Applied Thermal Engineering 236 (2024) 121910

Contents lists available at ScienceDirect

Applied Thermal Engineering

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

Research Paper

Innovative experimental approach for the dynamic Multi-Variable

investigation of Pulsating heat Pipes
Mauro Abela a, b, *, Mauro Mameli b, Sauro Filippeschi b, Brent S. Taft c
a
Service de Physique de l’Etat Condensé, CEA, CNRS, Université Paris-Saclay, CEA Saclay, Gif-sur-Yvette Cedex, France
b
Department of Energy, Systems Land and Construction Engineering, University of Pisa, Largo L. Lazzarino, Pisa, Italy
c
U.S. Air Force Research Laboratory, Kirtland Air Force Base, NM 87117, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Pulsating Heat Pipes (PHP) are passive two-phase heat transfer devices characterized by a simple structure and
Pulsating Heat Pipes high heat transfer capabilities. Despite this, their large-scale application is still hindered by the actual unpre­
Startup dictability of their dynamic behavior during the startup and the thermal crisis. An innovative experimental
Condenser temperature
apparatus is designed to systematically investigate the above-mentioned phenomena. It consists in a square loop
Number of evaporator sections
made of four borosilicate transparent glass tubes joined at corners by means of brass connectors. The external
Number of condenser sections
Fluid Distribution tube surface is coated with several transparent Indium Tin Oxide heaters. The device is used to topologically
Fluid Visualization reproduce 5, 7, and 11 evaporator/condenser sections PHPs with a 2 mm inner diameter tube, filled with pure
Indium Tin Oxide ethanol and tested in horizontal position. First a steady-state characterization is performed. The condenser
temperature is varied from 10 ◦ C to 40 ◦ C; the input power goes from 10 W to 40 W. Results show that both the
increase of number of evaporator sections and condenser temperature enhance the PHP performance in terms of
equivalent thermal resistance. Then, the effect of the condenser temperature on the initial liquid phase distri­
bution is analyzed. It is observed that at low condenser temperatures (10 ◦ C, 20 ◦ C), the liquid phase tends to
gather in the condenser sections rather than the evaporator sections and it is arranged in short liquid plugs;
conversely, for high condenser temperatures (30 ◦ C, 40 ◦ C), the amount of liquid phase in evaporators is higher
and it is arranged in long liquid plugs. Finally, the transient behavior is studied in terms of startup time. It is
observed that the parameter influencing most the startup time is the initial fluid distribution; the startup time
increases with the increase of evaporator sections volume occupied by liquid phase and with the length of liquid
plugs; on the other hand, the startup time decreases with the increase of number of menisci in the evaporator
sections.

different authors. Indeed, the PHP startup is affected by several pa­

rameters such as: the number of heating sections (often ambiguously
1. Introduction called “number of turns” in the literature; in the present work it is indeed
replaced with “number of evaporator sections”), the orientation with
The need of efficient and reliable heat transfer devices channeled the respect to gravity, the filling ratio, the condenser temperature, the wall
researchers’ attention towards a relatively new technology: the Pulsat­ to fluid heat flux at the evaporator, the initial fluid distribution, and so
ing Heat Pipe (PHP) [1]. It consists in a capillary tube bent in several on. For example, if on one hand it is generally accepted there exists a
turns, vacuumed, and partially filled with a working fluid. Despite its critical number of evaporator sections that ensures the startup occur­
advantages with respect to the standard Heat Pipes (simple structure, rence, different authors find different critical numbers and generally the
ability to cover complex and wide geometries, foldability and so on), the analysis lacks of a clear physical explanation.
PHP is still facing the hesitation of the industry since its operational Khandekar et al. [2] conducted experiments on a PHP made of
boundaries are not clearly defined in terms of optimal start-up condi­ copper using water, ethanol and R-123 as working fluids. The device is
tions. The attempts done so far in the literature to define quantitative tested in various orientation and results show that it does not operate in
criteria for the PHP startup do not consider the interplay of all the horizontal mode because of the small number of evaporator sections and
variables affecting the PHP operation bringing to different outcomes by

* Corresponding author.
E-mail address: [email protected] (M. Abela).

https://doi.org/10.1016/j.applthermaleng.2023.121910
Received 6 June 2023; Received in revised form 26 September 2023; Accepted 1 November 2023
Available online 5 November 2023
1359-4311/© 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
M. Abela et al. Applied Thermal Engineering 236 (2024) 121910

Nomenclature ε Emissivity [-]

T Temperature [◦ C] Superscripts
E evaporator
Q̇ Input Power [W]
C condenser
n Number [-]
FL fluid
Cp Heat Capacity [J/K]
sys Systematic
S Cross Section [m^2]
rnd random
h Heat Transfer Coefficient [W/m^2 K]
Req Equivalent Thermal Resistance [K/W] Subscripts
P Pressure [Pa] c convective
I Current [A] exp experimental
V Volume [m^3], Voltage [V] eff effective
t Time [s] a ambient
llmax Maximum Liquid Plug length [m] su startup
k Thermal conductivity [W/m K] t tube
C Tuning Parameter [J/m^2 K] e external
d Diameter [m] i internal
ITO Indium Tin Oxide
Greek Symbols G Borosilicate glass
ρ Density [kg/m^3]
σ Stefan-Boltzmann constant [W/m^2 K^4]

low operating pressures. Charoensawan et al. [3]studied the effect of higher than the vapor collapse the system is more prone to dry-out
orientation on PHP filled with different working fluids (ethanol, R123a, phenomena. The study also introduces a mathematical model predict­
water) for various number of evaporator sections (from 5 to 23). They ing the ideal startup temperature, which showed a 16 % error range
report critical values of around 16, needed for the full activation of PHP compared to experimental data. Patel et al. [10] experimentally inves­
with an inner diameter of 2 mm. The same authors [4] report that the tigate the working fluid influence on the startup performance of a closed
critical number of evaporator sections depends on evaporators temper­ loop PHP. Authors charge the device at 50 % with 11 different working
ature and the inner diameter of the tube; the higher the number of fluids and conduct tests in vertical bottom heated position. The best
evaporator sections the lower the evaporators temperature needed for results in terms of startup heat flux are shown by acetone while water
the startup. Lee et al. [5] perform a similar study on a PHP filled with has the highest startup heat input. The addition of pure fluids or sur­
FC72 showing that performance becomes independent from orientation factant to the water leads to reduced startup heat flux since they reduce
when the number of evaporator sections is 20. The main reason behind the surface tension. Hao et al. [11] studied the effect on the startup
the apparent incoherency of results reported by different authors is performance of different surface wetting properties (hydrophilic and
related to the fact that results are obtained from different test cells with hydrophobic surfaces). Authors report that the startup performance is
different features and results are often difficult to scale or compare; improved in hydrophilic PHP due to the low flow resistance. In fact, the
therefore, those results were important to trace a roadmap but still lack hydrophilic PHP showed a startup temperature reduced by 4.41 ◦ C and
of general validity. Many other authors deal with the study of the rela­ startup time was reduced by 100 s with respect to the non-treated sur­
tionship between different parameters and the startup occurrence and face PHP. At the same time, the hydrophobic PHP showed a startup
characteristics both experimentally and numerically. Xu et al.[6] tested temperature increase of 2.37 ◦ C and startup time was increase of 100 s
a copper PHP and filled with FC72 observing the existence of two rather compared to the non treated surface PHP. Recently, Zhou et al. [12]
different startup processes: the first is dominated by latent heat ex­ investigated the impact of graphene oxide nanofluid and filling ratios on
change and characterized by sudden temperature overshoot in the the startup performance of a water-based Pulsating PHP. Results show
evaporator; the second, where sensible heat exchange is predominant, is that the addition of graphene nanoparticles could improve the startup
characterized by a smooth oscillations period. Similar observations are performance in terms of temperature and time. Moreover, for constant
reported by Khandekar et al. [7] who built a single loop PHP and tested nanoparticles concentration, the startup performance is enhanced by
it for long durations. Absolute pressure measurements along with tem­ increasing the working fluid filling ratio. Daimaru et al. [13] present a
perature measurements are taken along the tube. They observed two methodology for the numerical modeling of pulsating heat pipe with
types of startups: a sudden startup and a gradual one. The occurrence of check valves. Authors perform the model validation by comparing
a specific type is likely linked to the local void fraction distribution and simulations with on orbit data of a flat plate PHP. Main objective is to
the special two-phase distribution of working fluid. Liu et al. [8] per­ understand the influence of the initial phase distribution on the startup
formed a comprehensive experimental study of PHP startup and used the since experimentally “reproducing and specifying initial vapor–liquid dis­
control theory to make a quantitative estimation of startup performance; tribution in OHPs is virtually impossible”. Results suggested that difficult
effects of filling ratio, fluid thermophysical properties and inclination startup are due to concentration of liquid plugs in the condenser zone as
angle are reported. The optimal filling ratio for water, ethanol and result of cooling during the eclipse. Nikolayev [14] performed a theo­
methanol is reported to be respectively 41 %, 52 % and from 35 % to 41 retical study of a single branch PHP startup considering as primary cause
%. Moreover, a working fluid with small dynamic viscosity, small spe­ the meniscus/film evaporation rather than boiling. Among other pa­
cific heat and large saturation pressure gradient versus temperature can rameters, he identified the temperature gradient along the tube as the
improve the startup performance. Soponpongpipat et al.[9] propose a most important. The following year Nekrashevych and Nikolayev [15]
concept for the optimal startup of a closed-loop oscillating heat pipe extended the previous study to a multibranch PHP. One of the main
(CLOHP) based on direct fluid visualization. They suggest that a proper results is that there is no startup without bubble generation. They report
startup depends on the balance between vapor expansion in the evap­ that without this effect, after initial oscillation the liquid gathers in a
orator and vapor collapse in the condenser. When vapor expansion is single plug and the motion declines quickly. The same model is then

