Journal Pre-proofs

Comparative study on natural convection melting in square cavity using lat‐

tice Boltzmann method

Ziliang Rui, Juan Li, Jie Ma, Han Cai, Binjian Nie, Hao Peng

PII: S2211-3797(20)31741-1
DOI: https://doi.org/10.1016/j.rinp.2020.103274
Reference: RINP 103274

To appear in: Results in Physics

Received Date: 24 April 2020

Revised Date: 16 July 2020
Accepted Date: 24 July 2020

Please cite this article as: Rui, Z., Li, J., Ma, J., Cai, H., Nie, B., Peng, H., Comparative study on natural
convection melting in square cavity using lattice Boltzmann method, Results in Physics (2020), doi: https://
doi.org/10.1016/j.rinp.2020.103274

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover
page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version
will undergo additional copyediting, typesetting and review before it is published in its final form, but we are
providing this version to give early visibility of the article. Please note that, during the production process, errors
may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

© 2020 Published by Elsevier B.V.

 Comparative study on natural convection melting in square cavity
using lattice Boltzmann method
Ziliang Ruia, Juan Lib, Jie Maa, Han Caia, Binjian Niec, Hao Penga*
a Jiangsu Key Laboratory of Process Enhancement and New Energy Equipment Technology, School of Mechanical and Power

Engineering, Nanjing Tech University, No. 30 Pu Zhu South Road, Nanjing 211816, P. R. China
b School of Mechanical and Electrical Engineering, Nanjing Forestry University, No. 159 Long Pan Road, Nanjing 210037,

P. R. China
c School of Chemical Engineering, University of Birmingham, Birmingham B15 2TT, UK

Abstract:
The phase change process is of significant importance in the application of phase change material (PCM). In

this paper, the natural convection melting in a square cavity was investigated and the enthalpy-based lattice

Boltzmann model (TLBM) combined with the pseudo-potential LB model was developed to trace the solid-liquid

interface. Four cases with different boundary conditions were calculated and the corresponding temperature

contours, velocity vectors and average liquid fraction at various time were compared.

Results show that TLBM has good accuracy in simulating the solid-liquid phase transition process. Variation

in boundary conditions has appreciable impact on the natural convection process and melting behavior. Comparing

model.1 (single heating wall) with model.2 (double adjacent heating walls with cold wall), the maximum liquid

fraction is increased 16.7% in model.2 and the additional top heating wall acts to accelerate the average melting

rate by 16.7%. The comparison between model.3 (double opposite heating walls) and model.4 (double adjacent

heating walls without cold wall) shows that the natural convection heat transfer is intensive in model.3 and

resulting in a 280% faster melting rate than that in model.4. Comparing the melting rate between the model.5 (all

heating walls) and model.3, the increase of 54% due to the more heating walls in model.5 is observed.

Key words: Natural convection, phase change material, Lattice Boltzmann, liquid fraction, melting behavior

Corresponding author: Hao Peng

Email address: phsight1@hotmail.com (H. Peng)

1
 Nomenclature

c lattice speed [m∙s-1] Fo Fourier number

ci discrete velocity in direction i [m∙s-1] Pr Prandtl number

cp specific heat [kJ∙kg-1∙°C-1] Ste Stefan number

density distribution function in direction

fi Ra Rayleigh number
i [kg∙m-3]

equilibrium density distribution

fieq Greek symbols
function [kg∙m-3]

Fi natural convection force [N] α thermal diffusivity [m2∙s-1]

fl liquid fraction β coefficient thermal expansion [ºC-1]

G gravity acceleration [m∙s-2] λ thermal conductivity [W∙m-1∙°C-1]

temperature distribution function in

gi u velocity[m∙s-1]
direction i [kJ∙kg-1]

equilibrium temperature distribution

gieq ν kinematic viscosity [m2∙s-1]
function in direction i [kJ∙kg-1]

