Professional Documents
Culture Documents
Microscale - Capson -Commen- Modified Presure Equation - Final11
Microscale - Capson -Commen- Modified Presure Equation - Final11
Microscale - Capson -Commen- Modified Presure Equation - Final11
abc
Department of Mechanical Engineering, Indian Institute of Technology Delhi, New Delhi,
Delhi, India 110016
* e-mail: om.saran7@gmail.com
1
Development and validation of an exact analytical second-order slip
velocity model for flow in three-dimensional microchannel Array
1. Introduction
The study of fluid flow through microchannels has gained attention nowadays due to the
application of microdevices in the cooling of electronic circuits, computer chips, fuel cells,
and medical devices [1]. Microscale devices have characteristic dimensions in micron,
which is comparable to the mean free path of the gas molecule even at atmospheric
conditions. The gas flow through these devices is called rarefied flow. The rarefaction is
quantified as the ratio of the mean free path (λ) of gas molecules to the characteristic
dimension of the system (l), which is called the Knudsen number (Kn) [2]. The range of
Knudsen numbers between 0.001 to 0.1 causes slight rarefaction due to which slip occurs at
the wall. It is called a slip regime. Flow problems of this regime can be modelled using
continuum fluid mechanics with slip boundary conditions. Gad-Al-Huk [3] had done flow
regime classification using the Knudsen number and studied boundary conditions in the slip
regime. In rarefied gas flow, fractional reduction of tangential momentum of gas molecules
occurs due to momentum exchange with the wall. This reduction in the momentum of gas
2
Some researchers [4]–[7] used Boltzmann and Burnett equations to model the rarefied gas
flow . Many authors [8]–[11] obtained experimental data and solved the problem of slip
regime analytically using the Navier-Stokes equation with slip boundary condition. Arkilic
et al. [9] used the perturbation method and solved the Navier-Stokes equation with the first-
order slip boundary condition. The authors also validated it against the obtained
experimental results. Some other notable works of researchers [12]–[15] have compared
their experimental results with the analytical solution of Arkilic et al. [9]. Zahid et al. [16]
Besides the plane model, Ebert et al.[17] and Morini et al.[18], [19] suggested a new
Apart from molecular dynamics, Lattice Boltzmann approach, computational fluid dynamics
is also used to simulate the gas flow through microchannel using the Navier-Stokes
equation. The Navier–Stokes equation with first-order slip and temperature jump boundary
condition was solved by Chen et al.[20] using a finite-difference methodology. The results
of pressure distribution were validated with Pong et al. [21] and the mass flow rate with
Arkilic et al. [9]. Roy et al. [22], [23] developed 2D code using the Navier -Stokes equation
with first-order slip velocity and temperature jump boundary condition. The authors solved
it using the Finite Element Method. The simulation was carried out by Morini et al. [18]
using the Navier-Stokes equation as a governing equation with a first-order slip model. They
analysed the effect of rarefaction on the flow. Cai et al. [24] further extended the work done
by Arkilic et al.[9] with the inclusion of the energy equation and the removal of the
Navier-Stokes equation with first-order slip boundary and temperature jump boundary
3
Besides 2D Analysis, some limited studies of the gas flow in the 3D channels are also
solved 3D rarefied gas flow using first-order slip boundary condition and studied the
To check the Navier-Stokes equations validity in the transition regime requires a higher-
order boundary condition. Towards this, initially, Schamberg et al. [29] obtained second-
order boundary conditions using Burnett approximate solution of the Boltzmann equation.
Although it was observed that the solution did not always have an agreement with
experimental data of Schaaf et al. [30] and in many cases, even Navier -Stokes equation
worked better than that. Deissler et al. [31] and Cercignani et al. [32] derived second-order
boundary conditions for slip flow at a higher Knudsen number. Dongari et al. [33] used
second-order boundary condition and obtained analytical solution for higher Knudsen
number flow. Few researchers performed experiments and validated their experimental data
using second order slip boundary condition. Sreekanth et al. [34] proposed a modified
boundary condition and given the general form of boundary condition containing
independent slip coefficient C1 and C2, based on experimental data obtained the flow in
circular tube. In addition to this, the authors also carried out the experimental measurement
of gas flow through the microchannel. Colin et al. [35], [36] used Deissler’s second-order
boundary condition and obtained mass flow rate expression. The measured data was
validated with analytical solution up to Knudsen number 0.22. The plane model was also
variation between the plane model and rectangular model results for channel of aspect ratio
(ar) of 0.01 and Kn = 0.053. The increase in the aspect ratio (ar) causes more deviation of
models.
4
Maurer et al. [37] have obtained experimental result up to mean Knudsen number 0.8 for
helium and upto 0.6 for nitrogen respectively. They obtained nondimensional mass flow rate
coefficients was obtained fitting the experimental data with polynomial of second order.