2
M. Abela et al. Applied Thermal Engineering 236 (2024) 121910

Table 1
varying and Fixed Parameters.
Parameters Varying/ Details Category
Fixed

Tube material Fixed ITO coated Pyrex / brass MATERIAL

Internal diameter Fixed 2 mm
External diameter Fixed 5 mm
Total length Fixed 3380 mm
Condenser length Varying From 5 to 700 mm
Adiabatic length Varying From 5 to 700 mm GEOMETRY
TOPOLOGY
Evaporator length Varying From 5 to 700 mm
Max n. of evaporators Varying up to 16 per tube
per tube
Max n. condensers Varying Depends on the cond.
per tube length
Pattern of eva/cond Varying –
Fluid* Varying Mainly low pressure fluids
Input Power* Varying Depends on tube choice
Filling ratio* Varying 10 % to 90 % (vol.)
Orientation/g-level* Fixed horizontal OPERATION
Condenser Varying 10 ◦ C to 40 ◦ C
temperature*
Initial phase Varying In terms of what phase fills
distribution the evaporator.

Fig. 1. Schematic representation of a Pulsating Heat Pipe (top) and its unbent
configuration (bottom).

used [16] to simulate the startup of a PHP tested in microgravity con­

ditions. It is pointed out that the initial phase distribution has a major
impact on the startup performance. For this reason, authors report that
the initial distribution is one of the parameters to be known to perform a
complete validation of the PHP model. This is one of the most difficult
initial conditions to obtain from experiments since most of the tested
PHPs in literature are not fully transparent. The objective the present
work is contribute to the definition of the Pulsating heat pipe opera­
tional limits by investigating 1) the effect of the number of evaporator
sections and condenser temperature on the steady state performance; 2)
the relationship between various parameters (condenser temperature,
liquid plugs length, phase distribution) and the startup performance. To
obtain results that are independent from the specific PHP, an innovative
approach is adopted. A modular test cell is designed and built; it allows
to perform a multi-variable analysis and to quantitatively investigate the
effects of various parameters on the PHP performance in a well-defined
framework of initial and boundary conditions. Moreover, transparent
tubes and innovative transparent evaporators open to new analysis
possibilities in terms of phase distribution tracking and visualization of
the thermofluidic phenomena. Within such system, three different
Fig. 2. a) Picture of the experimental test cell in the 11 heated/cooled sections
configurations are studied (5, 7, 11 evaporator/condenser sections); two
configuration. b) Indium Tin Oxide heater. c) Cooling unit.
kinds of experimental investigations are performed: i) the steady state
characterization where each configuration is tested at various input
powers (from 10 W to 40 W) and condenser temperatures (from 10 ◦ C to respect to previous literature. A clear relationship between the initial
40 ◦ C); ii) the transient investigation where the influence of various fluid distribution and the startup time is observed; the analysis takes into
parameters on the startup time is analyzed. The steady state results account the local evaporator/condenser phase distribution, the length of
confirm that the increase of number of evaporator sections is beneficial liquid plugs and the number of vapor/liquid interfaces in the condenser
in terms of global performance, i.e., thermal resistance; similarly, the and evaporator sections. Moreover, the condenser temperature is found
increase of condenser temperature ensures better performance. The to have a major influence on the initial phase distribution inside the
transient analysis is focused on the startup performance investigation in PHP.
terms of startup time, trying also to define it more rigorously with Experimental Apparatus.

3
M. Abela et al. Applied Thermal Engineering 236 (2024) 121910

Table 2
Experimental Matrix.
Working Fluid Filling ratio (%) Number of heated/cooled sections (-) Condenser Temperature (◦ C) Input Power Power profile Test
(W) (-) Type
(-)

Ethanol 50 5, 7, 11 10, 20, 30, 40 10, 20, 30, 40 step-up/ Steady state
step-down
Ethanol 50 11 10, 20, 30, 40 10 – startup

