AIAA JOURNAL

Vol. 30, No. 10, October 1992

Experimental and Numerical Investigations of Low-Density

Nozzle and Plume Flows of Nitrogen
lain D. Boyd*
Eloret Institute, Palo Alto, California 94303
Paul F. Penkot
NASA Lewis Research Center, Cleveland, Ohio 44135
and
Dana L. MeissnerJ and Kenneth J. DeWitt
University of Toledo, Toledo, Ohio 43606
Downloaded by TECHNION - ISRAEL INST OF TECH on June 13, 2015 | http://arc.aiaa.org | DOI: 10.2514/3.11247

Experimental and numerical investigations are performed and compared for the flow of nitrogen in a small
nozzle and in the near field of the plume resulting from expansion into near-vacuum conditions. The experimen-
tal data obtained were in the form of pressure measurements using a pitot tube, in the nozzle-exit plane and near
field of the plume. Since the flow regimes vary from continuum, at the nozzle throat, to rarefied, in the plume,
two different numerical studies are undertaken: the first employs a continuum approach in solving the
Navier-Stokes equations, and the second employs a stochastic particle approach through the use of the direct
simulation Monte Carlo (DSMC) method. Comparison of the experimental data and the numerical results at the
nozzle exit reveals that the DSMC technique provides the more accurate description of the expanding flow. It
is discovered that the DSMC solutions are quite sensitive to the model employed to simulate the interaction
between the gas and the nozzle-wall surface. It is concluded that the simple fully diffuse model is quite
satisfactory for the present application. The study provides the strongest evidence to date that the DSMC
technique predicts accurately the flow characteristics of low-density expanding flows.

Nomenclature thrust level of these engines is usually quite small, and it was
De = diameter at nozzle exit thought previously that they presented few problems of inte-
Dp = diameter of probe gration with the spacecraft. However, as discussed by Dettleff1
Dt = diameter at nozzle throat in his excellent review, such rockets may cause a number of
Mx - Mach number ahead of a shock deleterious effects which reduce significantly the operational
Px = static pressure ahead of a shock lifetime of a spacecraft. Although the firing times of the
P0 = total pressure thrusters are usually only a few seconds, it must be remem-
Poy = total pressure behind a shock bered that they are fired repeatedly for a number of years.
Px = static pressure ahead of a shock This may lead to the gradual accumulation of contamination
Rep - probe Reynolds number from the plume on sensitive surfaces such as solar arrays. In
T0 - total temperature addition, the plume can cause heating or electrical charging of
rref = reference temperature the spacecraft, which can alter the thermal balance and dam-
Tw = temperature of nozzle wall age scientific instruments. Thrust loss and disturbance torques
Tx = static temperature ahead of a shock are also unwanted effects which can result from the firing of
Ty = static temperature behind a shock the control thrusters. The magnitude of these effects is very
U0 = thermal velocity at total temperature much dependent on both the thruster design and the satellite
t/oo = freestream velocity configuration.
7 = ratio of specific heats The assessment of the interaction between the spacecraft
jLtref = viscosity given by reference temperature and the plume requires an accurate description of the expand-
Hy = viscosity given by temperature behind a shock ing flowfield. Therefore, a research program has been estab-
a? = viscosity temperature exponent lished with the objective of advancing the theoretical predic-
Poo = freestream density tions of small rocket expansions through comparison with
experimental data. The major phenomena to be investigated
include the flowfield of low-Reynolds-number nozzles, the
Introduction resultant plume flowfields and impacts, and the ambient envi-
ronment effects which arise during testing in ground-based
F OR the control of satellites and large space structures in
orbit, a number of propulsive devices are often used. The facilities. The approach adopted is to develop and compare
various theoretical and numerical techniques used in predic-
Received Sept. 16, 1991; revision received Jan. 28, 1992; accepted tions and to obtain accurate experimental data for code devel-
for publication Feb. 6, 1992. Copyright 1992 by the American opment and verification. Although the program encompasses
Institute of Aeronautics and Astronautics, Inc. No copyright is as- all low-thrust propulsion concepts, including the neutral flows
serted in the United States under Title 17, U.S. Code. The U.S. of resistojets, the plasma flows of arcjet thrusters, and small
Government has a royalty-free license to exercise all rights under the chemical rockets, the initial emphasis is directed toward an
copyright claimed herein for Governmental purposes. All other rights
are reserved by the copyright owner.
understanding of the nozzle and plume flows of resistojets.
* Research Scientist, mailing address: NASA Ames Research Cen- The purpose of the present study is to compare recently
ter, MS 230-2, Moffett Field, CA 94035 acquired experimental data with predictions obtained using
tAerospace Engineer. Member AIAA. numerical methods. The region of study is the nozzle flow,
jGraduate Student, Department of Mechanical Engineering. and the near-field expansion of the plume. The gas employed
Professor, Department of Chemical Engineering. Member AIAA. is nitrogen due to its simple diatomic structure. The aim is to
2453
 2454 BO YD ET AL.: LOW-DENSITY PLUME FLOWS OF NITROGEN

