International Journal of Heat and Mass Transfer 175 (2021) 121389

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

journal homepage: www.elsevier.com/locate/hmt

Investigation of transient heat transfer in multi-scale PCM composites

using a semi-analytical model
Adam Dobri a, A. Tsiantis b, T.D. Papathanasiou b,∗, Yanwei Wang a,∗
a
Department of Chemical and Materials Engineering, School of Engineering and Digital Sciences, Nazarbayev University, Kabanbay Batyr Ave. 53, Nur-Sultan
010000, Kazakhstan
b
Department of Mechanical Engineering, University of Thessaly, Volos, Greece

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents a semi-analytical model to describe the transient heat transfer in phase change ma-
Received 4 August 2020 terial (PCM)-based composites. One such example is micro-encapsulated paraffin in gypsum plaster walls
Revised 11 February 2021
for building applications. Without the availability of an analytical solution for the problem at hand, and
Accepted 19 April 2021
given the computational intensity of full-scale numerical solution of the problem, there is a need for
Available online 8 May 2021
alternate approaches that can be used in the design of PCM walls. The key assumption underlying the
Keywords: proposed model is that the spherical paraffin particles are small enough relative to the thickness of
Phase change material (PCM) the wall, and therefore at each time instant each particle is surrounded by a spatially uniform, albeit
Building application time-dependent matrix temperature. This evolving matrix temperature is used as a boundary condition
Numerical heat transfer in order to solve for the analytical temperature distribution at the particle-scale, from which the heat
Transient thermal response flow into the particle can be determined. This procedure avoids spatial discretization of the micro-scale
Multi-scale composites
and results in a macro-scale model in which the paraffin particles appear as sinks/sources at each nodal
point. The heat equation is then solved using the Method of Lines, which reduces a parabolic partial dif-
ferential equation into a set of ordinary differential equations. The proposed model is used to simulate
the constant flux wall conditions often seen in thermal characterization experiments of PCM walls and
structures. Simulation results elucidate the impact of the particle radius and interfacial resistance on the
transition at the end of the thermal management phase. Simulations of cyclic environmental tempera-
tures more relevant to building applications show that PCM volume loadings as low as 5% can reduce the
energy demands of an HVAC system by 15 to 20%. Moreover, the model is also shown to provide excel-
lent agreement with the work of Šavija and Schlangen, who simulated the transient thermal response of
hardening concrete using the commercial finite element package FEMMASSE.
© 2021 Elsevier Ltd. All rights reserved.

1. Introduction and cooling systems. Full-sized test rooms have been built using
plaster containing encapsulated PCM and tested for a year. The
Phase Change Materials (PCM) are materials whose phase rooms with PCM integrated into the gypsum plaster showed lower
changes occur favourably under certain operating conditions. Sys- daily maximal temperatures in relation to the reference gypsum-
tems based on these materials are of utmost importance in many walled rooms [13]. During a hot day, a portion of the heat flow-
applications, ranging from thermal insulation and thermal en- ing from the environment into the wall drives the phase change of
ergy storage for buildings [1–10] to optics and energy/data stor- the PCM. The melting of the PCM slows the increase in the wall
age [11,12]. The energy required to drive the phase change itself temperature, leading to air in the room heating more slowly. Dur-
is often the principal motivator, for example, latent heat can store ing a cool night, the reverse process happens as the PCM solidi-
more energy over a smaller temperature range than sensible heat fies and releases the heat stored in the wall to the environment
in thermal energy storage systems. and the room. This also leads to the PCM room being warmer at
Building designers are very interested in integrating PCM into night than the reference room. The transient nature of the heat-
their designs in order to reduce the energy demands on heating ing/cooling cycles and the desire to ensure that the PCM volume
fraction and size are adequate provide the motivation to study the
∗
transient thermal response of such materials.
Corresponding authors.
E-mail addresses: athpapathan@mie.uth.gr (T.D. Papathanasiou),
There are several different ways to integrate PCM into build-
yanwei.wang@nu.edu.kz (Y. Wang). ing materials, ranging from using panels of PCM to different types

https://doi.org/10.1016/j.ijheatmasstransfer.2021.121389
0017-9310/© 2021 Elsevier Ltd. All rights reserved.
A. Dobri, A. Tsiantis, T.D. Papathanasiou et al. International Journal of Heat and Mass Transfer 175 (2021) 121389

