IAC-05-C4.2.

08
Design Criterion of the Dual-Bell Nozzle Contour
Hirotaka Otsu, Masafumi Miyazawa, and Yasunori Nagata
Shizuoka University, Johoku 3-5-1, Hamamatsu, Shizuoka 432-8561, JAPAN

Downloaded by UNIV OF CALIFORNIA LOS ANGELES on May 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.IAC-05-C4.2.08

E-mail; thootu@ipc.shizuoka.ac.jp

Abstract
In this study, we performed a CFD analysis to determine the optimum dual-bell nozzle contour. Especially, we investigated the effect of the deflection angle at the wall inflection on the separation point transition. Based on our present
analysis, we found that the deflection angle at the wall inflection should be larger than the angle determined by a simple
Prandtl-Meyer expansion. Also, the time to accomplish the separation point transition from the wall inflection to the
extension nozzle exit is estimated to be less than 10 ms when our design criterion of the dual-bell nozzle contour is
applied to the booster engine of H-2A launch vehicle. This time interval is considered to be essentially instant from the
practical point of view of the safe nozzle operation.

1 Introduction
The dual-bell nozzle is an advanced rocket engine nozzle
concept with an altitude-adapted capability.1 This nozzle
concept is promising for application to existing and nearterm launch vehicles with a simple modification to the current bell-type nozzle configuration. As shown in Fig. 1, the
dual-bell nozzle has a base nozzle and an extension nozzle
connected at the wall inflection where a forced, steady and
symmetrical separation takes place at low altitudes. The
wall inflection point, therefore, acts as an effective nozzle
exit with a small nozzle area ratio at low altitudes. At high
altitudes, nozzle flow is attached to the wall through the
extension nozzle exit, thereby providing the use of the full
area ratio. When applied to booster engines to be ignited
on the ground, this nozzle is expected to provide a significant overall propulsion and flight performance gains over
the conventional bell-type nozzle. This is due to its capability to adapt the nozzle exhaust flow to ambient pressure
at low and high altitudes during the ascent flight.
In our previous studies,2, 3 we analytically evaluated the
propulsion and flight performance gains of the dual-bell
nozzles with taking account of weight penalties of dualbell nozzles.2 In that analysis, theoretical-optimum performance gains are obtained by utilizing the flow separation
model in the nozzle proposed by Romine.4 In this model,
the flow separation inside the nozzle is controlled by the
nozzle flow adjustment to ambient pressure, not by the wall
boundary layer. Using this model, we calculated the separation pressure and location through a triple-point shock

Fig. 1: Sketch of the dual-bell nozzle

structure, resulting in good agreements with experimental
data. Figure 2 shows the gain of the trajectory-averaged
specific impulse for LE-7A (LOX/LH2 engine) and RD180 (LOX/kerosene engine). From this figure, we can see
that the trajectory-averaged specific impulse for both LE7A and RD-180 increases by about 10 [s] with the area
ratio of 100.
From the point of view of practical application, we need
to assess the design issues to assure safe and reliable nozzle operations during the launch vehicle ascent phase. A
forced, steady, and symmetrical separation has to take place
at the wall inflection point of the dual-bell nozzle during
the low altitude mode operations. On the other hand, at an

1

1.5

LE-7A (Equilibrium)
LE-7A (Frozen)
RD-180 (Frozen)

12

Radial Distance Y [m]

Trajectory-averaged Isp gain [s]

Downloaded by UNIV OF CALIFORNIA LOS ANGELES on May 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.IAC-05-C4.2.08

14

10

Wall Inflection
1.0

θip

θB
0.5

Extension
Nozzle

Base Nozzle

8

0.0
0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Axial Distance X [m]

6

Fig. 3: Dual-bell nozzle contour for α = 1.0
4
2
0

40

50

60

70

80

90

100

Area Ratio of Extension Nozzle
Fig. 2: Estimated gain of the trajectory-averaged specific impulse for LE-7A and RD-180 (Ref. 2)

altitude where the separation criterion is cleared, the separation point must jump to the extension nozzle exit within
a sufficiently short period of time to avoid unfavorable side
loads and/or vibration hazards. Also, the separation conditions have to be uniquely determined for the safe operation
of this nozzle.
To resolve the above issues on the dual-bell nozzle,
we analyzed the dual-bell nozzle flow with a NavierStokes CFD code applied to laminar axis-symmetric nozzle
flows.3 Especially, we focused on the effect of the deflection angle at the inflection point on the dual-bell nozzle
flow. According to our CFD analysis, it was found that
the deflection angle should be larger than the angle evaluated using a simple Prandtl-Meyer expansion theory and
the time to complete the separation point transition was estimated to be less than 0.1 [s]. Additionally, the theoretical
separation criteria to be applied to the dual-bell nozzle were
also assessed.
In the present study, we tried to narrow down the overexpansion level at the wall inflection through our CFD
analysis. Additionally, the effect of the nozzle length on
the separation point transition was also investigated.