Ewart et al. [38] obtained experimental data with helium for Knudsen number from 0.03 to
50, and results compared with solution obtained from kinetic approach. Moreover, slip
coefficient was extracted by comparing the non-dimensional mass flow rate using first order
slip boundary for Knudsen number 0.03 to 0.3 and for 0.03 to 0.7 using polynomial of
second degree. In the transition and free molecular regime kinetic approach is used for slip
coefficient calculation. Graur et al. [39] measured the highly accurate data for different
gases for flow through microchannel to obtain the slip coefficient and study the flow
behaviour for gas flow in slip to free molecular regime (0.03-30). They have obtained
nondimensional mass flow rate expression for first and second order slip boundary condition
using continuum approach for Knudsen number 0-0.03. The slip coefficients for various
gases were obtained. The experimental data for slip to free molecular regime has been
compared with kinetic approach and analytical solution has been fitted using value of
variable slip coefficient as fitting parameter. They have concluded that irrespective of
Knudsen number and modelling approach TMAC decreases with increase in molecular
weight of gases. The values of TMAC are below unity this indicates that using full
accommodation coefficient is not right approach to deals with gas surface interaction.
Beskok et al. [40] proposed universal model to study the validity for wide range of Knudsen
number (0<Kn<∞) and compared it with DSMC and experimental data. They concluded that
model has good agreement for Kn ≤ 1, however the deviation between experimental and
simulation results was found 21% for Kn>1. Model used are based on diffused reflection
which is one of the limitations for accepting it as general slip model. The different slip value
5
are listed by Barber and Emersion [41] for model having two slip coefficients.
Vadiraj et al.[42] carried out an experimental measurement of the gas flow through a
trapezoidal microchannel having dimensions with a depth of 103µm, top and bottom width
as 1143µm, 998µm with the length of 2cm and bonded using quartz (dissimilar material).
They have performed experiment and obtained slip coefficient variation using the
Srinivasan et al. [43]–[45] fabricated microchannels with different aspect ratios (0.002, 0.01
and 0.1), channels were etched on the silicon wafer and bonded with another wafer using
fusion bonding. This fabrication method led to a controlled surface structure microchannel.
They have performed experiments and simulation modelling using second order boundary
condition and studied the flow behaviour. Reference [43] new analytical model was
developed using second-order slip boundary conditions to check the validity of Navier–
Stokes equations from slip to transition regime Authors model the experimental data up to
outlet Knudsen number 1.01 for nitrogen gas flow also determined the slip coefficient. In
Reference [44] further they modelled up to free molecular regime and concluded that up to
Kn ≤1 measured and experimental mass flow rate deviation are within 2% using slip
coefficients C1=1.31 and C2=0.11, and for Knudsen number ≥1 normalised volume flow rate
had a 3.30% deviation with other available experimental data. The Knudsen minimum was
Moreover, mass flow rate, pressure distribution and velocity distribution were plotted to
study fluid flow behaviour. Reference [45] they have studied flow behaver through smooth
and rough channel. Due to low roughness slip is high and TMAC value predicted as
6
Few researchers [36], [37], [42] obtained experimental data where channel was bonded with
dissimilar material and researchers [38], [39] with channel bonded with same material
(silicon to silicon) used the first and second-order boundary condition checked the
applicability and
using second order boundary condition was derived without any approximation to check the
validity of previous proximate model and further used to model the three-dimensional
rarefied gas flow through microchannel. To replicate the actual experimental condition,
microchannel having a trapezoidal cross-section with inlet and outlet plenums (as a
reservoir) is used for 3D simulation. The simulation results are compared with experimental
layer or kinetic layer and, its thickness is the order of one mean free path. When the mean
free path is comparable to the characteristic dimension of flow, the flow is called rarefied
flow. In rarefied gas flow, the collision frequency in Knudsen layer regions is insufficient to
bring quasi thermodynamic equilibrium. The partial exchange of momentum and energy in
A first-order slip boundary condition is initially used to model this problem, which has
accurately modelled the flow up to Kn = 0.10. It is evident from Figure 1 that there is a
deviation between actual slip velocity and extrapolated first-order slip velocity.
Furthermore, the increase in Knudsen number leads to a higher mean free path which causes
For higher Knudsen number, many researchers used Navier-Stokes equation with second-
order slip boundary condition and modelled the rarefied flow taking approximation in
7
expression as included in the introduction part. Although the second-order solution was the
approximate solution, its modelling ability is better than the first order. This indicates that
the magnitude of second-order slip velocity lies between actual and extrapolated slip
For the analytical modelling, we can refer to Srinivasan et al. [43]. Nondimensionalised
governing equations and second-order slip boundary conditions are used here directly.
The mathematical analysis steps are included to obtain slip velocity at the wall using the
(1)
(2)
y
Velocity profile
z x
Knudsen Layer
(O) λ
Q
Wall Actual
First order
8
The streamwise (x-direction) velocity profile is obtained after nondimensionalization, order
(3)
Now, the boundary conditions that were applied in equation (3) are given below.
(i) At
(ii) At
Implementing the boundary conditions, the value of constants is evaluated and given as:
(4)
In the above equation, Kn is variable and function of x coordinate. Knudsen number is based
on channel height, and it can correlate the local Knudsen number to the outlet Knudsen
Where is the Knudsen number based on the outlet condition of the microchannel.
(5)
9
The value of constant is evaluated using the boundary condition for v along the axis. From
the literature survey, Roy et al. [22] found that the velocity along the y-direction will vanish
at the centre and asymmetric along the x-axis. Applying the boundary conditions.
at , hence C = 0
(6)
Putting the boundary condition in equation (6) that the wall-normal velocity must be zero
(no penetration condition) at the wall and taking bottom wall condition.