fully contained within the structure. A solution to this problem is

Nomenclature to use composites where discrete regions of PCM are surrounded
by the building material. This can be done at the macro-scale,
cj heat capacity of phase j with PCM introduced directly into the holes of the commonly-used
Fo particle Fourier number, Fo ≡ α pt/R2 hollow bricks [16], or at the micro-scale, with the fabrication of
f melt fraction of PCM particles plaster boards containing PCM [13,18]. Microencapsulated PCM are
Hf PCM latent heat preferable as the encapsulation reduces their reactivity and makes
i node index in bulk scale discretization them easier to handle, while the small size increases the surface
Kj thermal conductivity of phase j area across which heat transfer can occur [19].
L wall thickness Based on the testing procedures and the desired applications,
N number of nodes in the bulk scale discretization understanding the transient thermal response of these compos-
n0 number of ψn terms in the summation ites is paramount. Thermal characterization methods use a fit to
Q heat flux into the room at x = L an initial transient to extract properties such as the thermal dif-
Q pm heat exchange from the particles to the matrix fusivity; misunderstanding of these transients will introduce errors
within the control volume V into the reported properties. Transients are also important in the
QPCM heat flow into the room at x = L for PCM-based final applications; engineers would like to be able to control heat
composite wall flow rates to and from heat storage systems [20]. The cyclic nature
Qref heat flow into the room at x = L for the reference of the day/night temperature changes in cities imposes constraints
gypsum wall during the selection of building materials [19], where gypsum-PCM
QPCM relative change in the heat flow into room for a walls have been shown to reduce the maxima of the daily temper-
PCM wall vs the reference atures [13,21].
Qsensible amount of heat required to bring the entirety of the An HVAC system is a heating, ventilation, and air-conditioning
particle to the melting temperature after the onset system. Computer modeling of building HVAC systems is com-
of melting monly used during the design of new buildings; however, these
R particle radius programs rely on the underlying assumptions about heat trans-
r radial position fer in the materials. As the properties of a composite’s constituent
Tmv matrix temperature averaged over the control vol- parts diverge, or in the presence of an interfacial thermal resis-
ume V tance (as e.g. between encapsulated PCM and wall material), the
Tmelt PCM melting temperature traditional one-temperature models are no longer valid [15,22,23].
Tp temperature in the particle phase A composite with spherical PCM inclusions is one of these cases
Tp,avg volume-averaged temperature of the particle where considering the material to be homogenous with an effec-
TS temperature at the surface of the particle, Tp (r = R ) tive thermal conductivity can lead to errors. As there is no simple
t time analytical solution to heat transfer in these multi-scale composites,
t1 melting onset time and fully numerical approaches are too computationally intensive,
t2 melting completion time alternative models are required. Younsi et al. modeled the transient
V control volume response of PCM in plaster walls by considering the wall to be ho-
x axial position in wall mogenous and then solving a two-part enthalpy equation in a con-
Z (r, t ) time evolution of a particle under standard condi- trol volume in order to take into account the phase change [24].
tions   Šavija and Schlangen used a temperature-dependent specific heat
αj thermal diffusivity of phase j, α j = K j / ρ j c j capacity and a non-isothermal phase change to take into account
γ interfacial thermal resistance the latent heat [25].
ε particle volume fraction Recently, Jayalath et al. [26] presented a multi-scale analysis on
ρj density of phase j thermal properties of cement-based materials containing micro-
τ reduction in Km due to tortuosity encapsulated phase change materials. They considered the com-
ψn n th time-function related to the exchange posite to be a single, homogeneous material (a one-temperature
( )m value in the matrix phase model) without any resistance between the two phases and used
( )p value in the PCM particle phase computationally expensive 3D finite element method (FEM) simu-
( )i value at node i lations of smaller unit cells to determine the effective properties of
the composite.
In this study, we present a multi-scale, semi-analytical model
of composites. One of the drawbacks of a pure PCM system is that does not require micro-scale discretization to describe the
that PCM tend to be poor thermal conductors [14]. In practice, the transient response of composite materials comprised of microen-
low thermal conductivity leads to slower characteristic response capsulated PCM for applications such as building materials. We use
times, making solid PCM impractical for many storage applications. a two-temperature model that can take into account the interfacial
Many works have been produced about increasing the thermal resistance that is likely to be present between the two phases. The
conductivity by impregnating the pores of metal foams with PCM analytical solution to the temperature profile in a particle is used
[15,16] or dispersing high conductivity fillers into a block of PCM. to avoid the discretization at the particle-scale. We use the Method
In these composites the PCM provides the heat storage capacity of Lines, a very efficient method to solve problems in transient
while the metal foam or fillers increase the thermal conductivity heat conduction, as it converts an initial-boundary value problem
in order to increase the speed by which the heat is transported into a system of ODEs, which can be solved by readily available
to the PCM. Chiu et al. showed a much faster thermal response algorithms[27].
and increased heat accumulation rate in a composite slab when The outline of this paper is as follows. In Section 2, the model
the PCM was intercalated with expanded graphite [17]. development is presented, first at the particle-scale and then at
Composites based on metal foams or intercalated PCM can be the macro-scale. In Section 3 the model is used to describe the
quite useful in latent heat storage systems, however they are im- thermal response of a slab under a constant flux boundary condi-
practical for use as construction materials because the PCM is not tion, similar to the conditions often found in materials characteri-

2
A. Dobri, A. Tsiantis, T.D. Papathanasiou et al. International Journal of Heat and Mass Transfer 175 (2021) 121389

is the heat capacity, K is the thermal conductivity and r is the ra-

dial position. The subscript p indicates the particle phase, which is
the PCM material.
 2 
∂ Tp ∂ Tp 2 ∂ Tp
ρpc p = Kp + (1)
∂t ∂ r2 r ∂r
The internal symmetry boundary condition is given by Eq. (2)

∂ Tp 
=0 (2)
∂ r r=0
When the particle is small relative to the bulk dimensions,
the surrounding medium can be considered to be of uniform, al-
beit time-evolving temperature, Tmv . The surface boundary condi-
tion for heat flow from the particle to the surrounding medium
in the presence of an interfacial thermal resistance, γ , is given by
Fig. 1. Nodal representation of a composite plaster wall with spherical PCM inclu-
sions, with an enlargement of node i, where Tmv denotes the matrix temperature Eq. (3) [28]. As the particles are encapsulated, the interfacial re-
averaged over the control volume. sistance is expected to play a role in the thermal response [29].
TS is the surface temperature of the particle and is equivalent to
Tp (r = R ). This equation represents the heat flow per unit area
zation. In Section 4, the model is applied to simulations of cyclic from the particle to the matrix and provides the coupling between
environmental temperatures more relevant to building applications the micro and macro-scales.
of PCM-based composites. In Section 5 we compare the results of  
∂ Tp TS − Tmv
our model with those of Šavija and Schlangen [25] for the case of −K p = (3)
microencapsulated phase change materials in hardening concrete. ∂r r=R
γ
Section 6 concludes and summarizes the work. We use the analytical solution to ∂ Tp /∂ r (Eq. (3), above) in or-
der to avoid discretization at the particle-scale and evaluate this
2. Model development boundary condition. The analytical solution is obtained by using
Duhamel’s Theorem to solve Eq. (1) subject to the time-varying
Consider the case of a composite slab which is comprised of boundary condition, TS . Z (r, t ) is the time evolution of the parti-
a plaster matrix with the volume fraction, ε , of spherical inclu- cle temperature for an initial temperature of zero and a surface
sions of PCM. The wall is discretized into N nodes, as shown in temperature
  of unity, where the thermal diffusivity, α p , is given
Fig. 1, where the control volume, V, associated with each nodes is by K p / ρ p c p .
small relative to the bulk scale. The control volume is small enough ∞
2R  (−1 )n
 r  α pt
that the matrix temperature does not vary significantly within Z (r, t ) = 1 + sin nπ exp −n2 π 2 2 (4)
it yet large enough to include a representative sample of parti- πr n=1
n R R
cles. The matrix temperature averaged over the control volume is
For a time-varying surface temperature, the transient profile is
denoted by Tmv . The semi-analytical model is a two-temperature
given by Eq. (5).
model where the analytical solution to the temperature profile in