2 Design of Dual-Bell Nozzle Contour
In this section, we describe how to design the dual-bell nozzle contour in the present study. The area ratio at the inflection point is set to be 37.5, which is same as that of LE-7A.
At the extension nozzle exit, the area ratio is set to be 100.
The contour of both the base nozzle and the extension nozzle is a parabola. The detail is as follows.

• The base nozzle length and the total length of dualbell nozzle are assumed to be 80% of a conical nozzle
with a half-angle of 15 [deg.].
• The parabola for the base nozzle is determined using
the coordinates of nozzle throat and inflection point
along with the deflection angle θB at the base nozzle
exit.
• The parabola for the extension nozzle is determined
using the coordinates of inflection point and extension
nozzle exit along with the deflection angle θip at the
inflection point.
• The deflection angle at the inflection point, θip , is determined using the Prandtl-Meyer function.
θip = θB + α (νE − νB )

(1)

,where Prandtl-Meyer angles, νE and νB , are calculated using one-dimensional isentropic flow relation as shown below.


γ+1
γ−1
−1
ν =
tan
(M 2 − 1)
γ−1
γ+1

− tan−1 M 2 − 1
The magnitude of the inflection angle is expressed in terms
of over-expansion factor, α, defined as the multiple factor with respect to the constant extension nozzle designated
wall pressure case in which the inflection angle equals the
difference of Prandtl-Meyer angles between the base and
extension nozzle exits. This factor corresponds to the degree of over-expansion at the inflection point. In the present
analysis, this factor, α, is selected to be 0.8, 1.0, 1.1, 1.2,
1.25, 1.5, and 2.0 to determine the effect of θip on the
behavior of the shock-induced flow separation. Figure 3
shows the dual bell nozzle contour with the over-expantion
factor of 1.0.

3 CFD Analysis and Conditions
The governing equations for the present CFD analysis
are the full Navier-Stokes equations for laminar axis-

2

For our CFD code, we selected AUSM-DV scheme with
MUSCL technique to obtain a higher order accuracy.5 The
governing equations are discretized using the finite volume
method and integrated in time by the Euler explicit method.

Radial Distance Y [m]

The viscosity coefficient and thermal conductivity are
calculated using Sutherland’s formula. Although this formula is valid for pure air, the effect of viscosity on the
shock-induced flow separation inside the nozzle is expected
to be negligible. In fact, we investigated the effect of viscosity on the flow separation with varied viscosity coefficients from 1% to 10000% of the original value. The results indicated no significant difference. The conditions in
the combustion chamber and the exhaust gas are taken from
Table 1.

1.5

1.0

0.5

0.0

0

1

2

3

4

Axial Distance X [m]

Fig. 4: Computational grid inside the dual-bell nozzle
40

Radial Distance Y [m]

30

20

10

0

0

25

50

75

100

Axial Distance X [m]

As for boundary conditions, along the nozzle surface,
the adiabatic wall is assumed and the inflow condition is
imposed on the nozzle throat. The physical properties at
the nozzle throat are calculated from the combustion chamber conditions using the isentropic relations with the Mach
number of 1.0. On the axis of symmetry, the symmetric
condition is applied. At the end of the computational domain, 0th order extrapolation is used for all physical properties. In order to determine the pressure where the separation point transition takes place, the ambient pressure
ranges from 100 [kPa] to 1 [kPa] and the pressure was determined to an accurary of 0.01 [kPa]. Figure 6 shows the
applied boundary condition on each boundary of the computational domain.