So, at ,
(7)
(8)
The presence of a logarithmic term in the pressure equation makes the equation highly
nonlinear. In literature, reference [43] has done an order of magnitude analysis and
compared different terms in pressure equation (10) for slip regime using slip regime
experimental data. In literature reference [34], [37] it was found that C1 is of order one and
Reference [43] used nondimensional pressure variation data of Roy et al. [22] slip regime
data and performed the order of magnitude comparison using values C 1=1.13 and C2=0.13.
Based on the magnitude comparison of each term in the pressure equation, they replaced
C2ln~
p with a constant because the magnitude of the logarithmic term is negligible compared
10
The present work solved a highly nonlinear pressure equation (9) using the Newton Raphson
scheme to get pressure profile variation along the flow direction. It has seen from the
equation (9) for low Knudsen number influence of the third term 24Kn o2C2ln~
p is negligible.
However, as the Knudsen number increases, the logarithmic term will affect the pressure
equation and slip velocity, suggesting that it should be retained in expression at a higher
The pressure boundary conditions are used, and values of constants D and E are calculated.
Inlet and outlet pressure boundary conditions have been given as:
(i) At
(ii) At
(9)
(10)
(11)
Plot of nondimensional pressure variation with approximation and exact solution are shown
in Figure 2 for slip and transition regimes data. It was found that in slip regime both profiles
11
are quite matching shown in Figure 2 (i) for Kno=0.067, pressure ratio 1.84. However, in
early transition regime both profile are different the plot shown in Figure 2 (ii) for
Kno=0.22, pressure ratio 1.95, both data was taken from reference [47]. Another data of end
of transition regime taken from reference [44] shown in Figure 2 (iii) for Kno=0.90 pressure
ratio 1.93 the deviation clearly indicates two different profiles and it indicates that profile
(i)
(ii)
(iii)
analysis. Further using exact analysis slip velocity profile has been developed to apply on
12
Substituting the value of pressure gradient in equation (4) from equation (11), we get
(12)
To evaluate the slip velocity on the top and bottom wall, the value of ~y=± 1/2 have
(13)
Here
H- Depth of channel
-Nondimensional pressure
P- Pressure ratio
The slip velocity is calculated from the above equation (13) at the top and bottom wall,
Slip velocity at the wall was evaluated using equation (13) along the flow direction. The slip
is denormalised and implemented using UDF in Fluent with a suitable slip coefficient and
Fabrication, characterisation, and bonding have been carried out by reference [43] . Here
microchannel system used for the study of 3D flow is etched on the silicon wafer and
13
bonded to other silicon wafer using fusion bonding. The dimensions of the microchannel are
The creation of the geometry and mesh generation was done in ICEM CFD 18.1. The fine
mesh was used near the wall to capture the sharp gradient and coarse mesh as it approaches
the fluid core. The finite volume method was used to solve the governing equations in each
cell. The second-order upwind scheme is used in discretising momentum and energy
equation. The pressure discretisation is also second-order accurate. The convergence criteria
for continuity was obtained by dividing the experimental mass flow rate with the number of
cells present in the cross-section of the computational domain. The validation in gas flow
through the 3D microchannel was obtained for first-order and general second-order slip
boundary conditions. The first-order slip boundary condition was validated with Arkilic et
al. [9] experimental work and general second-order boundary condition with Colin et al.
experimental data [36] and [48]. A two-dimensional simulation is also performed for
flow with the slip velocity profile using the ideal gas model. The channel was having a
trapezoidal cross-section due to selective wet etching. The angle (θ) between the top wall
14
and sidewall of the channel is 54.7° shown in Figure 3. However, the present test section has
an array of twenty channels with a common inlet and outlet plenum. It was assumed that in
steady-state flow conditions, the flow was uniformly distributed to each channel. Therefore,
a single microchannel with a trapezoidal cross-section having an inlet and outlet plenum was
used for numerical simulation. The fluid domain of the microchannel with the inlet and
outlet plenum was discretised with structured hexahedral mesh. The 3D steady compressible
Navier-Stokes equation and the continuity equation were used as the governing equation.
The general second-order slip boundary condition was used on the top and bottom walls.
The viscosity of the working fluid is assumed to be constant. The equation of state was used
z x
θ H
Continuity equation:
(14)
(15
15
)
(16
(17
)
Equation of state:
(18)
This flow is compressible because of the substantial pressure drop along the flow direction
due to friction. So, to obtain the pressure in the flow field ideal gas model is used. In the
numerical simulation, the experimental mass flow rate through a single channel was taken as
the inlet boundary condition and pressure as the outlet boundary condition. Therefore,
general second-order slip velocity boundary condition was applied only on the top and
bottom walls. In the CFD simulation, general second-order slip boundary conditions cannot
be applied directly. So, to implement this boundary condition, the slip velocity profile was
solved analytically and implemented on the top and bottom wall using a user-defined
function.
To implement the slip velocity on sidewall outlet Knudsen number is calculated based on
channel width as characteristic dimension. The Knudsen number based on channel width
was 0.00021, which lies in the continuum regime. Therefore, no-slip boundary condition is
16
applied on the sidewalls of the channel. The experimental flow was assumed to be
isothermal because the material of silicon wafer has thermal conductivity ̴ 38 W/mK at room
temperature.
(19)
The numerical simulation was carried out because the momentum equations introduce
substantial complexity in the solution of flow problems due to the presence of the nonlinear
advection term and the second-order derivative viscous term. There is no general analytical
solution for the system of the equations because it cannot be directly converted into an
ordinary differential equation for 3D flow even with no-slip boundary conditions.
