Rocketry MIT

16.
512, Rocket Propulsion

Prof. Manuel Martinez-Sanchez
Lecture 1: Introduction
Types of Rockets (Engines)
- Depending on gas acceleration mechanism/force on vehicle mechanism.
“Thermal” Gas pushes directly on walls by P (pressure) forces

Nozzle accelerates gas by P forces
(most large rockets, chem, nuclear, some electric…)
JG
Electrostatic Ions accelerated by E field
(a) Electrostatic force (push) on electrodes
(Ion engines) G
(b) Force (push) on magnetic coils through gas j
(Hall thrusters)
G JG
Electromagnetic Gas accelerated by j × B forces
Force (push) on coils or conductors
(MPD thrusters, PPT’s)
16.512 concentrates on Thermal
- Depending on energy source:
Solid Propellant
Chemical (always “thermal”) Liquid Propellant Monopropellant

Bipropellant
Hybrid
Nuclear (Thermal)
Nuclear (Electric) can be Thermal, ES or EM
Solar (Thermal)
Solar (Electric) can be Thermal, ES or EM
16.512 deals mostly with Chemical.
- Depending on Thrust level (per unit mass)
- High thrust ( ≥ 1g) for launch, fast space maneuvering (16.512)

- Low thrust (10-5 – 10-2 g) for efficient in-space maneuvers (16.522)
16.512, Rocket Propulsion Lecture 1

Prof. Manuel Martinez-Sanchez Page 1 of 3
Performance Measures
Specific Impulse
Isp =
F
mg ( F
or c =
m )
(sec) (m/sec)
Dominant for chem. Rockets, range 200-500 sec

Trade-off vs. mass for EP, range 500-6000 sec
Thermal Efficiency ηth = Jet kinetic power

Thermal input power
(Thermal Rockets)
Also for electrical thrusters
Power to jet
η =
Input electrical power
~ 30-80 %
ηth
Very close to 100% in chem. (non-issue) important in solar thermal (60-80%)
electrothermal, etc.
Thrust/weight F/W
Very large ~ (20-100) for Chem.

Medium (5-20) for Nuclear
Very low (~10-3) for (Solar, EP, power limited)
Others (design selection factors)
- “Life”, most meaningful in total impulse capacity

- Re-start capability
- Throttleability
- Dispersion
- Cost
Rocket Selection Guide (by mission)

1) Non-Space missions Rocket Type
Atmospheric/Ionospheric Sounding Solid Propellant, 1-4 stages
Tactical Missile Solid Prop., 1-2 stages
Medium-Long Range Missiles Solid or Liquid Prop., 2-3
stages (very high
acceleration)
2) Launch to space Solid, liquid or combinations,

2-4 stages (2-4g)
Possible: hybrid, 2-4 stages
3) Impulsive ∆V in space Small Solid Prop. (Apogee

(time-critical maneuvers, kick, etc)
energy change from elliptic orbits, Bi-propellant (storable)
plane change from elliptic orbits, liquids, Monopropellant
non-fuel-limited situations...) (storable) liquids,
∆V ≤ 1000 m/s Future: Nuclear thermal
4) Low-Thrust ∆V in space
(Mass-limited missions ∆V ≥ 2000 m/s Solar-electric systems:
non time-critical missions, Arcjets (a bit faster, less Isp)
small, continuous orbit corrections Hall, Ion (slower, higher Isp)
near-circular orbits...) PPT (precision maneuvers)
Nuclear-electric systems
Direct solar-thermal

16.512, Rocket Propulsion,
Lecture 2: Rocket Nozzles and Thrust
Rocket Thrust (Thermal rockets)
i
m= ∫∫ ρ u dA
Ae
n e
∫∫ P dSx − ∫∫ P dA
e ex = ∫∫ u x ( ρ un )dAe
Solid int . Ae Ae
surfaces
i
(Tanks included) dm
Note: ∫∫s.,int
PadSx − ∫∫ P dA
Ae
a ex = 0, so subtract,
∫∫ ( P − Pa ) dSx = ∫∫ ( P e − Pa ) dAex + ∫∫ ρu u dA
x n e
Solid int . Ae Ae
Thrust ≡ F

∫∫ ρu u dA
Ae
x n e
In general then, define ue =

∫∫ ρu
Ae
n dAe
∫∫ P dA
Ae
e ex
and P e =
Aex
( )
i
⇒ F = m u e + P e − Pa Aex
If things are nearly constant on spherical caps, modify control volume to

spherical wedge:
i
m= ∫∫ ρu dA
Ae
r
∫∫ ( P − Pa ) dSx − ∫∫ ( P
e − Pa ) dAex = ∫∫ ( ρu ) u dA
r x
int . Ae Ae
solids.
dAex = dA cos θ ux = ur cos θ

Define
ue =
∫∫ Ae
ρ ur ux dA
; Pe =
∫∫ Ae
Pe dAex
i Aex
m
and use
dA = 2π r sin θ rdθ
For ideal conical flow, ρ , ur , P are constant over Ae . Then
α 1
ue =
ρ ur2 ∫∫ Ae
cos θ dA
= ur
∫0
2π r sin θ cos θ dθ
= ur 2
sin2 α
α
ρ ur ∫∫ Ae
dA
∫0
2π r sin θ dθ
1 − cos α
or
1 + cos α
ue = ur
2
Also, since Pe = const on the exit surface, P e = Pe

∫ ( P − P )dA
s.s.
a x + ∫ (P
Ae
e − Pa )dAx = ∫ ( ρu dA)u
Ae
n x
( )
i
F = m u e + P e − Pa Aex
i
m= ∫ Ae
ρundA
ue =
∫Ae
ρ unux dA
∫ Ae
ρ un dA
Pe =
∫ Ae
Pe dAx
∫ Ae
dAx
Ax = ∫ Ae
dAx
At design, Pe = Pa (and parallel flow beyond). Also ue x

i
Then uniform → F = m uex

Energy Considerations
So, momentum balance gives the Thrust Equation. What does an Energy Balance
give?
Start with a near-stagnant flow in the upstream plenum (“combustion chamber”, or

“nuclear heater” or “arc heated plenum”). The total specific enthalpy
1
htc = hc + υc2 ≅ hc may be different for different streamlines, due to combustion
2
“streaks:, arc constriction, etc., But along the flow expansion in the nozzle, ht is
conserved for each streamline. At the exit,
1 2
he + υe = hto (each streamtube)
2
or ( )
υe = 2 htc − he ≅ 2 ( hc − he )
For a well-expanded nozzle, with large area ratio, he → o by adiabatic expansion, and
υe tend to a max. υe MAX = 2 h tc . In any real, finite expansion, he ≠ o, so some of
htc is wasted as thermal energy in the exhaust. Define a nozzle efficiency.
htc − he he h
ηN = =1− ≅1− e
htc htc hc
γ −1
h T ⎛P ⎞ γ
For ideal gas, e = e = ⎜⎜ e ⎟⎟ . But, in any case,
hc Tc ⎝ Pc ⎠
υe2 2
υe = υe MAX ηN = ηN 2 htc (i.e., ηN = )
htc
Since Pe ≅ uniform, so is ηN , even when htc is not. Also, υe is non-uniform if htc is

(in proportion to htc ).
The Jet Power is the kinetic energy flow out of the nozzle
PJet =
1
2
( )
m htc − he = ηN htc m

Effect of Stagnation Enthalpy Non-uniformities
Consider a case where htc varies from streamtube ( dm ) to streamtube (but

Pe=const., so ηN = const.). Then
F = ∫∫ υ e dm + ( Pe − Pa ) Ae
For Pa = o (vacuum operation) and PaAe << F (large expansion), (or if Pe = Pa)
F ≅ ∫∫ υ e dm = 2 ηN ∫∫ htc dm (1)
and the input power is P = ∫∫ h tc dm (2)
⎧⎪P is minimum(For a given F, m)⎫⎪

It can be shown that ⎨ ⎬ if the flow is uniform
⎩⎪or F is maximum ( given P, m) ⎭⎪
( htc =const.). If it were, we would have
FUNIF . = 2 ηN m htc ; PUNIF = m htc
2
⎛ F ⎞ F2 F2
Eliminating htc , PUNIF = m ⎜ UNIF ⎟ = UNIF =
⎜ 2η m ⎟ 2 ηN m 2 ηN m
⎝ N ⎠
PUNIF
Define an “efficiency” ηUNIF = (for a given thrust)
PACTUAL
Now, express in general F by (1) and P by (2)
( )
2
ηUNIF =
2 ηN ∫∫ htc dm
2 ηN ( ∫∫ dm)( ∫∫ h tc dm )
Define “generalized vectors” u = 1 υ = htc in the space of the dm values.
(u i υ)
2
Then ηUNIF =
u2
i υ 2
≤1 ( = cos 2
)
θu iυ .

Equality applies only when υ is a constant, i.e., htc =const. This proves the “ansatz”.
50% of flow has htc = 0.5 htc
Example:
50% of flow has htc = 1.5 htc
2
⎛1 1 ⎞
⎜ 2 0.5 + 2 1.5 ⎟
ηUNIF = ⎝ ⎠ = 0.933 (6.7% energy loss due to nonunif.
⎛1 1
(1) ⎜ 2 0.5 + 2 1.5 ⎞⎟
⎝ ⎠
Important in arcjets, less in film-cooled chemical rockets.

16.512, Rocket Propulsion
Lecture 3: Ideal Nozzle Fluid Mechanics
Ideal Nozzle Flow with No Separation (1-D)
- Quasi 1-D (slender) approximation

- Ideal gas assumed
F = m ue + ( Pe − Pa ) Ae
F
CF ≡
Pc At
Optimum expansion: Pe = Pa
Ae
- For less , Pe > Pa , could derive more forward push by additional
At
expansion

Ae
- For more , Pe < Pa , and the extra pressure forces are a suction,
At
backwards
Compute m = ρ uA at sonic throat:
1 γ +1
⎛ 2 ⎞ γ −1 ⎛ 2 ⎞ ⎛ 2 ⎞ 2 (γ −1) Pc At R
m = ρc ⎜ ⎟ γ RgTc ⎜ ⎟ At = g⎜ ⎟ ; Rg =
⎝ γ + 1⎠ ⎝ γ + 1⎠ ⎝ γ + 1⎠ RgTc M
call Γ 2
3
RgTc Pc At
call c* = (“characteristic velocity”) → m =
Γ (γ ) c*
⎛P ⎞ A
Can express ue , Pe , Ae , etc in terms of either Me or ⎜⎜ e ⎟⎟ or e :
⎝ Pc ⎠ At
Pe 1
= γ
; γ +1
Pc 2 ( γ −1)
⎛ γ −1 2⎞ γ −1
⎜ 1 + 2 Me ⎟
⎝ ⎠
1+ 1
⎛ γ −1 2 ⎞2 γ −1
1+ Me
Ae ⎛ Pt ⎞ ⎛ ut ⎞ ⎛ Pt ⎞ 1 Tt 1 ⎜ 2 ⎟
= ⎜⎜ ⎟⎟ ⎜⎜ ⎟⎟ = ⎜⎜ ⎟⎟ = ⎜ ⎟
At ⎝ Pe ⎠ ⎝ ue ⎠ ⎝ Pe ⎠ Me Te Me ⎜ γ +1 ⎟⎟
⎜
⎝ 2 ⎠
γ
Pe ⎛ Te ⎞ γ −1
= ⎜⎜ ⎟⎟ ,
Pc ⎝ Tc ⎠
Te 1
and = ,
Tc γ −1
1+ Me2
2
ue2 γ M2 γ
Because c pTe + = c pTc → RgTe + e γ RTe = RT
2 γ −1 2 γ −1 g c

m ⎛ P − Pa ⎞ Ae u ⎛P P ⎞ Ae
CF = ue + ⎜⎜ e ⎟⎟ = *e + ⎜⎜ e − a ⎟⎟
Pc At ⎝ Pc ⎠ At c ⎝ Pc Pc ⎠ At
Tc
Me γ Rg
γ −1 γ +1
1+ Me2
ue 2 ⎛ 2 ⎞ 2 (γ −1) Me
= =γ⎜ ⎟
⎝ γ + 1⎠
*
c RgTc γ −1
1+ Me2
Γ 2
In vacuum,
( Pa = 0)
γ +1 γ 1
−
γ +1 ⎛ γ − 1 2 ⎞ 2 (γ −1) γ −1 −
2
⎜ 1+ Me ⎟
ue Pe Ae ⎛ 2 ⎞ 2 (γ −1) γ Me 1 ⎝ 2 ⎠
CFv = + =⎜ ⎟ + γ +1
⎝ γ + 1⎠
* Pc At Me
c γ −1
1+ Me2 ⎛ γ + 1 ⎞ 2 (γ −1)
2 Me ⎜ ⎟
⎝ 2 ⎠
1
γ +1 γ Me +
⎛ 2 ⎞ 2 (γ −1) Me
(CF )v =⎜ ⎟
⎝ γ + 1⎠ γ −1
1+ Me2
2
and otherwise,

⎛ A ⎞ Pa
CF = CF v − ⎜⎜ e ⎟⎟
⎝ At ⎠Me Pc
Note:
For Pe = Pa ,
γ +1
ue ⎛ 2 ⎞ 2 (γ −1) γ Me
(CF )Matched = =⎜ ⎟
⎝ γ + 1⎠
*
c γ −1
1+ Me2
2
γ +1
2 ⎛ 2 ⎞ 2 (γ −1)
For Pe = Pa = 0 (CF )Max,Vac =γ
γ −1
⎜ ⎟
⎝γ + 1⎠
Choice of Optimum Expansion For a Rocket Flying Through an Atmosphere

( Pa varying)
F
The thrust coefficient CF = was derived in class in the form
Pc At
Pa ⎛ Ae ⎞
CF = CFvac − ⎜⎜ ⎟⎟ (1)
Pc ⎝ At ⎠
γ +1
γ Me + 1
⎛ 2 ⎞ 2 (γ −1) Me
CFvac =⎜ ⎟ (2)
⎝ γ + 1⎠ γ −1
1+ Me2
2
and we also found
γ +1
⎛ γ −1 2 ⎞ 2 (γ −1)
1+ Me
Ae 1 ⎜ 2 ⎟
= ⎜ ⎟ (3)
At Me ⎜ γ +1 ⎟⎟
⎜
⎝ 2 ⎠
The thrust-derived velocity increment is
tb F tb CF
ΔVF = ∫0 m
dt = Pc At ∫0 m
dt (4)

where CF = CF ( t ) due only to the variation of Pa in (1), while m = m ( t ) because of
Ae
mass burnout. The quantities CFvac and depend on Me (or nozzle geometry), but
At
are time-invariant. Substituting (1), (2) and (3) into (4),
⎡ tb dt ⎛ Ae ⎞ tb Pa dt ⎤
ΔVF = Pc At ⎢CFvac
⎣⎢
∫0
−⎜ ⎟
m ⎜⎝ At ⎟⎠ ∫ 0
⎥
Pc m ⎦⎥
or
tbPa dt
ΔVF ∫0 Pc m Ae
= CFvac − (5)
tb dt tb dt At
Pc At ∫
0 m ∫0 m
We now make the approximation that the trajectory will change little when we vary
A
Me (and hence CFvac , e ). We can then regard the time integrals in (5) as fixed
At
quantities while we optimize Me . Define the non-dimensional variables
tbPa dt
ΔVF ∫
0 Pc m
v = ; p= (6)
tb dt tb dt
Pc At ∫0 m ∫0 m
so that (5) becomes
⎛A ⎞
v = CFvac ( Me ) − p ⎜⎜ e ⎟⎟ ( Me ) (7)
⎝ At ⎠
and we can now differentiate v w.r.t Me (holding p=const.)
⎛A ⎞
∂CFvac ∂⎜ e ⎟
∂v A
−p ⎝ ⎠ =0
t
= (8)
∂Me ∂Me ∂Me
γ +1
⎛ 2 ⎞ 2 (γ −1)
From (2) and (3), the factor ⎜ ⎟ appears in both terms of (8) and can be
⎝ γ + 1⎠
ignored. We then have

⎡ γ +1 ⎤
⎛ ⎞ ⎢ ⎛ 1 + γ − 1 M 2 ⎞ 2 (γ −1) ⎥
⎜ γ Me + 1 M ⎟ ⎢ ⎜⎝ e ⎟ ⎥
∂ ⎜ e ⎟= p ∂ 2 ⎠
⎢ ⎥
∂Me ⎜ γ −1 2 ⎟ ∂Me ⎢ M e ⎥
⎜ 1+ Me ⎟ ⎢ ⎥
⎝ 2 ⎠
⎣⎢ ⎦⎥
γ +1
γ −1 ⎛ γ − 1 2 ⎞2(γ −1) ⎡ γ −1 ⎤
γ− 1 2Me ⎜1 + 2 Me ⎟ 2Me
Me2⎛ 1 ⎞ 1 ⎢ γ +1 1 ⎥
− ⎜⎜ γ Me + ⎟⎟
2 = p⎝ ⎠
⎢
2 − ⎥
⎢ 2 ( γ − 1) 1 + γ − 1 M2 Me ⎥
Me ⎠ 2 ⎛ 3 Me
γ −1 2 ⎝ γ −1 2 ⎞ 2
1+ Me
⎜1 + 2 Me ⎟
e
2 ⎣⎢ 2 ⎥⎦
⎝ ⎠
3
⎛ γ −1 2⎞ 2 γ +1 1 γ
Multiply times ⎜1 + Me ⎟ , and note that + =
⎝ 2 ⎠ 2 ( γ − 1) 2 γ − 1
γ
⎛ γ −1 2⎞ γ −1
⎛ 1 ⎞⎛ γ −1 2⎞ γ −1 ⎜1 + 2 Me ⎟ ⎡ γ − 1 2 ⎞⎤
⎜ γ − 2 ⎟ ⎜1 +
⎜ M ⎟⎝ 2
Me ⎟ −
⎠ 2
γ Me2 + 1 = p ⎝ ( M 2
⎠
) ⎣
⎛
⎢( γ + 1) Me − ⎜1 +
2
⎝ 2
Me ⎟ ⎥
⎠⎦
⎝ e ⎠ e
Expand & simplify
γ
⎛ γ −1 2⎞ γ −1
γ ( γ − 1) 2 γ − 1 γ ( γ − 1) 2 γ − 1 ⎜1 + 2 Me ⎟
γ +
2
1
Me − 2 −
Me 2
−
2
Me −
2
= p ⎝
Me2
⎠
(M2
e −1)
1444444444442444444444443
1 Me2 − 1
1− =
Me2 Me2
Me2 − 1
Cancel the factor ( Me = 1 is clearly not an optimum!)
Me2
γ γ −1
⎛ γ −1 2⎞
1 = p ⎜1 + Me ⎟
⎝ 2 ⎠
or

γ −1 ⎡ γ −1 ⎤
γ −1 ⎛1⎞ γ 2 ⎢⎛ 1 ⎞ γ ⎥
1+ Me2 = ⎜ ⎟ Me = ⎢ ⎜ ⎟ − 1⎥ (9)
2 ⎝ p⎠ OPT γ −1 ⎝ p⎠
⎢⎣ ⎥⎦
Notice that the exit pressure is given by
Pe 1
= γ
(10)
Pc
⎛ γ −1 2⎞ γ −1
⎜ 1 + 2 Me ⎟
⎝ ⎠
and so the optimum exit pressure turn out to be
⎛ Pe ⎞
⎜⎜ ⎟⎟ =p (11)
⎝ Pc ⎠OPT
Pao
However, if p < 0.4 , this would imply Pe < 0.4Pao , and there would be flow
Pc
separation at the highest Pa (on the ground). To avoid this, the optimality condition
must be amended to
⎛ Pe ⎞ ⎪⎧ Pa ⎪⎫
⎜⎜ ⎟⎟ = Greater of ⎨ ρ , 0.4 o ⎬ (12)
⎝ Pc ⎠OPT ⎪⎩ Pc ⎪⎭
with a similar expression for Me :
⎧ ⎡ γ −1 ⎤⎫
⎪ ⎡ γ −1 ⎤
⎪ 2 ⎢⎛ 1 ⎞ γ ⎥ 2 ⎢⎛ Pc ⎞ γ ⎥ ⎪⎪
Me = Least of ⎨ ⎢ ⎜ ⎟ − 1⎥ , ⎢⎜ 2.5 ⎟ − 1⎥ ⎬ (13)
OPT
⎪ γ − 1 ⎢⎝ p ⎠ γ − 1 ⎢⎝ ⎜ Pao ⎟ ⎥⎪
⎥⎦ ⎠
⎪⎩ ⎣ ⎣⎢ ⎦⎥ ⎪⎭
The limiting condition in which the whole burn occurs at Pa o is simple.
We then obtain
tb Pao dt
∫0 Pc m Pao
p= tb dt
= (14)
Pc
∫ 0 m

and the optimality condition (12) yields ( Pe )OPT = Pao , i.e., the nozzle should be
pressure-matched, as expected.
Pao
As more and more of the burn shifts to higher altitudes, p decreases from . As
Pc
Pao
long as it still remains above 0.4 , equation (11) gives some intermediate
Pc
Pao
optimum design, and if p drops below 0.4 , the nozzle should be designed to be
Pc
on the verge of separation on the ground.
Nozzle Flow Separation Effects
Rule of thumb (to be explored later):
Flow separates at the point in the nozzle where
P 0.4Pa (Summerfield criterion)
So, if Pe > 0.4Pa (even if Pe < Pa ), no separation
After separation, roughly parallel flow, at P = Pa (no strong p gradients in “dead

water” region to turn flow).
So zero thrust contribution Performance with separation at that of a nozzle

with exit pressure Pe' = 0.4Pa
So,
Pe (full nozzle)
(a) Pa < ,
0.4
Pa Ae
CF = CFvac −
Po At
f ( Me ) g ( Me )

Pe (full nozzle)
(b) Pa > ,
0.4
⎪⎪ e (
⎧M ' = M P ' = 0.4 P
e a )
calculate ⎨ A' A'
⎪ e = e Me'
⎪⎩ At At
( )
Pa Ae'
( )
then CF = CFvac Me' −
Po At

Lecture 4-5: Nozzle Design: Method of Characteristics
The Method of Characteristics (Ideal Gas)

(Ref. Phillip Thompson Compressive Fluid Dynamics, McGraw Hill, 1972, Ch. 9)
2-D or axisymmetric
ur
Homentropic as well as isentropic → ∇×u = o
ur
plus ∇ (ρ u ) = 0
0
ur ur 1 ⎛ u2 ⎞ ur ur 1
and u ∇u + ∇p = o ⇒ ∇ ⎜ ⎟ + ω × u + ∇p = 0
ρ ⎝ 2 ⎠ ρ
Use intrinsic co-ordinates
r ∂u 1 ∂p
Eq. of motion along s : u + =o (1)
∂s ρ ∂s
r u2 1 ∂p ∂u2 2
Eq. of motion along n : = =−
R ρ ∂n ∂n
1 ∂ϑ ∂ϑ ∂u ∂u ∂ϑ
Now ≡ u2 = −u +u =o (2)
R ∂s ∂s ∂n ∂n ∂s
(good no p in it)
ur
(can also get this from ∇ × u = o )
∂
Continuity: ( ρ u 2πr δn ) = o
∂s

Prof. M. Martinez-Sanchez Page 1 of 20
1 ∂ρ 1 ∂u 1 ∂r 1 ∂δn
+ + + =o
ρ ∂s u ∂s r ∂s ∂n ∂s
∂r 1 ∂ϑ
( δs ) sin ϑ = ( ∂s ) sin ϑ
∂s r ∂n
∂δn ⎛ ∂ϑ ⎞
ds = ⎜ − δn ⎟ δs
∂s ⎝ ∂n ⎠
1 ∂ρ 1 ∂u ∂ϑ sin ϑ
+ − =−
ρ ∂s u ∂s ∂n r
⎛ 2 ⎛ ∂p ⎞ ⎞
Homentropic: dp = c 2 d ρ ⎜⎜ c = ⎜ ⎟ ⎟⎟
⎝ ⎝ ∂ρ ⎠s ⎠

1 ∂p 1 ∂u ∂ϑ sin ϑ
so + − =−
ρ c ∂s u ∂s ∂n
2
r
1 ∂p ∂u ∂p
and, from s eq. of motion, = −u so you can eliminate :
ρ ∂s ∂s ∂s
u ∂u 1 ∂u ∂ϑ sin ϑ 1⎛ u2 ⎞ ∂u ∂ϑ sin ϑ
− + − =− ⎜ − 1 + ⎟ + =
c 2 ∂s u ∂s ∂n r u⎝ c 2 ⎠ ∂s ∂n r
2
M -1
M 2 − 1 ∂u ∂ϑ sin ϑ
+ = (4)
u ∂s ∂n r
1 1 1
Introduce the Mach angle μ = sin−1 = tan−1 tan μ =
M 2
M −1 M2 − 1
M 2 − 1 ∂u ∂ϑ
Then (2) tan μ + =o
u ∂n ∂s
M 2 − 1 ∂u ∂ϑ tan μ sin ϑ
And (4) + tan μ =
u ∂s ∂n r
Introduce the Prandtl-Meyer function ω (M) by
du
dω = M2 − 1 (to be integrated later)
u

∂ω ∂ϑ
tan μ + =o
∂n ∂s
then
∂ω ∂ϑ tan μ sin ϑ
+ tan μ =
∂s ∂n r
add and subtract to obtain the “Characteristics form” (single differential operator per
equation)
⎛ ∂ ∂ ⎞ tan μ sin ϑ ⎛ ∂ ∂ ⎞ sin μ sin ϑ

⎜ ∂s + tan μ ∂n ⎟ ( ω + ϑ ) = r ⎜ cos μ ∂s + sin μ ∂n ⎟ ( ω + ϑ ) = r
⎝ ⎠ ⎝ ⎠
⇒
⎛ ∂ ∂ ⎞ tan μ sin ϑ ⎛ ∂ ∂ ⎞ sin μ sin ϑ
⎜ ∂s − tan μ ∂n ⎟ ( ϑ − ω) r ⎜ cos μ ∂s − sin μ ∂n ⎟ ( ϑ − ω) = − r
⎝ ⎠ ⎝ ⎠
r + ⎧cos μ ⎫ r − ⎧ cos μ ⎫
In (s, n) coordinates, 1 = ⎨ ⎬ 1 =⎨ ⎬
⎩ sin μ ⎭ ⎩− sin μ ⎭
∂ ∂ r+ ∂
cos μ + sin μ =1 ∇ =
∂s ∂n ∂m+
so
∂ ∂ r− ∂
cos μ − sin μ =1 ∇ =
∂s ∂n ∂m−
(m+, m- are lengths along characteristics)
⎧ ∂ sin μ sin ϑ ⎫
⎪ ∂m+ ( ϑ + ω) = + r ⎪
⎛ m− inclined ϑ + μ ⎞ ⎪⎪ ⎪⎪
⎜⎜ + ⎟⎟ ⎨ ⎬
⎝ m inclined ϑ − μ ⎠ ⎪ ∂ sin μ sin ϑ ⎪
− (
⎪ ϑ − ω) = − ⎪
⎩⎪ ∂m r ⎭⎪
For 2-D, r → ∞ ⇒ ϑ + ω =const. along m+ ≡ I+ (inclined ϑ − μ )

ϑ − ω =const. along m- ≡ I- (inclined ϑ + μ )

2-D Simple Regions
Consider a uniform region; the flow from it enters

some disturbed region, like a wall turning. One of
the m families originate in the uniform region (m+
in example) and carries a constant invariant, e.g.
ϑ + ω = ϑo + ωo everywhere downstream. Along one
of the other characteristics (m- here), we carry a
constant (which varies from ch. To ch. of that
family); we can evaluate it at the wall, for instance
ϑ − ω = ϑw − ωw along each m- line and

ωw = ϑo + ωo − ϑw so ϑ − ω = 2ϑw − ( ϑo + ωo )
2ϑ = ϑo + ωo + ϑw − ωw = 2ϑw along each m− → ϑ = ϑw

then
2ω = ϑo + ωo − ( ϑw − ωw ) along each m −
→ ω = ϑo + ωo − ϑw = ωw
so ϑ and ω (and hence M, μ) are constant along each m-, and these m- lines are
straight (const. ϑ + μ ).
This is a Simple Region (one of the invariants is constant). The turning of the flow is
dictated by how ϑ changes on a m+ line, as different m- lines are crossed. Since on a
m+ we have ϑ + ω = ϑo + ωo , changes of ϑ are equal and opposite to those in ω (ω
increases as M does in the expansion, so ϑ decreases (increase negatively) at the
same rate.) So, we can interpret ω as the magnitude of the isentropic flow turning in
a simple region, i.e., when nothing varies along one characteristic family.
Calculation of ω (M)
du ⎛ dM 1 dT ⎞
dω = M2 − 1 = M2 − 1 ⎜ +
u ⎝ M 2 T ⎟⎠
but

⎛ γ −1 2 ⎞
M
M2 − 1 ⎜ 2 ⎟ M2 − 1 dM
dω = ⎜1 − ⎟ dM =
M ⎜⎜ γ −1 2 ⎟⎟ M γ −1 2
1+ M 1+ M
⎝ 2 ⎠ 2
Integrates (with ω =0 at M=1) to
M2 − 1 ⎛ γ +1⎞
ω = K tan−1 − tan−1 M 2 − 1 ⎜⎜ K = ⎟
K ⎝ γ − 1 ⎟⎠
π π π π⎛ γ +1 ⎞
For M → ∞ ω→K − = ( K − 1) = ⎜ − 1⎟
2 2 2 ⎜
2 ⎝ γ −1 ⎟
⎠
γ 1.2 1.25 1.4 5/3

ω (M → ∞ ) 208o 180o 130.5o 90o
(Max. turning from M=1)
So rocket exhaust ( γ 1.2 − 1.3 ) can turn backwards at a sonic nozzle exit to
vacuum)
(but very long density, long mfp, so continuum

approach fails at some point in the expansion,
molecules then continue in straight line).
Actually, one starts from some high Me>1, so

the actual turning is through ω∞ − ω (Me ) , not
ω∞ , even in a vacuum:

Example of Application: Ideal 2-D plug Nozzle at Design Condition
Along m− , ϑ − ω = ϑa − ωa
Simple region (Prandtl-Meyer fan

centered at lip L)
In particular, at inlet M=1, so

ϑthroat = −ωa
r dϕ 1
tan μ = =
dr M2 − 1
π
From geometry, μ − ϑ + ϕ = ϑt
2
π
Sub. ϑ = ω − ωa μ − ω + ωa + ϕ = + ωa
2
π
ϑt = −ωa ϕ= +ω−μ
2
π M2 − 1 1 ⎛ γ +1⎞
ϕ= + K tan−1 − tan−1 M 2 − 1 − tan−1 ⎜⎜ K = ⎟
2 K 2
M −1 ⎝ γ − 1 ⎠⎟
π
−
2

M2 − 1
ϕ = K tan−1
K
ϕ M2 − 1 1
tan = =
K K K tan μ
ϕ ⎛ ϕ⎞
sin d ⎜ cos ⎟
r dϕ 1 dr K dϕ = −K 2 ⎝ K ⎠
⇒ =
ϕ
=K
ϕ ϕ
dr r
K tan cos cos
K K K
const.
r = K2
⎛ ϕ⎞
⎜ cos K ⎟
⎝ ⎠
ht
For ϕ = 0,r = ht (throat height) r = K2
⎛ ϕ⎞
⎜ cos K ⎟
⎝ ⎠
Ma2 − 1
In particular, at end of expansion M = Ma → ϕa = K tan−1 , then
K
ht ra
ra = and ha = ra sin μa =
⎛ ϕa ⎞
K2
Ma
⎜ cos K ⎟
⎝ ⎠
1
xa = ra cos μa = ra 1 −
Ma2

Numerical Application
γ ⎡ γ −1
⎤
Po ⎛ γ − 1 2 ⎞ γ −1 2 ⎢⎛ Po ⎞ γ ⎥
= 100 = ⎜ 1 + Ma ⎟ ⇒ Ma = ⎜ ⎟ − 1⎥
Pa ⎝ 2 ⎠ γ − 1 ⎢⎝ Pa ⎠
⎣⎢ ⎥⎦
2 ⎡ 0.3
⎤
Take γ = 1.3 , Ma = ⎢100 − 1⎥ = 3.554
1.3
0.3 ⎣ ⎦
⎛ M2 − 1 ⎞ γ +1 2.3
Then ωa = K tan−1 ⎜ ⎟ − tan
−1
M2 − 1 K= = = 2.769
⎝ K ⎠ γ −1 0.3
ωa = 67.35o ⇒ ϑt = −67.35o
⎛ M2 − 1 ⎞
Also ρa = K tan−1 ⎜ ⎟ = 141.01
o
⎝ K ⎠
ra 1 1 ra
and so = = = 34.41
ht ⎛ ρa ⎞
K2
⎛ 141.01 ⎞
2.7692
ht
⎜ cos K ⎟ ⎜ cos 2.769 ⎟
⎝ ⎠ ⎝ ⎠
ra
From the geometry, ha=ra sin μa =
Ma
ha 1 ha
= 34.41 =9.684
ht 3.554 ht

2
xa ⎛ 1 ⎞ xa
and = ra cos μa = 34.41 1 − ⎜ ⎟ =33.02
ht ⎝ 3.554 ⎠ ht
Very long and pointy, should be truncated.
Non-Simple Regions. When the characteristics of both families intersect some

upstream disturbance, they affect each other’s invariant, and characteristics are no
longer straight, and no longer carry constant flow properties (except for their own
invariant)
0-1-4 is a simple region
0-4-41 is uniform
1-4-10 is non-simple
4-41--11 is a simple region
Then we need to calculate in a step-by-step manner, carrying to each point the two
invariants I+, I- from neighboring upstream points, along the m+, m- lines from them
to us. After this is done, we know the new segments of m+, m- from our point (slopes
ϑ + μ , ϑ − μ ), so we can extend the grid as we go. Notice the flowfield properties can
be found first everywhere and only then we need to come back and place the points
geometrically.
Example: Design a 2-D ideal nozzle to expand from near sonic conditions (M0 = 1.1)
to Me = 3. Use only 4 characteristics. Use a corner expansion as a starter,
γ =1.25

2.25
K = =3
0.25
1.21 − 1
ωo = 3 tan−1 − tan−1 1.21 − 1 = 1.435o ; ϑo = 0
3
9 −1
ωe = 3 tan−1 − tan−1 9 − 1 = 59.413o ; ϑe = 0
3
At inlet: I − = ϑ − ω = ϑo − ωo = −1.435o (also at 4)
At exit: I + = ϑ + ω = ϑe + ωe = 59.413o (also at 4,

7, 9, 10)
2ϑ4 = −1.435 + 59.413 = 57.978 ϑ4 = 28.989o

At 4, then, 2ω4 = 59.413 + 1.435 = 60.848 ω4 = 30.424o

Select: I1+ = 1.44o I2+ = 20.76o I3+ = 40.09o I4+ = 59.41o
Point M I+ = ϑ + ω ()
o
I- = ϑ − ω ()
o
ω ()
o
ϑ ()
o
μ ()
o
ϑ+μ ()
o
ϑ−μ ()
o
1 1.1 +1.44 -1.44 1.44 0 65.38 65.38 -65.38o

2 1.439 20.76 -1.44 11.10 9.66 44.02 53.68 -34.36
3 1.726 40.09 -1.44 20.77 19.33 35.41 54.74 -16.08
4 (and 4w) 2.013 59.41 -1.44 30.43 28.99 29.79 58.78 -0.80
5 1.726 20.76 -20.76 20.76 0 35.41 35.41 -35.41
6 2.013 40.09 -20.76 30.43 9.67 29.79 39.46 -20.12
7 (and 7w) 2.315 59.41 -20.76 40.09 19.33 25.59 44.92 6.26
8 2.315 40.09 -40.09 40.09 0 25.59 25.59 -25.59
9 (and 9w) 2.641 59.41 -40.09 49.75 9.66 22.25 31.91 -12.59
10 (and 11) .3 59.41 -59.41 59.41 0 19.47 19.47 -19.47
Notes
Points 1-4: same ϑ − ω
Point 5: Here ϑ = 0 (a boundary condition) and ϑ + ω = 20.76
Point 6-7: same ϑ − ω
Point 8: Here ϑ = 0 and ϑ + ω = 40.09
Point 9: same ϑ − ω
Point 10: Here ϑ = 0 and ϑ − ω = 59.41
Note the very shallow angles of the m+ lines (from 4, 7, 9, 10) which will put point
10 far to the right, and point 11 (at slope (m+) of 19.47o even farther.
Locating the points geometrically
(xc − x a ) tan α + ( x c − x b ) tan β = y a − y b
ya − yb + xa tan α + xb tan β
xc =
tan α + tan β

yc − yb = ( xc − xb ) tan β = tan β
( ya − yb ) + ( xa − xb ) tan α
tan α + tan β
yb tan α + ya tan β + ( xa − xb ) tan α tan β

yc =
tan α + tan β
put xa, ya into 1, 2
tan α into 3 Run 2 → xc (st. in 7)
xb, yb into 4.5 Run 3 → yc (st. in 7)
tan β into 6
Finding the (x,y)’s is very laborious. Accuracy can be increased by averaging

together the angles at the ends of each segment, which can be done because those
α + αc β + βc
angles come from the first pass. For instance, α = a , β = b .
2 2

Boundary with Prescribed Pressure
Pc
P= r
,
⎛ r −1 2⎞ r −1
⎜1 + 2 M ⎟
⎝ ⎠
and M = M ( w ) ,
so P = P ( w ) (given Po )
or w = w (P )
So, if P is fixed on a boundary (contact surface), we can assign w there (just as we

assign θ on a solid boundary)
From known point a,
θ − w = θa − wa
So θ = w (P ) + θa − wa
and this determines the slope of the boundary, and that of the “reflected” m+
θ + w = 2w (P ) + θa − wa
More General Contact Surface Condition
If the outside fluid is also supersonic, we must solve on both sides of the contact
surface, making sure P and θ are common at each boundary point
16.512, Rocket Propulsion Lecture 4-5

P1 ( w1 ) = P2 ( w2 )
θ1 = θ2
From m1+ know I1+ = θ + w1+ (P )
From m2− know I2− = θ − w2− (P )
So, I1+ − I2− = w1+ (P ) + w2− (P )
⇒ Solve for P ⇒ w1 , w2
known now
I1+ + I2− w1+ (P ) + w2− (P )
and =θ+
2 2
Modifications for Axisymmetric Conditions
∂ sin μ sin θ ⎫
+
( θ + w) =r ⎪
∂m ⎪
⎬
∂ sin μ sin θ ⎪
( θ − w) = − ⎪⎭
∂m− r

(1) Calculate ( x, r )c from the angle ( μ − θ )a , ( μ + θ )b
x a tan α + xb tan β + ra − rb
( x − xa ) tan α + ( x − xb ) tan β = ra − rb ⇒x−
tan α + tan β
and r − rb = ( x − xb ) tan β
x − xa
(2) Δm+ =
cos α
x − xb
Δm− =
cos β
(3) Advance invariants θ + w , θ − w based on μ, θ at a, b:
sin μa sin θa ⎫
θ + w = ( θ + w )a + Δm+ ⎪
ra ⎪
⎬ (r at c, computed in (1))
sin μb sin θb
θ − w = ( θ − w )b − Δm− ⎪
rb ⎪
⎭
(4) M = M ( w ) , μ = μ (M)
(5) Iterate from (1) with α, β averaged between (a, b) and c:
( μ − θ )a + ( μ − θ )c
α→
2
( μ + θ )b + ( μ + θ )c
β→
2

ra + rb⎫
r→ ⎪
2 ⎪
⎪
μa + μb ⎪⎪
μ→ ⎬ better
2 ⎪
⎪
⎪
θ + θb ⎪
θ→ a
2 ⎪⎭
(6) Continue to new point.
So, computation of ( θ,M) is now coupled to that of ( x, r ) , whereas in 2-D ( θ,M)

can be found first. But the actual amount of computation is not much more (only
the iteration stops).
On the axis, r → 0 , but θ = 0 , so
1. x = xa + ra tan α
ra
2. Δm+ =
sin α
sin μa sin θa
3. θc = 0; wc = ( θ + w )a +
ra
ra , θa ,
4. Mc = M ( wc ) , μc = μ (Mc )
( μ − θ )a + μ c
5. α → , go to (1) (once)
2
6. Continue
Extension to Cases with Weak Shocks
ΔPshock
Since the entropy jump in a shock increases only as the cube of , the
ρu20
isentropic assumption can be approximately extended when characteristics of our
family show some mild convergence (in principle, that is always indicative of shock
formation, because they carry conflicting information). When is the convergence too
strong? Since characteristics are discretized, for weak convergence the ones of our
family will converge, but not cross, and as long as they don’t, it should be OK. Of
course, with finer resolution they will cross, but the loss of accuracy in ignoring that
is of the same order as that increased by the coarse discretization in the first place.
This allows us to calculate off-design nozzle performance, like an overexpanded plug

nozzle.

2-D Spike Nozzle with Pa < Padesign
Pressure forces from hot-gas bathed surfaces are the same as at design. Net thrust
is increased because the Pa contribution PaA e is reduced:
( )
F = Fdes. + Pades. − Pa A e
(just as for a bell)
and so Fvac = Fdes. + Pades. A e
“Practical” Nozzle Designs
Ideal nozzle are too long, last portion has small wall angle, so small thrust
contribution. With small ?, maybe negative contribution. So, options:
(a) Constrain length, ask for contour that gives highest thrust/given L.
Methods of calculus of variations (Raw nozzle, Ref : “ Exhaust Nozzle
Contour from Optimum Thrust”, Jet Propulsion 28 (June 1958): 377-382.
Exit flow non-parallel, non-uniform, computationally high.
(b) Ad-hoc method (widely used) is to truncate contract an ideal nozzle.
(1) Design wall contour for desired Pe Pc

(2) If longer than desired L truncate it to some intermediate area ratio.
(3) Contract this truncated nozzle to desired length
x L
x'a = xb' c = xb' desired
xd L ideal
(4) Translate profile to right metal kink at P smoothes out.

Gives “adequate” performance but less than ? nozzle. Not really
justified. (Hoffman, J. D. J. of Propulsion 3 (March-April 1987): 150-
156.

