1
Vitrification, nucleation and crystallization in phenyl-2-hydroxybenzoate (salol) studied by Differential Scanning Calorimetry (DSC) and Thermally Stimulated Depolarisation Currents (TSDC)

2
The behaviour of phenyl-2-hydroxybenzoate (salol) with respect to crystallization and vitrification was studied by Thermally Stimulated Depolarisation Currents (TSDC) and Differential Scanning Calorimetry (DSC).

3
It was found that this compound can be easily supercooled, and vitrified by further cooling.

4
The degree of supercooling can be high, even for low cooling rates.

5
The fragility index was found to be m ≅ 60 by DSC and by TSDC, a value that is lower than the values previously reported.

6
Furthermore, glassy salol crystallises on heating (cold crystallisation).

7
Two different polymorphs were detected, and the conditions of formation of each of them have been clearly established.

8
Finally, it was shown that the β-relaxation of salol, that is difficult to observe by Dielectric Relaxation Spectroscopy, was very clearly detected by TSDC.

9
This mobility, that is believed to assist the occurrence of the nucleation process, was found to be narrowly distributed in activation energy, with a mean value around Ea = 46 kJ mol−1.

10
The mean characteristic time at the glass transition temperature of these molecular motions was found to be of the order of τβ(Tg) ≅ 10−2 s.

Introduction

11
Salol, or phenyl-2-hydroxybenzoate, or phenyl salicylate (see Fig. 1) are three names for an ester with diversified applications, namely as a stabilizer for cellulosic and vinyl plastics, as an ingredient for suntan preparations, and also as an analgesic and antipyretic.

12
Furthermore, it is a fragile glass-forming substance and, in this context, it is considered as a model substance for the study of the glass transition and of the molecular mobility in the supercooled liquid state and in the glass.1

13
It was found, a long time ago, that crystalline salol can exist in, at least, two different crystal structures,2–4 one stable that melts at Tfus = 41.7° C, and the other, instable, with Tfus = 28.5° C.

14
Furthermore, the crystallization behaviour of salol has been the subject of investigation for many years.3–6

15
Salol is easily supercooled, that is to say, it has a low tendency to crystallize on cooling and it is relatively easy to vitrify.

16
It was found7 that the crystallization due to crystal growth occurs in the temperature interval between 254 and 313 K (between −19 and +40° C) with a maximum at 297 K (26° C).

17
More recently, a different crystallization mechanism was identified by Oguni et al8,9. in the vicinity of the glass transition temperature (Tg = 211 K = −52° C at 10 K min−1 obtained by DSC), and it was demonstrated that the crystals formed at 220 K and at 300 K have the same crystal structure.

18
It is recognised that the homogeneous nucleation rate has a maximum near Tg in many cases, given that the development of structured clusters of molecules is enhanced as the temperature decreases.

19
In particular, this is expected to be the case for fragile liquids, given that they have more opportunities for the development of structured cluster-like configurations.

20
Below Tg, however, the motional rearrangement of molecules becomes sluggish, so that the nucleation process should be kinetically reduced.

21
Despite this general idea, the existence of a slow nucleation mechanism below Tg in salol was proposed in a recent paper.10

22
According to this work, the crystal formed as a consequence of this mechanism is a metastable polymorph that melts at Tfus = 300 K = 27° C, and the β-molecular relaxation process is believed to be the motional process that controls the nucleation and growth mechanisms under such conditions.11,12

23
In the present work we report some results obtained by Differential Scanning Calorimetry (DSC) that provide us with a better understanding of the thermal behaviour of salol.

24
In this context we will characterise the glass transition of salol, and we will determine the fragility index of this glass-forming system for the heating rate dependence of the onset temperature of the glass transition DSC signal.

25
The conditions for nucleation, as well as the conditions that enable the crystallization of salol to generate one or the other of the two crystalline polymorphic forms, will be discussed.

26
The technique of Thermally Stimulated Depolarisation Currents (TSDC) will also be used with two purposes.

27
One of them is to monitor the cold crystallization process of salol, the other is to study the β-relaxation process that is believed to assist the nucleation process.

Experimental

28
Salol (C13H10O3, CAS no.

29
118-55-8), was purchased from Aldrich (99% purity).

30
It was purified by sublimation at room temperature under a pressure of 1.3 × 10−1 Pa.

31
The calorimetric measurements were performed with a 2920 MDSC system from TA Instruments Inc. Dry high purity He gas with a flow rate of 30 cm3 min−1 was purged through the sample.

