Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A new calorimetric technique for phase change materials and its application to alkane-based PCMs

A new calorimetric technique for phase change materials and its application to alkane-based PCMs Mater Renew Sustain Energy (2016) 5:4 DOI 10.1007/s40243-016-0068-y ORIGINAL PAPER A new calorimetric technique for phase change materials and its application to alkane-based PCMs 1,2 2 2 1 • • • • Jan Leys Benoıˆt Duponchel Ste´phane Longuemart Christ Glorieux Jan Thoen Received: 30 October 2015 / Accepted: 21 January 2016 / Published online: 11 February 2016 The Author(s) 2016. This article is published with open access at Springerlink.com Abstract The correct determination of the phase transi- Introduction tion behaviour of phase change materials (PCMs) is para- mount for assessing their application potential. In this The increasing concern about the finiteness of the planet’s work, the merits of a novel calorimetric technique, Peltier- resources stimulates research in very diverse fields. In the element-based adiabatic scanning calorimetry (pASC), for case of energy, two major angles can be distinguished: the PCM characterisation are investigated, especially in com- development or improvement of sustainable energy sources parison with the commonly-used differential scanning and increasing the effectiveness of energy use. One calorimeter (DSC). A comparative study of two alkane- approach for the latter is the temporary storage of heat (or based PCMs, the commercial mixture RT42 and the pure cold) for later use, or to dampen or delay undesired tem- alkane tricosane (C23) with these two techniques shows perature variations. The research into the so-called phase that pASC provides data at much higher resolution than change materials (PCMs) aims to provide us with materials DSC, due to its operation in thermodynamic equilibrium. with suitable properties for these tasks. It is evident that Specifically the rate-dependence and deformation that is apart from such aspects as environmental safety, economic inherent to DSC experiments is absent in pASC. In addi- feasibility and ease of application, correct knowledge of tion, the enthalpy of the PCM is directly obtained. pASC the physical properties of the candidate materials is results easily conform to the accuracy limits that are pro- required, in particular the thermal properties [1]. posed in literature for the transition temperature and stor- PCMs as available and studied today rely on the presence age capacity of PCMs. of a temperature-driven phase transition to achieve their energy storage or release. The conversion from one phase to Keywords Phase change materials  Heat capacity and another requires that the material acquires or releases energy enthalpy  Phase transitions  Adiabatic scanning to its environment, and this can be used, for example, to calorimetry  Differential scanning calorimetry  Alkanes maintain a constant temperature of a room in a building, to name one of the prototypical applications for PCMs. But science knows a multitude of different phase tran- sitions, and the involved transition heats vary widely. For example, in liquid crystals, the transition from the nematic & Jan Leys jan.leys@fys.kuleuven.be phase (a fluid state with one-dimensional orientational order) to the isotropic liquid phase has a heat of a few J g Stephane Longuemart longuema@univ-littoral.fr associated with it [2], about 100 times smaller than the transition from ice to water (333 J g ). Thus, a careful Soft Matter and Biophysics, Departement of Physics and choice of the transition is important. In general, the larger Astronomy, KU Leuven, Celestijnenlaan 200D box 2416, the difference in order between the two phases, the larger 3001 Leuven, Belgium 2 the transition heat will be, although the intermolecular ´ ´ ´ Unite de Dynamique et Structure des Materiaux Moleculaires ´ ˆ interactions will also play an important role. As a (UDSMM), Universite du Littoral Cote d’Opale, 145 avenue M. Schumann, 59140 Dunkerque, France 123 4 Page 2 of 16 Mater Renew Sustain Energy (2016) 5:4 consequence, transitions from solid to liquid and from limited thermal conductance of the heat-flux sensor. Thus, liquid to vapour have high transition heats involved, but for the registered heat flow does not correctly relate to the practical applications, the large difference in volume properties of the sample, but rather to the properties of the between liquid and vapour is a problem (a short discussion heat-flux sensor, a situation which persists until all transi- and further references can be found in Reference [1]). tion heat has been delivered. Even in power-compensated Therefore, attention has been focussed at the solid–liquid DSCs, which actively provide heat to the sample, this transition [1, 3, 4], and to a lesser extent at solid–solid cannot be avoided, as the instantaneous demand for heat transitions [5–8]. during the transition may exceed the power of the heater In terms of material properties and application feasi- system. bility, alkanes and salt hydrates are generally considered As a consequence, a DSC can only provide a deformed the most promising candidates for PCMs, although a lot of effective heat capacity curve in the vicinity of a phase different materials have also been studied, such as eutectic transition, leading to incorrect values of the transition mixtures, fatty acids, polymers, metals, and specific high- temperature and of the spread of transition heat over the temperature materials. All of these categories have their temperature region of the phase transition. In general, a specific advantages and disadvantages, and books and DSC heating experiment will broaden the transition region, reviews give systematic overviews of these [1, 3, 4, 9–13]. overestimate the transition temperature and associate more The use of ice as a material for cold storage is an age-old heat with the high-temperature part of the transition region. and common example of PCM use [14]. In our view, while a DSC gives a useful first picture of a In this paper, we draw attention to another aspect of phase transition, high-quality data cannot be obtained with PCM research, namely the characterisation of the thermal it. This is realised by researchers in the field of PCMs [15, properties. The application potential of a PCM is deter- 16], and careful DSC use [17] and improved analysis of mined by a limited number of thermal parameters, corre- DSC data [18, 19] have been proposed. sponding to some evident questions: (1) at which But also a number of alternative methods have been put temperature will it store/release heat? (2) How much heat forward. Among the examples, we first note the possibility will it store/release? (3) How efficiently will it exchange of running a DSC in an isothermal (step) mode [1, 16], heat with its environment? The latter can be quantified in operating much like a relaxation calorimeter [20–23]. A different ways, but most commonly, thermal conductivity common method is the so-called T-history method [1, 16, is taken as the measure. In this paper, we will only be 24–26], in which a PCM sample is studied in conditions concerned with questions (1) and (2). Question (1) asks for that mimic the operational use of PCMs: its response to a the determination of the phase transition temperature, changing environment is recorded. A relatively large PCM whereas (2) relates to the transition heat, and the heat sample (for example 20 ml [16] as compared to a DSC capacity over the temperature range of application. sample of few ll) is placed inside a temperature-controlled Commonly, the transition temperature T , the transition chamber together with a reference sample. The temperature tr heat Dh and the specific heat capacity c of materials are of the chamber is then suddenly changed, and the tem- perature response of sample and reference is recorded. A determined by means of differential scanning calorimetry (DSC). This technique allows for a fast and relatively easy comparison of the two curves allows the extraction of information about T , Dh and c . Closely related to this determination of these thermal properties, but it has a tr p fundamental flaw for the characterisation of PCMs: it is method is the three-layer-calorimeter [27, 28], in which fundamentally unable to provide correct data for transitions even larger samples (typically 100 g) in hermetically sealed with large heats involved. This deficiency of DSC is caused bags are studied. by its very measurement principle. In a simplified picture In this paper, we present a calorimetric method that of a heating run on the common heat-flux DSC, the furnace provides another alternative to DSC, in the particular case of the DSC is heated up at a fixed scanning rate. The of PCMs as well as in general. The adiabatic scanning sample cell, in thermal contact with this furnace by means calorimeter allows the determination of the same parame- of a heat-flux sensor, follows the temperature evolution of ters as a DSC, combined with a direct measurement of the the furnace with a small delay, as the heat needs some time enthalpy curve h(T), but it does not suffer from the phase to flow into the sample. When a transition is reached, the transition deformation because of a fundamentally different sample will need the transition heat at once, in order to measurement principle [29–32]. In adiabatic scanning keep up with the furnace, but this is not possible due to the calorimetry (ASC), the sample is heated by a controlled, constant amount of power. The sample uses this power, either to heat itself up, or, at the appropriate temperature, to In fact, it is the thermal effusivity e that is the correct measure here, pffiffiffiffiffiffiffiffiffiffi undergo phase conversion. In the latter case, the sample since e ¼ jqc , with j the thermal conductivity, q the density and c the specific heat capacity. 123 Mater Renew Sustain Energy (2016) 5:4 Page 3 of 16 4 remains at the same temperature until the conversion is mg. The first DSC sample mass was 7.03 mg, the second complete. As such, the sample remains in equilibrium all one 7.15 mg. the time, and the results do thus not suffer from deforma- tion like in a DSC. We will provide a more detailed DSC description of the technique below. For PCMs, we highlight two main advantages of ASC, A Q1000 DSC (TA Instruments, New Castle DE, USA) especially when comparing with DSC. First, ASC provides was used for the DSC measurements. This instrument is a the researcher with a correct, undeformed, equilibrium heat heat-flux DSC using the so-called Tzero technique, which capacity, and also with direct enthalpy data. Second, the allows for a separate calibration of the heat capacity and use of a constant power mimics much better the actual thermal resistance of the sample and reference platforms of thermal behaviour of a PCM in operational conditions than the DSC, leading to an improved baseline stability and the forced temperature evolutions in a DSC. In addition, better heat capacity determination [35]. The samples were ASC can readily be performed at lower temperature rates, measured under nitrogen flow. For this work, the calibrations were performed as rec- for example at 1 K h , which is a more realistic rate in ommended by the instrument’s manuals. First, the temper- many real-life situations than the 1 or 10 K min typically ature scale was corrected on the basis of the melting points used in DSC. In this paper, we investigate the difference of cyclopentane (93.43 C), diphenyl ether (26.86 C) and between ASC and DSC for the measurement of a com- indium (156.60 C) as determined at 10 K min . Diphenyl mercial alkane-based PCM, as well as of a pure alkane with ether was specifically used for these experiments to have a a comparable melting point. calibration point inside the temperature region of interest in this work, 20 to 70 C. The Tzero calibration (involving measurement of the empty DSC and of (nearly) identical Experimental details sapphire disks) was made at 20 K min , determining the RT42 Samples heat capacity and thermal resistance of the sample and ref- erence platforms. The thermal resistance of the sample pans The PCM RT42 was obtained from Rubitherm Technolo- was calibrated based on the melting curve of indium. For the gies GmbH, Berlin, Germany. Although not specified on calibration of the heat capacity, a 12 mg piece of sapphire the current data sheet [33], older versions mention that this was placed in a sample pan and measured at 10 K min . material is a mixture of alkanes. It is described as a solid– The ratio between the obtained and literature value was used liquid PCM with a melting region of 38–43 C, and it to calculate c . stores 174 kJ kg between 35 and 50 C. We found that obtaining a correct value for c from the Three samples were prepared. For the ASC measure- DSC measurements is a far from trivial task. These cali- ments, a sample of 80.0 mg was placed inside a Mettler– brations are in principle sufficient for runs at 10 K min . Toledo 120 ll stainless steel Medium Pressure Crucible. Accurate determination of c at different rates would Two samples, of 8.41 and 8.95 mg, in TA Standard Alu- require these to be redone at each of these rates, a proce- minum Hermetic cells were used for the DSC dure which is very time-consuming, particularly for slow measurement. rates. Therefore, instead of performing the calibration extensively at all rates, a carefully conducted heating run at C23 Samples 10 K min was taken as a reference run, and the c values at other rates away from the transitions were additively C23 is a shorthand notation for the linear alkane tricosane, shifted until they matched with these reference data away C H . The high-purity C23 used in this work was 23 48 from the transitions. Since a DSC uses the ratio of the heat obtained from Sigma-Aldrich: tricosane analytical standard flow and the temperature rate for the calculation of c , this (product number 91,447), and the CoA mentioned a purity additive shift corresponds to correcting the level of zero of 99.8 % as obtained from gas chromatography. Literature heat flow during the experiment (an experimental correc- data indicate that this compound has multiple phase tran- tion in the DSC software that was performed for the ref- sitions, the two most important ones are the crystal to erence runs, but not for the other ones). Since the detected rotator phase transition at about 42.5 C and the rotator to heat flow (as calibrated by the Tzero method) is not altered liquid transition at 47.3 C; these transitions involve 66.9 in this calculation, transition enthalpy values as calculated and 163.7 J g of transition enthalpy, respectively [34]. directly from integration of the heat flow or from integra- Three samples were prepared, using the same types of tion of c are identical. Effectively, this procedure corre- DSC cells as for RT42. The ASC sample mass was 59.8 sponds to a correction of the DSC baseline. 123 4 Page 4 of 16 Mater Renew Sustain Energy (2016) 5:4 ASC P (t) T (t) Adiabatic scanning calorimetry [30] obtains the heat capacity C and enthalpy H of the sample by applying a tr constant power P to the sample, and continuously regis- tering the temperature evolution with time T(t). Then, C and H are calculated on the basis of the following equa- t t t t t t _ 0 i f 0 i f tions, where the temperature rate T  dT=dt is introduced: C ðTÞ¼ ; p ð1Þ TðTÞ T (T ) 0 0 HðTÞ¼ H þ C ðT ÞdT ð2Þ 0 p H − H = P dt dt 0 0 HðTÞ¼ H þ P dT ð3Þ dT = P (t − t ) T 0 tðTÞ T T 0 tr HðTÞ¼ H þ Pdt ð4Þ tðT Þ HðTÞ¼ H þ PðtðTÞ tðT ÞÞ: ð5Þ 0 0 P C = H denotes the enthalpy of the sample at the starting temperature T of the experiment. One can see that the constant power allows for a straightforward determination H(T ) C (T ) of the enthalpy of the system: a continuous curve with the same temperature resolution as for the direct T(t) data and as for the heat capacity is obtained. The data processing during the course of a typical experiment is illustrated in Fig. 1, for the case of a weakly first-order transition, a transition with a small latent heat T T T T 0 tr 0 tr and considerable pre-transitional contributions, as com- monly observed in liquid crystals [2]. Starting at a time t , ∂H a constant power P is applied to the sample, leading to an C = ∂T evolution of its temperature T(t). In the low-temperature phase, the temperature increases, but gradually slower when approaching the transition. During the phase transi- Fig. 1 An illustration of the data and data processing in an ASC experiment on a sample with a phase transition with a small latent tion between t and t , all added power is used for the phase i f heat and substantial pre-transtional contributions to the heat capacity. conversion and the temperature does not change. Once the A constant power P is applied to the sample at starting time t , this conversion is complete and the high-temperature phase is leads to a variable temperature evolution T(t) of the sample; during reached, the temperature increases again. Via numerical the phase transition at T the temperature stays constant between t tr i and t . These two quantities P and T can be used to calculate directly _ f differentiation, the rate T is calculated from T(t), display- the enthalpy H(T), or with the rate T as intermediate, the heat capacity ing the decreasing value close to the transition; the rate C ðTÞ becomes zero at the transition. Division of P and T leads to C , showing the increase in the pre-transitional wings. In pASC the enthalpy, the phase transition is visible as the jump, of which the height corresponds to the latent heat. In order to The challenge of ASC resides in the implementation. arrive at the specific values, the pre-calibrated contribution Currently, there are two possibilities. The ‘‘classical’’ ASC from the sample addenda needs to be subtracted and the implementation will not be discussed here, as these sample mass taken into account. instruments are difficult to operate and require rather large samples (typically 0.5 g or more), but they were very In the text, capital letters for C and H will refer to a total heat successful in high-resolution studies of phase transitions capacity and enthalpy, expressed in J K and J, and lower case letters [29, 30, 36]. The successor of the classical ASC is the 1 1 1 c and h to specific values, expressed in J g K and J g . 