XRS-2200 Engine Data
Thrust, lbf
At Sea Level 206,500

In Vacuum 268,000
Specific Impulse, sec.
At Sea Level 339

In Vacuum 439
Propellants O2, H2
Mixture Ratio (O/F) 5.5
Chamber Pressure, psia 857
Cycle Gas Generator
Area Ratio 58
Throttling, Percent Thrust 40-119
Differential Throttling +/- 15%
Dimensions, Inches
Forward End 133 high x

88 wide
Aft End 46 high x

88 wide
Forward to Aft 79

Lecture 6: Heat Conduction: Thermal Stresses
Effect of Solid or Liquid Particles in Nozzle Flow
An issue in highly aluminized solid rocket motors.
3
2Al + O2 → Al2O3
2
m.p. 2072 C , b.p. 2980 C
In modern formulations, with ∼ 20% Al by mass, the Al2O3 mass fraction of the
exhaust can be 35-40%. This material does not expand, so there must be a loss in
exit velocity, hence in Isp.
i i
Assume mass flows mg (gas) ms (solids), non-converting.
The momentum equation is
i i
mg dug + ms dus + Adp = 0
Call ρs the (mass of solids)/(volume) (not the density of the solid, theory)
ρg ug dug + ρs us dus + dp = 0
Define a mass flux function
i
ms ρsus
x= =
i i ρgug + ρsus
mg + ms
⎛ x ⎞
⇒ ρgug ⎜ dug + dus ⎟ + dp = 0
⎝ 1 − x ⎠
dp x
ugdug = − − ugdus
ρg 1−x
The energy equation is similarly,
(1 − x ) ( cpgdTg + ugdug ) + x ( csdTs + usdus ) = 0

Substitute here ugdug from above:
dp x x
cpgdTg −
ρg
−
1−x
ugdus +
1−x
( csdTs + usdus ) = 0
dp x ⎡
ρg
= cpgdTg +
1−x ⎣
( )
cs dTs + us − ug dus ⎤
⎦
γ
dp P ⎛ T ⎞ γ −1
with no particles (x=0), this gives R g T = cp dT → = ⎜⎜ ⎟
P P0 ⎝ T0 ⎟⎠
With particles, we need to know the history of the velocity slip us − ug and of the
temperature slip Ts − Tg . This is a difficult problem, requiring detailed modeling of
the motion and heating/cooling of the particle. But we can look at the extreme cases
easily.
(a) Very Small Particles → good contact. For sub-micro particles (not a bad
representation of reality), we can say that
us ug = u ,
Ts Tg = T . Then
dp ⎛ x ⎞
= ⎜ cpg + cs ⎟ dT
ρg ⎝ 1 − x ⎠
Note that the mean specific heat ( cpg and cs are per unit mass) is
cp = (1 − x ) cpg + xcs ⎫⎪
( )
⎬ R g = c p − c v = (1 − x ) cpg − cvg = (1 − x ) R g
and also c v = (1 − x ) cvg + xcs ⎪⎭
dp cp
So that = dT
ρg 1−x
dp cp dp cp dT
RgT = dT → =
P 1−x P Rg T
cp
and defining an effective γ by the usual γ = ,
cv
γ
P ⎛ T ⎞ γ −1
= ⎜⎜ ⎟
P0 ⎝ T0 ⎟⎠

The equation of motion is now
( ρg + ρs ) udu + dp = 0
Or
ρg
udu + dp = 0
1−x
ρg P P
= =
1−x R g T (1 − x ) R g T
P
udu + dp = 0
RgT
From the two boxed equations we see that everything from here can proceed
as if the gas were simple, but with molecular mass
Mg
M=
1−x
(or R g = (1 − x ) R g ),
and with
c p = (1 − x ) cpg + x cs .
For example,
⎡ γ −1 ⎤
γ ⎢ ⎛P ⎞ γ ⎥
ue = 2 R g Tc ⎢1 − ⎜⎜ e ⎟⎟ ⎥ Tc , Pc in chamber etc.
γ −1 ⎢ ⎝ Pc ⎠ ⎥
⎣ ⎦
For sensitivity analysis it may be of interest to linearize this for x 1 . The

algebra is tedious, but one gets,
ue 1 ⎪⎧ ⎡
≅ 1 − × ⎨1 − c ⎢1 +
(1 − η0 ) ln (1 − η0 ) ⎤ ⎪⎫
⎥⎬
ue0 2 ⎪ ⎢⎣ η0 ⎥⎦ ⎪⎭
⎩

γ −1
cps ⎛P ⎞ γ
with c = , η0 = 1 − ⎜⎜ e ⎟⎟
cpg ⎝ Pc ⎠
and, of course,
⎡ γ −1 ⎤
γ ⎢ ⎛P ⎞ γ ⎥
ue0 = 2 R g Tc ⎢1 − ⎜⎜ e ⎟⎟ ⎥
γ −1 ⎢ ⎝ Pc ⎠ ⎥
⎣ ⎦
We see from this that if c < 1 ( cps < cpg , which is common), then ue < ue0 (and
∼ ∼
vice-versa).
For a numerical example, look at Problem 2 (attached)
(b) Very Large Particles Hard to quantify, but probably for diameter > 100 µm or
∼
so, the particles have too much inertia (and thermal inertia) to follow the gas
acceleration and cooling. We then have
dus dug ; Ts Tc ( ≅ Tg at chamber)
or dus 0; dTs 0
dp
Returning to the equation, it now looks as if there were no particles:
ρg
dp
= cpgdTg
ρg
(i.e. , particles just do not participate in the dynamics or in the thermal

γ γ −1
P ⎛ T ⎞
balances). So, we still have =⎜ ⎟ . This does not mean zero
P0 ⎜⎝ T0 ⎟⎠
performance effect, though. We do not get the full gas exit velocity
⎡ γ −1 ⎤
γ ⎢ ⎛ Pe ⎞ γ ⎥
ue = 2 R g Tc ⎢1 − ⎜⎜ ⎟⎟ ⎥
γ −1 ⎝ Pc ⎠
⎢⎣ ⎥⎦
but the particulates do not contribute to thrust, because they exit at

us ue :
i i
mg ue + ms us
g Isp = i i
= (1 − x ) g Isp0
mg + ms

This is actually more loss than in the small particle case (about twice as much,
depending on c).
From the example, this is a serious loss in solid rockets.
Criterion for Slip
4 2
πRp3ρs
3
2 2
dup 2 Rp ρs dup 2 ρsRp
mp
dt
(
= 6πµgRp ug − up ) 9 µg dt
= ug − up τR =
9 µg
dup ug − up
= call ug − up = s up = ug − s
dt τR
dug ds s ds s dug
− = + =
dt dt τR dt τR dt
t
dug −
τR
Say τR and = ag are constant → s ag τR + C e
dt
s ( 0 ) = 0 → C = −ag τR
⎛ ε2 ⎞
1 − ⎜1 − ε + ....⎟
⎜ 2 ⎟
= ⎝ ⎠ = 1 − ε + ...
ε 2
⎛ −
t ⎞
⎜
s = ag τR 1 − e τR ⎟
⎜ ⎟
⎝ ⎠
and ug = ag t
1 t
t τR 1− ...
2 τR
⎛ −
t ⎞
s τR ⎜ τR ⎟
= 1−e
ug t ⎜ ⎟
⎝ ⎠
1
t τR
t τR

So, small slip for t τR
L
τR
u
2
L 2 ρs Rp
ug 9 µg
9 µgL
Rp
2 ρs ug
Say
µg ∼ 3 × 10−5 Kg / m / s
L ∼ 0.3 m 3 × 10−5 × 0.3
Rp 4.5 = 0.9 × 10−11 = 3 × 10−6 m = 3µm
ρs ∼ 3 × 103 Kg / m3 3 × 103 × 1.5 × 103
ug ∼ 1.5 × 103 m / s
Rp 3µm → no slip
So,
Rp 3µm → full lag

Problems
Problem 2
As noted in class, the effect of carrying a mass fraction x of fine solid particles in the
expanding gas in a rocket nozzle can be accounted for by using an average specific
heat ratio
(1 − x ) cpg + xcs
γ =
(1 − x ) cvg + xcs
and an average molecular mass
Mg
M =
1− x
For Al2O3 the high temperature specific heat is cs = 1260 J/Kg/K.

Consider a solid rocket with γ = 1.17 (1.25) , Pe Pc = 0.01, M g = 18 g / mol.
For a 20% aluminum loading in the propellant, x is of the order of 37%.
Calculate the matched specific impulse of the rocket and compare to what it would
be for the same Tc = 3300 K , but with no particles.

Problem 2
Specific heat of clean gas

1.25
r R 1.17 8.314
cpg = = = 3180 J / Kg / K
r − 1 M 0.17 0.018 2309
0.25
2309
cpg 3180 1.25
cvg = = = 2718 J / Kg / K
r 1.17
1845
The specific heat of the solid (or liquid) Al2O3 is cs = 1260 J / Kg / K.
The average specific heat ratio is then
2309
(1 − x ) cpg + x cs (1 − 0.37) × 3180 + 0.37 × 1260
r= = = 1.1336
(1 − x ) cvg + x cs (1 − 0.37) × 2718 + 0.37 × 1260 1.1795
1845
And the average molecular mass ( Ms ∞ ) is
Mg 18
M= = = 28.57 g / mol
1−x 1 − 0.37
Pe
The exit speed for = 0.01 and Tc = 3300 K is then
Pc
⎡ γ −1 ⎤
γ ⎢ Q ⎛P ⎞ γ ⎥
ue = 2 Tc ⎢1 − ⎜⎜ e ⎟⎟ ⎥ = 2613 m / sec
γ −1 M ⎢ ⎝ Pc ⎠ ⎥ 2521
⎣ ⎦
NOTE: Alternatively, and easier to do, you can use
⎡ γ −1 ⎤
⎢ ⎛P ⎞ γ ⎥
ue = 2 Cp Tc ⎢1 − ⎜⎜ e ⎟⎟ ⎥
⎢ ⎝ Pc ⎠ ⎥
⎣ ⎦
with C p = (1 − 0 .3 7 ) 3 1 8 0 + 0 .3 7 × 1 2 6 0 = 2 4 6 9 J/Kg/K
(so M is not really needed)

As a check,
γ R 1.1336 8.314
= = 2469 J/Kg/K
γ −1 M 0.1336 0.02857
as it should.
Under pressure-matched conditions, there is no exit pressure contribution to thrust

or Isp , and hence
2613
Isp = = 266.3 sec
9.81
257.3 s
Without particulate but with the same Pc , Pe and Tc , we would obtain
⎡ γ −1 ⎤
2γ Q ⎢ ⎛ Pe ⎞ γ ⎥
ue0 = Tc ⎢1 − ⎜⎜ ⎟⎟ ⎥ = 3199 m / sec
γ − 1 Mg ⎢ ⎝ Pc ⎠ ⎥ 3029
⎣ ⎦
3199
and Isp0 = = 326.1 sec
9.81 309.1
266.3 ⎞ 16.8%
⎛
There is therefore a loss of ⎜1 − × 100 = 18.3% in Igp
⎝ 326.1 ⎟⎠
257.3
1−
309.1
It is interesting to test the accuracy of the linear approximation given in class for
small x:
γ −1
ue x⎡
1 − ⎢1 −
c ⎛ (1 − η0 ) ln (1 − η0 ) ⎞ ⎥⎤ ; η = 1 − ⎜⎛ Pe ⎞
⎟
γ
⎜1 + ⎟
ue0 2⎢ cpg ⎜ η0 ⎟⎥ 0
⎜P ⎟
⎣ ⎝ ⎠⎦ ⎝ c ⎠
ue
We find η0 = 0.4878 , and then = 0.837 (16.3% loss)
ue0
(not too different, despite large x)

NOTE: Alternatively, and easier to do, you can use
⎡ γ −1 ⎤
⎢ ⎛P ⎞ γ ⎥
ue = 2 Cp Tc ⎢1 − ⎜⎜ e ⎟⎟ ⎥
⎢ ⎝ Pc ⎠ ⎥
⎣ ⎦
with C p = (1 − 0 .3 7 ) 3 1 8 0 + 0 .3 7 × 1 2 6 0 = 2 4 6 9 J/Kg/K
(So M is not really needed)
γ R 1.1336 8.314
As a check, = = 2469 J/Kg/K as it should.
γ −1 M 0.1336 0.02857

Lecture 7: Convective Heat Transfer: Reynolds Analogy
Heat Transfer in Rocket Nozzles
General
Heat transfer to walls can affect a rocket in at least two ways:
(a) Reducing the performance. This tends to be a 1-3% effect on Isp only, and is
therefore secondary.
(b) Creating great difficulties in the design of hot-side structures that have to
survive heat fluxes in the 107 − 108 w / m2 range.
The principal modes of heat transfer to nozzle and combustor walls are
convection and radiation. Of these, convection dominates, and radiation tends to be
important only for particle-laden flows from solid propellant rockets.
Convective Heat Transfer
We will review here the compressible 2D boundary layer equations in order to extract
information on wall heat transfer.
The governing equations are (in the B.L. approximation)
∂ ( ρu ) ∂ ( ρv)
Continuity + =0 (1)
∂x ∂y
∂u ∂u ∂p ∂τxy ∂ ⎛ ∂u ⎞
X-Momentum ρu + ρv + = = ⎜µ ⎟ (2)
∂x ∂y ∂x ∂y ∂y ⎝ ∂y ⎠
∂P
Y-Momentum =0 (3)
∂y

∂ht ∂h ∂ ∂ ⎛ ∂T ⎞
Total enthalpy ρu
∂x
+ ρv t =
∂y ∂y
( )
uτxy + ⎜k ⎟
∂y ⎝ ∂y ⎠
(4)
u2
where ht = h + is the specific total enthalpy, and µ is the viscosity. For a
2
laminar flow, µ = µ ( T ) is a fluid property. Rocket boundary layers are almost always
turbulent, and µ is then the “turbulent viscosity”, where momentum transport is
effected by the random motion of turbulent “eddies”. If these eddies have a velocity
scale u' and a length scale l ' , we have, in order-of-magnitude.
µ turb. ∼ ρ u'l' (5)
where u' is some fraction of the local u, and l ' tends to be of the order of the wall
distance y. The important points about (5) are
(a) µ turb. µ , mostly because l ' mean free path and

(b) µ turb. is proportional to density (whereas µ is not, because the m.f.p. is inversely
proportional to ρ ).
Similarly, the last term on the right in the energy balance, representing the
convergence of heat flux, contains the “turbulent thermal conductivity” K ∼ ρ cpu'l' .
Once again, we notice that K is here proportional to density. We also note that the
“turbulent Prandtl number”
µ t cp
Pr = ∼ 1 (from the orders of magnitude)
kt
It is of some interest to note the origin and composition of the viscous term in
equation (4). If we collect the dot products u i τ t around a fluid element as shown (in
B.L. approximation),

∂
we obtain the term
∂y
(
uτxy ) as written in (4). This can be expanded as
2
∂ ∂τxy ∂u ∂ ⎛ ∂u ⎞ ⎛ ∂u ⎞
∂y
(
uτxy = u )
∂y
+ τxy
∂y
=u ⎜µ ⎟ + µ⎜
∂y ⎝ ∂y ⎠
⎟ (6)
⎝ ∂y ⎠
The 1st term in (6) is just the velocity times the viscous net force per unit volume, so
it is the part of the total viscous work that goes to accelerate the local flow. The
second term in (6) is positive definite, and it is the rate of dissipation of energy into
heat due to viscous effects. We will return later to this heating effect.
Approximate Analysis Let us manipulate the right hand side of equation (4):
∂ ⎛ ∂u ⎞ ∂ ⎛ ∂T ⎞ ∂ ⎡ ⎛ ∂u K ∂T ⎞ ⎤
⎜ uµ ⎟+ ⎜K ⎟= ⎢µ ⎜ u + ⎟⎥
∂y ⎝ ∂y ⎠ ∂y ⎝ ∂y ⎠ ∂y ⎣ ⎝ ∂y µ ∂y ⎠ ⎦
∂h ∂T
and, since = cp , this yields
∂y ∂y
⎡ ⎛ u2 ⎞⎤
∂ ⎢ ⎜∂ 2 1 ∂h ⎟ ⎥ ⎛ µcp ⎞
µ + ⎜⎜ Pr ≡ ⎟
∂y ⎢ ⎜⎜ ∂y Pr ∂y ⎟⎟ ⎥ ⎝ k ⎟⎠
⎢⎣ ⎝ ⎠ ⎥⎦
We note here that, both for laminar and turbulent flows, Pr is a constant,
independent of P and T to a good approximation. In fact, as we noted before, it is
also of order unity ( ∼ 0.9 for turbulent flows). So, the RHS of the energy equation
becomes
∂ ⎡ ∂ ⎛ h u2 ⎞ ⎤
⎢µ ⎜ + ⎟⎥ (7)
∂y ⎢⎣ ∂y ⎜⎝ Pr 2 ⎠⎟ ⎥⎦
If we made the approximation Pr = 1 , then this would reduce

∂ ⎛ ∂ht ⎞ u2
further to ⎜µ ⎟ with ht = h + . If in addition, we made approximations
∂y ⎝ ∂y ⎠ 2
∂P
the “flat plate” approximation 0 , then the pair of equations
∂x
(1), (4) would become
∂u ∂u ∂ ⎛ ∂u ⎞ ⎫
ρu + ρv = ⎜µ ⎟ ⎪
∂x ∂y ∂y ⎝ ∂y ⎠ ⎪
⎬ (8)
∂h ∂h ∂ ⎛ ∂ht ⎞⎪
ρu t + ρ v t = ⎜µ ⎟⎪
∂x ∂y ∂y ⎝ ∂y ⎠⎭

These are identical equations for u and ht . The same equation would also govern the
linearly transformed variables
u ht − htw
u= ; h= (9)
ue hte − htw
where the ( )e subscript denotes the value of a variable in the local “external”
flow (just outside the boundary layer). Both u and h satisfy identical boundary
conditions:
uw = hw = 0; ue = he = 1 (10)
and, as noted, identical governing equations. We conclude that, under the

assumption
⎛ ∂P ⎞
⎜ Pr = 1, ∂x = 0 ⎟ ,
⎝ ⎠
ht − hw u
= (11)
hte − hw ue
u2w
where we also noticed htw = hw + = hw . This similarity relation between velocity
2
and total enthalpy profiles is known as Crocco’s analogy.
Approximate heat flux at the wall
We are interested in the magnitude of the wall heat flux
⎛ ∂T ⎞
qw = ⎜ K ⎟ (12)
⎝ ∂y ⎠w
⎛ K ∂h ⎞ ⎛ K ∂h ⎞
qw = ⎜ ⎟ =⎜ µ t ⎟
⎜ cp ∂y ⎟ ⎜ ⎟
⎝ ⎠w ⎝ µcp ∂y ⎠w
0, since uw = 0
⎛ ∂h ⎞ ∂ ⎛ u ⎞ 2
⎛ ∂h ⎞ ⎛ ∂h ⎞
where we used ⎜ t ⎟ = ⎜⎜ h + ⎟⎟ = ⎜ ⎟ + ⎜ u ⎟
⎝ ∂y ⎠w ∂y ⎝ 2 ⎠
w ⎝ ∂y ⎠w ⎝ ∂y ⎠w

K 1
The group = should be set equal to unity, for consistency with the stated
µcp Pr
approximations. Thus
⎛ ∂h ⎞
qw = ⎜ µ t ⎟
⎝ ∂y ⎠w
Use now equation (11):
⎛ ∂ ⎡ u ⎤⎞ ht − hw ⎛ ∂u ⎞
qw = ⎜ µ
⎜ ∂y ⎢
⎣
hw + hte − hw (
⎥ ⎟ = e
ue ⎦ ⎟⎠
) ue
⎜µ ⎟
⎝ ∂y ⎠w
⎝ w
⎛ ∂u ⎞
and notice that ⎜ µ ⎟ is the wall shear stress, τw . So
⎝ ∂y ⎠w
hte − hw
qw = τw (13)
ue
which is also called Reynolds analogy. A more compact form of this can be written in
terms of the Friction Coefficient
τw
cf ≡ (14)
1
ρ u2
2 e e
and the Stanton number
qw
St = (15)
(
ρeue hte − hw )
with the result (from (13))
cf
St = (16)
2
One important point can be made about the result (13):
The heat flux to the wall is driven by the enthalpy (or temperature) difference
between Total external and Wall values, not between static values. This can be non-
intuitive. Consider the situation near the exit of a highly expanded space nozzle,
where the bulk temperature Te may have dropped to, say, 300K due to the strong
expansion from a chamber temperature of, say, 3000K. The wall could be made of
Tungsten so as to be able to sustain relatively high temperature and cool itself by
radiation to space, so Tw could be, say, 1500K. Is the nozzle wall being heated or
cooled by the 300K gas? The answer is that it is being heated, because

Tte = Tc = 3000K, while Tw = 1500 < Tt . Are we violating the 2nd Principle of
e
Thermodynamics. Read on.
Simplified Profiles, Across the Boundary Layer
To better understand this situation, let us return to Crocco’s analogy (equation 11)
2
and write ht = h + u , and solve for h:
2
u2
(
h = hte − hw ) uu
e
−
2
(17)
This is a quadratic relationship between h and u. For low subsonic flows h ht , so

the last term is not strong, and the relationship becomes linear in the limit.
The relationship between slopes at the wall flows from (17):
0
⎛ dh ⎞ 1 ⎛ du ⎞ ⎛ ∂u ⎞
⎜
⎝ dy ⎠w
(
⎟ = hte − hw
u
)⎜
e ⎝ dy
⎟ − ⎜u ⎟
⎠w ⎝ ∂y ⎠w
⎛ dh ⎞ ⎛ hte − hw ⎞
or ⎜ du ⎟ = ⎜⎜ u ⎟ (18)
⎝ ⎠w ⎝ ⎟
e ⎠
We can use (17) and (18) to sketch h vs. u across the boundary layer. For a case
with he > hw , this looks like

Note that whenever ue > hte − hw there develops an intermediate temperature
maximum. But in any case, the wall slope is as if the line were coming from h t e , not
from he . The case when he < hw is more revealing even:
Now the wall slope is seen to be positive (heat into the wall), despite he < hw (as
long as ht > hw )
e
So, the quadratic portion of the Crocco relationship is responsible for the extra wall
heat; this can in turn be traced to viscous dissipation, which accumulates in the
boundary layer and elevates its temperature, so that the wall is heated even when
the outside temperature is low (as long as the flow has high speed).
Modification for Pr ≠ 1
We leave for now the issue of the non zero pressure gradient, except to note that it
introduces small modifications down to the throat. The deviations of Pr from unity
are small, and, for gases Pr < 1 (~ 0.9 for turbulent flow). This breaks the perfect
balance between dissipation and conduction responsible for Crocco’s analogy, in the
sense of favoring conduction of the dissipated heat. As a second consequence, the
temperature overshoot is reduced, and so is the wall slope of T and the heat flux to
the wall. The direct effect of higher conduction ( Pr < 1 ) is accounted for
approximately by modifying Reynolds analogy to
cf
St = (19)
2Pr 0.6
The secondary effect (reduced overshoot) is accounted for by replacing the driving
enthalpy difference ht − hw by haw − hw , where haw is the “Adiabatic-wall enthalpy”,
e
defined as

u2e
haw = he + r ;r 0.9 (20)
2
(turbulent)
and r is the “Recovery factor”.
With these changes, the heat flux is now
cf
qw = ρeue (haw − hw ) (21)
2Pr 0.6
The Bartz heat flux formula
A very crude, but surprisingly effective representation for the friction factor cf is that
supplied by the well-studied case of fully developed turbulent flow in a pipe.
0.046 ρeueD
cf = ; Re = (22)
R 0.2
e
µe
where R e is the Reynolds number based on diameter D, and µe is the laminar

viscosity. Putting also h = cp T + constant, equation (21) now gives
0.2
0.023 ⎛ µe ⎞ 0.026
qw = ρeuecp ( Taw − Tw ) ⎜ ⎟ = ( ρeue )0.8 µ0.2
e cp ( Taw − Tw ) (23)
Pr 0.6 ⎜⎝ ρeueD ⎟⎠ D 0.2
0.026
It is common practice to define a heat transfer “gas-side film coefficient”, hg (not an
enthalpy!) by
qw
hg ≡ (24)
Taw − Tw
And, so far, we have
0.026
hg = ( ρeue )0.8 µ0.2
e cp (25)
D0.2
At this point we note that the formulation so far has ignored the strong variations of
ρ and µ across the boundary layer since these quantities depend on temperature as
1
ρ ∼ (at P=constant) ; µ ∼ T w ( w 0.6 ) (26)
T
A commonly used approach to including these variations is to replace ρe and µe in

equation (25) by their values at some intermediate temperature <T>:

w
Te ⎛< T >⎞
ρe → ρe ; µe → µe ⎜⎜ ⎟⎟ (27)
<T> ⎝ Te ⎠
and <T> can be evaluated by several empirical rules. For Mach numbers not much
higher than 1, we can simply use
Te + Tw
<T> (28)
2
Making the replacements of (27) in equation (25), we obtain
0.8 − 0.2w
0.026 ⎛ Te ⎞
( ρeue )
0.8
hg = ⎜ ⎟ µ0.2
e cp (29)
D0.2 ⎝< T >⎠
which is one form of Bartz’ formula. A more useful form follows from the continuity
equation:
i
m P At R g Tc
ρeue = = c , with ,
A c* A Γ (γ)
and where A is the local cross-section, and A t the throat cross-section. Substituting
2
A ⎛D ⎞
in (29), and using t = ⎜ t ⎟ , the final form is
A ⎝D⎠
0.8 1.8 0.8 − 0.2w

0.026 ⎛ Pc ⎞ ⎛ Dt ⎞ ⎛ Te ⎞
hg = ⎜ ⎟ ⎜ ⎟ cp µ0.2
e ⎜ ⎟ (30)
D0.2
t ⎝ c *⎠ ⎝D⎠ ⎝< T >⎠
Several important trends and observations can be made now:
⎛ 1 ⎞
(a) Smaller throat diameter leads to larger heat flux ⎜ ∼ 0.2 ⎟ . This comes
⎜ D ⎟
⎝ t ⎠
straight from the Reynolds no. dependence of cf .
(b) Heat flux is almost linear in chamber pressure ∼ Pc0.8 . This limits the ( )
feasibility of high chamber pressures, which are otherwise very desirable.
⎛ ⎛ D ⎞1.8 ⎞
(c) Maximum heat flux occurs at the throat ⎜ ∼ ⎜ t ⎟ ⎟ . One critical design
⎜ ⎝D⎠ ⎟
⎝ ⎠
consideration is therefore the thermal integrity of the throat structure.

(d) Lighter gasses lead to higher heat fluxes, through the combined effects of cp
⎛ 1 ⎞
and c * ⎜ hg ∼ 0.6 ⎟
⎝ M ⎠
0.8 − 0.2w 0.68

⎛ T ⎞ ⎛ Te ⎞
(e) The factor ⎜ e ⎟ ⎜ ⎟ is greater than unity. This
⎝< T >⎠ ⎝< T >⎠
enhancement of heat flux follows mainly from the fact that the gas in the
boundary layer is mostly cooler than in the core, hence denser. We showed
before that the turbulent heat conductivity is proportional to density.
Example
Consider the Space Shuttle Main Engine (SSME), which is a Hydrogen-Oxygen rocket
with (roughly) these characteristics:
Pc = 220 atm 2.2 × 107 Pa
Tc = 3600 K
M = 15g / mol
r 1.25
R g Tc
c* = 2600 m / s
Γ (γ)
γ R
cp = 2800 J / Kg / K
γ −1 M
µe 3 × 10−5 Kg / m / s
2
Tthroat = Tc 3200 K = Te
γ +1
Tw = 1000 K
We calculate then
Te + Tw 3200 + 1000
< T >= = = 2100 K (at the throat)
2 2
0.8 − 0.2w 0.61

⎛ Te ⎞ ⎛ 3000 ⎞
⎜ ⎟ ⎜ 2100 ⎟ 1.3
⎝< T >⎠ ⎝ ⎠

and so, using equation (30),
hg 160, 000 w / m2 / K
and qw (
160, 000 Tawt − 1000 )
1
⎛ γ −1 2⎞
( Taw )t = Tt ⎜1 + r
2
µ ⎟ = 1.057 × 3200 3400 K (slightly less than Tc )
⎝ ⎠
0.9
qw 160, 000 × 2400 = 3.8 × 108 W / m2
This is a very high level of heat flux. To visualize the implications, suppose this qw
had to be transmitted through a thin metal plate (thickness δ , thermal conductivity
k).
∆T
One would have qw = K where ∆T is the temperature drop through the metal .
δ
As an initial guess, suppose the metal were stainless steel (K 20 W / m / K ) , and
δ =1mm. Then
qw δ 3.8 × 108 × 10−3

∆T = = = 19, 000 K !!
K 20
Obviously, this is unacceptable. Try using Copper instead, with K 400 W / m / K

(twenty times better). This gives ∆T =950K, still not acceptable (copper would be
very soft then). The plate would have to be thinner and made of copper. Not an easy
problem.

More rationally
The St or hg should depend on x, distance from start of nozzle, since the B.L. is still
developing (not fully developed). In addition, there should be some accounting for
• acceleration
• property variation through B.L.
• cylindrical geometry
The article by Rubsin and Inonye (ch. 8 in Rosenhow and Hartnett’s Handbook of
Heat Transfer, McGraw-Hill, 1973) gives a general formula for turbulent B.L. In an
cylinder, with acceleration:
A
St ( x ) = n
(and hg = ρeuecpSt )
⎛ρu x ⎞
s ⎜ e e eff ⎟ Fc1−n FRnθ
⎝ µe ⎠
cf 2
s= 1 found walls.
ch
A⎫
⎬ = constants, depending on Reynolds no. based on mom. th.
n⎭
1
R eθ > 4000 , A = 0.0131, n =
7
1
R eθ < 4000 , A = 0.0293 , n =
5
Fc ⎫⎪
⎬ = Factors for property variability. Can take several nearly equivalent forms. A
FR θ ⎪
⎭
simple one from Eckert, is
ρe <Τ > ⎫
Fc = = ⎪
ρ (< Τ >) Te ⎪
⎪< Τ > T T
⎬ = 0.28 + 0.50 w + 0.22 aw
w ⎪ Te Te Te
µe ⎛ Te ⎞ ⎪
FRθ =
µ (< Τ >)
⎜ ⎟ (w 0.6 ) ⎪
⎝< Τ >⎠ ⎭
u2e ⎛ γ −1 2⎞
Taw = Te + r = Te ⎜ 1 + r Me ⎟
2 ⎝ 2 ⎠

r 0.9 (recovery factor)
The “effective distance” xeff is related to the actual distance x through an integral
(accounting for memory of past acceleration)
x f ( x ')
xeff ( x ) = ∫ dx '
0 f (x)
where f =
(
ρeue zRµne ) 1−n
n
Fc FR1 −n
θ
haw − hw ⎛ u2 ⎞
z= ⎜ haw = he + r e ⎟
hte − hw ⎜ 2 ⎟
⎝ ⎠
R=R(x)= body radius at x.
For a quick estimate of R eθ , we can simplify further to the flat-plate case, in which
dθ c f
= ,
dx 2
with
cf
⎧0.0128 R1 4
⎪
=⎨
eθ (R eθ < 4000 ) , and with dθ
=
dR eθ
2 ⎪0.0065 R1e 6
⎩ θ
(R eθ > 4000 ) dx dR ex
⎧ 0.0128 ⎧0.0366 R 4 5 R eθ < 4000

⎪ ⎧4 5 4 ⎪ ex
⎪⎪ 5 R eθ = 0.0128 R ex
14
dR eθ ⎪ R eθ
dR ex
=⎨
⎪ 0.0065
→⎨
⎪ 6 R 7 6 = 0.0065R
→ R eθ
⎪
⎨
⎪
(R ex < 1.99 × 101 )
⎪ R1 6 ⎪⎩ 7 e ex ⎪0.0152 R 6e 7 R eθ > 4000
⎩ eθ ⎩ x
⎧ 0.0366
⎪ 0.2
R eθ θ ⎪ R ex
or = =⎨
R ex x ⎪ 0.0152
⎪ R1 7
⎩ ex

For large rockets, R ex tends to be ∼ 107 − 108 , so the high R e formulas should
throat
be better, despite the common use of Bartz’s formulae, which are based on the low
R e formulation. Fortunately, differences tend to be small, and are marked often by
other uncertainties (surface films, fluid properties).
Example and Comparisons:
R 2t 1 R ± R 2 − R 2t
Consider nozzle R = x tan x + x=
4 tan α x 2 tan α
R c − R 2c − R 2t
with origin at x = xc =
2 tan α
Rc
and = 1.5 , α = 15o
Rt
Rt
and going through throat at x = xt =
2 tan α
Using γ=1.25 and the R eθ > 4000 option, we find
xt
Rt 1.875 0.9
⎛ xeff ⎞ 5 ⎛ 1.125 ⎞ ⎛ 0.6979 ⎞ ⎛ x ⎞
⎜⎜
⎝ Rt
⎟⎟
⎠throat
= ∫
xc
M 12 ⎜⎜
⎝ 1 + 0.125M
2 ⎟⎟
⎠
⎜⎜
⎝ 0.6515 + 0.0464 M
2 ⎟⎟
⎠
d ⎜⎜
⎝ Rt
⎟⎟
⎠
Rt
1
R x 4 R
where
Rt
=
Rt
tan15o + ⇒
R
(x)
⎛ x ⎞ o t
⎜ ⎟ tan15
⎝ Rt ⎠
2.25
1 ⎛ 1 + 0.125M2 ⎞ R
and 1 ⎜ ⎟⎟ = ⇒ M (x)
⎜ 1.125 Rt
M 2⎝ ⎠
⎛x ⎞
The integration gives ⎜⎜ eff ⎟⎟ = 1.0892
⎝ R t ⎠throat
x t − xc ⎛ xc ⎞
Compared to = 1.153 ⎜⎜ and = 0.713 ⎟⎟
Rt ⎝ Rt ⎠

Since xeff appears to the 1 power, the memory/acceleration effect (up to the
7
throat) is insignificant.
The throat St is then
(0.9) ( Taw = 3263 )

Te 2
(using r=1, Tw = 1000 K , Tc = 3300 K, = 3300 = 2933K )
t 2.25
0.0131
(St )throat = 1
⎛ ρ u ( x t − xc ) ⎞ 7
0.771
⎛< T >⎞
⎜⎜ ⎟⎟ ⎜ ⎟
⎝ µ ⎠throat ⎝ Te ⎠throat
(3263)
(0.6952)
<T > 1000 3300
= 0.28 + 0.50 + 0.22 = 0.6979
Te 2933 2933
Take Pc = 2 × 107 N / m2 ,
M= 25 g/mol
8.314
R g Tc 3300
* 0.025
c = = 2.25
= 1592 m / s
Γ
⎛ 2 ⎞ 0.5
1.25 ⎜ ⎟
⎝ 2.25 ⎠
Pc
⇒ ( ρu ) t = *
= 12560 Kg / s / m2
c
xt 1 xc 1.5 − 1.52 − 1
= = 1.866 = = 0.7128
Rt 2 tan15o Rt 2 tan15o
and, with R t = 0.3 m ,
0.6
⎛ T ⎞
µ 6.8 × 10−5 ⎜ ⎟ ⇒ µthroat = 6.70 × 10−5 Kg m sec
⎝ 3000 ⎠

One gets,
(St )throat = 0.00133 (0.00124 using xt instead of xt − xc )
Using the R eθ < 4000 option (small rockets)
0.0293
(St )throat =
0.2 0.68
⎛ ρ uxeff ⎞ ⎛< T >⎞
⎜ ⎟ ⎜ ⎟
⎝ µ ⎠throat ⎝ Te ⎠throat
and using again xeff = xt − xc , etc,
we get ( St )throat = 0.00102 (0.000933 using xt )
For comparison, the “fully developed pipe flows” formulation would give
0.2 0.8 − 0.2w 0.9

Rg 0.026 ⎛ c* ⎞ ⎛ Te ⎞ ⎛ At ⎞
St = = ⎜⎜ ⎟⎟ µ0.2
e ⎜ ⎟ ⎜ ⎟
ρ ue cp D0.2
t ⎝ Pc ⎠ ⎝< T >⎠ ⎝ A ⎠
1 at throat
(St )throat = 0.000958
This is close to the R θ < 4000 results above (and, indeed, the coefficients are for
R θ < 4000 ). But this appears coincidental, based on the fact that for most nozzles,
∆x ∼ R t .

Lecture 8: Convective Heat Transfer: Other Effects
Overall Heat Loss and Performance Effects of Heat Loss
(1) Overall Heat Loss
i
The local heat loss per unit area is qw = ρucp ( Taw − Tw ) St , and using m = ρuπR2 , the
integrated heat loss is
L 2
⎛ dR ⎞
Qw
x =0
∫qw 2πR ds ; ds = 1 + ⎜ ⎟ dx
⎝ dx ⎠
dx (small angles) (1)
L i L
m i dx
Qw ∫ πR 2
cp ( Taw − Tw ) St 2πR dx = m cp ( Taw − Tw ) St 2
∫ (2)
R
0 0
For an approximate evaluation, assume the quantity cp ( Taw − Tw ) St is a weak

function of x, and treat it as a constant. We then obtain
L L
Qw Taw − Tw dx ⎛ Tw ⎞ dx
i Tc
2St
R (x)∫ ⎜⎜1 −
⎝
⎟ 2St
Tc ⎟⎠ ∫ R (x) (3)
m cp Tc 0 0
L
⎛L ⎞ dx Tw 1 1
For many rockets, ⎜ ⎟ ≡
⎝ R ⎠eff ∫ R (x)
0
is of the order of 6-10, and
Tc
∼ − , so the
4 3
Qw
ratio i
(heat loss divided by total enthalpy flux) is of the order of 8-16 times
m cp Tc

the Stanton number. As we found before, St is itself ~ 0.001, leading to fraction
heat losses of the order of 1-2%. While this is a small fraction, its absolute value
may be large, because the total thermal power is enormous. As an example, for the
SSME engine
i F 2 × 106 N J
m cp Tc = cp Tc × 2770 × 3600 K ,
C 4500 m s KgK
i
or m cp Tc = 4.4 × 109 W (the output power of four large power stations).
A 1.5% fraction of this means 66 MW lost to the walls (some 80,000 HP).
(2) Effect on Performance
As a starting guess, we could imagine that all of the losses ( Qw ) are reflected in an
equal amount of kinetic energy loss in the exhaust. If ue0 is the exist velocity with
no losses,
i ⎛u u2 ⎞
2
e
m⎜ 0 − e ⎟ Qw (4)
⎜ 2 2 ⎟
⎝ ⎠
But a little reflection shows that the kinetic energy loss must be less than Qw .
Indeed, heat losses that occur near the nozzle exit plane are almost irrelevant for
performance, because the thermodynamic efficiency of the remaining expansion
from the point of loss to the exhaust is very small, so very little of that loss is
reflected in a kinetic energy decrease.
So, for the time being, we simply acknowledge this by writing
i ⎛u u2 ⎞
2
e
m ⎜ 0 − e ⎟ < Qw (5)
⎜ 2 2 ⎟
⎝ ⎠
2Qw ue 2Q
or ue > u2e0 − or > 1− i w (6)
i ue0
m mu2e0
⎡ γ −1 ⎤
⎢ ⎛P ⎞ 2 ⎥
Remembering that u2e0 2 = cp ( Tc − Te ) = cp Tc ⎢1 − ⎜⎜ e ⎟⎟ ⎥,
⎢ ⎝ Pc ⎠ ⎥
⎣ ⎦

⎛ i ⎞
⎜ Qw m cp Tc ⎟
ue
> 1− ⎝ γ −1
⎠ (7)
ue0
⎛P ⎞ γ
1−⎜ e ⎟
⎝ Pc ⎠
Qw
If the fractional loss i
is of order 1.5% and the expansion efficiency
m cp Tc
γ −1
⎛P ⎞ γ u 1 0.015
η = 1 − ⎜⎜ e ⎟⎟ is of order 75%, then e > 1 − = 1 − 0.01 (i.e., a loss of less
P
⎝ c⎠ u e0
∼ 2 0.75
than 1% in specific impulse, ignoring the exit pressure contribution).
The calculation can be made more precise by tracking the evolution of the gas
temperature. The total energy equation, accounting for the losses, is
Tc
i dht ⎛ dT du ⎞
m = ρuA ⎜ cp +u ⎟ −qw 2πR = − ρucp ( Taw − Tw ) St 2πR
dx ⎝ dx dx ⎠
dT du 2St
or cp +u =− cp ( Tc − Tw ) (8)
dx dx R
But the momentum equation (ignoring, somewhat inconsistently, the effects of

du dp du 1 dp
friction), gives ρ u + = 0 , or u =− . Substituting in (8),
dx dx dx ρ dx
dT 1 dp 2St
cp = − cp ( Tc − Tw )
dx ρ dx R
1 dp γ − 1 1 dp
Divide by cp T and note that =
ρ cp T dx γ p dx
1 dT γ − 1 1 dp 2St Tc − Tw
= − (9)
T dx γ p dx R T
γ −1
T ⎛P ⎞ γ
Without the heat loss term, this would integrate to = ⎜⎜ ⎟⎟ , the ideal flow
Tc ⎝ Pc ⎠
(isentropic) relation. More generally now,

γ −1
T ⎛P ⎞ γ ⎡ L 2S T − T ⎤
= ⎜⎜ ⎟⎟
Tc ⎝ Pc ⎠
exp ⎢ −
⎢ R ∫
t c
T
w
dx ⎥
⎥
(10)
⎣ 0 ⎦
and since the exponent is a small number,
γ −1
T ⎛P ⎞ γ ⎡ L
2St Tc − Tw ⎤
= ⎜⎜ ⎟⎟
Tc ⎝ Pc ⎠
⎢1 −
⎢ ∫ R T
dx ⎥
⎥
(11)
⎣ 0 ⎦
To evaluate the correction term, we use for T the undisturbed T, as if no heat loss
had happened. This gives
γ −1
Tc − Tw T − Tw Tc ⎛ Tw ⎞ ⎛ Pc ⎞ γ
= c ⎜⎜1 − ⎟⎜ ⎟
T Tc T ⎝ Tc ⎟⎠ ⎝ P ⎠
⎛ T ⎞
and we also assume that St ⎜⎜1 − w ⎟⎟ is nearly constant:
⎝ Tc ⎠
γ −1
⎡ L
γ −1 ⎤
T ⎛P ⎞ γ ⎢ ⎛ Tw ⎞ ⎛ Pc ⎞ γ dx ⎥
Tc
⎜⎜
⎝ Pc
⎟⎟
⎠
⎢1 − 2St
⎢⎣
⎜⎜1 −
⎝
⎟
Tc ⎟⎠ ∫⎜ ⎟
0⎝
P ⎠ R ⎥
⎥⎦
(12)
and, in particular, at the exit plane,
γ −1
⎡ L
γ −1 ⎤
Te ⎛ Pe ⎞ γ ⎢ ⎛ Tw ⎞ ⎛ Pc ⎞ γ dx ⎥
Tc
⎜⎜
⎝ Pc
⎟⎟
⎠
⎢1 − 2St
⎢⎣
⎜⎜ 1 −
⎝
⎟
Tc ⎟⎠ ∫⎜ ⎟
0⎝
P ⎠ R ⎥
⎥⎦
(13)
We now express the exit kinetic energy as
(
u2e = 2cp Tte − Te ) (14)
where both Tte and Te are affected by the losses. For the total energy loss, we have
( )
i
m cp Tc − Tte = Qw
and so

L
Tt e Qw ⎛ T ⎞ dx
Tc
=1− i = 1 − 2St ⎜⎜1 − w ⎟⎟
⎝ Tc ⎠ ∫R (15)
m cp Tc 0
where we have used the result in equation (3). For the loss of static energy, we have
the result in (13). Using both in (14),
⎡ L
γ −1
L
γ −1 ⎤
⎢ ⎛ Tw ⎞ dx ⎛ Pe ⎞ γ ⎛ T ⎞ ⎛ Pe ⎞ γ dx ⎥
∫ ∫
2
ue = 2cp Tc ⎢1 − 2St ⎜⎜1 − T ⎟⎟ − ⎜⎜ ⎟⎟ + 2St ⎜⎜ 1 − w ⎟⎟ ⎜ ⎟
R ⎝ Pc ⎠ Tc ⎠ P ⎠ R ⎥⎥
⎢ ⎝ c ⎠ 0 ⎝ 0⎝
⎣ ⎦
⎧ γ −1
L ⎡ γ −1 ⎤ ⎫
⎪ ⎛P ⎞ γ ⎛ T ⎞ ⎢ ⎛ Pe ⎞ γ ⎥ dx ⎪
or u2e = 2cp Tc ⎨1 − ⎜⎜ e
⎪ ⎝ Pc
⎟⎟
⎠
− 2St ⎜⎜ 1 − w ⎟⎟
⎝ Tc ⎠ ∫ ⎢1 − ⎜ P ⎟
0 ⎢ ⎝ ⎠
⎥ R ⎬
⎥⎦ ⎪
(16)
⎩ ⎣ ⎭
γ −1
⎛P ⎞ γ
We see now that the factor 1 − ⎜ e ⎟ occurring in the integral of (16) is just the
⎝P ⎠
“thermodynamic relief” we had mentioned earlier, which makes the loss of kinetic
energy be less than the heat loss. Indeed, this factor becomes zero as P → Pe , so, as
anticipated, heat losses near the exit are irrelevant.
⎡ γ −1 ⎤
⎢ ⎛P ⎞ γ ⎥
To simplify the writing, use u2e0 = 2cp Tc ⎢1 − ⎜⎜ e ⎟⎟ ⎥
⎢ ⎝ Pc ⎠ ⎥
⎣ ⎦
and define
γ −1 γ −1
⎛P ⎞ γ ⎛ P ⎞ γ
η0,e = 1 − ⎜⎜ e ⎟⎟ and ηx,e =1−⎜ e ⎟ :
⎜ P (x) ⎟
⎝ Pc ⎠ ⎝ ⎠
L
u2e ⎛ T ⎞ ⎛ ηx,e ⎞ dx
u2e0
= 1 − 2St ⎜⎜1 − w ⎟⎟
⎝ Tc ⎠ ∫ ⎜
⎜η
0 ⎝ 0,e
⎟
⎟ R
⎠
(17)
⎛ ηx,e ⎞
and, again, the right-hand-side minus the ⎜ ⎟ factor would be the relative heat
⎜ η0,e ⎟
⎝ ⎠
loss (equation 3). Numerical evaluation shows that the modified integral in (17) is

L
dx
about 2
3
of the original integral ∫R
0
. Remembering our earlier estimate of the
relative Isp loss ( < 1%), we conclude that a better estimate is about 0.67%. This
∼
amounts to 3 sec. out of Isp 400s .

