International Journal of Rotating Machinery, 10: 1–13, 2004

Copyright
c Taylor & Francis Inc.

ISSN: 1023-621X print
DOI: 10.1080/10236210490258016

Pumping Speed Measurement and Analysis

for the Turbo Booster Pump
R. Y. Jou, S. C. Tzeng, and J. H. Liou
Department of Mechanical Engineering, Chien Kuo Institute of Technology, Changhua,
Taiwan, Republic of China

also made with other turbo pumps. The compared results

This study applies testing apparatus and a computational show that the turbo booster pump presented here has good
approach to examine a newly designed spiral-grooved turbo foreline performance.
booster pump (TBP), which has both volume type and mo-
mentum transfer type vacuum pump functions, and is capa- Keywords CFD, DSMC, Flow meter method, Transition flow, Turbo
ble of operating at optimum discharge under pressures from booster pump
approximately 1000 Pa to a high vacuum. Transitional flow
pumping speed is increased by a well-designed connecting
element. Pumping performance is predicted and examined INTRODUCTION
via two computational approaches, namely the computa- Review of the Turbo Pump Designs
tional fluid dynamics (CFD) method and the direct simu-
Plasma-based etch and chemical vapor deposition (CVD) de-
lation Monte Carlo (DSMC) method. In CFD analysis, com-
pend on the reaction of gas molecules and reactive ions on the
parisons of measured and calculated inlet pressure in the
wafer surface. The concentration, arrival rate and directionality
slip and continuum flow demonstrate the accuracy of the
of reactive gases and ions determine the etching rate, deposi-
calculation. Meanwhile, in transition flow, the continuum
tion rate, step coverage, process uniformity, and etching profile.
model of CFD is unsuitable for calculating such rarefied
Meanwhile, the parameters of concentration, arrival rate and di-
gas. The pumping characteristics for a full 3D model on a
rectionality are in turn determined by flow rates and chamber
rotating frame in transition and molecular regimes thus are
pressure, as well as by plasma energy and the distance between
simulated using the DSMC method and then confirmed ex-
the plasma and wafer. The above two processes differ from alter-
perimentally. However, when the Knudsen number is in the
natives in that they require high flow rates, and relatively shallow
range 0.5 < Kn < 0.1, neither CFD computation nor DSMC
vacuum levels. Specifically, pressures of around 10 mTorr are re-
simulation is suitable for analyzing the pumping speed of
quired, along with flow rates of up to 800 SCCM. Those require-
the turbo booster pump. In this situation, the experimental
ments become increasingly stringent with the move to 300 mm
approach is the most appropriate and effective method for
wafer and 0.25 to 0.18 µm designs. Consequently, pumps used
analyzing pumping speed. Moreover, the developed pump
for pumping CVD or etching equipment in the IC industry re-
is tested using assessment systems constructed according to
quire superior discharge characteristics in conditions ranging
ISO and JVIS-005 standards, respectively. Comparisons are
from atmospheric pressure to a high vacuum.
Momentum transfer from the high-speed rotating blades to
the gas molecules affects the action of a turbomolecular pump
Received 19 June 2002; accepted 31 December 2002. (TMP). In molecular drag pump (MDP) designs, the pumped
The National Science Council under the contract number of NSC90-
2212-P-270-001 provides supports for this work. All authors deeply fluid generally moves parallel to the moving wall, while in TMP
appreciate the experimental support of Precision Instrument Develop- designs the movement is perpendicular. Maurice (1974) pro-
ment Center, National Science Council in Taiwan. Support of Ms. Y. W. posed a significant improvement to the MDP design that en-
Chang in DSMC simulation and Dr. H. P. Cheng in CFD calculation hanced pump compression ability by reducing the influence of
is also highly appreciated. Both of them have done much excellent internal leakage. Moreover, Henning (1988) reviewed 30 years
research in this field.
Address correspondence to R. Y. Jou, Department of Mechani- of turbo type and turbodrag type pump design to trace the devel-
cal Engineering, Chien Kuo Institute of Technology, Changhua 500, opment of turbo pumps. Furthermore, Duval et al. (1988) dis-
Taiwan, Republic of China. E-mail: ryjou@ckit.edu.tw cussed the development history of molecular pumps. Duval et al.