Table 1: Combustion chamber conditions for LE-7A
Mixture Gas Molecular Mass [g/mol]
Heat Ratio
Chamber Pressure [kPa]
Chamber Temperature [K]

13.0
1.24
12,258
3,600

Fig. 5: Computational grid for the whole computational
domain
0th order Extrapolation

Adiabatic Wall
Non-slip Condition

0th order Extrapolation

The computational domain is separated into two areas:
one is inside the nozzle and the other is outside the nozzle
as shown in Fig. 4 and 5. Inside the nozzle, the computational grid has 300 points along the nozzle wall and 100
points along the line normal to the nozzle centerline. Outside the nozzle, the computational grid has 150 along the
line of symmetry and 130 points along the line normal to
the line of symmetry. The computational grid is shown in
Fig.4 and 5.

Non-slip Condition

Downloaded by UNIV OF CALIFORNIA LOS ANGELES on May 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.IAC-05-C4.2.08

symmetric flow. The exhaust gas inside the nozzle is assumed to be perfect gas.

M=1
Throat㧔‫ޓޓ‬㧕
Symmetric Condition

Fig. 6: Applied boundary condition on each boundary

4 Results and Discussions
Flow features inside the dual-bell nozzle
Figure 7 shows the Mach number distribution inside the
nozzle for both low and high altitude modes. When the
ambient pressure is relatively high, which corresponds to
the low altitude mode, the flow separation occurs at the inflection point. In the low altitude modes, the effective area
ratio is limited to the area ratio at the inflection point, in
other words, the base nozzle exit. Additionally, the normal
shock and the triple point occur near the extension nozzle
exit. When the ambient pressure is relatively low, which
corresponds to the high altitude mode, the flow separation
occurs at the extension nozzle exit. This means that the effective area ratio is the area ratio at the extension nozzle
exit in the high altitude modes.
From this figure, we can see that the flow pattern inside

3

Inflection Point

LE-7A Dual-Bell Nozzle

2

Nozzle Exit

55

Mach Number

Pa = 50 [kPa]

50

1

Triple Point

0

Wall Pressure [kPa]

Radial Distance [m]

45

0 1 2 3 4 5 6

Low Altitude Mode
Pa = 100 [kPa]

35
Pa = 30 [kPa]

30

Pa = 25 [kPa]

25
20

Pa = 19 [kPa]

15
10

Pa < 15 [kPa]

-1

5
1.5

High Altitude Mode

2.0

2.5

3.0

3.5

Axial Distance [m]

Pa = 19 [kPa]

Inflection Point

Nozzle Exit

55

-2

Pa = 50 [kPa]

50
45

1

2

3

Axial Distance [m]

4

5

Fig. 7: Mach number distribution inside the dual-bell
nozzle for both low and high altitude modes. The overexpansion factor, α, is 1.0.

Wall Pressure [kPa]

0

Pa = 40 [kPa]

40
35

Pa = 30 [kPa]

30

Pa = 25 [kPa]

25
20

Pa = 20 [kPa]
15
10
Pa < 15 [kPa]
5
1.5

Effect of over-expansion factor on transition
As expected, the over-expansion factor, α, the degree of
over-expansion at the wall inflection, should be larger than
1.0 to ensure sufficiently fast transition. However, excessively large factors may introduce difficulty in manufacturing the nozzle wall near the inflection point. Based on this
considereation, we tried to look for the optimum range of
this factor that provides a sufficiently fast transition without
any problems in design and manufacture.
Figure 8 shows pressure distributions along the nozzle
wall with varied ambient pressure for α = 0.8, 1.0, 1.2, and
1.5. In all cases, the separation point moves from the inflection point to the extension nozzle exit with decrease in
the ambient pressure.
For α = 0.8, the separation point starts to move from
the inflection point at the ambient pressure of 30 [kPa].
The separation point reaches the extension nozzle exit at
the ambient pressure of 20 [kPa]. Therefore, in this case,
the ambient pressure where the separation point transition
starts, pa1 , is 30 [kPa] and the ambient pressure where the
separation point transition ends, pa2 , is 20 [kPa]. Thus, the
pressure difference between pa1 and pa2 is relatively large.
For α = 1.0, the trend is almost similar to that for α =0.8.

2.0

2.5

3.0

3.5

Axial Distance [m]
Inflection Point

Nozzle Exit

55
Pa = 50 [kPa]

50
45
Wall Pressure [kPa]

the nozzle changes drastically according to the change of
the ambient pressure due to the shock-induced flow separation. This indicates that the ambient pressure where the
flow pattern changes is one of the key properties to estimate
the dual-bell nozzle performance gain.

