in the Φ non curve is significantly larger than in the Φ rev curve. Figure 4. Left: Simplified phase diagram of a sucrose and water mixture. The path taken by the measurement is marked by red arrows. Right: Curves obtained from a TOPEM measurement. Differentiation of different stable crystallites PET crystallizes on cooling at 0.5 K/min from the equilibrated melt. In the following heating measurement at 0.5 K/min, the total heat flow curve exhibits a double peak (Figure 6). The TOPEM measurement shows that the reason for the double peak lies in the existence of crystallites of different stability. In the first peak, the reversing heat flow predominates. Smaller crystallites melt reversibly close to their equilibrium melting temperature. In the second peak, crystallites melt with superheating. This peak is almost entirely to be found in the non-reversing heat flow. The melting process is fundamentally different to that observed at a temperature about 10 K lower. 16

17 Conclusions When TOPEM is used to investigate melting processes, attention must be paid to the linearity of the measurement. The linear region is determined beforehand in preliminary experiments. The range in which TOPEM curves are valid can be established by comparing the total heat flow and the mean value of the measured heat flow. Processes that take place under conditions of local equilibrium can be detected in the reversing heat flow because they more or less follow the temperature modulation. Processes that take place far from equilibrium do not follow the temperature modulation and thus contribute to the non-reversing heat flow. Literature [1] J. Schawe, UserCom 24, 11. [2] UserCom 22, 6. [3] J. Schawe, T. Hütter, C. Heitz, I. Alig, D. Lellinger, Thermochimica Acta 446 (2006) 147. [4] J. Schawe, UserCom 22, 16. [5] J. Schawe, UserCom 23, 6. [6] J. Schawe, Thermochimica Acta 451 (2006) 115. Figure 5. Heat flow curves in the melting region of PET that had been crystallized for 10 min at 170 C. The underlying heating rate was 0.3 K/min. Figure 6. The melting behavior of PET that had been previously cooled at 0.5 K/min shows two different melting processes. 17

18 Characterization of delivery systems by thermogravimetry Applications Dr. V. Normand, K. Aeberhardt, Firmenich S.A., Geneva, Switzerland Introduction A good perfume is of course expected to provide a pleasant and distinctive odor. At the same time, the fragrances should remain perceptible for as long as possible at a constant level of intensity. For this reason, the fragrances in perfumes are now being encapsulated in so-called delivery systems. The release of fragrances then occurs under control, allowing the perception of the perfume to be optimized with respect to intensity and lasting effect. The encapsulation of fragrances in suitable delivery systems is therefore a topic of great importance for producers of perfumes. To identify the most suitable delivery system from the very large number of possible carriers available requires a rapid analytical screening technique that can describe the stability and release performance of a fragrance from the delivery system. Thermogravimetry (TGA) is an excellent technique for this purpose. In this article, the release of Romascone from three different delivery systems has been investigated using thermogravimetry. Romascone is a fragrance that finds application in women s perfumes. The delivery systems utilized three different types of polymeric nanoparticles based on crosslinked vinyl acetate. Theoretical principles Evaporation of pure liquids If volatile compounds (such as fragrances) are measured in the TGA, a continual loss of mass is expected because the furnace is open and an equilibrium state is never reached. This is because there is a steady transfer of molecules from the liquid phase to the gas phase. Molecules that reach the boundary layer between the liquid and the gas phase are swept out of the TGA furnace by the purge gas. Under isothermal conditions, this results in a constant rate of loss of mass which is determined by the vapor pressure of the compound and the mass transfer at the boundary layer, i.e. where m is the mass, k is a constant that describes the mass transfer at the boundary layer between the liquid and the gas phase and P vap is the vapor pressure. Evaporation of a compound from a binary mixture In a mixture of two compounds, the chemical potentials of the two compounds in the mixture are reduced compared to the chemical potentials of the individual pure compounds. Here P 1 and P 2 are the partial pressures of the two compounds, x 1 and x 2 are their mole fractions and P 1 0 and P 2 0 are the vapor pressures under normal conditions. For real compounds and assuming that only one substance is volatile, the Flory approximation applies and the partial pressure of the volatile component is given by: Here f 1 stands for the volume fraction of the volatile component (solvent) and c for the so-called Flory interaction parameter. For mixing of the two components to occur spontaneously, the mixing enthalpy (expressed here by the Flory interaction parameter, c) must be small. Typically c varies between 0 (for good solvents, athermic mixing) and 0.5 (for bad solvents, endothermic mixing). If the interaction parameter is greater than 0.5, demixing of the system is expected. If the density of the two compounds in the mixture is about the same, the volume fraction of the solvent (f 1 ) equals its mass fraction (w 1 ). The rate of mass loss is then given by the equation: Experimental details The investigations described here were performed with a METTLER TOLEDO TGA851/SDTA e equipped with the small furnace. Samples masses of typically 8 mg (fragrance and nanoparticles together) were measured in aluminum crucibles. The mass fraction of the nanoparticles made up 40% of the total mass. The purge gas was nitrogen at 20 ml/min. The measurements were performed isothermally at different temperatures. In a binary mixture of two ideal noninteracting compounds with molecules of equal size, Raoult s law predicts that the partial pressure of each compound is proportional to the mole fraction of each species in the mixture, that is: If the first derivative of the TGA signal is plotted as a function of the mass fraction of the solvent, the parameters k P vap and c can be determined from a fit of this curve with the function according to eq 4. If several isothermal measurements of the evaporation behavior are performed, the temperature dependence of the two parameters can be investigated. The fol- 18

