Method of characteristics to evaluate steady 2D flow in

a supersonic nozzle
Harsharaj Parmar, Purdue University, West Lafayette, IN, 47907
Submitted as a project report for ME510 Gas Dynamics

1 Model description
The flowchart shown on the next page gives a brief overview of the algorithm used to find
the parameters involved in the method of characteristics. A detailed explanation of how the
parameters are calculated and the inverse Prandtl-Meyer function is now provided for closure.

Before beginning, a crucial nomenclature used in the flowchart that is fronts is explained
using Fig.1. Fronts in the current study depict the group of points like [1,2,3,4,5,6,7],
[9,10,11,12,13,14] in Fig. 1 and so on. Thus, if we have n expansion waves in our computation
then we have n fronts in our plane with the first front being the group [1,2...,n] and progressed
to the right.

Figure 1: Illustrative plot of characteristics [Anderson].

1.1 Initialization
• Using Mexit find θmax . Consider Fig. 1 where we see that point 34 has no expansion
wave interactions downstream of it so Mexit has been achieved there. Also a C−
characteristic runs from point a to 34 which implies that,

K− (a) = K− (34) =⇒ θmax + νa = θ34 + ν34

Since 34 lies on centerline (θ34 = 0) and expansion at a is from sonic conditions

(∆θ = νa ) we get,

νe
2θmax = νe =⇒ θmax =
2
where νe is obtained from Mexit and Prandtl-Meyer relation.

1
Figure 13: Flowchart depicting the computational steps

13
• Based on the number of expansion waves n we partition θmax and calculate the total
number of points ntotal from,

(n + 1)(n + 2)
ntotal = −1
2

1.2 First front and boundary evaluations

• In the first front we know that the expansion has taken place from sonic inlet and so
from θ − ν relation we have θi = νi . This implies that for the first front both θ and ν
values are known and we can find the K+ and K− values using,

K+ = θ − ν K− = θ + ν

• Notice that all the C− characteristics originate from nozzle corner ”a” so is we know the
K− values at the first front then we can propagate them downstream. This downstream
sweep assigns K− values to all the points in the domain.

• Now at the centerline boundary we have θ = 0 and using the K− values obtained before
we can find the other parameters,

ν = K− K+ = −K−

At this point we know the centerline boundary parameters and the first front parameters.
Also, note that nozzle wall has the same properties as the point preceding it on the
front so the wall values will be updated at the end once all the interior values have
been evaluated.

1.3 Interior evaluations

• Moving to interior points we note that all the C+ characteristics arise from the centerline
boundary points and since we have already evaluated the K+ values at these boundary
points we can use them for K+ values at all the interior points.

• Considering the second front in Fig. 1 for instance, we know the K+ at 9 (since its
a boundary point) and hence we can propagate it along the front to [10,11,12,13,14].
Now both K− and K+ values are known for interior points of the second front and we
can find θ and ν from,

(K− + K+ ) (K− − K+ )
θ= ν=
2 2
This procedure is carried out for all the fronts and so we get properties at all the
interior points.

2
 1.4 Nozzle wall evaluations
• Properties at the nozzle wall are identical to those at points before it on the fronts
because no expansion wave interaction occurs between them. Since all interior points
have been evaluated the nozzle wall calculations are relatively easy and straightforward.

1.5 Inverse Prandtl-Meyer function and Mach angle

• At this point we have the values of θ and ν at all the points in the domains and we
move on to finding the Mach number. For a given Mach number we have a relation to
obtain ν but the inverse relation cannot be solved for analytically. Many studies use
polynomial data-fitted relations to find M corresponding to a particular ν. Here the
polynomial is adopted from ”Inversion of Prandtl-Meyer Relation”,Hall, 1975 given as,

1 + Ay + By 2 + Cy 3
M=
1 + Dy + Ey 2
2/3
π √

ν

where y = and ν∞ = 6 − 1 with the constants given for γ = 1.4 as,
ν∞ 2

A 1.3604
B 0.0962
C -0.5127
D -0.6722
E -0.3278

The accuracy of approximation used is quite good with maximum error less than 0.01%
in the range of ν (0 − 50◦ ) presented by the current problem as shown in Fig. 2.

Figure 2: Error plot for Hall approximation with ν [I. M. Hall, 1975 ]

3
• Once the Mach number is obtained at each point in the domain then the Mach angle
can be calculated at each point by using,
 
−1 1
µ = sin
M
All property evaluations have been completed at this stage and the only step remaining
is to define the location of points in the domain.

1.6 Locating points in the domain

• For simplicity we set the origin to be the nozzle corner point ”a” in Fig. 1 from which
expansion waves emanate and the throat half length to be 1 unit. The throat half
length gives us the y coordinates for all the centerline boundary points given the origin
at a.