understand the fluid mechanics without complication from

other physical processes such as vibrational and chemical
nonequilibrium. In this article, description is made of the
experimental facility and the type of measurements which
have been undertaken at the exit of the nozzle and in the
plume. The application of two different numerical techniques
is also described. One method provides a continuum analysis
from solution of the Navier-Stokes equations. Because of the
rarefied nature of the expanding flow, the second method uses
a numerical technique that simulates particle behavior at the
molecular level, specifically, the direct simulation Monte
Carlo (DSMC) method. Although the DSMC method is more
numerically intensive than the continuum technique, it is bet-
ter able to capture a number of potentially important nonequi-
librium effects. Previously, it has been found by Boyd and Fig. 1 Nozzle diagram.
Stark2 that the DSMC method offers results which yield better
correspondence to experimental data in the rarefied flow
regime of an expansion plume. In addition, the improper use
of continuum methods in the simulation of such plumes can
Downloaded by TECHNION - ISRAEL INST OF TECH on June 13, 2015 | http://arc.aiaa.org | DOI: 10.2514/3.11247

lead to significant underestimation of important impingement

characteristics such as heat transfer and drag force.3 It is
therefore an aim of this study to assess any deficiencies in the
Navier-Stokes solutions. Particular attention is given to the
physical and computational limitations of these methods I F'ow ^-H
[Controller />>"
based on the assumptions inherent to their formulation. This
investigation provides the opportunity for assessment of the ^4 Accumulator }J [ |
DSMC technique in modeling an expanding gas. In addition, DC Power
Supply
it will be shown that a portion of the flowfield is particularly
sensitive to the model assumptions employed for the interac-
tion between the gas and the surface of the nozzle wall. The
coupling of experimental measurements to the computations
offers a unique opportunity to investigate the validity of these
models.
Fig. 2 Schematic of experimental apparatus and test section.
Reference Problem
For this study, a reference problem was defined that typi- jet.4 The nozzle was made sufficiently large to minimize probe
fied the nozzle flow of an electric propulsion device and also size relative to the exit diameter and does not duplicate neces-
matched the flow capacity of the vacuum pumps on the exper- sarily in size actual thruster hardware. As previously men-
imental facility. A conical nozzle was selected for both analyt- tioned, nitrogen was selected as the test gas for analytical
ical and machining simplicity, but also represents the geome- convenience and does not represent particularly an actual fluid
try of a typical small-thruster nozzle. The nozzle for the of spacecraft propulsion system.
problem is illustrated in Fig. 1, with details given in Table 1.
The nozzle flow conditions are listed in Table 2 for two Experimental Investigation
configurations that correspond to two different experimental Previous experimental investigations of expanding flow
apparatus and two slightly different operating conditions. from small rockets include those performed by Bailey5 with
The nominal Reynolds number of 850, based on throat carbon dioxide, and by Legge and Dettleff5 with hydrazine.
diameter and the viscosity of the gas at the inlet stagnation The test gases and flow conditions examined in Refs. 5 and 6
temperature, is characteristic of a high-performance resisto- made detailed numerical analysis difficult either in terms of
physical modeling or computational expense. The present in-
Table 1 Nozzle geometry vestigation is unique in that the experiment was designed
Inlet diameter Dt 22.1 mm partly on the constraints imposed by the numerical modeling
Throat diameter Dt 3.18 mm techniques employed in the study. The specific intent was to
Exit diameter De 31.8 mm provide a direct correlation between the model predictions and
Longitudinal radius RL 3.18 mm experimental data.
Wall thickness tw 1.65 mm
Lip thickness ti 0.25 mm Test Facility
Inlet half-angle 0/ 45 deg
Exit half-angle 0/ 20 deg The experimental tests for this study were conducted in a
Area ratio (De/Dt)2 100 space-simulation vacuum facility. The experimental apparatus
was mounted in a 0.9-m-diam by 1.6-m-long test section that
was attached to a tank 4.9 m in diameter and 19 m in length as
Table 2 Nozzle flow conditions for the two illustrated in Fig. 2. The vacuum system for the facility con-
_____experimental configurations_____
sisted of 20 oil-diffusion pumps in series with four blowers
Configuration and four roughing pumps.7
1 2 The facility pumping system maintained a vacuum pressure
in the test section on the order of 10 ~ 2 Pa for the nozzle
Total pressure P0 6400 Pa 6210 Pa experiments. The vacuum pressure was monitored with a hot-
Total temperature T0 699 K 700 K cathode ionization gauge that was mounted on the test section
Mass flow m 6.8 X 10 - 5 kg/s 6.8 x 10 ~ 5 kg/s
Reynolds number Rea 850 850 and connected to a digital readout.
Wall temperature Tw\ 551 K 507 K
Experimental Apparatus
Wall temperature TW2 539 K 498 K
a
The experimental apparatus for producing the flowfields
Re = 4m/fi0Dt, where \n0 is the gas viscosity at T0. consisted of a heat exchanger and nozzle (as previously de-
 BO YD ET AL.: LOW-DENSITY PLUME FLOWS OF NITROGEN 2455