123 Shield heater T recorder cell Mater Renew Sustain Energy (2016) 5:4 Page 5 of 16 4 Adiabatic shield zero temperature gradient over the Peltier element, this is exactly the power used to heat the sample. Cooling runs are performed by imposing (by the PID control loop) a fixed temperature gradient over the Peltier element: this leads to a constant power drawn from the sample, which is mea- Sample cell sured by the Peltier, now also acting as a heat-flux sensor. Cell Cell For this work, two pASC instruments were used. RT42 thermometer heater was measured by a pASC with a water thermostat as the base heat bath, allowing an operational temperature range from 5 to 90 C. C23 was measured by a pASC using a Peltier temperature-controlled air chamber as the base heat bath, element providing a temperature range from 30 to 120 C . For improved adiabatic conditions, the internal volume of both calorimeters was kept vacuum. Comparison with other methods Constant power source pASC versus DSC Both (p)ASC and DSC implement Eq. (1) for the deter- Controller mination of the heat capacity. The key difference between the two is that in ASC, P is kept constant and the varying T Fig. 2 A sketch of the core part of a pASC. The sample cell is placed is measured, exactly opposite to DSC where T is constant on top of a platform with a thermometer and heater. The Peltier and P is measured. The consequences for the sample are element acts as a differential thermometer between the sample and the sourronding adiabatic shield. While a constant power is applied to the enormous: in a DSC, the sample is forced to follow T, even sample, the temperature of the cell is recorded; the output of the if its own thermodynamics dictate that it should remain at Peltier element is used to control the temperature difference between the same temperature to undergo phase conversion. Thus at sample and shield a phase transition, the sample is driven out of thermal and thermodynamic equilibrium, and the instrument output is a novel Peltier-element-based adiabatic scanning calorimeter mixture of the sample properties and the instrument char- (pASC), which provides greater user-friendliness and also acteristics. In contrast, ASC only provides the sample with allows smaller samples [31, 32, 37, 38]. an amount of power that it can freely use, and the At the core of the pASC, as can be seen in the schematic calorimeter follows the behaviour of the sample. The depiction in Fig. 2, there is a sample platform equipped sample is in no way forced and remains in thermodynamic with an electrical heater and a thermistor (NTC resistance equilibrium. One can state that in an ASC, the sample is in thermometer). This sample platform accommodates an control of the calorimeter. pASC allows to perform such sample cell, which is airtight to allow for operation under equilibrium experiments on samples of the order of 10 mg vacuum. The platform itself is mounted on top of a Peltier up to several g. element, which acts as a differential thermometer for the Another important consequence of the constant-power temperature difference between the sample and the sur- approach is that rate-dependence, an intrinsic feature of rounding adiabatic shield. If this difference is maintained at DSC, does not exist in ASC. Provided that the experiment zero, then all heat provided by the heater goes only to the takes place at an applied power such that there are no sample (and its addenda), and hence the power required for thermal gradients in the sample, the result is independent of Eqs. (1) and (5) is exactly known. This is achieved by the rate: upon approaching a phase transition, the increas- feeding the Peltier output into a PID control system for the ing c of the sample will lead to a decreasing rate, and adiabatic shield, resulting in a temperature stability better p than 50 lK. This performance can only be attained because regardless of the applied power, the sample will ‘‘slowly’’ pass the transition. the adiabatic shield is itself surrounded by an additional thermal shield and a heat bath, each maintained at a slightly A fundamental difference between ASC and DSC is the opposite relation between measurement speed and sensi- lower temperature; the details of these additional elements tivity. In a DSC, a lower rate leads to a lower heat flux to depend on the specific model of pASC. the sample, and thus a decreasing sensitivity [1, 39]. In an The description above is valid for heating runs: all ASC, even a small power (for example 50 lW) remains power to the sample is provided electrically, and due to the 123 4 Page 6 of 16 Mater Renew Sustain Energy (2016) 5:4 always easy to apply, so that this does not influence the environment, which depends on the changing temperature data quality. For the two techniques, slower runs generate difference between sample/reference and the environment. more data points for the same temperature range, achieving The power itself is obtained from the simultaneous cali- a higher resolution. bration experiment with the reference. As such, the main An apparent disadvantage of ASC in comparison with advantage of ASC, the sample controlling the experiment, DSC is the longer time needed for the experiments. is also present here: the heat leak provided by the tem- Whereas DSC experiments typically are performed at 10 K perature difference between sample and environment is used by the sample according to its own thermodynamics. min (although lower rates are recommended for experi- ments with PCMs [1, 16, 17, 40]), ASC experiments are The difference with ASC is the variable power. In this respect, the methods can be seen as the differential typically performed at a few K h away from the transi- equivalent of the nonadiabatic scanning calorimeter [21]. tions. But this disadvantage is offset by the higher resolu- While obviously much larger temperature steps are used tion of the ASC data and the fact that a complete for the T-history method and the power is determined by understanding of a sample from DSC experiments requires a differential approach, the essence of the methods is the in principle runs at several rates. Additionally, the required same. A notable difference between ASC and the T-his- calibration efforts for ASC are negligible compared to tory methods is the requirement of a known reference those of a DSC: once constructed and after an initial cali- material in the latter. This means that a determination of bration, an ASC does not require periodical recalibration. the properties of an unknown PCM depends on the quality Finally, we note the ability of ASC to directly measure of the set-up and literature data, but for ASC only on the the enthalpy of the sample, independent of the heat set-up. capacity. In DSC, enthalpy must be calculated as the temperature integral of the heat capacity, itself calculated Requirements for thermal storage data from the measured power and the applied (or measured) rate. Alternatively, the measured heat flow can be inte- Although from a fundamental perspective, experimental grated with respect to time. In ASC the integration is data cannot be ‘‘too good’’, the cost of experiments in absent: as the applied power is constant, the enthalpy terms of time, manpower or money puts practical limits on depends only on P and t, Eq. (5). The difference between data quality. Therefore, it is good to ask the question what ASC and DSC was discussed in Reference [41]. An the required data quality would be for PCMs, particularly extensive illustrated discussion of this issue is made in in view of the typical applications of these data. For the Reference [31], where thermal data for the melting of PCMs, these questions have been considered explicitly, and gallium as obtained by ASC and DSC are compared. have, for Germany, resulted in a quality assurance standard pASC versus T-history method published by RAL [40]. Several papers communicating aspects of the research underlying this document are The T-history method [1, 16, 24] and the largely equivalent available, of which we draw attention to some papers three-layer-calorimeter [27, 28] use rather large samples, pertaining to the determination of heat storage capacity arguing that for inhomogeneous systems typical DSC-sized [16, 42], as well as to the description of this subject in samples are not representative. These methods derive from Reference [1]. a more general approach for determining transition tem- When we look at some of the sources for the standard, peratures. In its simplest form, the material in a container is we find that, for the purpose of using these values in heated up above its phase change, and allowed to cool practical applications, the accuracy of the stored heat as a down while its temperature is monitored. When the latent function of temperature should be better than 10 % for the heat is released, the temperature stays constant. The inno- specific enthalpy and better than 0.5 K for the temperature vation of Zhang et al. was the introduction of a second [42]; in another source also 1 K is indicated for the latter container with a reference material of known thermal [16]. Most of the standard was created to assure that this properties [24]. Comparison of the temperature evolution accuracy can be achieved, mainly with, sometimes elabo- of the sample and the reference allows to estimate the heat rate, procedure descriptions to overcome the intrinsic that leaks out of the sample. Combined with the duration of problems a DSC has in measuring a PCM. the transition plateau, this allows the calculation of the With respect to the phase transition temperature the transition heat. RAL document implies that the temperature interval over From a conceptual point of view, these methods are which the stored heat is specified should be given, both for rather close to ASC. The difference is that instead of a melting and crystallisation, rather than a single tempera- controlled and predefined power like in ASC, an unknown ture; if present, the extent of supercooling should be variable power is used: the heat leak with the specified. The main document does not specify the required 123 Mater Renew Sustain Energy (2016) 5:4 Page 7 of 16 4 accuracy for the values of the temperature and stored heat 50 (a) heating (in contrast to the values of the thermal conductivity). cooling However, an accompanying document on the website of heating RAL (only available in German), gives more details on cooling these requirements. For constant-rate dynamic measure- ments (such as DSC), a ‘‘sufficiently low’’ heating and cooling rate should be used. This rate is the slower one of the following two possibilities: either the rate at which the temperatures of the inflection point in the enthalpy for repeated heating runs at this rate coincide with each other within 0.2 K and idem for repeated cooling runs, or the rate where the difference between the inflection point for a (b) heating and cooling run is less than 0.5 K. Thus, an accuracy of 0.5 K is recommended for the transition tem- perature. Further in the document, a rate of 1 K min is recommended. Similar definitions are formulated for the T- history method. For the accuracy of the enthalpy, no value 41 41.25 41.5 is specified. Results pASC 10 20 30 40 50 T / C RT42 Fig. 3 ASC runs on RT42. a Specific heat capacity c ðTÞ, showing the presence of two smaller transitions in addition to the large one at For RT42, several heating and cooling runs were per- 42 C. b Specific enthalpy h(T), normalised at 300 J g at 45 C, formed. Because the rate in an ASC run is variable, the indicating that most energy is associated with the main transition. speed of the runs is quantified by the average rate in the Inset: enlarged view on the enthalpy during the supercooling of the 1 main transition. The data for the two heating or cooling runs nearly liquid phase, this was of the order of 1.3 K h (0.022 K coincide, showing the reproducibility, while the difference between 1 1 1 min ) for heating runs and -1.2 K h (0.020 K min ) heating and cooling behaviour is small for cooling runs. A selection of representative runs is shown in Fig 3. Comparison of the heating and cooling runs shows that The first observation is that this PCM not only has the RT42, while having a strongly first-order main transition (large latent heat), does not exhibit strong supercooling of transition at 42 C after which it is named, but also two other ones, at 27 C and at 14 C. Both of these extra this 42 C transition. However, the manufacturer’s claim that there is no supercooling [33] is not entirely valid. It is transitions are relatively broad, stretching over several K. They are much less prominent than the main transition, but very clear that there is a difference in the high-temperature edge of the c ðTÞ curves. This is further supported by the the low-temperature one contains enough enthalpy to make it clearly visible in the h(T) plots in Fig. 3b, and the middle sharp peak in the cooling c ðTÞ at that side of the transi- transition can be discerned. tion. This peak indicates the temperature at which the The reproducibility of the data can easily be asserted: supercooled liquid suddenly starts to crystallise; at this the two cooling runs essentially coincide over the entire moment, it immediately releases the latent heat associated temperature range, whereas the heating runs do so every- with that part of the transition it has already passed, and the where except at the lowest transition. The latter difference temperature of the sample increases for a short time. But, is a consequence of the deviating thermal history of the while the supercooling is clearly asserted, the less than 0.1 runs combined with the transition being close to the lower K effect will have hardly any repercussions for practical temperature limit of the calorimeter. Because the transition applications. cannot be entirely completed within the instrument’s range, For the two smaller transitions, we note that there is no the initial states of the heating runs are not the same. hysteresis evident for the mid-temperature transition. For Similar effects were also observed for other heating runs the low-temperature transition, on the other hand, hys- with different starting points and thermal histories. teresis is evident, even with the different shape of the two −1 h /Jg −1 −1 c /Jg K p 4 Page 8 of 16 Mater Renew Sustain Energy (2016) 5:4 50 20 normal (a) fast Cry R R R Liq I V I II 10 20 30 40 50 (b) T / C Fig. 4 Comparison of the specific heat capacity c of RT42 obtained 1 1 _ _ in a normal (T  1.3 K h = 0.022 K min ) and a fast (T  32 liq liq 200 1 1 Kh = 0.53 K min ) ASC heating run. The transitions in the fast run as slightly shifted and broadened due to the thermal gradient in the sample heating runs taken into consideration. The peak maxima heating differ about 0.5 K, the high-temperature edges about 2–3 cooling K. 36 38 40 42 44 46 48 50 Figure 3b shows the h(T) curves. Like the c ðTÞ data, T / C they also show the reproducibility of the experiments. For the interpretation of h(T) for practical applications, one Fig. 5 ASC runs on C23. a Specific heat capacity c . b Specific must select two temperatures T and T , the difference 1 2 1 enthalpy h, normalised to 300 J g at 49 C. The data show the dh ¼ hðT Þ hðT Þ is the energy that the PCM can store 2 1 presence of six phases; the transitions all show hysteresis or supercooling. Most of the energy is associated with the Cry –R this temperature interval. The PCM has a good stability II V and R –Liq transitions II over the range from 10 to 50 C: there is no difference in dh for the four displayed runs over this total temperature range. However, the main transition at 42 C shows some names of the phases for easy referral to the different difference in the heating and cooling h(T): while having the transitions. In the heating run, below about 39 C, the same dh between 30 and 45 C, the distribution over the sample is a crystalline phase (Cry ), which is followed by a temperature range is different. narrow Cry phase. At 41 C, the alkane undergoes its first II In addition to these slow runs, also a fast ASC heating major transition, from the crystalline region to the rotator 1 1 run was made at 32 K h  0.5 K min in the liquid region. In the rotator phases, the positional order of the phase, to allow for a comparison with a slow DSC run (see crystal is retained, while the orientational is lost (the Sect. 4.3). Figure 4 shows c ðTÞ in comparison with a elongated molecules can rotate around their long molecular normal-speed ASC heating run. While all the features are axis). Different rotator phases are observed: between 41 retained, the three transition peaks are shifted to higher and 42 C, the sample is in the R phase, followed by the temperature. This is not a consequence of a failure of the R phase up to 45 C. The next one is the R phase, and I II measurement method, like in DSC, but an indication that a finally the liquid is reached at 47 C, after passing through substantial temperature gradient is generated inside the the most prominent transition in C23. sample. Comparison of heating and cooling data reveals hys- teresis for all transitions. This is clearly visible for all of C23 them, but there is some dependence on thermal history (in the