H total enthalpy [kJ∙kg-1] ωi weight coefficient in direction i

total enthalpy corresponding to the

Hl ρ density [kg∙m-3]
liquid temperature [kJ∙kg-1]

total enthalpy corresponding to the solid dimensionless relaxation time for density
Hs τf
temperature [kJ∙kg-1] distribution function

dimensionless relaxation time for

L latent heat of the phase change [kJ∙kg-1] τg
temperature distribution function

t time [s] Acronyms

T temperature [ºC] LB Lattice Boltzmann

Tc temperature of cold wall [ºC] PCM Phase change material

Th temperature of heating wall [ºC] Subscripts

Tl temperature of liquid phase [ºC] i direction

Tm melting temperature [ºC] l liquid

Ts temperature of solid phase [ºC] s solid

2
1. Introduction

The phase change materials (PCM) are receiving significant attention for thermal energy storage in recent

years [1]. A small volume of PCMs can store massive energy during solid-liquid phase-change process[2]. With

the ability to absorb and release latent heat when undergoing solid-liquid phase change process, PCM have been

recommended to apply in waste heat recovery[3], liquid-air energy storage system(microchannel[4], tank[5],

packed bed units[6] ), thermochemical energy storage system[7], cryogenic thermal energy storage system[8] and

other thermal management systems [9]. Furthermore, Microencapsulated Phase Change Material (MPCM) based

thermal energy storage technology has the potential to reduce the energy consumption of the buildings[10], by

improving the indoor thermal comfort, when integrated directly into the building materials[11]. The solid–liquid

phase transition is the key issue in analyzing PCM thermal energy storage system [12] especially the latent heat

thermal system [13]. The moving boundary of solid–liquid interface[14] and the complicated heat transfer

mechanisms of multiphase is one of the main obstacles to solving the problem [15]. Consequently, many numerical

models are established to investigate phase transition process and obtain accurate results. Among a series of 2D

and 3D models such as T-shaped enclosure[16], triangular chamber[17], and porous enclosure[18] or chamber[19],

the 2D square cavity is one of the most used classic models. Due to its high computing efficiency and intuitive

results, the model has been applied to solve many complicated thermal systems.

The Lattice Boltzmann method, as a relatively new numerical model, is reliable in solving solid-liquid phase

change problem. Compared with traditional numerical models, its obvious advantages such as the mesoscopic

kinetic background, easy boundary treatment and inherently parallelizable computation property make it an

efficient way for solving the energy equations.

Three types of models have been applied in phase transition simulation: phase field LB method, interfacial

tracking method and enthalpy based LB method. In the phase field LB model, a governing equation for the order

parameter is introduced to capture the phase interface. Based on the phase field lattice Boltzmann, Abadi and

Rahimian [20]applied a hybrid lattice Boltzmann finite difference model to simulate the phase transition in a

ternary system. The model was validated against available results. In the interfacial tracking model, the movement

of phase interface was explicitly tracked by the Stefan condition. Huang and Wu [21]proposed an Immersed

Boundary-LBM for the simulation of solid-liquid phase change problem. The momentum force term and energy

force term were introduced to treat the solid-liquid interface. The simulation results were satisfactorily consistent

with analytical solutions. Li and Wu [22]established the LBM with interfacial tracking method to investigate the

melting and solidification of thin gold film and compared the results with those of Two-Temperature Model (TTM).

3
The results of the two models agreed well with the experimental data. The phase field model and interfacial

tracking model can basically solve the phase transition problem, however, the treatments of velocity and

temperature boundary conditions on the phase interface were quite complex in these two models.

The enthalpy-based LB model was introduced to simplify the phase interface treatment. Jiaung et al.

[23]proposed the first enthalpy-based LB model for the phase transition problem with heat conduction. In this

model, the total enthalpy was divided into two different variables: latent heat and sensible enthalpy. The latent

heat term was calculated as a heat source term in temperature field governing equation and an iteration procedure

was introduced to solve the implicit lattice Boltzmann equation. The phase interface and total enthalpy were

updated simultaneously. Several phase change simulations of different dimensions were established to validate

this model and the results were convincing. Xu et al. [24]extended the two-dimensional enthalpy-based lattice

Boltzmann model to treat the axisymmetric solid-liquid phase change problems and obtained highly accurate

results. Based on the enthalpy-based LBM, Lu et al. [25]proposed an optimal two-relaxation-time (OTRT) Lattice

Boltzmann Equation (LBE) and studied five cases including one-dimensional to three-dimensional solid-liquid

phase change process. Results showed that the OTRT LBE has a smaller numerical error than other LBE in a large

range of relaxation time. Ren et al. [26]used the enthalpy-based multiple-relaxation-time LBM to study the phase

change process of PCM in a latent heat thermal energy storage system and found that replacing rectangle fins with

triangle fins improved the charging rate of PCM. Eshraghi and Felicelli [27]proposed an implicit LB model in

order to simplify the calculating procedure. They employed their model to simulate two-dimensional solidification

of a binary alloy and compared computational performance with FEM. The results of implicit LBM agreed well

with those of FEM. Huang and Wu [28]proposed the total enthalpy-based LB model, and the latent heat term was

combined with the time derivative term in this model in order to improve the stability and computational efficiency.