Lecture 9: Liquid Cooling
Cooling of Liquid Propellant Rockets
We consider only bi-propellant liquid rockets, since monopropellants tend to be small

and operate at lower temperatures. In a bi-propellant rocket, both the oxidizer and
the fuel streams are in principle available for cooling the most exposed parts of the
chamber and nozzle prior to being injected. This is called “regenerative cooling”,
because the heat loss from the gas is recovered (“regenerated”) into the liquid, so no
heat escapes. This is not to say no thermodynamic loss is incurred, though (heat is
transferred from very hot gas to cool liquid, which implies irreversibility and loss of
work potential).
Of the two streams, the fuel is normally used for cooling. This is for two reasons:
(a) Fuels tend to have higher specific heats, so more heat is removed for a given
∆T of the coolant, and
(b) Leakage from an oxidizer stream into the normally fuel-rich combustion gas
can produce a local flame that can be catastrophic, whereas leakage from a
fuel line into the same fuel-rich gas is inert. In addition, exposing hot metal
to oxygen or strong oxidants always carries some risk of accelerated chemical
attack, or even ignition. Some exceptions do exist where oxidizers are used
for cooling, though.
A typical arrangement is as shown below (Figure 1).
The fuel at high pressure from the fuel pump (FP) is sent through a series of narrow
passages carved into the nozzle and chamber walls, picks up the wall heat flux from
the gas, and is delivered eventually to the injector manifold. Since the nozzle region

is the most thermally loaded one, often the coolant flow is split, with one part
entering at the nozzle exit and another providing extra cool fluid by entering just
downstream of the throat.
A typical construction for the cooling channels is shown in Fig. 2.
The load-bearing part of the structure is milled longitudinally with channels of

varying depth and width (to obtain varying liquid velocity), and a high thermal
conductivity thin layer of a Copper alloy is then brazed on the inside.
Design Considerations
Two aspects need to be verified in the design of the cooling system:
(a) The coolant should have sufficient thermal capacity to absorb the heat load
without exceeding some critical temperature, which may be a chemical
decomposition limit (thermal cracking for hydrocarbons) or the boiling point
(although, with care, boiling can be sometimes tolerated or exploited for its
strong heat absorption properties).
Suppose QLOSS is the calculated total heat loss from the gas. As seen in a
i
previous lecture, this amounts to 1-3% of m cp Tc , more for the smaller
i
engines. Suppose also the fuel only is used as coolant, with a flow rate mF .
i i i
i i m− mF i m
The O F ratio is defined as O F = mox mF = , and so mF = . If
1+O
i
mF F
the liquid fuel has a specific heat ccool , its temperature rise ∆T from inlet to
exit of the cooling circuit will be given by
i
QLOSS = mF ccool ∆T (1)
i
i ⎛Q ⎞ mF
m cp Tc ⎜⎜ LOSS ⎟⎟ = ccool ∆T
⎝ QTOT ⎠ 1 + OF

⎛Q ⎞ ⎛ cp ⎞⎛ O⎞
∆T = ⎜⎜ LOSS ⎟⎟ ⎜⎜ ⎟⎟ ⎜ 1 + ⎟ Tc (2)
⎝ QTOT ⎠ ⎝ Ccool ⎠⎝ F⎠
QLOSS cp
which must be kept within limits. For example, say =0.02, =1,
Q TOT ccool
O =4, T =3000K; we obtain ∆T = 0.02 × 1 × 5 × 3000 = 300 K . This may or
F c
may not be acceptable; for a cryogenic coolant it would most likely be, but a
hydrocarbon fuel, exiting the pump at 300K will then leave the cooling circuit
at 600K, probably too high for chemical stability.
(b) The local cooling rate at the most exposed location (the throat) must be
sufficient to avoid decomposition or boiling even at the contact point of the
liquid with the wall. The thermal situation in a cut through the front wall of
the cooling passages is as schematizes in Fig. 3.
The Taw temperature is shown dashed because, as we know it is not the

actual gas temperature outside the gas boundary layer, but is the one driving
heat. The liquid bulk is at a temperature Tl , which is below that of the wetted
wall ( Twc ), because heat has to be driven through according to
q = hl ( Twc − Tl ) (3)
where hl is the liquid-side film coefficient, that can be calculated, for instance,
from Bartz formula using liquid properties. The same heat flux is supplied
from the gas through the gas-side film coefficient:

q = hg ( Taw − Twh ) (4)
and yet the same flux must cross the wall by conduction:
Twh − Twc
q=k (5)
δ
where k is the wall thermal conductivity, and δ its thickness.
⎧ q
⎪ Taw − Twh = h
⎪ g
⎪⎪ δ
Re-writing (3)-(5) as ⎨ Twh − Twc = q
⎪ k
⎪ q
⎪ Twc − Tl = h
⎪⎩ l
and adding, we obtain
⎛ 1 δ 1⎞
Taw − Tl = ⎜ + + ⎟q;
⎜ hg k hl ⎟
⎝ ⎠
Taw − Tl
q= (6)
1 1 δ
+ +
hg hl k
which we can then use to calculate intermediate temperatures from (3), (4),
(5). Clearly, what we have done is adding the series “thermal impedances”
1 δ 1
, and of the gas boundary layer, the metal, and the liquid.
hg k hl

Stresses in Cooled Nozzle Walls
Ribs carry weak in-

plane stress.
Wall must carry

the large hoop
stress due to Pg ,
Pl . In addition, hot
side will expand
more, forcing back
(cold) side to
higher tension. So,
use high σul steel
Fig. 1
for back (1) layer.
Front (2) layer needs to be good thermal conductor; use Cu or W-Cu alloy (higher
strength). Cu has higher expansion coefficient αCu > αsteel , which adds to the effect
of higher T, and ends up putting this layer in compression. This can be relieved by
hot assembly, so that the Cu is pre-stretched when cold.
Plane strain At any z, within one of the materials.
σ (z)
ε = (1 − ν ) + α ⎡⎣ T ( z ) − T0 ⎤⎦ (1)
E
where T0 can be interpreted as the temperature at which the strain ε is defined to

be zero, with zero stress. Since the shape remains planar, ε = constant (at least
within the layer).
Write (1) for both layers. We now “assemble” them with a tight fit, but zero stresses,
at T0 , which from now on means the assembly temperature. Upon heating or cooling,
thermal stresses will arrive, even with no loading or T gradients.
ε =0 by definition at
Both layers now have the same (constant) ε
assembly
σ1 ( z )
(1 − ν1 ) E1
+ α1 ⎡⎣ T1 ( z ) − T0 ⎤⎦ = ε (2)
σ2 ( z )
(1 − ν2 ) E2
+ α2 ⎡⎣ T2 ( z ) − T0 ⎤⎦ = ε (3)
For metals, ν varies little, so take ν1 = ν2 = ν .

The temperature T1 ( z ) will be nearly constant (adiabatic outer condition). T2 ( z ) will
vary linearly with z, according to
dT2
−k 2 =q (4)
dz
We write (3) at z = 0 , z = t2 and subtract:
σ2c − σ2h
(1 − ν ) E2
= α2 ⎡⎣ Twh − Twc ⎤⎦ (5)
and integrate (4) to
Twh − Twc
q = k2 (6)
t2
so that
α2 E2 t2
σ2c − σ2h = q (7)
1 − ν k2
Also (2) reads
σ1
ε = (1 − ν ) + α1 ⎣⎡ Tl − T0 ⎦⎤ , which can be combined with
E1
σ2c
ε = (1 − ν ) + α2 ⎡⎣ Twc − T0 ⎤⎦
E2
⎛σ σ ⎞
to give (1 − ν ) ⎜ 1 − 2c ⎟ = α2 Twc − α1 Tl − ( α2 − α1 ) T0 (8)
⎝ E1 E2 ⎠
Heat transfer from wall 2 to liquid gives
q = hl ( Twc − Tl ) (9a)
q
or Twc = Tl + (9b)
hl
Substitute into (8)
⎛σ σ2c ⎞ q
(1 − ν ) ⎜ E1 −
E2 ⎠
⎟ = ( α2 − α1 ) ( Tl − T0 ) + α2
hl
⎝ 1

σ1 σ2c α2 q ( α2 − α1 ) ( Tl − T0 )
or − = + (10)
E1 E2 1 − ν hl 1−ν
In all of this, q is taken as a given. It can be calculated from the given Taw , Tl , plus
hg , hl and t2 , k2 :
Taw − Tl
q= (11)
1 1 t
+ + 2
hg hl k 2
Equations (7) and (10) relate σ2h , σ2c , σ1 . We need one more equation
Force balance
Since T2 ( z ) is linear in z, so will σ2 ( z ) . For force calculations, then, we can use the
mean value
σ2c + σ2h
σ2 = (12)
2
The net balance then is
2 σ1 t1 + 2 σ 2 t 2 = PgD + 2Pl t l (13)

which is our 3rd equation, together with (7), (10). Notice that, since tl D , Pl has
only a minor effect on stresses (except in the ribs).
From here, we either solve for σ2h , σ2c , σ1 (given geometry, Pg , Pl , q) or, for design,
go the reverse route and decide on the geometry for assigned stresses. We will
pursue here the second approach.
Design
As noted, σ1 will be positive and high, whereas σ2h (and less so σ2c ) will be
negative, and probably high too. We then take the view that
σ1 = σ1ult,tens. S (14)
(S=safety factor ~ 1.5) and
⎧⎪σ2ult,comp. S (15a)
( −σ2h ) = least of ⎨
⎪⎩σ2buckling S (15b)
and use these conditions to determine t1 , t2 . The selection of t2 is complicated by

the fact that ( −σ2h ) will decrease with t2 , but so will (quadratically) σ2buckling :
For buckling, use a simple clamped-beam formulation:

E2Iπ2 4π2E2I
Fc = =
( l 2)2 l2
1
where I = Ht32 .
12
Dividing by A = Ht2 ,
⎛ 1 ⎞
4π2E2 ⎜ Ht23 ⎟ 2
⎝ 12 ⎠ π2 ⎛ t2 ⎞
σ2buckling = σ2buckling = ⎜ ⎟ E2 (16)
l 2 (Ht2 ) 3 ⎝ l ⎠
To proceed, start by eliminating σ2c between (7) and (10):
σ1 σ2h α t2 α2 q ( α2 − α1 ) ( Tl − T0 )
− − 2 q= +
E1 E2 1 − ν k2 1 − ν hl 1−ν
E α E ⎛1 t ⎞ E2 ( α2 − α1 ) ( Tl − T0 )
( −σ2h ) = − E2 σ1 + 1 2− ν2 ⎜ + 2 ⎟q+ (17)
1 ⎝ hl k2 ⎠ 1−ν
or, recalling (11),
1 t
+ 2
E2 α2E2 hl k2 ( α2 − α1 ) E2 ( Tl − T0 )
( −σ2h ) =−
E1
σ1 +
1−ν 1 1 t
( Taw − Tl ) +
1−ν
(18)
+ + 2
hg hl k2

This displays several effects:
σ1ult.
(a) Once σ1 = is fixed, −σ2h will increase with t2 , although weakly, since
S
t2 1 1
, (the Cu wall offers little thermal impedance, compared to the
k2 hg hl
two boundary layers).
(b) This ( σ2h ) (middle term in (18)) arises from the heating of layer 2 relative to
1, due to heat flowing in.
(c) The positive stress σ1 (to counter Pg mostly) relieves this tendency to
compress layer 2, and might even reverse it.
(d) The last term in (18) arising from differential expansion coefficients α2 − α1 ,
could be used as a design aid. If 2=Cu, 1=steel, α2 = 2.3 × 10−5 K −1 ,
α1 = 1.4 × 10−5 K −1 , so α2 − α1 > 0 .
Then, if −σ2h is still too high despite σ1 , we could increase T0 , if possible above Tl ,
to reduce −σ2h . This implies assembly at high temperature.
⎧⎪(18 ) ⎫⎪
We can use ⎨ ⎬ to construct a plot like Figure 3, and select a viable t2 . Once this
⎩⎪(16 ) ⎭⎪
is done, equation (7) gives σ2c , equation (12) gives σ 2 , and equation (13) gives t1 .
Some data:
Material E ( Pa )
(at 500K)
( )
d k −1 σult ( Pa )
(at 500K)
ν
Z=
(1 − ν ) σult.
( K)
o
Eα
Cu 0.95 × 1011 2.3 × 10−5 1.1 × 108 0.3 35
St. Steel 302 1.61 × 1011 1.8 × 10−5 4.6 × 108 0.3 111
Ti 1.63 × 1011 1.7 × 10−5 5.4 × 108 0.3 136
Alloy Steel 1.09 × 1011 1.4 × 10−5 5.1 × 108 0.3 234
(SAE x4130)
The last column is a “figure of merit” extracted from equation (18) to give a
preliminary rough idea of materials expansion stress. The higher Z, the higher the
∆T to reach σult. in a double strip of this material subject to differential heating ∆T .

Example Pg = 100 atm 107 Pa (neglect Pe effect)
D=0.3m; l=4mm
hg = 24000 W / m2 / K ;
hl = 2.76 × 105 W / m2 / K ;
k2 = 360 W / m / K
Taw = 3200 K , Tl = 400 K
5.1 × 108
σ1 = Pa ( alloy steel) ; E1 = 1.09 × 1011 Pa ; α1 = 1.4 × 10−5 K −1
1.5
1.1 × 108
−σ2ult,comp. = Pa ( Cu) ; E2 = 0.95 × 1011 Pa ; α2 = 2.3 × 10−5 K −1
1.5
Substituting into (18),
1.303 × 10−3 + t2
−σ2h = −2.96 × 108 + 8.740 × 109 + 1.221 × 106 ( 400 − T0 ) (19)
16.30 × 10−3 + t2
and, from (16)
σ2buckling = 1.95 × 1016 t22 ( t2 in m.)
Following are some calculated results:
COMMENTS
t2 (mm) 0 0.2
1.1 × 108
T0 = 297 K −σ2h (Pa ) 5.33 × 108 6.26 × 108 Since −σult,com = , assembling at
1.5
−σ2buckling (Pa ) 0 7.80 × 108 room T0 is not acceptable for any t2 .
t2 (mm) 0 0.2
T0 = 500 K −σ2h (Pa ) 2.81 × 10 3.78 × 108
8
Closer, but still no solution
−σ2,buckl. (Pa ) 0 7.80 × 108

t2 (mm) 0 0.2 0.1 With assembly at
−σ2h (Pa ) 8
0.36 × 108 1.34 × 10 0.85 × 10
8 T0 =700K, a very thin
t2 0.1mm is
T0 = 700 K −σ2,buckl. (Pa ) 0 7.80 × 108 1.95 × 108
acceptable, but may
+σ2c (Pa )
( )
q w / m2 = 6.15 × 107 +4.48 × 108
4.40
be questionable on
t1 (mm) robustness. Also, σ2c
is too high
t2 (mm) 0 0.2 0.3 If T0 =800K, is

−σ2h (Pa ) 8
−0.86 × 10 0.12 × 10 0.60 × 10
8 8
feasible, then
T0 = 800 K −σ2,buckl. (Pa ) 0 8
7.8 × 10 1.76 × 10 8 t2 = 0.3 mm is
acceptable. σ2c also
q = 6.07 × 107 +σ2c (Pa ) 9.79 × 107
OK (in tension)
t1 (mm) 4.48
With the assumed l = 4 mm , buckling is not a problem in any case, but compressive
failure is hard to avoid. It may be possible to exceed the elastic limit and go into
plastic compressive yield if ductility is high enough to ensure no rupture. But this
means no reusability.

Lecture 10: Ablative Cooling, Film Cooling
Transient Heating of a Slab
Typical problem: Uncooled throat of a solid propellant rocket
Inner layer retards heat flux to the heat sink. Heat sink’s T gradually rises during
firing (60-200 sec). Peak T of heat sink to remain below matl. limit. Back T of heat
sink to remain below weakening point for structure.
Prototype 1-D problem:
Can be solved exactly, or can do transient 1-D numerical computation. But it is

useful to look at basic issues first.
Thermal conductance of B.L.=hg

k1
Thermal conductance of front layer =
δ1
ki
Thermal conductance of layer i = ( δi = thickness, ki = thermal conductivity)
δi

k1
Want layer 1 to have hg to protect the rest.
δ1
⎛ k1 1 W / m / K ⎞ k1 W
(Say, porous, Oriented graphyte, ⎜⎜ ⎟⎟ → = 330 2 compared to
⎝ δ1 = 3 mm ⎠ δ1 mK
W
hg ∼ 50, 000 2 , so OK here).
mK
Also, from governing equation
∂T ∂2 T ∂T ∂2 T
ρc =k 2 → =α 2
∂t ∂x ∂t ∂x
k
(α = , thermal diffusivity, m 2 / s )
ρc
we see that
x2
x2 ∼ αt , or x ∼ αt , or t ∼ .
α
δ12
So the layer 1 will “adapt” to its boundary conditions in a time t ∼ .
α1
J Kg 1
Say, c 710 and ρ 1100 3
( solid graphyte),
KgK m 2
1
so α = = 1.3 × 10−6 m2 / s .
710 × 1100
(3 × 10 )
2
−3
δ2
The layer “adapts” in t ∼ −6
= 7.0 sec (more like = 1.8 sec ).
1.3 × 10 4α
⇒ Treat front layer quasi-statically, i.e., responding instantly to changes in heat flux:
T (t) −T (t)
q (t)
wh1 wc1
k1
δ1
This also means we can lump the thermal resistances of BL and 1st layer in series:
1 1 δ
+ 1
(hg ) eff hg k1

k1
and since hg ,
δ1
k1
(hg )eff ∼
δ1
hg
For layer 2 (the heat sink), k2 is high (metal) and hg ( )eff is now small (thanks to 1st
layer) so, very likely,
k2
δ2
(hg )eff
W
(For instance, say Copper, k2 360 , with δ2 = 2 cm. We now have
mK
k2 W k W k
(hg )eff δ2
= 350 2 , but 2 = 36, 000 2 , so indeed, 2
mK δ2 mK δ2
(hg )eff ).
Under these conditions, the heat sink is being “trickle charged” through the high
thermal resistance of layer 1. Most likely, heat has time to redistribute internally, so
that T2 is nearly uniform across the layer. We can then write a lumped equation.
dT2 k1
ρ2c2 δ2
dt
= q = hg( )eff ( Taw − T2 ) δ1
( Taw − T2 )
ρ2c2 δ1δ2 dT2

Define τ =
k1
τ
dt
+ T2 = Taw ( T2 (0 ) = T0 )
t
−
T2 = Taw − ( Taw − T0 ) e τ
J
For our example, say ρ2 = 8900 Kg / m3 (Copper), c2 = 430 , δ2 = 2 cm
KgK
8900 × 430 × 3 × 10−3 × 2 × 10−2

τ= = 230 sec
1

This is comfortable. Suppose Taw = 3300 K , T0 = 300 K , and we fire for 120 sec:
(60)
(60)
120 (989)
−
T2 (120 ) = 3300 − 3000 e 230 = 1520 K May need 4 cm
which is still (OK) for Copper (melts at 1360K, but no stress bearing, so can go to
~900. Also OK for steel on Carbon str member).
NOTE:
δ22
=
(0.02) 2
= 1.1 sec , so, indeed, layer 2 “adapts” quickly to B.C.’s

4α2 4 × 9.4 × 10−5
k2 360
→ uniform = = 9.4 × 10−5 m2 / s .
ρ2c2 8900 × 430
A More Exact Solution
Consider Taw “turned on” at t=0. The B.L. has a film coefficient hg , and the first
hg k1
layer has δ1 , k1 , so that hg ( )eff =
δ
∼
δ1
. Layer 2 has thickness δ2 , and has
1 + hg 1
k1
k2 , ρ2 , σ2 , α2 . The back is insulated.
Then one can prove that layer 2 has a temperature distribution

α2t
Taw − T2 ( x, t ) ∞ −λn2
⎛δ −x ⎞
Taw − T0
= ∑a e
n =1
n
δ22
cos ⎜ 2
⎝ δ2
λn ⎟
⎠
2 sin λn
where an =
λn + sin λn cos λn
and λn (n=1,2,…) are the roots of
(hg )eff δ2 k1 δ2
λn tan λn =
k2 k2 δ1
Graphically,
k1 δ2
For small ∆ ≡ , small λ1 , so tan λ1 λ1 , so
k 2 δ1
k1 δ2
λ12 ∆ λ1 ∆ =
k2 δ1
k /ρ c
α2 k1 δ2 2 2 2 = k1
and also a1 1 λ12 ≡τ
δ22 k δ1 2
δ2 ρ2c2 δ1δ2
2
from before
So, leading term is then
Taw − T2 ( x, t ) −
t
⎛δ −x ⎞
e τ cos ⎜ 2 λ1 ⎟
Taw − T0 ⎝ δ2 ⎠
1

which is what we found before. The other terms are much smaller, except at very
small time.
For thermal protection of solid rocket nozzles read sec. 14.2 (pp. 550-563) of
Sutton-Biblarz, 7th ed., especially, pp. 556-563.
A key concept is ablative materials. They contain a C-based homogeneous matl.

embedded in reinforcing fibres of strong (anisotropic) C. Best is C/C, strong
expensive fibre since nozzle can get to 3600 K, can be 2D or 3D. Also good is C or
Kelvin (Aramid) fibres +phenolic plastic resins (for large nozzles)
For the shuttle RSRM, the throat insert (C cloth phenolic) regresses ~ 1 inch/120 sec,
and the char depth is ~ 0.5° inch/120 s.

Film Cooling of Rockets
For application of data on slot-injected films, we need to define the initial film
thickness s, velocity uF , density ρF , or at least mass flux uF ρF .
i i i i
Assume we know the flow rates mc and mF , where mc is the “core” flow and mF the
“film” flow. We also know the fully-burnt temperatures and molecular weights
( Tc , TF ; Mc , MF ).
The areas occupied at the “fully burnt” section are not known; let them be Ac , AF .
From continuity,
i i
mc mc R
uc A c = = Tc (1)
ρc P Mc
P = Pc is common to both

i i
mF mF R
uF AF = = TF (2)
ρF P MF
and the total cross-section is known:
Ac + AF = A (3)
We need some additional information to find uF . The two momentum equations are
(neglecting friction):
duc dP ⎫
ρcuc + = 0⎪
dx dx ⎪⎪ duc duF
⎬ ρcuc = ρFuF
⎪ dx dx
duF dP
ρFuF + = 0⎪
dx dx ⎪⎭
uF duF ρ
= c (4)
uc duc ρF
Both, ρF and ρc , have been evolving as drops evaporate and burn. We make now
the approximation of assuming their ratio to remain constant (equal to the fully-
burnt value). Then (4) integrates to
uF2 ρc uF ρc
= = (5)
u2c ρF uc ρF
Substitute into the ratio (2)/(1)
i i
ρFuF AF mF ρ ρc AF mF
= i → F = i
ρcuc Ac ρc ρF Ac
mc mc
i
AF mF ρc
or = i (6)
Ac ρF
mc
ρFuF ρF
and also = (7)
ρcuc ρc
⎛ρu ⎞
This last ratio ⎜⎜ F F ⎟⎟ is called the “film cooling parameter”, MF :
⎝ ρcuc ⎠

ρF MF Tc
MF = = (8)
ρc Mc TF
The film thickness s (at complete burn up) follows from
AF = π ⎡D2 − (D − 2s ) ⎤ ⎫
2
⎢⎣ ⎥⎦ ⎪ 2
⎪ AF ⎛ D ⎞ 4s
⎬ = ⎜ ⎟ −1 (if s D)
⎪ A c ⎝ D − 2s ⎠ D
A c = π (D − 2s )
2
⎪
⎭
i
D AF D mF ρc
s = (9)
4 Ac 4 i ρF
mc
From Rosenhow & Hartnett, Chapter 17-B, we characterize film cooling by the
change it induces to the driving temperature ( Taw ) for heat flow. In the absence of a
0
film, Taw
⎛
= Tc ⎜1 + r
⎝
γ −1 2⎞
2
Mc ⎟ , and we calculate ( qw )No Film = hg Taw
⎠
0
( )
− Tw . The film
0 F F
changes Taw to Taw (lower, presumably). The lowest we could Taw to get is TF , so
we define a film cooling efficiency
0 F
Taw − Taw
η= (10)
Taw − TF
⎧⎪η = 0 F
if Taw 0
= Taw (no effect)
Limits: ⎨
F
⎪⎩η = 1 if Taw = TF (max imum effect)
If we can predict η , then
F
Taw 0
= Taw 0
− η Taw (
− TF ) (11)
and then
F
qw = hg Taw(− Tw ) (12)
where hg is computed as if there were no film. To predict η , we first computes the

parameter

1
−
x ⎛ µF ⎞ 4
ζ= ⎜⎜ ReF ⎟ (13)
MF s ⎝ µc ⎟⎠
where x is the distance downstream of the film injection (here we assume this is
from the burn-out section), and
ρFuF s
ReF = (14)
µF
and ρFuF = MF ( ρcuc ) , from before
From ζ , there are several semi-empirical correlations for η . A recommendation from

R & H is
2
1.9 Pr 3
η= (15)
⎛ cp ⎞ 0.8
1 + 0.329 ⎜ c ⎟ζ
⎜ cp ⎟
⎝ F ⎠
(or η = 1 if this gives >1)
which is supported by air data of Seban.
Example
TF 1 M ρ 0.8
Say = ; F = 0.8 → F = = 1.6 → MF = 1.6 = 1.265
Tc 2 Mc ρc 0.5
i i
mF 1 mF
= 0.1 → i =
i 9
m (0.01) mc
(0.0101)
Say D=0.5m x t − x com pl. com b = 0.5 m
P=70 atm=7.09 × 106 N / m2 ⎫

⎪ 7.09 × 106 × 0.020
Tc = 3200 K ⎬ ρc = = 5.33Kg / m3 ; ρF = 8.53Kg / m3
⎪ 8.314 × 3200
Mc = 20 g / mol; γ c = 1.2 ⎭
Mc = 0.2

8.314
uc = 0.2 1.2 × × 3200 = 253 m / s
0.02
1
uF = 253 = 200 m / s
1.6
8.53 × 200 × s
Say µF = 2 × 10−5 Kg / m / s → ReF = −5
= 8.53 × 107 s
2 × 10
ReF = 9.37 × 105
s=
D mF
i
ρc
=
0.5
×
1 1
= 0.0110 m
(8.51 × 10 ) 4
4 i ρF 4 9 1.6
mc OR 0.0101
0.000998
0.6
µF ⎛ TF ⎞
=⎜ ⎟ = 0.50.6 = 0.660
µc ⎜⎝ Tc ⎟⎠
1
0.5
( )
−
ζ= 9.37 × 105 × 0.660 4 = 1.282
1.265 × 0.0110
0.000998
(
8.51 × 104 ) (25.74)
epc µF
= 0.8 (say, rF rc ), Pr = 0.8
c pF µc
2
1.9 × 0.8 3
η= = 1.24 → η = 1
1 + 0.329 × 0.8 × 1.2820.8 0.368 0.361
(25.74 )0.8
So, this offers perfect film cooling, meaning
F Tc
Taw = TF = = 1600 K
2
(3200-0.361(3200-700)=2296 K)
If the wall is made of Cu, and is at Tw = 700K , the reduction in heat flow is
qFw 1600 − 700

= = 0.360
q0w 3200 − 700
⎛ 2296 − 700 ⎞
⎜ 3200 − 700 = 0.638 ⎟
⎝ ⎠
which can be decisive.

(This example shows one could get good film cooling with much less than 10% flow
in the film, maybe with only 2%).

Lecture 11: Radiation Heat Transfer and Cooling
Radiative Losses
At throat of a RP1-LOX rocket, evaluate radiation heat flux
Pc = 70 atm
Dc = 0.21m
Tc = 3500 K
O / F = 2.2
γ=1.25
M=25 g/mol
x co 0.3 8
xH2O = 0.3 1
xH2 0.1 4
x co2 0.1 1
γ
⎛ 2 ⎞ γ −1
Pthroat = ⎜ ⎟
⎝ γ + 1⎠
Pc = 38.85 atm
Pco = 14.8 atm
PH2 O = 12.0 at m
PH2 = 5.4 atm
Pco 2 = 4.3 at m
2
Tthroat = Tc = 3111K = 5600 R
γ +1
Assume slab if thickness L=0.9 Dt = 0.191m = 0.63 ft
(PL )co = 9.2 ft atm

(PL )H O = 7.5 ft atm
2
(PL )H = 3.4 ft atm

2
(PL )co = 2.7 ft atm

2

CO
Fig 4-22 for CO gas only to 2400 R

2 ft atm
So, extrapolated to 9.2 ft atm, ε (2400 R ) 0.1 at 2400 R. But ε falls rapidly with T.
If we conservatively extrapolate linearly in Log ε (T). εco would appear to go to ~
0.005 or so. Hence, even though the gas is CO-rich, radiation by CO is negligible.
H2O
At PT = 1atm , PL = 7.5 ft atm , Fig 4.15 gives εH2 o (5000 R ) = 0.18 , and extrapolating
a bit to T=5600R, εH2O 0.15 .
Fig 4.15 gives εw for Pw → 0 ,PT = 1atm . To correct for finite Pw and higher PT , use
Pw + PT
4.16. Here, for PwL = 7.5 ft atm , there is some significant effect of . We have
2
Pw + PT 12 + 38.9
= = 25.5 atm way beyond the graph.
2 2
Pw + PT
β= cw
2
0.5 1 (1)
0.8 1.18 (1.14)
1 1.23 (1.21)
1.2 1.28 (1.28)
0 0.57

( )
n
cw = c P + B
1 ⎫
n
0.57 = c Bn ⎫ ⎛ B + 0.5 ⎞ ⎛ 1 ⎞ ⎛ 1.28 ⎞
⎪⎜ ⎟ = ⎪ ln ⎜ ⎟ ln ⎜ ⎟
⎪⎝ B ⎠ 0.57 ⎪ ⎝ 0.57 ⎠ = ⎝ 0.57 ⎠
1 = c (B + 0.5)
n
⎬ ⎬ n=
n ⎛ 0.5 ⎞ ⎛ 1.2 ⎞
n ⎪ ⎛ B + 1.2 ⎞ 1.28 ⎪ ln ⎜ 1 + ⎟ ln ⎜ 1 +
1.28 = c (B + 1.2 ) ⎪⎭ ⎜ ⎟ = ⎪ ⎝ B ⎠ ⎝ B ⎟⎠
⎝ B ⎠ 0.57 ⎭
⎛ 1.2 ⎞
ln ⎜ 1 +
0.562 0.809 ⎝ B ⎟⎠ B = 0.105
n= = = 1.439 c = 1.175
⎛ 0.5 ⎞ ⎛ 1.2 ⎞ ⎛ 0.5 ⎞ n = 0.321
ln ⎜ 1 + ⎟ ln ⎜ 1 + ln ⎜ 1 +
⎝ B ⎠ ⎝ B ⎟⎠ ⎝ B ⎟⎠
( )
0.321
cw = 1.175 P + 0.105
Then, for P = 25.5 , cw = 3.33 → εw = 3.33 × 0.15 = 0.499 Suspect!

too much
CO2
For PL=2.7 ft atm, T=5600 R ε CO2 0.056
From fig 4.14, Correction cc 1.1 → εCO2 0.062
For interference, use Fig 4.17
Pw 12
= = 0.736
Pw + Pc 12 + 4.3
(Pr + Pw ) L = 7.5 + 2.7 = 10.2 ft atm
⇒ ∆ε 0.06
So εgas 0.499 + 0.062 − 0.06 0.5
This is likely to be an over estimate, because εH2O must saturate as PT increases, not
grow as PT0.3 .

With this ε g σ Tt4 = 0.5 × 5.67 × 10 −8 × 31114 = 2.66 × 106 W / m2
Compare to Convection:
Rs T
Say Tw = 1000 K c* =
δ
( ρu)0.8 cp µ0.2
hg = <> <>
(0.026 )
D0.2
t Pr0.6
8.314
× 3500
0.025
c* = = 1640 m s
0.658
3111 + 1000
< T >= = 2056
2
Pc 70 × 105
( ρu)e =
c*
=
1640
= 4269 Kg / m2 / s
3111
( ρu)<> = 4269
2856
= 6460 Kg / m2 / s
0.6
⎛ 2056 ⎞
µ<> 6 × 10−5 ⎜ ⎟ = 4.8 × 10−5 Kg / m / sec
⎝ 3000 ⎠
Sp = 1663 J / Kg / K Pr = 0.8
( )
0.2
64600.8 × 1663 × 4.8 × 10−5 × 0.026
hg = 0.2 0.6
= 345, 000 × 0.026 = 8960 W / m2 / K
0.21 0.8
( qw )conv 8960 (3500 − 1000 ) = 2.24 × 10 7 W / m 2
qrad
So = 0.12
qconv

As PT increases, each individual emission line is broadened by collisions, and ε
( )
increases. However, when PL is relatively large > 2.5 ft atm , Figure 4.14 shows the
∼
effect is small; this is because at that PL the bands are largely overlapped already
and only the broadening of their edges matters anymore. So, we ignore the PT effect.
Effect of Particulates 2R
2πR
For 1 , geometrical optics
λ
4
2πR ⎛ 2πR ⎞
For 1 , Rayleigh regime, particle appear to be smaller by ∼ ⎜ ⎟ .
λ ⎝ λ ⎠
For 3000 K, peak of spectrum at λ ∼ 1.2 µm
λ
R cross over = ∼ 0.4 µm
2π
Particles tend to be near (somewhat below) this value. For conservatism, assume
geometrical occultation.
L
−
mfp −npQPL
α = Prob. of absorption = 1-Prob. of transmission= 1 − e =1−e
x ρgas
nP = Qp = πRp2
1 − x 4π 3
R ρ
3 p s
x ρg 3 L
−
1 − x ρs 4 Rp
εp 1−e

Say L = R t = 0.3 m R p = 1 µ m = 10 −6 m
38.9 × 105 × 0.025

ρg = = 3.75 Kg / m3
8.314 × 3111
ρs = 3000 Kg / m3
x=0.3
0.3 3.75 3 0.3

−
εp 1−e 0.7 3000 4 10−6
= 1 − e−120 = 1
4 4
⎛ 2πRp ⎞ ⎛ 0.1 ⎞ 1
Now, suppose R p = 0.1 µm instead. Exponent has a factor ⎜ ⎟⎟ = ⎜ ⎟ = 256 ,
⎜
⎝ λ ⎠ ⎝ 0.4 ⎠
and has a 1 , which is another 1
= 1 0 → 2 5 .6 sm aller → 1 − e − 4.69 = 0 .9 9 1 still ∼ 1
Rp 0.1
In this case
(a) In flame looks solid (black body radiator)

(b) Radiative losses double ( ε =1 instead of 0.5), to ~20% of loss

Lecture 12: Review of Equilibrium Thermochemistry
Thermochemistry of Combustion for Propulsion Devices
12.1.1 Introduction
The general principles that govern chemical reactions (species conservation,

enthalpy balance and chemical equilibrium) have been introduced in previous
subjects. In the next three Sections we will briefly examine the application of these
principles to Propulsion.
The sketch in Fig 12.1 indicates the main regions.
Fig. 12.1. Main regions in a liquid-fuel combustor
In a combustion chamber which could belong to a ramjet or rocket, or, with

minor variations, to many other technical devices which utilize hot gas flows. It is
frequent to supply fuel (and sometimes oxidizer as well) in liquid form, and so the
first region involves a series of colliding liquid jets or sprays, designed to break up
these liquids into fine droplets. These now find themselves immersed in the hot
combustion gases, and evaporate rapidly, a step which must necessarily precede
actual chemical reaction. Mixing of the oxidizer and fuel vapors typically proceeds via
the strong turbulence existing in the combustor, and once this mixing has actually

achieved intimate molecular level interdiffusion within the turbulent eddies, reaction
occurs usually on a time scale which is short compared to those for vaporization and
mixing.
The combustion chamber design must take into account the kinetics of these
steps to arrive at a dimension which guarantees completion of the combustion
reactions without waste of volume. If this is so, the last portion of the combustor
contains at any time a mixture of combustion products which have reached a state of
thermodynamic equilibrium, and which are about to flow into their working devices,
be it a simple nozzle, a turbine (also preceded by its nozzles) or other configurations.
Equilibrium does not necessarily mean complete combustion since, at high

temperatures, various decomposition reactions are thermodynamically favored. In a
high speed Ramjet with a subsonic combustion chamber the incompleteness of the
combustion is so extreme –due to the high temperature prevailing- that, beyond a
certain flight speed, no net heat release can be obtained.
It is important to distinguish between this kind of combustion incompleteness,

which is unavoidable at high temperatures, and is due to the activation of reverse
reactions, and a more common but avoidable incompleteness, which occurs when the
combustion chamber is inadequately sized and the vaporization-mixing-reaction
processes are not completed in the residence time allowed to each fluid particle.
A first objective of our analysis is to determine the actual gas conditions

(temperature, composition) at the exit of an equilibrated combustor, for a set of
prescribed reactant flow rates. These flow rates, together with the discharge nozzle
geometry, also determine a combustor pressure, which we will here assume as an
independently prescribed quantity.
Beyond this point the flow typically expands and cools more or less
adiabatically. As it does so, there is at least a tendency to shift its chemical
composition, usually in the sense of completing the incomplete combustion, since the
lower temperature now inhibits decomposition reactions. Thus, there may be a sort
of afterburning effect along the nozzle expansion, and this sometimes contributes
significantly to increase the performance of rockets or ramjets. It is worth noting
here that, even if all the reactions that were not completed in the chamber were
completed in the nozzle, thus still releasing all the heat of combustion, the
performance level achieved would be lower than if no decomposition had ever
occurred in the first place; this is because in any thermodynamic cycle less work can
be extracted from heat added at lower than at higher temperatures. In any event, it
often happens that the nozzle recombination reactions are too slow (due to their
occurring at a reduced temperature, which slows down reaction rates) to keep pace
with the rapidity of the nozzle expansion process. In that case the flow is said to be
chemically frozen, and it comes out of the nozzle exit in a state of substantial non-
equilibrium, maintaining a composition close to what it had in the chamber. This is
the opposite extreme to the case when all the recombination reactions actually do
occur along the flow (equilibrium expansion). Actual performance always falls
between these two extremes (more performance than that for frozen expansion, less
than that for equilibrium expansion), and since the spread between them may not be
too large, it is useful to bracket the actual result by performing both frozen and
equilibrium flow calculations, both of which, as we will see, involve only consideration
of the end points of the expansion. Anything in between entails much more complex

calculations, due to the need to follow in detail the kinetics of a variety of chemical
reactions.
12.1.2 Combustors Far from Stoichiometry
Stoichiometric proportions of fuel and oxidizer are those for which, if the reaction
were complete, no excess of either would ensue. It is clear that this would then
ensure maximum heat release per unit mass, and hence highest adiabatic flame
temperature. Even under conditions of incomplete reaction, maximum temperature
usually does occur near stoichiometric conditions, and falls off for either fuel-rich or
fuel-lean conditions.
For some important types of combustors the fuel/air ratio is purposely selected to be
very far from stoichiometric, so as to keep the final temperatures below acceptable
materials limits. Examples are:
(a) Jet engine combustors (operated fuel/lean)
(b) Gas generators for rocket turbopumps (operated typically very fuel-rich).
In cases such as these the final temperature and other properties of the reacted
gases can be obtained from simple mass and energy balances, since their
composition can usually be obtained by inspection by assuming complete combustion
with excess oxidizer or fuel, as the case may be. This is because decomposition
reactions responsible for combustion incompleteness are not very active at these
moderate temperatures.
Example
The liquid hydrogen fuel in the Space Shuttle main engine is pressurized to the very
high combustion chamber pressure by a turbopump which is driven by a turbine. The
gas driving this turbine is generated in a “preburner”, into which the flow rates of
oxygen and hydrogen are (at maximum power)
• •
mL O X = 43 Kg/sec mL H = 40.2 Kg/sec
Since this is very fuel-rich, we can safely assume all of the O2 is consumed, and we
have the reaction
1
x H2 + O2 → H2O + (1 − x ) H2
2
where the x is given by the oxidizer/fuel ratio:
1
32
43 2 8
= = → x = 7.5
40.2 x .2 x
1
So that 7.5 H2 + O2 → H2O + 6.5H2
2

The adiabatic flame temperature corresponding to this reaction is obtained by stating
that the combustion process is adiabatic:
Enthalpy before reaction = Enthalpy after reaction
Before reaction, assuming the H2 and O2 enter the preburner very cold (say, for an
approximation, at O K), we read in Ref. 12.1,
ho2 = -8682 J/mole, hH2 = -8468 J/mole, for a total enthalpy
1
hbefore = 7.5 x (-8468) + (-8682) = -67,850 J
2
The temperature after reaction is such as to give this same enthalpy. With
some hindsight (we know the turbines are unlikely to be designed for more than
1100-1200K), we try 1000, 1100 K, and using the tables in VW-S (Ref. 12.1), App. A.
11 we obtain
T (K) 1000 1100
hH2O -241,827 + 25,978 -241,827 + 30167

hH2 20,678 23,723
h = hH2O + 6.5hH2 -81,390 -57,461
Interpolating linearly gives
13,540
T = 1000 + 100 = 1057 K
23, 939
The mean molecular mass of the gas is simply
18 x 1 + 2 x 6.5
M= = 4.13 g/mole
1 + 6.5
We can also calculate easily the mean specific heats cp and cv and their ratio, γ .
⎛ ∂h ⎞
For each of the components, H2o and H2 , we can approximate cp = ⎜ ⎟ as
⎝ ∂T ⎠
h (1100 ) − h (1000 )
cp . This gives ( cp ) = 41.89 J/mole K,
100 H2O
(c )
p H
2
= 30.37 J/mole K, so that
41.89 x 1 + 30.37 x 6.5

cp = = 31.91 J/mole K
7.5

Since for each constituent c v = cp − R , we also have
cv = cp − R = 31.91 − 8.31 = 23.60 J/mole K, giving
cp 31.91
γ = = = 1.352
cv 23.60
(Notice γ is not the average of the individual γ ’s)
12.1.3 Reactions Close to Stoichiometry (or with Preheated Reactants)
In other applications, where materials limits are either absent or can be overcome by
appropriate cooling of walls, the reactants may be in proportions close to
stoichiometric, so as to generate very high gas temperatures. Examples are:
(a) Rocket combustors (operated typically on the fuel-rich side, but only
moderately so)
(b) Ramjet combustors (where, in addition, the air is strongly heated by the
ram effect)
(c) Piston engines (fuel-lean for gasoline engines, but near-stoichiometric in
Diesel engines)
(d) Welding torches
In these cases one cannot easily predict the final composition of the burnt gases,
since the relatively complex molecules like H2o , CO2 (or even H2 , O2 ) tend to
break down into simpler ones which are more stable at high temperatures (such as
OH, H, O, etc.). Thus, in addition to mass and energy conservation, one needs to use
also the laws of chemical equilibrium to determine the temperature and other
properties of the gas products. In most real situations, many molecular species may
be present at the high temperatures of the flame, and so more than one equilibria
are simultaneously to be considered. In these cases, iteration is required to obtain
consistent solutions. A procedure which, with some judgment, usually works, is as
follows:
(1) Identify the possible molecular species in the products, classify them into
major and minor (minor species would be those that are expected to be
present in small concentration).
(2) Neglect at first the “minor” species, use procedures similar to those in the
previous example to obtain the “major” species concentration and an
approximate temperature.
(3) Write chemical reactions which would produce one “minor” species at a
time from only “major” ones. Calculate their equilibrium constants at the
approximate temperature of step (2), and use the equilibrium laws to
calculate the concentrations of the “minors”.
(4) Correct the “majors” concentration, and do a new enthalpy balance to

obtain an improved temperature.