The experimental apparatus is designed in a modular way to repro­ by applying a voltage difference. Heaters are designed to deliver a
duce different PHP layouts. The PHP is ideally unbent and reconfigured maximum radial wall to fluid local heat flux of 4 W/cm2; each heater has
as single loop (Fig. 1) where it is possible to play with the number/po­ an overall resistance of 50 Ω (20 Ω/sq). It is possible to increase the
sition/length of heating and cooling zones. A similar approach is used by heaters size by jumping adjacent heaters (as for the case in picture
Okazaki et al. [17] who compared a similar unbent configuration with shown in Fig. 2b) or decrease the heater size by shortening the distance.
an equivalent PHP with meanderings and demonstrated that they have By changing heaters position and/or length (by jumping two adjacent
the same thermal behavior. Therefore, the number of turns in the heaters or shortening the distance between two electrodes,), it is
evaporator section of a standard meandering PHP, correspond here to possible to reproduce a high number of different heating patterns. The
the number of heating sections. Within such a system, many parameters heat dissipation system in the condenser zones (in blue boxes in Fig. 2b
can be varied; Table 1 shows a summary of all parameters characterizing and in Fig. 2c) consists in a series of aluminum spreaders sandwiched on
the PHP functioning; in the second column it possible to appreciate the the borosilicate tubes and coupled with aluminum cold plates connected
high number of parameters that it is possible to vary. The versatility of in a loop to a thermal bath (HAAKE® 0024276). The thermal contact
the proposed design is even more evident considering that the PHPs between the tube and the spreader is enhanced by placing a thin wet
tested so far in the literature allow to vary only few parameters (denoted paper sheet on the tube external surface. In a similar way to heaters,
with * in Table 1). Note that the initial fluid distribution can be varied condensers are designed to be repositioned along the tube. When a
acting on the condenser temperature but can’t be fully controlled. cooled section needs to be placed at a joint location, cold plates are
The experimental test cell (Fig. 2 a) is a planar closed loop consisting employed to cool down the joint, as for the case of Fig. 2a. Condenser
of four transparent borosilicate tubes (I.D. 2 mm, O.D. 5 mm, tube length temperature is measured by a T-type thermocouple (RS® pro 363–0266,
765 mm, total channel length 3460 mm) joined at the corners by four bead diameter 1 mm). The data acquisition system is composed of a
brass joints (denoted with with A B C D in Fig. 2a), designed to guarantee National Instruments® FPGA chassis (NI-cRIO-9074) and modules (NI-
the channel continuity between the borosilicate tubes. Each brass joint 9213, NI-9472, NI-9205). A labVIEW ™ software is used to acquire
hosts a custom-made pressure transducer for the fluid pressure mea­ temperature signals at 50 Hz; pressure and power are acquired at 100
surement (Keller® PAA-33X, 3 bar abs, 0,2%FSO accuracy); moreover, Hz. A 12 MP camera (Ximea® MC124MG-SY) is mounted on a support
joints B and C host a micro-thermocouple for the fluid temperature that can slide on a rail running parallel to the tubes (at 20 cm) to capture
measurement (TC Direct® 406–590, bead diameter 0.25 mm, response the plugs position. A ruler is placed on top of the tubes as reference (see
time 0.1 s; denoted as TC in Fig. 2a). Joints B and D are also equipped Fig. 2b).
with a vacuum/filling micro-metering valve (IDEX® Upchurch Sc. P-
447). Indium Tin Oxide ohmic (ITO) heaters represent one of the main 1.1. Experimental procedure
innovations of the proposed test cell (Fig. 2 b); this solution already
proved to be successful in boiling experiments [18]. Each tube is Main objective of this work is perform a sensitivity analysis to assess
equipped with a series of 16 transparent ITO heaters (length: 40 mm) the influence of geometrical/topological and operational parameters 1)
vacuum deposited on the external surface by Diamond Coatings®, UK. on the PHP steady-state performance; 2) on the PHP startup character­
The heaters (in the red boxes in Fig. 2a and in Fig. 2b) are powered by a istics. As widely discussed in the introduction, the governing parameters
programmable power supply (ELC® ALR3206D) by means of custom- that may affect the PHP startup are: number of heating/cooling sections,
tailored copper-foil electrodes (see Fig. 2b). The temperature of each condenser temperature, input power, fluid initial distribution. The
evaporator is measured by a T-type thermocouple (RS® pro 363–0266, experimental matrix (summarized in Table 2) is defined to obtain a wide
bead diameter 1 mm) clamped by means of kapton tape on its external range of combinations but also to confine the investigation to a sus­
surface (see Fig. 2b). The ITO technology allows to manufacture trans­ tainable number of experiments.
parent heaters with a precise resistance to provide the desired heat flux The working fluid is ethanol, and the PHP is charged at 50 %

Fig. 3. Topology of 5/7/11 heated/cooled sections configurations.

4
M. Abela et al. Applied Thermal Engineering 236 (2024) 121910

volumetric filling ratio; three different geometrical configurations are Table 3

tested at different condenser temperatures and input powers. Before Maximum Uncertainties.
each refill, the device is emptied and vacuum-tested using a helium leak Measured Quantity Uncertainty
detector (Agilent® G8610A); moreover, before each test, the internal Cond/ Evap temperature δT
E,C
= 0.8K
pressure is checked to make sure it is equal to the saturation pressure (T-Type)
Fluid Temperature (K- FL
= 1.7 K
corresponding to the measured fluid temperature. All wired ITO heaters δT
Type)
resistances are checked making sure they correspond to the expected Fluid Pressure δPi = 300 Pa
values. Finally, the device is closed into a box to account only for free Input Current I δI = 0.016 A
convection losses. Before the heat power is supplied to the evaporator Input Voltage V δV = 0.6 V
sections, the thermal bath is set to the target value of condenser tem­ Startup time tsu δtsu = 2s
Liquid plug length δllmax = 1mm
perature TC* and the operator waits until the condenser temperature
n δQ̇(Q̇ = 10W) δQ̇(Q̇ = 20W) δQ̇(Q̇ = 30W) δQ̇(Q̇ = 40W)
reaches TC* +/ − 2 C and the others (fluid and evaporators) reach TC* +
◦

/ − 5 C. Two kinds of test are performed. The first (Figures from 5 to 14)
◦ 5 0.4506 0.7868 0.7726 0.9339
7 0.4768 0.6645 0.8107 0.9473
investigates the effect of the number of evaporator sections and of the 11 0.6365 0.8795 1.0478 1.2107
condenser temperature on the device activation (partial or full startup) n δQ̇net (Q̇ = 10W) δQ̇net (Q̇ = 20W) δQ̇net (Q̇ = 30W) δQ̇net (Q̇ = 40W)
and on the steady-state performance. Fig. 3 shows the topology of the 5 0.7718 1.0364 1.0432 1.1803
three configurations being studied, projected along a straight axis. The 7 0.7748 0.9277 1.0509 1.1771
cooled sections are colored in blue and the heated sections in red. The 11 0.8782 1.0763 1.2259 1.3754

joints are depicted in black, while the ITO patches are shown in pink.
The 11 heated/cooled sections configuration, corresponds to the one losses due to natural convection are correctly take into account. In this
shown in Fig. 2a; the x = 0mm and x = 3380mm of Fig. 3 coincide with way, the net power is obtained (see eq. A(2).
the point labeled A in Fig. 2a. The typical power profile is plotted on the
bottom of Fig. 6: the heat load is supplied to the evaporators by 10 W 1.3. Uncertainty analysis
steps up to 40 W every 20 min and then decreased by 10 W steps every
20 min. If the temperature of an evaporator exceeds 110 ◦ C (maximum The uncertainty analysis is performed using the methods discussed
allowed temperature to prevent ITO degradation) the power is by Moffat [19]. Thermocouples are calibrated by comparison to a
decreased by 10 W steps every 20 min. reference RTD PT100 (Instek® GOM-804/805/802); the value of the
The objective of the second test (Figures from 15 to 25) is to study the sys
systematic uncertainty component (δTi ) and random uncertainty
effect of the condenser temperature on the phase distribution and the
component (equal to the standard deviation, δTirdm ) are obtained for
effect of the latter on the startup time. This analysis is performed on the
each thermocouple; the resulting uncertainty is then given by eq (4). The
11 heated/cooled sections configuration; the power is provided as a 10
average evaporator (eq. (2) and condenser temperatures (eq. (3) un­
W step and it is kept constant for 10 min. Before the heat power is E C
supplied to the evaporator sections, the liquid plugs position/length is certainties (δT and δT respectively) are given by equations (5) and (6).
measured. The camera is slid along the rail and 12 picture per tube are To evaluate the uncertainty of the net power (eq. (7), it is needed to
taken (an example is Fig. 2b); once the images are acquired, the power consider the uncertainties of all the terms of equation (A.2); note that
supply is switched on and the reference time is set to t = 0s. A direct the uncertainty of the power supplied depends on the power supply
visual analysis of these images is conducted to identify the coordinates current and voltage (since Q̇ = VI, eq. (7). Finally, the thermal resis­
of the menisci and the precise locations of the heated/cooled sections. tance uncertainty is given by equation (9). Fluid pressure uncertainty is
This data are the input of a Matlab® script which correlates the positions obtained from the pressure transducer datasheet and the acquisition
of the menisci and heated/cooled sections, providing the global phase module datasheet. For what concerns the liquid plugs length measure­
distribution. ment, since the distance between the camera and the reference ruler is
equal to the distance between the camera and the channel axis (see
Fig. 2b), no parallax error is present; the major source of uncertainty
1.2. Net power
comes from the ruler used as reference. The startup time identification
algorithm is based on the pressure signals analysis (as discussed in the
To give a precise estimation of the equivalent thermal resistance Req
next section). The pressure signals are highly responsive (high frame
(eq. (1), it is necessary to evaluate the effective power supplied to the
E C
rate, 100 Hz) and accurate (the uncertainty is 300 Pa); therefore, the
system Q̇net -i.e., excluding convection losses-. With T and T are uncertainty on the startup time come from the algorithm itself rather
denoted respectively the steady state average evaporator temperature than the uncertainty of the pressure transducer; since the algorithms
and the steady state average condenser temperature; their expression is identifies a temporal window of 2 s in which the startup takes place.
given respectively in equations (2) and (3) where n is the number of both Therefore, the double of the temporal window is assumed as a conser­
heating and cooling sections. vative startup time uncertainty.
E C The uncertainties are reported in Table 3 (note that the form used for
T − T
Rth = (1) representing a quantity and its uncertainty is M = Mm ± δM where Mm is
Q̇net the measured quantity and δM its uncertainty). The thermal resistance
uncertainties are reported as error-bars in the plots.
1 ∑
n
(2)
E
T = TE √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
n 1 i δTi = (δT sys
2 rdm 2
(4)
i ) + (2δT i )