The temperature field equilibrium distribution function was modified to recover the correct governing equation.

Hu et al. [29]incorporated this model with D3Q19 lattice and simulated several three-dimensional solid-liquid

phase change problems. The model was validated under the condition that the solid phase is unfixed. Huang and

Wu [30]investigated the phase interface effects for this LB model and then improved the total enthalpy-based

model by introducing the multiple-relaxation-time (MRT) collision term to solve the velocity equation [31]. Luo

et al. [32] used the total enthalpy-based LB model to investigate the influence of various numbers and arrangements

of tubes on the melting dynamics of shell and tube models under different Rayleigh and Stefan numbers. The

improved model was demonstrated to be accurate and efficient by studying a few melting cases.

Although the above LB models have been used widely for phase change problem, it can be found that those

4
simulation cases for the 2D phase transition in square cavity were based on simple boundary condition which

includes top adiabatic wall, bottom adiabatic wall and heating wall with opposite cold wall. The influences of the

number and position of heating wall on melting behavior were still insufficient and the physical principles that led

to those effects were not specifically explained. In addition, the effect of cold wall was usually ignored, further

investigations were needed. Therefore, in present work, the cases of natural convection melting process in a square

cavity with four different numbers and positions of heating walls were studied. The difference in the numbers and

positions of cold walls was also considered in these models. The enthalpy-based TLBM model combined with the

pseudo-potential LB model was developed to trace the solid-liquid interface and the BGK approximation with

D2Q9 Lattice was applied to simulate the temperature and the velocity fields. The temperature contours, velocity

vectors and average liquid fraction at various times in these different cases were discussed and analyzed. The

physical principles behind the effect of different boundary conditions were clearly revealed.

5
2. Lattice Boltzmann model

2.1 Basic thermal model

A computational domain of a square cavity was constructed in this work.

Adiabatic

Liquid

T=Th

T=Tc
Solid
Model.1
Adiabatic

Fig.1 Model of phase transition in square cavity

The basic model (model.1) is shown in Fig.1, the cavity is filled by solid pure substance. The melting

temperature of the pure substance is Tm. The initial temperature of the cavity is Tc, the temperature of left boundary

is Th, and the top and the bottom walls are adiabatic.

The basic assumptions are:

(1) the liquid phase of the pure substance is incompressible.

(2) physical properties such as density, specific heat and thermal conductivity of both liquid and solid phase

are constant and equal.

(3) the liquid viscosity is constant.

(4) the solid phase is fixed to the boundary

For the comparative study, four models(model.2-5) with different boundary conditions were established, as

shown in Fig.2

T=Th Adiabatic

Liquid Liquid Liquid

T=Th

T=Th
T=Th

T=Tc

Solid
Solid
Model.2 Model.3
Adiabatic Adiabatic

(a)Double adjacent heating walls (with cold wall) (b) Double opposite heating walls

6
 T=Th T=Th

Liquid Liquid

Adiabatic

T=Th

T=Th
T=Th

Solid

Solid

Model.4 Model.5
Adiabatic T=Th

(c) Double adjacent heating walls (no cold wall) (d) All heating walls

Fig.2 Schematic of the four models

2.2 Conservation equations

Enthalpy-based heat equation can be written as [32]:

   c pT     Lf l 
     c pTu      T        Lf l u  (1)
t t

For a pure substance, the third term on the right side of Eq (1) is usually ignored, Eq (1) can be simplified to:

H 
     c pTu      T  (2)
t

The total enthalpy is calculated as [32]:

H  c pT  Lf l (3)

2.3 Lattice Boltzmann equation for the thermal field

In present study, 2D Lattice Boltzmann model with 9 velocity vectors was used(D2Q9). The corresponding

discrete velocities are [30]:


(0, 0) i =0

 k  k 
ci  (cos(  ),sin(  ) i =1-4 (4)
 2 2 2 2
 k 9 k 
 2(cos( 2  4 ),sin( 2  4 )) i =5-9

and the weight factors are [30]:

4 / 9 i0

i  1/ 9 i  1-4 (5)
1/ 36 i  5-9


The Enthalpy-based lattice Boltzmann methods ‘TLBM’ is used for solid-liquid phase transitions. TLBM

uses temperature T as basic evolution variable in the LB equation, and the phase transition term is considered as

7
the sink term in collision step.