(5) Iterate back to step (3) until convergence is obtained.
We illustrate these principles below.
Ref.12.1.1 G. Van Wylen and R. E. Sonntag, Fundamentals of Classical

Thermodynamic, Appendix A. 11, 3rd Ed., SI Version, J. Wiley & Sons, 1985.
12.2 Example Calculation of Thermochemical Equilibrium in a Rocket Chamber
12.2.1 Introduction
In the Space Shuttle’s Main Engines (SSME’S), the rocket combustion chamber
operates at p = 210 atm and an oxidizer-to-fuel mass ratio O/F = 6, the reactants
being cryogenic H2 and O2 . The stoichiometric ratio for full combustion
⎛ 1 ⎞
⎜ H2 + 2 O2 → H2O ⎟ would be O/F = 8, so this is now only slightly fuel-rich, and we
⎝ ⎠
expect a very high combustion temperature. In the next few sections we will analyze
this case step-by-step, in order to bring out the important points of any such
calculation. Practical problems of this nature are routinely solved by means of
complex computer programs based on the present methods or on more generalized
iteration schemes, but, of course, entailing the same basic physics. One such
standard code is the so-called CEC, Chemical Equilibrium Code, developed and
maintained by NASA (Ref. 12.2), which in its normal configuration, can handle about
200 species simultaneously.
12.2.2 Atomic Species Concentration
In order to more easily keep track of mass conservation, it is useful to work with the
numbers of moles, ni of the various chemical product species in an arbitrary total
mass of gas. This same mass of gas originates from a set of numbers of moles n j of
reactants, and we must impose that, no matter what reaction shifts occur, the
number of atomic moles for each kind of atom (O, H, N, C, ……) is the same before
and after reaction. As an example, suppose we have combustion of hydrogen in
oxygen, with unknown amounts of water, molecular and atomic oxygen, molecular
and atomic hydrogen and hydroxyl (OH) radicals formed. The oxidizer/fuel ratio is
prescribed to be, say, R, so that, if we arbitrarily select NO2 = 1/2 , then
32 x 1 / 2 8
=R ; NH2 = (12.2-1)
2NH2 R
The reaction is
8 1
H2 + O2 → nH2O H2O + nO2 O2 + nO O + nH2 H2 + nH H + nOH OH (12.2-2)
R 2
Note that the list of candidate reaction products is arrived at from our
background knowledge of what species actually do occur in significant quantities. For
instance, we might not include O2 or O if R were low enough to make the mixture
fuel-rich, such that under those conditions, oxygen species would be mainly depleted.

However, if we still did include them, with the proper equilibrium constants,
we would still find some small concentrations of these molecules present. We would
simply ignore them for practical purposes, as we ignore other exotic, but potentially
real species, like HO2 , H3 , etc.
The atomic species are H and O and in order to conserve atoms, we must impose:
16
H-conservation: = 2nH2O + 2nH2 + nH + nOH (12.2-3)
R
O-conservation: 1 = nH2O + 2nO2 + nO + nOH (12.2-4)
12.2.3 Equilibrium Relations
These are two equations involving the six unknown ni ’s. Clearly because of
the presence of the “decomposition products” O, H and OH, (plus O2 if fuel-rich or
H2 if lean), conservation alone cannot tell us the composition of the burnt gas. We
need four extra relationships, and these are the equilibrium conditions (laws of
mass-action) for four elementary reactions selected so as to involve these
decomposition products, as well as some of the “major” products, H2O , CO2 , H2 .
These could be, for example,
1
H2O R H2 + OH (12.2-5)
2
H2 R 2H (12.2-6)
1
H2O R H2 + O2 (12.2-7)
2
O2 R 2O (12.2-8)
or any linear combination of them.
For each of the elementary reactions above, an equilibrium relationship must

be satisfied. According to Ch. 16, these take the form
PH12/ 2 POH
= K1 ( T ) (12.2-9)
PH2O
P H2
= K2 ( T ) (12.2-10)
PH
PH 2 PO12/ 2
= K3 ( T ) (12.2-11)
PH2O

PO2
= K4 ( T ) (12.2-12)
PO2
where each Pi represents a partial pressure. These can be expressed in terms of the
given total pressure P and the mole numbers ni :
nH2O
PH2O = P ( n = nH2O + nO2 + nH2 + nH + nOH )
n
nO2
PO2 = P
n
nOH
PO H = P (12.2-13)
n
.
.
.
And therefore the equilibrium relations become
n1H2/ 2 nOH n
= K1 ( T ) (12.2-14)
nH2O p
nH2 n
= K2 ( T ) (12.2-15)
nH2 p
nH2 n1O2/ 2 n
= K3 ( T ) (12.2-16)
nH2O p
n2O n
= K4 ( T ) (12.2-17)
nO2 p
The equilibrium constants K1 , K2 , K3 , K4 still depend on temperature. If

temperature is known a-priori (or assumed at some step in an iteration process), are
can either look their values up in some compilation of chemical properties (such as
Table A.12 of Ref. 12.1), or more rationally, given the many possible combinations of

elementary reactions, they can be built up from tabulated values of the standard
chemical potentials (or Gibbs free energies) of the intervening species. The
equilibrium constants are given in our example by
⎛ 1 0 0 0 ⎞
⎜ 2 µH2 + µH O − µH2O ⎟
K1 = exp ⎜ − ⎟ (12.2-18)
⎜⎜ RT ⎟⎟
⎝ ⎠
⎛ 2µH0 − µH0 ⎞
K2 = exp ⎜ − 2 2
⎟ (12.2-19)
⎜ RT ⎟
⎝ ⎠
⎛ 0 1 0 0 ⎞
⎜ µH2 + 2 µO2 − µH2O ⎟
K3 = exp ⎜ − ⎟ (12.2-20)
⎜⎜ RT ⎟⎟
⎝ ⎠
⎛ 2µ0O − µH0 ⎞
K 4 = exp ⎜ − 2
⎟ (12.2-21)
⎜ RT ⎟
⎝ ⎠
where each µi0 ( T ) is the chemical potential of species i at 1 atm. pressure (standard
chemical potential), at the given temperature T, and R is the universal gas constant
in consistent units.
12.2.4 Conservation of Energy
Since the reactions occur at constant pressure, if we assume a well-insulated

combustor, the enthalpy of the products must equal that of the reactants. If hi ( T ) is
the molar enthalpy of the ith species at temperature T, we must have in our
example:
8 1
hH2 ( Ti ) + hO2 ( Ti ) = nH2O hH2O ( T ) + nO2 hO2 ( T ) + nO hO ( T )
R 2
+ nH2 hH2 ( T ) + nH hH ( T ) + nOH hOH ( T ) (12.2-22)
8
Notice the left hand side of this equation represents the enthalpy of the reactants (
R
1
moles of H2 , mole of O2 ) at their injection temperature, Ti , whereas the right
2

hand side represents that of the products, with each hi evaluated at T, the final
equilibration temperature, to be found. This is what is often called the “adiabatic
flame temperature” for the given O/F and pressure.

Lecture 13: Examples of Chemical Equilibrium
13.1 Numerical Iteration Procedure
For the Shuttle main engine, we take R = O/F = 6, and P = 210 atm. To get started,
we neglect nO2 , nO , nH and nO H , so that Eqs. (12.2-3) and (12.2-4) reduce to
8
2 n H2O + 2 n H2 =
3
nH2O = 1
which give nH2O = 1 , nH2 = 1/3 and (since all other ni ’s are taken to be zero, n = 4/3).
The enthalpy before reaction was (assuming very cold reactants, i.e. Ti = 00 K ),
4 1 4 1
H= hH2 ( O ) + hO2 ( O ) = -15,632 J (for the moles of H2 , mole of O2 ).
3 2 3 2
For the products to have the same enthalpy, the temperature must be very
high. Using the table in Ref. 12.1 (pp. 692-693) we have the trial values tabulated
below:
T(K) 3400 4000 4200
hH2O (J/mole) -92,973 -58,547 -46,924

hH2O (J/mole) 103,738 126,846 134,700
1 -58.394 -16,265 -2,024
hH2O + hH
3 2
Since we must obtain an enthalpy of -15,632 J, these values indicate a

temperature very close to 4000 K. Linear interpolation between 4000 and 4200 K
gives
T = 4009 K
We can now use this temperature to calculate the four equilibrium constants
K1 through K4 . These are given in Table 12 of Ref. 12.1. With some interpolation, we
obtain:
K1 = 0.974 ( atm)
1/2
K2 = 2.61 ( atm)
K3 = 0.589 ( atm)
1/2
K 4 = 2.286 ( atm)

At this point we can obtain our first nonzero estimate of the “minor” species
concentrations. From Eqs. (12.2-14) through (12.2-17),
n H2O n 1 4 /3
nO H = K1 = 0.974 = 0.138
n 1H2/ 2 p 1/3 200
nnH 2 4 /3 x 1/3
nH = K2 = x 2.61 = 0.076
P 200
2
⎛ nH O ⎞ n 2 ⎛ 1 ⎞2 4 / 3
(0.589 ) = 0.021
2
nO2 =⎜ 2 ⎟ K =
⎜ nH ⎟ p 3 ⎜⎝ 1 / 3 ⎟⎠ 200
⎝ 2 ⎠
nnO2 4 / 3 x 0.021
nO = K4 = x 2.286 = 0.018
P 200
We are now at the end of the first loop of our iteration procedure. We can
next re-calculate nH2O and nH2 from the atom conservation equations, including the
new “minor” ni ’s we just computed. With the new complete set of ni ’s, we can re-
calculate the enthalpy at a few temperatures and interpolate for a new set of K j ’s,
etc. We will do one more cycle in some detail to illustrate the nature of the typical
results and the way they may tend to oscillate or diverge. Beyond that, a tabulated
summary of the succeeding results will suffice.
Corrected nH2O , nH2 , n: From Eqs. (12.2-3) and (12.2-4),
4 0.076 + 0.138
nH2O + nH2 = − = 1.226
3 2
nH2O = 1 − 2 x 0.021 − 0.018 − 0.138 = 0.802
Hence nH2 = 1.226 − 0.802 = 0.424
and n = 1.226 + 0.138 + 0.076 + 0.021 + 0.018 = 1.480

Corrected temperature:
Since the partial decomposition of H2O and H2 into the minor species is an
endothermic process, the new temperature will be lower. Using the tables of
enthalpies and the computed mole numbers we calculate
h(2800 K) = 0.802 x (-126,533) + 0.424 x 81,370 + 0.138 x 121,729 + 0.076 x

269,993 + 0.021 x 90,144 + 0.018 x 301,587 = -22,338 J
h(3000 K) = 0.802 x (-115,466) + 0.424 x 88,743 + 0.138 + 129,047 + 0.076 x

274,148 + 0.021 x 98,098 + 0.018 x 305,771 = -8,769 J
Since we still want h = -15,632 J, linear interpolation gives a temperature
200
T = 2800 + ( −15, 632 + 22, 338) = 2899 K
−8, 769 + 22,338
It is clear that at this new, much lower temperature, there will be much less of the
minor species, since the new equilibrium constants, ( K1 through K4 ) will be smaller.
Since it was the presence of these minor species that forced a reduction from 4009 K
to 2889 K, we would next obtain a refined T again relatively high, and the process is
likely to overshoot at each iteration step. This suggests that we can accelerate
convergence by adopting as a new trial temperature the average of the last two:
4009 + 2899
T = = 3454 K
2
Corrected minor species: The new equilibrium constants (at 3480 K) are:
K1 = 0.251 ( atm)
1/2
K2 = 0.316 ( atm )
K3 = 0.184 ( atm) K 4 = 0.454 ( atm)

1/2
and, proceeding as before,
nO H = 0.0250 nH = 0.0297 nO2 = 0.0079 nO = 0.00105
Table 13.1 summarizes these two iterations, and shows the next few iterations as
well, leading to the practically converged values of the last line. The mole numbers
are then converted to mole fractions by simply dividing each by n (Table 13.2).
Notice that the numbers of mole given in Table 13.1 have been converted to mole/kg,
by dividing the n numbers used so far (which correspond to
8
x 0.00108 + 1 x 0.016 = 0.01869 Kg of reactants) by this mass 0.01869 Kg.
3

ITERATION NH2O NO2 NO NH2 NH NO H T (K) T + TOLD K1 K2 K3 K4
(mol/kg) 2 ( atm)
1/2
( atm ) ( atm)
1/2
( atm )
1 53.57 0 0 17.86 0 0 4009 0.974 2.610 0.589 2.286

2 42.96 1.125 0.964 22.71 4.071 7.393 2899 3454 0.236 0.281 0.173 0.190
3 52.09 0.423 0.056 17.87 1.591 1.339 3784 3619 0.356 0.587 0.259 0.432
4 50.42 0.199 0.172 18.76 1.912 2.587 3644 3632 0.370 0.619 0.267 0.458
5 50.48 0.182 0.171 18.66 2.025 2.555 3642 3637
Revised Table 13.1
The mean molecular weight of the gas is then obtained from Table 13.2 by:
M= ∑
i
xi Mi = 13.48 kg / Kmole
An approximate molar, specific heat for each species can also be obtained as
cpi ⎡⎣h (3800) − h (3400) ⎤⎦ / 400 , and these can also be averaged to obtain
J cal
cp = ∑x c
i
i pi = 50.62
( mole ) K
= 12.11
( mole )K
and then the “average” specific heat ratio is
cp 50.62
γ= = = 1.196
cp − R 5.62 − 8.31
Finally, for purposes of calculating the flow in the rocket nozzle, it is useful to
determine the entropy of the reacted gaseous mixture. We can refer this to a unit of
mass, or, alternatively, to the same arbitrary amount of mass we have so far worked
4 1
with, i.e., moles of H2 and mole of O2 (we should not calculate it per mole,
3 2
since the number of moles in this much mass may change from the present value n
= 1.384, due to reactions occurring in the nozzle). Table A. 11 of Ref. 12.1 gives
values of the specific entropies s0 of the various species at 1 atm, as a function of
temperature. We can easily correct these to the proper pressures by using
Si ( T,Pi ) = Si0 ( T ) − R ln Pi (13.1)
where Pi = P xi is the corresponding partial pressure. The results are summarized in

Table 13.3.

TABLE 13.3 Entropies at 3630 K
Species Entropy at 1 atm Partial Pressure Entropy at own

KJ / (Kmole)K atm pressure KJ /
(Kmole)K
H2O 293.905 142.8 252.66
H2 210.070 52.9 177.08
OH 261.771 7.4 245.13
H 166.668 5.9 151.91
O2 292.223 0.58 296.75
O 213.715 0.48 219.82
The total entropy in our control mass (4/3 moles of H2 plus 1/2 mole of O2 , in the
form of the Ni ’s listed in the last line of Table 12.2.1) is then
S= ∑s N
i
i i = 17.050 KJ / Kg / K
Ref. 12.2 “Computer Program for Calculation of Complex Chemical Equilibrium

Compositions, Rocket Performance, Incident and Reflected Shocks and Chapman –
Jongnet Dectorations”, by S. Gordon, NASA Accession No. M84-10621, NASA Lewis
Research Center.

Lecture 14: Non-Equilibrium Flows
Reacting Nozzle Flow
14.1 Introduction
As indicated in the introductory discussion of section 12.1, the actual

expansion process in a rocket or ramjet nozzle is intermediate between the extremes
of “frozen” and “equilibrium” flow, with the latter producing higher performance due
to recovery of some of the chemical energy tied up in the decomposition of complex
molecular species in the chamber - a kind of afterburning effect.
The two limits, “frozen” and “equilibrium” flow share an important property:
both are isentropic flows (if we ignore friction or heat losses). This is because, in the
frozen case no chemical change during expansion), there are no rate processes at all
occurring, the molecules preserving their identity all the way, while in the equilibrium
case, in which reactions do occur, their rate is so high (compared to the expansion
rate) that conditions adjust continuously to maintain equilibrium at the local pressure
and enthalpy level, with the result that the whole process can be regarded as
reversible (and hence isentropic). For any more realistic intermediate conditions, in
which reactions may proceed at rates comparable to that of the expansion, these
finite rate reactions produce an irreversibility, and a consequent entropy increase.
In this Section we will discuss in detail an example of each, frozen and

equilibrium expansion. We will use for this purpose the results of the equilibrium
chamber calculations of Sec. 13.1, relative to the space Shuttle Main Engine.
14.2 Frozen Flow Calculation
The simple ideal gas model we have used throughout most of this course
(constant molecular mass, constant specific heats) is an example of a frozen flow
model, since it is implied by these assumptions that no chemical change takes place.
Thus all of our “constant γ ” results belong in this category, and can be used as a first
approximation for nozzle flow calculations (using for instance the value of γ and M
computed for the combustor). However, even with no chemical changes, the specific
heats of the various molecules do change with temperature, generally decreasing as
temperature decreases in the range encountered in nozzles. This means that
γ = Cp Cv = Cp ( Cp − R g ) is not a constant, since R g , the gas constant, does not vary
due to the constancy of the molecular mass.
For a more precise calculation, we make use of the constant-entropy

condition for frozen flow. Suppose, for instance that the nozzle exit pressure is
specified. This pressure, together with the chamber entropy are enough to determine
all other thermodynamic variables of the frozen gas at the exit plane, in particular its
enthalpy he per unit mass. Since the chamber enthalpy ho is already known, we can
find the exit velocity by the steady-state energy equation:

1 2
h0 = he + ue
2
ue = 2 (h0 − he ) (14.1)
The process by which h is found once P and S are given depends on the data
available. Tables or graphs (Mollier charts) are the simplest method, but these are
unlikely to exist for the particular composition of interest. More fundamentally, one can
repeat the steps at the end of Sec. 13.1, when S was calculated after the gas
composition was converged upon. In this case, the composition (i.e., xH2 , xH , xOH , etc)
is fixed throughout ("frozen"), and for each P, we will try various temperatures until S,
the entropy, equals the chamber value.
In addition to the exit conditions, it is usually of interest to calculate the

throat conditions, since it is the throat area that determines the mass flow rate. Two
alternative procedures can be followed for this:
(a) Try a range of pressures about 1/2 the chamber pressure, calculate the
corresponding u = 2 (ho − h) and density ρ , and look for a maximum of ρu.
This occurs at the throat.
(b) For the same P range, calculate the local speed of sound, find where it is
equal to the local velocity u. The local speed of sound is given in general by
∂P
a= (14.2)
( ∂ρ )s
and for frozen flow this can be shown to have the familiar form
a = γR g T (14.3)
1
Conditional Throat : A =
ρu
,u= (
2 hc − h )
dA dρ du
= − − = 0 at throat
A ρ u
S = const.
1
dρ
du 1 dh 1 ρt dρ
= − = − = − 2
ut 2 hc − ht 2 hc − ht ρ t ut

dρ du dρ dp ut
So, = − = 2
= ut
2
→ =1 Mt = 1
ρt ut ρ tut dρ ⎛ ∂P ⎞
⎜ ∂ρ ⎟
⎝ ⎠s
(both, equil. and frozen)
Frozen and Equil. Speed of Sound
2 ⎛ ∂ρ ⎞ 1 RsT
a = ⎜ ∂ρ ⎟ c ρ dT = Tds + dρ = dρ
⎝ ⎠s ρ ρ
P
and = R gT
ρ
dp dρ dT
Frozen: R g = const − =
p ρ T
⎛ dP dρ ⎞ R g T dp dρ
Cρ T ⎜ P − ρ ⎟ = P dP ( −Rg + Cp ) = Cp
ρ
⎝ ⎠ p
Cv
⎛ dp ⎞ Cp P ⎡ 2 ⎛ ∂ρ ⎞ ⎤
⎜ dρ ⎟ = C ρ = γ R g T → ⎢a = ⎜ ⎟ = γ ( T ) R g T ⎥
⎝ ⎠s v ⎣ ⎝ ∂ρ ⎠s ⎦
γ (T)
R dp dρ dΜ dT
Equil. : R g = − + =
M p ρ Μ T
⎛ dp dρ dΜ ⎞ R g T
CP T ⎜ p − ρ + Μ ⎟ = p dp
⎝ ⎠
dp ⎛ dρ dΜ ⎞
Cv = CP ⎜ ρ − Μ ⎟
p ⎝ ⎠

⎛ dp ⎞ P ⎛ ρ dΜ ⎞
⎜ dρ ⎟ = γ ( T ) ρ ⎜ 1 − Μ dρ ⎟
⎝ ⎠s ⎝ ⎠
In equil., during expansion Μ ↑ (recombination)

dΜ P
and ρ ↓ (expansion) →
2
< 0, a > γ
dρ ρ
where the only novelty is that γ itself is variable, and must be taken to be the local
γ.
A general calculation procedure can be as follows:
(1) Given P0, T0, S0, all xi’s, R

(2) Specify an exit pressure P0
(3) Calculate T0=T(P0, s0; xi)
(4) Calculate he, ρe, ue= 2 (h0 − he )
(5) Calculate the specific impulse. For a rocket,
F mue + (Pg − Pa ) A e Pg − Pa
g Is p = •
= = ue + (14.4)
ρe ue A e ρe ue
m
where Pg is the external (ambient) pressure? All the above is independent

i
F
of size. If a particular thrust F is desired (or an m = ), we must also
gIs p
find the required throat area A* :
(6) Locate Px using one of the procedure (a) or (b) above.

(7) Calculate ρ *u* = ( ρ u)MAX , i.e., from method (a) above.
•
(8) A =* m
;
Ag
=
( ρ u)MAX .
A* ρe ue
( ρµ ) MAX
F
(9) Thrust coefficient = oF =
P0 A*

14.3 Frozen Flow example (Shuttle Main Engine)
Using a simple computer program, and starting from the chamber conditions
established in Sec 13.1, the procedures above lead to the results shown in Table
14.1. We have assumed matched exit conditions throughout (Pg = Pe)
Pe/P0 0.001 0.0005 0.0002

Pe (atm) 0.21 0.105 0.042
Te (K) 998 862 707
ue (m/sec) 4227 4320 4421
Isp (sec) 431 440 451
Ae/A* 62.9 106.3 213.0
P* (atm) 118.6 118.6 118.6
T* (K) 3316 3316 3316
TABLE 14.1 Frozen flow performance of nozzle,

from P0 = 210 atm, T0 = 3640K, (LOX – LH2, O/F = 6).
For reference, the Shuttle nozzle has an area ratio of A e A* = 76.5, which
would give a (matched) frozen specific impulse of about 435 sec, according to Table
14.1.
If we had simply used the “constant γ ” approximation, using (from Sec.

13.1) γ =1.191 and M = 13.48 g/mole, the exit velocity would have been simply:
⎡ γ −1
⎤
2γ R ⎢ ⎛ Pe ⎞ γ ⎥
ue = T0 1 − ⎜ ⎟
γ −1 M ⎢ ⎥
⎢⎣ ⎝ P0 ⎠ ⎥
⎦
For the case of Pe / P0 =0.0005, this would have given an exit velocity = 4316
m/sec, which is quite close to the 4320 m/sec shown in Table 14.1. Thus, in this
case, constant γ is a good model, but this may not be true in even higher
temperature cases, like in electrically heated gases (arc-jet rockets).
14.4 Equilibrium Flow Calculation
Here we must impose at each pressure below P0 the same equilibrium

conditions that were used in Sec. 13.1, with the difference that the entropy, not the
enthalpy, is now prescribed. The mole fractions ( xH2 , xH , xOH , etc .) are now
variable along the expansion, and so are therefore the molecular mass, the specific
heat and γ . The calculational difficulty resides precisely in the need to perform these
repeated equilibrium calculations, but, once again, computer programs are available
to ease the burden.
Otherwise, the procedure is entirely analogous to that outlined in Sec. 14.2

(steps (1) through (9)). The one noteworthy difference is that the speed of sound, if

required, cannot be calculated as the local value of γ R g T , but must be found from
the basic Eq. (14.2). Also, of course, the xi values are understood to be unknown in
principle, and are, in fact results of each local equilibrium calculation.
14.5 Equilibrium Flow Example (Shuttle Main Engine)
For the same chamber conditions as in Sec. 14.3 and once again assuming
matched exit conditions throughout, the results of a calculation using a simple
computer program are as shown in Table 14.2:
Pe/P0 0.001 0.0005 0.0002

Pe (atm) 0.21 0.105 0.042
Te (K) 1216 1058 877
(X )
H2 e
0.250 0.250 0.250
(x )
H2O
e
0.750 0.750 0.750
(x )
O2
e
0 0 0
( x0 )e 0 0 0
( x0H )e 0 0 0
( x H )e 0 0 0
Ug (m/sec) 4407 4511 4626

Isp (sec) 449 460 472
Ae / A* 69.7 118.5 239.4

P* (atm) 120.7 120.7 120.7
T* (K) 3422 3422 3422
(x ) 0.2502 0.2502 0.2502

*
H2
(x ) 0.6968 0.6968 0.6968

*
H2O
(x ) 0.0015 0.0015 0.0015

*
O2
( x0 ) 0.0010 0.0010 0.0010

*
( x0H ) 0.0282 0.0282 0.0282

*
(x H ) 0.0222 0.0222 0.0222

*
TABLE 14.2 Equilibrium flow nozzle performance

LOX-LH, O/F = 6, P0 = 210 atm, T0 = 3640 K.

Once again, for the Shuttle, (Ae/A* = 76.5), we would obtain Isp 451 sec in this
case, compared with 435 sec for the frozen flow case. If we want to find the specific
impulse in vacuum rather than at the matched pressure point, we would simply add
Pg
ρg ug g
(see Eq. (14.4)) and we would obtain ISP( )
vac, equil
= 465 sec. This is indeed
quite close to the actual performance of the Shuttle main engines, indicating that
equilibrium does prevail during expansion. This is a result of the high pressure and
large size of these engines, and would most likely not be the case in a smaller, lower
pressure engine.
In addition to the higher exit velocity in the equilibrium case, other interesting
differences between the results of Tables 14.1 and 14.2 are:
(a) The exit temperatures are higher in the equilibrium case by about 200 K
(a result of the afterburning).
(b) The exit area ratios are larger in the equilibrium case (for a given
pressure ratio). This is also due to the same “reheating” effect, which
produces more volume increase.
Notice, finally, that the “constant γ ” approximation would still give ug = 4316
m/sec for the case of Pe/P0 = 0.0005; this is now well below the 4511 m/sec shown
in Table 14.2 for this case. Thus, constant γ may seriously underestimate
performance in large, high pressure rockets.

Lecture 15: Selection of Propellant Mixtures
Solid Propellant Rocket Fundamentals
Solid Propellants Rockets
Read Sutton ch.11 → basic performance
⎛•⎞
Surface regression rate ⎜ r ⎟ , empirically correlated to gas pressure as
⎝ ⎠
• • Pc A t
r = apnc m = ρp a Pcn Ab =
c*
(n < 1)
1
⎛ A ⎞ 1− n
Pc = ⎜ ρp a c* b ⎟ n<1 for stability
⎝ Ad ⎠
Booster motor
Space Engine (IUS)
Tactical motor

B. Double – bore propellants: Homogenous, Nitroglicerine/Nitrocellulose + additive
C. Composite propellants: Ammonium Per chlorate (AP) + Rubber binder (fuel) +

Aluminum
Ex. Fig 11.7 Sutton. AP-CMDB 30% AP (150 µm ) n ~ 0.4
0.4
⎛ P ⎞
r (100 atm) 1 cm/s r 0.01 ⎜ c 7 ⎟ a=1.58 x 10-5
⎝ 10 ⎠
C* = 1600 m/s ρp = 0.0636 lb/m2 = 1760 Kg/m2
Want Pc = 50 atm 5 x 106 N/m2
( )
0.6
Ab 5 × 106
= = 235 cannot use end – burn
At 1760 × 1.58 × 10−5 × 1600
Grain configuration
For Solid Rocket Components and Motor Design
Read Sutton, Chapter 14

Lecture 16: Solid Propellants: Design Goals and Constraints
Solid Propellants
Read Sutton’s, Chapter 12
Double Base (DB) Nitrocellulose + Nitroglycerine + Additives (for opacity, plasticity,

…). Both NC and NG are explosives, dangerous sometimes
JPN NC 51.5%, NG 43%, Diethyl phthalate 3.2%, Ethyl centralite 1%, H2SO4 1.2% +
carbon black + candelilla wax
Composite Modified Double base (CMBD) DB + Ammonium perchlorate (AP) or

Aluminium (Al)
Composite (C) AP (sometimes A Nitr) + Synthetic Rubber binder (fuel) + Al. Safer
than DB
Other Composite contain nitramine explosive (RDX, HMX), replacing some AP
Type Isp (s) at Tc (K) ρp Al % r (cm/s) η Fabrication

1000/14.7 psi 3 @1000 psi
(g/cm )
DB 220-230 2530 1.60 0 1.14 0.3 Extruded

DB-AP-Al 260-265 3866 1.79 20-21 1.98 0.4 Extruded
CTPB/AP/Al 260-265 3370-3490 1.76 15-17 1.14 0.4 Cast
HTPB/AP/Al 260-265 3370-3490 1.85 4-17 1.02 0.40 Cast
The addition of Aluminum is not necessarily beneficial, as the following example

shows:
Problem 3. Adding Aluminum to the formulation of a solid rocket propellant increases

the gas temperature, but incurs performance penalties related to the solid particles
that are generated.
Consider a simple model for the effect of adding a mass fraction xAl of Aluminium, of
the form
Tc
= 1 + rx ; x= 1.85 x Al (1)
Tc o
where r ≅ 1.41 is a separately calculated coefficient, Tco is the flame temperature

without aluminum (~2500K), and x is the solids fraction in the gas (the 1.85 factor
accounts for the oxygen in the Al2O3 particles).

Consider also a linearized model for the effect of the particulates, of the form
ue
= 1 − fx (at fixed Tc ) (2)
ueo
where f is as derived in class:
⎧ γ −1
⎪⎪ 1 − cs ⎡1 + (1 − η) ln(1 − η) ⎤ ; η = 1 − ⎛ Pe ⎞
γ
f = ⎨ 2 2c ⎢⎢ ⎥ ⎜ ⎟ (small particles)
pg ⎣ η ⎥⎦ ⎝ Pc ⎠
⎪
⎪⎩ 1 − − − − − − − − − − − − − − − − (l arg e particles)
(a) Show that the optimum loading is given by
r − 2f xO PT
xO PT = ; ( x Al )OPT = (3)
3 rf 1.85
Pe J
(b) For = 0.01, γ g = 1.25, Mg = 18g / mol, cs = 1260 , calculate ( x A l ) for
Pc KgK O PT
both small and large particulates. Comment on results.
Problem 3 – Solution
Ignoring the exit pressure effect (or at matched conditions),
⎡ γ −1
⎤
⎛ P ⎞ γ
g Isp = ve = 2Cp Tc 1 − ⎜ ⎟ ⎥
⎢ e
⎢ ⎥
⎢⎣ ⎝ Pc ⎠ ⎥
⎦
which is proportional to Tc . The rest of the dependence (γ, cp ) are affected by

particulates, but that is counted separately in the loss analysis. So we have
(counting both effects)
(a) ve ∼ (1 − fx ) 1 + rx
To optimize, take the logarithmic derivative and equate to zero

−f 1 r
+ =0
1 − fx 2 1 + rx
2f (1 + rx) = r (1 − fx) r - 2f = 3rfx
r - 2f xOPT
xOP T = and then ( x Al )OPT =
3rf 1.85
(b) For small particulate f =

1 ⎪⎧ cs ⎡ (1 − η) ln (1 − η) ⎤ ⎪⎫ , and using the given
⎨1 − ⎢1 + ⎥⎬
2 ⎪⎩ cpg ⎢⎣ η ⎥⎦ ⎭⎪
values,
1.25 8.314
η = 1 - (0.01)0.25/1.25 = 0.6012 ; Cpg = = 2309 J / Kg / K
0.25 0.018
1 ⎧ 1260 ⎡ 0.3988 ln(0.3988) ⎤ ⎫

f = ⎨1 − ⎢1 + ⎥ ⎬ = 0.3934
2 ⎩ 2309 ⎣ 0.6012 ⎦⎭
1.41 − 2 × 0.3934
xO P T = xO P T = 0.3746
3 × 1.41 × 0.3934
and then the Aluminum fraction should be
0.3746
( X Al )O P T = = 0.2025 20.3% Al loading
1.85
For the case of larger particle, the class derivation showed f=1, and so
1.41 − 2 × 1
xOPT = < 0
3 × 1.41 × 1
This nonsensical result simply means there is no good Al loading in this case.
The losses due to the particles are stronger than the gains due to increased
temperature, so no Aluminum should be added.
Fortunately, the particles are small, not large.

Lecture 17-18: Solid Propellants: Other Topics
Combustion of Solid Propellants
For a general discussion, read Sutton, Chapter 13. A detailed model of combustion of
composite propellants is presented next.
Combustion of Composite Propellant

(Ref: Guy Lengellé, Jean-Robert Duterque, Jean-Claude Godon, Jean-Francois
Trubert, ONERA, “Solid Propellant Steady Combustion – Physical Aspects”. In
AGARD-LS-180-Combustion of Solid Proplellants,1991 (TL507.N867, no. 180))
Composite propellants are heterogenous mixtures of oxidizer grains and

powdered aluminum fuel, both embedded in a rubber-like binder, which is also a fuel.
The most common oxidizer by far is Ammonium Perchlorate (AP), (CℓO4NH4), a
crystalline substance with ρ = 1.95 g/cm3, cp = 0.31 cal/g/K, thermal diffusivity =
( )
dp = 2.5x10−3 − 4.55x10−6 T 0C cm2/sec, and an estimated m.p of 835K. AP is ground
to sizes from a few to around 100 µm . The finer grades are dangerous, so grinding is
done just prior to fabrication. AP has M = 116.5 g/mole and 55% by mass is oxygen.
The aluminum is also ground to similar sizes at the last minute. Al is a very
exothermic fuel, producing Aℓ2O3 which is liquid at the flame temperature ( ≈ 3500K),
and condenses later to a solid.
The binder is often polybutadiene (synthetic rubber), either Carboxyl

Terminated (CTPB) or Hydroxyl Terminated (HTPB). The composition of CTPB is
ρp
cal / sec
C7H11.24O0.2 , with thermal conductivity λ = 3.6x10−4 , = 0.97 g / cm3 ,
p
(cm)(K )
cp = 0.39 cal / g / K
Best performance is obtained with very high percentage of AP, although

mechanical properties require a minimum of binder, and AP concentration ranges
from ≈ 70% by mass when there is Aℓ ( ≈ 16%), the balance being binder (14%), to
about 80%-85%, with no Aℓ (as in “smoke-less” compositions), the balance then
being all binder.
Overview of Combustion Mechanism – The burning of AP-binder propellants (no Aℓ) is

a complex series of phenomena, and the detailed geometry of the grains does matter
(size, particularly). At P ≥ 20 atm, AP itself can deflagrate exothermically, and it
decomposes partly in a thin liquid layer on the surface of a grain, partly in an “AP
flame” a few µm above it. The heat of decomposition raises T to ≈ 1205K by itself;
heat from the outer flame (more below) can raise the AP flame temperature well
above this, however.
Around the AP grains, the heat from the main flame decomposes the binder,
which generates a mixture of short-chain hydrocarbons, while absorbing about 360
cal/g, plus the energy to heat it to the surface temperature Ts ≅ 1000K-1100K.

The O2-rich gas generated in the AP flame co-flows outwards with the binder
decomposition products, with interdiffusion along the way. This is a “diffusion flame”,
and the final combustion takes place in it, raising T to about 3540K.
When there is Aluminum, the Aℓ Particles

Diffusion flame
are ejected when the binder holding them
(Main flame ~ recedes; they then burn at several
AP flame 3450 k) hundred µm from it; during this burning
(1200-1600 k) they agglomerate to several tens of
AP µm (they are liquid, m.p=930K, but
liquid binder remain “encased” in Aℓ2O3 until this shell
breaks at 2300K, then the Aℓ spews ≈
solid
1 µm microparticles which burn quickly)
t=0 t = 2x10-4s t = 4x10-4s t = 6x10-4s t = 8x10-4s 10-3sec
From the surface regression point of view, Aℓ “burns” instantly, as its particles are
ejected.
Overall Burn Rate from Burn Rates of Constituents
Let υ p be the mean surface regression speed (cm/sec), and υ AP , υb the

corresponding rates for the AP and the binder individually and in isolation. Although
the geometry is more complex, we can idealize the propellant as a layered medium,
with alternative thicknesses δ AP , δ b . The time to burn through both δ AP and δ b is
δ AP + δ b δ δ
t = = t AP + tb = AP + b
υp υ AP υb
δ AP δb
Calling ξ AP = , ξ b = 1 − ξ AP = the volume fractions of the constituents
δ AP + δ b δ AP + δ b
1 ξ AP ξb
= + (1)
υp υ AP υb

The mass flux (g/cm2/sec) burnt is
m
•
p = ρp υ p (2)
where ρp is the mean density
mAP + mb 1 α AP α b
ρp = = + (3)
mAP mb ρp ρAP ρb
+
ρAP ρb
where α AP is the mass fraction of AP, and α b = 1 − α AP is that of the binder.
Note: ρAP = 1.95 g/cm3, ρb = 0.91 g/cm2
The mass fluxes of the individual constituents are, similarly,

m
• •
AP = ρAP υ AP ; b = ρb υb (4)
From (1) and (2),
ρAP ρ
ξ AP
ξb b
1 1 ρ p ρp
= = +
m
ρpυ p • ρAP υ AP ρb υb
p
ρi ⎛ Vi ⎞ ⎛ Mi ⎞ M Mi
and since ξ i = ⎜ ⎟/ = = αi
ρp ⎜⎝ V ⎟⎠ ⎝ Vi ⎠ V M
1 α AP αb
=• +• (5)
m
•
p AP b
AP For aluminum loaded propellants, we noted that

the Aℓ particles are ejected when the binder
AP b
holding them burns through. Looking at the
AP
simplified geometry below, it can be seen that the
mean burning speed υ p would be the same if the
Al b
binder were really “filling in” for the Aℓ.