Many fluid flow problems with complex geometry and boundary conditions have no
analytical solution and were solved using the numerical technique. Nevertheless, the
problem with the analytical solution provides a significant relationship among the solution
parameters if it can be obtained. The combination of the numerical and analytical solution
was used to model the problem because, by default, Fluent has a no-slip boundary condition
at the wall. The slip boundary profile obtained along the streamwise direction analytically is
implemented on the top and bottom surface of the channel using UDF. For the slip velocity
profile, prerequisites are pressure gradient and slip coefficients; the pressure gradient also
optimal mesh selected for simulation has 0.025% of relative error of inlet pressure with
respect to the next level of refined mesh. A mesh having 5219900 nodes was selected in the
grid independence test. The convergence criteria of order 10-16 was taken for continuity.
17
Figure 4. Grid independency test for simulation of present experimental data.
required. The appropriate value of slip coefficient C 1 was obtained using the velocity profile
of the first-order slip boundary condition on the top and bottom wall of the channel.
Furthermore, using the slip coefficient C1 and a guess value of C2, the second-order slip
velocity was obtained and again simulated for the appropriate value of C2.
Steps involves for calculating the slip coefficients for first, general second-order boundary
conditions:
Step I
18
Step II
Step III
Step IV
Step V
Step VI
In every step, expression of the first-order slip boundary condition is written in the first row,
and the second one is related to the general second-order slip boundary condition. In the first
step, streamwise velocity profiles obtained from the analytical solution is given, which is a
function of streamwise (~
x ), its normal coordinates ~y and slip coefficients. In the second
step, the velocity profile normal to streamwise distribution is obtained using a continuity
equation. The aim of calculating this profile is to get the pressure profile. Now obtained
pressure profile expressions were given in the third step. In the fourth step, pressure gradient
expressions are given. The streamwise velocity is given in step fifth and slip velocity at wall
19
Estimation of first-order slip coefficients C1:
Using the guess value of slip coefficient C1 in the first-order pressure expression, the
pressure gradient is calculated using the fourth step. The calculated pressure gradient is used
in first-order velocity expression in the first step. Now, step fifth has been achieved after
substituting the pressure gradient in the first step. This first-order velocity expression is only
velocity as a function of ~
x is obtained are shown in step sixth.
Now in step sixth, we have got the first order slip velocity expression as a function ~
x only.
obtained. Now denormalisation was done to obtain the actual value of slip velocity, and
UDF is used to implement it on the wall of the channel. After, that 3D simulations are
performed (mass flow rate as inlet and pressure as outlet boundary conditions), and inlet
pressure was obtained and compared with experimental results. This process has been
repeated until we got appropriate values of first-order slip coefficient C 1. From Table 2, We
can see the simulated inlet pressure values and their percentage deviation (%δ) with respect
to experimental data using the first-order slip boundary condition. Here, C 1=1.18 is taken as
Table 2. Percentage deviation between measured and simulated inlet pressure for different
first-order slip coefficients for data of P=1.58.
20
Estimation of slip coefficients C1, C2:
The slip coefficient C1 (obtained from iterative method) and a guessed value of second slip
coefficient (C2) is substituted in pressure expression obtained from the second-order analysis
given in step three. Now pressure gradient has been obtained using step fourth. The obtained
pressure gradient is substituted in second-order velocity expression in step first and the
attained streamwise velocity expression shown in step fifth. Now in step fifth, velocity
velocity along the streamwise direction has been calculated. Now denormalising the velocity
and using UDF, this slip profile is implemented on the walls of the channel (top and
bottom). After implementing the second-order slip velocity profile, simulation was done
(mass flow rate as inlet and pressure as outlet boundary conditions), and inlet pressure was
obtained. This simulated pressure was compared with experimental results. This process was
repeated for different values of slip coefficient C 2. From Table 3, for the second-order
boundary condition lowest deviation is with the value of C 1 =1.18, C2 =0.15. However,
keeping the value of C1 =1.18, and varying value of C2 for all experimental data of the same
Knudsen number, the simulations have been performed and obtained deviations are shown
in Table 4. It can be observed that C 1=1.18, C2 =0.34 has minimum deviation. The
coefficients C1 =1.18, C2 =0.15 have not been selected because the reduction in C 2 might be
capturing uncertainty in experimental data. That is why it has more accurate for single
experimental data but for all six data average deviation more. So, finally, C 1=1.18 and
21
Table 3. Percentage deviation between measured and simulated inlet pressure for a fixed
[∑ ( ( ( ]
k=6
%δ=average P i simulated −Pi experimental ) /P experimental )∗100 )k
k =1
Table 4. Average percentage deviation between simulated and measured inlet pressure for a
fixed value of C1 and different value of C2 for all six data of Kno= 0.1028.