Pa = 40 [kPa]

40
35

Pa = 30 [kPa]

30
25

Pa = 23.31 [kPa]

20
Pa = 23.30 [kPa]

15
10
5
1.5

Pa < 15 [kPa]
2.0

2.5

3.0

3.5

Axial Distance [m]
Inflection Point

Nozzle Exit

55
Pa = 50 [kPa]

50
45
Wall Pressure [kPa]

Downloaded by UNIV OF CALIFORNIA LOS ANGELES on May 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.IAC-05-C4.2.08

Pa = 40 [kPa]

40

Pa = 40 [kPa]

40
35

Pa = 30 [kPa]

30
25

Pa = 22.61 [kPa]

20
Pa = 22.60 [kPa]

15
10
5
1.5

Pa < 15 [kPa]
2.0

2.5

3.0

3.5

Axial Distance [m]

Fig. 8: Pressure distribution along the nozzle wall for α
= 0.8, 1.0, 1.2, 1.5 (top α=0.8, bottom α=1.5, LD =80%)
On the other hand, in the case of α=1.2 and 1.5, the pressure difference between pa1 and pa2 is less than 0.01 [kPa],

4

55

Table 2: Effect of the over-expansion factor on the transition time for LD = 80%

Inflection Point

Nozzle Exit
Pa = 50 [kPa]

50
45

Pa = 40 [kPa]

40

0.80
30
19
9.1869
12.170
6767

1.00
25
20
10.389
11.912
3360

1.20
23.31
23.30
10.865
10.866
3.111

1.50
22.61
22.60
11.114
11.115
3.114

Wall Pressure [kPa]

α
pa1 [kPa]
pa2 [kPa]
H1 [km]
H2 [km]
ttr. [ms]

35
Pa = 30 [kPa]

30

Pa = 25 [kPa]

25
20

Pa = 20 [kPa]

15

Pa = 15 [kPa]

10

Pa = 10 [kPa]

5
0
1.6

55

H-2A Nominal Flight
Transition Start
Transition Complete

2.0

2.2

2.4

2.6

11.5

3.0

α=1.2

α=1.0

10.0

Nozzle Exit
Pa = 50 [kPa]

Pa = 40 [kPa]

40

10.5

35
Pa = 30 [kPa]

30
25

Pa = 20 [kPa]

20

Pa = 17 [kPa]
15
Pa = 12 [kPa]

10

9.5

5

Transition Time for α=0.8
9.0
47

2.8

45

α=1.5

11.0

Inflection Point

50

Wall Pressure [kPa]

Flight Altitude [km]

0
1.6

48

49
50
51
52
53
Flight Time from Liftoff [s]

54

Pa < 5 [kPa]
1.8

2.0

55

2.2

which is very small compared to that in the case of α = 0.8,
1.0. This indicates that in the case of α= 1.2 and 1.5 the
time necessary to switch from low altitude mode to high
altitude mode is expected to be much shorter than that in
the case of α = 0.8, 1.0.
Next, we estimated the transition time from low altitude
mode to high altitude mode. From our CFD analysis, we
detected the ambient pressures where the separation point
transition starts and ends. These pressures, pa1 and pa2 , can
be related to corresponding altitudes, H1 and H2 . H1 is the
altitude where the separation point transition is estimated to
start and H2 is the altitude where the separation point transition is estimated to end. The results are shown in Fig. 9
and Table 2. From the results, we observe that the transition time for α=1.2 and 1.5 is about 3.0 [ms]. This value is
much smaller than that for α = 0.8 and 1.0. Therefore, from
a viewpoint of transition time, the over-expansion factor α
should be at least 1.2.

Mechanism of the instant transition
The trend about the separation point transition can be characterized by the pressure gradient inside the extension nozzle. When the over-expansion factor is less than 1.0, the

2.6

55

Inflection Point

2.8

3.0

Nozzle Exit
Pa = 50 [kPa]

50

Fig. 9: Effect of the over-expansion factor on the transition time for LD = 80%

2.4

Axial Distance [m]

45
Pa = 40 [kPa]

40
Wall Pressure [kPa]

Downloaded by UNIV OF CALIFORNIA LOS ANGELES on May 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.IAC-05-C4.2.08

1.8

Axial Distance [m]

12.5
12.0

Pa < 5 [kPa]

35
Pa = 30 [kPa]

30
25

Pa = 20 [kPa]

20
15

Pa = 13.14 [kPa]

10

Pa = 13.13 [kPa]

5
0
1.6

Pa < 10 [kPa]
1.8

2.0

2.2

2.4

2.6

2.8

3.0

Axial Distance [m]