For the TLBM, the evolution equation for the temperature distribution function is expressed as [28]:

1 L
gi ( x  ci t , t  t )  gi ( x, t )    gi ( x, t )  gieq ( x, t )   twi  f l ( x, t )  f l ( x, t  t ) 
 (6)
g cp

the relaxation time τg is calculated by [28]:

3 1
g   (7)
c t 2
2

Equilibrium distribution function for temperature field can be expressed as [28]:

ci  u (ci  u ) 2 u 2
gieq  wiT  (1    2) (8)
c2 2c 4 2c

and temperature T can be evaluated from [28]:

9
1
T
cp
g
i 0
i (9)

The local enthalpy is obtained by the liquid fraction [28]:

H  c pT  f l  t  t  L (10)

The liquid fraction is obtained by the linearly interpolating of enthalpy [28]

0 H  H s  c pTc

 H  Hs
f l  x, t    H s  H  Hl  H s  L (11)
 Hl  H s
1 H  Hl


2.4 Lattice Boltzmann equation for the velocity field

Natural convection is considered and the Lattice Boltzmann equation for the natural convection source term

can be expressed as [28]:

1
f i ( x  ci t , t  t )  f i ( x, t )  ( f i eq ( x, t )  f i ( x, t ))+tFi ci (12)
f

the relaxation time for the velocity field is [28]:

3 1
f   (13)
c 2 t 2

The equilibrium density distribution function f i eq is calculated by [28]:

ci  u (ci  u ) 2 u 2
f i eq ( x, t )  wi  (1    2) (14)
c2 2c 4 2c

The natural convection source term Fi in vertical direction is computed by [28]:

8
 Th Tc
Fi  3wi  G  (T   ) (15)
2 2

3. Results and discussions

3.1 Melting behaviors for single heating wall model

The Prandtl number (Pr), Rayleigh number (Ra) Stefan number (Ste) and Fourier number (Fo) are the four

key dimensionless parameters to characterize the melting behavior in the present work, as listed below [33]:

Pr   /  (16)

Ra  G  (Th  Tc)l 3 /  (17)

Ste  c p (Th  Tc) / L (18)

Fo   t / l 2 (19)

To be noticed, in TLBM, the temperature can locally decrease below the initial solid temperature when the

sink term dominates at low Stefan number (St), which is in contradiction with the phase change associated with

pure substance melting. To avoid this situation, the value of St needs to be controlled at suitable level. Furthermore,

the pseudo-potential LB model was combined with TLBM model so the boundary treatment on the phase interface

was not required. To validate the present LB model, the results obtained in the present work was compared with

that in Mencinger’s work[34]. In his work, the Ra, Pr, and Ste number were 25000, 0.02, and 0.01 respectively,

and the dimensionless parameters in present model were set to the same values. Comparison of the average liquid

fraction(fl) between present LB model and Mencinger’s results is shown in Fig.3. The maximum error of fl is

1.96% and the average error is around 1.54% which proves that the present LB model is acceptable.

1.0
Mencinger's work[27]
0.8 Present model

0.6
fl

0.4

0.2

0.0
0 2 4 Fo 6 8 10

Fig.3 Comparison of the average fl between present LB model and previous work

For all the following models, the dimensionless parameters were set to be Ra=25000, Pr =0.02 and Ste=0.5.