t
a particle with a single, time-varying surface condition is used to dTS
Tp (r, t ) = Z (r, t − λ ) dλ (5)
evaluate the heat flux from the particle to the matrix phase. This 0 dλ
heat flux is then treated as a heat source/sink in the matrix-phase
Substituting Eq. (4) into (5) and evaluating the integral leads to
heat equation in the control volume. This discretization at the bulk
Eq. (6), where the time-functions, ψn (t ), are given by Eq. (7).
scale leads to a system of ordinary differential equations (ODEs)
that can be solved using readily available solvers.
∞
2R  (−1 )n
 r
Tp (r, t ) = TS + sin nπ ψn (t ) (6)
The following assumptions are used: πr n=1
n R
• the particles are small relative to the bulk scale, R << L, where
  
α t t dTS
R is the radius of a PCM particle, and L is the thickness of a

αpλ
PCM-based composite slab ψn (t ) = exp −n2 π 2 p2 exp n2 π 2 2 dλ , n = 1, 2, 3, · · ·
R 0 dλ R
• both the matrix and the PCM phases are homogeneous,
(7)
isotropic
• thermal conductivity, heat capacity and density of a given It is more convenient to rewrite Eq. (7) in its differential form,
phase are temperature invariant
d ψn n2 π 2 α p dT
• phase change occurs isothermally =− ψn + S (8)
• the Rayleigh number of the liquid-phase PCM is small enough dt R2 dt
that conduction dominates over natural convection where the initial condition is ψn (t = 0 ) = 0. The analytical solution
• the volume change during PCM phase change is ignored to the temperature profile in the particle, Eq. (6), can now be used
• heat transfer is one-dimensional at the bulk scale as the thick- to evaluate the ∂ Tp /∂ r term and rewrite the particle boundary con-
ness of the wall is small relative to its height and width dition as Eq. (9).
∞
2 T − Tv
−K p ψn = S m (9)
2.1. Heat equation in the particle phase R
n=1
γ
Consider a spherical particle of radius R, isolated within From Eq. (8), it is evident that the ψn functions are first-order dif-
a matrix. The heat equation in a single particle is given by ferential equations, subject to the forcing function dTs /dt. For large
Equation 1 [28], where T is the temperature, ρ is the density, c enough values of n, ψn can be reasonably approximated by its

3
A. Dobri, A. Tsiantis, T.D. Papathanasiou et al. International Journal of Heat and Mass Transfer 175 (2021) 121389

quasi-steady state approximation. This allows the summation to be

split into two parts, the first n0 terms are given by the solutions to
the ODEs in Eq. (8) and the second set of terms is approximated
by the quasi-steady state approximation, leading to Eq. (10).
 
∞
2  
n0
R2 dTs TS − Tmv
−K p ψn + = (10) Fig. 2. Example of the constant flux boundary condition often found in testing
R
n=1
n π 2 α p dt
2
n=n0 +1
γ [18,40].

This may be rearranged into a differential equation for the surface Table 1
temperature in Eq. (11). Thermal properties of the three phases involved in this
0 work.
dTs 2γ K p ( m
R
T v − TS ) − nn=1 ψn
= 2
 π 2 n 0 1  (11) Matrix Solid PCM Liquid PCM
dt R
− K, Wm−1 K−1
2π αp 6 n=1 n
2 0.25 0.34 0.17
ρ , kgm−3 1140 900 750
2.2. Heat equation in the matrix phase c, J kg−1 K−1 1000 2500 2900

The presence of the spheres, shown in Fig. 1, affects the heat

equation in two ways, as shown in Eq. (12). First, the thermal con- by Eq. (14). The heat flux into the particle is obtained from the
ductivity, Km , is reduced via the (1 − τ ) term due to the tortuosity matrix-phase heat exchange term in Eq. (13). This flux is used to
of the matrix. Second, there is a heat generation term that repre- estimate the melting time; the phase change continues until time
sents the exchange of heat from the matrix to the particles, Q pm . t2 when enough heat has entered the particle to completely melt it
[36]. At t1 , the volume average temperature, Tp,avg , of the particle,
∂ Tmv Km (1 − τ ) 2 v Q pm defined by Eq. (15), is still below Tmelt .
= ∇ Tm + (12)

∂t ρm c m ( 1 − ε ) ρm c m 1 R
Tp,avg = Tp (r )4π r 2 dr (15)
This form is similar to the forms used by Zhang et al. [23] and 4
3
π R3 0
Li [15] who have considered Km (1 − τ ) to be equal to the effec-
This average can be calculated using Tp (r, t = t1 ) as given by
tive thermal conductivity of the matrix. The Km (1 − τ ) term can
Eq. (6).
be extracted from experimental data whenever possible [30] or it
The difference between the melting temperature and Tp,avg is
can be calculated from correlations such as those given by Maxwell
used to calculate the sensible heat that is required to bring the
[31] or Felske [32]. Felske’s approximations are used in this work
entirety of the particle to the meting temperature, Qsensible , and is
as Ricklefs et al. showed a good agreement between experimental
given by Eq. (16):
measurements of steady state thermal conductivity in cementitious   