first heat/cool cycle, depicted in Fig. 5, the hysteresis is Figure 5 shows that C23 has no less than five transitions nearly absent (except for Cry –Cry ), so we refer here to I II between 35 and 50 C. We refer the reader to Refer- subsequent runs). Typical values for the hysteresis, ences [34, 43–45] for a detailed description of the different expressed as the difference between the high-temperature phases of this compound. The exact nature of the phases is edges, are 1 K for Cry –Cry , 3 K for Cry –R , 0.3 K for I II II V not relevant for the discussion here, but we introduce the R –R , 0.4 K for R –R and 0.15 K for R –Liq. V I I II II −1 −1 c /Jg K −1 h /Jg −1 −1 c /Jg K Cry II Mater Renew Sustain Energy (2016) 5:4 Page 9 of 16 4 Like RT42, C23 shows a minute supercooling of its for RT42 and for C23 the results were essentially identical; Liq–R transition. The effect is a bit larger here. Earlier hence, only data for one sample are presented. II ASC measurements on a series of alkanes showed that the Liq–Rot transition hardly supercools, the authors gave an RT42 upper limit of 0.03 K [44]. Our value here is a bit higher, but this may be a consequence of samples of different size, Figure 6 shows three phase transitions over temperature geometry and purity used here, as well as a difference in range between 5 and 45 C. A small transition can be found measuring speed. For the Rot–Cry transition, the situation in the region from 8 to 18 C, visible as a broad largely is different for odd and even alkanes. While both are sus- symmetric peak. There is a clear hysteresis between heat- ceptible to supercooling, the effect is limited for odd ones, ing and cooling. The second transition is a small feature but much larger for the even ones [44], an observation that between 24 and 27 C, more of a broad bump than a clear we already earlier made from ASC data for C24 [46]. peak. The main transition is a large peak between 30 and The transition from R to Cry in the cooling run dis- 45 C. V II plays a large supercooling, terminating at 39 C with a The data show all the typical features expected from a sudden end-of-supercooling heat release. Due to the large comparison of DSC runs at different rates: higher rates lead amount of enthalpy in this transition, this heat drives the to a broadening of transition (if the results are presented sample back to more than 40 C, a temperature well within with T on the x axis), an effect which becomes especially the two-phase region of this transition, as can be seen from pronounced for transitions with large heats involved. For the substantial effective c values in the continuation of the 10 K min heating run, this is obvious: the transition this curve. In fact, while suddenly crystallising, the sample releases so much heat that it nearly converts itself back to the rotator state. −1 (a) −0.5 K min The h(T) curves in Fig. 5b clearly show that most of the −1 0.5Kmin −1 enthalpy in C23 is associated with the Cry –R and R – II V II −10 K min −1 Liq transitions, in a ratio of about 1:2. The other transitions 10 K min are hardly visible on the scale of this figure, and are thus of minimal importance for the application of C23 as a PCM. The supercooling of the R –Cry transition is also V II 20 clearly visible in h(T): a marked discontinuous step in the T axis is displayed. It is notable that after the transition is completed at about 38 C, the values of h(T) for heating and cooling coincide again. This means that no heat has gone undetected in the run, and it is a feature of ASC that −20 0 20 40 60 this is possible: the suddenly released heat at the end of the supercooling remains inside the sample cell due to the (b) adiabatic conditions, and is afterwards gradually extracted under the normal experimental conditions. DSC DSC experiments were performed at rates of 10, 2 and 0.5 K min in all cases. Before starting the scans, the samples were heated to 60 or 70 C, in the liquid phase, to make sure that the contact between the sample and the cell was optimal. Several cooling–heating combinations with the different rates were then made. For clarity of presentation, 30 32 34 36 38 40 42 44 T / C only runs at 10 and 0.5 K min are given in the figures. The results of the 2 K min runs generally show the Fig. 6 DSC runs on RT42. a Specific heat capacity c . b Detailed expected intermediate behaviour between the 10 and 0.5 K view of the main transition. The curves show the typical rate min ones. To verify the reproducibility, two nearly dependence of DSC data, as well as differences between heating and cooling runs identical samples of each material were prepared, and both −1 −1 −1 −1 c /Jg K c /Jg K p p 4 Page 10 of 16 Mater Renew Sustain Energy (2016) 5:4 seems to continue up to 46 C, 4 K above the nominal not very well visible in the DSC data. In general, the Cry – Cry transition becomes a small shoulder on the low- value for RT42. In fact, the transition peak at 10 K min II temperature side of the Cry –R transition, while it is reaches its maximum when the 0.5 K min transition peak II V more pronounced and much more clearly separated in is completely subsided. This clearly learns that PCMs ASC. The transition from R to R is difficult to assess. should be not be measured at the conventional high rates, a V I The R to R transition is present in all runs, although the conclusion shared with the authors of Reference [17]. I II exact characteristics differ from run to run. The main transition shows a distinct shape in the cooling The two main transitions, Cry –R and R –Liq, are curves. One would expect some onset behaviour, and a II V II clearly visible. In contrast to the sharp transitions in ASC, gradual increase of c , instead, a rather sudden increase is the smearing out due to the measurement concept is par- found in this case. This is combined with a double-peak ticularly clear. In this case, even reducing the rate does not feature, minute for the fast run, pronounced for the slow help very much. The DSC data for the cooling runs show runs. Similar to the ASC data, this is the signature of supercooling for both transitions and the strong end-of- supercooling the sample: the sudden heat release results supercooling jump that was already seen in the ASC data also in DSC in an anomalous temperature evolution (T is for the Cry –R transition. Even in the DSC, where the II V not a constant value). Due to the closeness of the temper- thermal contact between sample and environment is sup- ature sensor to the sample cell in the design of the Q1000 posed to be very efficient, so much heat is released that the as well as the strong thermal coupling [35], this evolution temperature of the sample rises substantially. The R –Liq II is registered in T(t). transition does not display a strong jump at the end of supercooling, although all the runs do show the spike at the C23 high-temperature side of the transition that we saw earlier in RT42 and found to be indicative of a small supercooling Two samples of C23 have been studied, and the results effect. were essentially identical. Figure 7 shows the results for scans at their full scale, indicating the notable difference in height of the transition peaks in comparison with RT42. Discussion The figure shows that much higher c values are reached with the transition peaks than for RT42, indicating that the Stored heat transition enthalpy is concentrated in a smaller temperature region in C23. The other common observations about DSC Generally the transition heat Dh of a phase transition is measurements, rate dependence and transition broadening, reported, which is the heat needed for the phase conversion are also clearly present, to a larger extent than in RT42. only. In terms of DSC analysis, it is the integral of the heat Due to the large heats involved and the closeness of the flux above the transition’s baseline. We have compared Dh transitions, there is considerable overlap between the values from our experiments with those available in the broadened transition peaks for the 10 K min data. literature for RT42 [47] and C23 [34, 44] and obtained Figure 8 displays several selections from the DSC data always good agreement with these values, both for ASC for C23. As a first observation, the smaller transitions are and DSC. However, in PCM applications, the transition heat is not −1 the most relevant way of describing the properties: it only −0.5Kmin −1 takes into account the heat that is associated with the phase 0.5Kmin −1 −10Kmin conversion, and not the heat that is needed for the heating 300 −1 10Kmin of the sample itself. For PCMs both of these contribute to the storage capacity, and therefore the total energy content over a temperature interval, the stored heat Q , is the stored relevant quantity. This is the change in enthalpy over the 100 temperature interval. The stored heat was calculated in two different ways, depending on the measurement technique. The enthalpy h(T) is a direct result of an ASC experiment, see Eq. (5), 20 30 40 50 60 and the calculation of the stored heat is trivial: the differ- T / C ence in enthalpy at two temperatures is the stored heat for that temperature interval. For DSC, separate experiments Fig. 7 Overview of the DSC runs for C23, with c data presented at (also used for the absolute value determination of c ) were full scale for comparison with the RT42 data −1 −1 c /Jg K p Mater Renew Sustain Energy (2016) 5:4 Page 11 of 16 4 Fig. 8 Detailed views on the (a) (b) −1 −1 DSC runs for C23. a Cooling −0.5 K min 0.5 K min runs in the transition region. −1 −1 −10Kmin 10 K min b Heating runs in the transition region. c 10 K min runs, heating and cooling. d 0.5 K min runs, heating and cooling (c) (d) 30 35 40 45 50 30 35 40 45 50 55 ◦ ◦ T / C T / C performed in which the instrument was allowed to stabilise quite good, provided one compares the same temperature extensively before starting the run, the same was done regions. When only the region directly surrounding the afterwards so that the zero-level of the heat flow was transition peak is taken into account, ASC gives smaller accurately known. In order to further reduce potential values, because of the smaller contribution of the sensible errors, only data from 10 K min heating runs will be heat over such smaller interval. Thus, even though most of the stored heat is contained in the transitions, a consider- discussed, as higher rates result in larger heat flows. For these runs, the measured contribution from the spurious able amount is still present in the form of sensible heat, a consequence of the rather high heat capacity in the rotator heat flow was less than 1 % compared to the heat flow phases (see Sect. 4.2). away from the transitions, and evidently much less than Mehling and Cabeza have proposed the stored heat over that during the transitions. We have not integrated c ðTÞ to a series of small temperature intervals (as opposed to a obtain h(T), because calculating c ðTÞ requires an extra single Q value for the entire transition region) as a calibration step and thus introduces an extra error; instead, stored convenient representation of the storage capacity of a PCM the heat flow was directly integrated. It should be noted in a given temperature region [1]. In terms of ASC data, that DSC is here at a double disadvantage compared to this is equivalent to reporting the difference of the enthalpy ASC: there is not only the deformation of the heat flow/ at the end and the beginning of each such interval. For heat capacity, but also, accurate measurements of the DSC, the same calculation is made on the basis of the absolute value of the heat flow must be made at high rates, integrated heat flow. The results of these calculations are where the deformation is worse. The results are sum- presented in Figs. 9 and 10. marised in Table 1. For RT42, we limit the view to the main transition, but For RT42, DSC seems to overestimate Q a bit with stored the observations for the two other transitions are exactly respect to ASC. On the other hand, the value given by the the same. Away from the transitions, the values for the supplier for the Rot–Liq transition is almost the same as the stored heat are more or less the same for both techniques. ASC value. As this value is obtained by a three-layer- There is an important difference for the phase transition: calorimeter and not a DSC, it acts as an independent result where ASC shows all heat concentrated between 37 and 43 here. Likely, the heat-flux calibration during the DSC runs C, the DSC makes the transition double as wide, with was slightly off. The correspondence for the stored heat of most of the heat present in the upper half of the transition. C23 for the large transitions between ASC and DSC is −1 −1 −1 −1 c /Jg K c /Jg K p p 4 Page 12 of 16 Mater Renew Sustain Energy (2016) 5:4 Table 1 Stored heats for selected phase transitions in RT42 and C23 ASC DSC Transition Source Tstart Tstop Q stored ð C) ( C) (J g ) RT42 Cry–Rot ASC 7 22 60.8 DSC 7 22 65.8 Rot–Rot ASC 24 32 29.4 DSC 24 32 33.7 Rot–Liq ASC 35 50 172.8 DSC 35 50 184.8 [33] 35 50 174 35 40 45 50 55 C23 T / C Cry –R ASC 40 41 70.3 II V Fig. 10 C23 stored heat in 0.5 K intervals. The ASC run was stopped 40 45 94.6 at 52 C. The DSC stored heat does not coincide with the ASC results DSC 40.5 45.5 94.1 for either of the main transitions RII–Liq ASC 46.5 47.5 167.1 46.5 51 175.5 the position of the transitions, a second observation is that 46.5 60 197.1 DSC 47 60 195.8 the DSC width of the transition is seriously increased. Where the effect is reasonably limited for the lower tran- DSC determinations have been made only for 10 K min heating sition, the higher transition is substantially widened: in runs, and with respect to a common linear background for all tran- ASC, the transition is limited to a single 0.5 K interval, sitions at once. Due to the closeness of the transitions, separate backgrounds could not be obtained whereas it takes six or seven of these in DSC. The data from 51 to 60 C were obtained through extrapolation These data sets illustrate clearly that if a transition is sharp in reality (as observed by ASC), DSC fails to cor- rectly obtain the transition temperature and profile. This is ASC a consequence of the fact that in the DSC more heat should DSC be provided at the same time interval for a sharp peak, which is made impossible by its construction, while for ASC, the transition takes just more time, but the temper- ature of the sample does not change. Solid heat capacity As mentioned by Charvat et al., several sources disagree on the value of the heat capacity of RT42 at either side of the main transition [48]. Our earlier ASC work gave 3.5 J g 30 35 40 45 50 1 1 1 K for the low-temperature phase and 2.3 J g K for T / C the liquid phase [47]; these values are confirmed by the Fig. 9 RT42 stored heat in 0.5 K intervals. The distribution of the present results. The data sheet that Rubitherm currently stored heat is wider in the DSC results 1 1 provides [33] mentions 2 J g K for both liquid and solid phase, but an earlier version of it mentioned 1.8 and 1 1 2.4 J g K for solid and liquid respectively; these are In fact, the DSC data indicate that most of the storage the literature values that appear in Reference [47] and capacity is present in a temperature region where the ASC probably also in Reference [10]. data say that there is no phase-change-enhanced storage Before discussing this issue in RT42, we will first dis- any more. cuss the situation for C23: RT42 is a mixture of alkanes, For C23, all transitions except the Cry –R and R –Liq II V II and therefore one should first review the situation in a pure are invisible, as they are overshadowed by these two main compound. As seen above, C23 has additional phases transitions. In comparison with RT42, the separation between the crystal and the liquid. The presence of these between the ASC and DSC is even worse: there is no so-called rotator phases is a common phenomenon for overlap between the transitions. Apart from the disparity in −1 Q /Jg stored −1 Q /Jg stored Mater Renew Sustain Energy (2016) 5:4 Page 13 of 16 4 alkanes of intermediate chain length n [34, 45, 49]. In (a) −1 particular the region between n ¼ 20 and n ¼ 30 has ASC, ≈ 22 mK min −1 received considerable attention, and detailed x-ray and ASC, ≈ 0.5 K min calorimetric data have clarified the nature of the five rotator −1 DSC, 10 K min phases [43, 44]. Although most attention has been paid to −1 DSC, 0.5 K min the properties of the phase transitions, the thermal data from different calorimetric techniques show that the heat capacity in the rotator phases is higher than that of the (normal) solid and liquid phases [44, 46, 50, 51]. This can be understood from the additional degrees of freedom present in the rotator crystals. The present results for C23 fall in line with the earlier observations: instead of values 1 1 around 2 J g K commonly seen for crystals and liquids, 1 1 values above 5 J g K are observed for most of the 500 (b) 1 1 −1 rotator region. A value of about 2.3 J g K is found in ASC, 20 mK min −1 the crystal phase. DSC, −10 K min −1 Our current measurements and the fact that RT42 is an DSC, −0.5 K min alkane mixture now sheds some light on the differing c values. The rotator phases of the pure alkanes persist in mixtures of alkanes, at least provided the differences in chain length of the components are not too different; also, the temperature range in which the rotator phases exist is broadened [34, 52, 53]. Combining this information with our measurements, we conclude that RT42 displays a normal crystalline phase below 10 C, a first rotator phase between 16 and 26 C, and a second rotator phase between 29 and 36 C, finally followed by the liquid phase above 42 10 20 30 40 50 C. Calorimetric data do not allow the identification of the T / C phases, but a