A new numerical model is required to validate with experimental and numerical data
available in the literature. In the present work, two sets of simulations were performed based
on data available in previous research works. The first set of simulations was performed to
validate the first order, and the other one was for validating the second-order general slip
boundary conditions. To provides the slip on walls of the channel, the slip velocity profile
was developed along the height and width of the channel separately. The 3D simulation for
a first-order slip model was validated against empirical results of Arkilic et al. [9] for Kno =
22
0.156. The second-order general slip boundary conditions for Kn o = 0.09 and 0.47 were
validated with experimental data (inlet pressure data) of Colin et al. [36], [48].
boundary conditions for the pressure ratio 1.15 to 2.48, outlet pressure 100.8 kPa, and outlet
Knudsen number 0.156. The working fluid used was helium. The 3D simulations with no-
slip boundary conditions have been performed on different mesh sizes for the grid
independence test. The optimal mesh having 0.01% of relative error in the value of inlet
pressure to the next level of refined mesh has been selected for simulations. The variation of
pressure at the microchannel inlet with respect to mesh size is shown in Figure 5 (i) and
Figure 5 (ii). A mesh having 450000 nodes was selected in the grid independence test to
(i) (ii)
Figure 5. Grid independency test for first-order validation (i) with reference [9] and (ii) with
reference [20].
The convergence criteria for continuity have been found of order 10 -15. So, to achieve this,
the mass flow rate is taken as inlet and pressure as outlet boundary conditions. For the
Knudsen number 0.156, the slip velocity profile was obtained for each pressure ratio flow.
The default no-slip boundary condition is replaced with a slip velocity profile using UDF to
23
implement slip boundary conditions on the channel walls. The simulation was performed for
the first-order model validation, and inlet pressure was plotted with respect to experimental
data of mass flow rate in Figure 6. It was found that the maximum deviation in the data of
Chen et al.[20] and Roy et al.[22] is 3.89% and 5.14%, respectively and, the maximum
deviation for the present simulation was 3.83% for inlet pressure.
Figure 6. Variation of mass flow rate with inlet pressure for slip condition at = 0.156
reference [9], and comparison with [20], [22].
Moreover, validation has also been done against numerical data of Roy et al. [23] for
velocity distribution along the centreline direction and transverse velocity of streamwise
direction. The 2D simulations were performed on different mesh sizes. The optimal mesh
selected for simulation was having 0.014% of the relative error of the mass flow rate at the
outlet with respect to the next level of refined mesh. A mesh having 41000 quadrilaterals
was selected in the grid independence test. The variation of mass flow rate at the inlet of
microchannel along with mesh size is shown in Figure 5(ii). The convergence criteria for
continuity was taken as 10-12 (for pressure boundary condition as the lowest criteria for
velocity are within 1.30% and 1.68%, respectively. The plots are shown in Figure 7 (i) and
(ii).
24
(i)
(ii)
Figure 7. Comparison of (i) Centreline velocity and (ii) transfer velocity along the length of
the channel for pressure ratio 1.34 and 2.70 with reference [23].
meshes having 450000 nodes for Kn=0.09 shown in Figure 8 (i) and 330000 nodes for
kn=0.47 shown in Figure 8 (ii) have been selected for the grid-independence test. For both
cases the mesh chosen has a relative error below of 0.03% of inlet pressure with respect to
the next level of refined mesh. The convergence criteria of the order 10-16 were taken for
continuity.
25
(i) (ii)
Figure 8. Grid independency test for validation of experimental data of (i) reference [36]
and (ii) reference [48].
So, to achieve convergence criteria, mass flow rate as inlet and pressure as outlet boundary
condition was used. The simulations were performed by applying the second-order
boundary conditions, and a converged value of pressure at the inlet of the microchannel was
Colin et al. [36] performed an experiment for gas flow through the microchannel and
developed an analytical model for the same. They found that for a flow-through channel
having an aspect ratio (ar) ≥0.01, the plane model underestimates the flow rate for Knudsen
number (Kn)≥0.05. They have also reported that the lateral wall affects the channel flow for
Furthermore, they found that their analytical model based on Deissler’s boundary conditions
have good agreement between experimental and analytical solutions for TMAC=0.93, up to
An exact analytical solution was used to obtain the slip velocity boundary profile in the
present work. The simulations were performed for flow through microchannel with slip
coefficients C1 =1.15 for the first order and general second-order boundary conditions with
slip coefficients C1 =1.15 and C2 =0.34. For Kn =0.09, the present simulation is in good
26
agreement with the experimental data, as shown in Figure 9 (i). The maximum deviation of
the first-order and Colin’s analytical solution are 0.50% and 2.36% w.r.t experimental value
and, the deviation of second-order simulation is 1.43%. The first-order, general second-
order and Colins model have good agreement with experimental data of slip regime. Figure
9 (ii) shows that Colin’s rectangular model overestimates the mass flow rate, due to which
(i)
(ii)
Figure 9. Comparison of inlet pressure against mass flow rate for (i) channel having an
aspect ratio of 0.055 at Kno= 0.09 (ii) channel having an aspect ratio of 0.01 at Kno= 0.47.