dTs 
composites with up to 30% volume fraction micro-encapsulated 
n0 ∞

4 6 ψn (t1 ) R2
Qsensible = π R3 ρ p c p + (16)
PCM and the approximations [33]. 3 π2 n2 n4 π 2 α p dt 
n=1 n=n0 +1 t=t1
The interphase heat exchange, Q pm , in Eq. (12) is given by
Eq. (13), The completion of melting, t2 is considered to occur when the
3ε total integrated heat flow to the particle is equal to the sum of the
Q pm = (TS − Tmv ) (13) latent heat term and Qsensible given by Eq. (17), where H f is the
Rγ
latent heat of fusion of the PCM particle.
where it is considered to be the product of the number of particles
t2   4
per volume, the surface area per particle and the heat flux from 4π R2 γ −1 v (λ ) − T
Tm,i S,i d λ = π R3 ρ p H f + Qsensible (17)
the particle to the matrix. In the present study, the heat flux from t1 3
the particle to matrix is evaluated using the particle-scale bound- When the particle Fourier Number (Fo ≡ α pt/R2 ) is large, the
ary condition given by Eq. (3). This provides the coupling between volume average temperature of the particle will be approximately
the micro- and macro-scales. equal to the surface temperature and the contribution of the sen-
The partial differential equation given by Eq. (12) is solved by sible heat term will be negligible. After the melting of the parti-
the Method of Lines, where the x-direction is discretized into N cles at a given node is complete, the particle surface temperature
nodes, with Eq. (14) describing the ODE for the volume-averaged is again allowed to vary based on Eqs. (8) and (11), using material
matrix temperature at node i. parameters for the PCM in the liquid phase.
 
v
dTm,i v v v
Km (1 − τ ) Tm,i+1 − 2Tm,i + Tm,i−1 3ε TS,i − Tm,iv
= + (14) 3. Simulations for constant flux materials characterization
dt ρm c m x 2 R ( 1 − ε ) ρm γ c m
Eqs. (11) and (8) describe the ODEs for TS,i and ψn,i , respectively. The set of boundary conditions used to demonstrate an appli-
This system of N (n0 + 2 ) ODEs can be solved using standard nu- cation of the model is that of a constant flux on one surface of
merical solvers; in this work we used MATLAB’s ode15s. We have the wall and heat transfer to an environment on the other sur-
previously validated an analogous approach for non-PCM compos- face, as shown in Fig. 2. The temperature evolution within or at
ites based on cylindrical inclusions using OpenFOAM simulations one extremity of a material while the other surface is subject to a
[34]. constant heat flux is a typical configuration in thermal character-
ization. This flux condition may be applied via a resistive heating
2.3. Melting element [15,17,37,38] or a heating lamp [18,39].
Table 1 summarizes the thermal properties of the three phases
The onset of melting occurs at time t1 as TS,i = Tmelt , where incorporated into the 2.5 cm thick wall. Unless otherwise noted,
Tmelt is PCM’s melting temperature. During melting, the common the melting temperature, Tmelt , is 23◦ ; the radius R is 25 μm; the
assumption of an isothermal phase change [17,24,35] implies that latent heat of fusion, H f , is 240J g−1 ; the interfacial thermal resis-
Eq. (11) is reduced to dTS,i /dt = 0. As the phase change is isother- tance, γ is 0.1 m2 KW−1 , and the heat transfer coefficient between
mal, the ψn,i values are irrelevant. Tm,i
v continues to be described the wall and the environment is set to be 10 Wm−2 K−1 .

4
A. Dobri, A. Tsiantis, T.D. Papathanasiou et al. International Journal of Heat and Mass Transfer 175 (2021) 121389

Fig. 3. (a) Transient response of the matrix temperature at the right side of the wall for ε from 0 to 0.15 and (b) the relative increase in the length of time required to reach
26◦ vs. the volume fraction, ε (Parameters: R = 25 μm and γ = 0.1 m2 KW−1 ).

Fig. 4. Transient response of the matrix temperature at the right side of the wall for (a) R from 2.5 to 250 μm (Parameters: ε = 0.1 and γ = 0.1 m2 KW−1 ) and (b) γ from
10−4 to 1 m2 KW−1 (Parameters: ε = 0.1 and R = 25 μm).

The effect of the PCM particle loading on the thermal response A sufficient number of N terms should be determined for each
at the opposite side of the wall is shown in Fig. 3, with particle set of model parameters. For example, in order to properly re-
volume fraction ε = 0.15 leading to a 350% increase in the time solve the evolution of the melt front, larger N values are re-
required for the wall surface to reach 26◦ . The trend agrees with quired for smaller particles with lower interfacial resistance. This
the experimental studies that have also shown significant reduc- is due to the ratio between the conduction (∝ N −1 ) and exchange
tions in the heating rate of PCM composites, leading to flattening (∝ R−1 γ −1 ) terms in the discretized matrix phrase heat equation
of the response in the “thermal management” phase [15,18,39,41]. in Eq. (14). When an insufficient number of nodes are used, the
Upon completion of melting, the sample heats more rapidly as the exchange term approaches the magnitude of the conduction term.
latent heat is no longer dominating the response. All the heat flowing into the node contributes to the melting and
The abruptness of this transition out of the thermal manage- each node melts completely before the following node. In practice
ment phase depends on the particle radius and the interfacial re- the limits where γ → 0 or R → 0 are not expected to occur due
sistance, with smaller R and γ values leading to the most abrupt to the encapsulation layer around the PCM [35] and the discrete
transition, as shown in Fig. 4. The thermal management effect can nature of the PCM encapsulation process.
be shown using relatively short measurements and, as such, many
tests are not maintained for long enough to effectively capture this 4. Simulations for building applications
transition [15,18]. In practice, thermal characterization should be
long enough to capture the shape of the transition. The boundary conditions presented in Section 3 are applica-
As the method is semi-analytical, the number of nodes in the ble to many materials characterization methods; however, they are
bulk discretization and the number of ψn terms to be evaluated not typically found in building applications. A more relevant set of
using Eq. (8) need to be selected so that they do not impact the boundary conditions are shown in Fig. 6, where the external en-
accuracy of the simulation. The impact of N on the time to reach vironmental temperature varies from 20◦ to 35◦ over the course
26◦ at the right hand side and the computation runtime are shown of a day. The interior environment contains an HVAC system that
in Fig. 5a. N = 51 typically provides a good balance between con- works to maintain the room at 20◦ . The heat transfer coefficient
vergence and solver run time; odd values are used to ensure that between the wall and the air is considered to be the same on both
a node is located precisely at x = L/2. Fig. 5b shows the conver- sides. The amount of heat that needs to be removed by the HVAC
gence of the time to reach 22◦ at the left hand side. The temper- system in order to maintain a comfortable temperature during the
ature lower than the melting temperature of the PCM was chosen day is an important parameter. When comparing walls, building
because the ψn terms are only relevant during sensible heating of designers would like to reduce energy usage and therefore costs.
the PCM particles. The relative change for a PCM wall vs. the reference gypsum wall,
QPCM , is given by Eq. (18) where QPCM and Qref are the heat flows
into the room at x = L for PCM-based composite and gypsum walls,