1
2 R. Y. JOU ET AL.

noted that modern MDP possess some interesting characteris- (1913), Holweck (1923), Siegbahn (1943), and Becker (1961)
tics, particularly a high compression ratio in the viscous flow are being proposed. These designs generally improve pumping
range, enabling MDP to operate as a booster and to operate in performance in certain ways, including either pumping speed
the 10−2 to 10−1 mbar at full pumping speed. Sawada et al. (1990, or compression ratio. Because of uncertainty regarding natu-
1992, 1993) proposed an analytical method for investigating the ral and treatment difficulties of the transitional flow problem, it
pumping performance of spiral grooved visco-vacuum pumps is difficult or even impossible to improve pumping speed and
with arbitrary groove cross-sections. Moreover, Sawada et al. compression ratio simultaneously, especially at higher exhaust
also experimentally assessed this theoretical model. pressure. Jou et al. (2000) presented a turbo booster pump (TBP)
Cleanliness is an important consideration in vacuum turbo with smooth and good performance in the transition regime. As
pump design. The most common method of ensuring cleanli- shown in Figure 1, this design has a compact structure com-
ness involves increasing the tolerable discharge pressure to allow pared with other available designs and differs significantly from
the use of oil-free backing pumps. The compound turbo pump, available methodologies in terms of design principles.
which complements the conventional TMP design by incorpo-
rating a turbo-drag section on the same rotation shaft, permits
discharge pressures of 1000 to 2000 Pa. Meanwhile, Levi (1992) Analysis and Simulation for Turbo Pump Designs
developed another type of pump, called a hybrid pump, which Regarding the numerical analysis of turbo vacuum pumps,
combines different impellers on the same axis without increasing Lee and Lee (1996) investigated the compression characteris-
the number of stages. Furthermore, Hablanian (1994) reviewed tics for a simplified drag pump model. Moreover, Schneider
the development history of turbine-type, high vacuum pumps et al. (1998) built an empirical and 2D numerical calculation for
and noted that the optimization process for designing a general- a single stage turbomolecular pump. Additionally, Nanbu et al.
purpose turbopump for high-vacuum use is incomplete. Finally, (1991) applied DSMC to design spiral grooves on a turbomolec-
Heldner and Kabelitz (1990) compared the design and perfor- ular pump rotor. Heo and Hwang (2000) also used DSMC to
mance characteristics of the most widely used rolling element calculate the blade row of a TMP in the molecular and transition
bearing system to more advanced bearing systems and noted that flow regions. Heo and Hwang concluded that the maximum com-
fully magnetic bearing designs are growing in popularity, and pression ratio and pumping speed depend strongly on the Knud-
becoming a design standard for IC industry application. Table 1 sen number in the transition region, and also depend weakly
briefly lists some basic characteristics of the TMP, MDP, and on the Knudsen number in the free molecular flow region. In
compound molecular pump (CMP). those studies, when dealing with the simulation of turbomolec-
The main advantage of compound turbo pumps is that they ular pumps, it is usual to fix the analysis model on the rotating
can exhaust at relatively high pressures, so high that simple, dry rotor and make some assumptions and simplifications. Owing to
diaphragm pumps can be used for backing. Turbo pumps also are the geometrical characteristics of the turbo booster pump rotor,
often used as an upstream booster or “supercharger” for a Roots as depicted in Figure 2, Chang et al. (2001) proposed a full 3D
blower, which has poor throughput in low milliTorr pressure DSMC model on a rotating frame that included the design pa-
ranges, but pumping speed still tends to decline. Currently, vari- rameters of clearance between the rotor and the stator and blade
ous pumps combining the traditional design principles of Gaede thickness. Additional treatments related to Coriolis acceleration

TABLE 1
Characteristic List of TMP, MDP, and CMP
TMP MDP CMP (TMP + MDP)
Rotor configuration

Pumping speed High Relatively small High

Foreline pressure Very low (<1 Pa) High (1000 to 2000 Pa) High (1000 to 2000 Pa)
Transitional flow performance Poor Good Good
Structure Complex Compact Complex; with TMP element
Compact; without TMP element
Manufacture Difficult Easy Difficult; with TMP element
Easy; without TMP element
 PUMPING SPEED MEASUREMENT AND ANALYSIS 3

pumps. However, to provide preliminary testing results for re-

search, this study adopts the original standard specifications for
turbomolecular pump testing.
To test the entire operating range of the turbo pumps, a sys-
tem that combines the measurement functions of the flow meter
and orifice methods should be developed. This approach would
facilitate the calculation of pumping speeds under a constant
inlet pressure condition based on the stable chamber pressure,
mass flow into the chamber, conductance of the orifice, and
the final base pressure of the system. The equation for eval-
uating pumping speeds adopted in ISO/CD5302 is presented
below.

Flow Meter Method

The steady pressure method is adopted to measure the volume
flow rate S by the determined gas throughput Q. Given constant
pressure P1 in a given area of the test dome (Figure 3a), which is
measured by a pressure gauge, the volume flow rate S is obtained
by the relationship

Q
S= [1]
P1 − P0

where P0 denotes the ultimate operational pressure in the test

dome.