We then get the approximate expression
1 ξ AP ξ b + ξ AA
= + ,
υp υ AP υb
identical to the Eq. (1), even with Aℓ particles. Eq. (5) also follows, with 1 − α AP in
m
•
place of α b . The only difference is that the mean density ρp in p = ρp υ p is modified
by the Aℓ
⎛1 α α α ⎞
⎜⎜ = AP + b + AA ⎟
⎝ ρp ρAP ρb ρAI ⎟⎠
Separate Burning of AP
Heat penetrates into the receding AP particle to a small depth only. This can be seen
from the heat balance written in the receding frame:
Surface
fixed
Ts
AP moving
υAP
T = To
dT d ⎛ dT ⎞ λ
ρAP υ AP c AP − ⎜ λAP =0 Define d = (heat diff., cm2/sec)
dx dx ⎝ dx ⎟⎠ ρc
dT d (T − T0 )
υ AP T − d AP = υ AP T0 ; d AP = υ AP (T − T0 )
dx dx
υ AP ⎛ dAP ⎞
x T − T0 x / ⎜⎜ ⎟
υ ⎟
T − T0 = c e dAP → = e ⎝ AP ⎠
Ts − T0
Note: dp ≅ 1.2 x 10−3 cm2 / s, c ≅ 0.31 cal / g / K .

This shows an exponential temperature decay into the solid, with a
⎛d ⎞
characteristic thermal thickness x* = ⎜ ⎟ . Assuming υ = 1 cm / s,
⎝ υ ⎠ AP
dAP ≅ 1.2 × 10 cm / sec , this gives x = 1.2 × 10−3 cm = 12 µ m .
−3 2 *
The top few microns of the AP are molten when its temperature exceeds
TAP ≅ 835 K (which may not happen if the pressure is so low, under ≈ 20atm, that the
flame-surface distance is too large to provide sufficient heating to it. Note this rate is
⎛ dT ⎞
qs = −λAP ⎜ ⎟
⎝ dx ⎠o
( )
= − λAP / x * (Ts − T0 ) = − ρAP υ AP c AP (Ts − T0 ) , so
qs MIN
= ( ρ υ c ) AP ( 835 − 298 ) typically.
In this molten layer, about 70% of the AP undergoes complete decomposition

to final (onydizing) gaseous products, according to
NH4CℓO4 → 0.285N2 + 0.12N2O+ 0.23NO + 1.62H2O + 0.76HCℓ + 0.12Cℓ2 + 1.015O2

(7)
whereas the remaining 30% sublimates as a mixture of ammonia, NH3, and

perchloric acid, HCℓO4; this mixture then completes the decomposition to the final
m
products of (7) in a premixed (AP) flame about 1µ from the surface.
Let us look at the energetics of these effects:
(a) Enthalpy per gram to bring AP to its surface (molten) temperature (including
some intermediate phase transitions):
hH
∆ = 266 + 0.328 (Ts, AP − 835) (cal/g) (8)

,
A
P
(b) Heat of the sublimation into NH3 + HCℓO4

hs
∆ = 476 to 510 cal/(g/sublimed) (493 average) (9)

,
A
P
(c) Heat of “combustion” of NH3 with HCℓO4

hC
∆ = −850 to − 885 cal/(g.reacted) (heat released) (10)

,
A
P
(d) Heat released per gram of AP directly degraded in the liquid phase:
hD
∆ = −375 cal / g (11)

,
A
P

hs
Note: We must have ∆ + ∆hc = ∆hD )
We can combine these values to obtain the Adiabatic Flame Temperature Tfad
, AP of the
“AP flame”:
∆hD,AP
hH
hD
hs
hC
∆ ,
A
P
+ 0.7∆ + 0.3 ( ∆ +∆ ) + c (T ad
)
Ts, AP = 0 (12)
,
A
P
,
A
P
,
A
P
g f , AP
which, using for the gas a specific heat
cg = 0.3cal / g / K
gives Tfad
, AP = 1205 K (13)
Rates and AP flame structure. The rate of pyrolization is found experimentally to

depend on the surface temperature according to an Arrhenius-type expression
m
•
− Es , AP / RTs
AP = AS , AP e (14)
with ES , AP = 20 Kcal/mol AS , AP = 96000 g/cm2/sec
The reaction rate for the premixed AP flame obeys a similar law, except that, being
primarily a bimolecular reaction, its rate is proportional to p2:
ω
•
− Eg , AP / RTf , AP
= p2 Ag, AP e (15)
ω
•
with p in atm, in g/cm3/sec, and with
Eg
= 15Kcal / mol; Ag, AP = 650g / cm3 / sec / atm2

,
P
If the velocity υ of the gas normal to the surface were known, Eq. (15) would allow
g
xf
calculation of the flame standoff distance, . The time to “burn” the gas is
,
A
P
ω
•
τ = ρg / , and then
c
h
υg ρg
Eg , AP
xf
• +
= υ gτ ch = υ g ρg / ω =
RTg
e (16)
,
AP 2
p Ag, AP
•
(and notice ρg υ g = m AP )

A separate expression for x f can be obtained from the fact that the AP flame has to
supply to the vaporizing surface the required heat of reaction, so if the flame moves
too far, the reaction is too slow, and vice versa. The net (convection + conduction)
heat flux in the gas between surface and flame is constant (no heat evolves there)
•
dT d ⎛ dt ⎞
⎜ λg dx ⎟ = 0 ( cg ≅ 0.3cal / g / K , λg = 1.9 x 10 cal / sec / cm / K ) (17)
−4
mAP cg −
dx dx ⎝ ⎠
with the boundary conditions T=TS,AP at x=0 and
⎛ dT ⎞ • •
⎜ λg dx ⎟ = mAP Qc = m AP ( ∆hH , AP + QS ) (18)

⎝ ⎠ x =0
Here Qs is the net heat required to take the AP from liquid at Ts to gaseous products
(before the AP “flame”)
Qs = 0.3 x 493 + 0.7(-375) = -115 cal/g (19)
Integrating (17) with (18),
•
dT • •
m AP cgT − λg = m AP cgTS,AP − mAP Qc
dx
• •
dT m AP
dx
−
λg
(T − TS, AP ) = mλAP Qc
g
Qc
and imposing T= TS, AP at x = 0 again, c = , so
cg
⎡ mAP cg ⎤
•
Q ⎢ λ x ⎥
T - TS, AP = c ⎢e g − 1⎥ (20)
cg
⎢ ⎥
⎣ ⎦
The “AP flame” is at x = x f, AP, where T reaches TfAP (adiabatic Tf for AP alone, but
maybe higher if there is heat supply from the main flame). Solving for xf,AP then,
λg ⎡ cg (Tf , AP − Ts, AP ) ⎤
xf, AP = •
An ⎢1 + ⎥ (21)
⎢ Qc ⎥⎦
m AP cg ⎣
This must be the same as (16). Equating them,

• Eg, AP
m AP RT λg
2
e f, AP = • An ⎡⎣⋅ ⋅ ⋅⎤⎦
p Ag, AP m AP cg
cg (Tf , AP − Ts, AP ) ⎤
Eg, AP
• λg A −
RTf, AD
⎡
or m AP = p g , AP
e An ⎢1 + ⎥ (22)
cg ⎢⎣ Qc ⎥⎦
So, for AP alone, where Tf , AP ≅ 1205 K is known, and Ts,AP can be estimated (not too
much higher than 835 K), the regression rate is proportional to pressure ( nAP ≅ 1 ).
Pyrolysis of Inert Binder. A similar formulation can be used for the calculation of the
heating rate due to the main flame, which serves to pyrolyze the surface of the
binder (and also to elevate the temperature of the AP flame).
Wf Tf The paper by Lengellè et al.

Simplifies the model by
treating for this purpose the
flow of gas as 1-D (even
Xf
though it really is 2-D or 3-
D due to the heterogeneity
Wf, AP
of the surface. It also
TF,AP
assigns a uniform mass flux
XF, AP mp above both, AP and
Ts, AP AP Binder binder. This is questionable,
Ts,b
but we’ll press on.
Similar to Eq. (17), we have now
•
dT d ⎛ dT ⎞
mp cg − λg =0
dx dx ⎜⎝ dx ⎟⎠
dT
and, defining q = λg the magnitude of the (surface-directed) conduction heat flux,
dx
•
dq mp cg
− q=0 (23)
dx λg
•
We integrate the condition mp Qf = q ( xf ) (24)

where xf is the location of the main flame, at which the energy of combustion
•
Qf ≅ 700 cal / g is released (q=0 above it, q = mp Qf below it):
•
m p c g ( xf − x )
• −
λg
q = mp Qf e (25)
At x=0, on the binder surface, the heat flux q(0) is used to pyrolyze the binder at the
• •
rate mb (p.u. area), and if Qc,b is the required heat (cal/g), we obtain q(0) = mb Qc ,b ,
or
•
m p cg xf
• • Qf −
λg
mb = m p e (26)
Qc ,b
The pyrolisation heat Qc,b is composed of that required to heat the binder form T0 to
Ts,b, plus the heat of decomposition Qs=360 cal/g:
Qc ,b = cb (Ts,b − T0 ) + 360 ( cb ≅ 0.39 cal / g / k ) (27)
For Ts,b=1100K, Qc,b=675 cal/g.
Eq. (25) is also useful to estimate the rate of arrival of heat from the main flame at
the location of the AP flame (x=xf,AP):
m
•
− (
cg xf − xf , AP )
m
p
•
λg
qf , AP = p Qf e (28)
and, from this, the AP flame temperature, using a modification of Eq. (12):
•
m AP ⎡⎣ ∆hH , AP + ∆hD, AP + cg (Tf , AP − Ts, AP ) ⎤⎦ = qf , AP (29)
or
•
( ad
m AP cg Tf ,AP − Tf,AP = qf ,AP ) (29b)
ad ad
where Tf,AP is the value of Tf,AP with no qf,AP present ( Tf,AP ≅ 1205K , as we found).

Estimating the Main Flame Distance. Let Dox be the diameter of the oxidizer particle.
The surface of the particle evolves oxidizing gas (after AP flame), while the binder
around it generates fuel gas.
These inter-diffuse to form a

diffusion-flame, similar to that
from a Bunsen burner (but inside-
Xf out). The radial distance covered
fuel by a substance diffusing with a
gas
oxidizer diffusivity D, in a time t, is of the
gas order of 2Dt . We say the
flame’s end is at the point when
this equals Dox/2 (times some
factor of order 1, to account for
AP
real geometry):
Dox Dox
≅ Ad1 / 2 2Dt ( Ad ≈ 1)
2
ρg •
Equating the time t to xf / υg = xf •
(notice we again use the overall mass flux m p
mp
here),
Dox •
≅ Ad1 / 2 2D ρ g xf / mp
2
or
•
2
mp DOX
xf = (31)
8 Ad ( ρ g D )
The diffusivity D is inversely proportional to the gas density, so that ρg D is

independent of P at a given T (and weakly dependent on T). Thus, (31) gives a main
flame distance which is independent of pressure, and scales with the square of AP
particle diameter.
To a good approximation, the mass diffusivity D is equal to the heat diffusivity:

λg λg
D ≅ dg = ⇒ ρg D ≅ (32)
ρ g cg cg
which can be substituted into (31):
•
mp cg 2
xf ≅ DOX (32)
8 Ad λg
The final gas temperature (after the main flame) can be estimated now from
Qf
α AP Tf ,AP + α bTs ,b + = Tf (34)
cg
Using α AP = 0.8, α b = 0.2, Tfad

,AP = 1205 K,
Ts ,b = 1100 K, cg = 0.3cal/g/K,
Qf = 700 cal/g. We calculate Tf = 3250 K.
Actually, Eq.(33) is appropriate only when the diffusion time (distance) is much more
than the reaction time (distance), as at high P (short reaction times), and/or large
AP particle diameters (long diffusion times). In the opposite limit, xf is really dictated
by the chemical reaction time (distance); similar to Eq. (16), this distance can be
written as
• Eg ,f
mp +
xf ,r = 2 e RTf (35)
p Ag ,f
The values of Ag,f and Eg,f are not given by Langellé et al. This is especially
regrettable for Eg,f which is very sensitive. We here take tentatively
Ag,f = 650g/cm3/sec/atm2, as for AP, and determining Eg,f by matching approximately
one of the small-diameter data points quoted in the paper. This leads to
Eg ,f ≅ 29.4Kcal / mol (36)
For the general case, then, we take xf to be the sum of the reaction and diffusion
distances:
⎡ c
•
1 +
Eg ,f
⎤
g
x f = mp ⎢ 2
Dox + 2 e RTf ⎥ (37)
⎢⎣ 8 Ad λg p Ag ,f ⎥⎦
•
Solution Procedure. Given P ,T0 , Dox , α AP , we want to calculate mp , as well as several
of the intermediate variables. The equations are fairly complex, so some iteration
must be devised. First, from the averaging law (Eq. (5)) with α b = 1 − α AP ,

•
mb 1 − α AP
•
= •
(38)
mp mp
1 − α AP •
m AP
•
mb
But •
is also calculable from Eq. (26); this involves xf, which is given by Eq. (37):
mp
•
•
−
m p cg • ⎡ c
mp ⎢
g
D2 +
(
exp + Eg ,f / QTs ) ⎤⎥
mb Q λg ⎢ 8 Ag λg ox p2 Ag ,f ⎥
•
= f e ⎣ ⎦
(39)
Qc ,b
mp
•
Equate (38) and (39) and solve for mp :
⎡ • •
⎤
Q 1 − α m p / m AP
λg A n ⎢ c AP ⎥
•
⎢ Qc ,b 1 − α AP ⎥
mp = ⎣ ⎦ (40)
⎡ cg exp ( − E g ,f / RTf ) ⎤⎥
cg ⎢ Dox2
+
⎢⎣ 8 Ad λg p2 Ag ,f ⎥⎦
•
This is similar to the expression Eq. (22) for m AP . In fact, the ratio of both equations
•
mp
provides a relationship between •
, and Tf,AP with Dox and p as parameters:
m AP
⎡ • •
⎤
⎢ Qf 1 − α AP m p / m AP ⎥
An
•
E g , AP / RTf , AP ⎢ Qc ,b 1 − α AP ⎥
mp e ⎣ ⎦• 1 (41)
=
cg (Tf ,AP − Ts ,AP ) ⎤ exp ( + E g ,f / RTf )
•
Ag ,AP ⎡ cg
m AP An ⎢1 + ⎥ p2 Dox
2
+
⎢⎣ Qc ⎥⎦ 8 Ad λg Ag ,f
•
mp
The AP flame temperature Tf,AP depends, in turn, on •
, as can be seen by
m AP
combining Eqs. (28) and (29):
−
(
mp cg xf − xf , AP )
• •
m AP ⎡⎣ ∆hH , AP + ∆hD, AP + cg (Tf , AP − Ts, AP ) ⎤⎦ = mp Qf e
λg
(42)
•
m p cg xf
−
λg
Here, we notice that the factor e is

•
m p cg xf •
− Q mb Q 1 − α AP
e
λg
= c ,b • = c ,b •
(43)
Qf Qf
mp mp
1 − α AP •
m AP
•
mp cg xf ,AP
+
λg
Also, the factor e can be expressed, using (21) as
• • • •
cg (Tf ,AP − Ts ,AP ) ⎤

mp m AP cg xf, AP mp / m AP
− • λg ⎡
e m AP
= ⎢1 + ⎥ (44)
⎢⎣ Qc ⎥⎦
Combining (42), (43) and (44)
•
mp
cg (Tf ,AP − Ts ,AP ) ⎤ mAP ⎛ mp ⎞
•
1 − α AP ⎡ •
∆hH ,AP + ∆hD ,AP + cg (Tf ,AP − Ts ,AP ) = Qc ,b ⎢1 + ⎥ ⎜

× • ⎟ (45)
•
⎢⎣ Qc ⎥⎦ ⎜ ⎟
mp ⎝ m AP ⎠
1 − α AP •
m AP
Here we recall that Qc depends on Ts,AP (Eqs. (18), (19) and (8))
Qc ≅ 266 + 0.328 (Ts ,AP − 835) − 115 = 0.328Ts ,AP − 123 ( cal / g ) (46)
For the present purposes, it is sufficient to assume a value for Ts,AP somewhat above
835K; this could be refined later by matching heat fluxes at the liquid surface. We
adopt for most of the following Ts,AP=925K.
• •
Given Ts,AP, we can see that Eq. (45) uniquely relates to Tf,AP to mp / m AP . The
following calculational procedure can than be followed:
• •
(a) Take a value of mp / m AP (typically in the range 0.2-0.6)
(b) Solve (45) for Tf,AP (non-linear equation, requires some internal iteration)
(c) Solve (41) for the group
cg e
+ E g ,f / RTf
( pDox )
2
ψ ≡ + (47)
8 Ad λg Ag ,f
and, hence, for the product pDox. Given pDox, This gives p.
•
(d) Calculate m AP from Eq. (22).
⎛ •
⎞
(e) m p = ⎜ m
• •
p
⎟ m AP
•
⎜ ⎟
⎝ m AP ⎠

•
•
mp
This produces a curve of mp vs. p, with •
as the running parameter (for a fixed
m AP
Dox, particle diameter).
•
Notice how Tf,AP and m AP depend on the combination pDox rather than on the
•
separate variables. However, since m AP contains the factor p directly, the final
•
burning rate of mp is of the form
•
mp = p f(p Dox) (48)
where f is some complicated function. This is an important scaling law (not pointed
out in Lengelle et al’s paper). To verify it, we can use the data reported in Fig. 31 of
the paper.
100
80% 5µm AP-CTPB

Burning Rate (mm/s)
80% 90µm AP-CTPB
Computed burning rate

10
1
10 100 1000
Fig.3 Pressure (atm)
For Dox=5 µm , we have

P(cm) 10 30 100 300
υ p (mm/sec) 7.8 21 61 92
PDox (atm µm ) 50 150 500 1500
υ p / p ( mm / sec / atm ) 0.78 0.70 0.61 0.307
For Dox=90 µm ,
P(cm) 10 30 100 300

υ p (mm/sec) 5 7.2 10.2 20.0
PDox (atm µm ) 900 2700 9000 27,000
υ p / p ( mm / sec / atm ) 0.5 0.24 0.102 0.0667

Plotting now both curves as ( υ p /p) vs. (pDox), they do coincide:
υp / P (mm/sec/atm)
1
0.5
Dox = 5µm Dox = 90 µm
0.2
pDox (atm x µm)
0.1 200 500 1000 2000 5000 10,000
100

Lecture 19: Liquid Propellants
Propellant Selection
Read Sutton, Ch. 7
Fuels
Hydrazine LH MMH RP-1 UDMH
N2H4 H2 N2H3-CH3 CH1.97 N2H2-(CH3)2
m.p. (K) 274.5 14.0 220.7 225 216
b.p. (K) 386.7 20.4 360.6 460-540 336
⎛ Kcal ⎞ 0.80 1.75 0.7 0.45 0.65

Sp. ht ⎜ ⎟
⎝ Kg K ⎠
ρ (g/cm3) 1.0 0.071 0.86 0.6-0.8 0.6-0.8
∆Hf @ 250 C +12.0 (ℓ) -2.4 (@20K) +12 .7 -5.9 (ℓ) +12.7 (ℓ)
(Kcal/mol) (ℓ)
Unstable Cryogenic Unstable Unstable
Oxidizers
NA NTO LOX
Nitric Acid Nitrogen Liquid Oxygen Hydrogen
Tetroxide Peroxide
HNO3 N2O4 O2 H2O2
m.p. (K) 231.6 261.7 54.4 272.8
b.p. (K) 355.7 294.3 90.0 423.5
⎛ Kcal ⎞ 0.4-0.16 0.37 0.4

Sp. ht ⎜ ⎟
⎝ Kg K ⎠
ρ (g/cm3) 1.5 1.37 1.1-1.2 1.46
∆Hf @250 C -41.4 0 −15.7 Kcal -44.8 (ℓ)

−3.8
(Kcal/mol) 4.18 mol
(@90K)
Corrosive Toxic Reactive, Cyrogenic Unstable

Desirable: Low m.p., high ρ , c, b.p. (for high ∆H in cooling)
Stability (negative ∆Hf )
Low vapor pressure (high b.p.)
Hypergolicity
Non-toxicity
Storability
Non-corrosive
Examples of use:
Hydrazine LH MMH RP-1 UDMH
Monoprops. J-2 Space biprops Atlas Titan II (50% N2H4)
SSME Shuttle OMS Thor Lunar Lander
Vulcain (Ariant) Delta (St.1) Proton?
J7 (Japan) Titan I
RL-10 (Centans) Saturn
RS-68(EELV) Delta IV RD-170
RD-180
Fastrac (X-34)
RS-76
NA NTO Lox H2O2
Titan II Atlas X-15
Shuttle OMS Jupiter
Space biprops RD-170

With
Proton? RP-1 RD-180
Delta (St.1)
Fastrac(X 34)
RS-76
SSME
With Vulcain
LH
RS-68
RL-10

NA NTO L ox H2O2
Titan II Atlas X-15
Shuttle OMS Jupiter
Space biprops RD-170

With
Proton? RP-1 RD-180
Delta (St.1)
Fastrac(X 34)
RS-76
SSME
With Vulcain
LH
RS-68
RL-10

Lecture 20: Combustion of Liquid Propellants
Simplified Drop-Wise Liquid Combustion Theory
1. Introduction
In this lecture we examine in more depth the fluid phenomena which

dominate the operation, and hence the design, of a liquid rocket. We start by
considering the two phase aspects of combustion, which usually determine the
combustor size. Next we examine the stability of combustor operation and the
design modification required to achieve it. Heat transfer to walls, wall
compatibility and wall cooling form the next topic. We conclude with some
notions in nozzle aerodynamic design and nozzle optimization.
2. Drop Vaporization and Spray Combustion
Returning to Fig. 12.1 (lecture 12), we note again the existence of a zone AC
where the propellants go from a series of liquid jets issuing through a multiplicity
of small injector holes, through breakup of these jets into droplets, impingement
(in some designs) of jets or droplet streams on each other, dispersion of the
droplets into a recirculating mass of combustion products, evaporation of the
droplets, interdiffusion of the vapors and kinetically controlled combustion. These
are obviously complicated processes, and a comprehensive analysis good enough
for first principles design requires large-scale computation [18, 19, 20]. In fact,
the largest number of existing liquid rocket combustors, those dating from before
1970, were developed mainly through empirical methods, supplemented by very
extensive testing. Improved modeling and computational capabilities have more
recently permitted a more direct approach, with fewer hardware iterations, but
theory is still far from completely developed in this area, and serves at this point
mainly to ascertain trends and verify mechanisms. For an in-depth discussion of
liquid propellant combustion, see Refs. 21 and 22. Here we will only review the
fundamental concepts which underlie current spray combustion models.
2.1 Single-Drop Combustion
Even when liquids are originally injected into the chamber, actual combustion
takes place in the gas phase, following vaporization of both, oxidizer and fuel. In
fact, vaporization is usually the rate–controlling step in the whole process,
although its rate is itself affected by the reactions occurring near each droplet. In
a few instances, both propellants enter the chamber as gases or easily vaporized
liquids, and then gas phase mixing is the limiting step. (An example is the Shuttle
SSME, where the hydrogen is vaporized in the cooling circuit, and the oxygen is
partially vaporized by the hydrogen in a co-flowing heat exchanger arrangement
ahead of the injectors, the rest being atomized and evaporated shortly after
injection).

In a fuel- rich environment, vapor from a fuel droplet diffuses without
reacting into the surrounding gas, while heat diffuses inwards to supply the latent
heat of evaporation. In the same environment, vapor from an oxidizer drop
comes in contact, and reacts with the fuel-rich gas at some distance from the
droplet, forming a spherical flame, from which heat diffuses both ways, while
reaction products diffuse outwards and also stagnate in layer around the droplet.
A similar description applies to a locally oxydizing environment. It will be
sufficient to analyze, for example the case of a fuel drop in an oxydizing
environment, from which we can easily derive other cases by suitable re-
interpretations.
As a preliminary, some relationships will be derived between the overall

stoichiometry (oxidizer/fuel ratio, or OF, by mass), and the mass fractions of
products and excess oxidizer or fuel to be found in the burnt gas far from the
droplets. Let the actual mass ratio be OF, and the stoichiometric mass fraction be
OFs (later called i for short), Assuming excess oxidizer, and complete combustion
(a rough assumption), the reaction can be mass balanced as
1 (Fuel) + OF (Oxidizer) → (1+OFs) (Products) + (OF-OFs) (Oxidizer)
and so the mass fraction of excess oxidization at infinity is
OF − OFs OF − OFs
Yox,∞ = =
(1 + OFs ) + (OF − OFs ) 1 + OF
Similarly, for fuel-rich operation,
1 ⎛ 1 ⎞ ⎛ 1 1 ⎞
(Fuel) + I (oxidizer) → ⎜1 + ⎟ (Pr oducts ) + ⎜ − ⎟ (Fuel)
OF ⎝ OFs ⎠ ⎝ OF OFs ⎠
and the mass fraction of excess fuel at infinity is
1 1
−
OF OFs OFs − OF
YF,∞ = =
⎛ 1 ⎞ ⎛ 1 1 ⎞ ( + OF ) OFs
1
⎜1 + ⎟+⎜ − ⎟
⎜ OFs ⎠ ⎝ OF OFs ⎟⎠
⎟ ⎜
⎝
which could be also formally obtained from YOX, ∞ by reversing the roles of fuel
1
and oxidizer (including OF → )
OF

Consider a spherical fuel droplet of radius R and density ρl . Its rate of
vaporization is
dR
Γ = −4 π R 2 ρρ
dt
and the main purpose of the analysis is the calculation of Γ from given R and gas
properties. Let Yf (r ) , Ypr (r ) by the mass fractions of fuel and of combustion
products in the region R < r < rFl between the drop and the flame location (no
oxidizer is to be found there, so Yf + Ypr = 1). These two constituents are both
convected by the mean outwards flow, and diffuse according to their
concentration gradients. Since the products do not penetrate the drop surface,
the net outflow of fuel at radius r is equal to the total mass flow, and is given by
dYf
Γ Yf − 4 π r2 ρ Df =Γ
dr
where Df is the diffusivity in the products. The net flux of products remain zero in
this region (but not their concentration Ypr = 1 − Yf ). Imposing that Yf approaches
zero at the flame radius, the above equation integrates to
⎛1 1 ⎞
−λ f ⎜⎜ − ⎟⎟
⎝ r rFl ⎠
Yf = 1 − e (R < r < rFl)
where
Γ
λf =
4 π ( ρ Df )
and the product ρ Df is taken to be constant.
While this gives information about the composition of the gas near the
droplet, the more important results come from a similar mass balance outside the
flame radius (r > rF).

k
Assume cp cp cpprods. , k+ k− k , ρg D
f OX
cP
For r > rF , the oxidizer mass balance is
dYox
Γ Yox = 4 π r 2 ρ D = -i Γ (1)
dr
or
dYox Γ iΓ
+ Yox = − (2)
d (1 r ) 4 π ρ D 4 π ρD
Γ Γ Cp
Define λ= (3)
4 π ρD 4 πk
and impose Yox (r = ∞ ) = Yox,∞ (4)
λ
Yo x = (i + Yox, ∞ ) e
−
r
−i (5)

λ
Impose also Yox (r = rF ) = 0 , solve for :
rF
λ ⎛ Y ⎞
= ln ⎜ 1 + ox,∞ ⎟ (6)
rF ⎝ i ⎠
which is a relationship between Γ and rF (rF turns out to be proportional to Γ ,

i.e., a “blowing” effect)
The distribution of both, mass fluxes and mass fractions are displayed below:
We now need the heat flux balance, both inside and outside the flame:
R < r <rF - Taking T = Tv as the (temporary) enthalpy zero, the convected heat
flow inside the flame, where only fuel vapor moves, is Γ cp ( T − Tv ) . The conductive
dT
flow is −4 π r 2 k , The total, flowing inward (negative) is used at r = R to vaporize
dr
liquid, at L (latent heat) per unit mass:
dT
Γ cp ( T − Tr ) − 4 π r 2 k = −Γ L (7)
dr
dT Γ Cp ΓL
+ ( T − Tv ) = − (8)
d (1 r ) 4 π k 4 πk
λ again

Imposing T(R) = Tv , this integrates to
L ⎡ λ ⎛⎜ R1 − 1r ⎞⎟ ⎤
T - Tv= ⎢e ⎝ ⎠
− 1⎥ (9)
cp ⎢⎣ ⎥⎦
and, in particular, at r = rF
L ⎡ λ ⎛⎜⎜ 1 − 1 ⎞⎟⎟ ⎤
TF - Tv = ⎢e ⎝ R rf ⎠ − 1⎥ (10)
cp ⎢ ⎥
⎣ ⎦
which also relates Γ (in λ) to rF, but also introduces TF , the flame temperature,
as a third unknown. We will need the heat flux balance for r > rF to complete the
formulation. For this purpose, we now move the enthalpy zero to T=TF, the flame
temperature. The total heat flow (convection plus diffusive) is now
dT
Γ cp ( T − TF ) − 4 π r2 k
dr
and this must be the heat released at the flame front, minus the heat sent
inwards from the flame to the liquid. The former is Γ Q (Q= heat of combustion
per unit fuel mass). The latter would be Γ L , except for the extra Γ cp ( TF − Tv )
due to the change of enthalpy references. Altogether,
dT
Γ cp ( T − TF ) − 4 π r 2 k = Γ ⎡⎣Q − L − cp ( TF − Tv ) ⎤⎦ (11)
dr
dT Γ cp Γ
or + ( T − TF ) = ⎡Q − L − cp ( TF − Tv ) ⎤⎦ (12)
d (1 r ) 4 π k 4 πk ⎣
Integrating with T ( ∞ ) = T∞ as a boundary condition,
Q − L − cp ( TF − Tv ) ⎛ − ⎞
λ
−
λ
T − TF = ⎜1 − e r ⎟ + ( T∞ − TF ) e r (13)
cp ⎝ ⎠
and particularizing at r = rF (T=TF),

Q − L − cp ( TF − Tv ) ⎛ − rλF ⎞
TF − T∞ = ⎜e − 1⎟ (14)
cp ⎜ ⎟
⎝ ⎠
which is the needed extra relationship to complement (6) and (10)
λ
Eqs. (10) and (14) can be simplified if is substituted from (6):
rF
⎛ λ
⎞
L ⎜ eR ⎟
TF − Tv = ⎜ − 1⎟ (10’)
cp ⎜ Yox,∞ ⎟
⎜1 + ⎟
⎝ i ⎠
Q − L − cp ( TF − Tv ) Yox,∞
TF − T∞ = (14’)
cp i
Eliminate TF between these two equations:
⎛ ⎞ ⎡ ⎛ ⎞⎤
L ⎜ eR
λ
⎟ ⎢Q − L L ⎜ eR
λ
⎟⎥ Y
Tv + ⎜ − 1 ⎟ − T∞ = ⎢ − ⎜ − 1 ⎟ ⎥ ox,∞
cp ⎜ Yox,∞ ⎟ ⎢ cp cp ⎜ Yox,∞ ⎟⎥ i
⎜1 + ⎟ ⎢ ⎜1 + ⎟⎥
⎝ i ⎠ ⎣ ⎝ i ⎠⎦
λ
L e R ⎛ Yox,∞ ⎞ Q Yox,∞ L
⎜1 + ⎟ = + + T∞ − Tv (15)
cp Yox,∞ ⎝ i ⎠ cp i cp
1+
i
λ Γ cp
Solve for , and use λ = :
R 4 πk
4 πk ⎡ Q Yox,∞ i + cp (T∞ − Tv ) ⎤
Γ= R ln ⎢1 + ⎥ (16)
cp ⎣ L ⎦

In the literature, this is written as
4 πk
Γ= R ln ⎡⎣1 + B⎤⎦ (17)
cp
Q Yox,∞ i + cp ( T∞ − Tv )
with B ≡ (18)
L
(the Spaulding parameter).
We have solved for Γ , and can go back to (6) to get the flame radius
ln (1 + B )
rF = R (19)
⎛ Y ⎞
ln ⎜1 + ox,∞ ⎟
⎝ i ⎠
and to (10’) to get TF. From (15), (18), lλ R = 1 + B , and then, from (10’),
L B − Yox ,∞ i
TF = Tv + (20)
cp 1 + Yox ,∞ i

Order of Magnitude:
Assume an oxidizer–rich LOX-kerosene burner, with (O/F)s=i=2.4, and O/F =3.

3 − 2.4
This gives Yox,∞ = = 0.15 . The heat of reaction is roughly Q=4.3 x 107
1+3
J/Kg/fuel and the fuel latent heat is L 2.5 × 105 J / Kg . Assume the surrounding
gas is at T∞ − Tv = 2000K , and take Cp 2000J / Kg / K . We calculate
Yox,∞ 0.15
= = 0.0625
i 2.4
4.3 × 107 × 0.0625 + 2 × 103 × 2 × 103

B= = 26.8
2.5 × 105
⎛ Yox,∞ ⎞ Yox,∞
In view of these numbers, ln ⎜1 + ⎟ 1 , but B 1 , so (19) and (20)
⎝ i ⎠ i
can simplify further to
i
rF R ln (1 + B ) (19’)
Yox,∞
L Q Yox, ∞ i
TF Tv + B = T∞ + (20’)
cp cp
Vaporization time - The droplet is actually being consumed, so its radius varies
(on a slower time than the gas transit time), according to
dR
ρl 4 π R 2 = −Γ (21)
dt
and since Γ is given by (17)
dR k
ρe 4 π R 2 = −4π R ln (1 + B )
dt cp
R 02 − R 2 k 2 k ln (1 + B )
ρl = ln (1 + B ) t R = R0 1 − t
2 cp ρl cp R 20

R t ρl cp R 20
or = 1− ; τv = (22)
R0 τv 2 k ln (1 + B )
τv is the deep vaporization time
For an order of magnitude, assume R 0 = 50 µm = 5 × 10−5 m, ρl = 800 Kg / m3 ,

K 0.1 W / m / k, plus the numbers used in the previous examples. We calculate
( )
2
800 × 2000 × 5 × 10−5
Γv = = 6.0 × 10−3 s = 6 ms.
2 × 0.1 × ln (1 + 26.8 )
On the other hand, since τv ∼ R20 , if R0 were instead 10 µm , τv would be 25

times shorter, or 2.4 × 10−4 . To see the significance of this, assume the gas
moves in the combustion chamber at M 0.2 , or ug ∼ 240 m / s . A droplet of
radius 50 µm would then require a combustor length no less than
240 × 0.006 = 1.44 m , which is excessive from the mass point of view. A drop of
R 0 = 10µm, on the other hand, require only 6 cm to evaporate. This brings out
clearly the importance of good atomization.

Other cases
For a fuel drop in a fuel-rich background, everything is the same if we make

Q=0 (no flame), and, of course, ignore rF and TF . In fact, the only equation
needed is the r < rF heat flow balance (Eq. 9), which, using T(∞) = T ∞ , gives
L
T∞ − Tv = ⎡eλ R − 1⎤⎦
cp ⎣
and so, finally
λ ⎛ cp ( T∞ − Tv ) ⎞
= ln ⎜ 1 + ⎟⎟
R ⎜ L
⎝ ⎠
4 πk
or Γ= R ln (1 + B ) (as in (17)),
cp
with
cp (T∞ − Tv )
B= (F / F Case) (23)
L
Again, this is as in (18) if Q=0.
So, the fuel drop evaporates faster if it is in an oxidizing environment (larger

B). The reason, clearly, is the extra heating from the nearly flame in that case.
For an oxidizer drop in an oxidizing background (0/0), we have a very similar

situation (no flame), so, once again, Γ ox is given by (17), with B given by (23),
but now Tv and L must be reinterpreted to be those for the oxidizer liquid.
Finally, for an oxidizer drop in a fuel-rich background, we return to the first case
(fuel drop in oxidizing gas), and switch the roles of fuel and oxidizer throughout.
For, example, Q must be the heat of reaction per unit oxidizer mass, Yox,∞ must
be replaced by Yf,∞ , i = (O / F)st. becomes 1 i , and (Tv , L) refer to the oxidizing
liquid.

CALCULATION OF CHAMBER LENGTH (2000 version)
(Generalizing Spaulding’s )
ρl cp R 20
The drop vaporization time Tv = depend on B. For a fuel-rich
2 k ln (1 + B )
background, as normally seen in rocket, we have two types of droplets:
cp ( T∞ − Tv,f )
Fuel drop in fuel-rich medium: BF,F = (23)
Lf
Q Yf,∞ + cp ( T∞ − Tv,ox )
Oxidizer drop in fuel-rich medium: BO,F = (24)
L ox
(notice Q is still the heat of reaction per unit fuel mass). In many cases L ox < L f
(especially for Lox), and also Q is a large quantity appearing in BO,F , but not in
BF,F . So, BO,F BF,F , meaning the oxidizer drops are likely to evaporate much
earlier than the fuel drops. For chamber length calculations, then, we focus on
the fuel drops mostly.
Assume fuel drops are injected into the chamber at a velocity uDo. Oxidizer
drops are assumed to vaporize very fast (or to be fully evaporated at injection, as
when Lox is used). At any section prior to fuel vaporization, the fuel liquid
fraction (by flux) is
4 A
γ= π R 3ρl nD uD • (25)
3
mF
The number-flux of drops nDuD A is conserved, so γ ∼ R3 and from (22), since the
initial γ is unity,
32
⎛ t ⎞
γ = ⎜1 − ⎟ ( τv = fuel drop vap. time) (26)
⎝ τv ⎠
The vaporized fuel fraction is 1- γ , so the gas mean velocity is
• •
ug =
(1 − γ ) mF + mox
(27)
ρg A

and, at the end of the vaporization, this becomes
•
m
u2 = (28)
ρg 2 A
If the molecular mass difference between fuel and oxidizer fuel and oxidizer
vapors is ignored (consistent with earlier approximations), and ρg , Tg are nearly
uniform, then we treat ρg as a constant, and so
• •
ug mF mox 1 − γ + O /F
•
(1 − γ) + •
= (29)
u2 1 + O /F
m m
u2 ⎡ ⎛ t ⎞ ⎤
32
or ug ⎢1 + O / F − ⎜1 − ⎟ ⎥ (30)
1 + O /F ⎢
⎣ ⎝ τv ⎠ ⎥⎦
O /F
Notice that (29) implies an initial gas mean velocity ug ( 0 ) u2 , which is
1 + O /F
due to the vaporized oxidizer flow.
The particle velocity, uD is different from ug in general, so there is slip, and

therefore drag force. Assuming small Reynolds number around each drop, Stokes
flow occurs, so
4 duD
π R 3 ρl = 6 π R µ g (ug − uD ) (31)
3 dt
or, defining a velocity relaxation time
2 R 20 ρl
τRe l = (32)
9 µg
2
⎛ R ⎞ duD
τRe l ⎜ ⎟ = ug − uD (33)
⎝ R 0 ⎠ dt
We now introduce non dimensional variable
τv t uD
=b ; 1− =θ ; =υ (34)
τRu τv u2

and write (33) as
θ dυ 1 + O / F − θ3 2
=− +υ (35)
b dθ 1 + O /F
This is a linear, first-order equation for υ(θ) , although with non-constant

coefficients. It can be integrated by standard methods, like parameter variation.
uD
Imposing uD ( t = 0 ) = uDo , or υ ( θ = 1) = 0 , the solution is found to be
u2
⎡u 2b ⎤ b 2b 32
ν = 1 + ⎢ Do − 1 − ⎥ θ + (3 − 2b)(1 + O / F) θ (36)
⎣ 2 (3 − 2b)(1 + O / F) ⎦
One more integration will yield the vaporization length:
τv 1
L = ∫ u dt = u τ ∫ υ ( θ ) dθ
0
D 2 v
0
(37)
After simplification, this gives the result
L 1 ⎡ uDo ⎛ 25 ⎞ ⎤
= ⎢ + ⎜1 − b⎥ (38)
u2 τv 1 + b ⎣ u2 ⎝ 1 + O / F ⎟⎠ ⎦
Notice that, from (22) and (32), the parameter b can be written as
τv ρl cp R 20 1 9 ( µgcp / k )
b= = =
τRe l 2 k ln (1 + B ) 2 ρl R 20 4 ln (1 + B )
9 µg
9 Pr
or b= (39)
4 ln(1 + B)
For B=30 and Pr=0.8, this gives b=0.52.
Spaulding defined a related parameter ξ = 2b and presented results for the

monopropellant case. We can reduce to that case by setting O/F=0. Result are
uD0
shown in the next figure for ξ = 0.5 (b = 0.25). The parameter χ0 is , and
u2

x
ξ= .
2 u2 τv
As an example, for the middle graph χ0 = 0.5 , so using O/F = 0 in Eq. (38)
L 1
we calculate ξ2 = = (0.5 × 0.6 × 0.5) or ξ2 = 0.267 . This coincides
2 u2 τv 2 × 1.5
with the result in the figure.
Notice how the droplets are initially slowed down by gas drag, but later, as
the gas evolves and accelerate, they are pulled along. Near the end of
vaporization, the rates accelerate, because the droplets are so light.
Evolution of evaporating droplet size and velocity. Here the reference vref for
velocity is that of the gas after complete evaporation, and for length, lref = 2 vref τv

References-Cont.
18. R.D. Sutton, W.S. Hines and L.P. Combs, “Development and Application of a
Comprehensive Analysis of Liquid Rocket Combustion”, AIAA JI., Vol. 10, no.
2, Feb. 1972, pp. 194–203.
19. P-Y Liang, S. Fisher and Y.M. Chang, “Comprhensive Modelling of Liquid
Rocket Combustion Chamber”. J. of Propulsion, Vol. 2, no.2, March-April
1986, pp 97-104.
20. M.S. Raju and W. A Sirignano, “Multicomponent Spray Computations in a

Modified Centerbody Combustor”, J. Propulsion Vol. 6, No. 2, pp 97-105,
March-April 1990.
21. F.A. Williams, Combustion Theory, Benjamin/Cummings Publishing Co., Inc.,

2nd Ed., 1985.
22. W.A. Sirignano, “Fuel Droplet Vaporization and Spray Combustion Theory”,
Progress in Energy and Combustion Science, Vol. 9, 1983, pp.291-322.
23. D.B. Spaulding, Aero. Quarterly, 10, 1. (1959). See also Ref. 21, Sec. 11.15.
24. H.H. Chiu, H.Y. Kim and E.J. Croke,”Internal Group Combustion of liquid
Droplets”, Proc. 19th International Symposium on Combustion, 1983.
25. P.Y. Liang, R.J. Jensen and Y.M. Chang, “Numerical Analysis of SSME
Preburner Injector Atomization and Combustion Process”. J. of Propulsion,
Vol. 3, No. , Nov-Dec.1987, pp. 508-514.
26. D.T. Harrje and F.H. Reardon (Editors). Liquid Propellant Rocket Combustion
Instability, NASA SP-194, 1972.
27. R. Bhatia and W.A. Sirignano, “One dimensional Analysis of Liquid-Fueled

Combustion Instability”, J. of Propulsion, Vol 7, No. 6, Nov-Dec. 1991, pp
953-961.
28. G.A. Flandro, ”Vortex Driving Mechanism in Oscillatory Rocket Flows”, J. of

Propulsion, Vol. 2, No. , May-June 1986, pp. 206-214.
29. R.J. Priem and D.C. Guentert, “Combustion Instability Limits Determined by a
Non linear Theory and a One-dimensional Model”. NASA TND-1409, Oct.
1962.
30. M. Habiballa, D. Lourme and F. Pit. “PHEDRE-Numerical Model for

Combustion Stability Studies Applied to the Ariane Viking Engine”. J.
Propulsion, Vol. 7, no. 3, May-June 1991, 322-329.
31. T. Poinsot, F. Bourienne, S. Candel and E. Esposito, “Suppression of

Combustion Instabilities by Active Control”. J. Propulsion, Vol.5, No. 1, Jan-
Feb. 1989, pp. 14-19.
32. NASA SP-8113, Liquid Rocket Combustion Stabilization Devices, Nov. 1974.