best guess based on the phase diagram of the pure alkanes with similar melting points suggests that the Fig. 11 Comparison of RT42 c as measured by ASC and DSC. lower rotator is the R phase and the upper one the R a Heating runs. b Cooling runs I II phase. It is likely that similar conclusions can be drawn for many other Rubitherm RT PCMs, and more generally for For the RT42 heating runs, the two ASC runs differ by a any alkane-based PCM. factor 20 in nominal rate, and so do the DSC runs. The The assertion that the phase below the main transition in ASC data undergo some marginal broadening, whereas the RT42 is not a ‘‘normal’’ crystal, in combination with the shape and width of the transition remain unaffected. For relative proximity of the phase transition, explains the high the same change of rate, the DSC data show a peak that is value of c that we observed earlier [47], a value that is more than double as wide and completely deformed. This essentially confirmed in the present work. The values illustrates the advantage of the constant power/variable rate reported in the other sources should be treated as more approach in ASC: irrespective of the amount of power approximate values, suitable for approximate PCM appli- applied, the calorimeter will slow down for a transition, cation calculations, but, in fact, outside the uncertainty allowing the phase conversion to take place in equilibrium. limits that have been put forward, which demand an The same broadening is also apparent in Fig. 12 for the uncertainty of less than 10 % [16]. C23 DSC data, but, due to the intrinsically sharper transi- tions, the effects are even worse. Where for RT42 the 0.5 K ASC versus DSC min DSC data are comparable to the ASC data, in C23 this is not the case. The R -Liq transition is in heating II Figures 11 and 12 make a comparison between ASC and more than two times as broad in the 0.5 K min run than DSC data for the two PCMs. As already noted in the dis- in the ASC data, while it does not fit within the chosen cussion of the measurement results, the shapes of the c temperature axis for the 10 K min run. The effect is more curves are quite different, illustrating the opposite mea- pronounced for the Cry –R transition, because it is even II V surement principles. sharper in the ASC runs. −1 −1 −1 −1 c /Jg K c /Jg K p p 4 Page 14 of 16 Mater Renew Sustain Energy (2016) 5:4 −1 −1 indicate such ranges as 0.1–100 K min . For the particular ASC,≈≈ 9.2 mK min −6.5mKmin −1 −1 case of PCMs, where the interest lies with large transitions, DSC, 10 K min −10 K min −1 −1 the higher rates of DSCs, combined with the measurement DSC, 0.5 K min −0.5 K min principle, cause problems because the results do not cor- (a) rectly represent the h(T) behaviour that a PCM will exhibit in typical applications. As an example, we point to a study where DSC data at different rates were inserted in a model for an application and compared with actual measurements, it was found that the data from the slowest runs (0.1 K min ) gave the best correspondence [54]. The authors also note that better data are necessary, which can be provided by ASC, not by using lower rates as in DSC, but by a measurement principle that is closer to the actual applica- tion conditions. Typical PCM applications consist of placing the PCM in an environment from which they need to extract heat (or 300 give to). In a DSC, due to the forced temperature rate, the (b) temperature gradient between the sample and the furnace will increase while passing the transition, a situation which is more or less the opposite of, for example, a solid PCM in a wall board of a warm room: the PCM melting will extract heat from the room, reducing the temperature gradient. In that respect, an ASC (with constant power) or the T-history method (with variable power) provide experimental cir- cumstances much closer to real-life situations. We also recorded approximately the time that was needed for the experiments. When referring to the pure measurement time, the time that the sample of interest is inside the instrument, DSC is clearly the faster method. A 0 measurement cycle over 60 K, with heating and cooling 30 35 40 45 50 55 runs at the three rates (0.5, 2 and 10 K min ) takes about 6 T / C h, including temperature stabilisations, thus amounting to a single working day. However, this turns out to be far from Fig. 12 Comparison of C23 c as measured by ASC and DSC. the complete picture: the execution of this study has shown a Heating runs. b Cooling runs that the careful operation of a DSC, with the attention needed to the calibration of the heat flow and heat capacity Another important issue is the sensitivity trade-off that is inherent to DSC. High rates are required for large heat at low scanning rates consumes an enormous amount of manpower and time. Due to the rate dependence of the flows, however, they enhance the broadening and shift of the transitions. For the wide and separated transitions in DSC, these calibrations need to be performed at all mea- surement rates, and as there is a need for periodic recali- RT42, this is not too problematic, but for C23, with its five bration, this is a serious burden on the DSC operator. The phase transitions over a 10 K range, including two with a large transition heat, it becomes dramatic. In the 10 K full supplier-suggested calibration at 10 K min requires about one working day, and this time needs to be multi- min experiments, only the two large transitions can be considered unambiguously present. Even in the 0.5 K plied with the number of rates needed (and more time is needed for lower rates). And ideally, such calibration min runs, the detection of the small transitions remains should be regularly repeated. Alternatively, in case c is problematic, with Cry –Cry and R –R clearly present, p I II I II determined with an approach in which three curves (empty but with noticeably different features than in the ASC pans, reference material, sample) are measured [39], then experiments. R –R is not detected. V I (supposing a 60 K range, heating and cooling runs ? 30 The dynamic range of ASC does not overlap much with min for mounting, stabilisation,...) about three times 45 that of a DSC. Current implementations of pASC cover min, 2.25 h is needed for 10 K min and three times 2.5 h, rates (away from transitions) from 0.1 to 60 K h (0.001 K 1 1 about a day for 0.5 K min (and because the operator may min to 1 K min ), while commercial DSCs usually −1 −1 −1 −1 c /Jg K c /Jg K p p Mater Renew Sustain Energy (2016) 5:4 Page 15 of 16 4 need to be present to exchange the samples, this may need thermodynamic equilibrium requires the use of slow to be done over multiple working days). Thus, although the scanning rates, for which DSC is intrinsically unsuitable, in direct measurement time is short, the full measurement contrast to ASC which is made for it. This makes the time including calibrations becomes substantial. advantage of ASC with respect to DSC clear for any From our experience in this work, a ‘‘pre-calibrated thermal measurement with higher demands than fast pASC’’ (factory-calibrated as would be the case if the screening. instrument was commercially available) gives a finalised Acknowledgments The authors thank Patricia Losada-Pe´rez for her result like presented in Fig. 3 in about 1 h of operator time assistance with the C23 pASC measurements. This research was for the sample mounting, about 100 h of measuring time supported by the Research Council of KU Leuven through IOF (requiring no user intervention) and about 15 min for Leverage Project IOF-HB/11/022 and Research Project OT/11/064. analysis. J.L. thanks ULCO for a post-doctoral fellowship. While a DSC may objectively still be the faster instru- Open Access This article is distributed under the terms of the ment, the difference is not that large once the complete Creative Commons Attribution 4.0 International License (http:// picture is taken into account. In addition, this work shows, creativecommons.org/licenses/by/4.0/), which permits unrestricted in agreement with literature, that correct thermal data on use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a PCMs should be obtained in equilibrium, which can only link to the Creative Commons license, and indicate if changes were be achieved at the lower rates at which a pASC naturally made. operates. The quality of these results further motivates the use of a somewhat slower method. References Conclusion 1. Mehling, H., Cabeza, L.F.: Heat and cold storage with PCM—an up to date introduction into basics and applications. Springer, We believe that for accurate determination of heat capacity Berlin, Heidelberg (2008). doi:10.1007/978-3-540-68557-9 profiles, and the derived enthalpy profiles and stored heat 2. Van Roie, B., Leys, J., Denolf, K., Glorieux, C., Pitsi, G., Thoen, J.: Phys. Rev. E 72(4), 041702 (2005). doi:10.1103/PhysRevE.72. per temperature interval tables, DSCs will need to be operated at rates below 0.5 K min . This is a regime 3. Zhou, D., Zhao, C.Y., Tian, Y.: Appl. Energy 92, 593 (2012). where DSCs are generally not readily used, and due to their doi:10.1016/j.apenergy.2011.08.025 construction, also do not perform particularly well. The 4. Hyun, D.C., Levinson, N.S., Jeong, U., Xia, Y.: Angew. Chem. Int. Ed. 53(15), 3780 (2014). doi:10.1002/anie.201305201 effort needed in terms of calibration, combined with the 5. Busico, V., Carfagna, C., Salerno, V., Vacatello, M., Fittipaldi, longer measurement times make it worthwhile to look at F.: Solar Energy 24(6), 575 (1980). doi:10.1016/0038- other measurement techniques. While the virtues of the T- 092X(80)90356-4 history method (and comparable techniques) are well- 6. Li, W., Zhang, D., Zhang, T., Wang, T., Ruan, D., Xing, D., Li, H.: Thermochim. Acta 326(1–2), 183 (1999). doi:10.1016/S0040- established for the study of PCMs, we believe that this 6031(98)00497-3 paper has also made clear that pASC is a suitable technique 7. Pielichowska, K., Pielichowski, K.: Polym. Adv. Technol. for this type of research. 22(12), 1633 (2010). doi:10.1002/pat.1651 The pASC instrument resembles in several aspects a 8. Alkan, C., Gu¨nther, E., Hiebler, S., Ensari, O.F., Kahraman, D.: Solar Energy 86(6), 1761 (2012). doi:10.1016/j.solener.2012.03. DSC, while it allows to incorporate an important improvement in PCM research that has been added by the 9. Zalba, B., Marı´n, J.M., Cabeza, L.F., Mehling, H.: Appl. Therm. T-history method: experimental circumstances that mimic Eng. 23(3), 251 (2003). doi:10.1016/S1359-4311(02)00192-8 the operational use of PCMs, and which are close to ther- 10. Kenisarin, M., Mahkamov, K.: Renew. Sustain. Energy Rev. 11(9), 1913 (2007). doi:10.1016/j.rser.2006.05.005 mal and thermodynamic equilibrium, at least much closer 11. Kenisarin, M.M.: Renew. Sustain. Energy Rev. 14(3), 955 (2010). than can be achieved in a DSC. As such, one can regard doi:10.1016/j.rser.2009.11.011 pASC as a T-history-like method, but operating in a much 12. Cabeza, L.F., Castell, A., Barreneche, C., de Gracia, A., Fer- tighter controlled environment. As a consequence, smaller ne´ndez, A.I.: Renew. Sustain. Energy Rev. 15(3), 1675 (2011). doi:10.1016/j.rser.2010.11.018 samples can be used, although the technique also allows 13. Oro´, E., de Gracia, A., Castell, A., Farid, M.M., Cabeza, L.F.: them to be sufficiently large to correctly represent the large Appl. Energy 99, 513 (2012). doi:10.1016/j.apenergy.2012.03. bulky application quantities. Although ASC requires longer measurements than DSC, 14. Morofsky, E.: Energy efficient building design and thermal energy storage. In: Paksoy, H.O. (ed.) Thermal energy storage for the difference is not that large due to the time needed for sustainable energy consumption. NATO science series, vol. 234, the required DSC calibration work (absent for pASC). In pp 103–132. Springer, Dordrecht (2007). doi:10.1007/978-1- addition, the determination of thermal parameters in 4020-5290-3_6 123 4 Page 16 of 16 Mater Renew Sustain Energy (2016) 5:4 15. He, B., Martin, V., Setterwall, F.: Energy 29(11), 1785 (2004). 36. Thoen, J., Cordoyiannis, G., Glorieux, C.: Liq. Cryst. 36(6–7), doi:10.1016/j.energy.2004.03.002 669 (2009). doi:10.1080/02678290902755564 16. Gunther, E., Mehling, H., Hiebler, S., Redlich, R.: Int. J. Ther- 37. Thoen, J., Leys, J., Glorieux, C.: Adiabatic scanning calorimeter. mophys. 30(4), 1257 (2009). doi:10.1007/s10765-009-0641-z Int. Appl. PCT/BE2011/000042, patent pending (2010) 17. Lazaro, A., Pen˜alosa, C., Sole´, A., Diarce, G., Haussmann, T., 38. Losada-Pe´rez, P., Mertens, N., de Medio-Vasconcelos, B., Slen- Fois, M., Zalba, B., Gshwander, S., Cabeza, L.F.: Appl. Energy ders, E., Leys, J., Peeters, M., van Grinsven, B., Gruber, J., 109, 415 (2013). doi:10.1016/j.apenergy.2012.11.045 Glorieux, C., Pfeiffer, H., Wagner, P., Thoen, J.: Adv. Condens. 18. Albright, G., Farid, M., Al-Hallaj, S.: J. Therm. Anal. Calorim. Matter Phys. 2015(4), 1 (2015). doi:10.1155/2015/479318 101(3), 1155 (2010). doi:10.1007/s10973-010-0805-x 39. Ho¨hne, G.W.H., Hemminger, W.F., Flammersheim, H.J.: Dif- 19. Gibout, S., Franquet, E., Mare´chal, W., Dumas, J.P.: Ther- ferential scanning calorimetry—an introduction for practitioners. mochim. Acta 566, 118 (2013). doi:10.1016/j.tca.2013.04.023 Springer, Berlin, Heidelberg (1996). doi:10.1007/978-3-662- 20. Ema, K., Uematsu, T., Sugata, A., Yao, H.: Jap. J. Appl. Phys. 03302-9 32(4R), 1846 (1993). doi:10.1143/JJAP.32.1846 40. Gu¨tegemeinschaft PCM e.V. RAL-GZ 896: Phase change mate- 21. Yao, H., Ema, K., Garland, C.W.: Rev. Sci. Instrum. 69(1), 172 rial. RAL Deutsches Institut fu¨rGu¨tesicherung und Kennzeich- (1998). doi:10.1063/1.1148492 nung e.V. (2013) 22. Lashley, J.C., Hundley, M., Migliori, A., Sarrao, J.L., Pagliuso, 41. Thoen, J.: Int. J. Mod. Phys. B 9(18–19), 2157 (1995). doi:10. P.G., Darling, T.W., Jaime, M., Cooley, J.C., Hults, W.L., Mor- 1142/S0217979295000860 ales, L., Thoma, D.J., Smith, J.L., Boerio-Goates, J., Woodfield, 42. Mehling, H., Ebert, H.P., Schossig, P. In: Proceedings of the 7th B.F., Stewart, G.R., Fisher, R.A., Phillips, N.E.: Cryogenics IIR Conference on phase change materials and slurries. Dinan 43(6), 369 (2003). doi:10.1016/S0011-2275(03)00092-4 (2006) 23. Quantum Design PPMS with heat capacity option. http://www. 43. Sirota, E.B., King Jr, H.E., Singer, D.M., Shao, H.H.: J. Chem. qdusa.com/products/ppms.html (2016). Accessed 7 Feb 2016 Phys. 98(7), 5809 (1993). doi:10.1063/1.464874 24. Zhang, Y., Jiang, Y., Jiang, Y.: Meas. Sci. Technol. 10(3), 201 44. Sirota, E.B., Singer, D.M.: J. Chem. Phys. 101(12), 10873 (1994). (1999). doi:10.1088/0957-0233/10/3/015 doi:10.1063/1.467837 25. Marın, J.M., Zalba, B., Cabeza, L.F., Mehling, H.: Meas. Sci. 45. Dirand, M., Bouroukba, M., Briard, A.J., Chevallier, V., Petit- Technol. 14(2), 184 (2003). doi:10.1088/0957-0233/14/2/305 jean, D., Corriou, J.P.: J. Chem. Thermodyn. 34(8), 1255 (2002). 26. Kravvaritis, E.D., Antonopoulos, K.A., Tzivanidis, C.: Meas. Sci. doi:10.1006/jcht.2002.0978 Technol. 21(4), 045103 (2010). doi:10.1088/0957-0233/21/4/ 46. Leys, J., Glorieux, C., Thoen, J.: High-resolution adiabatic 045103 scanning calorimetry applied to the phase transitions in the 27. Laube, A.: Das 3-Schicht-Kalorimeter—Ein einfaches, aber pra- n-alkanes tetradecane and tetracosane. In: Kiss, L., St.-Georges, zises Verfahren zur Bestimmung der Speicherkapazita¨t von L. (eds.) Proceedings of the 31th International Thermal Con- Latentwa¨rmespeichermaterial. http://www.waermepruefung.de ductivity Conference and 19th International Thermal Expansion (2016). Accessed 7 Feb 2016 Symposium, pp. 57–66. DEStech Publications Inc., Lancaster 28. Kenfack, F., Bauer, M.: Energy Procedia 46, 310 (2014). doi:10. (2013) 1016/j.egypro.2014.01.187 47. Losada-Pe´rez, P., Tripathi, C.S.P., Leys, J., Cordoyiannis, G., 29. Thoen, J., Marynissen, H., Van Dael, W.: Phys. Rev. A 26(5), Glorieux, C., Thoen, J.: Int. J. Thermophys. 32(5), 913 (2011). 2886 (1982). doi:10.1103/PhysRevA.26.2886 doi:10.1007/s10765-011-0984-0 30. Thoen, J.: High Resolution adiabatic scanning calorimetry and 48. Charva´t, P., Klimesˇ, L., Ostry´, M.: Energy Build. 68(part A), 488 heat capacities. In: Wilhelm, E., Letcher, T.M. (eds.) Heat (2014). doi:10.1016/j.enbuild.2013.10.011 capacities: liquids, solutions and vapours, pp. 287–306. The 49. Broadhurst, M.G.: J. Res. Natl. Inst. Stand. Technol. 66A(3), 241 Royal Society of Chemistry, London (2010). doi:10.1039/ (1962). doi:10.6028/jres.066A.024 9781847559791-00287 50. Roble`s, L., Mondieig, D., Haget, Y., Cuevas-Diarte, M.A.: J. 31. Leys, J., Losada-Pe´rez, P., Glorieux, C., Thoen, J.: J. Therm. Chim. Phys. 95(1), 92 (1998). doi:10.1051/jcp:1998111 Anal. Calorim. 177(1), 173 (2014). doi:10.1007/s10973-014- 51. Zammit, U., Marinelli, M., Mercuri, F., Paoloni, S.: J. Phys. 3654-1 Chem. B 114(24), 8134 (2010). doi:10.1021/jp102609y 32. Leys, J., Losada-Pe´rez, P., Slenders, E., Glorieux, C., Thoen, J.: 52. Ungar, G., Masˇic´, N.: J. Phys. Chem. 89(6), 1036 (1985). doi:10. Thermochim. Acta 582, 68 (2014). doi:10.1016/j.tca.2014.02.023 1021/j100252a030 33. Rubitherm Technologies GmbH. Data sheet RT42 (2013) 53. Sirota, E.B., King Jr, H.E., Shao, H.H., Singer, D.M.: J. Chem. 34. Dirand, M., Bouroukba, M., Chevallier, V., Petitjean, D., Behar, Phys. 99(2), 798 (1995). doi:10.1021/j100002a050 E., Ruffier-Meray, V.: J. Chem. Eng. Data 47(2), 115 (2002). 54. Arkar, C., Medved, S.: Thermochim. Acta 438(1–2), 192 (2005). doi:10.1021/je0100084 doi:10.1016/j.tca.2005.08.032 35. Danley, R.L.: Thermochim. Acta 395(1–2), 201 (2003). doi:10. 1016/S0040-6031(02)00212-5 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Materials for Renewable and Sustainable Energy Springer Journals

A new calorimetric technique for phase change materials and its application to alkane-based PCMs

Loading next page...