Figure 3: Schematic showing slopes of characteristic lines [Anderson]

• Using the slope relations given in Fig. 3 for C− and C+ characteristics we can find the
location of point 3 given the coordinates of 1 and 2 (note that the points in Fig. 3 do
not represent the actual case but are solely for representation).
• The angles for nozzle wall points are a little different from those in the interior, precisely
the C− characteristic slope differs and is given by,
θi + θmax θi + θi−1
tan(θC− ,i ) = f or i = 1 tan(θC− ,i ) = f or i > 2
2 2
where i represents the front under consideration and i − 1 the previous front.
• Referring to Fig. 1, we first determine location of 1 using C− slope and the throat half
length. From there we proceed up the first front and determine the locations using
(0, 0), previously calculated coordinates and slopes given in Fig. 3.
• Once all points on a front are calculated we move to the next front and march from
the boundary point up towards the nozzle wall. Finally, we get the coordinates for all
points in the domain and post process the data to visualize the expansion waves and
nozzle surface design.

4
2 Turning angle at Mach 3.0

Figure 4: Rough schematic of nozzle with four expansion waves

Let us consider the rough schematic in Fig. 4 to understand the calculation of maximum
turning angle at the corner. In particular focusing on points 14, 13 and a we see that point
14 is at the exit condition and since there are no expansion waves between 13 and 14, the
same exit conditions persist at 13.

ν14 = ν13 = ν(3.0)

Moreover, point 13 lies on the centerline and so,

θ13 = 0 =⇒ K−,13 = θ13 + ν13 = ν(3.0)

Now a C− characteristic runs from a to 13 (a − 4 − 8 − 11 − 13) and along a C− characteristic

the value of K− remains constant.

K−,a = K−,13 =⇒ θa + νa = ν(3.0)

The maximum turning at a results in θa = θmax and since the expansion at a is a Prandtl-Meyer
expansion from initially sonic conditions we have νa = θmax which gives us,
ν(3.0)
θmax =
2
Using the Prandtl-Meyer function with γ = 1.4 or Table A.5 from Anderson we get,

ν(3.0) = 49.7574◦ =⇒ θmax = 24.8787◦

We get the maximum turning angle to be θmax = 24.8787◦ and this would be partitioned
to obtain the θ for each expansion wave. In most computations the θ for first expansion
wave (a-1 in Fig. 4) is ensured to be as close to 0 as possible. This is because 1 lies on
the centerline and so by default has θ = 0 but we supply a non-zero value and this causes
an inconsistency. In this study θ1 is set to 0.2787◦ and the rest of 24.6◦ are partitioned as
needed.

5
3 Results
Computational results are obtained for the nozzle with four, five, seven and nine expansion
waves. As mentioned before in each computation θ1 is set to a small value of 0.2787◦ to
avoid any inconsistencies and the rest of 24.6◦ are divided as per requirement. Some notable
observations can be listed as follows,

• Increasing the number of expansion waves results in better resolution of the flow since
more characteristics now span θmax giving a closely spaced net across the nozzle.

• Good correspondence is obtained with quasi-1D flow area ratio at the outlet even when
only four expansion waves are used.

As a consequence of these observations the nozzle geometry, wave patterns, centerline

and wall Mach number results are all plotted for the case of nine expansion waves where the
flow is best resolved.

3.1 Characteristic lines and geometry

The coordinates mentioned in the plot given in Fig. 5 describe the last point in the domain
and help in determining the nozzle length and numerical area ratio at the outlet of the nozzle.

lmin = xe = 16.9309

Figure 5: Characteristic lines and nozzle surface profile for nine expansion waves

6
3.2 Mach numbers at wall and centerline
Mach number are plotted along nozzle length at the wall and centerline with M set to 1 at
x = 0 as shown in Fig. 6. Notice that at the centerline we only have points till x = 5 and
so the plot terminates earlier than the wall case where the entire length is covered.

Figure 6: Plot of centerline and wall mach numbers along the length of the nozzle for nine
expansion waves

3.3 Convergence analysis

Convergence in the solution can be examined from the plots given in Fig. 7 and 8 by checking
their separation. Firstly, the results in both plots are very close to each other which suggests
that the method works well even at lower number of expansion waves. Secondly, looking into
Fig. 7 for centerline mach numbers we observe visible dispersion in Mach number values from
x = 3 − 4. The lowest line in purple represents 4 expansion waves and differs considerably
from values at 7 and 9 expansion waves case which are very close.This shows that as we
increase the number of waves the plots move closer to each other to represent a singular
solution which is the essence of convergence.

Similarly, considering the plot in Fig. 8 we note a considerable variance from x = 10 − 14

and the case of 4 expansion waves slightly under predicts the Mach number. Here again
the results for 7 and 9 expansion waves are very close to each other and represent the
notion that as we increase the number of waves we would get a converged solution. The
spreading decreases as we move to higher number of expansion waves (4-5-7-9) and thus
denotes convergence.

7
Figure 7: Plot of centerline mach numbers along the length for different number of expansion
waves.

Figure 8: Plot of centerline mach numbers along the length for different number of expansion
waves.