scribed) that simulated a thruster as shown schematically in and rotational movements. The range of travel was 24 mm in
Fig. 3. Two heat-exchanger assemblies of different design the radial direction (R), 36 mm in the axial direction (Z), and
were used in the tests and are illustrated in Figs. 3a and 3b. A 360 deg in rotation (0). The tables were positioned manually
traversing mechanism was used to survey the plume with a through mechanical links between the rotary handles on each
pressure probe. table and handles outside the vacuum chamber. Position of
In the design illustrated in Fig. 3a, and denoted as configu- traverse was monitored by linear-variable-differential trans-
ration 1, the nitrogen was heated in a 3.2-mm-diam tube that formers (LVDTs) mounted on each of the tables. Output
was coiled around a cartridge heater of diameter 15.9 mm. In signals from the LVDTs were read on a digital voltmeter.
the design illustrated in Fig. 3b, and denoted as configuration A pitot tube of 1.02 mm o.d. and 0.15-mm wall thickness
2, the nitrogen was heated in an annular area comprised of a was used to measure pressure and flow angle in the nozzle
12.7-mm-diam cartridge heater contained in a tube of 17.3 plume. The probe had a 30-deg chamfer on the internal diame-
mm i.d. The heater for configuration 2 had a lower power ter at the tip to form a sharp lip at the leading edge. The pitot
density and lower surface temperature than that of configura- tube was attached directly to a capacitance manometer having
tion 1, but the heat exchangers produced the same gas temper- a range of 0-1.33 kPa and a listed manufacturer's accuracy of
ature. The design of configuration 2 produced nozzle-wall 0.1% of full scale to measure the pressure. The manometer
temperatures about 40 K lower than configuration 1 as listed was mounted directly on the rotary table. The pitot tube was
in Table 2. The heat exchanger on the apparatus was changed bent in a U-shape so that the tip was located on the axis of the
from the design of configuration 1 to the design of configura- rotary table as shown schematically in Fig. 3b.
tion 2 to lengthen the operational life of the heating element. Temperatures of the nozzle wall were measured by two
Downloaded by TECHNION - ISRAEL INST OF TECH on June 13, 2015 | http://arc.aiaa.org | DOI: 10.2514/3.11247

For both configurations, the pressure and temperature of chromel-alumel thermocouples that were tack-welded to the
the flow were measured upstream of the nozzle in a 22.1-mm- outer wall surface. As shown in Fig. 1, a thermocouple was
diam plenum. The measurements were effectively nozzle-inlet located about midway between the nozzle throat and exit
stagnation conditions as the ratio of area at the measurement plane r wl , and another at the exit plane Tw2.
station to the throat area was about 48:1. The pressure was
measured with a capacitance-manometer transducer having a Test Procedure
range of 0-13.3 kPa. Temperature was measured with a half The test section containing the experimental apparatus was
shielded, chromel-alumel thermocouple where the hot junc- first evacuated without gas flow to establish zero settings for
tion was located at the centerline of the measurement station. the capacitance manometers. The vacuum pressure without
The thermocouple was connected to a digital voltmeter with flow was about 10 ~ 4 Pa which served as the zero-reference
self-contained, cold-junction compensation. The flow of ni- pressure. The flowmeter was zeroed while containing nitrogen
trogen was measured with a transducer of the type that relates at the accumulator supply pressure. After the instruments
thermal changes in a capillary tube to volumnetric flowrate. A were zeroed, flow was established in the apparatus and main-
schematic of the flow system is shown in Fig. 2. tained by the flow controller. Simultaneously, 70 V at about
Although the nozzles for each configuration were machined 0.9 A was applied to the cartridge heater and time allowed for
from the same specifications, the plenum pressures differed the system to equilibrate at a nozzle-inlet temperature of 700
slightly for equivalent flowrates and gas temperatures (see K.
Table 2). For a flowrate of 6.8 x 10 ~ 5 kg/s and a gas temper- After the system reached steady state, pressure scans were
ature of 700 K, configuration 1 had a plenum pressure of 6400 made by moving the pitot probe to a given location and then
Pa which was slightly higher than the 6210 Pa of configuration rotating the probe to determine the point of maximum pres-
2. The difference in plenum pressures implied a difference in sure. Following Bailey,5 the maximum pressure was taken as
the throat diameters of about 1 % with the throat diameter of the pitot pressure reading for the particular location. The
configuration 2 being the larger of the two. angle 6 through which the probe was rotated to obtain the
The experimental apparatus was designed specifically for maximum pressure was taken as the flow-angle reading. A
making measurements in an expanding flow by use of a typical rotary scan, in this case at the exit plane (Z/De = Q)
traversing mechanism. The mechanism consisted of a rotary- and a radial position, R/De = 0.38, is shown in Fig. 4. The
traverse table mounted atop two linear-traverse tables, with flow angles were measured with respect to the nozzle axis. The
travel axes mutually perpendicular to provide radial, axial, error in the pressure measurement was estimated to be 5 Pa,
and in the flow angle, 2 deg.

Cartridge Heatei
Radiation TQ
Shielding (Shielded
Thermocouple)

Fig. 3a Hardware schematic of configuration 1.

1.02 mm O.D.
with 30 Chamfer"^

Cartridge Heater

Probe Angle (deg.)

Fig. 4 Measured pitot pressure vs probe angle at Z = 0 mm and

Fig. 3b Hardware schematic of configuration 2. R = 12 mm.
 2456 BOYD ET AL.: LOW-DENSITY PLUME FLOWS OF NITROGEN