9
All the results are obtained on the 50×50 square grid. The Fourier number (Fo) is used to represent the time step.

The LBM codes of this paper were solved in Visual Basic. Fig.4(a) shows the temperature contours and velocity

vector graphs (the black lines) of model.1. As shown in this figure, at Fo=0.4, the phase interface is nearly vertical

and the distribution of the velocity vector is generally regular. The reason may be attributed to the following. The

conduction effect plays the main role and the natural convection term can be neglected at the initial stage of the

melting process. While Fo=1, the phase interface is crooked and the solid material in the upper part of the cavity

melts faster than that in the lower part. It can also be noticed that a whirlpool caused by the natural convection

appears in the upper part of the flow field. While Fo increases to 2 and 4, the difference of the melting rate in the

upper and the lower part becomes pyramidally obvious, and the whirlpool in the flow field increases sharply. It

can be concluded that as the melting process goes on, the effect of natural convection becomes significant. Fig.4(b)

shows the average liquid fraction in single heating wall melting process. It can be found that the melting rate slows

down with the increase of the distance between solid-liquid interface and heating wall. Moreover, since the

temperature of cold wall is lower than the melting temperature, the solid materials in the cavity cannot melt

completely and the maximum value of average liquid fraction is about 0.6.

Temperature
1. 0
0.9
0.8
0.7
Fo=0.4 Fo=1
0.6
0.5
0.4
0.3
0.2
0.1
0.0

Fo=2 Fo=4

Fig.4(a) Temperature contours and velocity vector graphs of model.1

10
 1.0

0.8
model.1

0.6

fl
0.4

0.2

0.0
0 1 Fo 2 3 4

Fig.4(b) Average fl in model.1: single heating wall melting process

3.2 Melting behaviors for double heating walls models

In this section, the effect of double heating walls (model.2, model.3 and model.4) on the melting behaviors

(temperature contour, velocity vector and average liquid fraction) was investigated. The corresponding boundary

conditions were shown in Fig.2(a), Fig.2(b) and Fig.2(c), respectively.

In Fig.2(a), the top and the left wall were both heating walls (model.2). Fig.5 shows the simulation results of

the temperature and velocity distribution. When Fo=0.4, the melting process in the upper area of the cavity is

significantly accelerated compared with Fig.4(a). The natural convection phenomenon in the velocity field is

clearly observed at Fo=1. As illustrated in figures of Fo=1, 2 and 4, the temperature contours and velocity vectors

in the lower region are almost the same with those in Fig.4(a), which indicates that the melting behavior in the

lower area in model.2 is similar with that in model.1. It can be concluded that the top heating wall accelerates the

melting rate of the upper solid phase but it has little influence on the phase transition in the lower area.

Temperature
1. 0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
Fo=0.4 Fo=1 0.0

11
 Fo=2 Fo=4

Fig.5 Temperature contours and velocity vector graphs of model.2

In Fig.2(b), the right cold wall was replaced by a heating wall. As seen from the simulation results in Fig.6,

the temperature contours and velocity vector graph are vertically symmetrical. While Fo=0.4, the temperature

contour and velocity vector in the left half region of model.3 are almost the same as those of model.1. Isotherms

in central area are in shapes of peak curve when Fo=0.6. Two whirlpools can be seen at this moment too, which

means the convection term acts on both left and right sides. While Fo=0.8, the left and right liquid phase of the

upper region gets in contact with each other and the solid-liquid interface is in shape of peak curve. While Fo=1

the cavity is nearly filled with liquid which means the phase changing process is almost complete.

Temperature
1. 0
0.9
0.8
0.7
0.6
Fo=0.4 Fo=0.6
0.5
0.4
0.3
0.2
0.1
0.0

12
 Fo=0.8 Fo=1

Fig.6 Temperature contours and velocity vector graphs of model.3

In Fig.2(c), the right cold wall is replaced by an adiabatic wall. The top and right walls are heating walls.

From the results in Fig.7, while Fo≤2, the temperature contour, velocity vector and phase interface are diagonally

symmetric and the shapes of both phase interface and isotherms are getting closer to quarter circle. However, the

symmetry does not exist anymore at Fo=3, and the cavity is almost filled with liquid. This is because that the

gravity force only has little effect on liquid phase while the solid phase exists. With the melting process pushing

on, the area of solid phase decreases and the effect of gravity increases. Furthermore, the whirlpool in this model

is obviously larger than other models due to the influence of top heating wall and right adiabatic wall.

Temperature
1. 0
0.9
0.8
0.7
Fo=0.4 Fo=1
0.6
0.5
0.4
0.3
0.2
0.1
0.0

Fo=2 Fo=3

Fig.7 Temperature contours and velocity vector graphs of model.4

In order to quantitatively analyze the melting behaviors, the comparison of average liquid fraction fl for

models 1~4 is depicted in Fig.8.