32
Cooling was accomplished with the liquid nitrogen cooling accessory (LNCA) which provides automatic and continuous programmed sample cooling down to −150° C.

33
The baseline was calibrated scanning the temperature domain of the experiments with an empty pan.

34
The temperature calibration was performed taking the onset of the endothermic melting peak at a heating rate of q = 10 K min−1 of several calibration standards: n-decane (Tfus = 243.75 K), n-octadecane (Tfus = 301.77 K), hexatriacontane (Tfus = 347.30 K), indium (Tfus = 430.61 K) and tin (Tfus = 506.03 K).

35
The organic standards were high purity Fluka products, while the metal standards were supplied by TA Instruments Inc. The correction of the temperature for different heating rates was performed on the basis of results obtained with indium.

36
The onset temperature of the melting peak of indium, (Ton)fus, was obtained at several heating rates, qh, and the obtained values were fitted to a third order polynomial, (Ton)fus = a × q3h + b × q2h + c × qh + d.

37
The temperature correction for each heating rate allows the extrapolation to the melting of indium at infinitely slow rate conditions.

38
In this context, the correction for each heating rate, (ΔT)qh = a × q3h + b × q2h + c × qh, was subtracted from the experimental values in order to correct the data for the effect of the heating rate on the temperature of the DSC signals.

39
The enthalpy was also calibrated using indium (melting enthalpy ΔfusH = 28.71 J g−1).

40
Thermally Stimulated Depolarisation Current (TSDC) experiments were carried out with a TSC/RMA spectrometer (TherMold, Stamford, CT, USA) covering the range from −170 to +400 °C.

41
In order to analyse specific regions of the TSDC spectrum, different methods of polarising the sample can be used, namely the so-called TSDC global experiment and the Thermal Sampling (TS) experiment (often called thermal windowing or cleaning, or partial polarisation).

42
The TS method, where the polarising field is applied in a narrow temperature interval, enables one to resolve a global peak into its individual relaxation modes.

43
The physical background of the TSDC technique is presented elsewhere.13,14

44
The basic description of the TSDC experiment, and the discussion of the nature of the information it provides, is presented in detail in recent publications15,16.

Results and discussion

45
As underlined before, an important feature of the thermal behaviour of salol is that the liquid can be cooled down well below the melting temperature without crystallizing.

46
This can be the case even for cooling rates as low as 1 K min−1, and the supercooling can be as deep as 80° below Tfus.

47
It is to be underlined that nucleation and crystal growth are somewhat probabilistic events, so that our statements have in this context a considerable statistical or probabilistic signification.

48
In the case of the behaviour of liquid salol upon cooling, that means that the crystallisation on cooling from the isotropic melt is a relatively infrequent occurrence and that, in most cases, the metastable liquid presents a high stability, enabling an easy handling.

49
As will be seen later, crystalline phenyl salicylate presents two polymorphic forms.

50
One of them has an onset melting temperature at (Tfus)on = 30° C independent of the heating rate, and a temperature of maximum intensity slightly dependent on the heating rate (Tfus)max = ∼32° C.

51
The melting enthalpy of this polymorph was found to be approximately ΔfusH = ∼16.5 kJ mol−1.

52
The other polymorph has an onset melting temperature at (Tfus)on = 41° C which is also independent of the heating rate, and a temperature of maximum intensity slightly dependent on the heating rate (Tfus)max = ∼43° C.

53
The melting enthalpy of this polymorph was found to be ΔfusH = 18.6 ± 0.2 kJ mol−1, in good agreement with the value of 19.2 kJ mol−1 reported in the literature.9

The glass transition

54
The glass transition temperature was found to be Tg = 221 K = −52° C at 10 K min−1 (onset temperature of the DSC glass transition signal), and the heat capacity jump at the glass transition is ΔCp(Tg) = (111 ± 2) J K−1 mol−1, in excellent agreement with the literature values.9

55
The influence of the heating rate on the onset temperature, Ton, of the DSC glass transition signal was used to evaluate the activation energy at the glass transition temperature, and the fragility index of this glass-former.17,18

56
The obtained results, after correction of the temperature axis for the different heating rates, are plotted in Fig. 2.

57
The slope of the regression straight-line is −30.3 (with a confidence interval of ±0.8).

58
From this slope, the value of 252 kJ mol−1 was obtained for the activation energy of the structural relaxation, leading to the value of m = 60 for the fragility index of salol, that confirms the fragile character of the compound.