Orifice Method
The method adopted to measure the volume flow rate S is the
steady pressure “standard conductance” method in which a thin
FIGURE 1 orifice plate divides the test dome (Figure 3b) into two volumes.
Schematic drawing of the spiral-grooved turbo booster pump If the pressure in each volume is measured by pressure gauges
with a five DOF actively controlled magnetic bearing with the same sensitivity, the volume flow rate is given by
(AMB) system.
P1
S=C [2]
and centripetal acceleration are done using the DSMC method. P2 − 1
Finally, Cheng et al. (2000) used CFD calculation to examine
the flow field of this turbo booster pump. Comparisons of mea- where C denotes the calculated conductance taking account of
sured and calculated inlet pressure in the slip and continuum flow the orifice size and gas properties. The orifice conductance is
demonstrated that the calculation was quite accurate. However, evaluated by following equation
the continuum model of CFD in unsuitable in terms of transition
π RT 1
flow for calculating such rarefied gas. C= d2 [3]
32M (1 + L o /d)

where L denotes the thickness of the orifice plate, the universal

PERFORMANCE TESTING APPARATUS gas constant R equals 8314 N-m/(mol-K), and the molecule
Numerous standards exist for evaluating turbomolecular weight of air is 28.8 × 10−3 Kg/mol. Finally, the term 1/(1 +
pump performance, including ISO/CD5302, AVS 4.1, 4.2, 4.3, L o /d) represents a correction-factor which can be defined as the
DIN28428, JVIS-005 and so on. However, these testing spec- average throughput probability.
ifications contain some discrepancies. One issue is especially Jou et al. (2000) designed a testing system for investigat-
important in turbo pump testing. Since the operating pressure ing the performance of the turbo booster pump under high inlet
range of turbomolecular pumps differs from that of compound pressure. The system complies with ISO standards, and can be
or hybrid turbo pumps, the elongated operating foreline pres- applied to pumping speed testing using the flow meter method.
sure of composite-type pumps requires detailed consideration. Figure 4 displays the schematic layout of this system. The sys-
The present standard is only applicable to turbomolecular pump tem includes a testing chamber, auxiliary pumping system, vac-
testing, and is unable to assess functions of compound turbo uum gauges, mass flow meters, and an RGA. Table 2 presents
4 R. Y. JOU ET AL.

(a)

(b)
FIGURE 2
(a) Rotor configuration of the spiral-grooved turbo booster pump. (b) Computational domain and boundary conditions for each
surface of the turbo booster pump’s rotor.

the major testing functions of the system. A testing appara- PUMPING ANALYSIS BY CFD
tus constructed by Kashiyama Co. Ltd. in Japan is adopted AND DSMC METHODOLOGIES
to compare the testing functions of the ISO and JVIS stan- Numerical simulation of vacuum pump flow provides de-
dards. Table 3 briefly compares these two testing systems. In tailed information on fluid flow field during flow through the
these apparatus, a series of precision flow meters with differ- pumping channels. Important benefits of the numerical approach
ent measurement range are installed to adjust the rate of ni- include sufficient accuracy for engineering designs, fast
trogen flow into the chamber. Furthermore, pressure gauges of response to design changes (for example through eliminating
a B-A gauge for ultimate pressure indication in high vacuum a lengthy trial-and-error process), and saving on prototype de-
measurements, and a set of Baratron gauges for low vacuum velopment costs since the design performance can be simulated
measurements are attached to the system to measure the equi- and displayed on a computer before actual pump manufacture.
librium pressure inside the testing chamber. Pumping perfor- Depending on operating pressure ranges, two approaches are
mance can easily be assessed by flow meter and pressure gauge commonly used to examine and characterize a pumping flow
readings. in detail. Specifically, the computational fluid dynamics (CFD)
 PUMPING SPEED MEASUREMENT AND ANALYSIS 5

(a) methodology is used to analyze continuum flows and the di-

rect simulation Monte Carlo (DSMC) method is used for the
molecular flow simulation.
The Knudsen number, Kn in Table 4, which is defined as a
ratio of mean free path to a characteristic rotor length (such as
the groove depth), is used to distinguish the flow regimes and
identify the most appropriate method for adoption. No clear dif-
ference exists between the suitability of the CFD and DSMC
methods, both of which suffer limitations in transitional flow
regimes (in the approximate range 0.1 < Kn < 10). Conse-
quently, computations should be conducted to cover this range
as large as possible. Figure 5 outlines the entire procedure in
flowchart form. According to the design requirements, analyses
are conducted sequentially from low to high vacuum levels.
When inlet pressure follows a viscous flow regime (normally
Kn < 0.1, but calculations are performed for the upper limit
of Kn), based on a set of prescribed inlet conditions and outlet
mass flow conditions, the CFD approach is used to calculate the
inlet pressures achieved by the pump. The turbo booster pump
analysis makes the following basic assumptions.
(b)
1. Flow field is a laminar.
2. Gases satisfy the Newtonian constitutive law.
3. Gases are ideal and satisfy the state equation for an ideal gas.
4. Flow is isothermal.