Lecture 21: Liquid Motors: Injection and Mixing
This is a very complex issue, which continues to receive a great deal to research
attention, due to its importance for:
(a) Performance, and

(b) Stability of combustion.
The single-droplet models of the previous lecture only begin to scratch the surface of
this problem. A recent review by Sirignano is included and recommended for reading
in order to appreciate the current state of this topic.

Lecture 22: Liquid Motors: Combustion Instability (Low Frequency)
Combustion Stability
1. General Discussion
Elimination of instabilities has been historically one of the largest components

of all new liquid rocket development programs. This is because there has been little
reliable methodology to ensure stability through design, and also because of the
potentially catastrophic consequences of instability. The situation has improved to
some extent because of the vastly enlarged simulation capabilities existing now. As
we will see, this area is intimately related to that of spray combustion, where a
similar situation has prevailed. As in the combustion area, the advances in recent
years are promising, but not yet sufficient to provide reliable tools for a priori design,
at least against the high-frequency instability problem.
In a system with the very large energy density of a rocket combustor, there
are bound to be many mechanisms by which a small fraction of this energy can be
channeled into undesirable oscillations. The resulting instabilities are usually
categorized into “low frequency” and “high frequency” types. The former involve
pressure oscillations which are slow enough compare to the acoustic passage time
that the whole chamber participates in phase, while the latter exhibits acoustic
behavior, with different parts of the combustor oscillating with different phase or
amplitude. In either case, the instability arises when energy (or sometimes mass)
can be added to the gas at, or near, peaks in its pressure oscillation (this is the
classical Rayleigh criterion for instability, as explained) for example in Refs. (21) and
(26)*. This implies a synchronization of two mechanisms, one which generates the
energy or mass in some unsteady manner, and another one which allows the gas
pressure and other thermodynamic quantities to oscillate. These two mechanisms
must have comparable time constants for the mutual feedback to develop.
The acoustic modes of typical rocket combustors have wavelengths which are
some fraction of their linear size, and a wave speed (the speed of sound) of the
order of 2000 m/sec. Thus their frequencies are in the KHz range, and hence the
oscillation time constant is 10-4 to 10-3 sec, and potential couplings are to be sought
to energy or mass release mechanisms with time constants of that order. To be more
precise, it must be noted that a time lag in the release of combustion energy (such
as, for example atomization, evaporation, mixing or reaction delays) will not per se
provide coupling with the acoustic field. There must also be a sensitivity of the heat
release rate, and hence of the time delay, to the pressure or other wave quantity.
Since, in general, only some of the additive time lags show such a sensitivity, and
then maybe only for a fraction of the effective lag, it is found that the excited
acoustic modes have time constants which are a fraction of the time lags
themselves. As an example, Eq. (22 of lecture 19) gives for a droplet with ρA =1000
kg/m3, ln (l + B)=5 and ρg D ≅ 3 x 10-5 kg/m/sec a vaporization time of 8.2 msec.
Yet, this size droplets could de-stabilize acoustic modes of about 1KHz frequency, as
reported, for instance in Ref. 27.
* See Ref. List and end of Lecture 19

Other excitation mechanisms for high frequency waves have been reported.
Gas-phase propellant mixing delays in rapidly evaporating cryogenics may be of the
required order and show sensitivity to acoustic fields. Conversion of vortex energy to
acoustic dipole radiation upon impingement of a vortex street or a turbulent shear
layer on an obstacle can also be a destabilizing factor [28], although this mechanism
may also work as a damper, if the phase is arranged appropriately. On the other
hand acoustic frequencies are normally too high to allow involvement of elements
outside the chamber itself, such as the liquid-handling equipment or the overall
rocket structure. Also, chemical reaction times in well-mixed gases tend to be
shorter than 0.1-1 msec, and are not usually involved either.
The low frequency types of instability, by contrast, usually results from

coupling of oscillations of pressure in the whole chamber to elements outside of it.
These can be the injectors, the liquid lines (or the liquid in them), the turbopumps,
and, in the instability called POGO, the whole rocket structure. There is substantially
more understanding of these than of the fast instabilities, and analytical techniques
were developed early on [26] to help avoid them in the design process. Many of the
low-frequency instabilities can be described in terms of a single, non-sensitive,
combustion delay; as discussed, this delay can be several times longer than its
sensitive fraction. In others cases, the mechanism may be more complex, perhaps
involving different delays for the two propellant streams.
The literature also mentions an intermediate and benign type of instability,

which is a hybrid sometimes described as “entropy wave”. Acoustic oscillations may
modulate differently the oxydizer and fuel injection rates, creating patches of varying
stoichiometry and temperature. These convect with the mean flow, and, upon
impingement on the converging nozzle, may synchronize with and reinforce the
acoustic wave, which then propagates back to the injector region. Since part of the
mechanism is acoustic and part convective, the frequency falls between the first
acoustic longitudinal mode and the flow passage time.
The effects of these instabilities range from simple vibratory and acoustic
disturbances to very rapid burnout of walls. The most severe effects are associated
with high-frequency instabilities of the tangential chamber modes. These create
sloshing gas motions which completely disrupt the boundary layer near the injector
end of the chamber, and can lead to failure in a matter in seconds. Thus, most effort
has gone into understanding and eliminating tangential mode instabilities. Low
frequency instabilities, if allowed to proceed, can also be structurally damaging, but,
as noted, they can often be designed out, and if not, there are reliable
countermeasures against them.
2. Methods of Analysis for Low Frequency Instabilities
The compilation by Harrje and Reardon (Ref. 26)* remains the best source of
information on this topic. The analysis typically separates into a chamber portion,
balancing the gas generation, storage and disposable, plus a series of submodels
detailing the response of various other parts of the system, such as injectors,
pumps, etc. For preliminary analysis, linear perturbation models are used, and Ref.
(26)* shows a variety of subsystem response functions adequate for this purpose.
* - D.T. Harrje and T.H. Reardon (Editors) Liquid Propellant Rocket Combustion
Instability, NASA SP-194, 1972.

As a simple prototype, we discuss here the coupling of chamber and injector
alone (single stream). It is assumed that vaporization of the liquid happens a time τv
after injection, which may be a reflection of the time delay for atomizing into
droplets, plus the droplet evaporation time given by Eq. 22 (Lecture 19), although
this part would be better modeled as a distributed, rather than a lumped lag.
Let PC and TC be the chamber pressure and temperature, both of which are
taken to be uniform, and VC the chamber volume. The vaporization rate is:
• •
mv ( t ) = mi ( t − τv ) (1)
•
while the gas disposal rate through the nozzle is as given by m = Pc A t / c∗ . We
then have
d ⎛ Pc ⎞ • P A
⎜⎜ Vc ⎟ = mi ( t − τv ) − c ∗ t (2)
dt ⎝ R g Tc ⎟ c
⎠
The injection rate through an effective injector orifice area Ai, with a pressure
P0 behind it, is
•
mi = 2 ρL (P0 − Pc ) Ai (3)
where ρl is the gas liquid destiny. Linearization of these equations, assuming

constant Rg, Tc, P0 and c*, gives
• ⎡
1 Pc ( t − τv ) Pc' ⎤
'
Vc dPc'
= m ⎢− − ⎥ (4)
R g TC dt ⎢⎣ 2 Po − Pc Pc ⎥⎦
Pc A t
( )
•
where m = = 2ρl P0 − P c Ai is the mean mass flow rate, P c is the mean
c∗
chamber pressure, and Pc' its perturbation. We can define a “chamber evacuation
time”
Pc Vc
τc = (5)
R g Tc •
m
and an injector overpressure parameter
P0 − Pc
∆= (6)
Pc

and assume pressure variations of the form
Pc'
= ĉ eiωt (7)
Pc
Substituting into (2.25) yields an equation for the complex ω :
e−iωτv
iωτc = − −1 (8)
2∆
and an instability will result when the imaginary part of ω as given by this equation
is negative. The stability threshold occurs when ω is real. Imposing this and
eliminating ω between the real and imaginary part of Eq. (8) yields the threshold
condition
τv 2∆
= ⎡⎣ π − cos−1 ( 2∆ ) ⎤⎦ (9)
τc 1 − 4∆ 2
at which the oscillations occur with a frequency given by
ωτv = π − cos−1 ( 2∆ ) (10)
Notice that this instability threshold cannot be reached if ∆ > 1 . This is because of
2
the strong energy-dissipating role of the injector, which acts as a damper for the
oscillations. Relativity large injector pressure drops are routinely used, of the order
of ∆ ≅ 0.2 . Values above ∆ = 1 would lead to excessively and lossy pressurization
2
and injection systems.
NOTE:
(at threshold)
m
1 ⎛⎜ P c ⎞ ⎛ P 'c ⎞ ⎛ P 'c ⎞
'
• '
iωt
⎟ eiωt −iωτc 1 ⎜ ⎟ei( π−ωτc ) 1 ⎜ ⎟ ei cos ( 2 ∆ )
−1
mv e = − e = =
2∆ ⎜ PL' ⎟ 2∆ ⎜ Pc ⎟ 2∆ ⎜ Pc ⎟
⎝ ⎠ ⎝ ⎠ ⎝ ⎠
Hence, mass addition occurs at an advance phase angle cos-1(2 ∆ ) w.r.t. pressure
peaks. If ∆ is small, nearly on top.

Allowing a non zero (positive or negative) imaginary part of ω , we can obtain the
results shown in Fig. 1. In general, stability is enhanced by increasing injector drop
or chamber residence time (in relation to the evaporation lag). As shown in Fig. 2.3b,
the stable or neutral oscillations have frequencies of the order of fR ≈ (1 − 2 ) / 4τv . As
an example of the kinds of delays encountered in LOX-LH rockets, we reproduce Fig.
2 (from Ref. 26).

So, three cases:
•
mv 1 ⎛ P 'c ⎞
1. ∆ = 0.5 → = ⎜ ⎟
•
2∆ ⎝ Pc ⎠
m
Gas mass production in phase with gas expulsion (as with P 'c ).
•
m'v 1 ⎛ P 'c ⎞ −i ( − )
2. ∆ > 0.5 → = ⎜ ⎟e
• 2∆ ⎝ Pc ⎠
m
• •
m'v lags m'ont , chamber begins emptying before new gas is introduced → Stable.
•
mv 1 ⎛ P 'c ⎞ +i ( )
3. ∆ < 0.5 → = ⎜ ⎟e
• 2∆ ⎝ Pc ⎠
m
• •
m'v leads m'ont , New gas starts accumulating before it can be expelled. This
can be unstable, if the τc is small enough compared to τv .
Notice that the “sensitive” (to pressure) part of this time lag is 0.1 to 0.2 of the total
lag. According to the results in Fig. 1, the marginally stable oscillations with the total
delay of Fig. 2, (and with ∆ ≈ 0.2 ) would be at approximately a frequency
1 1.7
f = ≈ 270 Hz
2π 10-3
Fig. 2 Experimental values of combustion time lag for supercritical chamber pressure

Lecture 23: Liquid Motors: Stability (High Frequency); Acoustics
Combustion Instability: High Frequency
Methods of Analysis for High Frequency Instabilities
Prior to the advent of large-scale computations, the most successful

theoretical development in this area was the “sensitive time lag theory” of L. Crocco
[26]. More than a detailed physical theory, this was a model in which a few basic
parameters were introduced from intuitive considerations, and then used to correlate
experimental observations on stability thresholds. The principal parameter was the
sensitive time lag, during which the various rates which eventually resulted in
vaporization at the total time lag τ T after injection were assumed to vary with
pressure, velocity, stoichiometry, etc. This variation was characterized by means of
other important parameter, the “sensitivity index”. For pressure sensitivity, this is
∂ ln ( Rates )
n= (1)
∂ ln P
and the definition of τ is such that the variations in gas generation rate due to this
sensitivity are given locally by
• •
m− m ∂τ P '(t ) − P '(t − τ )
=− =n (2)
• ∂t P
m
Similar sensitivity indices can be introduced for velocity, etc. Once this parameterization
is accepted, it is only a matter of mathematical modeling to obtain the stability limits of
a given acoustic wave or cavity. This modeling could be linear or even allow for non-
linearities in the gas dynamics. It is one of the strengths of this theory that the acoustic
part of the problem, namely, combustor geometry, steady state combustion and heat
release, etc. are separated from the unsteady combustion effects, which allows for
generalization of test results and accumulation of meaningful stability data.
The results of calculations using the linear sensitive time lag theory are
displayed as shown in Fig. 1 (Ref. 26).

Fig 1. An n - τ diagram, showing instability zones for various modes.
Here are shown the loci of marginal stability for severe modes of a combustor,
on a map of interaction index n vs. sensitive lag τ . Each point on one of the lines
corresponds to a particular oscillation frequency, and these frequencies are found to
be within ± 10% of the undisturbed acoustic frequency of the mode. The goal of the
designer is to manipulate the factors influencing n and τ in order to place the
operating point outside all the stable regions of the various modes.
The parameters τ and n (into which is lumped the modifying effects of

velocity or other sensitivities) are basically empirical, and a large data base has been
laboriously accumulated on their dependencies upon many design factors. As an
example, Figs. 2(a) and 2(b) (Ref. 26) show data on τ for coaxial injectors, for
which n ≅ 0.5 throughout.

Starting with the early of Priem and Guentert [29], these methods have been
progressively replaced by detailed numerical time-dependent simulations of the
combustion process. The complexity of these processes is such that, even now these
simulations still contain large elements of modeling and approximation, and are
generally limited to one or two selected spatial dimension plus time.
A recent example of this approach is described in Ref. 30. Here the emphasis
is on the tangential acoustic modes. A detailed 2-D fluid mechanical model (in the
transverse plane) is coupled to a series of empirically derived droplet vaporization
laws for UDMH and N2O4. The flow is turbulent, modeled using a K- ε approximation,
and the drops are allowed to slip, exerting drag forces which are computed from
empirical drag coefficients, and also modifying the vaporization rates due to
convective heat transfer. The drop-heating transients are ignored. The computations
yield detailed time histories of all the fluid parameters, and comparisons to limited
test data on parametric effects of pressure and injector type are found to be
favorable.
A similar computation, but for longitudinal modes only, is described in Ref. 27.
Here the spatial dependence is on one dimension only (axial), but the droplet
interactions are calculated in somewhat more detail, including drop thermal inertia.
These calculations show strongest instability when the ratio of the acoustic period to
the droplet vaporization time is 0.15, which, as noted before, can be interpreted as
indicating a “sensitive” time lag which is a fraction (0.1-0.2) of the total vaporization
time. The calculations also show cases of entropy wave excitation, for which the
frequency corresponds closely to the convective time in the chamber.
In what follows, we provide a simplified analysis of the acoustic effects of

several combustion-related phenomena (heat release, mass addition, etc.), and then
use this in conjunction with Crocco’s theory for an assessment of stability in a simple
1-D situation. Some general conclusions about stability and destability effects are
also drawn from the acoustic analysis.

Acoustic equations with heat and mass addition, fluid forces, and molecular mass
changes.
∂ ∂
= =0
∂z ∂y
∂ρ ∂ ( ρ u )
1. + =m (mass addition p.u. volume, p.u. time)
∂t ∂x
∂u ∂u ∂p
2. ρ + ρu + = f (force p.u. volume)
∂t ∂x ∂x
⎛ ∂T ∂T ⎞ q ⎛ ∂ (1 / p ) ∂ (1 / p ) ⎞
3. cv ⎜ +u ⎟ = − p⎜ +u ⎟⎟ (q=heating rate p.u. volume)
⎝ ∂t ∂x ⎠ p ⎜ ∂t ∂x
⎝ ⎠
p R
4. = T
ρ M
Assume small perturbations about a uniform steady background, (with µ, m ,f, q):
ρ = ρ + ρ ' , u = u + u' , p = p + p' , T = T + T ' , µ = µ + µ ' (5)
u =0 (6)
0
D() ∂ () ∂ () ∂ ()' ∂ ()' ∂ ()'
Because of (6), ≡ + u = + (u + u ') (7)
Dt ∂t ∂x ∂t ∂x ∂t
2nd order
Linearizing then,
∂ρ ' ∂u '
+ρ =m (8)
∂t ∂x
∂u ' ∂p ' ∂ 2u ' ∂2 p ' ∂f

ρ + =f (9) ρ + 2
= (12)
∂t ∂x ∂x ∂t ∂x ∂x
∂T ' p ∂p '
ρ cv =q+ (10)
∂t ρ ∂t
p' ρ' T ' M' ⎛ p' ρ ' µ ' ⎞
− = + (11) T ' = T ⎜⎜ − + ⎟ (13)
p ρ T M ⎝ p ρ µ ⎟⎠

(13) into (10):
⎛ 1 ∂p ' 1 ∂ρ ' 1 ∂M ' ⎞ p ∂ρ '

ρ c v T ⎜⎜ − + ⎟⎟ = q +
⎝ p ∂t ρ ∂t M ∂t ⎠ ρ ∂t
p' R
= T = Rg T
p M
⎛R
T⎜
⎞ ∂ρ '
+ cv ⎟ = −q +
c v ∂p '
+p
c v ∂ M '/ M ( )
⎝M ⎠ ∂t R g ∂t Rg ∂t
cp
∂ρ '
=
⎡
1 ⎢ 1 ∂p '
−q+
⎤
p ∂ M '/ M ⎥ ( ) (14)
∂t c p T ⎢ γ − 1 ∂t γ −1 ∂t ⎥
⎣ ⎦
Sub. Into (8)
⎡
1 ⎢ 1 ∂p '
−q+
⎤
p ∂ M '/ M ⎥(+ρ
∂u '
=m
) (15)
⎢
c p T γ − 1 ∂t γ −1 ∂t ⎥ ∂x
⎣ ⎦
Differentiate w.r.t. t:
⎡
1 ⎢ 1 ∂ 2 p ' ∂q
− +
2
(
⎤
p ∂ M '/ M ⎥
+ ρ
∂ 2u '
=
)
∂m (16)
⎢
c p T γ − 1 ∂t
2
∂t γ −1 ∂t 2
⎥ ∂ x ∂t ∂t
⎣ ⎦
∂ 2u ' ∂ 2 p ' ∂f
but, from 12, ρ = − + substitute in (16)
∂x ∂t ∂x 2 ∂x
⎡
1 ⎢ 1 ∂ 2 p ' ∂q
− +
2 ⎤
−
(
p ∂ M '/ M ⎥ ∂ 2 p ' ∂f
+ =
∂m ) (17)
⎢
c p T γ − 1 ∂t
2
∂t γ −1 ∂t 2
⎥ ∂x 2
∂x ∂t
⎣ ⎦
2
Now ( γ − 1 ) c p T = γ R g T = c ( c =speed of sound) (18)
∂2 p '
− c
2
2 ∂ p'
=
⎡
∂ ⎢ 2
c m + ( γ − 1 ) q − p
∂ µ '/ µ ⎤ (
⎥ − c 2 ∂f
) (19)
∂t 2
∂x 2
∂t ⎢ ∂t ⎥ ∂x
⎣ ⎦

( γ − 1) C p T ×
∂ (15 )
→ c ρ
2 ∂2u '
+
∂2 p '
= ( γ − 1 )
∂q
+ c
2 ∂m
− p
(
∂2 µ ' µ ) (20)
∂x ∂x 2 ∂x ∂t ∂x ∂x ∂t ∂x
∂(9) ∂ 2u ' ∂2 p ' ∂f

→ ρ 2
+ = (21)
∂t ∂t ∂x ∂t ∂t
Subtract, divide by ρ :
∂ 2u '
− c
2
2 ∂ u'
=
1 ∂f
−
⎡
1 ∂ ⎢ 2
c m + ( γ − 1 ) q − p
∂ M' M ⎤
⎥
( ) (22)
∂t 2 ∂x 2 ρ ∂t ρ ∂x ⎢ ∂t ⎥
⎣ ⎦
2
Note the combination c m + ( γ − 1 ) q − p
(
∂ M' M )
∂t
Adding heat, adding mass, or having a decrease rate of molecular mass, all are (up
to factors) equivalent acoustic disturbances.
To close the problem, one needs to relate the perturbations (m, q, f, ∂µ ' ) to
∂t
the state variables ( p ', ρ ', u ',T ' ), by looking at the particular mechanisms
(vaporization, combustion, etc), and how they depend on pressure, velocity, and so
on.
General Conditions for Instability
Looking at Eq. (19), we see that the effects of gas mass generation m, heat
∂µ '/ µ
generation q and molecular mass change − are similar. All of them have the
∂t
effect of increasing the local volume, and we suspect therefore that when these
quantities (acting together) peak when the pressure also peaks, we will have
unstable condition.
To examine this, define the quantity
2
Q = c m + ( γ − 1) q − p
(
∂ M' M ) (23)
∂t
and ignore for now the local disturbing forces f. Then
∂2 p ' 2
2 ∂ p' ∂Q
2
− c 2
= (24)
∂t ∂x ∂t

For a sinusoidal wave of the type
⎡ ∧ i ω t − kx ) ⎤
p '(t ) = Re ⎢ p e ( ⎥
(25)
⎣ ⎦
with ω complex and k real, we then have
∧ 2 ∧ ∧
− ω 2 p + k 2 c p = iω Q (26)
Define ω 1 iQ (27)
ν ≡ ; h= ∧
kc 2k c p
and re-write (26) as ν 2 + 2hν − 1 = 0 (28)
which has the complex solution ν = − h ± 1 + h2 (29)
We now ask what form h should have for stability. First, we note that at the
stability threshold, ν is real, and so h = (1 − ν 2 ) / 2ν must also be real. From (27),
∧ ∧
this means that i Q must be in phase (or counter-phase) with p , namely, “volume
addition rate” must be 90o ahead of or behind pressure oscillations:
More generally, re-write (28) as ν + 2h − 1 = 0 , and put explicitly h = hR + ihI ,

ν
ν = ν R + iν I :

νR − iνI
ν R + iν I + 2 hR + 2 i hI − =0 (30)
ν R2 + i ν I2
Separate real and imaginary parts:
νR
ν R + 2hR − =0
ν R2 + ν I2
(31)
νI
ν I + 2hI + =0
ν R2 + ν I2
1 ⎡ 1 ⎤
From (31b), hI = − ν I ⎢1 + 2 2 ⎥
(32)
2 ⎣ ν R + iν I ⎦
This shows that whenever hI > 0 , ν I < 0 . Since p ' ∼ e iωt = e −ω t e iω t ,ν I < 0
I R
⎛ ∧
⎞
implies instability. Also, from (27b), hI > 0 implies ⎜ Q∧ ⎟ > 0 , which means that the
⎜ ⎟
⎝ p ⎠R
“volume addition rate” Q(t) must have a positive projection on pressure p’(t),
namely, a part in-phase with it. This is a confirmation of the physical intuition that
releasing “volume” when pressure is high must be de-stabilizing. We repeat that this
may mean heat addition, gas addition (vaporization) or molecular mass reduction
rate (decomposition of complex molecules).
Let us now consider briefly the effect of body forces f. Returning to (19) and
defining the complex quantity.
∧
if (33)
ρ = ∧
kp
we can see that (28) is expanded to the form
ν 2 + 2hν (1 − ρ ) = 0 (34)
For neutral conditions (stability threshold), we must have real ν (ν I = 0) , so,

taking the imaginary part of (34),
2ν hI + ρ I = 0 (35)
In the absence of forces, we found that hI > 0 would lead to instability. We see now
that if ρ I = −2ν hI , which is negative, then the process stabilizes at least to the point
⎛ ∧
⎞
of neutral stability. From (33), the conclusion is that ⎜ f∧ ⎟ < 0 is stabilizing, i.e., the
⎜ ⎟
⎝ p ⎠R
body forces should be in the backwards direction (against H velocity) when pressure
is high, and vice-versa.

Looking at one instant of time vs. distance:
So, the body forces are in quadrature with the pressure gradient force ( −∇ P ), and
are taking energy away from the wave motion.
The “Sensitive time lag” theory (Harrje and Reardon, p.1)
Basic scenario:
- Drops injected at t − τ T complete evaporation and combustion at
- However, from t − τ T to t − τ , no complete evaporation occurs, only

“precursor processes”. τ ≡ ”sensitive lag.”
- The rate of evaporation + combustion during τ is sensitive to pressure and/or

∂ln ( rates )
velocity. For p, ≡ n (1)
∂ln P
(n= “sensitivity index”).
- The duration τ of actual evap. + comb. changes in time, in response to these

rate changes. However, the total mass burnt is that of the drop, and it is
assumed their mass (and their number) are independent of P, υ in the
chamber.
Say, R is the relevant rate. Under “quiet” or “mean” conditions, R = R . When P ≠ P ,

⎛ p − p⎞
R R ⎜1 + n ⎟ (2)
⎜ p ⎟⎠
⎝
Express that the total mass burnt in ( t − τ , t ) is always the same, equal to
that burnt under mean condition
t
⎛ p − p⎞
t
∫τ R ⎜⎜ 1 + n ⎟ dt1 = ∫ R d t1 (3)
t− ⎝ p ⎟⎠ t −τ
t −τ t
P −P
∫τ
t−
R dt1 + R n ∫ P
dt1 = 0
t −τ
here replacing τ by τ is ok.

(2nd order error)
t
P −P
τ − τ = −n ∫ dt1 (4)
t −τ P
Let now, specifically, m be the rate of gas generation from liquid (local, per
unit volume). The liquid injection rate is constant, equal to m . The gas generated in
(t, t+td) is mdt. This gas originates from liquid that reached its “maturity” for
vaporization between t − τ and ( t − τ )+d ( t − τ ), and since liquid arrives at m ,
m dt = m d ( t − τ )
m−m dτ
= − (5)
m dt
dτ P (t ) − P (t − τ )
From (4) = −n
dt P
m−m P (t ) − P (t − τ )
and so =n (6)
m P
m' P ' (t ) − P ' (t − τ )

or, in terms of perturbations, =n (6’)
m P
NOTE: Eq. (3) is wrong in Harrje and Reardon.

Control of Instabilities
Despite all efforts at avoidance through design, the need remains for devices
that will help damp the many potential modes of instability in any given rocket
combustor. A new technology for active control is now evolving [31], in which
feedback controlled acoustic generators are used to cancel unstable waves. Since the
growth can be detected at small amplitude, it may not be necessary to inject very
large acoustic powers for this purpose. Nevertheless, the mainstay of current
practice is based on passive damping methods. A good description of these methods,
with design guidelines, is given in Ref. 32.
The most important high frequency stabilization devices are injector head
baffles and acoustic absorbers. Baffles are radial or circumferential barriers attached
to the injector head and extending 0.1-0.2 diameters in the axial direction. An
example of a baffled injector is that of the SSME (Fig. 3, from Ref. 2). The exact
mechanism by which baffles enhance stability is not well understood, which has led
to some divergence in design. It appears that the effect is related to the sensitivity
of the droplet velocity cross-over point, which occurs quite close to the injector face,
and it may involve disruption of the tangential gas motion associated with tangential
modes, or shifting of the local acoustic frequencies to values above the characteristic
drop vaporization frequencies.
Acoustic absorbers are cavities on the chamber walls with relatively narrow
connecting channels to the chamber, so as to dissipate power during pressure
oscillations in their vicinity. Their action is much better understood than that of
baffles, and designers can proceed with some confidence, using methods described,
for example in Ref. 32. Absorbers are often located on the cylindrical walls, near the
injector, or as “corner slots” between injector and cylinder. They can also take the
form of a continuous double wall with an array of holes periodically arranged to
connect to the chamber. The absorption coefficient of a well-designed absorber can
be high over a relatively wide frequency band, so as to contribute damping to the
most prevalent modes. Sometimes several different absorbers are used, each tuned
to a different frequency. Fig. 4 (Ref. 32) shows a baffled injector with corner
absorbers, and Fig. 5 (Ref. 32) shows an extended acoustic liner.
A simplified analysis of an acoustic absorber is described next.

Acoustic Absorbers: Resonators
The schematic shows a device

which is a variation on the
Helmholtz Resonator concept.
Consider first the half-

cycle during which flow is leaving
•
the resonator ( m >0). The
pressure inside the resonator is
PR, while at the entrance to the
narrow inlet duct, it is
1
PR' = PR − ρ v 2 , due to the
2
(subsonic) acceleration towards
the inlet. The density changes δρ
can be equated to δρ / c2 (c =
speed of sound in the resonator).
The mass balance is then
VR dPR •
2
= −m (1)
c dt
and the momentum balance in the duct is
d
dt
(
( ρ Ai Lv ) = Ai PR' − Pc )
•
or, using m = ρ A i v ,
• ⎛ • 2 ⎞
d m Ai ⎜ m ⎟ (2)
= P − P −
L ⎜⎜ 2 ρ A i2 ⎟⎟
R C
dt
⎝ ⎠
•
Consider next the other half-cycle, when the cavity is filling ( m <0). The
pressure at the chamber-side of the duct, which is now the flow inlet is
1
Pc' = Pc −
2
( )
ρ −v 2 , whereas that at the exit of the duct is now just PR. Eq. (1) still
holds, whereas the momentum balance is now
• ⎛ • 2 ⎞
d m Ai ⎜ m ⎟ (3)
= ⎜ P − P +
2 ρ A i2 ⎟⎟
R C
dt L ⎜
⎝ ⎠

We now differentiate (2) and (3), and unify the two half-cycle momentum equations
as
⎛ •
⎞
•
⎜ dP m •
⎟
d2 m Ai d m dPc
= ⎜ R − − ⎟
dt 2 L ⎜ dt ρ A i2 dt dt ⎟
⎜ ⎟
⎝ ⎠
Substituting Eq. (1) here,
•
• m •
2
d m dm Ai • A i dPc
+ + c2 m=− (4)
dt 2
ρ L A i dt LVR L dt
•
This equation is non-linear in m , but we can obtain reasonable results for steady
•
state operation if we replace m by its time average over one cycle. The analysis
•
reduces them to that of a forced linear oscillator, except that m needs to be
calculated self-consistently rather than being a prescribed quantity.
Assume solutions of the form

⎡ ∧• ⎤
•
⎡ ∧ ⎤
m = Re ⎢ m e iω t ⎥ , δPR = R e ⎢ δPR e iω t ⎥ ,etc. (5)
⎢ ⎥ ⎣ ⎦
⎣ ⎦
Eq. (4) becomes
⎛ • ⎞
⎜ m ⎟ ∧
⎜ 2 Ai ⎟ • Ai ∧
(6)
⎜ −ω + ρ L A i ω + c L V ⎟ m = − i ω L δ pc
2
⎜ i R ⎟
⎜ ⎟
⎝ ⎠
while Eq. (1) gives
∧
V ∧ •
i ω R2 δ PR = − m (7)
c
and solving (6) and (7) together,
c2 A i
∧
δ PR L VR (8)
∧
=
•
δ Pc m
c2 A i
+ i ω − ω2
L VR ρ L Ai
Ai
From this, the natural frequency of the oscillations is seen to be ωn = c (9)
L VR
and the damaging factor ζ is given by
• •
m m
L VR
2 ζ ωn = , or ζ = (10)
ρ L Ai 2 ρ L Ai c Ai
∧
•
For the “rectified sine wave” m (t), we can see that m = 2 m , and from (7) and (8),
• •
c2 A i
•
2 VR ∧
2 V L VR ∧
(11)
m = ω δ PR = ω R2 δ Pc
π c 2
π c •
m
c2 A i
+ i ω − ω2
L VR ρ L Ai

2
• ⎛ • ⎞
m ⎜ m 2 ⎟
2
⎛ c Ai
2
⎞
Using c A i − ω 2 + iω =
⎜
− ω2 ⎟ + ⎜
⎟ 2 , this yields a bi-quadratic
L VR ρ L Ai
⎜
⎜ L VR ⎟ ρLA ⎟ ω
⎝ ⎠ ⎜ i ⎟
⎜ ⎟
⎝ ⎠
• ∧
equation for m in terms of the amplitude δ Pc of the pressure fluctuations in the
chamber. The algebra is simpler in non-dimensional terms. Define
ω ω LVR
(12)
ν ≡ =
ωn c Ai
∧
δ Pc
2 VR (13)
ρ ≡
π L A i 1 ρ c2
2
2 2
• ⎛ ρ c Ai 2 2
⎞ ⎡
We then obtain m = 1 ⎜ ( ) (1 − ν ) ⎤
2 4
⎟ ⎢− 1 − ν 2 + 2
+ p2ν 4 ⎥ , and substituting in
2 ⎜⎝ ω VR ⎟ ⎣
⎠ ⎦
(10), the damping ratio is
p2ν 2 / 2 (14)
ζ2 =
( ) ( )
2 4
1 −ν2 + 1 −ν2 + p2ν 4
Several points can now be made:
(a) For any frequency, the damping increases with pressure fluctuation intensity.
This is a favorable circumstance because we need the damping most when
combustion is rough. Mathematically, this is a consequence of the non-
linearity of the equation. Physically, energy is dissipated both during
aspiration and during expulsion of gas from the cavity, (by the mixing out of
the jet kinetic energy), and it is clear that more energy is dissipated when the
driving pressure differences are stronger.
(b) Although the algebra is still tedious, differentiation of (14) shows that ζ is
maximum at ν = 1 , i.e., when the cavity is tuned in resonance with the
pressure fluctuations (by selecting parameters so that ωn = ω ). Putting ν = 1
in (14) gives
∧
δ Pc
VR
ζ MAX = p /2 = (15)
π L Ai 1
2
ρc2
which again shows ζ increasing with fluctuation intensity.

(c) Since we should design for resonance for the given ω , the dimensions should
Ai ω2
satisfy = 2 . Eq. (15) can then be put in the two equivalent forms
LVR c
∧ ∧
δ Pc δ Pc
1 2 VR 2
ζ MAX = = ω (16)
ωL π ρ Ai π ρ
This indicates we should use short inlets, large cavity volumes and
Ai ω2
small inlet areas (subject to = 2 ).
LVR c
As an indication of what is possible, consider 1% fluctuations on a

P P 8.3
pressure Pc such that c = RgTc is c = × 3000 = 1.25 × 106 m2 / s2 . The
ρ ρ 0.02
acoustic frequency to be damped is at f = 2000 Hz, or ω = 2 π f = 12 , 5000 m/s.
Then from the first of Eqs. (16), we can obtain critical damping ( ζ MAX = 1 ) by
choosing an inlet length
∧
δ Pc
1 2 1 2
L=
ω ζ MAX π ρ
=
12 , 500 π
(1.25 × 10 ) 4
= 7.2 × 10−3 m = 7.2 mm
which appears reasonable. To ensure resonance, then, we need to select.
VR c2 γ Rc Tc 1.2 × 1.25 × 106

= 2 = = 1.34 m
ω L ω2 L (12 , 500)
2
Ai × 7.2 × 10−3

A possible choice of geometry is then to take an inlet diameter of, say, 5mm, and a
cavity volume
π
VR = 1.34 0.0052 = 2.63 × 10−5 m3 = 26.3 cm2
4
If this is shaped as a cubic cavity, its side is about 3 cm. These dimensions are
sketched (roughly to scale) below.
As a final note, remember that the

“lumped parameter” idealization
used in the theory may not be very
precise. Only semi-quantitative
accuracy is to be expected, but the
trends should be correct and the
analysis can be used for preliminary
design, to be refined through
numerical simulation or physical
testing.
How many acoustic dampers:
1 •
δξ R = δ m v 2 + ρ δ PR VR =
2
• •
d (δξ R ) 1 • dδ PR m δm
= δ m v δv + VR v = δv =
dt 2 dt ρ Ai ρ Ai
With no dissipation
• •
d m Ai d δ m Ai L •
= ( PR − PC ) = δ PR δm
dt L dt L Ai
• VR dPR d δ PR c2 • VR
m=− =− δm δ PR
c 2 dt dt VR c2

d ⎛ L • δP ⎞ d ⎛ L 2 ⎛ • ⎞2 2⎞
⎜⎜ δ m + VR 2R ⎟⎟ = 0 × c2 ⎜ c ⎜ δ m ⎟ + VR (δ PR ) ⎟ = 0
dt A c dt ⎜A ⎝ ⎠ ⎟
⎝ i ⎠ ⎝ i ⎠
2
L 2⎛ • ⎞
c ⎜ δ m ⎟ + VR (δ PR )
2
2
Ai ⎝ ⎠ L ⎛ • ⎞ VR
2 (
δ PR )
2
ζR = = ⎜δ m⎟ +
2ρc 2
2 ρ Ai ⎝ ⎠ 2ρc
With dissipation
⎛ • • ⎞
• ⎜ m δ m⎟
dδ m A i ⎜ ⎟ c2 L •
= ⎜ δ PR − 2 ⎟
δm
dt L ρAi Ai
⎜ ⎟
⎜ ⎟
⎝ ⎠
d (δ PR ) c2 •
=− δm VR δ PR
dt VR
d ⎡ L 2 ⎛ • ⎞2 2⎤
c ⎜ δ m ⎟ + VR (δ PR ) ⎥
•
⎢ m
dt A
⎢⎣ i ⎝ ⎠ ⎦⎥ = ⎛ • ⎞
2
⎜ δ m ⎟ (energy dissipation rate) ζ R

2 ρ c2 2 ρ 2 A i2 ⎝ ⎠
3
•
m ∧
• •
2 •
Average εR = m = m
2 ρ 2 A2i π
∧
• Ai ∧ ρ L Ai ρ A2i ∧
At resonance, m = − δ Pc = − δ Pc
L • •
m m
2 ρ Ai
2
• ∧ •
2 ∧
m = δ Pc m = ρ A2i δ Pc
π •
π
m
3/2 3 3/2
•
1 ⎛2 ∧
⎞ 1 ⎛2⎞ 2 Ai ∧
εR = 2 ⎜
ρ A2i δ Pc ⎟ = ⎜ ⎟ δ Pc
2 ρ Ai ⎝ π
2
⎠ 2 ⎝π ⎠ ρ

2
∧
δ Pc
Average chamber acoustic energy ξc = Vc
ρ c2
• 3
1 ζR 1 N ⎛2⎞ 2 Ai ρ c2 2 NA i ρ
( ∆ωI )ch. =N = = c2
2 ξc 2 2 ⎜⎝ π ⎟⎠ ρ ∧ 2π 3/2
Vc ∧
Vc δ Pc δ Pc
contribution of ωI of
chamber, due to N
resonators Ai
Also, (ωn )c = (ωn )R = c
L VR
( ∆ωI )c 1 N Ai ρ 1 L VR 1 N ρ c2
∆ζ c = = c2 = L VR A i
ωn 2π 3 Vc δ Pc c Ai 2 π3 Vc ∧
δ Pc
δ Pc 1.25 × 104 1
Example ω = 2π × 2000 Hz = =
ρc 2
1.2 × 1.25 × 106
120
π
L = 7.2 × 10−3 m VR = 2.63 × 10−5 m3 Ai = 25 × 10−6 m2
4
and say Vc = 0.1 m3
1 N π
∆ζ c = 7.2 × 10−3 × 2.63 × 10−5 × 25 × 10−6 120 = 2.7 × 10−5 N
2π ∋ 0.1 4
So,for ∆ζ c = 0.1 , need 3730 dampers.

Say l = 4 cm
D = 40 cm
π D lT N l 2 3730 × 16
N= lT = = = 475 cm
l 2
πD π × 40
Not enough room
Can get ∼ ∆ζ 0.02 in lT = 1m

Lecture 24: Pressurization and Pump Cycles
Thrust Chamber Pressurization
1. Introduction
The most technically demanding part of a modern high pressure liquid

propellant rocket is the chamber pressurization system. Consider for example the
cluster of four RD-170 engines powering the Energia first stage. Their shared
oxygen pump must deliver 1792 Kg/sec (1.63 m3) of LOX at 614 atm pressure to
the combustors, for a power of 176,000 HP. The whole engine weighs 9500 Kg.
Similarly, the SSME LH pump (see Fig. 3 in lecture 25) delivers 73 Kg/sec of LH
(1.06 m3/sec) at 470 atm pressure, for a power of 76,000 HP. The SSME mass is
2870 Kg, of which the LH turbopump is 344 Kg. And these flow rates are both
dwarfed by those of the old F1 Saturn engine, which swallowed a small river (3.5
m3/ sec) of LOX, albeit at a more modest pressure.
Because of the importance of this system, we will devote this lecture to

engine pressurization cycles and components.
2. Gas Pressurization Systems
The simplest way to achieve the required thrust chamber pressure is to

provide a small, high pressure gas reservoir, which, at firing time, pressurizes the
propellant tanks. The tanks must then be thick enough to support the thrust
chamber pressure, plus the injector drop. In addition, the gas reservoirs are also
relatively heavy. Thus, the performance of a gas-pressurized rocket lags that of a
pump-pressurized rocket (where the tanks can be much lighter), as shown in Fig.
1 (from Ref. 40). We notice here that the relative size of the gains due to a turbo
pump system become overpowering for high ∆V rockets, but may not be worth
the extra complexity for small ∆V

Payload Weight/Gross Weight
0.60
Percent Improvement
with Pump-Fed Engine
200
0.40
Pump Fed
Pressure Fed
100
0.20
0 0
0 10,000 20,000 30,000 40,000
Mission Velocity (fps)
Fig. 1. Payload ratio for typical gas pressurization and pump pressurization
Thus, gas pressurization continues to play a role, particularly in small space-

based engines, such as monopropellant maneuvering and attitude control
thrusters. A relatively large and sophisticated example is the Shuttle OMS/RCS
system.
If a gas pressurization system is adopted, a choice must be made between

regulated or blow-down gas delivery. The regulated type is used most often,
because it avoids chamber operation over an extended pressure range, which
may be lead to stability problems. On the other hand, blow-down operation does
result in a simpler and lighter system, with less gas inventory, and is standard in
hydrazine monopropellant applications.
Consider a regulated gas feed system, schematized in Fig. 2. Let PG ( t ) be the

decreasing pressure in the gas reservoir, of constant volume VG and Pp the
regulated tank pressure, where Vp ( t ) is the increasing ullage volume in both
UG
+
UP
propellant tanks. The internal energy of the gas at some times t is and
the work of gas expansion in dt is Pp dVp . Therefore, integrating in (0,t),
(assuming adiabatic expansion)
UG + Up + Pp Vp = UGo (1)

NOTE: If isothermal exp., UG + UP = UGo and then PG ( t ) VG + Pp VP ( t ) = PGo VG
which can be re-written for an ideal gas as
⎛ Pv ⎞
⎜ U sin g U = ⎟ PG ( t ) VG + γ Pp VP ( t ) = PGo VG (2)
⎝ γ − 1⎠
This relates PG (t) to VP (t) at any time during operation. If the final gas pressure
is PGe (large enough to still overcome the regulator and injector drops), and if VTK
represents the whole tankage volume, we obtain for the gas reservoir volume
γ Pp
VG = VTK (3)
PGo − PGe
(no γ for isothermal case) (more generally in between)
The minimum mass of this gas reservoir, assumed spherical of radius RG, can
be estimated by noting that its wall thickness tG must be
Pgo
tG = RG (4)
2σGW
where σGW is the working stress of the wall material. If the density of this wall
material is ρGW , we obtain for the gas tank mass
3 ρGW γ Pp VTK
MGTK = (5)
2 σGW ⎛P ⎞
1 − ⎜ Ge ⎟
⎝ PGo ⎠
A similar calculation can be made for the mass of each of the propellant tanks. If,
for simplicity, both, oxidizer and fuel tanks are assumed to be geometrically
similar and made of the same material (a cylindrical body of length Lp, capped by
hemispheres of radius Rp, with equal Lp/Rp for both ), then the propellant tank
mass is
3 ρPW 1 + Lp / Rp
MPTK = Pp VTK
2 σPW 1 + 3 Lp / Rp
4

REGULATOR
PP PP
VP(t)
High
OXIDIZER
Pressure
FUEL
Gas
PG(t)
VG
Fig. 2. Schematic of gas pressurization system
We notice from (5) that the gas tank mass is nearly independent of its initial
pressure: a higher PGo allows a smaller volume VG (Eq. 3), but requires thicker
walls (Eq. 4). Values of PGo 5-10 times PP are common. To compare the mass of
PG
/
P
the gas tank to that of the main propellant tanks, assume = 0.2, γ = 5 3
e
G
o
(He gas) and Lp Rp = 6. For equal material ratio ( ρ σ ), Eqs. (5) and (6) then give
MGTK
= 1.64 , indicating that the gas reservoir is likely to be the heavier
MPTK
component (although adding anti-slosh baffles, insulation, etc. may modify this
conclusion).