For Knudsen number 0.47 using the same slip coefficient as C 1=1.15 and C2=0.34,
simulations were performed. The obtained inlet pressure has good agreement with
experimental data having deviation 0.59%. The Colin’s analytical solution obtained by the
rectangular model underestimated inlet pressure up to 7.47 %. This deviation is due to the
overestimation of the mass flow rate. Based on the above experimental and numerical
27
validation, it can be inferred that the present modelling technique can correctly predict the
5. Experimental Set-Up
The experimental set-up used in this work is developed in the turbomachinery Laboratory at
the Indian Institute of Technology Delhi. A cylinder filled with nitrogen gas up to 120 bar is
the source of the working fluid. The pressure regulator was connected to the gas cylinder to
provide the gas flow at constant pressure during experimental work. The Festo filters of
specifications (5, 1, 0.01µm) are used to filter the particle of bigger size and other impurities
present in the working fluid. Pressure transducer having an accuracy of ±0.1% of full-scale
is used for measuring inlet and outlet pressure of microchannel. The temperature
measurement was done by a resistance temperature detector (RTD) sensor with accuracy
±0.5% of full scale, and the temperature was maintained constant during the experiments (at
35.5℃). Mass flow meters of various ranges, like 0.5, 1, 2 Sccm was used. It showed
accuracy ±0.15% of full scale connected to the computer with data acquisition system. The
vacuum generator is used before the inlet of the microchannel to create low pressure. The
vacuum assembly is connected at the channel outlet to create a vacuum to achieve the
required Knudsen number. The whole experimental set-up is assembled on class 10 clean
bench.
recorded for outlet Knudsen number of 0.1028. The mass flow rate variations were
measured and recorded for various pressure ratios (1.24-1.96). The results of inlet pressure
28
6.1. The Methodology and Experimental Data
The detailed methodology is presented for the experiment of a gas flowing through the
microchannel. To achieve the required outlet Knudsen number, the vacuum is created inside
the vacuum chamber and pressure is provided at the inlet of the microchannel as per the
design pressure ratio. The system is first allowed to achieve a steady-state, which generally
takes about 30-45 min. The experimental data of pressure at the inlet, outlet and mass flow
For outlet Knudsen number 0.1028, outlet pressure is maintained at 6.51×10 4 Pa, and inlet
pressure varies from 8.07×104 Pa to 1.28×105 Pa. Experiments was performed for six
designed pressure ratios range from 1.24 to 1.96. Experimental data of nitrogen gas flow
In this work, the UDF has been developed and implemented in Fluent to provide slip
velocity at channel walls (top and bottom). The slip velocity is obtained using the exact
analytical solution with the appropriate independent slip coefficient C1=1.18 and C2=0.34.
The simulations are carried out for six different pressure ratios. The experimental mass flow
rate was taken as inlet boundary condition and pressure as an outlet boundary condition. The
converged values of inlet pressures were obtained using slip and no-slip boundary
conditions at the wall to verify the slip effect. An average deviation of error between
experimental and simulation results was calculated. The results of simulated inlet pressure
with no-slip boundary condition have a deviation of 17.44% with respect to the measured
value of inlet pressure. The positive error indicates that higher inlet pressure is required due
to the no-slip condition at the wall to maintain the same flow rate. This research focused on
29
the applicability of the second-order model with an exact analytical solution over a higher
having length 5400 µm, top width 488.7 µm, bottom width 487.26 µm and height 1.02 µm.
The simulation was performed by employing first-order and general second-order boundary
conditions with the appropriate slip coefficient (C 1=1.18 and C2=0.34). The uncertainty has
also been calculated for the accuracy of the experimental results, which is shown in plot of
Figure 10.
Figure 10. Variation of experimental and simulation inlet pressure with mass flow rate for
Kno= 0.1028.
The variation of simulated inlet pressure against experimental mass flow rate is plotted in
Figure 11. The first-order slip model results deviated from the experiment results by 2.77%,
30
Figure 11. Variation of experimental and simulation inlet pressure with mass flow rate for
Kno= 0.1028.
It indicates that the general second-order model can better predict the inlet pressure
conditions at a higher Knudsen number. This model also works better than Colin et al. [36]
For the study of microflow behaviour, pressure distribution along the streamwise direction is
presented in Figure 12. It shows its nonlinear nature along the flow direction and constant in
31
Figure 12. Pressure distribution along the length of the microchannel
X-component of velocity (u) is plotted in the x-y plane passing through centreline at
different locations of x-coordinate as x=1800µm, 2800µm and 4500µm and shown in Figure
13. It is evident from the plot that fluid is continuously accelerating along the flow direction.
are shown in Figure 14. Due to symmetry, velocity is zero at the centre, and because of no
penetration condition, velocity is again zero at the wall. The v velocity achieves maximum
32
The variation of z-component of velocity was plotted in x-z plane containing centreline axis
at different value of x coordinates which is shown in Figure 15. The velocity at the centre is
zero, and velocity at the wall is also zero because of no penetration condition. There is no
slip velocity at the sidewall along the flow direction because the Knudsen number based on
channel width as a characteristic dimension lies in the continuum regime for present case.
channel shows that the y and z-component of velocity are zero at centre and wall, achieving
maximum between centre and wall. The streamwise velocity is maximum at the centre and
has slip conditions on the top and bottom wall. In the plot, it is evident that each velocity
profile is continuously growing in the flow direction. This shows that this 3D flow is
developing in nature. Furthermore, an increase in lateral component and slip velocity in the
7. Conclusion
Experimental and numerical simulations was carried out to model the rarefied gas flow
through trapezoidal microchannel with inlet and outlet plenums with second order exact
analytical solution used for slip profile. The conclusions are summarised as follows.
1. The new exact analytical solution for the slip velocity is developed using a highly
nonlinear pressure gradient derived from the pressure equation. The pressure
33
technique developed is validated with the numerical solution and experimental data
result is 0.59%.
3. This model can be used for 3 dimensional modelling of rarefied flow in transition
regime.