The CFD code, UNIC, employs finite difference approxima-

tions to discretize transport equations on body-fitted grid-mesh
systems. Specifically, the simulation uses an adoptive third-order
upwind total variation diminishing (TVD) scheme to approxi-
mate the convective parts of the transport equations. Moreover,
the simulation uses a second-order central differencing scheme
for the viscous and source terms of the governing equations.
The present simulation also uses a pressure-based predicted/
corrector solution procedure to enhance the velocity-pressure
coupling and mass-conserved flow field solutions at the end of
each time step. Finally, an implicit Euler time-marching scheme
is used to enhance the convergence for steady-state flows. The
coordinates used for the analyses are fixed on the rotor, which
rotates at a constant speed. From the calculation, the gas pres-
sure distribution along the pumping channel, pressure differ-
ence between pump inlet and outlet, and gas pumping flow
rate can be predicted with reasonable accuracy. For compress-
ible flow computations such as vacuum pumping, a state equa-
tion of the following form is employed to close the governing
equations.

P
ρ= R  [4]
M
T

As illustrated in Figure 2b, the boundary conditions for the vac-

uum pumping flow analysis considered here include setting the
FIGURE 3 inlet condition to a fixed volume flow rate, and an absence of slip
Test domes of ISO/CD5302. (a) Flow meter method, (b) orifice motion of molecules on the wall. A periodic boundary condition
method. is subjected for each periodic computational domain.
6 R. Y. JOU ET AL.

FIGURE 4
Schematic layout of the measurement system designed by the flow meter method following the ISO testing specification.

DSMC simulation is utilized to calculate the pumping char- DSMC scheme is similar to that of the CFD scheme, as shown
acteristics of the turbo booster pump when its operating inlet in Figure 5.
pressure is in the molecular flow regime (normally Kn > 10, In DSMC analysis, for the sake of consistency, the follow-
but calculations must be extended to the lower limit of Kn). ing values are fixed; (a) throughput, which can be expressed as
The DSMC methodology involves the computerized simulation Q = Pin Sin = Pout Sout (where P denotes pressure, S represents
of molecular kinetics. The transport terms in the Navier-Stokes pumping speed, and subscripts in and out denote inlet and outlet
equations of the continuum gas dynamics fail when the gradi- respectively), and (b) pressure at the outlet, namely Pout . Given
ents of the macroscopic variables become so steep that their equal flow outside the inlet and outlet ports, the outlet influx can
scale length is of the same order as the average distance traveled be easily determined based on kinetic theory, yet the inlet influx

by the molecules between collisions, or the mean free path. Es- Nin,i cannot be determined because of the unknown Pin and Sin .
sentially, except for the solver, the computational procedure of Consequently, the following iterative procedures for adjusting

TABLE 2
Functions of the Turbo Pump Testing System

Items Instruments Testing ranges

Ultimate pressure GP360 7×10−4 ∼7.5 × 10−10 Torr
Ionization gauge
Pumping speed MKS Baratron 690 1 atm∼10−4 Torr
Capacitance diaphragm gauge
Throughput Kofloc 3920 <1SCCM, 0∼20 SCCM
Sierra Sidetrak 840 0∼200 SCCM, 0∼2 SLM
Mass flow meter 0∼20 SLM, 0∼500 SLM
Residual gas Blazers QME-200 0∼300 amu
Residual gas analyzer
 PUMPING SPEED MEASUREMENT AND ANALYSIS 7