 
/lp/springer-journals/a-new-calorimetric-technique-for-phase-change-materials-and-its-kqG29gBG9C
Publisher
Springer Journals
Copyright
Copyright © 2016 by The Author(s)
Subject
Material Science; Materials Science, general; Renewable and Green Energy; Renewable and Green Energy
ISSN
2194-1459
eISSN
2194-1467
DOI
10.1007/s40243-016-0068-y
Publisher site
See Article on Publisher Site

Abstract

Mater Renew Sustain Energy (2016) 5:4 DOI 10.1007/s40243-016-0068-y ORIGINAL PAPER A new calorimetric technique for phase change materials and its application to alkane-based PCMs 1,2 2 2 1 • • • • Jan Leys Benoıˆt Duponchel Ste´phane Longuemart Christ Glorieux Jan Thoen Received: 30 October 2015 / Accepted: 21 January 2016 / Published online: 11 February 2016 The Author(s) 2016. This article is published with open access at Springerlink.com Abstract The correct determination of the phase transi- Introduction tion behaviour of phase change materials (PCMs) is para- mount for assessing their application potential. In this The increasing concern about the finiteness of the planet’s work, the merits of a novel calorimetric technique, Peltier- resources stimulates research in very diverse fields. In the element-based adiabatic scanning calorimetry (pASC), for case of energy, two major angles can be distinguished: the PCM characterisation are investigated, especially in com- development or improvement of sustainable energy sources parison with the commonly-used differential scanning and increasing the effectiveness of energy use. One calorimeter (DSC). A comparative study of two alkane- approach for the latter is the temporary storage of heat (or based PCMs, the commercial mixture RT42 and the pure cold) for later use, or to dampen or delay undesired tem- alkane tricosane (C23) with these two techniques shows perature variations. The research into the so-called phase that pASC provides data at much higher resolution than change materials (PCMs) aims to provide us with materials DSC, due to its operation in thermodynamic equilibrium. with suitable properties for these tasks. It is evident that Specifically the rate-dependence and deformation that is apart from such aspects as environmental safety, economic inherent to DSC experiments is absent in pASC. In addi- feasibility and ease of application, correct knowledge of tion, the enthalpy of the PCM is directly obtained. pASC the physical properties of the candidate materials is results easily conform to the accuracy limits that are pro- required, in particular the thermal properties [1]. posed in literature for the transition temperature and stor- PCMs as available and studied today rely on the presence age capacity of PCMs. of a temperature-driven phase transition to achieve their energy storage or release. The conversion from one phase to Keywords Phase change materials  Heat capacity and another requires that the material acquires or releases energy enthalpy  Phase transitions  Adiabatic scanning to its environment, and this can be used, for example, to calorimetry  Differential scanning calorimetry  Alkanes maintain a constant temperature of a room in a building, to name one of the prototypical applications for PCMs. But science knows a multitude of different phase tran- sitions, and the involved transition heats vary widely. For example, in liquid crystals, the transition from the nematic & Jan Leys jan.leys@fys.kuleuven.be phase (a fluid state with one-dimensional orientational order) to the isotropic liquid phase has a heat of a few J g Stephane Longuemart longuema@univ-littoral.fr associated with it [2], about 100 times smaller than the transition from ice to water (333 J g ). Thus, a careful Soft Matter and Biophysics, Departement of Physics and choice of the transition is important. In general, the larger Astronomy, KU Leuven, Celestijnenlaan 200D box 2416, the difference in order between the two phases, the larger 3001 Leuven, Belgium 2 the transition heat will be, although the intermolecular ´ ´ ´ Unite de Dynamique et Structure des Materiaux Moleculaires ´ ˆ interactions will also play an important role. As a (UDSMM), Universite du Littoral Cote d’Opale, 145 avenue M. Schumann, 59140 Dunkerque, France 123 4 Page 2 of 16 Mater Renew Sustain Energy (2016) 5:4 consequence, transitions from solid to liquid and from limited thermal conductance of the heat-flux sensor. Thus, liquid to vapour have high transition heats involved, but for the registered heat flow does not correctly relate to the practical applications, the large difference in volume properties of the sample, but rather to the properties of the between liquid and vapour is a problem (a short discussion heat-flux sensor, a situation which persists until all transi- and further references can be found in Reference [1]). tion heat has been delivered. Even in power-compensated Therefore, attention has been focussed at the solid–liquid DSCs, which actively provide heat to the sample, this transition [1, 3, 4], and to a lesser extent at solid–solid cannot be avoided, as the instantaneous demand for heat transitions [5–8]. during the transition may exceed the power of the heater In terms of material properties and application feasi- system. bility, alkanes and salt hydrates are generally considered As a consequence, a DSC can only provide a deformed the most promising candidates for PCMs, although a lot of effective heat capacity curve in the vicinity of a phase different materials have also been studied, such as eutectic transition, leading to incorrect values of the transition mixtures, fatty acids, polymers, metals, and specific high- temperature and of the spread of transition heat over the temperature materials. All of these categories have their temperature region of the phase transition. In general, a specific advantages and disadvantages, and books and DSC heating experiment will broaden the transition region, reviews give systematic overviews of these [1, 3, 4, 9–13]. overestimate the transition temperature and associate more The use of ice as a material for cold storage is an age-old heat with the high-temperature part of the transition region. and common example of PCM use [14]. In our view, while a DSC gives a useful first picture of a In this paper, we draw attention to another aspect of phase transition, high-quality data cannot be obtained with PCM research, namely the characterisation of the thermal it. This is realised by researchers in the field of PCMs [15, properties. The application potential of a PCM is deter- 16], and careful DSC use [17] and improved analysis of mined by a limited number of thermal parameters, corre- DSC data [18, 19] have been proposed. sponding to some evident questions: (1) at which But also a number of alternative methods have been put temperature will it store/release heat? (2) How much heat forward. Among the examples, we first note the possibility will it store/release? (3) How efficiently will it exchange of running a DSC in an isothermal (step) mode [1, 16], heat with its environment? The latter can be quantified in operating much like a relaxation calorimeter [20–23]. A different ways, but most commonly, thermal conductivity common method is the so-called T-history method [1, 16, is taken as the measure. In this paper, we will only be 24–26], in which a PCM sample is studied in conditions concerned with questions (1) and (2). Question (1) asks for that mimic the operational use of PCMs: its response to a the determination of the phase transition temperature, changing environment is recorded. A relatively large PCM whereas (2) relates to the transition heat, and the heat sample (for example 20 ml [16] as compared to a DSC capacity over the temperature range of application. sample of few ll) is placed inside a temperature-controlled Commonly, the transition temperature T , the transition chamber together with a reference sample. The temperature tr heat Dh and the specific heat capacity c of materials are of the chamber is then suddenly changed, and the tem- perature response of sample and reference is recorded. A determined by means of differential scanning calorimetry (DSC). This technique allows for a fast and relatively easy comparison of the two curves allows the extraction of information about T , Dh and c . Closely related to this determination of these thermal properties, but it has a tr p fundamental flaw for the characterisation of PCMs: it is method is the three-layer-calorimeter [27, 28], in which fundamentally unable to provide correct data for transitions even larger samples (typically 100 g) in hermetically sealed with large heats involved. This deficiency of DSC is caused bags are studied. by its very measurement principle. In a simplified picture In this paper, we present a calorimetric method that of a heating run on the common heat-flux DSC, the furnace provides another alternative to DSC, in the particular case of the DSC is heated up at a fixed scanning rate. The of PCMs as well as in general. The adiabatic scanning sample cell, in thermal contact with this furnace by means calorimeter allows the determination of the same parame- of a heat-flux sensor, follows the temperature evolution of ters as a DSC, combined with a direct measurement of the the furnace with a small delay, as the heat needs some time enthalpy curve h(T), but it does not suffer from the phase to flow into the sample. When a transition is reached, the transition deformation because of a fundamentally different sample will need the transition heat at once, in order to measurement principle [29–32]. In adiabatic scanning keep up with the furnace, but this is not possible due to the calorimetry (ASC), the sample is heated by a controlled, constant amount of power. The sample uses this power, either to heat itself up, or, at the appropriate temperature, to In fact, it is the thermal effusivity e that is the correct measure here, pffiffiffiffiffiffiffiffiffiffi undergo phase conversion. In the latter case, the sample since e ¼ jqc , with j the thermal conductivity, q the density and c the specific heat capacity. 123 Mater Renew Sustain Energy (2016) 5:4 Page 3 of 16 4 remains at the same temperature until the conversion is mg. The first DSC sample mass was 7.03 mg, the second complete. As such, the sample remains in equilibrium all one 7.15 mg. the time, and the results do thus not suffer from deforma- tion like in a DSC. We will provide a more detailed DSC description of the technique below. For PCMs, we highlight two main advantages of ASC, A Q1000 DSC (TA Instruments, New Castle DE, USA) especially when comparing with DSC. First, ASC provides was used for the DSC measurements. This instrument is a the researcher with a correct, undeformed, equilibrium heat heat-flux DSC using the so-called Tzero technique, which capacity, and also with direct enthalpy data. Second, the allows for a separate calibration of the heat capacity and use of a constant power mimics much better the actual thermal resistance of the sample and reference platforms of thermal behaviour of a PCM in operational conditions than the DSC, leading to an improved baseline stability and the forced temperature evolutions in a DSC. In addition, better heat capacity determination [35]. The samples were ASC can readily be performed at lower temperature rates, measured under nitrogen flow. For this work, the calibrations were performed as rec- for example at 1 K h , which is a more realistic rate in ommended by the instrument’s manuals. First, the temper- many real-life situations than the 1 or 10 K min typically ature scale was corrected on the basis of the melting points used in DSC. In this paper, we investigate the difference of cyclopentane (93.43 C), diphenyl ether (26.86 C) and between ASC and DSC for the measurement of a com- indium (156.60 C) as determined at 10 K min . Diphenyl mercial alkane-based PCM, as well as of a pure alkane with ether was specifically used for these experiments to have a a comparable melting point. calibration point inside the temperature region of interest in this work, 20 to 70 C. The Tzero calibration (involving measurement of the empty DSC and of (nearly) identical Experimental details sapphire disks) was made at 20 K min , determining the RT42 Samples heat capacity and thermal resistance of the sample and ref- erence platforms. The thermal resistance of the sample pans The PCM RT42 was obtained from Rubitherm Technolo- was calibrated based on the melting curve of indium. For the gies GmbH, Berlin, Germany. Although not specified on calibration of the heat capacity, a 12 mg piece of sapphire the current data sheet [33], older versions mention that this was placed in a sample pan and measured at 10 K min . material is a mixture of alkanes. It is described as a solid– The ratio between the obtained and literature value was used liquid PCM with a melting region of 38–43 C, and it to calculate c . stores 174 kJ kg between 35 and 50 C. We found that obtaining a correct value for c from the Three samples were prepared. For the ASC measure- DSC measurements is a far from trivial task. These cali- ments, a sample of 80.0 mg was placed inside a Mettler– brations are in principle sufficient for runs at 10 K min . Toledo 120 ll stainless steel Medium Pressure Crucible. Accurate determination of c at different rates would Two samples, of 8.41 and 8.95 mg, in TA Standard Alu- require these to be redone at each of these rates, a proce- minum Hermetic cells were used for the DSC dure which is very time-consuming, particularly for slow measurement. rates. Therefore, instead of performing the calibration extensively at all rates, a carefully conducted heating run at C23 Samples 10 K min was taken as a reference run, and the c values at other rates away from the transitions were additively C23 is a shorthand notation for the linear alkane tricosane, shifted until they matched with these reference data away C H . The high-purity C23 used in this work was 23 48 from the transitions. Since a DSC uses the ratio of the heat obtained from Sigma-Aldrich: tricosane analytical standard flow and the temperature rate for the calculation of c , this (product number 91,447), and the CoA mentioned a purity additive shift corresponds to correcting the level of zero of 99.8 % as obtained from gas chromatography. Literature heat flow during the experiment (an experimental correc- data indicate that this compound has multiple phase tran- tion in the DSC software that was performed for the ref- sitions, the two most important ones are the crystal to erence runs, but not for the other ones). Since the detected rotator phase transition at about 42.5 C and the rotator to heat flow (as calibrated by the Tzero method) is not altered liquid transition at 47.3 C; these transitions involve 66.9 in this calculation, transition enthalpy values as calculated and 163.7 J g of transition enthalpy, respectively [34]. directly from integration of the heat flow or from integra- Three samples were prepared, using the same types of tion of c are identical. Effectively, this procedure corre- DSC cells as for RT42. The ASC sample mass was 59.8 sponds to a correction of the DSC baseline. 123 4 Page 4 of 16 Mater Renew Sustain Energy (2016) 5:4 ASC P (t) T (t) Adiabatic scanning calorimetry [30] obtains the heat capacity C and enthalpy H of the sample by applying a tr constant power P to the sample, and continuously regis- tering the temperature evolution with time T(t). Then, C and H are calculated on the basis of the following equa- t t t t t t _ 0 i f 0 i f tions, where the temperature rate T  dT=dt is introduced: C ðTÞ¼ ; p ð1Þ TðTÞ T (T ) 0 0 HðTÞ¼ H þ C ðT ÞdT ð2Þ 0 p H − H = P dt dt 0 0 HðTÞ¼ H þ P dT ð3Þ dT = P (t − t ) T 0 tðTÞ T T 0 tr HðTÞ¼ H þ Pdt ð4Þ tðT Þ HðTÞ¼ H þ PðtðTÞ tðT ÞÞ: ð5Þ 0 0 P C = H denotes the enthalpy of the sample at the starting temperature T of the experiment. One can see that the constant power allows for a straightforward determination H(T ) C (T ) of the enthalpy of the system: a continuous curve with the same temperature resolution as for the direct T(t) data and as for the heat capacity is obtained. The data processing during the course of a typical experiment is illustrated in Fig. 1, for the case of a weakly first-order transition, a transition with a small latent heat T T T T 0 tr 0 tr and considerable pre-transitional contributions, as com- monly observed in liquid crystals [2]. Starting at a time t , ∂H a constant power P is applied to the sample, leading to an C = ∂T evolution of its temperature T(t). In the low-temperature phase, the temperature increases, but gradually slower when approaching the transition. During the phase transi- Fig. 1 An illustration of the data and data processing in an ASC experiment on a sample with a phase transition with a small latent tion between t and t , all added power is used for the phase i f heat and substantial pre-transtional contributions to the heat capacity. conversion and the temperature does not change. Once the A constant power P is applied to the sample at starting time t , this conversion is complete and the high-temperature phase is leads to a variable temperature evolution T(t) of the sample; during reached, the temperature increases again. Via numerical the phase transition at T the temperature stays constant between t tr i and t . These two quantities P and T can be used to calculate directly _ f differentiation, the rate T is calculated from T(t), display- the enthalpy H(T), or with the rate T as intermediate, the heat capacity ing the decreasing value close to the transition; the rate C ðTÞ becomes zero at the transition. Division of P and T leads to C , showing the increase in the pre-transitional wings. In pASC the enthalpy, the phase transition is visible as the jump, of which the height corresponds to the latent heat. In order to The challenge of ASC resides in the implementation. arrive at the specific values, the pre-calibrated contribution Currently, there are two possibilities. The ‘‘classical’’ ASC from the sample addenda needs to be subtracted and the implementation will not be discussed here, as these sample mass taken into account. instruments are difficult to operate and require rather large samples (typically 0.5 g or more), but they were very In the text, capital letters for C and H will refer to a total heat successful in high-resolution studies of phase transitions capacity and enthalpy, expressed in J K and J, and lower case letters [29, 30, 36]. The successor of the classical ASC is the 1 1 1 c and h to specific values, expressed in J g K and J g . 