4. The average deviation of present experimental (Kno=0.1028) and simulation data for
inlet pressure using the general second-order slip boundary condition with slip
coefficient (C1=1.18, C2=0.34.) is 0.51%. The simulations results agree well with the
experimental data.
5. The pressure profile plotted along the centreline of the microchannel shows
34
Nomenclature
A Constants
ar An aspect ratio of the microchannel (H/W)
B, C Constants
C1 and Slip coefficient
C2
D, E Constants
H Depth of microchannel [m]
Kn Local Knudsen number
Kno Outlet Knudsen number
L Length of microchannel [m]
l The characteristic dimension of the system [m]
M Mach number
p Local pressure [Pa]
pi
P
Pressure ratio po
p Nondimensional pressure
pi Inlet pressure [Pa]
Outlet pressure [Pa]
R Gas constant []
35
δ Deviation in experimental and simulated data
Subscript
i inlet
o Outlet
rms Root mean square
s Surface
w Wall
36
Footnotes
Hariom Saran Singh email: om.saran7@gmail.com Department of Mechanical Engineering Indian
37
8. Reference
[1] M. Gad-el-Hak, The MEMS Handbook, Boka Raton. CRC Press, 1997.
[2] C. Shen, Rarefied gas dynamics: fundamentals, simulations and micro flows. Berlin:
microchannels,” Phys. Fluids, vol. 16, no. 9, pp. 3426–3437, 2004, doi:
10.1063/1.1764700.
[5] H. Xue, H. M. Ji, and C. Shu, “Analysis of micro-Couette flow using the Burnett
equations,” Int. J. Heat Mass Transf., vol. 44, pp. 4139–4146, 2001.
[6] B. Y. Cao, M. Chen, and Z. Y. Guo, “Rarefied gas flow in rough microchannels by
molecular dynamics simulation,” Chinese Phys. Lett., vol. 21, no. 9, pp. 1777–1779,
38
doi: 10.1109/84.585795.
[11] Y. Zohar et al., “Subsonic gas flow in a straight and uniform microchannel,” J. Fluid
[12] J. C. Shih et al., “Monatomic and Polyatomic Gas Flow Through Uniform
[13] S. Takuto, A., Soo, K. M., Hiroshi, I., and Kenjiro, “An Experimental Investigation of
Gas Flow in Microchannels,” J. Heat Transfer, vol. 126, no. October 2004, pp. 753–
[15] S. Hsieh, H. Tsai, C. Lin, C. Huang, and C. Chien, “Gas flow in a long
microchannel,” J. Heat Mass Trens., vol. 47, pp. 3877–3887, 2004, doi:
10.1016/j.ijheatmasstransfer.2004.03.027.
[16] W. A. Zahid, Y. Yin, and K. Zhu, “Couette Poiseuille flow of a gas in long
006-0108-5.
[17] W. A. Ebert and E. M. Sparrow, “Slip Flow in Rectangular and Annular Ducts,” J.
Basic Eng., vol. 87, no. 4, pp. 1018–1024, 1965, doi: 10.1115/1.3650793.
39
[18] G. L. Morini, M. Spiga, and P. Tartarini, “The rarefaction effect on the friction factor
of gas flow in microchannels,” Superlattices Microstruct., vol. 35, no. 3–6, pp. 587–
Experimental Thermal and Fluid Science, vol. 35, no. 6. pp. 849–865, 2011, doi:
10.1016/j.expthermflusci.2011.02.005.
microchannels,” Numer. Heat Transf. Part A Appl., vol. 33, no. 7, pp. 749–762, 1998,
doi: 10.1080/10407789808913964.
[21] Y. Pong, K. C., Ho, C., Liu, J., and Tai, “Non-Linear Pressure Distribution in
Uniform Microchannels.,” . Appl. Microfabr. to fluid Mech., vol. FED-197, pp. 51–
56.
flow through microchannels and nanopores,” J. Appl. Phys., vol. 93, no. 8, pp. 4870–
[23] R. Raju and S. Roy, “Hydrodynamic prediction of high speed microflows,” in 33rd
AIAA Fluid Dynamics Conference and Exhibit, no. June, doi: 2003, 10.2514/6. 2003-
4010.
[24] C. Cai and I. D. Boyd, “Compressible gas flow inside a two-dimensional uniform
microchannel,” J. Thermophys. Heat Transf., vol. 21, no. 3, pp. 608–615, 2007, doi:
10.2514/1.29362.
40
[25] C. S. Chen, “Numerical method for predicting three-dimensional steady compressible
different aspect ratio microducts,” Phys. Fluids, vol. 18, no. 10, 2006, doi:
10.1063/1.2354546.
[29] R. Schamberg, “The fundamental differential equations and the boundary conditions
for high speed slip-flow, and their application to several specific problems,” Ph.D.
boundary conditions for rarefied gases,” Int. J. Heat Mass Transf., vol. 7, no. 6, pp.
[32] C. Cercignani and A. Daneri, “Flow of a rarefied gas between two parallel plates,” J.
Appl. Phys., vol. 34, no. 12, pp. 3509–3513, 1963, doi: 10.1063/1.1729249.
41
[33] N. Dongari, A. Agrawal, and A. Agrawal, “Analytical solution of gaseous slip flow in
long microchannels,” in International Journal of Heat and Mass Transfer, 2007, vol.