Comparisons of vertical and horizontal scans of pressure forms of the constitutive relations for viscous stress and heat
across the nozzle diameter indicated that the flow was sym- transfer. This problem has received detailed investigation in a
metrical within experimental error. All reported pressure scans study relating to the problem of the normal shock wave.12 It
were made in the horizontal plane starting at the nozzle center- has been shown that solutions of shock waves obtained with
line (R/De =0) and extending outward radially. the Navier-Stokes equations give much poorer correspondence
to experimental data than computations performed with the
Numerical Investigations DSMC technique, or with the Burnett equations (a higher-or-
Previous numerical investigations of expanding flowfields der extension of the Navier-Stokes equations). Although Ref.
under rarefied conditions include those of Boyd and Stark,2'3 3 makes direct comparison between DSMC results and a solu-
Campbell,8 and Nelson and Doo.9 Each of the studies em- tion of the Euler equations (using the method of characteris-
ployed the DSMC technique, and Refs. 2, 3, and 9 also consid- tics) for an expanding gas under rarefied flow conditions, no
ered continuum methods, but all were unsatisfactory in differ- such study has been performed previously to investigate the
ent ways. For the different cases, the high computational time effect of the failure of the Navier-Stokes equations. This was
requirements of DSMC resulted in the analysis of only a small one of the primary aims of the present study.
part of the expansion, in the analysis of an unrealistic thruster
configuration, or in the use of numerical parameters that DSMC Computations
called into question the validity of the computations. As will The DSMC method13 is now a familiar tool employed in the
be discussed in the following sections, the present investiga- numerical modeling of flows involving rarefied gases. It has
Downloaded by TECHNION - ISRAEL INST OF TECH on June 13, 2015 | http://arc.aiaa.org | DOI: 10.2514/3.11247

tion has the advantage of using a numerically efficient DSMC been applied to expanding flows in a number of studies,
code which permits the computation of a real nozzle and although almost none of these have made direct comparison
plume flow without violating numerical restrictions. Another with experimental data. The present study therefore makes an
important aspect of the study is the availability of a contin- important contribution in that numerical and experimental
uum code which is discussed next. studies of an expanding flow have been performed simulta-
neously. The DSMC code employed in the present work is
Continuum Computations effectively structured for optimum use on vector supercom-
The continuum computation of the flow was made with the puters, and has been described in detail previously.14 The flow
numerical code termed the viscous nozzle analysis program is modeled axisymmetrically in two dimensions using the pro-
(VNAP2) that was originally developed for laminar, viscous cedures described in Chapter 9 of Ref. 13.
nozzle flow.10 It was later modified to allow computation of The code has been configured specifically to compute the
both internal and external flows, and, with the inclusion of flow in the nozzle, in the plume forward of the nozzle, and in
turbulence models, validated at higher Reynolds number.11 a small portion of the backflow. This region has been designed
The code solves the Navier-Stokes equations for a compress- to include a larger area than that investigated experimentally,
ible fluid in time-dependent, nonconservative form by the and to ensure that all flow contributing to that region has been
explicit, two-step MacCormack method which gives second- included. It is not, however, the purpose of this study to
order accuracy in space. With this scheme, the equations are provide a definitive computation of the backflow region. This
marched in space and iterated in time from a specified initial would be performed in a more numerically efficient way by
flowfield to a steady-state solution. For this problem, given starting a fresh computation at the nozzle-exit plane, using the
the relatively low Reynolds number, laminar flow was as- macroscopic results from the present study. This is proposed
sumed. as a topic worthy of future investigation. The computational
The code produces a body-fitted, nonorthogonal grid in domain extends to an axial distance of about 40 mm from the
physical coordinates that, by transformation, is rectangular in nozzle-exit plane, and to a radial distance of 50 mm. A com-
computational space. For the problem in this study, 31 radial putational grid of 760 x 50 nonuniform cells was required to
and 51 axial grid lines were specified. Furthermore, the grid meet the restriction of a cell length of one local mean-free path
was clustered in the nozzle throat and near the nozzle wall to in the flow direction. The DSMC computations are started at
capture adequately the steep gradients of the flow variables in a point just after the nozzle throat using the input of macro-
those regions. scopic properties from the continuum solution. Several differ-
The computation was started in the upstream, subsonic ent gas/surface interaction models have been considered in-
portion of the nozzle. The inflow conditions included specifi- cluding specular reflection, diffuse reflection, varying
cation of total pressure and temperature, and zero radial fractions of these two extremes, and the Cercignani-Lampis
velocity. The subsonic inflow boundary employed the method model implemented in DSMC by Lord.15 This model offers
of characteristics such that the axial velocity was determined separate accommodation coefficients for the reflection of the
from interior, downstream points, and evolved with the solu- normal and tangential velocity components, and would there-
tion from an initial profile. fore appear to offer greater versatility than combinations of
The stream pressure at the outflow surface in the subsonic the diffuse and specular models. For nitrogen, the rotational
portion of the flow was extrapolated rather than specified as is energy exchange model of Boyd16 was employed which in
normally done with the code. This scheme gave relatively essence varies the rotational collision number with tempera-
smooth profiles of properties in the vicinity of the wall at the ture. The exchange of vibrational energy is assumed frozen
exit plane, but did, in effect, impose an ambient pressure on due to the low temperatures encountered in the flowfield.
the flow. In cases where the subsonic exit-plane pressure was The limitations of the DSMC technique applied under the
explicitly fixed, the computed properties exhibited a disconti- present conditions are more of a computational nature, and
nuity across the sonic line where the flow decelerates from quite different from the physical difficulties experienced by
supersonic to subsonic in the plane normal to the wall. the Navier-Stokes equations. The numerical cost of DSMC
The code was iterated for 20,000 timesteps from an initial computations is linearly proportional to the density of the
condition corresponding to one-dimensional, isentropic flow. flow. This is because the size of a computational cell should be
Convergence of the scheme was not monitored as such, but scaled with the mean-free path, which is, of course, itself
conservation of mass was checked and used as a guide. At scaled inversely with the density. Therefore, a high-density
20,000 timesteps the mass flow was fairly consistent through- flowfield requires a large number of computational cells. This
out the nozzle. In the diverging section, it was within 1% of means that more particles must be simulated, and conse-
the value specified by the inlet conditions, and in the converg- quently more collisions computed. The computations of the
ing section, within 29/o. nozzle flow are quite expensive for the DSMC code because
It is well established that the Navier-Stokes equations break the conditions near the nozzle throat are those of a relatively
down at high Knudsen numbers due to the failure of the linear high-density, continuum regime. This computation represents
 BO YD ET AL.r LOW-DENSITY PLUME FLOWS OF NITROGEN 2457