The mass of the propellant itself is Mp = VTK ρp , where the densities of oxidizer
and fuel have been averaged to ρp . To compare to the minimum tankage mass,
assume steel construction, with σ ρ ≅ 3.4 × 104 (m / sec)2 and Pp = 20 atm. Eq. (5)
then gives MGTK / VTK = 184 Kg / m3 , and Eq. (3.6) gives MGTK / VTK = 112 Kg / m3 , for a
total of 296 Kg / m3. This is to be compared to the mean propellant density; for
N2O4-UDMH this is of the order of 1000 Kg/m3. Thus, not counting flanges valves,
regulators, etc., the tankage mass is 30% of the propellant mass, a figure which
is excessive for any ambitious mission. Advanced composite materials can reduce
this mass significantly, however, Titanium is used mostly.
An alternative method to reduce the required gas reservoir mass is to heat

the pressurizing gas prior to injection into the propellant tank. This can be
accomplished through heat exchanging, perhaps using the gas as a nozzle
coolant, or by including in the gas a very small amount of oxygen and hydrogen,
below the flammability limit, and then passing the gas through a catalytic bed
reactor. In either case, assuming a gas temperature rise ∆ T is accomplished, a
term MG cv ∆T needs to be added on the right hand side of Eq. (1), where (MG(t))
is the gas mass that has flowed through the heater. This yields eventually a
modified expression for the gas reservoir volume:
=
γP
p
V V (7)
Δ
G
T
K
+
⎛ T⎞
P ⎜1 ⎟-P
G
o
G
e
T ⎠
G
o
As an example, 1% O2 plus 2% H2 (by mole) in He yields upon reaction

∆T ≅ 230 K , which nearly cuts in half the required gas tank volume (and mass).
One problem with heating methods is the potential for erratic variations in feed
pressure if propellant sloshing cools the pressurizing gas in an unsteady manner.

Lecture 25: Basic Turbomachine Performance
Turbopump Pressurization Systems
1. Cycles
For higher performance, mechanical pumps must be used to feed the

combustion chamber. In turn, these pumps require drive power, which is always
provided by turbines using excess thermal energy in the propellants (although
electrical motors have been considered for small rockets). The manner in which hot
gas is provided to drive the turbines serves to distinguish among several
pressurization cycles, of which the most important are summarized in Fig. 1. (From
Ref. 41). The most common of these are the bi-propellant gas generator (G.G.)
cycle, the expander cycle and the staged combustion cycle.
(a) The Gas Generator Cycle
The GG cycle was used in the F1 engine, and is also in use in the Delta II,
Atlas and Titan rockets. In this cycle, a small fraction of the pressurized oxidizer and
fuel are diverted to a medium-temperature burner (Gas Generator), which produces
typically very fuel-rich gas to drive the turbine or turbines. These are designed with a
large pressure ratio, and their exhaust is either dumped overboard, or injected at
some point into the main nozzle to provide some extra thrust. Nevertheless, this
cycle is inherently somewhat lossy, in that the turbine gas is not fully utilized in the
main combustor. On the other hand, the power control is relatively straightforward,
and there is little interaction of the feed system with the rest of the rocket. Any
propellant combination can be used, all power levels are suitable, and any desired
pressure level can be obtained although the Isp loss increase with pressure (1.5-4 sec
per 100 atm). The mechanical power required to drive the pumps is
• ∆POP • ∆PFP
PP = mox + mF (1)
ηoP ρox ηFP ρ F
where ∆POP and ∆PFP are the pressure rise in the oxidizer pump (OP) and fuel pump
(FP), respectively which have efficiencies ηOP , ηFP . Also ρox , ρF are the liquid densities.
•
If the gas generator mass flow rate is mGG , the (single) turbine power is
•
PT = mGG ηT cP' TnηTT
'
(2)
in which ηT is the turbine isentropic efficiency (60-80%), η'TT is its thermodynamic

expansion efficiency
γ ' −1
ηTT
'
= 1 − ( Pie / Pn ) γ' (3)
which depends on pressure ratio Pie / Pn and GG gas specify heat ratio, γ ' . Also, cP' is
the specific heat of this gas, and Tti its temperature, which is controlled through
stoichiometry to values acceptable by uncooled turbines (700-1100K).
Ref. 41: Turbopump System for Liquid Rocket Engines. NASA SP-8107, Aug. 1974.

•
Equating PP and PT yields the required mGG . Suppose now the turbine exhaust is
expanded to the same exit pressure Pe as the main flow. The exhaust speed is then
⎡ γ ' −1
⎤
cGG = 2 cP' Tt .out ⎢1 − ( Pe / Pte ) γ ⎥ (4)
⎣ ⎦
where
(
Tt .out = Tn 1 − ηT ηTT
'
) (5)
This speed is generally lower than the main flow exit speed,
γ −1
c = 2 c p Tc η ; η = 1 − (Pe / Pc ) γ
(6)
where cp and η belong to the nozzle gas, and Pc , Tc , are the chamber pressure and
temperature. The relative Isp loss is:

•
δ Isp mGG ⎛ c ⎞
− ≅ • ⎜1 − GG ⎟ (7)
Isp c ⎠
m ⎝
and can be calculated once the turbine exhaust pressure Pte (and hence the turbine
pressure ratio) is selected. A tradeoff is involved here: if Pte is not very low, too
•
mGG
much mass must be diverted (large •
in Eq. 7), where as if Pte is too low, the
m
cGG
exhaust provides almost no additional thrust (small in Eq. 7). An optimum can
c
therefore be found.
•
As an example, consider the LOX-RP1 F-1 cycle, for which mox =1844 Kg/sec,
•
mF = 777 Kg / sec, ∆POP = 1.06 ×107 N / m2 , ∆PFP = 1.24 ×107 N / m2 , Tti = 1061 K and
ηOP = 0.746, ηFP = 0.726, ηT = 0.605. Also, ρox = 1145 Kg / m3 , ρF = 810 Kg / m3 . The GG
is estimated to have γ' = 1.35 and cp' = 2140 J / Kg / K , while the main gas has
γ = 1.25, cP = 2080 J / Kg / K . The chamber pressure is Pc = 95 atm, and at the exit, Pe
= 0.68 atm. Use of the equations above yields, after some searching, an “optimum”
δI
turbine pressure ratio of 23, for which sp = -0.0111. The actual F-1 engine had a
Isp
turbine pressure ratio of 16.4.
(b) The Expander Cycle
For engines, utilizing hydrogen (and possible methane) as fuel, the gas
generator can be eliminated. Instead, the fuel is simply routed from the exit manifold
of the nozzle cooling circuit to the turbine inlet. This is possible because hydrogen is
supercritical at the pump exit, and it simply expands smoothly into an ordinary gas
as it picks up heat. The resulting “Expander Cycle” is simple and efficient (the fuel is
fully utilized in the thrust chamber). This cycle is used in the RL-10 engine and in the
start-up sequence of the Japanese LE-5 engine (which then transitions to gas-
generatory operation).
The principle limitation of this cycle is the relativity small amount of heat
available from regenerative cooling, which limits applicability to chambe pressures
under approximately 70 atm. A simple analysis can demonstrate this point. The
pump power is given by Eq. (1) and the power derived from the fuel-driven turbine is
⎡ γ F −1
⎤
⎢ ⎛ Pinj ⎞ γF
⎥
PT = mFηT cPF Tι i ⎢1 − ⎜ ⎟ ⎥ (8)
P
⎢⎣ ⎝ ι i ⎠ ⎥⎦
where Pinj , the injector pressure, is also the turbine exhaust pressure. The turbine

inlet pressure Pn , is related to the fuel pump pressure rise ∆PFP and the cooling
circuit loss, ∆Pcool by
∆PFP = Pti + ∆Pcool − PTK (9)
while the oxidizer pump has ∆POP = Pinj − PTK . The shaft power balance then gives
1
⎛ ρF ⎞ 1 − δTK π T + δ cool − δTK ⎛ γ F −1
⎞
(O / F ) ⎜ ⎟ + =ψ ⎜1 − π T γ F ⎟
⎜ ⎟
(10)
⎝ ρox ⎠ ηOP ηFP ⎝ ⎠
where
Pinj ρ F cPF Tti

δ TK = PTK / Pinj , δ cool = ∆Pcool / Pinj , π T = , and ψ = ηT (11)
Pti Pinj
Assuming δ TK = 0.1, δcool = 0.2, ηOP = ηNF = 0.7, O / F = 5 and ρF / ρox = 69/1140 = 0.0605 ,
the relationship between πT and ψ , as given by (10) is shown plotted in Fig. 2. As the
figure shows, the turbine inlet pressure increases rapidly when ψ drops below about 30
(at which point 1 = 1.33 ). In fact, as also shown in Fig. 2, the quantity ψ has a
πT
minimum value of approximately 17.76 when 1 = 2.68, below which no solution
πT
exists. The turbine inlet temperature Tti is 200K in the RL-10. Using also
ηT = 0.7, ρF = 69 kg / m3 and CPF = 14,600 J/Kg/K, we calculate from (11) ψ ≅ 1390 /
Pinj (atm) and so the maximum Pinj is 1390/17.76=78 atm. For reference, the RL-10 has
PC = 32 atm, Pinj ≅ 40 atm, which corresponds to ψ = 35. Of course, as (11) indicates,
higher Pinj values could be achieved if Tti could be increased further.

(c) Staged Combustion Cycle
For rockets where high chamber pressure as well as high efficiency is desired,
the staged combustion cycle is the preferred choice. One could think of this as a
modified expander cycle, in which a small amount of oxidizer is added to the fuel
after the cooling circuit, thus increasing the available enthalpy for the turbine drive.
As in the expander cycle, all of the propellant is entirely used in the combustion
chamber. Unlike the expander, through, any oxidizer-fuel combination can be used.
Two prominent examples of this cycle are the Space Shuttle Main Engine (SSME),
and the Russian RD-170 booster engine (Figs. 3 and 4 from Ref. 2). In the SSME, the
pre-burners are incorporated into separate fuel and oxidizer turbopump assemblies,
and process most of the fuel (LH) with a small fraction of the oxidizer (LOX),
producing a light “vitiated hydrogen” turbine driving gas. In the RD-170, the pre-
burners process all of the oxidizer (LOX) and a fraction of the fuel (kerosene) to
produce a fuel-lean gas which drives the single central turbine. In both cases, the

turbine exhaust is ducted to the main combustor injectors, together with the
remaining LOX (SSME) or kerosene (RD-170). The choice of fuel-rich preburners is
precluded by carbon deposits on turbine blades and other surfaces when
hydrocarbon fuel is involved.

The staged combustion cycle provides the highest levels of rocket
performance, but at the cost of greatly increased complexity. This is both, because of
the many ducts and valves involved, and because of the very high pump exit
pressures. A secondary potential difficulty, which is shared by the expander cycle, is
that the turbines are unchoked and there is a possibility of low-frequency instability
developing.
Consider a simplified schematic (Fig. 5) of a fuel-rich cycle analogous to that

of the SSME. The injector pressure, Pi, the turbine inlet temperature, Tti , and the
overall O/F ratio, r, are prescribed. The heat input and pressure drop in the cooling
circuit (Qcool, ∆ Pcool) are also assumed given. We wish to determine the required
pressure rise ∆ PFP by the fuel pump, as well as the O/F in the pre-burners (rPB) and
the fuel split SF between them. The pressure rise in the oxidizer pump is simply
∆POP = Pi − PTK .

The analysis is best done iteratively. If ∆ PFP is temporarily assumed known,
then the turbine inlet pressure is
Pr i = PTK + ∆PFP − ∆Pcool − ∆PPB (12)

where ∆PPB is the pressure drop in the pre-burner (mainly associated with injection).
The required rise in the oxidizer booster pump is then
∆POBP = Pti + ∆PPB − Pi (13)
The energy balance in one of the pre-burners is written as
⎛ ∆POP ∆POBP ⎞ ∆PFP r

rPB ⎜ hOTK + + ⎟ + hFTK + + Qcool + PB h f = (1 + rPB ) CPti (Tti − T ref ) (14)
⎝ ρ0 ηOP ρ0 ηOBP ⎠ ρ F η FP rst
where hOTK and hFTK are the enthalpies of oxidizer and fuel in their tanks (this ignores
the low pressure booster pumps) rst is the stoichiometric O/F ratio and hf the fuel
heat value at the reference temperature Tref . The specific heat of the fuel-rich burnt
gas is cpti. Eq. (14) can be solved for rPB if ∆PFP is assumed known.

The shaft power balance for the fuel turbopump is
⎡ r ' −1
⎤
∆PFP ⎛
⎢1 − iP ⎞ r'
⎥
= S F (1 + rPB ) c pti Tn η FT (15)
ρ Fη FP ⎢ ⎝⎜ Pti ⎠⎟ ⎥
⎢⎣ ⎦⎥
where η FT is the turbine efficiency and γ ' = γ ti belongs to the pre-burner gas. A
similar balance can be written for the oxidizer turbopump, and, by division, we can
solve for the fuel split SF.
−1
⎡ η ρ η ⎛ ∆P ∆P ⎞ ⎤
S F = ⎢1 + FT F FP ⎜ r OP + rPB OPB ⎟ ⎥ (16)
⎣ ηOT ρ0 ∆PFP ⎝ ηOP ηOPB ⎠ ⎦
After rPB has been calculated from (14), SF is given by (16), and then (15) can be
used as a check on the assumed ∆PFP . In reality, an outer iteration loop is required
by the fact that γ ' and cpti themselves depend sensitively upon the preburner
stoichiometry, rPB. With the approximations cpH2 ≅ 7.67 cal / mol / K , and
c pH 2o ≅ 10.63 cal / mol / K
7.67+0.37rPB (17)
γ '=
5.68+0.37rPB
and then (for H2-O2),
γ' R
cρti = ; M’ = 2 (1+rPB) g / mol (18)
γ '− 1 M 1
As an example, Fig. 3.8 shows some results in which we have used
Tref = 0 K hOTK ≅ hFTK ≅ 0
and also
r = 6, rst = 8, POTK =2 atm PFTK =1 atm, h f = 1.21x108 J/Kg, Tti =1100 K, η FP = 0.741
ηop =0.781, ηOBP = 0.696, η FT = 0.790, ηOT = 0.729,
∆Pcool = 0.15 Pc , ∆PPB =0.05 Pc ,
Qcool =7.40×105 J/kg, ∆Pi =Pi -Pc = 0.1 Pc
As shown in Fig. 6, the pressure rise in the fuel pump increases more steeply than
the chamber pressure, just as it did in the expander cycle. However, very high
pressures are now attainable. The ∆PFP for the SSME is shown for reference. Notice
how the pre-burner stoichiometry is essentially fixed by Tti , and does not vary much
over the pressure range.

Lecture 26: Turbopumps
Turbopump Components
The mechanical components of the pressurization cycle (pumps and

turbines) are next to be considered. An excellent recent survey of this area is given
in Ref.40. A more comprehensive, but older survey is contained in a series of NASA
SP reports [41-43]. Pumps and turbines will first be discussed separately, and their
integration will then be examined.
(a) Pumps
Almost all existing rockets have centrifugal turbopumps. These deliver more
∆ P per stage than axial flow pumps, with only slightly less efficiency. Only if
multistaging becomes necessary is there a possible incentive to go to axial pumps;
this happens with LH2 fuel, where, due to the low density, the ∆ P per stage is limited
by the attainable rim speeds.
In general, the design attempts to maximize operating speed, since this

reduces the pump size, and hence the weight. Pump speed is limited by several
effects, most importantly cavitation at inlet. Others are centrifugal stresses (either at
the impeller or in the driving turbine), limiting peripheral speeds for bearing and
seals, and avoidance of critical speeds.
“Head rise” is used commonly instead of pressure rise to express the

performance of pumps. We can define head rise as the height to which one could
raise one Kg of fluid with the amount of ideal work per Kg done by the pump:
dp
∫
P2
H = ∆hs / g = P1 (1)
ρg
The rise is directly related to the pump work, even if the fluid has significant
compressibility:
gH
Work/mass= ∆h = (2)
ηp
and this is one of the advantages of its use. Obviously, if ρ = const.,
∆P Work ∆P
H= , = (3)
ρg mass ρ ηp

The head rise is directly to the peripheral speed of the impeller disk. The fluid enters
axially near they impeller hub, with no angular momentum; it leaves the impeller
with absolute tangential speed ωR2 – Vr2 tan β2, where β2 is the back-leaning blade
angle at the rim Fig 1, and Vr2 is the fluid radial exit velocity, related to the volume
flow rate as
Q = 2πR2b2 Vr2 (4)
The torque needed to drive the impeller is the net outflow rate of angular
momentum, and the work rate is this, times ω. Thus,
2 ⎛ V ⎞
i
Power = m ( ωR 2 ) ⎜1 − r2 tan β2 ⎟ (5)
⎝ ωR 2 ⎠
i
gH
and since we also have Power = m , the head rise is
ηp
( ωR 2 )
2
⎛ Vr2 ⎞
H = ηp ⎜1 − tan β2 ⎟ (6)
g ⎝ ωR 2 ⎠

⎛ V ⎞
The quantity ψ = ηp ⎜1 − r2 tan β2 ⎟ is sometimes called the “head coefficient.” In
⎝ ω R 2 ⎠
terms of ψ ,
( ωR2 )
2
H=ψ (7)
g
ψ is typically between 0.2 and 0.8. Values greater than unity could be obtained if
the blades were designed to learn forward ( β2 < 0 ), but then ∆ P would increase with
Q (through the effect of Vr2). This positive slope of the ∆ P vs. Q characteristic is
known to produce instabilities in the pumping system [44]. These are generally
dynamic in nature, and depend to some extent on the characteristics of the rest of
the system (free volume, throttling effects, etc), but it is relatively easy to
understand their origin from a quasistatic argument: if the pump temporarily delivers
more flow than can be disposed of in steady state by the downstream components,
∂ ( ∆P )
and if its characteristic has a positive slope , the pump pressure rise will also
∂Q
be higher than normal. The downstream pressure will therefore tend to increase for
both reasons, and a runaway situation ensues. In addition to this system instability,
there is also a tendency for flow maldistribution analogous to rotating stall, since the
flow is then unstable with respect to mass interchange between parallel streamtubes
[45].
Eq. (6) would predict a linear dependence of head on flow rate. In reality varying the
flow at a given speed will vary the internal flow angles with respect to blades, and
will therefore result in variations of the slope ∂ H ∂ Q . Examples of this behavior are
shown in Fig 2 (Ref. 41), where the flow coefficient is defined as
Q
∅2 = (8)
R2
2π
b2 ( ωR 2 )

The throttling range is defined by the point of maximum head, below which operation
is unstable. Similar (but stronger) restrictions apply to axial designs, such as that
used in the J-2 LH pump, and so these designs tend to be limited to applications
which require very limited throttling.
The pump is designed to specified head rise H or ( ∆ P) and volumetric flow Q.

These quantities can be used to construct the non dimensional quantities called
specific diameter and specific speed:

1
D(gH) 4
ds = 1
(9)
Q 2
1
ωQ 2
ns = 3
(10)
( gH) 4
In the U.S. literature, D is expressed in feet, Q in gpm and ω in rpm, and g is

omitted. This procedure, of questionable practically, result in related parameters Ds,
Ns given By Ns= 2728 ns, Ds = 0.01985ds.
Notice that
ωD
ns ds = 1
(11)
( gH) 2
and hence, from (7),
2
nsds = (12)
ψ
or, in English units, NsDs = 108.3 ψ . Since for centrifugal pumps, we found ψ <1,
we can see that their domain is a ( ns , ds ) map is ns ds >2 (or NsDs > 108 ).
The considerable empirical evidence on pump performance has been distilled

(Ref.46) by constructing ( ns , ds ) maps on which favorable regions are shown for
various types of machine. A very generalized example (taken from Ref. 41) is shown
in Fig. 3, where lines of constant peak efficiency are shown for a wide variety of
pumps. These use a set of clearance, tolerance, roughness, etc. factors, and are to
be taken only as indicative, since actual design may depart from adopted values. We
notice in Fig. 3 the line Ns Ds ∼ 100 , denoted as the “limit for dynamic pumps”, in
accordance to our discussion above. Radial and axial pump designs nearly merge,
although the axial type is indicated for the highest specific speed, which as Eq. 10
shows, may be simply a reflection of low head rise, as for example, in the inducer
stages featured commonly at the inlet of centrifugal pumps for cavitation
suppression. Table 1 gives the features of the high pressure SSME pumps, and the
resulting ( Ns , Ds ) points are included in Fig 3, where they are seen to lie roughly on
the η = 0.8 line.

The pumps must be designed so as to avoid cavitation, which both, degrades
performance, and causes damage to reusable articles. Cavitation risk exists at the
pump inlet, where the liquid pressure is lowest, and it increases with both, the fluid
velocity and the speed of the pump inducer blades meeting the fluid. The inducer
1
diameter needs to be large enough to reduce the inlet fluid dynamic head ρ C m2 to
2
some fraction of the excess inlet pressure over saturation P1-Psat, 1. This last quantity
is called the Net Positive Suction Pressure (NPSP), and is usually given as a suction
head NPSH= NPS/(ρg) .
Hydrogen Pump Oxygen Pump
∆ P (N/m2) 1.38 × 107 (per stage) 3.30 × 107

H (m) 20,000 2,930
Q (m3/sec) 0.96 0.229 (per side)
ω (rad/sec) 3674 3246
D (m) 0.308 0.16
ns (Ns) 0.386 (1050) 0.703 (1920)
ds (Ds) 6.61 (0.0131) 435 (0.0866)
Table 1. Characteristics of the SSME high-pressure fuel and oxidizer pumps
The ratio:
σ=
( NPSP ) =
( NPSP ) (13)
1 2
2
Cm / 2g
ρ Cm
2
is called the Thoma parameter, and empirical evidence [40,41] indicates that it
should be grater than 1 for LH, 2 for LOX and 3 for water and storable propellants.
The more favorable situation for hydrogen appears to be related to a greater vapor
suppression effect due to evaporate chilling when bubbles start to form.
Several other parametric representations of cavitation data exist. Thus, Refs

[40] and [41] use a “suction specific speed” Ss defined as in Eq (10), but with H
replaced by the NPSH, and with correction for flow blockage by the hub. This
parameter can be shown to be related to Thoma’s parameter σ and to ∅t = cm / ωR1
(R1= inducer radius) by
2.981
Ss = 3
(14)
σ 4 ∅t
In English units, the numerical factor is 8132. The data on cavitations onset
for a variety of liquids show that σ remains approximately constant for each, as
noted above. Independently, Ref. 47 shows cavitations results in the form
τ = f (Zt ) (15)

where
NPSP sin ϑ
τ= ; Zt = ∅t (16)
1 1 + cos ϑ
ρ ( ωR1 )
2
and ϑ is the inducer leading edge blade angle, which, at design conditions, is close
to tan−1 ∅t and is typically 50 − 100 . The data lies close to the line τ = 3Zt . With small
angle approximations for ϑ and ∅t , this can be shown to be equivalent to σ = 3 2 ,
intermediate between Stangeland’s recommendations [40] for LH and LOX.
With the inducer diameter chosen from the above criteria, the shaft speed is
limited [40] so as to keep the inducer tip speed below 170 m/sec (LOX) or 340
m/sec (LH). This is done in order to control cavitation in the blade-tip leakage vortex
[40]. This speed limitation may conflict with the desire to place the (rs , ds ) point on a
favorable spot in the efficiency maps. In that case, the NPSH must be raised, either
by partial pressurization of the tanks, or, as in the SSME, by the use of separate low-
pressure booster pumps. These are only rough guidelines, and, the precise allowable
limits depend upon detailed design of the inducer. Progress in inducer design has
been a pacing item in allowing turbopump speed to increase, thus reducing weight
(as well as increasing life).
References cited:
40. M. L. Joe Stangeland. “Turbopumps for Liquid Rockets Engines”. Ninth Cliff
Garrett Turbomachinery’ Award Lecture, April 7, 1992. SAE/SP-92/924.
41. Turbopump Systems for Liquid Rocket Engines, NASA SP-8107, Aug. 1974.
42. Liquid Rocket Engine Turbines, NASA SP-8110, Jan 1974.
43. Liquid Rocket Engines Centrifugal Flow Turbopumps, NASA SP-8109, Dec.
1973.
44. E.M. Greitzer, “The Stability of Pumping Systems”, ASME, Transactions, Jl. of
Fluids Engineering 103 (1981), pp.193-242.
45. J.L.Kerrebrock, Aircraft Engines and Gas Turbines, MIT press. 1992 (Sec.5-
7).
46. O.E. Balje, Turbomachines: A Guide to Design, Selection and Theory, J.

Wiley & Sons, New York, 1981.
47. L.B. Stripling, “Cavitation in Turbopumps”, pt. 2, Trans. ASME, Series D, J.

Of Basic Engineering (1962): 339.

Lecture 27: Turbines
1. Rocket turbine design emphasizes power density, because of the overriding

concern for mass saving. Efficiency, while clearly a consideration, takes a less
prominent role than in aircraft turbine design. This tends to favor low reaction
turbines and when multi-staging, velocity compounding over pressure
compounding.
The degree of reaction of a turbine stage (stator nozzles plus rotor blades)
is the fraction of the fluid static enthalpy drop which occurs in the rotor (see
section 2. of this Lecture). In an impulse turbine, the degree of reaction is zero,
meaning that the gas expands and accelerates as it turns in the stator passages,
and then is merely redirected at constant thermodynamic state by the moving
rotor blades. The stage velocity triangles are shown in Fig 1, which also includes
the case of a 50% reaction turbine for comparison. In both cases, the flow leaves
axially. The torque (and hence the power) is proportional to the change in the
tangential component of the absolute velocity. Fig 1 shows that this change is the
wheel speed ω R for the 50% reaction turbine, but is 2 ω R for the impulse
turbine. It is also clear, however, that flow velocities are higher in the impulse
case which will lead to larger viscous losses. Also, the lack of acceleration in the
rotor passage will favor flow separation on the suction side. Altogether, the more
powerful impulse stage is also less efficient.

If two stages are used, one can over-turn the flow in the first rotor, so that it
leaves with a component of velocity contrary to rotation. The second stator then can
re-direct this flow without accelerating it, and send it into yet another pure impulse
rotor. It can be shown that this velocity compounding (see Section 3. of this lecture)
will yield four times as much torque as the single impulse stage such as those in Fig.
1b so that the flow is re-expanded in the second nozzle (pressure compounding).
This only doubles the power of the single stage, but improves the efficiency. The
SSME fuel turbopump turbine is essentially of this type, although there is some non
zero reaction (R ≅ 0.22).
For preliminary design purpose, Ref [46] (Refer to the list of the References in
Lecture 25) presents (ns , ds ) diagrams similar to those in use for pumps. One of
these is reproduced in Fig. 2. The quantities Q and gH in the definitions of ds and ns
(Eqs. 3.34-3.35) are now to be specified more fully: the volumetric flow rate Q is at

turbine exit static conditions, and the energy gH is the total-to-static isentropic
enthalpy drop:
⎡ γ −1
⎤
⎢1 − ⎛ Pte ⎞ ⎥
γ
gH = c pTT .ti ⎜ ⎟ ⎥ (1)
⎢
⎝ Pt .ti ⎠ ⎥
⎣⎢ ⎦
Notice that, as in the case of a pump, the product ns ds has a simple connotation:
ωR 2
ns ds = 8 = (2)
C0 ψ
where c0 = 2gH is sometimes called the “spouting velocity”. The velocity ratio
1
ωR c0 = is often used as a design parameter. For optimum efficiency, it should
2ψ
be about 0.4 in an impulse stage, 0.2 for a velocity-compound double stage, or 0.7
for a 50% reaction stage [40]. The optimum rotational speed of the turbine was in
the past higher than that for the driven pumps, which required heavy gearing. The
steady advancements in inducer design have more recently allowed mounting both
turbine and pumps on one shaft, with major mass savings.
Whereas pump impellers (except for hydrogen) are rarely stress limited, the
reverse is true for turbines, where the blade root axial stress and/or the disc rim
hoop stress tend to limit speed. The radial root stress is approximately (see Sec. 4 of
this Lecture)
A
σ6 = ρb ω2 (3)
2π
where A is the annulus flow area and ρb is the density of the blade material, which is
dictated by the flow rate and inlet gas density. There is thus an incentive to use
materials, such as directionally solidified superalloys, that retain a high strength at
the uncooled operating conditions.

Fig. 2. nsds diagram for turbines calculated for minimum loss coefficients.
2. Some More General Performance Relationships for Turbines
∆P
Ignore inefficiencies: ∆h =
P
Define Degree of Reaction
( ∆h)Rot. ( ∆P )Rot.
R = (low p.r. per stage, ρ const.)
( ∆h)stage ( ∆P )stage

c=absolute vel.
ω=relative vel.
Assume axial exhaust

(design cond.)
Torque
i
= cIN sin α1 r
m
Power ∆Pstage
= ( ωr ) cIN sin α1 =
i
ρ
m
= ( ωr ) c OUT tan α1
cOUT = cIN cos α1
cx =const. through stage
( ∆P )stator c2IN − c2OUT c2 ⎛ 1 ⎞ c2

= = OUT ⎜ 2
− 1 ⎟ = OUT tan2 α1
ρ 2 2 ⎝ cos α1 ⎠ 2
( ∆P )stator 1 cOUT tan α1

1−R = =
( ∆P )stage 2 ωr
( ∆P P ) stage cOUT tan α1

Also, ψ = =
( ωr )
2
ωr
ψ
So, 1 − R = ψ = 2 (1 − R ) More work/stage with less reaction
2
R=0 (impulse turbine) → ψ = 2

R= 1 (50% reaction) → ψ = 1
2
ψ
tan α1 =
c c φ
Also, φ = x = OUT , so, For, 50% design, α1 = β2
ωr ωr 1
tan β2 =
φ
ωr
and tan β2 = For impulse, tan α1 = 2 tan β2
cOUT

3. Velocity Compounding: Follows this diagram starting from the exhaust (at
bottom)
St. 1
Rotor 1 Power =
i i
m(4ωr + 2ωr)Rω = 6 m(ωR)2
St. 2
Rotor 2 Power =
i i
m(2ωR)Rω = 2 m(ωR)2
i
Total = 8 m(ωR)2 =
4 × Single impulse
1st stator must give

4ωr tangential velocity.

4. Blade root stress:
d 1
Pb ω2 r A dr = −
dr
(
σ A dr) σ = σ1 −
2
Pb ω2 (r22 − r12 )
1
σ (r2 ) = 0 σ1 = Pb ω2 (r22 − r12 )
2
A
A = π (r22 − r12 ) σ1 = ρb ω2
2π

Lecture 28: Mechanical Design of Turbomachinery
Integration and Rotordynamics of Turbo Pumps
1. Integration and Mechanical Components
As noted before, turbines and pumps are often mounted on a common

shaft. If the oxidizer and fuel have similar densities, their respective pumps can
also be on one shaft. This is the case in the MK-3 Atlas and Delta II booster
turbopump, which, however, has a geared turbine-pump transmission. The
Russian RD-170 takes a further integration step by having a single turbine drive
both, fuel and oxidizer pump, all on a single shaft. In addition, these pumps feed
not a single thrust chamber, but a cluster of four in the case of the Energia
vehicle (Fig 4, Lec. 24). Engines using LH fuel require different speeds for the
oxidizer and fuel pumps. The first of these engines, the RL-10, had a single-stage
oxygen pump on one shaft, gear driven by a second shaft on which were
mounted a 2-stage hydrogen pump and the drive turbine. More recent engines
(LE-7, SSME) feature separate shafts for the oxygen and fuel, each carrying its
own drive turbine.
Bearing design and bearing placement have a significant impact on the

overall turbopump characteristics. Existing engines use roller element bearings,
and in recent designs, these are lubricated and cooled by the propellant being
pumped, which simplifies the construction. On the other hand, this departure
from traditional bearing practice has necessitated extensive research on
compatible materials. Ref [48] describes work on advanced ball bearings for the
future Space Transportation Main Engines (STME) and the SSME Alternate
Turbopump Development (ATD). The lubrication concept relies on sacrificial wear
of the bearing cage (bronze-60% Teflon), and its transfer from the rolling
elements to the raceway surface.
Roller element bearings can provide some stiffness, through angular contact
design, but the bulk of the axial thrust of the pumps and turbines must be
hydraulically balanced, either by back-to-back pump arrangements with no
feedback, or, as in the SSME turbopumps, by providing hydraulic feedback to
some surface acting as the balancing piston [40]. Future designs are likely to
feature hydrostatic bearings, which rely on a very thin fluid film to support the
rotor without solid contact with the casing. The advantages of these bearings are
summarized in Table 1 from Ref [40]. The most important are the removal of the
surface speed limitation of ball bearings and the much higher radial stiffness. The
surface speed is expressed by the “DN product” in conventional bearings, and, as
Table 2 indicates, is in the range of 1 − 2 × 106 (mm) × (Rpm). This limitation
forces the designer to seek bearing locations with the smallest possible diameter,
such as outboard of the pumps and turbines, but these bearing locations tend to
lower the 1st natural frequency, and to interfere with flow approach to the pumps.

Item Hydrostatic bearing Ball bearing
Speed limit None 2.0M DN LH2

1.75M DN LO2
Life limit Unlimited steady state = 2h
Transient rub concern
Design constraints Supply pressure availability Shaft diameter for torque

transmission
Direct stiffness 1 to >5M lb/in. 0.5 to 1M lb/in. for duplex

pair
Damping lb s lb s
50 to >500 2 to 5
in. in.
Rotor-dynamics No constraints for optimum Position constraints

position
No adjustable damping
Adjustable stiffness and
damping
Table 1. Hydrostatic bearing benefits
As in all turbo machines, seals are required to reduce or prevent leakage of

fluids around the shaft from high to low pressure areas. The high linear speeds of the
rocket turbopump surfaces, as well as, in some cases, the oxidizing nature of the
fluid, dictates the use of non-contact type seals, which, by their nature, allow a
nonzero leakage rate. Thus, in oxidizer turbopump with the fuel-rich turbine on same
shaft, there is a need to introduce some high-pressure inert gas into a region
separating the two fluids, with seals provided to minimize leakage (and hence
inventory) of this purge gas. Multi-tooth labyrinth seals are standard in jet engines,
and were incorporated at various points in the original SSME turbo pumps. However,
after a sequence of redesigns to correct vibration problems most of these have been
replaced by very low clearance smooth cylindrical seals, which have much higher
radial stiffness and significantly contribute to raising the lowest natural frequencies
of the rotor (in fact, these seals can be viewed as a transition to hydrostatic bearing
designs).

Lecture 29: Rotordynamics Problems
1. Turbopump Rotor Dynamics
Because of high power density and low damping in rocket turbopumps, these
machines exhibit in their most extreme form a variety of vibration effects, which are
either absent or masked by normal damping mechanisms in other turbo machines.
The low damping is especially prominent in liquid hydrogen pumps, because of the
very low viscosity and density of this medium. Oil squeeze film dampers are
precluded in any cryogenic medium.
The general frame work of Rotor Dynamics is now well established, through a
combination of classical analysis and detailed numerical simulation [49, 50, 51].
Intensive efforts on the application of these theoretical methods to a specific rocket
turbopump are described by Ek[52], and were instrumental in pointing the way to a
series of improvements that resolved a serious development problem in the SSME.
The greatest difficulty in this field remains the precise characterization of the fluid
forces acting on the rotor at components such as seals, turbines, or impellers. Once
these are specified, numerical models of great power and versatility can be brought
to bear for analyzing the dynamics of a given configuration. Because of the
remaining uncertainties in the basic forces. Ek’s 1978 recommendation [52] remains
valid today: “Prediction of stability in a new design must be viewed with skepticism.
A prediction of instability should, however, be taken very seriously”.
2. Forced and Self- Excited Vibrations
Three are two types of rotor dynamics problems:
(a) Resonances which usually occur when the rotating speed coincides with one of
the natural (“critical”) frequencies of the rotor (including its supporting
structure). These fall in the category of Forced Vibrations, in which an excitation
force produces deflection responses of an amplitude which increases as the
excitation frequency approaches a critical frequency. If the excitation is at exact
resonance, the amplitude grows linearly in time at first, and then, if viscous
damping exists, it approaches a limit which is inversely proportional to the
damping factor. Removal of the excitation removes the response. The exciting
forces are typically related to rotor mass imbalance or geometrical imperfections.
Resonances rarely pose serious problems, unless the steady operating point lies
very close to one critical. On the other hand, since the structure is made as light
as practical, many natural modes usually exist, several of them either below or
not far above the operating range. Efforts are made in the design phase to create
a relatively wide range of resonance- free speeds around the normal operating
point. Passage through criticals, if made rapidly enough, is not a severe
condition.
Table 3 (Ref [53]) shows the critical frequencies of the SSME fuel turbopump.
Notice that several of the shaft modes are split into adjacent pairs of critical
frequencies because of the lack of symmetry of the casing structural supports,

even though the bearing structure itself is symmetric. This asymmetry is in
general a beneficial effect, and provides a sort of effective damping [54].
(b) Self- Excited Vibrations These are autonomous oscillations, in which the shaft
vibrates at one of its natural frequencies (not equal to the shaft speed), and due
to some positive feedback mechanism, absorbs energy from an external source
(usually the fluid) into the vibrational mode. Exact balancing does not remove
this type of vibration. Once initiated, if damping is insufficient, the vibration will
increase exponentially in amplitude until some nonlinear mechanism intervenes,
or until rubbing occurs. Self-excited vibrations are also called “rotor-dynamic
instabilities” or “subsynchronous vibrations”. The latter designation is due to the
fact that they are most often observed in the lowest shaft mode when the
rotating speed is well above the frequency of this mode. For some mechanisms of
excitation, the ratio of the rotation speed at onset of instability to the frequency
of the vibration excited is a simple integer,
COUPLED HPFTP ROTOR AND CASE MODES
Mode Frequency(Hz) Description

1 - Rotor free spin, X
2 47.0 Case rocking, Y
3 100.8 Case rocking, Z
4 196.4 Case rocking + bending, Y
5 197.5 Diffuser torsion
6 249.8 Case rocking + bending, Z
7 287.6 Rotor translation, Y
8 321.0a Rotor translation, Z
9 375.1 Rotor rocking, Z
a
10 424.0 Rotor rocking, Y
11 463.7 Rotor axial, X
12 488.7 Case + rotor rocking, Z
13 513.6 Rotor bending, Y
a
14 515.0 Rotor bending, Z
15 554.8 Turbine case torsion, X
16 574.7 Diffuser bending, Y
17 577.4 Diffuser bending, Z
18 599.4 Case + rotor rocking, Y
19 650.4 Case torsion, X
20 657.3 Case axial, X
21 694.9 Turbine case bending, Z

or an excitation threshold Table 3 exists at an integer multiple of the
excited frequency [55]. However, thus is not universally the case, and, in fact, no
such simple ratio or threshold seems to exist for the most important mechanisms
(seal or turbine blade-tip effects). It is true, however, that the rate of energy
input into the vibration increases with power level of the turbopump, and hence
with its speed; thus, the machine damping may be sufficient to compensate for
this effect at low rotation speeds, but, as speed increases, a threshold will
appear, beyond which the operation is unstable.
The two types of vibration described can be easily distinguished in tests by

plotting a series of vibration spectra at increasing rotational speed (a cascade
diagram, Fig. 1). Here
Fig. 1. A cascade diagram showing one mode only. Forced vibrations related
to imbalance are seen at Ω = ω , with a resonance when ω = Ωcr . Also shown
is a sub-synchronous, self-excited vibration at Ω = Ωcr .
ω is the shaft speed, Ω is the angular frequency of the vibration, and Ωcr is the
critical or natural frequency (only one shown). As indicated, no self-excited
vibration is visible until ω reaches some threshold, and then the instability
becomes more and more prominent. The amplitude itself depends on conditions
when the spectrum is taken (variation rate of ω dwell time at the given condition,
etc.), but the frequency information is still quite useful.