[34] A. K. Sreekanth and M. Field, “Slip Flow Through Long Circular Tubes,” in
Proceedings of the 6th International Symposium on Rarefied Gas Dynamics, pp. 667–
680, 1969.
[35] C. Aubert and S. Colin, “High-order boundary conditions for gaseous flows in
Order Slip Flow Model in Rectangular Microchannels,” J. Heat Transf. Eng., vol. 25,
microchannels for helium and nitrogen,” Phys. Fluids, vol. 15, no. 9, pp. 2613–2621,
[38] T. Ewart, P. Perrier, I. A. Graur, and J. G. Méolans, “Mass flow rate measurements in
[40] A. Beskok and G. E. Karniadakis, “Report : A model for flows in channels, pipes, and
42
ducts at micro and nano scales,” Microscale Thermophys. Eng., vol. 3, no. 1, pp. 43–
microchannels: From slip to transition,” Heat Transf. Eng., vol. 27, no. 4, pp. 3–12,
momentum accommodation coefficient and slip coefficients for rarefied gas flow in a
microchannel,” Sadhana - Acad. Proc. Eng. Sci., vol. 43, no. 10, 2018, doi:
10.1007/s12046-018-0929-4.
Studies on Gas Flow Through Silicon Microchannels,” J. Fluids Eng. Trans. ASME,
and numerical study on gas flow through microchannels from slip to free-molecular
10.1088/1361-6439/aac4d5.
Phys. Fluids, vol. 15, no. 8, pp. 2352–2354, 2003, doi: 10.1063/1.1587155.
43
[47] K. Srinivasan, P. M. V. Subbarao, and S. R. Kale, “Experimental and Numerical
[48] S. Colin, “Rarefaction and compressibility effects on steady and transient gas flows in
0002-y.
44
Acknowledgement
Decalration of Interest
On behalf of all the authors, the corresponding author states that there is no conflict of interest.
45
Tables
Table 6. Dimensions of the microchannel system
46
Table 7. Percentage deviation between measured and simulated inlet pressure for different
first-order slip coefficients for data of P=1.58.
47
Table 8. Percentage deviation between measured and simulated inlet pressure for a fixed
48
Table 9. Average percentage deviation between simulated and measured inlet pressure for a
fixed value of C1 and different value of C2 for all six data of Kno= 0.1028.
Second-order slip
coefficients
%δ
S.N Kno C1 C2
o
1 0.1028 1.18 0.69 0.84
2 0.1028 1.18 0.34 0.51
3 0.1028 1.18 0.20 0.53
4 0.1028 1.18 0.15 0.53
5 0.1028 1.18 0.13 0.55
49
Table 10. Experimental condition for Nitrogen gas flow
50
List of figures
Figure 1.
y
Velocity profile
z x
Knudsen Layer
(O) λ
Q
Wall Actual
First order
51
Figure 2.
(i)
(ii)
52
(iii)
Figure 17. Variation of approximate and exact nondimensional pressure distribution (i)
Kn=0.667 presure ratio 1.84 (ii) Kn=0.22 pressure ratio 1.93 (iii) ) Kn=0.90 pressure ratio
1.93.
53
Figure 3
z x
θ H
54
Figure 4
Figure 19. Grid independency test for simulation of present experimental data.
55
Figure 5
(i) (ii)
Figure 20. Grid independency test for first-order validation (i) with Reference [9] and (ii)
with Reference [20].
56
Figure 6
Figure 21. Variation of mass flow rate with inlet pressure for slip condition at = 0.156
reference [9], and comparison with [20], [22].
57
Figure 7
(i)
(ii)
Figure 22. Comparison of (i) Centreline velocity and (ii) transfer velocity along the length of
the channel for pressure ratio 1.34 and 2.70 with reference [23].
58
Figure 8
(i) (ii)
Figure 23. Grid independency test for validation of experimental data of (i) reference [36]
and (ii) reference [48].
59
Figure 9
(i)
(ii)
Figure 24. Comparison of inlet pressure against mass flow rate for (i) channel having an
aspect ratio of 0.055 at Kno= 0.09 (ii) channel having an aspect ratio of 0.01 at Kno= 0.47.
60
Figure 10
Figure 25. Variation of experimental and simulation inlet pressure with mass flow rate for
Kno= 0.1028.
61
Figure 11
Figure 26. Variation of experimental and simulation inlet pressure with mass flow rate for
Kno= 0.1028.
62
Figure 12
63
Figure 13
64
Figure 14
65
Figure 15
66
Figures caption
Notes on contributors
Institute of Technology Delhi. He received his master’s degree from IIT BHU in 2011.
Pradesh (Guna) India. He joined IIT Delhi in 2014 and his research work includes
experimental and numerical modelling of rarefied gas flow through straight and complex
microchannels.
Amit Kakkar
Nainital in year 2002 (Gold Medalist). He then joined as a faculty member in the
Technology Delhi for pursuing his M.S. (Research) studies in 2010. His main research
67
Paruchuri M.V. Subbarao is a professor in Mechanical Engineering Department Indian
Institute of Technology Delhi. He received Phd degree from IIT Kanpur in 1994. His
computational study of high-speed gas dynamics, turbulent flow fluid mechanics and heat
transfer. He is author 4 book and of more than 100 articles. He is currently Ray W. Herrick
68