1 ∑
n √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
(3) √
C
T = TC √1 ∑ n
∂T
E 2
n 1 i δT = √
E
( E δT Ei ) (5)
n 1 ∂T i
To correctly evaluate Q̇net , empty-tube tests are conducted; in this way it
is possible to get rid of the influence of the fluid in the evaporator
heating. Then, the empty tube is numerically modeled (see Appendix A);
the simulation and the experimental results are compared to check if the

5
M. Abela et al. Applied Thermal Engineering 236 (2024) 121910

√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
√ 2
√1 ∑ n
∂T
C
δT = √ (6)
C
( C δT Ci )
n 1 ∂T i

̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
√( )2 ( )2̅
√
√ ∂Q̇ ∂Q̇
δQ̇ = √ δV + δI (7)
∂V ∂I

√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
⎞2̅
√⎛ ⎞2 ⎛ ⎞2 ⎛
√
√ ∂Q̇
⎜ net ⎟ ⎜∂Q̇ ⎟ ⎜∂Q̇ ⎟
∂Q̇net =√
√⎝ δQ̇⎠ + ⎝ net δT e ⎠ + ⎝ net δTa ⎠ (8)
∂Q̇ ∂T e ∂Ta

√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
⎛ ⎞2
√
√(
√ ∂Req E )2 (∂Req C )2 ⎜ ∂Req ⎟
∂Req = √
√ E δT + C
δT +⎝ δQ̇net ⎠ (9)
∂T ∂T ∂Q̇eff

2. Results

The results of the steady state tests are discussed first. For each
configuration (number of evaporator/condenser sections n = 5, 7, 11)
the device is tested at different heat input powers (Q̇ = 10 W, 20 W, 30
W, 40 W) and different condenser temperatures (TC = 10 C, 20 C, 30 C,
◦ ◦ ◦

40 C). The steady state behavior is qualitatively analyzed to highlight

◦

the impact of the above-mentioned varying parameters and to give some

insights obtained from the direct visual analysis of the fluid through the
transparent channel. Then, a more thorough quantitative comparison
between all the tests is provided in terms of equivalent thermal resis­
tance. In the second part of this section, the quantitative analysis of the
startup behavior is discussed.
To the authors’ knowledge literature does not provide a unified
quantitative description of startup time, therefore it is useful to intro­
duce here the definition of full startup time tsu that is used here as figure of
merit to characterize the transient behavior. In previous experiments
[20] the same research group used to denote those cases where “wall
temperature variations start to be perceivable within the evaporator section at
least in one channel” (i.e. the pressure oscillations are not present in all
branches or where the functioning is not continuous), as partial startup.
However, from an applicative point of view, it is of primary importance
to be able to distinguish the full activation that brings to a stable and
continuous operation from the intermittent one. Within this innovative
experimental apparatus, the high number of local measurements and the
full visualization allowed to conceive a robust algorithm to quantita­
tively extract the full startup time from the experimental measurements
without introducing any bias. The pressure signals (Pk , with k = 1,2..nP )
are found to be the most suitable for the identification of the startup time
since they are more responsive than the thermocouple signal (due to the Fig. 4. Startup time identification algorithm flow chart.
thermal inertia); moreover; pressures and fluid motion are intrinsically
linked. c. AND the mean value of the pressures in the interval
The algorithm for the startup time identification can be summarized t ∈ [t * , t * +50s] is at least the 95 % of the pseudo steady state
as follows: average pressure (Pk > 0.95P).
4) If all conditions (a,b,c) are satisfied, tsu = t* , otherwise no startup is
1) The derivative of the pressure signals (Ṗk (t)) is evaluated present
⃒ ⃒
⃒ ⃒
2) The temporal location of ⃒⃒Ṗk (t)⃒⃒ peaks with a minimum prominence
where nP is the number of pressure transducers, 4 in this case; Ṗk is
of 0.5 is determined the k-th pressure signal derivative; Ṗk is average value of the k-th pres­
3) The peaks are analyzed and
⃒ ⃒ sure signal derivative; P is the pressure steady state average value. Note
⃒ ⃒
a. IF all the signals derivatives (⃒⃒Ṗk (t) ⃒⃒, for k = 1,2..nP ) show a peak that the time is referred to the instant when the first power step is
supplied (t = 0s). The prominence is a measure of how distinct or sig­
in the interval t ∈ [t * − 1s, t* +1s] nificant a local maximum point is within its surrounding data points. It
⃒ ⃒
⃒ ⃒
b. AND the average value of the pressures derivative (⃒⃒Ṗk ⃒⃒) in the quantifies the relative height or magnitude of the local maximum
compared to the neighboring points. The value of 0.5 is empirically
interval t ∈ [t * , t* +50s] is at least the 95 % of the steady state deduced. Both a) and b) conditions ensure that fluid motion takes place
average pressure derivative in all branches; while c) and d) are needed to be sure that partial stop

6
M. Abela et al. Applied Thermal Engineering 236 (2024) 121910

Fig. 5. Temporal evolution of the measured variables for the n5/10 ◦ C configuration.

Fig. 6. Temporal evolution of the measured variables for the n5/30 ◦ C configuration.

overs are not mistaken for full startup. In fact, c) can be seen as a dy­
namic condition since it ensures that after the startup time oscillations
are present in all branches, d) ensures that the average value after the
startup is close to the steady state one. The latter is evaluated through
the analysis of the pressure signals during the last 5 min of each power
level. For each pressure signal, the average value (over the 5 min) and
moving average of both pressure amplitude (window size of 10 s) and
frequency (window size of 60 s) is calculated. If the moving average
value of both pressure amplitude and frequency are respectively in a +/-
5 % and +/- 30 % range with respect to the average values calculated
over the 5 min interval, those are considered quasi steady-state fre­
quency and average value. Note that the four signals are characterized
by almost the same frequency and amplitude when the device is oper­
ating. The flow chart of the startup time identification algorithm is
shown in Fig. 4.

2.1. Steady state analysis

From now on, each configuration (n = 5, 7, 11) coupled with a

condenser temperature will be shortened for the sake of brevity as fol­ Fig. 7. Detailed pressures evolution in a 200 s window for n5/30 ◦
C
lows: for instance, n5/20 ◦ C stands for n = 5 configuration tested at configuration.
20 ◦ C condenser temperature. Since the overall tube length is constant,
the variation of n implies that the length of the heated and cooled zones PHP tested at TC = 10 C. As soon as the power is supplied, the evapo­
◦

must vary too. The authors’ choice is to scale the heated/cooled zones in rators temperature rises and reaches a steady-state temperature; when
such a way that the total input power Q̇ and the radial heat flux (W/m2) the input power is increased the temperature increases accordingly and
per heated section are kept constant for all the configurations. reaches a new steady-state temperature which results from the balance
Fig. 5 shows the temporal evolution of the measured variables by of the heat transferred to the fluid, the heat diffusion along the tube and
means of superposed subplots: From top to bottom: the evaporators the heat lost as results of natural convection. The fluid temperature and
temperatures TiE , pressures Pi , and heat input power Q̇, of the n5/10 ◦ C