one of the most numerically intensive simulations undertaken This procedure was performed for each of the two different
with the DSMC technique. Reasonable computational execu- experimental configurations. The DSMC code has a numerical
tion times have been achieved through the vectorization of efficiency of l . l x l O " 6 CPU s/timestep/particle, and the
many parts of the algorithm for efficient performance on a computation required 3 CPU h on the Cray Y/MP. The
Cray-Y/MP supercomputer. execution time for the continuum code was 0.8 CPU h.
The first comparison of the two numerical solutions is made
Calculation of Pitot Pressure in Fig. 5 in which the velocity profiles close to the nozzle
throat are shown. Radial distance R and the velocity U have
To compare the numerical solutions with experimental data, been normalized by the throat diameter Dt and the thermal
pitot pressures were first calculated from the numerical results velocity in the stagnation chamber U0. Excellent agreement is
using the computed state variables. The calculated pitot pres- obtained for the two profiles, verifying that the continuum
sure can be viewed as the pressure that would be measured if and DSMC solutions are in agreement in the continuum-flow
a pitot tube were inserted into the stream computed by the regime near the nozzle throat.
numerical codes. Since the flow was both supersonic and
Significant differences were, however, observed in the two
rarefied, the calculation involved a two-step process: numerical solutions further downstream in the nozzle. In Fig.
1) The pressure ratio across a normal shock was calculated
6, the velocity profiles computed by the two methods are
with the Rayleigh pitot tube equation (c.f. Ref. 17, p. 154)
compared at the nozzle exit plane (the radial distance is here-
represented by the functional relation: after normalized by the nozzle exit diameter De). The velocity
predicted by DSMC is always greater than the continuum
(1)
Downloaded by TECHNION - ISRAEL INST OF TECH on June 13, 2015 | http://arc.aiaa.org | DOI: 10.2514/3.11247

Poy/Px =f(y,M2x) data. It is of particular importance to note that the velocity

given by DSMC at the nozzle wall is nonzero. This indicates
where y is the ratio of specific heats which for nitrogen is 1.4. that the no-slip boundary conditions normally employed in
2) The ideal pressure Poy was corrected for rarefaction ef- continuum computations are insufficient for this application.
fects by The slip effect in rarefied flows is an established phenomenon
(2)
and is important in the present study due to its significant
log (Pom/ Poy} = 0.089 - 0. 120 log (Rep) effect on the boundary-layer thickness. A further comparison
at the nozzle-exit plane is shown in Fig. 7 where temperature
This expression is employed for Rep < 5.6, otherwise

*P om '/P
* oy
-i10
w

where Pom is the corrected total pressure behind the shock.

The correction in Eq. (2) was obtained from a report on the
rarefaction effects of pitot tubes for measuring pressure in
rarefied, supersonic, wind-tunnel flow.18 In Ref. 18, the factor
P0m /Poy ^ 1 was plotted as a function of Rep, with total
temperature of the stream T0 as parameter. Equation (2) is a
fit of an interpolated curve for T0 = 700 K.
In Ref. 18, the probe Reynolds number was defined as

Rep = (3)

In the probe Reynolds number, the actual diameter of the

pitot tube on the experimental apparatus was used for Dp. The
values of px and Ux were obtained from the DSMC results.
The viscosity \iy was derived from the numerical results by
first calculating the temperature behind a shock Ty. This
quantity was computed from the normal-shock equation for U/Uo
the ratio of static temperature (c.f. Ref. 17, p. 118) repre-
sented by the functional relation Fig. 5 Comparison of computed velocity profiles at the inflow sur-
face of the DSMC computational domain.
Ty/Tx=g(7,M2) (4)

where the values of static temperature ahead of the shock Tx

and the Mach number Mx were obtained from the numerical
results. The viscosity was then found from the standard tem-
perature power-law for the viscosity of gases (c.f. Ref. 19, p.
28):

where jLtref was taken20 a s 2 . 5 8 x l O ~ 5 N - s/m2, the viscosity

of nitrogen at the reference temperature Trcf of 500 K, and o>
was 0.75.
Results and Discussion
The continuum code was executed using the input of Tables
1 and 2. The wall temperature distribution was estimated from
the values obtained experimentally with the thermocouples,
and was employed in both the continuum and DSMC compu-
tations. The continuum solution near the nozzle throat was U/Uo
then employed to obtain a solution of the nozzle flow and the Fig. 6 Comparison of computed velocity profiles in the nozzle-exit
near-field expansion region with the vectorized DSMC code. plane for configuration 1.
 2458 BOYD ET AL.: LOW-DENSITY PLUME FLOWS OF NITROGEN