19 lowing relationship is then expected for the interaction parameter: where W is the mixing enthalpy of the system, k the Boltzman constant and T the temperature. Evaporation limited by diffusion In eq 4 it is assumed that the evaporation rate of the volatile component is given solely by its vapor pressure. In a delivery system, the evaporation rate of the volatile component can also be limited by diffusion of the volatile molecules to the surface of the delivery system. In this case Fick s law applies and we obtain: Here D stands for the diffusion coefficient in the delivery system, which here depends on whether the delivery system is liquid, rubbery or glassy. A is the surface area of exchange, dc/dr the concentration gradient within the delivery system and a, a and a are constants that take into account the dependence of the diffusion on the volume or mass fraction of the volatile component in the delivery system. If the delivery system consists of nanoparticles as in the case described here, it can be assumed that the concentration gradient within the particles is constant after a short time. In this case, the rate of mass loss is proportional to the diffusion coefficient. Results and discussion In the delivery system investigated here, Romascone was encapsulated in nanoparticles made of different types of crosslinked vinyl acetate. With Samples A and B the degree of crosslinking was chosen so that the delivery system was in a rubbery elastic state. With Sample C, more heavily crosslinked nanoparticles were used so that the delivery system was in the glassy state. Evaporation of pure Romascone Figure 1 shows the isothermal TGA curves of pure Romascone measured at different temperatures. The curves show that although the rates of mass loss for the various temperatures are different, they do not change during the experiment. Physically, this means that the rate of evaporation is only determined by the temperature-dependent vapor pressure of the fragrance, which means that the evaporation process can be described by eq 1. If the slopes of the curves are plotted logarithmically as a function of the reciprocal temperature in Kelvin (red curve), the straight line expected from the Clausius-Clapeyron equation is obtained. Evaporation of Romascone from a rubbery delivery system In these experiments, the delivery system was in a rubbery-elastic state. The results (see Figure 2) show that in these samples a bend appears in the TGA curve after a certain time (except for the measurement at 25 C; in this case the measurement time was not sufficiently long). This bend occurs when evaporation is limited by the transport processes in the polymer (diffusion). The slope of the TGA curves before the bend is not constant either. According to eq 4, the evaporation rate depends on the Romascone fraction (the evaporation rate corresponds to the slope of the TGA curve). The evaporation rate as a function of the Romascone mass fraction can be calculated based on the amount of Romascone present after a certain time. The corresponding data for the measurement at 40 C together with the best fit curve according to eq 4 is displayed in Figure 3. In the same figure, the values found with the fit for k P vap for the three temperatures (25 C, 40 C and 70 C) were compared with the corresponding values for pure Romascone (calculated from the slopes of the mass loss curves in Figure 1). It was found that the values for k Pvap for the delivery system are systematically lower than those for pure Romascone. The temperature dependence of the interaction parameter, c, is plotted for two different rubbery delivery systems in Figure 4. The two samples (Samples A and B) differ in the degree of crosslinking of the nanoparticles used for the delivery system (nanoparticles in Sample A are more strongly crosslinked that those in Sample B). The figure shows the expected linear increase with increasing temperature (see eq 5). Evaporation of Romascone from a glassy delivery system In these measurements, the evaporation of Romascone from a glassy delivery system (Sample C) was investigated. The results of the measurements at different Figure 1. Evaporation of pure Romascone at various temperatures. Figure 2. Evaporation of Romascone from a delivery system in the rubbery-elastic state at various temperatures. Figure 3. Rate of loss of mass of Romascone from a rubbery-elastic delivery system at 40 C. Inserted diagram: the parameter k P vap for the delivery system and pure Romascone at various temperatures. Figure 4. Temperature dependence of the interaction parameter, c. The slope of the curve describes the interaction energy between Romascone and the nanopolymer. It can be seen that the interaction energy between the more lightly crosslinked nanoparticles and Romascone (Sample B) is larger than that between the more heavily crosslinked delivery system and Romascone. 19