The comparison between mode.1 and model.2 proves that the top heating wall has little effect on the variation

rate of the melting speed. It can also be noticed that the maximum fl of model.2 increases to 0.7 which means the

action of top heating wall increases the average melting rate by 16.7%. Comparing model.4 to model.2, when

13
Fo=3.1, the fl of model.4 and model.2 are 1 and 0.68 respectively. In other words, the average melting rate of

model.4 increases by 84.33% compared with that of model.2. That is because the cold wall seriously hinders the

melting behavior. According to the comparison between model.3 and model.4, the melting end time of model.3

and model.4 are Fo =3.1 and Fo=1.1 respectively. It can be observed that the average melting rate of opposite

heating walls melting is 280% faster than that of adjacent heating walls melting. The explanation can be given for

this phenomenon. The difference in temperature caused by left or right heating wall induce an extra driving force

in upward direction. The upward driving force accelerates the movement of liquid phase in the cavity and enhance

the natural convection heat transfer. However, the driving force arisen from the top heating wall only acts on the

liquid phase near the top wall because of gravity force. The conclusion can be drawn that the natural convection

heat transfer is intensive under the action of right heating wall. Moreover, the trends of average liquid fraction

show that the melting rate in model.3 accelerates during Fo=0.6 to 0.8 which is obviously different from those of

the other three models. The reason can be attributed to the natural convection phenomenon caused by left and right

heating wall becomes especially obvious when the left and right phase interface get in contact with each other.

1.0
model.4
0.8 model.3

0.6

model.2
fl

0.4 model.1
Fo=3.1
0.2
Fo=1.1

0.0
0 1 2 3 4
Fo

Fig.8 Average fl comparison of model.1: single heating wall melting, model.2: double adjacent heating walls
melting (with cold wall), model.3: double opposite heating walls melting and model.4: double adjacent heating
walls melting (no cold wall)

3.3 Melting behaviors for all heating walls model

As can be seen in Fig.2(d), the cavity is heating from all directions. Observing from the results in Fig.9(a),

while Fo=0.4, the isotherms near the center point and the solid-liquid interface are almost circular. As the Fo

increases to 0.6, the circular phase interface becomes elliptic because of the faster melting rate in the upper area

and the gravity force. In addition, from the comparison of average liquid fraction in Fig.9(b), the sudden change

in melting rate does not exist in all heating walls melting process. The finish time of model.5 and model.3 are 0.68

14
and 1.05 respectively which means the additional heating walls in model.5 increase the melting rate by 54%.

Temperature
1. 0
0.9
0.8
0.7
Fo=0.4 Fo=0.6
0.6
0.5
0.4
0.3
0.2
0.1
0.0

Fo=0.64 Fo=0.68

Fig.9(a) Temperature contours and velocity vector graphs of model.5

1.0

0.8

0.6 model.5
model.3
fl

0.4
Fo=1.1
0.2
Fo=0.68
0.0
0.0 0.5 1.0 1.5 2.0 2.5
Fo
Fig.9(b) Comparison of average fl between model.3: double opposite heating walls melting and model.5: all

heating walls melting

4.Conclusion

In this paper, the enthalpy-based TLBM model has been developed to simulate the 2D natural convection

15
melting process in square cavity. For all examples with mixed boundary conditions, good accuracy is demonstrated.

The key findings are shown below:

(1) In model.2 (double adjacent heating walls with cold wall), the top heating wall accelerates the melting

rate of the upper solid phase but has little influence on the phase transition in the lower area. The maximum liquid

fraction in model.2 has a 16.7% augment and the additional top heating wall expedites the average melting rate by

16.7%.

(2) The comparison between model.3 (double opposite heating walls) and model.4 (double adjacent heating

walls without cold wall) shows that the difference in diving forces arisen form different heating walls causes the

enhancement in natural convection heat transfer and a 280% faster melting rate in model.3. Moreover, when the

left and right phase interface get in contact with each other, the enhancement in heat transfer caused by left and

right heating wall becomes especially obvious.

(3) Comparing model.2 with model.4, the average melting rate of model.4 is 84.33% faster than that of

model.2 because the cold wall seriously hinders the melting behavior.