continuum data revealed that the solutions for Mach number

agreed in the near-continuum flow regime near the nozzle
throat. Proceeding further down the nozzle, however, showed
Wall -^7
the Mach numbers computed with DSMC becoming increas-
ingly larger than the corresponding continuum values. Careful
examination of the flow-angle data reveals that the flow near
the lip in the nozzle-exit plane is turned up towards the wall.
In this region, the flow-angle becomes greater than the nozzle
half-angle of 20 deg. This behavior was also observed in the
study by Campbell.8 As the gas expands further out away
from the nozzle, the flow-angle data begins to resemble the
Continuum flow from a point source.
DSMC Translational In Fig. 10, the pitot pressure profiles in the nozzle-exit plane
- DSMC Rotational
are shown for configuration 1 (pitot pressure, Pom, has been
normalized by the stagnation pressure Po). Four separate
profiles are shown: the experimental data, the continuum
solution, and two DSMC solutions employing two different
0 0.2 0.4 0.6 0.8 1 gas/surface interaction models. The DSMC computation
T/To which employs fully diffuse reflection at the wall gives the best
Downloaded by TECHNION - ISRAEL INST OF TECH on June 13, 2015 | http://arc.aiaa.org | DOI: 10.2514/3.11247

Fig. 1 Comparison of computed temperature profiles in the nozzle- correspondence to the experimental data. The continuum so-
exit plane for configuration 1. lution predicts higher pitot pressures in the isentropic core of
the expansion, and then lower pressures in the thick, laminar
boundary layer. The DSMC solution employing specular wall
400 reflection essentially provides a nonviscous Euler solution
with no boundary layer. The sensitivity of the exit-plane pro-
file to the surface model is readily seen by comparison of the

Fig. 8 Comparison of temperature contours computed with DSMC

for the translational (lower) and rotational (upper) energy modes.

profiles are shown. Both translational and rotational tempera-

tures are obtained from DSMC, whereas the continuum
method, predicated on thermal equilibrium, gives just one Fig. 9 Mach number (lower) and flow angle (upper) contours com-
temperature. All temperatures are normalized to the stagna- puted with DSMC.
tion value To. The continuum temperature is always higher
than both the DSMC temperatures. These latter values show
only a small degree of nonequilibrium in the low-density re-
gion close to the nozzle wall.
The nonequilibrium between the modes of translational and
rotational energy is more accentuated as the plume expands
away from the nozzle. This phenomenon is observed clearly in
Fig. 8 where the translational temperature, shown in the lower
portion of the figure, is compared with the rotational temper-
ature, shown in the upper portion. At the nozzle lip, the
translational temperature is turned into the lip, and seems to
curve around the blunt end of the wall. By comparison, the
rotational temperature freezes quite quickly in this region, due
to the relatively low density. The vertical lines from the nozzle
lip in Fig. 8 indicate a plane beyond which excessive statistical
scatter exists in the computed data. As stated earlier, the
backflow region of the flow will be computed in the future
with a separate simulation to be begun at the nozzle-exit plane.
The structure of the nozzle and plume flow computed with
the DSMC method is further considered in Fig. 9 where con- 0.0 0.1 0.2 0.3 0.5 0.6

tours of Mach number (lower portion) and flow angle (upper R/De
portion) are shown. The Mach number contours reveal that Fig. 10 Comparison of computed (DSMC and continuum) and mea-
the sonic line intersects the nozzle lip, as observed previously sured pitot pressure profiles for configuration 1 in the nozzle-exit
by Bird.21 Although not shown here, comparison with the plane at Z = 0 mm.
 BOYD ET AL.: LOW-DENSITY PLUME FLOWS OF NITROGEN 2459

two DSMC solutions. Further DSMC computations employ-

ing varying combinations of diffuse and specular reflection,
and the Cercignani-Lampis model,15 all gave profiles lying
between the two extremes indicated in Fig. 10. The fact that
the diffuse model agrees so well with the experimental data is
a strong indication that this model is well suited to the sub-
sonic flow along the nozzle wall. This study provides one of
the very rare occasions in which gas/surface interaction mod-
els for the DSMC technique have been assessed directly
through comparison with experimental flowfield data. The
sensitivity of the computations to the wall model indicates that
possibilities exist for the assessment of such phenomena
through carefully designed experiments, without recourse to
molecular-beam studies.
It is difficult to ascertain whether the differences between
the continuum and particle simulation results are due to the
failure of the Navier-Stokes equations in a rarefied expanding
flow, or instead to the no-slip, nozzle-wall boundary condition
employed in the continuum analysis and the outflow boundary
Downloaded by TECHNION - ISRAEL INST OF TECH on June 13, 2015 | http://arc.aiaa.org | DOI: 10.2514/3.11247

R/De
condition. A computation in Ref. 13 (p. 188) proposes that the
translational temperature is lower than the continuum temper- Fig. 13 Comparison of computed (DSMC) and measured pitot pres-
ature, and the rotational temperature higher, for a one-dimen- sure profiles for configuration 1 at Z = 36 mm.
sional steady spherical expansion at a similar Knudsen number
as the nozzle flow. The exact divergence of the translational
and rotational temperatures is very much dependent upon the