59
It is to be reported that this value is not very far from the value m = 63 obtained by Dielectric Relaxation Spectroscopy19 and from m = 68 obtained by Modulated Temperature Differential Scanning Calorimetry,20 but it is significantly lower than m = 76 obtained by Thermally Stimulated Depolarisation Currents,21 and from the fictive temperatures of differently quenched glasses obtained by DSC.22

60
Given the scattering of these values, we decided to repeat the determination of the fragility index of salol using the technique of Thermally Stimulated Depolarisation Currents (TSDC).

61
The sample we used now was from the same batch that was used for DSC measurements, previously purified by sublimation as described in the Experimental.

62
From these new TSDC results, that were treated according to the procedures described elsewhere,16 the value of 255 kJ mol−1 was obtained for the activation energy of the structural relaxation, which corresponds to a value m = 61 for the fragility index of salol.

Nucleation and crystallization

63
Nucleation is believed to be the first step of the crystallization process.

64
It is a stochastic process that depends on a large number of factors, namely size of the sample, type of cell or container used, cooling or heating rate, presence of impurities, etc… According to Oguni et al.,8 the homogeneous nucleation takes place in many cases above but near Tg.

65
This is so because the development of structured clusters of molecules is enhanced as the temperature decreases, and because below Tg the motional rearrangement of molecules becomes sluggish, so that the nucleation process should be kinetically reduced.8

66
Heterogeneous nucleation, on the other hand, depends on many factors and can occur in a wide temperature domain.

67
In order to characterize, as precisely as possible, the conditions for nucleation in salol, we carried out a series of successive experiments described as follows: 1—the sample melted at 60° C is cooled down at ∼25 K min−1 to a temperature T0; 2—the sample is heated at constant rate (10 K min−1) from T0 up to 60° C.

68
The temperature T0 varied from experiment to experiment, from −65° C (well below the glass transition temperature) up to +5° C.

69
In all the experiments cold crystallization was observed, indicating that nucleation is not restricted to a narrow temperature interval in the vicinity of the glass transition temperature, but that it takes place in a wide temperature interval above Tg.

70
Fig. 3 shows the result of different DSC experiments where the sample was heated up from the glassy state at T = −90° C up to above the melting temperature, at different heating rates.

71
Prior to each experiment the sample was vitrified by cooling from the melt (T = 60° C) down to T = −90° C.

72
The results of these experiments show three essential features: 1—the glass transition signal at Tg ≅ −52° C (not shown in Fig. 3); 2—an exothermic signal that corresponds to the cold crystallization (crystallization on heating) and is observed in the temperature interval between T = −25° C and the melting peak; 3—and the endothermic melting peak at Tfus.

73
It can be observed that the exothermic signal of the cold crystallization is strongly sensitive, in shape as well as in position, to changes in the heating rate.

74
In fact, for low heating rates the exothermic peak is relatively narrow and well separated from the melting peak.

75
The onset temperature of the crystallization process (that we identify with the onset of the exothermic signal at the lowest heating rate of 1 K min−1) is (Tcr)on = −25° C.

76
In contrast, for higher heating rates, the position of the exothermic signal shifts to higher temperatures and eventually interferes with the melting peak.

77
Another important conclusion that we can draw from the results shown in Fig. 3 is that, for low heating rates (lower than 6 K min−1), the cold crystallization leads to the most stable polymorph, that melts at Tfus = 42° C, while for high heating rates (higher than 9 K min−1), the cold crystallization tends to form the polymorph that melts at Tfus = 32° C.

78
The powder X-ray diffraction patterns of the most stable crystalline form of salol have been reported.8

79
Furthermore, it was shown that those patterns are identical for the crystal formed by annealing the metastable liquid at 300 K, and for the crystal formed from a long annealing (5 days) at 220 K (very slightly above Tg).

80
Since the latter probably arises from an homogeneous nucleation process,8 this finding, together with the results shown in Fig. 3 for the lowest heating rates, suggest that the crystallization of the most stable polymorph of salol is initiated by a relatively slow homogeneous nucleation taking place at low temperatures.

81
Note also in Fig. 3 that, for intermediate heating rates, a signal is often observed between the exothermic crystallization peak and the endothermic melting peak, as shown on the thermogram at 7 K min−1.

82
The signal, that presents an endothermic part followed by an exothermic one, corresponds to the interconversion between the two polymorphs: the lower temperature polymorph crystallized first, begins to melt at 32° C and, as it melts, the stable polymorph crystallizes to melt later at 42° C.