(4) The comparison between model.5 (all heating walls) and model.3 demonstrates that the additional heating

walls increase the melting rate by 54%.

Acknowledgement

The authors acknowledge the financial support provided by National Natural Science Foundation of China

(Grant No. 51776095) and Young Science Leaders Project of Jiangsu Province.

Reference

[1].Huo, Y. and Z. Rao, Lattice Boltzmann simulation for solid–liquid phase change phenomenon of phase change

material under constant heat flux. International Journal of Heat and Mass Transfer, 2015. 86: p. 197-206.

[2].Sheikholeslami, M., Numerical simulation for solidification in a LHTESS by means of nano-enhanced PCM.

Journal of the Taiwan Institute of Chemical Engineers, 2018. 86: p. 25-41.

[3].Salih, S.M., J.M. Jalil, and S.E. Najim, Experimental and numerical analysis of double-pass solar air heater

utilizing multiple capsules PCM. Renewable Energy, 2019. 143: p. 1053-1066.

[4].Ding, M., C. Liu, and Z. Rao, Experimental investigation on heat transfer characteristic of TiO2-H2O nanofluid

in microchannel for thermal energy storage. Applied Thermal Engineering, 2019. 160: p. 114024.

[5].Peng, H., et al., n-Alkanes phase change materials and their microencapsulation for thermal energy storage: a

Critical Review. Energy Fuels,2018. 32(7): p.7262-7293

16
[6].Peng, H., X.K. Shan, Y. Yang, and X. Ling, A study on performance of a liquid air energy storage system with

packed bed units. Applied Energy,2018. 211: p.126-135

[7].Chen, X.Y, et al., State of the art on the high-temperature thermochemical energy storage systems. Energy

Conversion and Management, 2018. 177: p.792-851

[8].Shen, T., H. Peng, and X. Ling, Experimental Measurements and Thermodynamic Modeling of Melting

Temperature of the Binary Systems n-C11H24–n-C14H30, n-C12H26–n-C13H28, n-C12H26–n-C14H30, and n-

C13H28–n-C15H32 for Cryogenic Thermal Energy Storage. Industrial & Engineering Chemistry Research, 2019.

58(32): p. 15026-15035.

[9].Shen, T., et al., Experimental study and thermodynamic modeling of solid-liquid equilibrium of binary systems:

Dodecane-tetradedcane and tridecane-pentadecane for cryogenic thermal energy storage. Fluid Phase Equilibria,

2019. 493: p. 109-119.

[10].Sheikholeslami, M., et al., Acceleration of discharge process of clean energy storage unit with insertion of

porous foam considering nanoparticle enhanced paraffin. Journal of Cleaner Production, 2020. 261:121206.

[11]. Pushpendra K.S.R., K.S. Shailendra, K.G. Naveen, Potential of microencapsulated PCM for energy savings

in buildings: A critical review. Sustainable Cities and Society,2020,53.

[12].Chekmyshev, K.E. and P.G. Ovcharenko, Numerical simulation of bimetallic casting cooling during the

process of lost foam casting. Journal of Crystal Growth, 2019. 527: p. 125243.

[13].Rathore, P.K.S. and S.K. Shukla, Potential of macroencapsulated PCM for thermal energy storage in buildings:

A comprehensive review. Construction and Building Materials, 2019. 225: p. 723-744.

[14].Fu, W., et al., Preparation and properties of phase change temperature-tuned composite phase change material

based on sodium acetate trihydrate–urea/fumed silica for radiant floor heating system. Applied Thermal

Engineering, 2019. 162: p. 114253.

[15].Charvát, P., et al., Solar air collector with the solar absorber plate containing a PCM – Environmental chamber

experiments and computer simulations. Renewable Energy, 2019. 143: p. 731-740.

[16].Izadi, M., et al., Coupled FHD–MHD free convection of a hybrid nanoliquid in an inversed T-shaped

enclosure occupied by partitioned porous media. Numerical Heat Transfer, 2019. 76(6): p. 479-498.

[17]. Izadi, M., Effects of porous material on transient natural convection heat transfer of nano-fluids inside a

triangular chamber. Chinese Journal of Chemical Engineering, 2020. 28: p. 1203-1213.

[18]. Izadi, M., et al., LTNE modeling of Magneto-Ferro natural convection inside a porous enclosure exposed to

nonuniform magnetic field. Physica A, 2019. 535: 122394.