6 6

0.2 0.4 0.6 0.8

R/De
Fig. 14 Comparison of measured pitot pressure profiles for configu-
rations 1 and 2 at Z = 12 mm.
0.4 0.6
R/De
Fig. 11 Comparison of computed (DSMC) and measured pitot pres- rotational-energy-exchange model employed in the simula-
sure profiles for configuration 1 at Z = 12 mm. tion. The two models used in Ref. 13 are different from that
employed in the present study. Also, the present application is
complicated by two-dimensional and boundary-layer effects.
However, it is reasonable to expect similar nonequilibrium
behavior as that which occurs in the one-dimensional, steady
expansion. Therefore, it is concluded that the continuum com-
putation does not predict the flow correctly because of its
inability to capture thermal-nonequilibrium effects and the
requirement for use of a slip boundary condition (both of
which are recognized deficiencies of the Navier-Stokes equa-
tions applied to rarefied flows).
In Fig. 11, the pitot pressure profiles at an axial distance of
12 mm from the nozzle-exit plane are shown for experimental
configuration 1. The two profiles obtained with the DSMC
technique show the effect of applying the correction of Eq. (2)
to the computed results. Once again, the DSMC solutions give
excellent correspondence to the experimental data. These same
trends are observed further downstream at distances of 24 and
36 mm from the nozzle-exit plane in Figs. 12 and 13. It is most
encouraging to find that the excellent agreement between the
DSMC computations and the experimental data is continued
into the plume where the correction factors employed in Eq.
(2) become larger as the density, and the probe Reynolds
Fig. 12 Comparison of computed (DSMC) and measured pitot pres- number, decrease. Although the data from Ref. 17 used for
sure profiles for configuration 1 at Z = 24 mm. the corrections has been corroborated to some extent by the
 2460 BOYD ET AL.: LOW-DENSITY PLUME FLOWS OF NITROGEN

example of this data is shown in Fig. 16 for configuration 2 at

the location of 12 mm from the nozzle exit plane. It is found
that the DSMC computations offer fairly good agreement
with the probe measurements, except in the vicinity of R/
De = 0.31, where the measurements exhibited some scatter. It
is the aim in the future to improve our confidence in the flow-
angle measurements through use of a conical probe designed
specifically for this task. The comparison between experiment
and DSMC analysis for flow angle further demonstrates the
capability of the numerical technique.

Concluding Remarks
By comparison with new experimental data, it has been
shown that the direct simulation Monte Carlo method pro-
vides an accurate description of a low-density flow in the
nozzle and the near-field expansion of a small rocket for two
slightly different experimental configurations. This study
thereby provides verification of the DSMC method in an
Downloaded by TECHNION - ISRAEL INST OF TECH on June 13, 2015 | http://arc.aiaa.org | DOI: 10.2514/3.11247

expanding flow for the first time. From a number of different

Fig. 15 Comparison of computed (DSMC) and measured pitot pres- gas/surface interaction models, it was found that fully diffuse
sure profiles for configuration 2 at Z = 24 mm. reflection gave the best agreement with experiment. Solutions
of the Navier-Stokes equations under the same conditions
were less successful in predicting the flow correctly. It is well
established that, for rarefied flows, the Knudsen number be-
comes so great that the assumptions invoked in developing the
Navier-Stokes equations are invalidated. In previous studies
of shock waves, it has been shown that the DSMC technique
offers better correspondence to experimental data under con-
ditions where the Navier-Stokes equations break down. It is
therefore concluded that the failure of the continuum solution
is affected by the physical limitations of the Navier-Stokes
equations. This is made worse in the present study because
low-density, slip effects along the nozzle wall are also impor-
tant.

Acknowledgment
Support for the first author by NASA Grant NCC2-582 is
gratefully acknowledged.

References
^ettleff, G., "Plume Flow and Plume Impingement in Space
Technology," Progress in Aerospace Science, Vol. 28, 1991, pp.
Fig. 16 Comparison of computed (DSMC) and measured flow angle 1-71.
2
for configuration 2 at Z = 12 mm. Boyd, I. D., and Stark, J. P. W., "Modeling of a Small Hydrazine
Thruster Plume in the Transition Flow Regime," Journal of Propul-
sion
3
and Power, Vol. 6, No. 2, 1990, pp. 121-126.
experimental data taken in the present study, the Reynolds Boyd, I. D., and Stark, J. P. W., "Assessment of Impingement
numbers reached in Fig. 13 were right at the lower limit of that Effects in the Isentropic Core of a Small Satellite Control Thruster
study. It is the intention to investigate these effects thoroughly Plume," Proceedings of the Institution of Mechanical Engineers,
in the future with the DSMC method. Vol.
4
203, 1989, pp. 97-103.
Yoshida, R. Y., Halbach, C. R., and Hill, S. C., "Life Test
The effect of the two different experimental apparatus is Summary and High Vacuum Tests of 10m Ib Resistojets," Journal of
now considered. Radial scans of pitot pressure for both exper- Spacecraft and Rockets, Vol. 8, No. 4, 1971, pp. 414-416.
imental configurations, at an axial distance of 12 mm from the 5
Bailey, A. B., "Flow-Angle Measurements in a Rarefied Nozzle
nozzle exit, are shown in Fig. 14. The pressures from configu- Plume," AIAA Journal, Vol. 25, No. 12, 1987, pp. 1301-1304.
6
ration 2 (triangular symbol) are lower in the core of the flow Legge, H., and Dettleff, G., "Pitot Pressure and Heat Transfer
near the center line, and the overall profile flatter than those Measurements in Hydrazine Thruster Plumes," Journal of Spacecraft
for configuration 1 (circular symbol). The pressures from both and7 Rockets, Vol. 23, No. 4, 1986, pp. 357-362.
configurations are nearly the same in the outer region of the Finke, R. C., Holmes, A. D., and Keller, T. A., "Space Environ-
plume. This figure is typical of the difference in pressure ment Facility for Electric Propulsion Systems Research," NASA TN-
D-2774, May 1965.
profiles between the two experimental configurations. A com- 8
Campbell, D. H., "Angular Variation of Flowfield Properties in
parison of the DSMC results computed for configuration 2, at Free Jet Expansions," Rarefied Gas Dynamics, edited by A. E.
an axial distance of 24 mm from the nozzle exit, is shown in Beylich, VCH Press, Weinheim, Germany, 1991, pp. 1019-1024.
9
Fig. 15. Once again, the numerical data offers excellent agree- Nelson, D. A., and Doo, Y. C., "Simulation of Multicomponent
ment with the experiment. It is significant that the DSMC Nozzle Flows into a Vacuum," Rarefied Gas Dynamics, edited by E.
computations have been successful in predicting the lower P. Muntz, D. P. Weaver, and D. H. Campbell, AIAA, Washington,
pressure at the centerline and the flatter profile observed ex- DC,10
1989, pp. 340-351.
perimentally for configuration 2. Cline, M. C., "Computation of Two-Dimensional, Viscous Noz-
The measurement of pitot pressure at locations away from zlen Flow," AIAA Journal, Vol. 14, No. 3, 1976, pp. 295-296.
Cline, M. C., and Wilmoth, R. G., "Computation of High Rey-
the plume axis involves rotation of the probe. As explained nolds Number Internal/External Flows," AIAA Journal, Vol. 21,
earlier, the experimental data is recorded as the maximum No.12 2, 1983, pp. 172-173.
pressure obtained during this rotation. This point also pro- Lumpkin, F. E., and Chapman, D. R., "Accuracy of the Burnett
vides the angle of the flow at each location in the flowfield. An Equations for Hypersonic Real Gas Flows," Journal of Thermo-
 BOYD ET AL.: LOW-DENSITY PLUME FLOWS OF NITROGEN 2461