5
A. Dobri, A. Tsiantis, T.D. Papathanasiou et al. International Journal of Heat and Mass Transfer 175 (2021) 121389

Fig. 5. (a) Convergence of the time to reach 26◦ at the right-hand side and the solver run time vs. N and (b) Convergence of the time to reach 22◦ at the left-hand side and
the solver run time vs. n0 (Parameters: R = 25 μm and γ = 0.1 m2 KW−1 ).

Fig. 6. Example of an external environment boundary condition where the interior room is maintained at a constant temperature by an HVAC system.

Fig. 7. (a) Transient response of the matrix temperature and heat flux at the right side of the wall for one cycle of a transient external temperature and (b) heat flux from
the walls to the room and the wall-averaged PCM melt fraction (Parameters: ε = 0.05, R = 25 μm, and γ = 0.1 m2 KW−1 ).

respectively. The effect of the PCM melting temperature and heat of fusion
on the QPCM during the first 12 h are shown in Fig. 8. Only the
 
QPCM dt − Qref dt heating phase during the day is considered as this is the most im-
QPCM (t ) =  (18) portant period in building applications such as offices. At night,
Qref dt
a higher air exchange rate can be used to accelerate the release
of the latent heat from the solidifying PCM [13] and many juris-
The thermal response of the wall to one cycle of a transient
dictions offer off-peak electricity rates. A larger H f leads to more
exterior temperature ramping from 20 to 35◦ and back is shown
energy savings, as more heat can be stored in the wall. A lower
in Fig. 7. The temperature on the inner side (x = L) in the PCM
melting temperature is also beneficial, until a threshold is reached
wall initially follows the gypsum reference wall before the onset of
where the QPCM is no longer limited by the transfer rate to the
the thermal management phase where the PCM melts. During this
particles.
phase, roughly from 4 to 8 h, the PCM wall temperature rises more
If the melting temperature is too low, or the PCM wall is other-
slowly. The lower PCM wall temperature leads to a lower heat flux
wise unable to discharge the stored heat overnight, it is considered
into the room in comparison with the reference wall. During cool-
to be over-saturated. In this case, the morning will start while part
ing the reference wall temperature and the associated heat flux to
of the wall is still in the liquid phase, so less of the solid PCM
the room decrease more rapidly than the PCM wall as the par-
is available to absorb the following day’s heat. An example of this
ticles solidify. The higher heat flux during the cooling period in
type of case is shown in Fig. 9 where 60 % of the wall remains in
the PCM wall shown in Fig. 7(b) is consistent with the experimen-
the liquid phase at the start of the second cycle. The wall tempera-
tal results where the nighttime temperatures in PCM-based rooms
ture and heat flux to the room are consequently higher on the sec-
were warmer than the gypsum reference [13].

6
A. Dobri, A. Tsiantis, T.D. Papathanasiou et al. International Journal of Heat and Mass Transfer 175 (2021) 121389

Fig. 8. QPCM as a function of the particle volume fraction for a series of (a) latent heat and (b) melting temperature values (Parameters: R = 25 μm and γ = 0.1 m2 KW−1 ).

Fig. 9. The heat flux and melt fraction as a function of time for two consecutive
days, where the PCM does not fully re-solidify after the night (Parameters: ε = 0.1, Fig. 10. Predicted temperature increase in time vs different levels of PCM loading
Tmelt = 21◦ , γ = 0.1 m2 KW−1 , and H f = 240J g−1 ). without taking into account the latent heat contribution. The dots are from a 30 ×
30 × 30 μm3 representative volume element in [25], and the lines are predictions
from our model.