123 Shield heater T recorder cell Mater Renew Sustain Energy (2016) 5:4 Page 5 of 16 4 Adiabatic shield zero temperature gradient over the Peltier element, this is exactly the power used to heat the sample. Cooling runs are performed by imposing (by the PID control loop) a fixed temperature gradient over the Peltier element: this leads to a constant power drawn from the sample, which is mea- Sample cell sured by the Peltier, now also acting as a heat-flux sensor. Cell Cell For this work, two pASC instruments were used. RT42 thermometer heater was measured by a pASC with a water thermostat as the base heat bath, allowing an operational temperature range from 5 to 90 C. C23 was measured by a pASC using a Peltier temperature-controlled air chamber as the base heat bath, element providing a temperature range from 30 to 120 C . For improved adiabatic conditions, the internal volume of both calorimeters was kept vacuum. Comparison with other methods Constant power source pASC versus DSC Both (p)ASC and DSC implement Eq. (1) for the deter- Controller mination of the heat capacity. The key difference between the two is that in ASC, P is kept constant and the varying T Fig. 2 A sketch of the core part of a pASC. The sample cell is placed is measured, exactly opposite to DSC where T is constant on top of a platform with a thermometer and heater. The Peltier and P is measured. The consequences for the sample are element acts as a differential thermometer between the sample and the sourronding adiabatic shield. While a constant power is applied to the enormous: in a DSC, the sample is forced to follow T, even sample, the temperature of the cell is recorded; the output of the if its own thermodynamics dictate that it should remain at Peltier element is used to control the temperature difference between the same temperature to undergo phase conversion. Thus at sample and shield a phase transition, the sample is driven out of thermal and thermodynamic equilibrium, and the instrument output is a novel Peltier-element-based adiabatic scanning calorimeter mixture of the sample properties and the instrument char- (pASC), which provides greater user-friendliness and also acteristics. In contrast, ASC only provides the sample with allows smaller samples [31, 32, 37, 38]. an amount of power that it can freely use, and the At the core of the pASC, as can be seen in the schematic calorimeter follows the behaviour of the sample. The depiction in Fig. 2, there is a sample platform equipped sample is in no way forced and remains in thermodynamic with an electrical heater and a thermistor (NTC resistance equilibrium. One can state that in an ASC, the sample is in thermometer). This sample platform accommodates an control of the calorimeter. pASC allows to perform such sample cell, which is airtight to allow for operation under equilibrium experiments on samples of the order of 10 mg vacuum. The platform itself is mounted on top of a Peltier up to several g. element, which acts as a differential thermometer for the Another important consequence of the constant-power temperature difference between the sample and the sur- approach is that rate-dependence, an intrinsic feature of rounding adiabatic shield. If this difference is maintained at DSC, does not exist in ASC. Provided that the experiment zero, then all heat provided by the heater goes only to the takes place at an applied power such that there are no sample (and its addenda), and hence the power required for thermal gradients in the sample, the result is independent of Eqs. (1) and (5) is exactly known. This is achieved by the rate: upon approaching a phase transition, the increas- feeding the Peltier output into a PID control system for the ing c of the sample will lead to a decreasing rate, and adiabatic shield, resulting in a temperature stability better p than 50 lK. This performance can only be attained because regardless of the applied power, the sample will ‘‘slowly’’ pass the transition. the adiabatic shield is itself surrounded by an additional thermal shield and a heat bath, each maintained at a slightly A fundamental difference between ASC and DSC is the opposite relation between measurement speed and sensi- lower temperature; the details of these additional elements tivity. In a DSC, a lower rate leads to a lower heat flux to depend on the specific model of pASC. the sample, and thus a decreasing sensitivity [1, 39]. In an The description above is valid for heating runs: all ASC, even a small power (for example 50 lW) remains power to the sample is provided electrically, and due to the 123 4 Page 6 of 16 Mater Renew Sustain Energy (2016) 5:4 always easy to apply, so that this does not influence the environment, which depends on the changing temperature data quality. For the two techniques, slower runs generate difference between sample/reference and the environment. more data points for the same temperature range, achieving The power itself is obtained from the simultaneous cali- a higher resolution. bration experiment with the reference. As such, the main An apparent disadvantage of ASC in comparison with advantage of ASC, the sample controlling the experiment, DSC is the longer time needed for the experiments. is also present here: the heat leak provided by the tem- Whereas DSC experiments typically are performed at 10 K perature difference between sample and environment is used by the sample according to its own thermodynamics. min (although lower rates are recommended for experi- ments with PCMs [1, 16, 17, 40]), ASC experiments are The difference with ASC is the variable power. In this respect, the methods can be seen as the differential typically performed at a few K h away from the transi- equivalent of the nonadiabatic scanning calorimeter [21]. tions. But this disadvantage is offset by the higher resolu- While obviously much larger temperature steps are used tion of the ASC data and the fact that a complete for the T-history method and the power is determined by understanding of a sample from DSC experiments requires a differential approach, the essence of the methods is the in principle runs at several rates. Additionally, the required same. A notable difference between ASC and the T-his- calibration efforts for ASC are negligible compared to tory methods is the requirement of a known reference those of a DSC: once constructed and after an initial cali- material in the latter. This means that a determination of bration, an ASC does not require periodical recalibration. the properties of an unknown PCM depends on the quality Finally, we note the ability of ASC to directly measure of the set-up and literature data, but for ASC only on the the enthalpy of the sample, independent of the heat set-up. capacity. In DSC, enthalpy must be calculated as the temperature integral of the heat capacity, itself calculated Requirements for thermal storage data from the measured power and the applied (or measured) rate. Alternatively, the measured heat flow can be inte- Although from a fundamental perspective, experimental grated with respect to time. In ASC the integration is data cannot be ‘‘too good’’, the cost of experiments in absent: as the applied power is constant, the enthalpy terms of time, manpower or money puts practical limits on depends only on P and t, Eq. (5). The difference between data quality. Therefore, it is good to ask the question what ASC and DSC was discussed in Reference [41]. An the required data quality would be for PCMs, particularly extensive illustrated discussion of this issue is made in in view of the typical applications of these data. For the Reference [31], where thermal data for the melting of PCMs, these questions have been considered explicitly, and gallium as obtained by ASC and DSC are compared. have, for Germany, resulted in a quality assurance standard pASC versus T-history method published by RAL [40]. Several papers communicating aspects of the research underlying this document are The T-history method [1, 16, 24] and the largely equivalent available, of which we draw attention to some papers three-layer-calorimeter [27, 28] use rather large samples, pertaining to the determination of heat storage capacity arguing that for inhomogeneous systems typical DSC-sized [16, 42], as well as to the description of this subject in samples are not representative. These methods derive from Reference [1]. a more general approach for determining transition tem- When we look at some of the sources for the standard, peratures. In its simplest form, the material in a container is we find that, for the purpose of using these values in heated up above its phase change, and allowed to cool practical applications, the accuracy of the stored heat as a down while its temperature is monitored. When the latent function of temperature should be better than 10 % for the heat is released, the temperature stays constant. The inno- specific enthalpy and better than 0.5 K for the temperature vation of Zhang et al. was the introduction of a second [42]; in another source also 1 K is indicated for the latter container with a reference material of known thermal [16]. Most of the standard was created to assure that this properties [24]. Comparison of the temperature evolution accuracy can be achieved, mainly with, sometimes elabo- of the sample and the reference allows to estimate the heat rate, procedure descriptions to overcome the intrinsic that leaks out of the sample. Combined with the duration of problems a DSC has in measuring a PCM. the transition plateau, this allows the calculation of the With respect to the phase transition temperature the transition heat. RAL document implies that the temperature interval over From a conceptual point of view, these methods are which the stored heat is specified should be given, both for rather close to ASC. The difference is that instead of a melting and crystallisation, rather than a single tempera- controlled and predefined power like in ASC, an unknown ture; if present, the extent of supercooling should be variable power is used: the heat leak with the specified. The main document does not specify the required 123 Mater Renew Sustain Energy (2016) 5:4 Page 7 of 16 4 accuracy for the values of the temperature and stored heat 50 (a) heating (in contrast to the values of the thermal conductivity). cooling However, an accompanying document on the website of heating RAL (only available in German), gives more details on cooling these requirements. For constant-rate dynamic measure- ments (such as DSC), a ‘‘sufficiently low’’ heating and cooling rate should be used. This rate is the slower one of the following two possibilities: either the rate at which the temperatures of the inflection point in the enthalpy for repeated heating runs at this rate coincide with each other within 0.2 K and idem for repeated cooling runs, or the rate where the difference between the inflection point for a (b) heating and cooling run is less than 0.5 K. Thus, an accuracy of 0.5 K is recommended for the transition tem- perature. Further in the document, a rate of 1 K min is recommended. Similar definitions are formulated for the T- history method. For the accuracy of the enthalpy, no value 41 41.25 41.5 is specified. Results pASC 10 20 30 40 50 T / C RT42 Fig. 3 ASC runs on RT42. a Specific heat capacity c ðTÞ, showing the presence of two smaller transitions in addition to the large one at For RT42, several heating and cooling runs were per- 42 C. b Specific enthalpy h(T), normalised at 300 J g at 45 C, formed. Because the rate in an ASC run is variable, the indicating that most energy is associated with the main transition. speed of the runs is quantified by the average rate in the Inset: enlarged view on the enthalpy during the supercooling of the 1 main transition. The data for the two heating or cooling runs nearly liquid phase, this was of the order of 1.3 K h (0.022 K coincide, showing the reproducibility, while the difference between 1 1 1 min ) for heating runs and -1.2 K h (0.020 K min ) heating and cooling behaviour is small for cooling runs. A selection of representative runs is shown in Fig 3. Comparison of the heating and cooling runs shows that The first observation is that this PCM not only has the RT42, while having a strongly first-order main transition (large latent heat), does not exhibit strong supercooling of transition at 42 C after which it is named, but also two other ones, at 27 C and at 14 C. Both of these extra this 42 C transition. However, the manufacturer’s claim that there is no supercooling [33] is not entirely valid. It is transitions are relatively broad, stretching over several K. They are much less prominent than the main transition, but very clear that there is a difference in the high-temperature edge of the c ðTÞ curves. This is further supported by the the low-temperature one contains enough enthalpy to make it clearly visible in the h(T) plots in Fig. 3b, and the middle sharp peak in the cooling c ðTÞ at that side of the transi- transition can be discerned. tion. This peak indicates the temperature at which the The reproducibility of the data can easily be asserted: supercooled liquid suddenly starts to crystallise; at this the two cooling runs essentially coincide over the entire moment, it immediately releases the latent heat associated temperature range, whereas the heating runs do so every- with that part of the transition it has already passed, and the where except at the lowest transition. The latter difference temperature of the sample increases for a short time. But, is a consequence of the deviating thermal history of the while the supercooling is clearly asserted, the less than 0.1 runs combined with the transition being close to the lower K effect will have hardly any repercussions for practical temperature limit of the calorimeter. Because the transition applications. cannot be entirely completed within the instrument’s range, For the two smaller transitions, we note that there is no the initial states of the heating runs are not the same. hysteresis evident for the mid-temperature transition. For Similar effects were also observed for other heating runs the low-temperature transition, on the other hand, hys- with different starting points and thermal histories. teresis is evident, even with the different shape of the two −1 h /Jg −1 −1 c /Jg K p 4 Page 8 of 16 Mater Renew Sustain Energy (2016) 5:4 50 20 normal (a) fast Cry R R R Liq I V I II 10 20 30 40 50 (b) T / C Fig. 4 Comparison of the specific heat capacity c of RT42 obtained 1 1 _ _ in a normal (T  1.3 K h = 0.022 K min ) and a fast (T  32 liq liq 200 1 1 Kh = 0.53 K min ) ASC heating run. The transitions in the fast run as slightly shifted and broadened due to the thermal gradient in the sample heating runs taken into consideration. The peak maxima heating differ about 0.5 K, the high-temperature edges about 2–3 cooling K. 36 38 40 42 44 46 48 50 Figure 3b shows the h(T) curves. Like the c ðTÞ data, T / C they also show the reproducibility of the experiments. For the interpretation of h(T) for practical applications, one Fig. 5 ASC runs on C23. a Specific heat capacity c . b Specific must select two temperatures T and T , the difference 1 2 1 enthalpy h, normalised to 300 J g at 49 C. The data show the dh ¼ hðT Þ hðT Þ is the energy that the PCM can store 2 1 presence of six phases; the transitions all show hysteresis or supercooling. Most of the energy is associated with the Cry –R this temperature interval. The PCM has a good stability II V and R –Liq transitions II over the range from 10 to 50 C: there is no difference in dh for the four displayed runs over this total temperature range. However, the main transition at 42 C shows some names of the phases for easy referral to the different difference in the heating and cooling h(T): while having the transitions. In the heating run, below about 39 C, the same dh between 30 and 45 C, the distribution over the sample is a crystalline phase (Cry ), which is followed by a temperature range is different. narrow Cry phase. At 41 C, the alkane undergoes its first II In addition to these slow runs, also a fast ASC heating major transition, from the crystalline region to the rotator 1 1 run was made at 32 K h  0.5 K min in the liquid region. In the rotator phases, the positional order of the phase, to allow for a comparison with a slow DSC run (see crystal is retained, while the orientational is lost (the Sect. 4.3). Figure 4 shows c ðTÞ in comparison with a elongated molecules can rotate around their long molecular normal-speed ASC heating run. While all the features are axis). Different rotator phases are observed: between 41 retained, the three transition peaks are shifted to higher and 42 C, the sample is in the R phase, followed by the temperature. This is not a consequence of a failure of the R phase up to 45 C. The next one is the R phase, and I II measurement method, like in DSC, but an indication that a finally the liquid is reached at 47 C, after passing through substantial temperature gradient is generated inside the the most prominent transition in C23. sample. Comparison of heating and cooling data reveals