physics and Heat Transfer (to be published); see also AIAA Paper pp. 1997-1999.
17
91-0771, Jan. 1991. Shapiro, A. H., The Dynamics and Thermodynamics of Com-
13
Bird, G. A., Molecular Gas Dynamics, Clarendon Press, Oxford, pressible Fluid Flow, Ronald Press, New York, 1953.
18
England, UK, 19786. Stephenson, W. B., "Use of the Pitot Tube in Very Low Density
14
Boyd, I. D., "Vectorization of a Monte Carlo Method For Flows," Arnold Engineering Development Center, Rept. AEDC-TR-
Nonequilibrium Gas Dynamics," Journal of Computational Physics, 81-11, Arnold, AFS, TN, Oct. 1981.
Vol. 96, Oct. 1991, pp. 411-427. 19
White, F. M., Viscous Fluid Flow, McGraw-Hill, New York,
15
Lord, R. G., "Application of the Cercignani-Lampis Scattering 1974.
20
Kernel to Direct Simulation Monte Carlo Calculations," Rarefied Gas Vargartik, N. B., Tables on the Thermophysical Properties of
Dynamics, edited by A. E. Beylich, VCH Press, Weinheim, Germany, Liquids and Gases, Wiley, New York, 1975, p. 459.
21
1991, pp. 1427-1433. Bird, G. A., "The Nozzle Lip Problem," Proceedings of the 9th
16
Boyd, I. D., "Analysis of Rotational Nonequilibrium in Standing International Symposium on Rarefied Gas Dynamics, Vol. I, DFVLR
Shock Waves of Nitrogen," AIAA Journal, Vol. 28, No. 11, 1990, Press, Porz-Wahn, Germany, 1974, pp. A22 1-8.
Downloaded by TECHNION - ISRAEL INST OF TECH on June 13, 2015 | http://arc.aiaa.org | DOI: 10.2514/3.11247

Recommended Reading from Progress in Astronautics and Aeronautics

UNSTEADY Unsteady transonic aerodynamics is

afield with many differences from its
counterpart, steady aerodynamics.
TRANSONIC The first volume of its kind, this
timely text presents eight chapters
AERODYNAMICS on Physical Phenomena Associated
with Unsteady Transonic Flows;
David Nixon, editor Basic Equations for Unsteady Tran-
sonic Flow; Practical Problems: Air-
planes; Basic Numerical Methods;
Computational Methods for Un-
steady Transonic Flow; Application
1989, 385 pp, illus, Hardback of Transonic Flow Analysis to Heli-
ISBN 0-930403-52-5 copter Rotor Problems; Unsteady
AIAA Members $52.95
Aerodynamics for Turbomachinery
Nonmembers $69.95
Order #: V-120 (830)
Aeroelastic Applications; and Alter-
native Methods for Modeling Un-
steady Transonic Flows. Includes
more than 470 references, 180 fig-
Place your order today! Call 1 -800/682-AI AA ures, and 425 equations.

American Institute of Aeronautics and Astronautics Sales Tax: CA residents, 8.25%; DC, 6%. For shipping and handling add $4.75 for 1-4 books (call
Publications Customer Service, 9 Jay Could Ct, P.O. Box 753, Waldorf, MD 20604 for rates for higher quantities). Orders under $50.00 must be prepaid. Please allow 4 weeks for
Phone 301/645-5643, Dept. 41 5, FAX 301/843-01 59 delivery. Prices are subject to change without notice. Returns will be accepted within 15 days.