TABLE 3 ulated molecules at each step. Consequently, the criterion for the
Comparisons of Testing Systems Designed by ISO and JVIS attainment of a steady value for Pin used in this investigation is
Standards that the relative difference between the average value of Pin in
the previous two thousand steps and the following two thousand
PIDC system (ISO) Kashiyama system (JVIS)
steps is below 0.5%.
Ion gauge (GP 360) B-A gauge (for ultimate pressure)
CDG (1 Torr, 10 Torr, Schultz gauge RESULTS AND DISCUSSION
1000 Torr) The rotor configuration presented in Figure 2 is the design
CDG examined in this study. This rotor is operated at 24000 rpm. The
1SCCM, 20SCCM, 30SCCM, 100SCCM, rotor diameter is 206 mm and the axial rotor length is 240 mm.
200SCCM, 2SLM, 500SCCM, 2SLM, 10SLM, Moreover, the rotor contains five blades. This design includes
20SLM 50SLM a group of spiral blades embedded in a cylinder, a pump cas-
ing, and a single cylindrical rotor coaxially arranged within the
the given conditions are used to provide the given throughput pump casing. Each blade comprises three different elements that
Q and outlet pressure Pout . (1) First, Pin is assigned an initial are continuously linked. The blade characteristics comprise a
value based on an estimate, then the inlet influx is determined dynamic-vane element on the inlet side, a spiral-grooved el-
by applying the continuity condition of mass at the boundary. ement on the exhaust side, and a connecting element linked
(2) Next, one time step of the DSMC calculation is performed, the dynamic-vane and the spiral-grooved element continuously
and the number of the incoming and outgoing molecules (denote together. TMP design philosophy is a basis for dynamic-vane
as Nin,i and Nin,o respectively) at the inlet entrance during the design, and viscous flow theory forms the basis of the spiral-
time step are recorded. (3) Finally, the inlet pressure of the next groove design. Each blade characteristic, such as section angle,
time step is calculated using the following relationships. The blade thickness, axial length, and so on, is carefully designed
inlet influx then can be determined. to satisfy the operating principles of molecular flow pumping,
viscous flow pumping, and smooth transition flow pumping. The
Nin,i − Nin,o design aims to optimize performance when gases are pumped
Uin = [5]
Ain n in t along the axial direction. Thus, even when gas flow is disturbed
Sin = Uin Ain [6] by sudden blade shape variation inside the pumping channel,
Q discharge efficiency is not compromised.
Pin = [7] To identify the discrepancies between the ISO and JVIS test-
Sin
ing standards, the turbo booster pump backed by a 1500 l/min
In the above equations, Ain denotes the inlet area, and Uin rep- foreline pump (SD90VIII) is tested in each system, respectively.
resents the free stream velocity outside the inlet port. As illustrated in Figure 6, the pumping speeds vary by 40% in
Steps (2) and (3) are repeated until the inlet pressure stabi- for inlet pressure of 0.01 Pa to 1 Pa. Moreover, when the inlet
lizes. Several methods of improving the convergence exist. In pressure exceeds 1 Pa, the deviation between these two testing
this work steps (2) and (3) are not performed until the initial flow results is below 10%. When inlet pressure is below 10−2 Pa,
is well developed and the incoming and outgoing molecules are since a GP360 ion gauge is installed on the ISO testing appara-
substituted by time average values in Equation (5). Because of tus, the measurement range in the pumping speed can reach the
the limited number of simulated molecules, the resulting Pin 10−3 Pa range. Moreover, when the inlet pressure condition is
of each step fluctuates slightly even when the steady state has below the 10−3 Pa range, an orifice method other than the flow
been reached. This fluctuation is statistical is nature, and should meter method used here should be adopted to measure the inlet
decrease with increasing number of incoming and outgoing sim- condition up to the ultimate pressure regime.

TABLE 4
Classification of Pumping Flow Regimes

Flow Knudsen number Kn Model

Viscous flow Kn < 0.01 Navier-Stokes eqns.
(No-slip BC)
Slip flow 0.01 < Kn < 0.1 Navier-Stokes eqns.
(Slip BC)
Transitional flow 0.1 < Kn < 10 Boltzmann eqn.
Free molecular flow 10 < Kn Collisionless
Boltzmann eqn.
8 R. Y. JOU ET AL.

FIGURE 5
Design and simulation procedures for the turbo booster pump development by using CFD and DSMC methodologies.
 PUMPING SPEED MEASUREMENT AND ANALYSIS 9

FIGURE 6
Schematic of the pumping speed testing by the flow meter method.

The pumping speed test involves testing the turbo booster pump used. The maximum pumping speed of this turbo booster
pump backed by different foreline pumps or combinations of pump occurs at the 1000 l/s level. From the test results, the ex-
backing pump and foreline pump. Notably, Figure 6 displays haust pressure can reach as high as 1000 Pa. This high exhaust
the schematic layout of the testing using the flow meter method. pressure characteristic is very important because it allows the
Moreover, Figure 7 plots the pumping speed curves tested us- use of a mechanical booster pump to be omitted, thus enabling
ing SD90VIII and SD120H as the foreline pumps. The pumping cheaper, more compact and more reliable systems.
speed specification of SD90VIII is 1500 l/min, while of SD120H To demonstrate the improvement of pumping performance
is 2000 l/min. As revealed in Figure 7, when inlet pressure ex- in the transitional flow of the designed turbo booster pump, a
ceeds 1 Pa, the testing data of these two system connections commercialized compound-type turbo pump (Osaka TG1133M
varies with pumping speed. This phenomenon reveals the ef- with magnetically levitated bearing, as shown in Table 1) is
fect of foreline pump size. Below this pressure a plateau region tested in the ISO system. Figure 7 displays the test results. This
exists for pumping speed, which is independent of the foreline design combines six stages of TMP blades in the inlet side with
10 R. Y. JOU ET AL.