7
M. Abela et al. Applied Thermal Engineering 236 (2024) 121910

in dry-out and follows the trend shown in Fig. 5. Thanks to the direct
visualization it is possible to give an explanation to this behavior: when
evaporator 5 is in dry-out (1900s < t < 3500s), a long vapor bubble is
stuck the evaporator 5 region; then (at t ≅ 3500s) the bubble is pushed
in the evaporator 1 region. The presence of a bubble in the same region
throughout the entire test indicates this region is not active, i.e., no
liquid plugs are able to penetrate it. As already reported by the same
authors in previous experiments, there exists a hysteresis with respect to
the heat input level: the device indeed does not work during the first step
(10 W) of the heat-up steps, while the PHP is clearly working at the same
heat input level of the heat-down steps.
A similar behavior is observed in for n7/10 ◦ C showed in Fig. 9; in
this case, a partial startup takes place for low input powers (Q̇ = 10W,
20W), adjacent evaporators (T5E , T E6 , T7E ) are in dry-out (no temperature
oscillations are present) while the others show an intermittent behavior
characterized by stopovers. At t ≅ 2300 s, the device stops working, no
pressure oscillations are present and all the evaporators temperatures
rise. At t ≅ 2400 s the input power is increased (Q̇ = 30W) and after ≅
Fig. 8. Detailed pressures evolution in a 200 s window for n7/30 ◦
C 300 s seconds, a partial startup takes place. From now on, only evapo­
configuration.
rator 7 is in dry-out (i.e. partial dry-out of the system). Similarly, to the
case shown in Fig. 6, the partial dry-out is due to the presence of a long
pressure trend clearly show that no startup takes place; this is also vapor bubble that occupies the entire length of one or more heated
confirmed by the absence of any visible fluid motion through the zones; for low input powers, liquid plugs are not able to penetrate this
transparent tubes. Therefore, the PHP works mainly as a conductive region. After the power increase, the pressures suddenly increase, and
medium. An almost identical behavior is obtained by the n5/20 ◦ C their unbalance is high enough to push liquid plugs into the evaporators
configuration. 5 and 6 region reducing the size of the bubble that from now on occupies
Fig. 6 shows the behavior of n5/30 ◦ C. In this case, the device shows only the evaporator 7 region. Once again, the presence of local non-
two trends: for t < 1900s, the flow motion is poor and the curves follow active regions, testify that the device operation is only partial.
the same trends of the previous case (see Fig. 5), meaning that there is The n7 configuration exhibits a full start up and a stable operation
not a stable operation; for t ≥ 1900s, a partial startup takes place and the only when the condenser temperature is increased to TC = 30 C; the
◦

device operates for the rest of the test. Before the partial startup, fluid device fully starts up and shows a continuous operation without stop­
temperature and pressure do not exhibit any notable changes; at t ≅ overs. In Fig. 8, a detail of the pressure temporal evolution of n7/30 ◦ C is
1900s a sudden pressure rise takes place, and it is shortly after followed plotted; at first glance it is possible to notice the difference with the case
by fluid temperature rise and evaporators temperature steep drop (this of n5/30 ◦ C, the pressure oscillation is continuous, and the amplitude is
phenomenon is known as temperature overshoot). The latter is a sign of higher. Moreover, all the regions are active. Fig. 10 shows the behavior
the wall to fluid heat exchange; in fact, when the fluid oscillations start, of n7/40 ◦ C; the device fully starts up right after the power is supplied.
cold fluid coming from the condenser sections wets the evaporators and Evaporators temperatures, fluid temperatures and pressures show
resulting in a sudden evaporator temperature decrease. Fig. 7 shows the continuous operation throughout the test.
detailed temporal evolution of the pressures of the n5/30 ◦ C configu­ The n11 configuration (see Fig. 11 and Fig. 12) exhibits the best
ration in a temporal window around the partial startup. After pressure thermal behavior. The device fully starts up at the lowest input power
peaks located at t ≅ 1840s, only intermittent and small amplitude and the evaporator temperatures are less dispersed (i.e. they oscillate in
pressure perturbations are present (also known as stopovers). Moreover, a narrower temperature range with respect to previous cases).
some evaporators dry-out during the operation. In fact, for Comparing the two (TC = 10 C and TC = 40 C) steady-state behaviors,
◦ ◦

1900s < t < 3500s, evaporator 5 (see T5E in Fig. 6) exhibits a high tem­ it is evident that the condenser temperature has a smaller influence on
perature (≅ 70◦ C) with respect to all the others (≅ 50◦ C); at t ≅ 3500s, the performance. On the other hand, a closer look to the startup
T5E starts decreasing while an adjacent evaporator 1 temperature (see T1E (t < 300s) operation reveals an effect of the condenser temperature on
in Fig. 6) starts increasing. For the rest of the test, evaporator 1 remains the transient behavior. In fact, it is observed that the increase of

Fig. 9. Temporal evolution of the measured variables for the n7/10 ◦ C configuration.

8
M. Abela et al. Applied Thermal Engineering 236 (2024) 121910

Fig. 10. Temporal evolution of the measured variables for the n7/40 ◦ C configuration.

Fig. 11. Temporal evolution of the measured variables for the n11/10 ◦ C configuration./condenser sections, 10 ◦ C condenser temperature.

Fig. 12. Temporal evolution of the measured variables for the n11/40 ◦ C configuration. /condenser sections, 10 ◦ C condenser temperature./condenser sections,
40 ◦ C condenser temperature.

condenser temperature determines initial fluid distributions that in turn of evaporator sections increases the level of internal perturbations and
determine longer transients; this is discussed in detail in the next section. consequently, the probability of a more efficient phase redistribution
To quantify the influence of the increase of number of evaporator sec­ during the operation. Moreover, since for the present case the total
tions and condenser temperature on the overall performance, the length of the channel is constant, the increase of evaporator sections
equivalent thermal resistance (see eq. (1) is chosen as figure of merit. corresponds to an increase of the thermal gradient between evaporator
The resistance is plotted (see Fig. 13 and Fig. 14) for the n7 and n11 and condenser sections. Therefore, the thermal gradients and pertur­
configurations; the n5, n7/10 ◦ C and n7/20 ◦ C configurations are not bations create a non-equilibrium pressure conditions that are the driving
taken into account since they do not show a stable operation throughout force for self-sustained oscillations and thermofluidic transport as
the test. From the Figs. 13 and 14, it is clear that increasing the number observed by Khandekar et al. [2]. In fact, for constant condenser tem­
of evaporator sections is beneficial in terms of global performance. In perature and input power the overall thermal resistance on n11 is lower
fact, as already observed by Charoensawan et al. [21], a higher number of 0.3 – 0.6 W/K than the one of n7. Moreover, the thermal resistance

9
M. Abela et al. Applied Thermal Engineering 236 (2024) 121910

Fig. 13. Thermal Resistance for ne = 7.

Fig. 15. Condenser volume occupied by liquid phase vs condenser
temperature.

Fig. 14. Thermal Resistance for ne = 11.

decreases at higher condenser temperatures. This effect is presumably Fig. 16. Evaporator volume occupied by liquid phase vs condenser tempera­
related to the change of the fluid physical properties and in particular ture. phase.
viscosity and density that are inversely proportional to temperature. As
already observed by Cai et al. [22], lower density liquid has smaller 2.2. Transient Analysis
inertia and reduces the temperature difference between evaporator and
condenser needed to sustain oscillations; lower viscosity contributes to This section describes the results of the PHP transient behavior and
the reduction of temperature difference because the decrease of viscous the fluid visualization analyses. The scope is to understand how the
drag reduces the driving pressure difference between evaporator and condenser temperature and the initial fluid distribution, which have
condenser. been rarely investigated in the literature, affect the PHP startup time.
The present steady state analysis demonstrates that the activation These two parameters are not independent, in fact it is observed that the
and stable operation of a horizontal PHP does not only depend on the fluid distribution is influenced by the condenser temperature. A
number of heated and cooled sections but also on the combined effect of comprehensive understanding of the fluid initial distribution is crucial
heat input level and condenser temperature. Consequently, the critical for comprehending the startup phenomenon. To better understand the
number of heated/cooled section (i.e. the minimum number of heated/ chain of phenomenological cause and effect, the parameters that char­
cooled section that allows the PHP activation and stable operation) is acterize the fluid initial are firstly identified. These are the local filling
not a constant value applicable to every PHP. Furthermore, the results ratio in the evaporator and condenser zones VlE /Vtot E
, VlC /Vtot
C
(i.e. the
confirms that the present assessment is able to replicate the thermo­ percentage of condenser and evaporator total volume occupied by the
fluidic behavior of multiturn PHPs even without the presence of local liquid phase for each condenser temperature maximum) the length of
pressure drops due to meanderings. the liquid slugs llmax , and number of liquid/vapor interfaces present in
the condenser and evaporator sections nEm . and nCm . Then, the effect of the
condenser temperature T C on the above-mentioned parameters is
shown. Finally, the influence of the above-mentioned parameters on the

10
M. Abela et al. Applied Thermal Engineering 236 (2024) 121910

Fig. 17. Maximum initial liquid plugs length vs condenser temperature.