83
The DSC technique thus shows that the polymorph that melts at Tfus = 32° C is unstable, which makes the X-ray analysis of its crystalline structure difficult.

84
It is necessary to emphasize that the behaviour depicted in Fig. 3, and described before, is not always strictly observed.

85
As pointed out before, nucleation and crystal growth are somewhat probabilistic events, and the consequence is that the experiments are not entirely reproducible.

86
For example, we observed that, if the sample crystallizes on cooling, the crystal formed is always the stable form that melts at Tfus = 42° C.

87
Moreover, if a very tiny portion of the sample crystallizes on cooling, the nucleation and crystal growth processes, that take place in a subsequent heating ramp, will lead also to the stable form, independently of the heating rate.

88
An important requirement for the behaviour shown in Fig. 3 to be observed is that no crystallization takes place in the process of cooling from the melt that precedes the linear heating ramp.

Temperature dependence of the kinetics of cold crystallization

89
It was shown before that for low heating rates the onset of the cold crystallization is at ∼−20° C.

90
In the following we will use the TSDC technique in order to evaluate the stability of the metastable supercooled liquid phase, and to qualitatively analyse the temperature dependence of the rate of the cold crystallization.

91
The TSDC glass transition peak will be used as a probe to quantitatively detect the glassy phase present in the sample.

92
Fig. 4 shows the results of two TSDC experiments.

93
In the first one, the sample of salol was cooled quickly from the liquid (50° C) down to the polarisation temperature TP = −57° C, and the experiment was then carried out.

94
The final temperature of the linear depolarisation ramp was Tf = 38° C, and the heating rate was q = 4 K min−1.

95
The result of this experiment shows two peaks.

96
The first, which appears in the glass transition region, is a component of the glass transition relaxation.

97
The second one shows a maximum intensity at ∼0–5° C.

98
Given the big difference between the temperature location of this peak and the polarisation temperature (TP = −57° C), it is not conceivable that it arises from a dipolar polarisation activated by the electric field.

99
This peak at ∼0–5° C is a consequence of a structural modification which occurred in the sample during the constant rate heating process, and that can be ascribed to the crystallisation of salol from the supercooled liquid state.

100
Note that this peak occurs precisely in the temperature region where the cold crystallisation occurs, beginning close to a temperature of ∼−25° C, in agreement with the DSC results reported previously.

101
The second experiment, whose result is shown as a flat dashed line in Fig. 4, was performed immediately after the first: the sample was cooled quickly directly from the final temperature of the heating ramp of the first experiment (Tf = 38° C) down to the same polarisation temperature, and it was submitted to the same TS experiment.

102
The disappearance of the glass transition peak indicates that, at the end of the first experiment, the sample is no more in the glassy state, reinforcing the idea that the peak at 0–5° C was due to cold crystallization.

103
Moreover, the absence of any peak in the 0–5° C region in the second heating ramp indicates that the sample is fully crystallized.

104
The TSDC and DSC data are thus in total agreement.

105
In order to obtain more detailed information on the kinetics of this transformation, we performed a series of TSDC global experiments, where the final temperature of the heating ramp, Tf, varied from series to series.

106
Doing so, the sample was allowed to anneal, for a given period of time, at temperatures in the vicinity of Tf, where the transformation of the supercooled liquid into the crystal is more or less effective.

107
Fig. 5 shows the results of such a series of global experiments where Tf = −21° C.

108
In the first experiment of the series (higher peak in Fig. 5) the sample was cooled quickly from the liquid state (50° C) down to the polarisation temperature (TP = −57° C), in order to prepare the sample in the glassy state.

109
In all the subsequent experiments, the sample was cooled down to the polarisation temperature from the final temperature of the linear heating ramp of the previous TSDC experiment (Tf = −21° C, as reported before, for the experiments shown in Fig. 5).

110
The intensity (or the area) of the glass transition peak probes the quantity of glassy phase present in the sample, so that the decrease of the intensity of the TS peaks observed in Fig. 5 from experiment to experiment is the signature of the progressive crystallisation of salol.

111
Fig. 6 shows how the intensity of the TSDC peaks evolves from experiment to experiment, in a different series of experiments with different Tf.

112
It can be concluded from Fig. 6 that the crystallisation from the supercooled liquid is slow at −25° C (in perfect agreement with the DSC results), and that the rate of crystallisation increases with increasing temperature, so that it becomes particularly fast for temperatures higher than −17° C.

113
The results reported in this section illustrate how the TSDC technique can be used in order to detect phase transformations, and to follow their time evolution as a function of temperature, thus providing useful information regarding the respective phase diagrams and the kinetics of phase transformations.