FIGURE 7
Pumping speeds testing of the turbo booster pump backed by a single SD90VIII or SD120H foreline pump.

an eight groove MDP section in the outlet side to enhance its design for vacuum pumping. The proposed turbo booster pump
compression ability under a viscous flow regime. The measured has a TMP element (which functions as a rotor) on its inlet
ultimate pressure is around 10−7 Pa. When inlet pressure ex- side, but lacks a stator, meaning that the pumping speed in this
ceeds 32 Pa, the compression capability of the MDP section is flow regime is less than that of the pump it is compared with.
lower than the turbo booster pump, and its foreline pressure is On the other hand, if inlet pressure exceeds 5 Pa, continuum
only 500 Pa, below the 1300 Pa of the turbo booster pump. This viscous flow dominates the vacuum pumping. A spiral-grooved
compound-type turbo pump requires a more powerful foreline rotor can maintain the pressure difference between inlet and
pump to back its pumping. Furthermore, when the inlet pressure exhaust. The turbo booster pump design presented in this study
is below 5 Pa, since the flow regime is a free molecular flow has a much larger cross section than the compared pump, as well
regime, the traditional TMP mechanism (see Table 1) with mul- as a faster pumping speed. Table 5 compares the performance
tiple interlaced stages of rotor and stator is the most effective of the turbo booster pump, pump A (Osaka TG1133M), and

TABLE 5
Performance Comparisons of Turbo Vacuum Pumps with Different Rotor Designs

Items Turbo booster pump Pump A Pump B

Size (mm) φ285 × 400 φ300 × 360 φ253 × 319
Inlet flange (mm) NW200 DN200CF ICF253
Rotation speed (rpm) 24000 33600 35000
Pumping speed (67 Pa) 80 l/s 50 l/s 22 l/s
(1.33 Pa) 645 l/s 850 l/s 500 l/s
(0.133 Pa) 860 l/s 1500 l/s 1000 l/s
Max. foreline pressure (Pa) 1300 500 266Pa
Vibration (µm) <0.01 <0.01 <0.01
Cooling water 1∼4 L/min Air Air/Water
Foreline pump SD90VIII SD90VIII ?
Note (Tested by PIDC’s system) (Tested by PIDC’s system) (Data extracted from product’s catalog)
 PUMPING SPEED MEASUREMENT AND ANALYSIS 11

FIGURE 8
Comparisons of experimental measurements and calculations by the CFD and the DSMC methodologies for pumping speeds of
the turbo booster pump backed by the SD120H foreline pump.

pump B (Seiko STP-H1000C). The comparison indicates that the of the analysis Meanwhile, Table 6 lists the relevant comparison
pumping speed of the turbo booster pump is best in the transition data. All analyses are conducted using a commercialized CFD
flow regime. package (UNIC) and a self-developed DSMC code based upon
Figure 8 compares the experimental data with the analysis re- Bird’s (1994) method. Moreover, calculations are based on the
sults of CFD and DSMC methodologies to verify the accuracy SD120H foreline pump specifications. In the CFD study, the
TABLE 6
Listing of Computational Results and Experimental Data for Pumping Speeds of a Turbo Booster Pump

Inlet pressure (Pa) Outlet pressure (Pa) Computed Experimental Relative error
Scheme (Computed) (Fixed conditions) results (l/sec) data (l/sec) %
DSMC 0.03 1.87 982.2 1114.8 11.89
0.06 3.60 943.26 1070.6 11.89
0.09 5.07 908.31 1049.1 13.42
0.13 6.27 878.38 1033.2 14.98
0.16 7.46 843.8 1021 17.36
CFD 0.65 19.06 729.01 873.5 16.54
1.63 33.33 615.27 734.1 16.19
4.47 55.85 435.66 356.9 -22.07
10.73 80.38 283.52 223 -27.14
56.87 164.49 119.03 99.2 -19.99
115.16 244.74 89.72 75.7 -18.52
557.05 800.60 59.29 48.4 -22.50
12 R. Y. JOU ET AL.