hys- teresis for all transitions. This is clearly visible for all of C23 them, but there is some dependence on thermal history (in the first heat/cool cycle, depicted in Fig. 5, the hysteresis is Figure 5 shows that C23 has no less than five transitions nearly absent (except for Cry –Cry ), so we refer here to I II between 35 and 50 C. We refer the reader to Refer- subsequent runs). Typical values for the hysteresis, ences [34, 43–45] for a detailed description of the different expressed as the difference between the high-temperature phases of this compound. The exact nature of the phases is edges, are 1 K for Cry –Cry , 3 K for Cry –R , 0.3 K for I II II V not relevant for the discussion here, but we introduce the R –R , 0.4 K for R –R and 0.15 K for R –Liq. V I I II II −1 −1 c /Jg K −1 h /Jg −1 −1 c /Jg K Cry II Mater Renew Sustain Energy (2016) 5:4 Page 9 of 16 4 Like RT42, C23 shows a minute supercooling of its for RT42 and for C23 the results were essentially identical; Liq–R transition. The effect is a bit larger here. Earlier hence, only data for one sample are presented. II ASC measurements on a series of alkanes showed that the Liq–Rot transition hardly supercools, the authors gave an RT42 upper limit of 0.03 K [44]. Our value here is a bit higher, but this may be a consequence of samples of different size, Figure 6 shows three phase transitions over temperature geometry and purity used here, as well as a difference in range between 5 and 45 C. A small transition can be found measuring speed. For the Rot–Cry transition, the situation in the region from 8 to 18 C, visible as a broad largely is different for odd and even alkanes. While both are sus- symmetric peak. There is a clear hysteresis between heat- ceptible to supercooling, the effect is limited for odd ones, ing and cooling. The second transition is a small feature but much larger for the even ones [44], an observation that between 24 and 27 C, more of a broad bump than a clear we already earlier made from ASC data for C24 [46]. peak. The main transition is a large peak between 30 and The transition from R to Cry in the cooling run dis- 45 C. V II plays a large supercooling, terminating at 39 C with a The data show all the typical features expected from a sudden end-of-supercooling heat release. Due to the large comparison of DSC runs at different rates: higher rates lead amount of enthalpy in this transition, this heat drives the to a broadening of transition (if the results are presented sample back to more than 40 C, a temperature well within with T on the x axis), an effect which becomes especially the two-phase region of this transition, as can be seen from pronounced for transitions with large heats involved. For the substantial effective c values in the continuation of the 10 K min heating run, this is obvious: the transition this curve. In fact, while suddenly crystallising, the sample releases so much heat that it nearly converts itself back to the rotator state. −1 (a) −0.5 K min The h(T) curves in Fig. 5b clearly show that most of the −1 0.5Kmin −1 enthalpy in C23 is associated with the Cry –R and R – II V II −10 K min −1 Liq transitions, in a ratio of about 1:2. The other transitions 10 K min are hardly visible on the scale of this figure, and are thus of minimal importance for the application of C23 as a PCM. The supercooling of the R –Cry transition is also V II 20 clearly visible in h(T): a marked discontinuous step in the T axis is displayed. It is notable that after the transition is completed at about 38 C, the values of h(T) for heating and cooling coincide again. This means that no heat has gone undetected in the run, and it is a feature of ASC that −20 0 20 40 60 this is possible: the suddenly released heat at the end of the supercooling remains inside the sample cell due to the (b) adiabatic conditions, and is afterwards gradually extracted under the normal experimental conditions. DSC DSC experiments were performed at rates of 10, 2 and 0.5 K min in all cases. Before starting the scans, the samples were heated to 60 or 70 C, in the liquid phase, to make sure that the contact between the sample and the cell was optimal. Several cooling–heating combinations with the different rates were then made. For clarity of presentation, 30 32 34 36 38 40 42 44 T / C only runs at 10 and 0.5 K min are given in the figures. The results of the 2 K min runs generally show the Fig. 6 DSC runs on RT42. a Specific heat capacity c . b Detailed expected intermediate behaviour between the 10 and 0.5 K view of the main transition. The curves show the typical rate min ones. To verify the reproducibility, two nearly dependence of DSC data, as well as differences between heating and cooling runs identical samples of each material were prepared, and both −1 −1 −1 −1 c /Jg K c /Jg K p p 4 Page 10 of 16 Mater Renew Sustain Energy (2016) 5:4 seems to continue up to 46 C, 4 K above the nominal not very well visible in the DSC data. In general, the Cry – Cry transition becomes a small shoulder on the low- value for RT42. In fact, the transition peak at 10 K min II temperature side of the Cry –R transition, while it is reaches its maximum when the 0.5 K min transition peak II V more pronounced and much more clearly separated in is completely subsided. This clearly learns that PCMs ASC. The transition from R to R is difficult to assess. should be not be measured at the conventional high rates, a V I The R to R transition is present in all runs, although the conclusion shared with the authors of Reference [17]. I II exact characteristics differ from run to run. The main transition shows a distinct shape in the cooling The two main transitions, Cry –R and R –Liq, are curves. One would expect some onset behaviour, and a II V II clearly visible. In contrast to the sharp transitions in ASC, gradual increase of c , instead, a rather sudden increase is the smearing out due to the measurement concept is par- found in this case. This is combined with a double-peak ticularly clear. In this case, even reducing the rate does not feature, minute for the fast run, pronounced for the slow help very much. The DSC data for the cooling runs show runs. Similar to the ASC data, this is the signature of supercooling for both transitions and the strong end-of- supercooling the sample: the sudden heat release results supercooling jump that was already seen in the ASC data also in DSC in an anomalous temperature evolution (T is for the Cry –R transition. Even in the DSC, where the II V not a constant value). Due to the closeness of the temper- thermal contact between sample and environment is sup- ature sensor to the sample cell in the design of the Q1000 posed to be very efficient, so much heat is released that the as well as the strong thermal coupling [35], this evolution temperature of the sample rises substantially. The R –Liq II is registered in T(t). transition does not display a strong jump at the end of supercooling, although all the runs do show the spike at the C23 high-temperature side of the transition that we saw earlier in RT42 and found to be indicative of a small supercooling Two samples of C23 have been studied, and the results effect. were essentially identical. Figure 7 shows the results for scans at their full scale, indicating the notable difference in height of the transition peaks in comparison with RT42. Discussion The figure shows that much higher c values are reached with the transition peaks than for RT42, indicating that the Stored heat transition enthalpy is concentrated in a smaller temperature region in C23. The other common observations about DSC Generally the transition heat Dh of a phase transition is measurements, rate dependence and transition broadening, reported, which is the heat needed for the phase conversion are also clearly present, to a larger extent than in RT42. only. In terms of DSC analysis, it is the integral of the heat Due to the large heats involved and the closeness of the flux above the transition’s baseline. We have compared Dh transitions, there is considerable overlap between the values from our experiments with those available in the broadened transition peaks for the 10 K min data. literature for RT42 [47] and C23 [34, 44] and obtained Figure 8 displays several selections from the DSC data always good agreement with these values, both for ASC for C23. As a first observation, the smaller transitions are and DSC. However, in PCM applications, the transition heat is not −1 the most relevant way of describing the properties: it only −0.5Kmin −1 takes into account the heat that is associated with the phase 0.5Kmin −1 −10Kmin conversion, and not the heat that is needed for the heating 300 −1 10Kmin of the sample itself. For PCMs both of these contribute to the storage capacity, and therefore the total energy content over a temperature interval, the stored heat Q , is the stored relevant quantity. This is the change in enthalpy over the 100 temperature interval. The stored heat was calculated in two different ways, depending on the measurement technique. The enthalpy h(T) is a direct result of an ASC experiment, see Eq. (5), 20 30 40 50 60 and the calculation of the stored heat is trivial: the differ- T / C ence in enthalpy at two temperatures is the stored heat for that temperature interval. For DSC, separate experiments Fig. 7 Overview of the DSC runs for C23, with c data presented at (also used for the absolute value determination of c ) were full scale for comparison with the RT42 data −1 −1 c /Jg K p Mater Renew Sustain Energy (2016) 5:4 Page 11 of 16 4 Fig. 8 Detailed views on the (a) (b) −1 −1 DSC runs for C23. a Cooling −0.5 K min 0.5 K min runs in the transition region. −1 −1 −10Kmin 10 K min b Heating runs in the transition region. c 10 K min runs, heating and cooling. d 0.5 K min runs, heating and cooling (c) (d) 30 35 40 45 50 30 35 40 45 50 55 ◦ ◦ T / C T / C performed in which the instrument was allowed to stabilise quite good, provided one compares the same temperature extensively before starting the run, the same was done regions. When only the region directly surrounding the afterwards so that the zero-level of the heat flow was transition peak is taken into account, ASC gives smaller accurately known. In order to further reduce potential values, because of the smaller contribution of the sensible errors, only data from 10 K min heating runs will be heat over such smaller interval. Thus, even though most of the stored heat is contained in the transitions, a consider- discussed, as higher rates result in larger heat flows. For these runs, the measured contribution from the spurious able amount is still present in the form of sensible heat, a consequence of the rather high heat capacity in the rotator heat flow was less than 1 % compared to the heat flow phases (see Sect. 4.2). away from the transitions, and evidently much less than Mehling and Cabeza have proposed the stored heat over that during the transitions. We have not integrated c ðTÞ to a series of small temperature intervals (as opposed to a obtain h(T), because calculating c ðTÞ requires an extra single Q value for the entire transition region) as a calibration step and thus introduces an extra error; instead, stored convenient representation of the storage capacity of a PCM the heat flow was directly integrated. It should be noted in a given temperature region [1]. In terms of ASC data, that DSC is here at a double disadvantage compared to this is equivalent to reporting the difference of the enthalpy ASC: there is not only the deformation of the heat flow/ at the end and the beginning of each such interval. For heat capacity, but also, accurate measurements of the DSC, the same calculation is made on the basis of the absolute value of the heat flow must be made at high rates, integrated heat flow. The results of these calculations are where the deformation is worse. The results are sum- presented in Figs. 9 and 10. marised in Table 1. For RT42, we limit the view to the main transition, but For RT42, DSC seems to overestimate Q a bit with stored the observations for the two other transitions are exactly respect to ASC. On the other hand, the value given by the the same. Away from the transitions, the values for the supplier for the Rot–Liq transition is almost the same as the stored heat are more or less the same for both techniques. ASC value. As this value is obtained by a three-layer- There is an important difference for the phase transition: calorimeter and not a DSC, it acts as an independent result where ASC shows all heat concentrated between 37 and 43 here. Likely, the heat-flux calibration during the DSC runs C, the DSC makes the transition double as wide, with was slightly off. The correspondence for the stored heat of most of the heat present in the upper half of the transition. C23 for the large transitions between ASC and DSC is −1 −1 −1 −1 c /Jg K c /Jg K p p 4 Page 12 of 16 Mater Renew Sustain Energy (2016) 5:4 Table 1 Stored heats for selected phase transitions in RT42 and C23 ASC DSC Transition Source Tstart Tstop Q stored ð C) ( C) (J g ) RT42 Cry–Rot ASC 7 22 60.8 DSC 7 22 65.8 Rot–Rot ASC 24 32 29.4 DSC 24 32 33.7 Rot–Liq ASC 35 50 172.8 DSC 35 50 184.8 [33] 35 50 174 35 40 45 50 55 C23 T / C Cry –R ASC 40 41 70.3 II V Fig. 10 C23 stored heat in 0.5 K intervals. The ASC run was stopped 40 45 94.6 at 52 C. The DSC stored heat does not coincide with the ASC results DSC 40.5 45.5 94.1 for either of the main transitions RII–Liq ASC 46.5 47.5 167.1 46.5 51 175.5 the position of the transitions, a second observation is that 46.5 60 197.1 DSC 47 60 195.8 the DSC width of the transition is seriously increased. Where the effect is reasonably limited for the lower tran- DSC determinations have been made only for 10 K min heating sition, the higher transition is substantially widened: in runs, and with respect to a common linear background for all tran- ASC, the transition is limited to a single 0.5 K interval, sitions at once. Due to the closeness of the transitions, separate backgrounds could not be obtained whereas it takes six or seven of these in DSC. The data from 51 to 60 C were obtained through extrapolation These data sets illustrate clearly that if a transition is sharp in reality (as observed by ASC), DSC fails to cor- rectly obtain the transition temperature and profile. This is ASC a consequence of the fact that in the DSC more heat should DSC be provided at the same time interval for a sharp peak, which is made impossible by its construction, while for ASC, the transition takes just more time, but the temper- ature of the sample does not change. Solid heat capacity As mentioned by Charvat et al., several sources disagree on the value of the heat capacity of RT42 at either side of the main transition [48]. Our earlier ASC work gave 3.5 J g 30 35 40 45 50 1 1 1 K for the low-temperature phase and 2.3 J g K for T / C the liquid phase [47]; these values are confirmed by the Fig. 9 RT42 stored heat in 0.5 K intervals. The distribution of the present results. The data sheet that Rubitherm currently stored heat is wider in the DSC results 1 1 provides [33] mentions 2 J g K for both liquid and solid phase, but an earlier version of it mentioned 1.8 and 1 1 2.4 J g K for solid and liquid respectively; these are In fact, the DSC data indicate that most of the storage the literature values that appear in Reference [47] and capacity is present in a temperature region where the ASC probably also in Reference [10]. data say that there is no phase-change-enhanced storage Before discussing this issue in RT42, we will first dis- any more. cuss the situation for C23: RT42 is a mixture of alkanes, For C23, all transitions except the Cry –R and R –Liq II V II and therefore one should first review the situation in a pure are invisible, as they are overshadowed by these two main compound. As seen above, C23 has additional phases transitions. In comparison with RT42, the separation between the crystal and the liquid. The presence of these between the ASC and DSC is even worse: there is no so-called rotator phases is a common phenomenon for overlap between the transitions. Apart from the disparity in −1 Q /Jg stored −1 Q /Jg stored Mater Renew Sustain Energy (2016) 5:4 Page 13 of 16 4 alkanes of intermediate chain length n [34, 45, 49]. In (a) −1 particular the region between n ¼ 20 and n ¼ 30 has ASC, ≈ 22 mK min −1 received considerable attention, and detailed x-ray and ASC, ≈ 0.5 K min calorimetric data have clarified the nature of the five rotator −1 DSC, 10 K min phases [43, 44]. Although most attention has been paid to −1 DSC, 0.5 K min the properties of the phase transitions, the thermal data from different calorimetric techniques show that the heat capacity in the rotator phases is higher than that of the (normal) solid and liquid phases [44, 46, 50, 51]. This can be understood from the additional degrees of freedom present in the rotator crystals. The present results for C23 fall in line with the earlier observations: instead of values 1 1 around 2 J g K commonly seen for crystals and liquids, 1 1 values above 5 J g K are observed for most of the 500 (b) 1 1 −1 rotator region. A value of about 2.3 J g K is found in ASC, 20 mK min −1 the crystal phase. DSC, −10 K min −1 Our current measurements and the fact that RT42 is an DSC, −0.5 K min alkane mixture now sheds some light on the differing c values. The rotator phases of the pure alkanes persist in mixtures of alkanes, at least provided the differences in chain length of the components are not too different; also, the temperature range in which the rotator phases exist is broadened [34, 52, 53]. Combining this information with our measurements, we conclude that RT42 displays a normal crystalline phase below 10 C, a first rotator phase between 16 and 26 C, and a second rotator phase between 29 and 36 C, finally