ond day, with QPCM being 19 % lower on the second day. Building
designers will need to carefully consider the local day and night-
time temperatures and the capacities of the HVAC systems before are isolated within the matrix, an assumption that will break down
selecting the PCM material and wall thickness to be used. This will at high particle loadings.
ensure that the walls meet the cooling requirements while mini- The effect of the latent heat is shown in Fig. 11 for 10% and
mizing their costs. 20% volume loading of PCM particles; in all cases, the PCM’s phase
transition delays the temperature increase. The decrease in adi-
abatic temperature increase after three days is summarized in
5. Comparison with an existing model Fig. 12.
The effect of our model’s assumptions about the melting phase
In this section the proposed semi-analytical model for tran- are shown in Fig. 13. The melting time estimation based on the
sient heat transfer in PCM-based composites is used to simulate heat flux into the particles is more accurate for lower heats of fu-
the temperature changes occurring in hardening concrete. The re- sion because for large H f values the movement of the melt inter-
sults are compared with an existing model described by Šavija and face becomes more important. This is not expected to be an is-
Schlangen [25]. The rate of heat generation of the concrete reaction sue in building applications as typical heats of fusion tend to be
is given by Figs. 4 and 14 in ref. [25] for the meso- and macro- closer to 90 − 180 than 360 − 900 kJkg−1 [19]. The abruptness of
scale simulations, respectively. This heat generation is taken into the isothermal phase change assumption in our model can also
account by adding a heat source in Eq. (14). The particles are as- be seen in comparison to Šavija and Schlangen who assumed that
sumed to have a radius of 25 μm and an interfacial resistance of the melting occurred over a small temperature window. In future
10−3 m2 KW−1 . studies, our model may be updated to accommodate for a non-
The first simulations presented by Šavija and Schlangen were isothermal phase change.
those of a 30 × 30 × 30 μm3 mesh under adiabatic conditions. This In their second set of simulations, Šavija and Schlangen sim-
cube is compared with a single averaging volume, without conduc- ulated the setting of a 1 meter-thick wall and its fixed concrete
tion, in our model. The dilution effect, of the PCM loading, without base that is open to the environment at sides and top. They used
taking into account the latent heat, is shown in Fig. 10. The points the FEMMASSE finite element model tool, which is based on the
correspond to the reference curves while the solid lines correspond concept of state parameter. The maximum temperature within the
to the matrix temperature predicted by the semi-analytical model wall, important in the calculation of thermal stresses and crack-
where the inclusions are treated as heat sinks. The semi-analytical ing, is tracked against time for these simulations. As our model
model reproduces the same trend, with good accuracy at lower presents a 1D slab, the two cases are not directly compared on the
volume loading. In Section 2 we considered that all the particles same axes.

7
A. Dobri, A. Tsiantis, T.D. Papathanasiou et al. International Journal of Heat and Mass Transfer 175 (2021) 121389

Fig. 11. Effect of the latent heat contribution on the transient response of an adiabatic control volume with a volume loading of (a) 10% and (b) 20% PCMs. The dots are
from [25], and the lines are predictions from our model.

and completion of the phase transitions. This is related to our as-

sumption of an isothermal phase change.
For a given volume fraction, a lower melting temperature PCM
leads to a greater suppression of the maximum temperature as
shown in Fig. 15. When the melting temperature is too high, for
example 50◦ in Fig. 15(b), the PCM does not melt and therefore
the added PCM is of little use in the final application. Again, the
main difference between the two models is the relative abruptness
of the transitions during the onset and completion of melting that
we attribute to our assumption of an isothermal phase change.
For a given volume fraction and melting temperature, a higher
heat of fusion of the PCM leads to a greater suppression of the
maximum temperature as shown in Fig. 16. When most of the heat
released during the setting reaction drives the phase change of the
PCM, it is possible to almost completely remove the early temper-
Fig. 12. Decrease in the simulated adiabatic temperature rise after 3 days of hydra- ature peak and its associated risk of cracking.
tion The dots are from [25], and the lines are predictions from our model. Overall the semi-analytical model developed in this work is
able to reproduce the thermal responses simulated using the com-
mercial finite element package FEMMASSE by Šavija and Schlangen
The effect of the amount of PCM on the maximum temperature [25]. Our simulations were completed on a desktop workstation
in the setting wall is shown in Fig. 14. The melting temperature is in run times on the order of tens of seconds, considerably faster
25◦ and the heat of fusion is 180 kJkg−1 unless otherwise speci- than typical transient FEM analyses. The agreement between the
fied. Šavija and Schlangen [25] showed that increasing the amount two models is better at lower volume fractions and lower heats of
of PCM lowers the maximum temperature occurring in the wall in fusion of the PCM. The main difference between the two sets of
the initial stages of the concrete setting reaction. This is desired results is the abruptness of the phase transitions that arise due to
in order to mitigate the risks of early age thermal cracking. Our the reference work considering a phase change temperature win-
model reproduces this trend, with the most notable difference be- dow while the presented model considers an isothermal phase
ing the relative abruptness of the temperature changes at the onset change.

Fig. 13. Influence of the (a) latent heat of fusion and (b) melting temperature of PCM microcapsules on the adiabatic temperature response of a volume loading of 10% PCM
cement. The dots are from [25] and the lines from our model.

8
A. Dobri, A. Tsiantis, T.D. Papathanasiou et al. International Journal of Heat and Mass Transfer 175 (2021) 121389

Fig. 14. The simulated evolution of the maximum temperature in a hardening concrete wall depending on the PCM microcapsule addition level: (a) Reference simulations
and (b) Our semi-analytical model.

Fig. 15. The simulated evolution of the maximum temperature in a hardening concrete wall depending on the melting temperature of the 10% microcapsules (a) Reference
Simulations and (b) Our semi-analytical model.

Fig. 16. The simulated evolution of the maximum temperature in a hardening concrete wall depending on the heat of fusion of the 10% microcapsules (a) Reference
Simulations and (b) Our semi-analytical model.

6. Conclusion and particle surface temperatures. The model captures the onset
and completion of the thermal management phase. While the PCM
A two-temperature semi-analytical model for the transient parameters are typically selected for their impact on the thermal
heating of PCM-based walls that avoids the discretization at the management phase itself, the transition to the fully-melted PCM
particle-scale by using an analytical solution for the intra-particle phase can provide more information about the composite. The par-
temperature subject to a spatially constant but time-dependent lo- ticle sizes and thermal interfacial resistances tend to modify the
cal matrix temperature, has been presented. The macro-scale PDE abruptness of the transition while increases in the volume frac-
is transformed into a system of ODEs using the Method of Lines. tion or latent heat tend to delay the transition. For these reasons,
The resulting system of ODEs can be solved using standard numeri- it is of upmost importance to conduct long thermal characteriza-
cal solvers, such as MATLAB’s ode15s, to determine the local matrix tion experiments in order to better interpret the results. The model