Fig. 19. Number of menisci in the evaporator vs evaporator volume occupied
by liquid phase.

Another parameter which is influenced by the condenser tempera­

ture is the maximum initial liquid plug length (llmax ) i.e., the length of the
longest liquid plug present in the channel before the power is supplied.
Fig. 17 shows the maximum length of the longest liquid plugs against the
condenser temperature. It appears that the length of liquid plugs in­
creases with the increase of condenser temperature T C . As discussed
before, after the thermal bath is switched on and before the system
reaches the thermal equilibrium, if the condenser temperature is lower
than the ambient temperature (T C = 10 C, 20 C) it is observed that the
◦ ◦

fluid tends to gather in the condenser sections; this distribution does not
significantly change during the isothermalization (before the power is
supplied). Conversely, for condenser temperatures higher than the
ambient temperature (T C = 30 C, 40 C), it is observed that the
◦ ◦

condenser regions are occupied by vapor bubbles while in the other

regions the fluid gathers in long liquid plugs; also in this case, the dis­
tribution does not change during the system isothermalization.
In Fig. 18, the number of menisci in the condenser sections is plotted
against the condenser total volume occupied by liquid phase. Again, the
plot can be divided into two parts depending on the condenser tem­
Fig. 18. Number of menisci in the condenser vs condenser volume occupied by perature. For T C = 30 C, 40 C a relatively small percentage of
◦ ◦

liquid phase. condenser volume is occupied by the liquid phase (VlC /Vtot C
< 40 %) and
the number of menisci is low. This is due to the fact that for high
startup time tsu is explained. The most suitable configuration to perform condenser temperatures, the condenser sections tend to be occupied by a
the startup analysis is the n11 since it exhibits a full startup also for the small amount of relatively long vapor bubbles ( 1 − 2); it follows that the
lower heat input level (10 W). The data showed in the next plots refers to number of menisci in each condenser section is usually very small
a data sample of 30 tests: 6 at T C = 10 C, 10 at T C = 20 C ◦ C, 7 at T C =
◦ ◦
( 0 − 3). On the other hand, for T C = 10 C, 20 C liquid phase tends to
◦ ◦

C
30 C, 7 at T = 40 C. Before each test, the condenser temperature level
◦ ◦
gather in the condenser sections and consequently the condenser volume
is set until the system is almost isothermal. occupied by liquid phase tends to be higher than the previous case. The
number of menisci decreases with the decrease of condenser tempera­
2.2.1. Initial phase distribution ture because vapor bubbles tend to condense for low condenser tem­
Since the initial phase distribution is affected by the condenser perature. In Fig. 19, the number of menisci in the evaporator sections is
temperature, the local filling ratio before the startup is firstly investi­ plotted against the evaporator total volume occupied by liquid phase.
gated with a statistical approach. In Figs. 15, 16, 17 the data is grouped For the reasons discussed above, for TC = 10 C, 20 C the liquid phase
◦ ◦

in boxplots. The bottom and the top of each boxplot are respectively the tends to gather into the condenser sections; it follows that the total
25th percentile and the 75th percentile, the red line is the median and evaporator volume occupied by liquid phase is low (VlE /Vtot E
< 20 %) and
the two extremes at the ends of the black dashed vertical lines are the consequently the number of menisci is low too (< 5). On the other hand,
maximum and minimum values.
for T C = 30 C, 40 C, the condenser sections tend to be occupied by
◦ ◦

In Figs. 15 and 16 the local evaporator and condenser filling ratios

vapor bubbles and therefore more liquid phase is present in the evapo­
are plotted against the condenser temperature. The liquid phase tends to
rator sections (25% < VlE /VtotE
< 45 %); in this case, the number of
gather in the condenser sections for T C = 10 C and T C = 20 C while it
◦ ◦

menisci increases with the percentage of total evaporator volume

tends to randomly occupy evaporator and adiabatic zones for TC = 30 C
◦

occupied by liquid phase.

and T C = 40 C.
◦

11
M. Abela et al. Applied Thermal Engineering 236 (2024) 121910

Fig. 20. Startup time vs condenser volume occupied by liquid phase. Fig. 22. Startup time vs maximum liquid plug length for n11/10 ◦ C and
n11/20 ◦ C.

Fig. 21. Startup time vs evaporator volume occupied by liquid phase.

Fig. 23. Startup time vs maximum liquid plug length n11/30 ◦ C- n11/40 ◦ C.
2.2.2. Startup time
Finally, the influence of fluid initial distribution on the startup time In Figs. 22 and 23, the startup time is plotted against the maximum
is analyzed. In Figs. 20 and 21, the startup time is plotted respectively liquid plug length. For low condenser temperatures (TC = 10 C, 20 C,
◦ ◦

against the condenser and evaporator volume occupied by liquid phase. see Fig. 22), the maximum liquid plugs lengths are smaller than 500 mm
In Fig. 20, for medium/high values of condenser volume occupied by and no evident influence on the startup time appears; in fact, the latter
liquid phase (45%< V Cl /VtotC
< 85 %) corresponding to TC = 10 C and
◦
shows a quasi-constant trend. A different trend emerges from Fig. 23
C
T = 20 C) the startup times are the lowest (50s < tsu < 120s). For low
◦ where the maximum liquid plugs lengths cover a bigger range of values
and where the startup time shows an increasing trend.
values of condenser volume occupied by liquid phase (0%< V Cl /Vtot C
<
C C To explain the behavior shown in Figs. 22 and 23, it is useful to refer
45 %, corresponding to T = 30 C and T = 40 C), the startup time is
◦ ◦

to Figures 15 16 17. Summarizing the results discussed so far, it emerges

higher (tsu > 150s). This trend is mirrored in Fig. 21. In fact, for small
that for low condenser temperatures (TC = 10 C,20 C) the liquid phase
◦ ◦

amounts of liquid phase in evaporator section (5%< V El /Vtot E

< 25 %,
tends to gather in the condenser sections rather than the evaporator
corresponding to T C = 10 C and TC = 20 C) the startup time is small.
◦ ◦

sections and the length of liquid plugs is relatively small (<500 mm); on
On the other hand, for 25%< V Cl /VtotC
< 40 %, the startup time is higher. the other hand, for high condenser temperatures (T C = 30 C,40 C), the
◦ ◦

It can be assessed that lower condenser temperatures are beneficial in amount of liquid phase in evaporators is higher and it is arranged in long
terms of startup time and that tsu for TC = 10 C is similar to the one for
◦
liquid plugs. Therefore, the increase of startup time with the increase of
T C = 20 C and results to be less dispersed with respect to the higher evaporator volume occupied by liquid phase and maximum liquid plugs
◦

temperatures cases. At higher condenser temperatures, the evaporator is length can be explained by i) the higher inertia of long liquid plugs and
more filled with liquid resulting in higher startup times but data are ii) the high bubble nucleation overshoot on smooth surface (i.e., energy
more dispersed and need further discussion that follows here below. barrier required for the bubble nucleation) with respect to the energy