20 Applications Figure 5. Evaporation of Romascone from glassy polymeric nanoparticles. temperatures are displayed in Figure 5. It shows that in each case the system asymptotically approaches a constant, temperature-dependent composition. This means that in this delivery system, part of the Romascone remains in the delivery system and is not released. The equilibrium concentration is however not reached with the timescale of the experiment. The analysis of the data according to the approach of Flory (eq 4) gives unrealistic values for the unknown parameters k P vap and c. In fact, in this case the evaporation rate is determined by the diffusion of the Romascone molecules to the surface of the nanoparticles practically from the beginning, so that the evaporation behavior of the fragrance is described by eq 6. Accordingly, a linear relationship between the logarithm of the evaporation rate and the volume fraction of the volatile components is expected. The results in Figure 6 show that apparently two curves with different slopes are required to describe the data. The reason for this behavior is that Romascone acts as a plasticizer for the nanoparticles. Depending on the Romascone fraction and the temperature, the delivery system is either in the glassy or rubbery state, which leads to the different slopes in the figure: the blue curves describe data for a delivery system in the rubbery elastic state, and the red lines for a delivery system in the glassy state. Whether the delivery system is in the glassy or rubbery state depends on the actual Romascone mass fraction at a particular temperature. The point of intersection of a red and the corresponding blue curve therefore corresponds to the critical Romascone mass fraction for the temperature: above this mass fraction the delivery system is rubbery elastic, and below it is glassy. The figure shows that at high temperatures and high Romascone fractions, the rate of mass loss is higher. In the same way, it is clear that the diffusion of Romascone in a glassy delivery system is appreciably slower than in the rubbery-elastic delivery system. The different slopes at the different temperatures therefore describe the temperature dependence of the diffusion coefficient. If the logarithm of the slopes of the curves in Figure 6 are displayed as a function of the reciprocal temperature, the curves shown in Figure 7 are obtained. The evaporation process can therefore be understood as an activated process according to the equation Conclusions The evaporation of volatile substances from delivery systems can be investigated by thermogravimetry. In the example, the evaporation of Romascone from nanoparticles based on crosslinked vinyl acetate was investigated. If the delivery system is in the rubbery-elastic state, the evaporation of Romascone can be described by the Flory theory (evaporation is limited by the volatility of the volatile substance). If the delivery system is in the glassy state, the evaporation is limited by the diffusion of the volatile substance within the nanoparticles and takes correspondingly longer. The method described here is simple and allows delivery systems to be quickly characterized and optimized. Literature [1] L. Ouali, G. Léon, V. Normand, H. Johnsen, A. Dyrli, R. Schmid and D. Benczédi, Mechanism of Romascone Release from Hydrolized Vinyl Acetate Nanoparticles, Polymers for advanced Technologies, 2006 (17), Figure 6. Evaporation of Romascone from a glassy delivery system at different temperatures. Figure 7. Arrhenius diagram for glassy (red curve) and rubbery elastic (blue curve) delivery systems. The activation energy for both delivery systems is about 17.2 kj/g. Here a(t) represents the slope of the mass loss curve, E a is the activation energy, R the gas constant and T the temperature. The blue curve describes the evaporation from the rubbery-elastic delivery system, and the red the evaporation from the glassy delivery system. The two curves have approximately the same slope. This indicates that the activation energy of the evaporation processes is independent of the state of the delivery system (rubbery-elastic or glassy), at whose surface the evaporation takes place. Evaluation of the data gives an activation energy of about 17.2 kj/g. 20

21 Detection and evaluation of weak sample effects in DSC Tips and hints Figure 1. Despite the large sample mass (52 mg aluminum oxide powder) and fast heating rate (20 K/min), the noise and slope can be reduced to the level of that obtained with empty sample and reference crucibles (rms noise < 0.5 uw, slope < 0.07 mw in the temperature range shown) by using a reference material (aluminum oxide powder). Marco Zappa In DSC measurements, a large sample mass is often used in order to detect weak sample effects. Despite the large mass, it is often still difficult to detect and evaluate very weak effects. This has to do with the high heat capacity of a large sample, which increases both the signal noise and the slope of the DSC curve. The noise and slope can be reduced by using a suitable reference material in the reference crucible. This largely balances out or compensates the effect of the heat capacity of the sample during the measurement. The sample effects are then much more clearly observed and evaluated due to improved signal-to-noise ratio and the much flatter DSC curve. Improving the signal-to-noise ratio reducing the slope The larger the difference of the heat flows between the reference and sample sides of the detector outside regions in which thermal events occur, the greater the noise and the steeper the slope of the DSC curve. The different heat flows are a result of thermal asymmetry between the sample and the reference sides due to