Fig. 10: Pressure distribution along the nozzle wall for
α = 1.2, 1.5, 2.0 (top α=1.2, bottom α=1.5, LD =70%)
pressure gradient is negative, in other words, the pressure at
the inflection point is higher than that at the extension nozzle exit. The typical shock-induced flow separation criteria
are represented by the pressure ratio between the ambient
pressure and the pressure at the shock. This means that the
pressure at the shock decreases with decrease in the ambient pressure. Therefore, when the pressure gradient is
negative, the location where the criterion is satisfied moves
from the inflection point to the nozzle exit with decrease in
the ambient pressure and the separation point moves from
the inflection point to the exit with decrease in the ambient pressure, as a result, the separation point transition can
not complete instantly. On the other hand, when the pressure gradient is positive, the prssure at the inflection point

5

4

10

5

L = 80%

Conclusions

D

Downloaded by UNIV OF CALIFORNIA LOS ANGELES on May 2, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.IAC-05-C4.2.08

Transition Time [ms]

L = 70%
D

3

10

2

10

1

10

0

10

0.8

1.0

1.2

1.4

1.6

1.8

Over-expansion factor, α

2.0

Fig. 11: Effect of the nozzle length on the transition time

is lower than the pressure at the nozzle exit. This means
that the separation criterion is satisfied at the nozzle exit
before the criterion is satisfied at the inflection point. This
is why the separation point moves from the inflection point
to the nozzle exit instantly.

Effect of the nozzle length
Finally, we investigated the effect of the nozzle length, LD ,
on the separation point transition. In this case, the total
nozzle length, LD , is set to be 70% of the conical nozzle of
the half angle of 15 [deg.] Other parameters for the dualbell nozzle contour was fixed.
Figure 10 shows the pressure distribution along the nozzle wall for α = 1.2, 1.5, 2.0. From this figure, it was
found that the pressure gradient inside the extension nozzle is negative even for α =1.2 and 1.5 and the gradient is
positive for α = 2.0. Thus, in the case of LD = 70 %, the
preffered over-expansion factor seems to be greater than
2.0.
Figure 11 shows the transition time for LD = 70%
and 80%. Although the preffered over-expansion factor
is greater than 1.2 for LD = 80%, the preffered overexpansion factor is expected be greater than 2.0 for LD =
70% to complete the separation point transition within 10
[ms]. The difference of the over-expansion factor is significant. This indicates that the shorter nozzle length may
require the larger over-expansion factor at the wall inflection. Thus, on design of the dual-bell nozzle contour, we
should pay attention to both the over-expansion factor and
the nozzle length.

In order to resolve design issues for the dual-bell nozzle
contour, we performed a CFD analysis of dual-bell nozzle
flows. From our present CFD calculations, we found that
essentially instant separation point transition takes place
when the over-expansion factor is greater than 1.2. The
transition takes place during the time interval well under
10 [ms] at the altitude of about 10 [km] for a simulated
H-2A boost flight.
Hence, we conclude that the over-expansion factor for
instant transition should be greater than 1.2 for LD = 80%.
But it was also found that the over-expansion factor needs
to be greater than 2.0 for LD = 70%. This means that
the nozzle length also has a significant impact on separation point transition and estimation of the preffered overexpansion factor for the dual-bell nozzle. On design of the
dual-bell nozzle contour we need, therefore, to pay attention to both the deflection angle and the nozzle length for
preffered features of the dual-bell nozzle.

References
[1] Hagemann, G., Immich, H., Nguyen, T. V., and Dumnov, G. E., “Advanced Rocket Nozzles,” Journal of
Propulsion and Power, Vol. 14, No. 5, 1998, pp. 620–
633.
[2] Miyazawa, M., Takeuchi, S., and Takahashi, M.,
“Flight Performance of Dual-Bell Nozzles,” AIAA
Paper 2002-0686, Jan., 2002.
[3] Miyazawa, M., and Otsu, H., “An Analytical Study
on Design and Performance of Dual-Bell Nozzles,”
AIAA Paper 2004-0380, Jan., 2004.
[4] Romine, G. L., “Nozzle Flow Seperation,” AIAA
Journal, Vol. 36, No. 9, Sep. 1998, pp. 1618–1625.
[5] Wada, Y. and Liou, M.-S. “A Flux Splitting Scheme
with High-Resolution and Robustness for Discontinuities,” AIAA Paper 94-0083, January, 1994.

6