3.4 Minimum nozzle length

The term minimum nozzle length is used because the nozzle design adopted here excludes
the expansion section which gradually changes the turning angle to θmax from the throat.
8
If the expansion section is included then the length of the nozzle increases attributed to
the gradual change in turning angle. In our computations the maximum turning angle is
applied straight away and so the length of nozzle is said to be minimum. As seen from the
convergence studies, solution obtained at 9 expansion waves is the most accurate and is very
close to the converged solution. Reverting back to Fig. 5 we observe that the coordinates of
last point in the domain are (16.9309,3.2443) which implies that,
lmin = 16.9309
where the unit can be defined based on given units of throat area. Here lmin = 16.9309 units
since half throat length is taken to be 1 unit. Also as seen in Fig. 9, 10 and 11 the nozzle
length increases with decrease in the number or expansion waves and so 9 expansion waves
indeed gives us the minimum calculated nozzle length.

Figure 9: Characteristic lines and nozzle surface profile for seven expansion waves.

3.5 Quasi 1-D flow comparison

In this section a comparison is laid out between the numerical results obtained from solution
of nine expansion waves in the nozzle and quasi 1-D flow results. Two comparisons are
shown one graphical and the other quantitative. Graphical comparison is drawn using the
numerical area ratio prevalent at the wall points and the computed area ratio from quasi
1-D flow relation at given wall Mach numbers.

3.5.1 Graphical area ratio comparison along nozzle length

Wall Mach number variation with length was considered in Fig. 8 but here we shall observe
the computed area ratio and the area ratio from quasi 1-D relations using Mach numbers at

9
 Figure 10: Characteristic lines and nozzle surface profile for five expansion waves.

Figure 11: Characteristic lines and nozzle surface profile for four expansion waves.

the wall points along nozzle length. Now for the computed area ratio we have the coordinates
of all the wall points and can easily calculate the ratio using,
 
A
= 1 + yi
A∗ num,i

10
 For the quasi 1-D flow area ratio we use the Mach number at the wall points and plot
them with length along with the numerical ratios.
 2   (γ+1)/(γ−1)
A 1 2 γ−1 2
= 1+ Mi
A∗ 1D,i Mi2 γ + 1 2

Figure 12: Area ratio comparison of numerical results with quasi 1-D flow results for the
case of nine expansion waves.

Interesting observations can be made from the plot shown in Fig 12.

1. For the same Mach number and throat area, computed nozzle area is greater than
the quasi 1-D flow area. Assuming equal stagnation temperatures, we have equal
temperature and subsequently equal flow velocities along the length of the nozzle for
both cases.
2. Moreover, the throat is choked in both the cases so mass flow rate is fixed and equal.
Now for equal flow velocity and mass flow rate, since computed area is greater the
density prevalent should be lower than the quasi 1-D flow case.
3. Lower density in the computations indicates lower pressures from ideal gas law and so
we have that pressures are lower inside the nozzle in the computations than quasi 1-D
results. Expansion is greater in computations than 1-D case.
4. This is in accordance to our expectations because expansion wave interactions result
in pressure drops in our computations and those are absent in 1-D analysis. Note that
expansion waves begin interacting at the wall after the first wall point and so deviations
are observed after the same.

11
 5. Finally as we move to end of the nozzle the expansion interactions reduce to the extent
that at the last wall point no expansion waves exist which has been seen earlier in the
report. As a result of this, the curves move closer to each other and converge at the
end to result in the same area ratio and pressure drop across the nozzle.

3.5.2 Outlet area ratio comparison

For the given Me = 3.0 we can get the area ratio at outlet from quasi 1-D flow analysis and
compare it with the numerical ratio. Using compressible flow calculator at M = 3.0 we have,
 
A
= 4.2346
A∗ 1D
 
A
From Fig. 5, 9, 10, 11 and = 1 + ye we get the following values,
A∗ num,e

Table 1: Outlet area ratio comparison

 
A
Number of %error
A∗ e
expansions
9 4.2443 0.2291
7 4.2513 0.3944
5 4.2765 0.9895
4 4.3194 2.0026

Table 1 clearly shows that the numerical model gives decent results for four expansion
waves and high fidelity solutions for seven and nine expansion wave cases (error drops below
0.5%).

4 Conclusion
In conclusion we can say that method of characteristics can accurately resolve steady two
dimensional flow problems in nozzle design and achieve high levels of accuracy. For the
current problem with Me = 3.0 we get the minimum nozzle length lmin = 16.9309 with nine
expansion
  waves spanning the maximum turning angle. Area ratio at the exit is found to be
A
= 4.2443 and exhibits an error of 0.2291% compared to quasi one dimensional flow
A∗ e
solution.

Convergence studies have been done from Mach number plots at wall and centerline and
can be assumed to be established after seven expansion waves for this problem. The solution
with nine expansion waves is the closest to the converged solution and can be deemed useful
for design.

12