rotor channel space is divided into 84585 body-fitted meshes. represents a useful and accurate design tools for prototype design
Calculations are based on a fixed outlet pressure, obtained from and performance assessment for new varieties of turbo pump.
the mass flow rate and the speed of the foreline pump, assum-
ing that foreline conductance can be ignored. Table 6 lists the CONCLUSION
pumping speeds calculated for this rotor under different outlet This investigation measured the pumping speed for the spiral-
conditions. The modeled gas is nitrogen. From the CFD com- grooved turbo booster pump with well-designed blade geometry
putational results, the simulated inlet pressures can be down to and simulates pumping performance using both the CFD and
0.65 Pa with reasonable accuracy, it is belonging to the slip flow DSMC approaches. The operating pressure range of the pump
regime. Furthermore, when the inlet pressure is further reduced, designed in this study extends from 10−5 Pa on the inlet side
the continuum assumption inherent to the fluid dynamics is not to several hundred Pascals on the exhaust side. Moreover, this
exactly satisfied owing to the more rarefied gas density, and thus study measured the performance of the pump presented here,
the DSMC simulation approach should be adopted. and compared it to that of other varieties of turbo pumps. The
In the DSMC simulations, the number of simulated molecules testing results reveal that this design has a higher foreline capa-
ranges between 600 thousands and 4 million, and the number bility up to 1000 Pa range than other pumps it was compared
of cells ranges between 6,000 and 8,000. After performing the with. Consequently, the pump examined here can omit the need
DSMC calculations, a three-dimensional steady-state flow field for a mechanical booster pump, thus simplifying the pumping
can be observed. Table 6 evaluates five speeds, under inlet pres- system and reducing costs. Since the designed rotor satisfies the
sures of 0.03 Pa, 0.06 Pa, 0.09 Pa, 0.13 Pa, and 0.16 Pa. Notably, pumping characteristics under different pressure ranges, it can
the relative errors compared to experimental testing are below be successfully operated at the molecular flow, transitional flow,
17%. and viscous flow regimes, respectively. Furthermore, this study
Figure 8 shows a pumping speed curve fitted by the calcu- also combines CFD and DSMC methodologies for simulating
lated speeds. This speed curve can then be compared with the the speeds of the turbo booster pump. The calculation results
experimental testing results to verify the accuracy of the anal- show that the proposed design procedure is a successful and ef-
yses. As Figure 8 shows, the CFD computations predict the fective tool in the turbo booster pump’s design. Unfortunately,
pumping speed with an error range from 16.19% to 27.14%. because of the inherent complexity of flow field analysis of the
As illustrated in Figure 8, for CFD computations, the relative turbo booster vacuum pump and the fundamental limitations of
error to experiment increases with Knudsen number. Computa- the CFD and DSMC approaches, both of these computational
tions failed when Kn > 0.1 range. That is, it exceeds the 25% methodologies are inadequate for providing an accurate pump-
criterion limit. When inlet pressure is below 0.65 Pa, inappro- ing speed trend in the 0.001 < Pin < 0.01 Pa range, as shown in
priate use of the CFD method to analyze the rarified gas flow Figure 8. Therefore, the experimental approach employed here
causes a significant increase in the number of computational is the only way of accurately and effectively exploring pump
errors. In this case, the DSMC simulation approach based on performance in the 0.001 < Pin < 0.01 Pa range. This investi-
molecular kinetic theory must be used to predict the pumping gation aims to demonstrate the effectiveness of the experimen-
speed in the molecular flow regime. Analyzing a flow in a transi- tal approach for investigating the pumping speed of the turbo
tional or viscous regime using the DSMC approach requires far booster pump, and the failure of using numerical methods for
greater computational resources and time than using the CFD turbo booster pump analyses in this pressure range. Until now,
approach. Consequently, a compromise between the CFD and the literature has neglected this subject. The methods developed
DSMC approaches represents the best method of achieving an here can be applied in the future to design new compound or
efficient analysis. When inlet pressure is between 10−5 Pa and hybrid turbo pumps.
1 Pa, the DSMC approach offers the best method of analyzing
the problem, and the analytical results reveal an error ranging be-
tween 11.89% and 17.36%. For DSMC simulations, decreasing NOMENCLATURE
Knudsen number the relative error to experiment is increased. Ain inlet area
Simulations failed when Kn < 0.7. That is, it exceeds 17% crite- C orifice conductance
rion limit. Interestingly, the pumping speed values predicted by D diameter of test dome
DSMC simulation are always smaller than the testing data. One D0 diameter of rotor
possible explanation for this phenomenon is that the calculated d diameter of the orifice
pressures in the analyzing model occur in slightly different loca- h in depth of groove channel at inlet side
tions to the measured pressures. When the Knudsen number in h out depth of groove channel at outlet side
the range of 0.5 < Kn < 0.1, neither the CFD computations nor L length of rotor
the DSMC simulation are suitable for analyzing the pumping Lo thickness of the orifice plate
speed of the turbo booster pump. M molecular weight
Finally, from the numerical results displayed in Figure 8, the n in number density at the inlet side
design methodology combining the CFD and DSMC approaches Nin,i , Nin,o number of the incoming and outgoing molecules
 PUMPING SPEED MEASUREMENT AND ANALYSIS 13