9
A. Dobri, A. Tsiantis, T.D. Papathanasiou et al. International Journal of Heat and Mass Transfer 175 (2021) 121389

also provides a method to estimate the energy savings that can [14] P. Zhang, X. Xiao, Z.W. Ma, A review of the composite phase change materials:
be attained by different PCM-based composite walls under a vari- fabrication, characterization, mathematical modeling and application to perfor-
mance enhancement, Appl. Energy 165 (2016) 472–510, doi:10.1016/j.apenergy.
ety of transient environments, with energy savings during summer 2015.12.043.
months being 15 to 20% for PCM loadings as low as 5% (volume [15] W.Q. Li, H. Wan, T.T. Jing, Y.B. Li, P.J. Liu, G.Q. He, F. Qin, Microencapsulated
fraction). Finally, on the temperature evolution in hardening con- phase change material (MEPCM) saturated in metal foam as an efficient hy-
brid PCM for passive thermal management: a numerical and experimental
crete with the addition of microencapsulated phase change mate- study, Appl. Therm. Eng. 146 (September 2018) (2019) 413–421, doi:10.1016/
rials, good agreement has also been shown between the presented j.applthermaleng.2018.10.006.
model and the published FEM simulations by Šavija and Schlangen [16] Y. Zhang, Q. Wang, Impact of phase change material’s thermal properties on
the thermal performance of phase change material hollow block wall, Heat
[25].
Transf. Eng. 40 (19) (2019) 1619–1632, doi:10.1080/01457632.2018.1480879.
[17] Y.J. Chiu, W.M. Yan, H.C. Chiu, J.H. Jang, G.Y. Ling, Investigation on the ther-
Declaration of Competing Interest mophysical properties and transient heat transfer characteristics of composite
phase change materials, Int. Commun. Heat Mass Transf. 98 (October) (2018)
223–231, doi:10.1016/j.icheatmasstransfer.2018.09.011.
The authors declare that they have no known competing finan- [18] L. Liu, B. Xiong, Y. Men, Facile preparation of porous plaster board containing
cial interests or personal relationships that could have appeared to phase change capsules using gel template, Energy Build. 156 (2017) 134–139,
doi:10.1016/j.enbuild.2017.09.074.
influence the work reported in this paper. [19] Y. Konuklu, M. Ostry, H.O. Paksoy, P. Charvat, Review on using microencapsu-
lated phase change materials (PCM) in building applications, Energy Build. 106
(2015) 134–155, doi:10.1016/j.enbuild.2015.07.019.
CRediT authorship contribution statement
[20] A.M. Khudhair, M.M. Farid, A review on energy conservation in building ap-
plications with thermal storage by latent heat using phase change materials,
Adam Dobri: Conceptualization, Methodology, Investigation, Energy Convers. Manage. 45 (2) (2004) 263–275, doi:10.1016/S0196-8904(03)
Writing - original draft, Writing - review & editing. A. Tsiantis: 00131-6.
[21] J.P. Bravo, T. Venegas, E. Correa, A. Álamos, F. Sepúlveda, D.A. Vasco, C. Bar-
Investigation, Writing - review & editing. T.D. Papathanasiou: Re- reneche, Experimental and computational study of the implementation of
sources, Writing - review & editing, Funding acquisition. Yanwei mPCM-modified gypsum boards in a test enclosure, Buildings 10 (1) (2020),
Wang: Conceptualization, Resources, Writing - review & editing, doi:10.3390/buildings10010015.
[22] M. Quintard, Heat transfer in composite materials and porous media: multi-
Project administration, Funding acquisition. ple-scale aspects and effective properties, Heat Transf. Polym. Compos. Mater.
(2016) 175–201.
[23] P. Zhang, Z.N. Meng, H. Zhu, Y.L. Wang, S.P. Peng, Melting heat transfer charac-
Acknowledgements
teristics of a composite phase change material fabricated by paraffin and metal
foam, Appl. Energy 185 (2017) 1971–1983, doi:10.1016/j.apenergy.2015.10.075.
This work was carried out with support from Grant Award [24] Z. Younsi, H. Naji, M. Lachheb, Numerical investigation of transient ther-
Number 090118FD5313 under the “Structure-Property Correlations mal behavior of a wall incorporating a phase change material via a hybrid
scheme, Int. Commun. Heat Mass Transf. 78 (2016) 200–206, doi:10.1016/j.
in Multi-Scale Composites” project from Nazarbayev University. icheatmasstransfer.2016.09.007.
The authors would like to acknowledge Dr. Branko Šavija from [25] B. Šavija, E. Schlangen, Use of phase change materials (PCMs) to mitigate early
the Delft University of Technology for making available the data age thermal cracking in concrete: theoretical considerations, Constr. Build.
Mater. 126 (2016) 332–344, doi:10.1016/j.conbuildmat.2016.09.046.
from his published model for the purposes of the comparison in [26] A. Jayalath, L. Aye, T. Ngo, P. Mendis, Multi-scale analysis on thermal prop-
Section 5. The authors would like to acknowledge the support of erties of cement-based materials containing micro-encapsulated phase change
the Greek Research and Technology Network (GRNET) for the com- materials, Constr. Build. Mater. 254 (2020) 119221.
[27] V.R. Subramanian, R.E. White, Semianalytical method of lines for solving el-
putational time granted in the National HPC facility ARIS. liptic partial differential equations, Chem. Eng. Sci. 59 (4) (2004) 781–788,
doi:10.1016/j.ces.2003.10.019.
References [28] H.S. Carslaw, J.C. Jaeger, Conduction of Heat in Solids, International Series of
Monographs on Physics, second ed., Oxford, Clarendon Press, 1959.