12
M. Abela et al. Applied Thermal Engineering 236 (2024) 121910

is symmetrical to the one discussed above, i.e., a small number of

menisci is in the evaporator because the liquid tends to gather in the
condenser sections; again, no relationship between startup time and
number of menisci emerges. However, for TC = 30 C and T C = 40 C,
◦ ◦

when the evaporator is filled with a higher amount of liquid and, more
likely, a higher number of menisci (as shown in Fig. 19), a different trend
appears. The startup time shows a decreasing trend as the number of
menisci increases. This is due to the fact that the meniscus microregion
(i.e., the region between the liquid bulk and the adsorbed film) is
characterized by a low thermal resistance which in turn leads to
extremely high evaporation rates, as discussed by Stephan and Busse
[25] and Höhmann and Stephan [26]. Therefore, the increase of menisci
in evaporator sections contributes to increase the pressures imbalances
that prime the PHP startup. It is worth noting that the acknowledged
chaotic nature of PHPs inherently poses challenges in defining repeat­
ability. The standard definitions of repeatability within this domain is
focused on overall metrics rather than local variables/phenomena. So
far in the literature, tests conducted with the same initial and boundary
conditions are considered repeatable when they show the same thermal
resistance values. Obviously, this overall approach is less sensitive to the
Fig. 24. Startup time vs number of menisci in condenser section. slight local discrepancies that may result from the system chaotic nature.
Moreover, the results are reported in terms of condenser/evaporator
sections volume occupied by liquid phase, maximum liquid plugs length
and total number of menisci in evaporator sections without taking into
account the exact fluid distribution. For this reason, even when the
number of menisci or the maximum liquid plug length remain constant,
any difference in their specific location within the evaporator sections
can yield marginally varied results. For these reasons, since it is virtually
impossible to exactly replicate a specific initial phase distribution using
the experimental apparatus discussed, it is not possible to strictly apply
the concept of repeatability when referring to the PHP startup/transient
operation.

3. Conclusions

An innovative modular experimental apparatus is developed to

perform a systematic multi-parametric investigation of the influence of
number of evaporator sections, condenser temperature and phase initial
distribution on the startup and the steady state performance of three
PHP configurations. Tests are performed on 5 – 7 – 11 evaporator sec­
tions PHP; the working fluid is ethanol at 50 % volumetric filling ratio.
The main outcomes can be summarized as follows:

Fig. 25. Startup time vs number of menisci in evaporator section. 1) The activation and stable operation of a horizontal PHP do not
depend only on the number of heated and cooled sections but also on
needed to evaporate the liquid at the menisci location. In fact, the two- the combined effect of heat input level and condenser temperature;
phase flow triggered by meniscus/film evaporation rather than the consequently, the critical number of heated/cooled section (i.e., the
bubble nucleation is the main startup mechanism reported in literature. minimum number of heated/cooled section that allows the PHP
For example, Qu et al. [23 24] in their study on the startup of micro PHPs activation and stable operation) is not a constant value applicable to
report the absence of bubble nucleation during the startup period. every PHP.
Finally, the influence of the number of menisci in condenser/evaporator 2) The effect of the condenser temperature on the initial fluid distri­
section and the startup time is investigated. In Fig. 24 the startup time is bution is analyzed. For low condenser temperatures (TC = 10 C,
◦

plotted against the number of menisci in the condenser section nCm . As 20 C), the liquid phase tends to gather in the condenser sections
◦

expected, for high condenser temperature (TC = 30 C, 40 C) vapor rather than the evaporator sections and the length of liquid plugs is
◦ ◦

bubbles tend to entirely cover the condenser sections and consequently smaller; moreover, more menisci are present in the condenser section
almost no menisci are present. On the other hand, for (TC = 10 C and with respect to the evaporator section. For high condenser temper­
◦

C
T = 20 C) liquid plugs more menisci are in the condenser section. It is atures (TC = 30 C, 40 C), the amount of liquid phase in evaporators
◦ ◦ ◦

worth noting that the number of menisci in the case of T C = 10 C tends is higher and it is arranged in longer liquid plugs; menisci are
◦

to be smaller with respect to the case TC = 20 C. This is a consequence

◦
distributed in adiabatic and evaporator section while almost none of
of the fact that in the first case the fluid tends to condense and to form a them is observed in the condenser section.
liquid plug that cover entirely each condenser section; with increasing 3) A relationship between the startup time and the initial fluid distri­
condenser temperature, vapor bubbles appear. In general, the number of bution emerged. The presence of long liquid plugs leads to an in­
menisci in condenser section has no significant influence on the startup crease of the startup time. On the other hand, the increase of number
time. In Fig. 25 the startup time is plotted against the number of menisci of menisci in the evaporator sections leads to a decrease of the
in the evaporator sections nEm . For TC = 10 C and T C = 20 C the situation
◦ ◦
startup time.

13
M. Abela et al. Applied Thermal Engineering 236 (2024) 121910

Fig. A1. Empty tube model scheme.

Declaration of Competing Interest

The authors declare that they have no known competing financial

interests or personal relationships that could have appeared to influence
the work reported in this paper.

Data availability

Data will be made available on request.

Acknowledgments

The Authors would like to express their deepest gratitude and

appreciation to the M.Sc. student Mattia Capuani, whose dedication and
meticulousness in conducting the experiments gave an invaluable
contribution to the realization of this research paper.
The present work is funded by GRANT AWARD FA9550-19-S-0003-
EUROPEAN OFFICE OF AEROSPACE RESEARCH AND DEVELOP­
MENT – Air Force Office of Scientific Research (USA) – Id. number
WS00317867 - Pulsating heat pipes understanding and prediction. The
Fig. A2. Comparison between empty tube simulation and experiment. views expressed are those of the authors and do not necessarily reflect
the official policy or position of the Department of the Air Force, the
The present results offer new insights for the understanding of the Department of Defense, or the U.S. government.
phenomena governing the PHP operation.

Appendix A

In order to estimate the net power supplied to the system, the heat losses need to be clearly identified. To do so, the experimental characterization
of the empty device is performed. Before the test, the device is vacuumed in order to exclude the heat exchange with the working fluid. During the test,
only one evaporator (length Le = 175mm, placed in the center of a tube of length Lt = 2La + Le = 765mm) is connected to the power supply; the power
Q̇ is supplied from 0 W to 5 W with steps of 1 W every 10 min (time needed to reach the thermal equilibrium). During the test, the environment (Ta )
temperature is acquired. The test temperature profile is plotted in red in Fig. A2. The experimental data are used to validate the heat loss model
described here. The tube is modeled using a 1D approach (see Fig. A1) with the following assumptions:

• The Temperature is uniform in radial direction since the tube wall thickness is small.
• No heat exchange takes place in the internal wall since no fluid is present (vacuum condition).
• The temperature at the two ends of the tube is equal to the ambient temperature T(0) = T(Lt ) = Ta ; this boundary condition is experimentally
deduced.
• The heat flux is supplied uniformly in the evaporator volume.

Under these assumptions, the thermal diffusion inside the solid is described by the following equation:
∂T kt ∂2 T πde
= + (h + C)(Ta − T) +
∂t ρCp ∂x2 ρCp St
⎧
⎪ Q̇ ( 4 )
1 ⎨ + πde σεITO Ta − T if x ∈ evaporator section
4
+ Le (A1)
ρCp St ⎪
⎩ ( )
πde σεG Ta 4 − T 4 otherwise

where ρCp are respectively the tube density and tube heat capacity; St = π(d2e − d2i )/4 is the tube cross section; T is the tube temperature; kt is the tube

14
M. Abela et al. Applied Thermal Engineering 236 (2024) 121910

thermal conductivity; h is the free convection coefficient for a horizontal cylinder evaluated using the Nusselt number correlation proposed by
Churchill and Chu [27];C is a tuning parameter to take into account additional convective losses due to the clamps that hold the evaporators electrodes
in place; Ta is the temperature of the environment; σ is the Boltzmann constant, εITO is the Indium thin Oxide film emissivity (equal to 0.3 as reported
by Tecchio ([28], p. 31) and εG is the borosilicate glass emissivity (0.9). Q̇ is the power supplied; de and di are respectively the external and in ternal
tube diameter. The equation (A1) is discretized in space using the central finite difference method and discretized in time using an explicit Runge-
Kutta formula, the Dormand-Prince pair [29]. In Fig. A2, the experimental and simulated temperature are compared. It is clear that the heat trans­
fer mechanisms are captured with a good accuracy. It follows that the net power supplied to the system is expressed by eq. (A2).
[ ( E ) ( 4 )]
(A2)
E
Q̇net = Q̇ − nE π de Le (h + C) T − Ta + σεITO T − Ta 4

where n is the number of heating sections and εITO itois the emissivity of the ITO film.