17
[19]. Izadi, M., et al., MHD thermogravitational convection and thermal radiation of a micropolar nanoliquid in a

porous chamber. International Communications in Heat and Mass Transfer, 2020. 110: 104409.

[20].Haghani Hassan Abadi, R. and M.H. Rahimian, Hybrid lattice Boltzmann finite difference model for

simulation of phase change in a ternary fluid. International Journal of Heat and Mass Transfer, 2018. 127: p. 704-

716.

[21].Huang, R. and H. Wu, An immersed boundary-thermal lattice Boltzmann method for solid–liquid phase

change. Journal of Computational Physics, 2014. 277: p. 305-319.

[22].Li, L., M. Wu, and L. Zhou, Lattice Boltzmann Simulations for Melting and Resolidification of Ultrashort

Laser Interaction with Thin Gold Film. International Journal of Thermophysics, 2018. 39(7).

[23].Jiaung, W.S., J.R. Ho, and C.P. Kuo, Lattice Boltzmann method for the heat conduction problem with phase

change. Numerical Heat Transfer Part B-Fundamentals, 2001. 39(2): p. 167-187.

[24].Xu, X., et al., Axisymmetric lattice Boltzmann simulation of the heat-exchanger method-based sapphire

crystal growth. International Journal of Heat and Mass Transfer, 2019. 129: p. 299-309.

[25].Lu, J.H., H.Y. Lei, and C.S. Dai, An optimal two-relaxation-time lattice Boltzmann equation for solid-liquid

phase change: The elimination of unphysical numerical diffusion. International Journal of Thermal Sciences, 2019.

135: p. 17-29.

[26].Ren, Q., H. Xu, and Z. Luo, PCM charging process accelerated with combination of optimized triangle fins

and nanoparticles. International Journal of Thermal Sciences, 2019. 140: p. 466-479.

[27].Eshraghi, M. and S.D. Felicelli, An implicit lattice Boltzmann model for heat conduction with phase change.

International Journal of Heat and Mass Transfer, 2012. 55(9): p. 2420-2428.

[28].Huang, R., H. Wu, and P. Cheng, A new lattice Boltzmann model for solid–liquid phase change. International

Journal of Heat and Mass Transfer, 2013. 59: p. 295-301.

[29].Hu, Y., et al., Lattice Boltzmann simulation for three-dimensional natural convection with solid-liquid phase

change. International Journal of Heat and Mass Transfer, 2017. 113: p. 1168-1178.

[30].Huang, R. and H. Wu, Phase interface effects in the total enthalpy-based lattice Boltzmann model for solid–

liquid phase change. Journal of Computational Physics, 2015. 294: p. 346-362.

[31].Huang, R. and H. Wu, Total enthalpy-based lattice Boltzmann method with adaptive mesh refinement for

solid-liquid phase change. Journal of Computational Physics, 2016. 315: p. 65-83.

[32].Luo, K., et al., Lattice Boltzmann simulation of convection melting in complex heat storage systems filled

with phase change materials. Applied Thermal Engineering, 2015. 86: p. 238-250.

18
[33]. Izadi, M., Ghalambaz, M., Mehryan, S.A.M., Location impact of a pair of magnetic sources on melting of a

magneto-Ferro phase change substance. Chinese Journal of Physics, 2020. 65: p. 377-388.

[34].Mencinger, J. and H. Wu, Numerical simulation of melting in two-dimensional cavity using adaptive grid.

Journal of Computational Physics, 2004. 198: p. 243-264.

19
 Credit author statement
Rui Ziliang: Conceptualization, Methodology, Software, Data Curation, Writing - Original
Draft.
Li Juan: Data curation, Formal analysis.
Ma Jie: Writing- Reviewing and Editing.
Han Cai: Writing- Reviewing and Editing.
Nie Binjian: Writing- Reviewing and Editing, Supervision, Validation.
Peng Hao: Writing- Reviewing and Editing, Resources, Project administration, Funding
acquisition

20
 Highlights
 The TLBM model was developed to simulate the 2D natural convection melting.

 Four cases with various numbers and positions of heating walls were calculated.

 For all cases with mixed boundary conditions, good accuracy is demonstrated.

 Variation in boundary condition affects the natural convection melting appreciably.

21