the heat capacity of the sample itself. This asymmetry is greater, the heavier the sample, the larger its specific heat capacity, and the faster the sample is heated or cooled. The unequal heat capacities can be compensated by using an appropriate reference material. In the ideal case, the differential heat flow signal outside regions in which thermal effects of the sample occur is then zero. Noise and slope are thereby reduced to the level of that obtained with empty sample and reference crucibles (see Figure 1). How to compensate the sample heat capacity In general, the reference is an empty crucible identical to the sample crucible. The reference can however be adapted to the sample properties in order to achieve better thermal symmetry, i.e. similar heat capacities with similar temperature dependence on the sample and reference sides. This is achieved by filling the reference crucible with a reference material whose thermal mass is equivalent to that of the sample. The compensation of the heat capacity of the sample by the reference material at a particular temperature is given by the equation The mass of the sample, m S, multiplied by its specific heat capacity, c ps, should be equal to the specific heat capacity of the reference material, c pr, multiplied by its mass, m R. Example: To compensate the heat capacity of a sample of polystyrene (30.0 mg) using aluminum oxide powder as reference material on the reference side: Sample: polystyrene, m S = 30 mg, c p = 1.17 J/g K (at room temperature) Reference material: aluminum oxide powder, m R, cp = 0.78 J/g K (at room temperature) The calculation according to eq (1) yields: 30.0 mg * 1.17 J/g K = m R * 0.78 J/g K The required mass of the reference material, m R, is therefore 45.3 mg. The ideal reference sample An ideal reference material: does not exhibit any thermal effects or any discontinuities in specific heat capacity in the temperature range investigated. does not react with the crucible material or with the surrounding atmosphere (especially in the temperature range of the sample transition). is easy to dispense (e.g. powder or liquid). in the case of liquids, has a higher melting and boiling point than the sample. Here it is advisable to use her- 21

22 Tips and hints Figure 2. Denaturation of proteins: the endothermic peak with a peak height of only 6 uw can be measured with confidence using a reference solution, despite the very low protein concentration (0.1 weight %). In contrast, without the reference solution, the presence of a peak is questionable, even though the slope of the DSC curve was mathematically corrected. metically sealed crucibles or to work under pressure to prevent vaporization of the reference material and the sample material itself. has a specific heat capacity whose temperature dependence is known or is the same or similar to that of the sample. In many cases, dry α-aluminum oxide powder (c p (25 C) = 0.78 J/g K) is a suitable solid reference material. For dilute aqueous solutions (e.g. sugar, starch or protein solutions, etc.), water or a suitable buffer solution (c p (25 C) = 4.18 J/g K) can be used. With highly filled samples, it is best to use the filler itself as the reference material. To achieve the best accuracy, it is advisable to adjust the temperature and heat flow with reference material in the sample and reference crucibles. Detecting denaturation in dilute protein solutions Heat capacity compensation with a reference solution improves the signal-tonoise ratio to such an extent that proteins can be studied in dilute solution. Figure 2 shows that the very weak endothermic peak of the denaturation of lysozyme in solution at a concentration of just 0.1 mass % can easily be detected if a reference solution is used. If the 105 mg sample mass is not compensated, the denaturation peak can no longer be detected with certainty. Evaluating weak glass transitions with confidence Composites reinforced with fibers contain only a small amount of the matrix resin and exhibit only a very weak glass transition. Often it is not possible to evaluate the transition with confidence because of the marked slope of the DSC curve. In this case, the slope is due to the relatively large temperature dependence of the specific heat capacity. The determination of the glass transition can be greatly improved if carbon fibers are used as the reference material. The advantage compared with aluminum oxide powder is that the specific heat capacity of the carbon fibers exhibits the same temperature dependence as a large part of the sample itself. The slope is therefore reduced in an ideal way and the glass transition is clearly observed. Summary Heat capacity compensation using a reference material in the reference crucible is a simple and effective method to reduce both the signal noise and the slope of the DSC curve. This is especially useful if very weak thermal effects in large samples at high heating rates have to be measured, i.e. under conditions that increase the noise and slope. Lower noise is equivalent to improved sensor sensitivity. Together with reduced slope, this allows weak effects to be detected and evaluated with confidence. Figure 3. Epoxy resin/carbon fiber composite material. The very weak glass transitions remain almost undetected in the non-compensated measurements. The unequivocal identification of a glass transition would be too uncertain. In contrast, in the compensated measurements, the glass transitions can be identified and evaluated with certainty. The results show that denaturation in very dilute protein solutions and weak glass transitions in reinforced carbonfiber composite materials can only be reliably detected and evaluated if compensation is used. 22