Appendix B. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.applthermaleng.2023.121910.

References [14] V.S. Nikolayev, Effect of tube heat conduction on the single branch pulsating heat
pipe start-up, Int J Heat Mass Transf 95 (Apr. 2016) 477–487, https://doi.org/
10.1016/j.ijheatmasstransfer.2015.12.016.
[1] M. Mameli, G. Besagni, P.K. Bansal, C.N. Markides, Innovations in pulsating heat
[15] I. Nekrashevych, V.S. Nikolayev, Effect of tube heat conduction on the pulsating
pipes: From origins to future perspectives, Appl Therm Eng 203 (Feb. 2022),
heat pipe start-up, Appl Therm Eng 117 (2017) 24–29, https://doi.org/10.1016/j.
https://doi.org/10.1016/j.applthermaleng.2021.117921.
applthermaleng.2017.02.013.
[2] S. Khandekar, N. Dollinger, M. Groll, Understanding operational regimes of closed
[16] M. Abela, M. Mameli, V. Nikolayev, S. Filippeschi, Experimental analysis and
loop pulsating heat pipes: An experimental study, Appl Therm Eng 23 (6) (2003)
transient numerical simulation of a large diameter pulsating heat pipe in
707–719, https://doi.org/10.1016/S1359-4311(02)00237-5.
microgravity conditions, Int J Heat Mass Transf 187 (May 2022), https://doi.org/
[3] P. Charoensawan, S. Khandekar, M. Groll, P. Terdtoon, Closed loop pulsating heat
10.1016/j.ijheatmasstransfer.2022.122532.
pipes - Part A: Parametric experimental investigations, Appl Therm Eng 23 (16)
[17] S. Okazaki, H. Fuke, H. Ogawa, Performance of circular Oscillating Heat Pipe for
(Nov. 2003) 2009–2020, https://doi.org/10.1016/S1359-4311(03)00159-5.
highly adaptable heat transfer layout, Appl Therm Eng 198 (Nov. 2021), https://
[4] P. Charoensawan, P. Terdtoon, Thermal performance of horizontal closed-loop
doi.org/10.1016/j.applthermaleng.2021.117497.
oscillating heat pipes, Appl Therm Eng 28 (5–6) (Apr. 2008) 460–466, https://doi.
[18] C. Baldassari, M. Mameli, M. Marengo, Non-uniform onset of nucleate flow boiling
org/10.1016/j.applthermaleng.2007.05.007.
of r-134a inside a glass minichannel, Experimental Heat Transfer 27 (4) (2014)
[5] J. Lee, Y. Joo, S.J. Kim, Effects of the number of turns and the inclination angle on
316–328, https://doi.org/10.1080/08916152.2013.849182.
the operating limit of micro pulsating heat pipes, Int J Heat Mass Transf 124 (2018)
[19] R. J. Moffat, “Describing the Uncertainties in Experimental Results.”.
1172–1180, https://doi.org/10.1016/j.ijheatmasstransfer.2018.04.054.
[20] L. Pagliarini, L. Cattani, M. Mameli, S. Filippeschi, F. Bozzoli, S. Rainieri, Global
[6] J.L. Xu, X.M. Zhang, Start-up and steady thermal oscillation of a pulsating heat
and local heat transfer behaviour of a three-dimensional Pulsating Heat Pipe:
pipe, Heat and Mass Transfer/waerme- Und Stoffuebertragung 41 (8) (Jun. 2005)
combined effect of the heat load, orientation and condenser temperature, Appl
685–694, https://doi.org/10.1007/s00231-004-0535-3.
Therm Eng 195 (Aug. 2021), https://doi.org/10.1016/j.
[7] S. Khandekar, A.P. Gautam, P.K. Sharma, Multiple quasi-steady states in a closed
applthermaleng.2021.117144.
loop pulsating heat pipe, International Journal of Thermal Sciences 48 (3) (Mar.
[21] P. Charoensawan, S. Khandekar, M. Groll, P. Terdtoon, Closed loop pulsating heat
2009) 535–546, https://doi.org/10.1016/j.ijthermalsci.2008.04.004.
pipes - Part A: Parametric experimental investigations, Appl. Therm. Eng. 23 (16)
[8] X. Liu, Y. Chen, M. Shi, Dynamic performance analysis on start-up of closed-loop
(2003) 2009–2020, https://doi.org/10.1016/S1359-4311(03)00159-5.
pulsating heat pipes (CLPHPs), International Journal of Thermal Sciences 65 (Mar.
[22] S. Cai, R.-L. Chen, and C. Chen, An Investigation of Evaporation, Boiling, and Heat
2013) 224–233, https://doi.org/10.1016/j.ijthermalsci.2012.10.012.
Transport Performance in Pulstating Heat Pipe. 2002. 10.1115/IMECE2002-33334.
[9] N. Soponpongpipat, P. Sakulchangsatjaati, N. Kammuang-Lue, P. Terdtoon,
[23] J. Qu, H. Wu, Flow visualization of silicon-based micro pulsating heat pipes, Sci
Investigation of the startup condition of a closed-loop oscillating heat pipe, Heat
China Technol Sci 53 (4) (2010) 984–990, https://doi.org/10.1007/s11431-009-
Transfer Engineering 30 (8) (Jul. 2009) 626–642, https://doi.org/10.1080/
0391-y.
01457630802656876.
[24] J. Qu, H. Wu, P. Cheng, Start-up, heat transfer and flow characteristics of silicon-
[10] V.M. Patel, Gaurav, and H. B. Mehta, Influence of working fluids on startup
based micro pulsating heat pipes, Int J Heat Mass Transf 55 (21) (2012)
mechanism and thermal performance of a closed loop pulsating heat pipe, Appl
6109–6120, https://doi.org/10.1016/j.ijheatmasstransfer.2012.06.024.
Therm Eng 110 (Jan. 2017) 1568–1577, https://doi.org/10.1016/j.
[25] P. C. Stephan and C. A. Busse, “Analysis of the heat transfer coefficient of grooved
applthermaleng.2016.09.017.
heat pipe evaporator walls,” 1992.
[11] T. Hao, X. Ma, Z. Lan, Effects of Hydrophilic and Hydrophobic Surfaces on Start-Up
[26] C. Höhmann, P. Stephan, Microscale temperature measurement at an evaporating
Performance of an Oscillating Heat Pipe, J Heat Transfer 140 (1) (Aug. 2017),
liquid meniscus, Exp Therm Fluid Sci 26 (2) (2002) 157–162, https://doi.org/
https://doi.org/10.1115/1.4037341.
10.1016/S0894-1777(02)00122-X.
[12] Y. Zhou, et al., Enhancement of start-up and thermal performance in pulsating heat
[27] S.W. Churchill, H.H.S. Chu, Correlating equations for laminar and turbulent free
pipe with GO/water nanofluid, Powder Technol 384 (2021) 414–422, https://doi.
convection from a horizontal cylinder, Int J Heat Mass Transf 18 (9) (1975)
org/10.1016/j.powtec.2021.02.021.
1049–1053, https://doi.org/10.1016/0017-9310(75)90222-7.
[13] T. Daimaru, H. Nagai, M. Ando, K. Tanaka, A. Okamoto, H. Sugita, Comparison
[28] C. Tecchio, “Experimental study of boiling : characterization of near-wall
between numerical simulation and on-orbit experiment of oscillating heat pipes,
phenomena and bubble dynamics.” [Online]. Available: https://theses.hal.science/
Int J Heat Mass Transf 109 (2017) 791–806, https://doi.org/10.1016/j.
tel-03859592.
ijheatmasstransfer.2017.01.078.
[29] J. R. Dormand and P. J. Prince, “A family of embedded Runge-Kutta formulae.”.

15