[1] R. Baetens, B.o.r.P. Jelle, A. Gustavsen, Phase change materials for building ap- [29] W. Li, S.-G. Li, S. Guan, Y. Wang, X. Zhang, X. Liu, Numerical study on melt
plications: a state-of-the-art review, Energy Build. 42 (9) (2010) 1361–1368. fraction during melting of phase change material inside a sphere, Int. J. Hy-
[2] B.P. Jelle, Traditional, state-of-the-art and future thermal building insula- drogen Energy 42 (29) (2017) 18232–18239, doi:10.1016/j.ijhydene.2017.04.136.
tion materials and solutions–properties, requirements and possibilities, Energy Special Issue on The 5th International Conference on Energy Engineering and
Build. 43 (10) (2011) 2549–2563. Environmental Engineering (ICEEEE2017), 15-16 April 2017, Xiamen, China
[3] B. Zalba, J.M. Marın, L.F. Cabeza, H. Mehling, Review on thermal energy stor- [30] G.C. Glatzmaier, W.F. Ramirez, Use of volume averaging for the modeling
age with phase change: materials, heat transfer analysis and applications, Appl. of thermal properties of porous materials, Chem. Eng. Sci. 43 (12) (1988)
Therm. Eng. 23 (3) (2003) 251–283. 3157–3169.
[4] S. Raoux, Phase change materials, Annu. Rev. Mater. Res. 39 (2009) 25–48. [31] J. Maxwell, Conduction through heterogeneous media, chapter ix–a treatise on
[5] A. Sharma, V.V. Tyagi, C. Chen, D. Buddhi, Review on thermal energy storage electricity and magnetisma, 1873.
with phase change materials and applications, Renew. Sustain. Energy Rev. 13 [32] J. Felske, Effective thermal conductivity of composite spheres in a continuous
(2) (2009) 318–345. medium with contact resistance, Int. J. Heat Mass Transf. 47 (14) (2004) 3453–
[6] V. Tyagi, S. Kaushik, S. Tyagi, T. Akiyama, Development of phase change mate- 3461, doi:10.1016/j.ijheatmasstransfer.2004.01.013.
rials based microencapsulated technology for buildings: a review, Renew. Sus- [33] A. Ricklefs, A.M. Thiele, G. Falzone, G. Sant, L. Pilon, Thermal conductivity of
tain. Energy Rev. 15 (2) (2011) 1373–1391. cementitious composites containing microencapsulated phase change materi-
[7] F. Kuznik, D. David, K. Johannes, J.-J. Roux, A review on phase change mate- als, Int. J. Heat Mass Transf. 104 (2017) 71–82, doi:10.1016/j.ijheatmasstransfer.
rials integrated in building walls, Renew. Sustain. Energy Rev. 15 (1) (2011) 2016.08.013.
379–391. [34] A. Dobri, Y. Wang, T.D. Papathanasiou, Transient heat transfer in fibrous multi-
[8] D. Zhou, C.-Y. Zhao, Y. Tian, Review on thermal energy storage with phase scale composites: a semi-analytical model and its numerical validation, Nu-
change materials (PCMs) in building applications, Appl. Energy 92 (2012) mer. Heat Transf. Part A 77 (9) (2020) 840–852, doi:10.1080/10407782.2020.
593–605. 1746154.
[9] S.A. Memon, Phase change materials integrated in building walls: a state of [35] T. Rhafiki, T. Kousksou, A. Allouhi, W. Benomar, H. Zennouhi, A. Jamil,
the art review, Renew. Sustain. Energy Rev. 31 (2014) 870–906. Y. Zeraouli, Numerical analysis of a micro-encapsulated PCM wallboard:
[10] P.K.S. Rathore, S.K. Shukla, Potential of macroencapsulated PCM for thermal en- fluxmeter applications, J. Build. Eng. 14 (2017) 127–133, doi:10.1016/j.jobe.2017.
ergy storage in buildings: a comprehensive review, Constr. Build. Mater. 225 10.008.
(2019) 723–744. [36] V. Alexiades, A.D. Solomon, Mathematical Modeling of Melting and Freezing
[11] M. Wuttig, N. Yamada, Phase-change materials for rewriteable data storage, Processes, Routledge, New York, 1959.
Nat. Mater. 6 (11) (2007) 824–832. [37] H. Roder, A transient hot wire thermal conductivity apparatus for fluids, J. Res.
[12] M. Wuttig, H. Bhaskaran, T. Taubner, Phase change materials for non-volatile Natl. BureauStand. 86 (1981) 457–493.
photonic applications, Nat. Photonics 11 (8) (2017) 465–476, doi:10.1038/ [38] M. Assael, K. Antoniadis, I. Metaxa, S. Mylona, Y. Assael, J. Wu, M. Hu, A
nphoton.2017.126. novel portable absolute transient hot-wire instrument for the measurement
[13] P. Schossig, H.-M. Henning, S. Gschwander, T. Haussmann, Micro-encapsulated of the thermal conductivity of solids, Int. J. Thermophys. 36 (2015) 3083–3105,
phase-change materials integrated into construction materials, Solar Energy doi:10.1007/s10765-015-1964-6.
Mater. Solar Cells 89 (2) (2005) 297–306, doi:10.1016/j.solmat.2005.01.017. Eu- [39] B. Srinivasaraonaik, L. Singh, S. Sinha, I. Tyagi, A. Rawat, Studies on the me-
roSun2004 chanical properties and thermal behavior of microencapsulated eutectic mix-

10
A. Dobri, A. Tsiantis, T.D. Papathanasiou et al. International Journal of Heat and Mass Transfer 175 (2021) 121389

ture in gypsum composite board for thermal regulation in the buildings, J. [41] A.M. Borreguero, I. Garrido, J.L. Valverde, J.F. Rodrguez, M. Carmona, Develop-
Build. Eng. 31 (September 2019) (2020) 101400, doi:10.1016/j.jobe.2020.101400. ment of smart gypsum composites by incorporating thermoregulating micro-
[40] L. Qian, X. Pang, J. Zhou, J. Yang, S. Lin, D. Hui, Theoretical model and finite ele- capsules, Energy Build. 76 (2014) 631–639, doi:10.1016/j.enbuild.2014.03.005.
ment simulation on the effective thermal conductivity of particulate composite
materials, Compos. Part B 116 (2017) 291–297, doi:10.1016/j.compositesb.2016.
10.067.

11