followed by the liquid phase above 42 10 20 30 40 50 C. Calorimetric data do not allow the identification of the T / C phases, but a best guess based on the phase diagram of the pure alkanes with similar melting points suggests that the Fig. 11 Comparison of RT42 c as measured by ASC and DSC. lower rotator is the R phase and the upper one the R a Heating runs. b Cooling runs I II phase. It is likely that similar conclusions can be drawn for many other Rubitherm RT PCMs, and more generally for For the RT42 heating runs, the two ASC runs differ by a any alkane-based PCM. factor 20 in nominal rate, and so do the DSC runs. The The assertion that the phase below the main transition in ASC data undergo some marginal broadening, whereas the RT42 is not a ‘‘normal’’ crystal, in combination with the shape and width of the transition remain unaffected. For relative proximity of the phase transition, explains the high the same change of rate, the DSC data show a peak that is value of c that we observed earlier [47], a value that is more than double as wide and completely deformed. This essentially confirmed in the present work. The values illustrates the advantage of the constant power/variable rate reported in the other sources should be treated as more approach in ASC: irrespective of the amount of power approximate values, suitable for approximate PCM appli- applied, the calorimeter will slow down for a transition, cation calculations, but, in fact, outside the uncertainty allowing the phase conversion to take place in equilibrium. limits that have been put forward, which demand an The same broadening is also apparent in Fig. 12 for the uncertainty of less than 10 % [16]. C23 DSC data, but, due to the intrinsically sharper transi- tions, the effects are even worse. Where for RT42 the 0.5 K ASC versus DSC min DSC data are comparable to the ASC data, in C23 this is not the case. The R -Liq transition is in heating II Figures 11 and 12 make a comparison between ASC and more than two times as broad in the 0.5 K min run than DSC data for the two PCMs. As already noted in the dis- in the ASC data, while it does not fit within the chosen cussion of the measurement results, the shapes of the c temperature axis for the 10 K min run. The effect is more curves are quite different, illustrating the opposite mea- pronounced for the Cry –R transition, because it is even II V surement principles. sharper in the ASC runs. −1 −1 −1 −1 c /Jg K c /Jg K p p 4 Page 14 of 16 Mater Renew Sustain Energy (2016) 5:4 −1 −1 indicate such ranges as 0.1–100 K min . For the particular ASC,≈≈ 9.2 mK min −6.5mKmin −1 −1 case of PCMs, where the interest lies with large transitions, DSC, 10 K min −10 K min −1 −1 the higher rates of DSCs, combined with the measurement DSC, 0.5 K min −0.5 K min principle, cause problems because the results do not cor- (a) rectly represent the h(T) behaviour that a PCM will exhibit in typical applications. As an example, we point to a study where DSC data at different rates were inserted in a model for an application and compared with actual measurements, it was found that the data from the slowest runs (0.1 K min ) gave the best correspondence [54]. The authors also note that better data are necessary, which can be provided by ASC, not by using lower rates as in DSC, but by a measurement principle that is closer to the actual applica- tion conditions. Typical PCM applications consist of placing the PCM in an environment from which they need to extract heat (or 300 give to). In a DSC, due to the forced temperature rate, the (b) temperature gradient between the sample and the furnace will increase while passing the transition, a situation which is more or less the opposite of, for example, a solid PCM in a wall board of a warm room: the PCM melting will extract heat from the room, reducing the temperature gradient. In that respect, an ASC (with constant power) or the T-history method (with variable power) provide experimental cir- cumstances much closer to real-life situations. We also recorded approximately the time that was needed for the experiments. When referring to the pure measurement time, the time that the sample of interest is inside the instrument, DSC is clearly the faster method. A 0 measurement cycle over 60 K, with heating and cooling 30 35 40 45 50 55 runs at the three rates (0.5, 2 and 10 K min ) takes about 6 T / C h, including temperature stabilisations, thus amounting to a single working day. However, this turns out to be far from Fig. 12 Comparison of C23 c as measured by ASC and DSC. the complete picture: the execution of this study has shown a Heating runs. b Cooling runs that the careful operation of a DSC, with the attention needed to the calibration of the heat flow and heat capacity Another important issue is the sensitivity trade-off that is inherent to DSC. High rates are required for large heat at low scanning rates consumes an enormous amount of manpower and time. Due to the rate dependence of the flows, however, they enhance the broadening and shift of the transitions. For the wide and separated transitions in DSC, these calibrations need to be performed at all mea- surement rates, and as there is a need for periodic recali- RT42, this is not too problematic, but for C23, with its five bration, this is a serious burden on the DSC operator. The phase transitions over a 10 K range, including two with a large transition heat, it becomes dramatic. In the 10 K full supplier-suggested calibration at 10 K min requires about one working day, and this time needs to be multi- min experiments, only the two large transitions can be considered unambiguously present. Even in the 0.5 K plied with the number of rates needed (and more time is needed for lower rates). And ideally, such calibration min runs, the detection of the small transitions remains should be regularly repeated. Alternatively, in case c is problematic, with Cry –Cry and R –R clearly present, p I II I II determined with an approach in which three curves (empty but with noticeably different features than in the ASC pans, reference material, sample) are measured [39], then experiments. R –R is not detected. V I (supposing a 60 K range, heating and cooling runs ? 30 The dynamic range of ASC does not overlap much with min for mounting, stabilisation,...) about three times 45 that of a DSC. Current implementations of pASC cover min, 2.25 h is needed for 10 K min and three times 2.5 h, rates (away from transitions) from 0.1 to 60 K h (0.001 K 1 1 about a day for 0.5 K min (and because the operator may min to 1 K min ), while commercial DSCs usually −1 −1 −1 −1 c /Jg K c /Jg K p p Mater Renew Sustain Energy (2016) 5:4 Page 15 of 16 4 need to be present to exchange the samples, this may need thermodynamic equilibrium requires the use of slow to be done over multiple working days). Thus, although the scanning rates, for which DSC is intrinsically unsuitable, in direct measurement time is short, the full measurement contrast to ASC which is made for it. This makes the time including calibrations becomes substantial. advantage of ASC with respect to DSC clear for any From our experience in this work, a ‘‘pre-calibrated thermal measurement with higher demands than fast pASC’’ (factory-calibrated as would be the case if the screening. instrument was commercially available) gives a finalised Acknowledgments The authors thank Patricia Losada-Pe´rez for her result like presented in Fig. 3 in about 1 h of operator time assistance with the C23 pASC measurements. This research was for the sample mounting, about 100 h of measuring time supported by the Research Council of KU Leuven through IOF (requiring no user intervention) and about 15 min for Leverage Project IOF-HB/11/022 and Research Project OT/11/064. analysis. J.L. thanks ULCO for a post-doctoral fellowship. While a DSC may objectively still be the faster instru- Open Access This article is distributed under the terms of the ment, the difference is not that large once the complete Creative Commons Attribution 4.0 International License (http:// picture is taken into account. In addition, this work shows, creativecommons.org/licenses/by/4.0/), which permits unrestricted in agreement with literature, that correct thermal data on use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a PCMs should be obtained in equilibrium, which can only link to the Creative Commons license, and indicate if changes were be achieved at the lower rates at which a pASC naturally made. operates. The quality of these results further motivates the use of a somewhat slower method. References Conclusion 1. Mehling, H., Cabeza, L.F.: Heat and cold storage with PCM—an up to date introduction into basics and applications. Springer, We believe that for accurate determination of heat capacity Berlin, Heidelberg (2008). doi:10.1007/978-3-540-68557-9 profiles, and the derived enthalpy profiles and stored heat 2. Van Roie, B., Leys, J., Denolf, K., Glorieux, C., Pitsi, G., Thoen, J.: Phys. Rev. E 72(4), 041702 (2005). doi:10.1103/PhysRevE.72. per temperature interval tables, DSCs will need to be operated at rates below 0.5 K min . This is a regime 3. Zhou, D., Zhao, C.Y., Tian, Y.: Appl. Energy 92, 593 (2012). where DSCs are generally not readily used, and due to their doi:10.1016/j.apenergy.2011.08.025 construction, also do not perform particularly well. The 4. Hyun, D.C., Levinson, N.S., Jeong, U., Xia, Y.: Angew. Chem. Int. Ed. 53(15), 3780 (2014). doi:10.1002/anie.201305201 effort needed in terms of calibration, combined with the 5. Busico, V., Carfagna, C., Salerno, V., Vacatello, M., Fittipaldi, longer measurement times make it worthwhile to look at F.: Solar Energy 24(6), 575 (1980). doi:10.1016/0038- other measurement techniques. While the virtues of the T- 092X(80)90356-4 history method (and comparable techniques) are well- 6. Li, W., Zhang, D., Zhang, T., Wang, T., Ruan, D., Xing, D., Li, H.: Thermochim. Acta 326(1–2), 183 (1999). doi:10.1016/S0040- established for the study of PCMs, we believe that this 6031(98)00497-3 paper has also made clear that pASC is a suitable technique 7. Pielichowska, K., Pielichowski, K.: Polym. Adv. Technol. for this type of research. 22(12), 1633 (2010). doi:10.1002/pat.1651 The pASC instrument resembles in several aspects a 8. Alkan, C., Gu¨nther, E., Hiebler, S., Ensari, O.F., Kahraman, D.: Solar Energy 86(6), 1761 (2012). doi:10.1016/j.solener.2012.03. DSC, while it allows to incorporate an important improvement in PCM research that has been added by the 9. Zalba, B., Marı´n, J.M., Cabeza, L.F., Mehling, H.: Appl. Therm. T-history method: experimental circumstances that mimic Eng. 23(3), 251 (2003). doi:10.1016/S1359-4311(02)00192-8 the operational use of PCMs, and which are close to ther- 10. Kenisarin, M., Mahkamov, K.: Renew. Sustain. Energy Rev. 11(9), 1913 (2007). doi:10.1016/j.rser.2006.05.005 mal and thermodynamic equilibrium, at least much closer 11. Kenisarin, M.M.: Renew. Sustain. Energy Rev. 14(3), 955 (2010). than can be achieved in a DSC. As such, one can regard doi:10.1016/j.rser.2009.11.011 pASC as a T-history-like method, but operating in a much 12. Cabeza, L.F., Castell, A., Barreneche, C., de Gracia, A., Fer- tighter controlled environment. As a consequence, smaller ne´ndez, A.I.: Renew. Sustain. Energy Rev. 15(3), 1675 (2011). doi:10.1016/j.rser.2010.11.018 samples can be used, although the technique also allows 13. Oro´, E., de Gracia, A., Castell, A., Farid, M.M., Cabeza, L.F.: them to be sufficiently large to correctly represent the large Appl. Energy 99, 513 (2012). doi:10.1016/j.apenergy.2012.03. bulky application quantities. Although ASC requires longer measurements than DSC, 14. Morofsky, E.: Energy efficient building design and thermal energy storage. In: Paksoy, H.O. (ed.) Thermal energy storage for the difference is not that large due to the time needed for sustainable energy consumption. NATO science series, vol. 234, the required DSC calibration work (absent for pASC). In pp 103–132. Springer, Dordrecht (2007). doi:10.1007/978-1- addition, the determination of thermal parameters in 4020-5290-3_6 123 4 Page 16 of 16 Mater Renew Sustain Energy (2016) 5:4 15. He, B., Martin, V., Setterwall, F.: Energy 29(11), 1785 (2004). 36. Thoen, J., Cordoyiannis, G., Glorieux, C.: Liq. Cryst. 36(6–7), doi:10.1016/j.energy.2004.03.002 669 (2009). doi:10.1080/02678290902755564 16. Gunther, E., Mehling, H., Hiebler, S., Redlich, R.: Int. J. Ther- 37. Thoen, J., Leys, J., Glorieux, C.: Adiabatic scanning calorimeter. mophys. 30(4), 1257 (2009). doi:10.1007/s10765-009-0641-z Int. Appl. PCT/BE2011/000042, patent pending (2010) 17. Lazaro, A., Pen˜alosa, C., Sole´, A., Diarce, G., Haussmann, T., 38. Losada-Pe´rez, P., Mertens, N., de Medio-Vasconcelos, B., Slen- Fois, M., Zalba, B., Gshwander, S., Cabeza, L.F.: Appl. Energy ders, E., Leys, J., Peeters, M., van Grinsven, B., Gruber, J., 109, 415 (2013). doi:10.1016/j.apenergy.2012.11.045 Glorieux, C., Pfeiffer, H., Wagner, P., Thoen, J.: Adv. Condens. 18. Albright, G., Farid, M., Al-Hallaj, S.: J. Therm. Anal. Calorim. Matter Phys. 2015(4), 1 (2015). doi:10.1155/2015/479318 101(3), 1155 (2010). doi:10.1007/s10973-010-0805-x 39. Ho¨hne, G.W.H., Hemminger, W.F., Flammersheim, H.J.: Dif- 19. Gibout, S., Franquet, E., Mare´chal, W., Dumas, J.P.: Ther- ferential scanning calorimetry—an introduction for practitioners. mochim. Acta 566, 118 (2013). doi:10.1016/j.tca.2013.04.023 Springer, Berlin, Heidelberg (1996). doi:10.1007/978-3-662- 20. Ema, K., Uematsu, T., Sugata, A., Yao, H.: Jap. J. Appl. Phys. 03302-9 32(4R), 1846 (1993). doi:10.1143/JJAP.32.1846 40. Gu¨tegemeinschaft PCM e.V. RAL-GZ 896: Phase change mate- 21. Yao, H., Ema, K., Garland, C.W.: Rev. Sci. Instrum. 69(1), 172 rial. RAL Deutsches Institut fu¨rGu¨tesicherung und Kennzeich- (1998). doi:10.1063/1.1148492 nung e.V. (2013) 22. Lashley, J.C., Hundley, M., Migliori, A., Sarrao, J.L., Pagliuso, 41. Thoen, J.: Int. J. Mod. Phys. B 9(18–19), 2157 (1995). doi:10. P.G., Darling, T.W., Jaime, M., Cooley, J.C., Hults, W.L., Mor- 1142/S0217979295000860 ales, L., Thoma, D.J., Smith, J.L., Boerio-Goates, J., Woodfield, 42. Mehling, H., Ebert, H.P., Schossig, P. In: Proceedings of the 7th B.F., Stewart, G.R., Fisher, R.A., Phillips, N.E.: Cryogenics IIR Conference on phase change materials and slurries. Dinan 43(6), 369 (2003). doi:10.1016/S0011-2275(03)00092-4 (2006) 23. Quantum Design PPMS with heat capacity option. http://www. 43. Sirota, E.B., King Jr, H.E., Singer, D.M., Shao, H.H.: J. Chem. qdusa.com/products/ppms.html (2016). Accessed 7 Feb 2016 Phys. 98(7), 5809 (1993). doi:10.1063/1.464874 24. Zhang, Y., Jiang, Y., Jiang, Y.: Meas. Sci. Technol. 10(3), 201 44. Sirota, E.B., Singer, D.M.: J. Chem. Phys. 101(12), 10873 (1994). (1999). doi:10.1088/0957-0233/10/3/015 doi:10.1063/1.467837 25. Marın, J.M., Zalba, B., Cabeza, L.F., Mehling, H.: Meas. Sci. 45. Dirand, M., Bouroukba, M., Briard, A.J., Chevallier, V., Petit- Technol. 14(2), 184 (2003). doi:10.1088/0957-0233/14/2/305 jean, D., Corriou, J.P.: J. Chem. Thermodyn. 34(8), 1255 (2002). 26. Kravvaritis, E.D., Antonopoulos, K.A., Tzivanidis, C.: Meas. Sci. doi:10.1006/jcht.2002.0978 Technol. 21(4), 045103 (2010). doi:10.1088/0957-0233/21/4/ 46. Leys, J., Glorieux, C., Thoen, J.: High-resolution adiabatic 045103 scanning calorimetry applied to the phase transitions in the 27. Laube, A.: Das 3-Schicht-Kalorimeter—Ein einfaches, aber pra- n-alkanes tetradecane and tetracosane. In: Kiss, L., St.-Georges, zises Verfahren zur Bestimmung der Speicherkapazita¨t von L. (eds.) Proceedings of the 31th International Thermal Con- Latentwa¨rmespeichermaterial. http://www.waermepruefung.de ductivity Conference and 19th International Thermal Expansion (2016). Accessed 7 Feb 2016 Symposium, pp. 57–66. DEStech Publications Inc., Lancaster 28. Kenfack, F., Bauer, M.: Energy Procedia 46, 310 (2014). doi:10. (2013) 1016/j.egypro.2014.01.187 47. Losada-Pe´rez, P., Tripathi, C.S.P., Leys, J., Cordoyiannis, G., 29. Thoen, J., Marynissen, H., Van Dael, W.: Phys. Rev. A 26(5), Glorieux, C., Thoen, J.: Int. J. Thermophys. 32(5), 913 (2011). 2886 (1982). doi:10.1103/PhysRevA.26.2886 doi:10.1007/s10765-011-0984-0 30. Thoen, J.: High Resolution adiabatic scanning calorimetry and 48. Charva´t, P., Klimesˇ, L., Ostry´, M.: Energy Build. 68(part A), 488 heat capacities. In: Wilhelm, E., Letcher, T.M. (eds.) Heat (2014). doi:10.1016/j.enbuild.2013.10.011 capacities: liquids, solutions and vapours, pp. 287–306. The 49. Broadhurst, M.G.: J. Res. Natl. Inst. Stand. Technol. 66A(3), 241 Royal Society of Chemistry, London (2010). doi:10.1039/ (1962). doi:10.6028/jres.066A.024 9781847559791-00287 50. Roble`s, L., Mondieig, D., Haget, Y., Cuevas-Diarte, M.A.: J. 31. Leys, J., Losada-Pe´rez, P., Glorieux, C., Thoen, J.: J. Therm. Chim. Phys. 95(1), 92 (1998). doi:10.1051/jcp:1998111 Anal. Calorim. 177(1), 173 (2014). doi:10.1007/s10973-014- 51. Zammit, U., Marinelli, M., Mercuri, F., Paoloni, S.: J. Phys. 3654-1 Chem. B 114(24), 8134 (2010). doi:10.1021/jp102609y 32. Leys, J., Losada-Pe´rez, P., Slenders, E., Glorieux, C., Thoen, J.: 52. Ungar, G., Masˇic´, N.: J. Phys. Chem. 89(6), 1036 (1985). doi:10. Thermochim. Acta 582, 68 (2014). doi:10.1016/j.tca.2014.02.023 1021/j100252a030 33. Rubitherm Technologies GmbH. Data sheet RT42 (2013) 53. Sirota, E.B., King Jr, H.E., Shao, H.H., Singer, D.M.: J. Chem. 34. Dirand, M., Bouroukba, M., Chevallier, V., Petitjean, D., Behar, Phys. 99(2), 798 (1995). doi:10.1021/j100002a050 E., Ruffier-Meray, V.: J. Chem. Eng. Data 47(2), 115 (2002). 54. Arkar, C., Medved, S.: Thermochim. Acta 438(1–2), 192 (2005). doi:10.1021/je0100084 doi:10.1016/j.tca.2005.08.032 35. Danley, R.L.: Thermochim. Acta 395(1–2), 201 (2003). doi:10. 1016/S0040-6031(02)00212-5

Journal

Materials for Renewable and Sustainable EnergySpringer Journals

Published: Feb 11, 2016

References