The β-relaxation of salol

114
According to the so-called heterogeneous model, a glass and a supercooled liquid are viewed as composed of structured clusters that sink in non-structured or amorphous regions.

115
In the context of this model, the α-relaxation is considered to originate from collective motions of the molecules within the structured clusters, while the β-process would arise from local motions in the gap between clusters.

116
Since the β-relaxation time is much shorter than the α-relaxation time near the glass transition temperature, the β-relaxation process is believed to assist the nucleation process.11

117
One of the techniques most widely used to study the mobility associated with the α and β-processes is Dielectric Relaxation Spectroscopy.

118
However, the use of this experimental technique for the analysis and characterization of the β-process is difficult in the case of salol, given that this process is disguised under the main α-relaxation, and appears as a so-called excess wing in the high frequency side of the dielectric α-peak.23

119
We recently showed that, in such cases, TSDC appears as an excellent technique to resolve both α and β-relaxational processes, and to enable a clear identification and characterization of the β-rearrangement.24

120
Fig. 7 shows the result of a series of TSDC experiments on the secondary β-relaxation of salol.

121
The peaks in the right-hand side, showing a strongly increasing intensity, indicate the coalescence between α and β motional processes, while the other peaks correspond to molecular motions of the β-process.

122
Fig. 8 shows the representation of the activation enthalpy, ΔH, associated with these TSDC peaks as a function of the respective temperature location (temperature of maximum intensity, Tm).

123
The line shown in Fig. 8 is the so-called zero entropy line.

124
It can be seen that a small number of points in the higher temperature side clearly deviate from the zero entropy line: they correspond to the higher intensity peaks in Fig. 7 and indicate the mergence of the α and β-motions.

125
On the other hand, most of the points in the low temperature side show very small deviations from the zero entropy line, indicating that they correspond to localized and non-cooperative molecular motions.

126
From the data included in Figs. 7 and 8 we can conclude that the TSDC technique provides very clear results on the β-relaxation of salol, allowing a very precise characterisation of the molecular mobility associated with it.

127
The β-relaxation shows a narrow distribution in activation energy with a mean value around Ea = 46 kJ mol−1.

128
The dynamics of this motional process is such that the mean characteristic time at the glass transition temperature is τβ(Tg) ≅ 1.6 × 10−2 s.

129
We want to note that this β-mobility of salol is not present in the TSDC spectrum of the crystallized sample, suggesting that this mobility is a feature of the glassy state.

130
Let us finally note that the results in Fig. 7 were obtained using standard vitrification conditions (cooling rates not higher than −5 to −20 K min−1), without need for fast quenching of the liquid25.

Summary and conclusions

131
We have investigated phenyl-2-hydroxybenzoate (salol) from the point of view of the molecular mobility, in relation with the problem of phase transformations.

132
The study of the glass transition by DSC (heating rate dependence of the onset temperature of the glass transition signal) provided the value of 252 kJ mol−1 for the activation energy of the structural relaxation, leading to the value of m = 60 for the fragility index of salol, that confirms the fragile character of the compound.

133
The TSDC results, obtained with the same sample batch used for DSC experiments, lead to similar values, suggesting that those previously reported in the literature are somewhat overestimated.

134
Two polymorphs of salol were identified.

135
One of them has a melting temperature at (Tfus)max = ∼32° C and a melting enthalpy of approximately ΔfusH = ∼16.5 kJ mol−1.

136
The other polymorph, the most stable, melts at (Tfus)max = ∼43° C and shows a melting enthalpy of ΔfusH = 18.6 ± 0.2 kJ mol−1.

137
It was demonstrated that the less stable polymorph can be obtained by cold crystallization from a fully amorphous sample in a linear heating ramp with rate higher than ∼9 K min−1.

138
It was shown how the cold crystallization process can be monitored using the dielectric technique of TSDC.

139
The TSDC results confirmed those obtained by DSC, and provided qualitative information about the temperature dependence of the cold crystallization kinetics.

140
Finally, TSDC was used to study the β-relaxation of salol, that is believed to assist the occurrence of the nucleation process.

141
It was shown that this β-process is detected by TSDC in a very clear way, allowing a very precise characterisation of this mobility.

142
The activation energy is narrowly distributed, with a mean value around Ea = 46 kJ mol−1.

143
The mean characteristic time at the glass transition temperature was found to be of the order of τβ(Tg) ≅ 10−2 s.