Nin,i inlet influx Gaede, W. 1913. Die Molekularluftpumpe. Ann. Phys. 41:337.
P0 , P1 , P2 ultimate pressure of the test dome, pressure at up- Hablanian, M. H. 1994. Comments on the history of high vacuum tur-
per chamber and pressure at lower chamber bopumps, in Vacuum Science and Technology—Pioneers of the 20th
Q throughput Century, AIP Press.
R universal gas constant Heldner, M., and Kabelitz, H. P. 1990. Reliability of turbomolecu-
lar vacuum pumps: A comparison of rolling element and magnetic
S volume flow rate (pumping speed)
bearing systems. J. Vac. Sci. Technol. A8(3):2772.
T temperature Henning, J. 1988. Thirty years of turbomolecuar pumps: A review and
t time step recent developments. J. Vac. Sci. Technol. A6(3):1196.
Uin free stream velocity outside the inlet port Heo, J. K., and Hwang, Y. K. 2000. DSMC calculations of blade rows of
X, Y, Z coordinates of inertial frame a turbomolecular pump in the molecular and transition flow regions.
x, y, z coordinates of rotating frame Vacuum 56:133–142.
Holweck, F. 1923. Pompe moléculaire hélicoidale. C.R. Acad. Sci.
177:43.
Greek Symbols Jou, R. Y., Cheng, H. P., Chang, Y. W., Chen, F. Z., and Iwane, M. 2000.
δ clearance between rotor and casing Designs, analyses, and tests of a spiral-grooved turbo booster pump.
ρ gas density J. Vac. Sci. Technol. A18(3):1016–1024.
φin blade angle at inlet side Lee, Y. K., and Lee, J. W. 1996. Direct simulation of compression
φout blade angle at outlet side characteristics for a simple drag pump model. Vacuum 47(6–8):807–
 rotational speed 809.
Levi, G. 1992. Combination of turbomolecular pumping stages and
molecular drag stages. J. Vac. Sci. Technol. A10(4):2619.
Maurice, L. 1974. A new molecular pump. Jpn. J. Appl. Phys. Suppl.
REFERENCES 2:21.
Becker, W. 1961. Zur Theorie der Turbomolekularpumpe. Vakuum Nanbu, K., Kubota, H., Igarashi, S., Urano, C., and Enosawa, H.
Techn. 10:199. 1991. Performance of spiral grooves on a rotor of turbomolecular
Bird, G. A. 1994. Molecular Gas Dynamics and the Direct Simulation pump. Transactions of Japan Society of Mechanical Engineers B
of Gas Flows. Clarendon Press, Oxford. 57(533):172–177.
Chang, Y. W., and Jou, R. Y. 2001. Direct simulation of pumping char- Sawada, T., and Nakamura, M. 1990. Spiral grooved visco-vacuum
acteristics in a full 3D model of a single stage TMP. Applied Surface pumps with various groove shapes. Vacuum 41(7–9):1833.
Science 169–170:772–776. Sawada, T., Nakamura, M., and Abumiya, A. 1992. Spiral grooved
Chang, Y. W., Jou, R. Y., Cheng, H. P., and Chen, F. Z. 2001. Pumping vacuum pump working in high pressure ranges. Vacuum 43(11):
performance analyses of a turbo booster pump by direct simulation 1097.
Monte Carlo method. Vacuum 60:291–298. Sawada, T. 1993. Improvement in the performance of spiral grooved
Cheng, H. P., Jou, R. Y., Chen, F. Z., Chang, Y. W., Iwane, M., and visco-vacuum pumps in coarse vacuum region. Vacuum 44(5–7):689.
Hanaoka, T. 2000. Three-dimensional flow analysis of spiral-grooved Schneider, T. N., Katsimichas, S., de Oliveira, C. R. E., and Goddard,
turbo booster pump in slip and continuum flow. J. Vac. Sci. Technol. A. J. H. 1998. Empirical and numerical calculations in two dimen-
A18(2):543–551. sions for predicting the performance of a single stage turbomolecular
Duval, P., Raynaud, A., and Saulgeot, C. 1988. Molecular drag pump: pump. J. Vac. Sci. Technol. A16(1):175–180.
Principle, characteristics, and applications. J. Vac. Sci. Technol. Siegbahn, M. 1943. A new design for a high vacuum pump. Ark. Math.
A6(3):1187. Astron. Fys. 30B:261.
 International Journal of

Rotating
Machinery

International Journal of
The Scientific
Engineering Distributed
Journal of
Journal of

Hindawi Publishing Corporation

World Journal
Hindawi Publishing Corporation Hindawi Publishing Corporation
Sensors
Hindawi Publishing Corporation
Sensor Networks
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

Journal of

Control Science
and Engineering

Advances in
Civil Engineering
Hindawi Publishing Corporation Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

Submit your manuscripts at

http://www.hindawi.com

Journal of
Journal of Electrical and Computer
Robotics
Hindawi Publishing Corporation
Engineering
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

VLSI Design
Advances in
OptoElectronics
International Journal of

International Journal of
Modelling &
Simulation
Aerospace
Hindawi Publishing Corporation Volume 2014
Navigation and
Observation
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
in Engineering
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014
Engineering
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2010
Hindawi Publishing Corporation
http://www.hindawi.com
http://www.hindawi.com Volume 2014

International Journal of
International Journal of Antennas and Active and Passive Advances in
Chemical Engineering Propagation Electronic Components Shock and Vibration Acoustics and Vibration
Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014