3. Sources of Rotor-Dynamic Excitation: Cross-Coupled Forces
The most important excitation mechanisms are related to the production of

cross-forces when the shaft is offset from its central location. Before describing some
specific examples, the generic dynamic effects of these cross-forces will be
discussed.
Qualitatively, if a cross-force Fy results from a shaft offset ex, this force exerts a
torque Fyex with respect to the nominal shaft center. It is well known that any linear
vibration in the x-y plane can be resolved into a forward and a backward circular
oscillations. Of these two, one (the forward component in Fig 2. is reinforced by the
resulting torque Fyex, which, being produced by the displacement itself, will remain
synchronized to it. This is the basic instability mechanism.
For a simple linearized analysis, suppose the fluid effects are such that a
⎛ • • ⎞
general transverse displacement (ex,ey) and displacement rate ⎜ ex , ey ⎟ of the shaft
⎝ ⎠
produces forces (Fx,Fy) according to
C xy ⎤ ⎪⎧ex ⎪⎫
•
⎧⎪Fx ⎪⎫ ⎡ −K xx K xy ⎤ ⎪⎧ex ⎪⎫ ⎡ −C xx
⎨ ⎬=⎢ ⎥⎨ ⎬+ ⎢ ⎥⎨ ⎬ (1)
⎩⎪Fy ⎭⎪ ⎣⎢ −K xy −K xx ⎦⎥ ⎩⎪ey ⎭⎪ ⎣⎢ −C xy −C xx ⎦⎥ ⎪ • ⎪
⎩ey ⎭

The coefficients K x x and K x y are the direct and cross-coupled stiffness,
respectively. Notice that the cylindrical symmetry has been exploited to reduce to
two the number of stiffness factors, and that a positive K x x is restoring, and will
augment the structural stiffness K0, although, in general, K xx << K0 . Similar
comments apply to the damping coefficients C xx , C xy , except that C xx may be of the
same order as any additional damping C0.
Define a complex displacement
ex + i ey = z (2)
⎧ 2
⎪ M d ex = − ( K0 + K xx ) ex + K xy ey
• •
− (C0 + C xx ) ex + C xy ey
⎪ dt 2
⎨
⎪ d 2 ey −( K + K )e
• •
then ⎪ M 2
= −K xy ex 0 xx y − C yy ex − (C0 + C xx ) ey (3 a, b)
⎩ dt
Add i x (3b) to (3a):
d2 z
M = − ( K0 + K xx ) ( ex + iey ) + K xy ( ey − iex ) − − − −
dt 2
z -i z
d2 z •
M 2
= −(K0 + K xx + iK xy )z − (C0 + C xx + i C xy ) z
dt
K C
C K
z+ z+ z =0 (4)
M M
∧
C K
Put now z = z e i Ωt (unstable if Ω I < 0) ⇒ + Ω2 − i Ω− =0 (5)
M M
2
C ⎛ C ⎞ K
Ω=i ± −⎜ ⎟ +
2M ⎝ 2M ⎠ M (6)
K xy K xy 1 C0 + C xx 1 C xy
Call K xx = K yy = ζ= η=
K0 K0 2 K0 M 2 K0 M

2
Ω Ω C0 + C xx + iC xy ⎛ C0 + C xx + iC xy ⎞ K + K xx + iK xy
= = −i ± ⎜ ⎟ + 0
Ω0 K0 2 K0 M ⎜ 2 K0 M ⎟ K0
⎝ ⎠
M
Ω
= = i(ζ + i η) ± (ζ + i η)2 + 1 + K xx + iK xy (7)
K0 M
Ω ⎛ K K xy ⎞
γ ζ , η 1, K xy 1 i ζ − η ± ⎜ 1 + xx + i ⎟
K0 M ⎝ 2 2 ⎠
K xx ⎛ 1 ⎞
1+ − η + i ⎜ ζ + k xy ⎟
+ 2 ⎝ 2 ⎠
Ω
= (8)
Ω0
-
K xx ⎛ 1 ⎞
−1 − − η + i ⎜ ζ − k xy ⎟
2 ⎝ 2 ⎠
1
For K xy > 0, if K xy > ζ , unstable (9)
2
K xy C0 + C xx
> K xy > Ω0 (C0 + C xx ) (For instability) (10)
K0 K0 M
Clearly, the fluid damping C xx , if positive, reinforces the other machine damping C0,
and promotes stability, whereas the cross-coupled stiffness K x y is equivalent to a
negative damping −K x y Ω0 . The cross-damping Cxy is seen to have relatively minor
effect on the dynamics, since, as Kxx, it only affects the natural frequency, and not
the growth or decay rate. In some instances, the fluid-related stiffness is not
negligible, and can be exploited to help relocate the critical frequencies away from
unfavorable ranges. This was, in fact, the approach taken in the SSME redesign [52].
The two principal avenues for improving rotordynamic stability-increasing damping
or raising the natural frequency are both exemplified on the right-hand side of Eq.
(10).

4. Some Examples of Cross-Coupled Force Generation
Among the mechanisms which have been identified as contributors to the

production of destabilizing cross-forces are (a) Pressure non-uniformity in labyrinth
and other seals; (b) Non-uniform driving force in turbines due to tip leakage effects
[59, 60]; (c) Non uniform pressure and driving force in pump impellers, also related
to leakage effects [55, 56]. Other effects are reviewed in Ref [57]. Rather than
attempting here a general coverage of this rather rich field, we will only explain in
some detail the labyrinth seal and blade–tip effects, which, aside from their
importance in practical cases, can be seen as prototypes of the relevant physics.
(a) Labyrinth Seals. The existence of flow swirl at the entrance to an offset labyrinth
seal (and, in modified form, to other seals as well) gives rise to a rotation of the
pressure pattern produced by the offset, and hence produces cross-coupled forces.
Two principal effects can be identified here [58]. One of these can be described as
follows: the fluid in the gland of the seal (a single-cavity labyrinth, for simplicity)
circulates azimuthally in the varying area created by the shaft offset (Fig. 3).
Fig. 3: Kinematic quantities associated with a labyrinth seal. The shaft is spinning at
an angular frequency ω , while simultaneously undergoing a circular precession of
amplitude e and frequency Ω .

If the tangential velocity V were constant (which is approximately true under
some conditions), the circulating flux ρ V l (h + δ) would need to be increasing with ϑ
at points where the depth (h + δ) (h= sealing strip height, δ (ϑ) = local gap) is
increasing. This implies an excess of leakage from upstream of the seal into the
cavity over that from the cavity to downstream of it, which is accomplished by a
locally depressed P (ϑ) . Thus, the pressure would be minimum at point C in Fig. 3.
and maximum at point A. The net integrated pressure force is then rotated 90 from
the shaft displacement in the direction of the swirl. Since the pressure pattern is
anchored to the whirling gap pattern, it is clear that, for a given inlet swirl, the effect
will depend on the whirling speed Ω and will, in fact, be zero when Ω R = V , because
the gland fluid will then not move tangentially with respect to the whirling gap
pattern. This argument is modified by the azimuthal variations induced on the
velocity V, which were so far neglected. A more complete analysis [58] gives the
force components parallel and perpendicular to the displacement as
π ⎛ lh ⎞ 2 eδ ⎛ lh ⎞ ∆P ρ
F = ⎜ (
⎟
4 ⎝ Rδ ⎠
l ∆P )
1+ f2
; f =⎜ ⎟
⎝ R δ ⎠ V − ΩR
(11)
π 2 eδ
F⊥ =
2
(
l ρ∆P V − ΩR
1+ f2
) (12)
where ∆ P = Pi − Pe , and we have assumed incompressible, inviscid fluid and a circular

whirling motion about the casing center with amplitude e. Aside from a negative K xx
(in the notation of (Eq. 1), we see from (12) that positive K xy and C xx are predicted.
The presence of higher order terms in Ω in Eq. 12 also indicated effective mass and
other effects, but these have little dynamic impact. The main factors are related in
this case through
R
C xx = K xy (13)
V
and the simple case in which velocity variations are neglected is recovered when
lh
1 in Eqs. (11) and (12)
Rδ
The second seal mechanism depends on the existence of friction between the
circulating fluid and the rotor and casing surfaces (although the resultant cross-force
is still a pressure, and not a frictional force). Because of friction, the mean tangential
velocity V is usually slightly less than the inlet tangential velocity V i and leakage
fluid entering the gland from upstream will continuously add tangential momentum

to the gland fluid as they mix, in a manner similar to the operation of an ejector
pump. This effect is strongest at point D (Fig. 3), where the gap is widest, and
∂P
weakest at point B. Thus, a positive pressure gradient will exist at D, and a
∂ϑ
negative one at B, again leading to a pressure maximum at A and to a cross-force
along the perpendicular axis ( ⊥ ) . Since this effect does not depend on fluid velocities
relative to the whirling frame, it does not show whirl velocity dependence, and does
not contribute to damping ( C xx ) . A simplified analysis, neglecting tangential velocity
variations, gives for this effect
( )
2
⎛ Rδ ⎞ ρ Vi − V
2 e
F = πR ⎜ ⎟ 2
(14)
⎝ hl ⎠ 1 + g h
F⊥ = πR2
(
ρ∆P Vi − V e ) ; g=
Rδ Vi − V
(15)
1 + g2 h hl ∆P ρ
Because of the importance of inlet swirl in promoting seal cross-forces, de-

swirling fins can be used ahead of the seal, if this is at all practical. Experiments [58]
have validated the above cross-force mechanisms, while also pointing out the
importance of other secondary effects, particularly for direct stiffness.
Both force components are greatly magnified in smooth seals with very small
clearance [52, 61]. Whether the added cross-coupling or the added stiffness is more
important when one of these is substituted for a labyrinth, must be directly assessed
through dynamic analysis for each specific case.
(b) Turbine Blade-Tip Effects. It was independently pointed by Alford [62] and
Thomas [63] that the sensitivity of blade-tip losses to blade-tip gap in turbines
should produce forward-whirling cross forces. The basic mechanism is simple: When
the turbine rotor is offset from its center, the blades on the side where the tip gap is
reduced will gain efficiency, and hence produce more tangential driving force than
average, and the opposite will happen on the side where gap opens up. Integrating
these forces around the periphery yields, in addition to the desired torque, a side
force in the forward-whirling direction (see Fig. 4). It is easy to translate this
−∂ η
argument into an explicit equation for the cross-force. Let β = be the
∂ (δ / H )
sensitivity of blade efficiency to relative tip clearance δ / H , where δ is the clearance
and H is the blade height. This factor is of the order of 1-5, depending on design and
operating point.

The tangential force f per unit length along the perimeter is then approximated as
⎛δ − δ⎞
f ≅ f − β⎜ (16)
⎜ H ⎟⎟
⎝ ⎠
although it must be pointed out that this equates work loss to efficiency loss, and
hence it ignores pressure ratio variations also induced by the offset. The clearance
perturbation δ − δ varies in proportion to the offset e, and sinusoidally with
azimuth ϑ , and so, upon projection in the direction perpendicular to the offset, and
integration, one obtains
e Q e
F⊥ = π R f β =β (17)
H 2R H
where Q = 2 π R f is the turbine torque. The direct force F is predicted to the zero.
The existence of these forces was confirmed experimentally in Ref. [64] and,
in more detail, in Refs [59,60] where it was found that, in addition to the above
mechanism, a second source of cross-forces is a tangentially rotated pressure
nonuniformity is acting on the turbine hub. The contribution of this component is
additive to the basic Alford/Thomas effect, and amounts typically to 40% of the total
cross-force. In addition, both mechanisms also give stiffening force components
(along with offset direction).

One important question which remains experimentally unanswered is whether
or not the Alford/Thomas forces show a significantly sensitivity to whirl speed, i.e.,
whether they provide C xx component for the stiffness matrix. The original argument
would predict no such sensitivity. A fairly obvious extension, accounting for change
Ω e cos ϑ to the basic blade speed ωR can be shown to yield a modified Alford factor
1 H Ω
β' = β − (18)
ψ R ω
•
where ψ = Qω m (ωR)2 is the turbine work coefficient, which is close to 2 for impulse
turbines. Since H R ∼ 0.2 and Ω ω (whirl to spin ratio) is ∼ 0.5 , we see that
β ' ≈ β − 0.05 , which would not be significant. Other velocity-dependent effects may
arise form time lags in the azimuthal redistribution of flow approaching a whirling
rotor; these would also be of order H R , but no direct experimental evidence exists
for their magnitude. Ref [60] does provide a theoretical model for this effect however
the Alford-Thomas forces can be very large, requiring damping log decrements of the
order of 0.1 in typical rocket turbopumps.

References Cited
48. F. Ferlitia et al. “Rotor Support for the STME Oxygen Turbopump”. Paper AIAA
92-3282. 28th Joint Propulsion Conference, July 1992.
49. J.P. Den Hartog, Mechanical Vibration, McGraw-Hill book Company, 1956.
50. Childs, Dara W. Turbomachine Rotor Dynamics: Phenomena, Modeling and

Analysis, Wiley Interscience, N.York, 1993.
51. J.M. Vance, Rotordynamics of Turbomachinery, Wiley, New York, 1988.
52. M.C. Ek “Solution of the Subsynchronous Whirl Problem in the High-Pressure

Hydrogen Turbomachinery of the Space Shuttle Main Engine”. J. of Spacecraft,
Vol. 17, No. 3.
53. G.R. Mueller, “Finite Elements Models of the Space Shuttle Main Engine” NASA
TM-78260, Jan. 1980.
54. F. Ehrich “The Role of Bearing Support Stiffness Anistropy In Suppression of

Rotordynamic Instabilities”. ASME 12th Biennial Conf. on Vibration and Noise.
Montreal, Canada, Sept. 1989.
55. Jery, B. Acosta, A. J., Brennen, C.E. and Caughey, T.K. “ Hydrodynamic
Impeller Stiffness, Damping and Inertia in the Rotordynamics of Centrifugal
Flow Pumps”. NASA CP-2338, 1984, pp.137-160.
56. Chamie, D.S., Acosta, A.J., Brennen, C.E. and Caughey, T.K. “Experimental
Measurements of Hydronamic Radial Forces Stiffness Matrices for a Centrifugal
Pump Impeller”, ASME JI. Of Fluids Engineering, Vol.107, No. 3, 1985, pp 307-
315.
57. F. Ehrich and D.W Childs, “Self-Excited Vibrations in High Performance

Turbomachinery”, Mech. Engineering, ay 1984.
58. Knox T. Millsaps, “The impact of Unsteady Swirling Flow in a Single-Gland

Labyrinth Seal on Rotordynmics Stability: Theory and Experiment”. Ph.D.
Thesis, MIT, 1992.
59. M. Martinez –Sandues, B. Jaroux, S.J. Song and S. Yoo “Measurement of

Turbines Blade-Tip Rotor dynamic Excitation Forces” J. of Turbomachinery, 117,
3, July 1995, pp. 384-393.
60. S.J. Srig and M. Matrinz-Sanchur, “Rotordyanamic Effects due to Turbine-Tip

Leakage: Part I, Black-Scale Effects. Part II, Radius Scale Effects and
Experimental Verification “. J. of Turbomachinery, 119, 4, Oct 1997, pp. 695-
714.
61. D.W. Childs, “Dynamic Analysis of Turbulent Annular Selas Based on Hir’s
Lubrication Equation” ASME J. of Lubrication Technology, V.105, No. 3, 1983,
pp. 437-445.

62. J.S. Alford, “Protesting Turbomachinery from Self-Excited Rotor Whirl”, J. of
Engineering for Power, Oct. 1965.
63. H.J. Thomas “Unstable Oscillations of turbine Rotors Due to Steam Leakage in
the Clearance Of the sealing Glands and the Buckets”. Bull. Scientifique A.J.M.,
Vol. 71, 1958, pp 1039-1063.
64. K. Urlicks, “Clearance Flow Generated Transverse Forces at the Rotors of

Thermal Turbomachines”. NASA TM -77292, Oct.1983. (Translated from
Doctoral Dissertation at the Technical Univ. of Munich, 1975).

Lecture 30: Dynamics of Turbopump Systems: The Shuttle Engine
Dynamics of the Space Shuttle Main Engine Oxidizer Pressurization Subsystems
Selected Sub-Model
In the complete SSME engine, all variables affect each other in complex ways.
In order to test our fault detection algorithms, a dynamic subsystem is desired, with
reduced order, but with the unmodelled states interacting as weakly as possible with
those modelled. Attention was focused on the liquid oxygen subsystem for two main
reasons:
(a) The O/F ratio at rated power is 6, so that the LOX dynamics should dominate
over the LH effects wherever they interact, and
(b) The turbopump pre-burners are run very fuel-rich in order to limit the turbine
inlet temperatures below the metallurgical limits of the uncooled blades;
small excursions of the LOX flow to the pre-burners are then immediately
translated into large and potentially critical turbine temperature excursions.
In our submodel we therefore focus attention on the LOX turbopump, which

feeds both, the main LOX injectors, and (after the boost stage) the two turbine pre-
burners. Also modeled are the dynamics of the LOX feeding line to the pre-burners,
as well as to the main LOX valve and main injector, plus the LOX pre-burner itself
and the main chamber pressure. Variations of the LH-related states should indeed
couple weakly to the LOX system, mainly through LH flow variations onto the pre-
burners (insensitive, since they are fuel-rich), and into the main chamber (choked
flow, no feedback).
The Dynamic Equations
There are three types of dynamic equations to be considered:
(1) Rotational dynamics of the LOX turbopumps.
(2) Equations expressing the liquid inertia under pressure difference variations
(analogous to inductance in electric circuits).
(3) Equations expressing the ability of cavities to store fluid due to its
compressibility under pressure fluctuations (analogous to capacitive effects).
(1) Rotational Dynamics
If IOTP is the moment of the inertia of the Oxidizer Turbo Pump (OTP) rotor, Ω 02 its
angular velocity, τ OT
2
the torque delivered by the OTP turbine, τ OP
2
the torque
absorbed by the main oxidizer pump stage, and τ OP
3
the torque absorbed by the
oxidizer booster pump, then

d Ω02
IOTP = τ OT2 − τ OP2 − τ OP3 (1)
dt
In a hybrid system where Ω02 is in rad/sec, t in seconds, and the torque in lb-
1
in, the constant IOTP has a value (which implies IOTP = 422 lbmin2).
0.916

(2) Inertia of LOX in pre-burner common supply line
This is a prototypical equation of type (2), and, in order to illustrate the

underlying physics, we well give here a brief derivation of it.
Consider a pipe of length L and cross-section A, fed by the booster pump

discharge at a pressure POD3, and having a mean pressure PPOS (Pre-burner Oxidizer
Supply). Frictional forces along the pipe, and at bends and restrictions, contribute a
total pressure drop λ ⎛ 1 ρ υ 2 ⎞ , where ρ and υ are the LOX density and velocity, and
⎜2 ⎟
⎝ ⎠
λ (of order unity) is a pressure loss coefficient. The liquid is then acted on in the
forward direction by a net force ⎛⎜ POD3 − λ 1 ρυ 2 − PPOS ⎞⎟ A. . The mass of liquid in the
⎝ 2 ⎠
pipe is ρ AL, and we must have
dυ ⎛ λ ⎞
( ρ AL ) = ⎜ POD3 − PPOS − ρ υ 2 ⎟ A
dt ⎝ 2 ⎠
i
Now, the flow rate in the pipe is mOP 3 = ρ υ A , so the equation can be re-written as
•
d mOP 3 ⎛ λ i 2 ⎞
L = ⎜⎜ POP 3 − PPOS − m OP 3 ⎟ A
⎟
2ρ A
2
dt ⎝ ⎠
•
or ⎛ L ⎞ d m OP 3 • 2
⎜ A ⎟ dt = POP 3 − PPOS − K m OP 3
(2)
⎝ ⎠
where
λ (3)
K =
2ρ A
2
•
In units of lbm / sec for m and lb / in2 for P, the constants have the values L
=
1 ,K=
A 100
2
0.000813. This implies L
= 3.86 in−1 and λ = 0.0261 ⎡⎣ A in ( )⎤⎦ .
A
(3)Fluid Capacitance in preburner LOX supply line
This is a prototypical equation of type (3), and we also provide a derivation

below:
•
Considering again the POS supply line, it receives LOX flow at a rate m OP 3
• •
from the booster pump, and discharges m FPO
into the fuel preburner (FP) and m OPO
into
•
the oxidizer pre-burner (OP), plus a small amount m OP 2 C which is diverted to cool the
pump. Under dynamic conditions, there is a (generally non-zero) net inflow

• • • • •
m OP 3 − m FPO − m OPO − m OP 2 C ≡ ∑ m. Let ρV be the mass stored in the pipe, where
π D2
V = L is the volume available. Then we must have
4
d (ρ V ) •
(4)
dt
= ∑m
Even though LOX is a liquid, it has finite compressibility at the very high
pressures involved here. This is measured by the thermodynamic parameter
1 dρ −10 −1
K = ≈ 5 × 10 Pa ≈ 3.4 × 10−6 in2 / lb
ρ dP
In general, the volume V also varies slightly under pressure fluctuations, but it
can be shown that this effect is secondary. We therefore rewrite (4) as
⎡ ⎛ 1 d ρ ⎞ ⎤ dPPOS • • • •
⎢( ρV )POS ⎜ ⎟ ⎥ = m OP 3 − m FPO − m OPO − m OP 2C (5)

⎣ ⎝ ρ dP ⎠LOX ⎦ dt
In the same units as before, the factor ⎡( ρV ) ⎛ 1 d ρ ⎞ ⎤ has a value of

⎢ POS ⎜ ⎟ ⎥
⎣ ⎝ ρ dP ⎠ LOX ⎦
1 1 dρ
. Using ≅ 3.4 × 10−6 in2 / lb, this implies a line volume V = 409 in , which
3
38,120 ρ dp
−1
combined with previously estimated L / A = 3.86 in , yields L ≅ 40 in, A ≅ 10 in .
2
These are not expected to be exact dimensions of the POS line, because the model
lumps together several secondary inertios and capacitances, but they do appear
reasonable. Incidentally, from the previous result λ ≅ 0.0261A2 in , we now estimate
( )
λ ≅ 2.8, again a reasonable value for a pressure loss factor.
The Complete Submodel
In addition to the three equations derived above, there are three others of the
fluid inertia type and three others of the fluid capacitance type. The complete
submodel, in the same unit used so far, is shown in Table….on next page.
Equations (6), (7), (8) are the ones just derived. Eq. (9) deals with the inertia
of the LOX moving through the valve and the injectors of the Fuel Preburner (FP)
under the fluctuating drive of the pressure difference PPOS − PFP , less the pressure
• 2
drops in the valve and in the injectors. These drops have the characteristic m form,
just as in Eq. (2), but, in addition the valve open area fraction A / AFPV appear squared
in the denominator, as it should according to Eq. (3). This area fraction will act as one
of our control variables.
Eq. (10) is identical in structure to Eq. (9), but refers to the fluid inertia in the
LOX dome of the Oxidizer Preburner (OP). Once again, the OP valve area fraction
A / AOPV appears here as a control variable.

The remaining inertia-type equation is Eq. (13), which refers to the LOX
moving through the Main Oxidizer Valve (MOV) under the drive of the difference
between the main oxidizer discharge pressure, POD 2 , and the main combustor
pressure, Pc, less the sum of the pressure drops in the MOV (assumed 100% open)
and the injectors.
The remaining three capacitance-type equations are Eqs. (11), (12) and (14).
Eq. (11) describes accumulation of gas in the Oxidizer Preburner (OP), with mass flow
• •
m OPF (the un-modelled fuel flow into the OP) plus mOPO (the LOX flow into the OP)
•
entering, and almost all of the gas flow input to the oxidizer turbine, m OT 2 , leaving.
Table: DYNAMICS OF LOX PRESSURIZATION SUBSYSTEM
Equation Equation Description Time

No. Constant
(sec.)
(6) 1 d ΩO2 Rotational 0.058

= τ OT 2 − τ OP 2 − τ OP 3 dynamics of
0.916 dt
OTP
(7) 1
din
OP 3 • 2 LOX inertia in 0.00013
= POD3 − PPOS − 0.000813 m OP 3 preburner
100 dt
supply line
(8) 1
dP
POS • • • • Mass storage 0.0020
= m OP 3 − m FPO − m OPO − m OP 2 C in preburner
38120 dt supply line
(9) 2 LOX inertia in 0.0048
dm
i
⎛ • ⎞ • 2
1 FPO m FPO injector dome
= PPOS − PFP − 0.02488 ⎜ ⎟ − 0.1948 m OPO
of FP
2 dt ⎜ A / AFPV ⎟
⎝ ⎠
(10) 2 LOX inertia in 0.0034
din ⎛ • ⎞ • 2
OPO m OPO injector dome
= PPOS − POP − 0.260 ⎜ ⎟ − 1.463 m OPO
of OP
dt ⎜ A / AOPV ⎟
⎝ ⎠
(11) 1
dP
OP • • • Mass storage 0.0143
= m OPF + m OPO − 0.9980 m OT 2 in Oxidizer
10, 000 dt Preburner
(12) 1
dP
F1 • • • Mass storage 0.0050
= m FT 1 + m OT 2 + MFT 2 − 1.085 m OT 2 in fuel ducts to
3000 dt injector
(13) dm
•
• 2
LOX inertia in 0.0078
1 MOV main injector
= POD2 − PC − 0.001715 m MOV
25 dt dome
(14) 1 dPc • • • Mass storage 0.00074
= m F 1 + m MOV − m CN in main
4000 dt
combustor

Equation (12) describes the gas accumulation in the two large ducts which
bring the partially oxidized hydrogen to the main chamber injector dome. Feeding this
volume are the (unmodelled) discharge flows on the main Fuel Turbine ⎛⎜ m FT2 ⎞⎟ and of
•
⎝ ⎠
•
the low pressure Fuel Turbine mFT1 , plus the discharge of the main Oxidizer Turbine,
⎛• ⎞ ; leaving this volume is basically the main Fuel Injector flow ⎛ • ⎞ , plus some
m
⎜ OT2 ⎟ ⎜ m FI ⎟
⎝ ⎠ ⎝ ⎠
•
smaller LH cooling flows, up to 1.085 m FI .
Finally, Eq. (14) governs the changes in the main combustor pressure, Pc ,
due to mass accumulation. The mass inputs are the fuel and oxidizer injector flows
• • •
(m ,mFI MOV
), while the mass loss is the nozzle flow rate mCN .
Characteristic Times
For each of the dynamic equations (6)-(14), we can estimate the

characteristic time constant, which provides same preliminary appreciation for the
dynamics of the system. For this purpose, we balance the rate term with one of the
dominant terms on the right; for instance, for Eq. (6), the time constant is
ΩO2
τ = , with ΩO2 and τ OT 2 evaluated at their nominal values (rated power). These
0.912τ OT 2
time constraints are included in Table…. The preburner supply line flow and the main
combustor pressure are seen to adjust rapidly (under 1 msec). Filling and emptying of
the Oxidizer Preburner is relatively slow (14 msec), and the shaft speed of the OTP is
very slow (58 msec). All other dynamics are comparable in speed with time constants
of a few msec.
Calculation of non-state variables
The sequence of algebraic computations (no additional dynamics) required to

calculate the right-hand-sides of Eqs. (6)-(14) is summarized in Appendix…. The data
for these calculations are the values of the nine state variables, the values of the
control variables (preburner valve openings), and a few unmodelled variables arising
from the fuel side of the overall system. The latter are generally kept at their nominal
values.

Appendix B. Steady State Values from the SSME Thermodynamic Model

Appendix C. State Variables in a simulated Throttling Sequence

Appendix D. Characteristic Times (approx.) for the SSME Dynamic Model

Lecture 32: Orbital Mechanics: Review, Staging
Mission Planning, Staging
The remaining lectures are devoted to Mission Planning and Vehicle Design, which in
reality occurs even before the rocket engines are fully specified (although iterations
continuously proceed throughout the process, and engine characteristics do affect
the mission plan).
Very roughly, the iteration steps in planning a launch mission are:
(a) Estimate the required ∆VTOTAL using impulsive thrusting formulae, plus add-
ons for gravity losses, drag losses, turning losses, etc.
(b) Distribute this ∆VTOT optimally among vehicle stages (since all orbit launches
so far require multiple stages in order to avoid carrying empty tankage in the
later stages).
(c) Using the mass fractions from (b), perform more detailed flight simulations
and refine the partial and total ∆V for the mission.
During stage (b), the total ∆V is assumed to be unchanged when the mass
distribution for the stages is varied. This is not strictly true, because often the
mission optimization leads to changes in the altitude and velocity at which the
various firings are executed and, as we will see, this may alter the various ∆V ’s.
This is the role of stage (c) above.
Another point to be made is that “stages” and “firings” may not map one-to-
one. A given stage may be turned off, allowed to coast, and then re-ignited. Or the
firing of two consecutive stages may occur with no interruption (or minimal
interruption), so that both can be idealized as occurring in the same place. As long
as the ∆V ’s are still regarded as insensitive to mission profile details (as per the
comment above), these distinctions do not impact the stage mass calculations, but
they can be of great practical importance nonetheless.
Impulsive Thrusting-Gravity Losses. Because of the large accelerations imparted by

rocket engines, their firings are usually short, from under one minute to about 10
minutes. In fact, there is a performance incentive in minimizing the firing time, as
long as the accelerations remain below structural or other limits. This can be most
easily seen in the context of a vertical ascent against gravity. The vehicle’s equation
of motion is then (ignoring drag)
dv
m = F − mg (1)
dt
•
dm
and F = m c = −c (2)
dt

dv d ln m
= −c −g (3)
dt dt
and integrating,
m0
∆V = V − V0 = c ln − gt (4)
m
m0
The “ideal”, or gravity-free velocity increment is the familiar ∆Videal = c ln (5)
m
But the presence of gravity reduces the velocity increment by ∆VGravity = gt (6)
which can be made insignificant if t is short, but can be very important otherwise. In
the limit when the thrust is barely enough to cancel weight, the vehicle just hovers
indefinitely with no velocity gain.
In practice, the significant item is the fuel used in the firing, which is
contained in the mass ratio m0/m.
The common procedure is then to first ignore gravity, as if the firing was
impulsive (t=0), and calculate the ∆V required for the mission under this
assumption. In our simple ascent example, the “mission” is to reach a velocity V,
starting at V0, and so the impulsive ∆V is simply V-V0. From (4) then
mo
c ln = ∆Vimp. + gt (6)
m
and so the extra ∆VGrav. = gt is added on as a correction, with the implication of

additional fuel being used for a given V-V0.
In a more general ascent trajectory (but still over a “flat Earth”, since gravity
losses occur only near the beginning of flight, before the path becomes nearly
horizontal) we would have
dv F
= − g sin γ (7)
dt m
dγ
V = −g cos γ (8)
dt

Here we assumed thrust to be aligned with velocity. This is called a gravity
turn, and is not the most general maneuver. It is, however, the most economical
strategy for turning, since any lateral component of thrust uses propellant without
adding flight energy.
Formal integration of (7) now gives
m0 t
m ∫0
∆V = c ln − g sin γ dt (9)
and so the gravity loss is
∆VGrav. = ∫ g sin γ dt
0
(10)
Of course, the particular γ ( t ) to be used here must come from simultaneously

solving (8) with (7). This solution cannot be done in simple analytical terms when
thrust is constant, since a nonlinear 2nd order differential equation is involved. But,
F
interestingly, there is a relatively simple solution when he thrust acceleration a =
m
is assumed constant (i.e., throttling down as mass is consumed). Although this is not
a very realistic option, it still is useful in giving information about the initial rotation
of the trajectory near the ground, which happens before the mass has time to
change much.
Eliminate time by dividing Eqs. (8) and (7) by each other, which separates the
variables V and γ :
dV a − g sin γ
=− dγ (11)
V g cos γ
a
We introduce = n and also change angle variable to
g
⎛π γ⎞ 1 − Γ2 2Γ
Γ = tan ⎜ − ⎟ ; sin γ = ; cos γ =
⎝4 2⎠ 1 + Γ2 1 + Γ2
2dΓ
dγ = − (12)
1 + Γ2
The variable Γ varies between 0 when γ = 900 (initial configuration) to 1 when

γ = 00 (orbit insertion). Substituting in (11) and simplifying,
dV dΓ 2Γ dΓ
= (n − 1) +
V Γ 1 + Γ2

which can be integrated to
(
V = C Γn−1 1 + Γ2 ) (13)
Here C is a constant of integration. The solution (13) satisfies V = 0 when Γ

= 0 (vertical start) for all C (n>1), so C must be calculated by imposing a particular
trajectory angle γ (or Γ ) at some specified velocity V (or, from later results, at
some time or altitude).
The time t is calculated from Eq. (8):
V dγ C
dt = − = Γn−2 1 + Γ2 dΓ
g cos γ g
( )
or, imposing t = 0 at Γ = 0.
C ⎛ Γn−1 Γn+1 ⎞ (14)

t= ⎜ + ⎟
g ⎝n − 1 n + 1⎠
dz
Similarly, the altitude z follows from = V sin γ :
dt
C2 2n−3 2n+1
dz = V sin γ dt =
g
Γ ( Γ dΓ )
C2 ⎛ Γ2n−2 Γ2n+2 ⎞
or, with z = 0 at Γ = 0 z = ⎜ − ⎟ (15)
g ⎝ 2n − 2 2n + 2 ⎠
We can use this model to calculate gravity losses. Starting from (10), and using the
relationships (12),
( )
Γ
1 − Γ2 C
∆VG = g ∫
0 1+ Γ 2
g
Γn−2 1 + Γ2 dΓ
C ⎛ Γn−1 Γn+1 ⎞
or ∆VG = ⎜ − ⎟ (16)
g ⎝n − 1 n + 1⎠
We could now use (14) to calculate the constant C by specifying the time to turn to a
given angle ( Γ ). Alternatively, we can eliminate C by division of (16) and (14):

n−1 2
1− Γ
∆VG = gt n +1 F (17)
n −1 2
1+ Γ
n+1 F
π
where ΓF = Γ ( γF ) , and γF is the angle reached at t, starting from γ = at t = 0.
2
As an example, say n = 3, γF = 200 ( ΓF = 0.7002). We find from (17)
∆VG
= 5.94 m/s2
t
and if t = 60 sec., ∆VG = 357 m / s , which is a substantial loss.
An alternative procedure would be to set the velocity VF reached when γ = γF .

Eliminating C now between (13) and (16) gives
1 Γ2
−
∆VG = VF n − 1 n2 + 1 (18)
1+Γ
Say n = 3, VF = 1,500 m/s, γF = 200 . We calculate ∆VG = 380 m / s in this case.
Maximum Dynamic Head (“Max-q”) During Ascent
1
Aerodynamic forces are proportional to q = ρ V2 . Initially, V 0 and ρ is high.
2
Later, V increases, but ρ decrease. There is a point of “max-q” in between, which is
important for design.
Assume Vertical flight . Neglect drag:
dv dv F ⎛ F ⎞
m = F − mg = − g = (n − 1) g ⎜n ≡ ⎟
dt dt m ⎝ mg ⎠
dz dv
=v or v = (n − 1 ) g
dt dz
Assume n = const. (F ∼ m)
v2
= (n − 1) gz
2

Atmospheric “Lapse Rate” “Adiabatic Lapse Rate”
g γ −1 g
Also, T = T0 − Γz Γ < Γa ≡ = ∼ 10 K / km
cp γ Rg
p dp g dz
and dp = −ρ g dz = − g dz =−
RgT p R g ( T0 − Γz )
g
dp g d ( T0 − Γ z ) p ⎛ Γ z ⎞ Γ Rg
=+ = ⎜1 − ⎟
p Γ R g T0 − Γ z p0 ⎝ T0 ⎠
g
−1
ρ ⎛ Γz ⎞ Γ R g
= ⎜1 − ⎟
ρ0 ⎝ T0 ⎠
g
−1
ρv 2 ⎛ Γ z ⎞ Γ Rg
q= = ρ0 ⎜1 − ⎟ (n − 1) gz (19)
2 ⎝ T0 ⎠
⎛ Γ ⎞
⎜− ⎟
dln q ⎛ g ⎞ ⎝ T0 ⎠ 1
For qMAX =0 ⎜⎜ − 1⎟ + =0
⎟
dz ⎝ Γ Rg ⎠1− Γz z
T0
Some altitude, regardless of

acceleration or lapse rate.
g Γ 1 Γ R g T0
− + + − =0 zqMAX = (20)
R g T0 T0 z T0 g
Air: R g = 287 J / Kg / K , T0 290 K , g = 9.8 m/s2
zqMAX = 8, 490 m
g
−1
⎛ Γ R g T0 ⎞ Γ Rg R g T0
Then qMAX = ρ0 ⎜ 1 −
⎜
⎟
⎟
(n − 1) g
T0 g g
⎝ ⎠
g
−1
⎛ ΓR g ⎞ Γ R g
qMAX = ⎜1 − ⎟ (n − 1) P0 (21)
⎝ g ⎠
(proportional to acceleration)

Γ Rg 0.006 x 287
Say Γ = 6 K / km = = 0.176
g 9.8
and n = 3
1
−1
qMAX = (1 − 0.176 ) 0.176
(3 − 1) P0 = 0.808 atm
1 atm ( )
= 0.808 x 14.7 x 122 = 1710 psf
R g T0 2
Also, then v2Max q = 2 (n − 1) g '
MMax 2
= (n − 1) (based on C0, at ground)
g q
γ
2 T 2 (n − 1) 1
2
Based on local T, MMax q = (n − 1) 0 =
γ T γ Γ R g T0
1−
T0 g
' 2
2 (n − 1) 2x2
MMax = M2 = MMax q = 1.862
q
⎛ ΓRg ⎞ 1.4 (1 − 0.176 )
γ ⎜1 − ⎟
⎝ g ⎠
Drag Losses: Like gravity losses, drag losses are important only near the ground,
peaking somewhat above z ( q MAX ) . Therefore, they should be estimated and added
to the 1st stage ∆V budget alone. The “drag loss” is defined by analogy to ∆VG as the
decrease in velocity due to the accumulated drag deceleration:
"∞"
D
∆VD = ∫
0
m
dt (22)
Drag is D = q CD A, where A is the frontal area, and CD varies with vehicle

shape and Mach number (from about 0.02 at low M to a peak of perhaps 0.15 in
transonic flow, then decreasing again). For estimation purposes only, we will use a
mean CD = CD and write (22) as
A CD m0 dz
∆VD =
M0 ∫q m v
(23)

Our estimate will be based on quantities evaluated at qMAX , and an effective
∆z ∼ 3 z ( qMAX ) :
A CD ⎛m ⎞ 3z ( qMAX )
∆VD qMAX ⎜ 0 ⎟ (24)
m0 ⎝ m ⎠qMAX v ( qMAX )
A CD
The “ballistic coefficient” can be related to the vehicle length L and its mean
M0
2
density ρ . Assuming an given shape with (Volume) = AL, we find
3
A CD 3 CD
= (25)
M0 2 ρL
v
m0 −
The mass ratio = e c can be estimated using v = 2 (n − 1) g and so
m
2(n−1)R g T0
⎛ m0 ⎞ −
⎜ ⎟ =e c
(26)
⎝ m ⎠qMAX
Using as well the values found previously for qMAX and z ( q MAX ) , and
simplifying, our approximate expression is
g
−1 2(n−1) R g T0
n −1 P ⎛ Γ R g ⎞ Γ Rg −
∆VD 4.5 CD R g T0 0 ⎜ 1 − ⎟ e c
(27)
2 ρ gL ⎝ g ⎠
For an example, take CD = 0.1 , n = 3, T0 = 290 K, ρ = 500 Kg/m3 (half the

water density), Γ = 6 K/km = 0.006 K/m, and c = 3,000 m/s. We calculate
1060
∆VD (m / s ) (28)
L (m)
For a large vehicle (say, L = 30 m) this is small ( ∆VD = 35 m / s ). But for a 3

m. vehicle this amounts to ∆VD = 353 m / s , a substantial loss. The difference can be
traced to the larger Area/Volume of the smaller vehicle.
To conclude, note the dependence ∆VD ∼ n − 1 , which shows that fast-

accelerating vehicles, like interception missiles, suffer more drag losses than slowly
accelerating ones. There is here a tradeoff with gravity losses, which vary in the
opposite manner.

Optimum Staging
Msi = εi Mi Mi+1 = MLi = λi Mi
∆Vi
−
Li
MLi + MSi = Mfi = e Mi
Mi+1
Mi+1 ∆V
− i ML Mn−1 M M
= e Ci − εi .... 2 = L
Mi Mn−1 Mn−2 M1 M1
n ⎛ − i ⎞
∆V
ML
= π ⎜ e Ci − εi ⎟
M1 i =1 ⎜ ⎟
⎝ ⎠
Maximize subject to ∑ ∆Vi = ∆V (assume εi is independent of Mi . In reality it may

i
depend on absolute mass.)
ML ⎡ ⎛ − ∆Vi ⎞ ⎤
⎛ ⎞
φ = ln
M0
− α ⎜ ∑ Vi ⎟ =
⎝ i ⎠
∑i ⎢⎢ln ⎜⎜ e Ci − εi ⎟⎟ − α ∆Vi ⎥⎥
⎣ ⎝ ⎠ ⎦
∆Vi
1 − ci
− e ⎛ ∆Vi
⎞
∂φ ci 1
For each i, = −α = −ci ⎜1 − εi e ci ⎟
∂∆Vi ∆V
− i α ⎜ ⎟
e ci
− εi ⎝ ⎠
⎛ 1 ⎞
1
∆Vi ⎜1 + α c ⎟
+ 1 = εie ci ∆Vi = Ci ln ⎜ i ⎟
αci ⎜ εi ⎟
⎜ ⎟
⎝ ⎠
⎛ 1 ⎞
n ⎜ 1 + αc ⎟
Then, find α from ∑
i=1
ci ln ⎜
⎜ εi
i ⎟ = ∆V , then find ∆Vi from
⎟
⎜ ⎟
⎝ ⎠

Assuming ci = c (same all stages), then
⎛ 1 ⎞ ⎡⎛ 1 ⎞ ⎤
n
⎜1 + α c ⎟ ⎢ ⎜1 + ⎟ ⎥
∆V n
αc⎠ ⎥
= ∑ ln ⎜ ⎟ = ln ⎢ ⎝
c i=1 ⎜ εi ⎟ ⎢ πεi ⎥
⎜ ⎟ ⎢ ⎥
⎝ ⎠ ⎢⎣ ⎥⎦
1
n 1 − 1
( ) ( )
∆V ∆V
⎛ 1 ⎞ n c + n
⎜ 1 + αc ⎟ = πi εi e nc
α= e nc < ε >G = π εi
⎝ ⎠ 1− < εi >G i
⎡ ⎛ 1 ⎞ ⎤ ⎡ ∆V ⎤ ∆Vi ∆V ε
∆Vi = c ⎢ln ⎜1 + ⎟ − ln εi ⎥ = c ⎢ln < ε > + − ln εi ⎥ = − ln i
⎣ ⎝ αc ⎠ ⎦ ⎣ nc ⎦ c nc <ε>
n
n ⎡ −⎜
⎛ ∆V ε ⎞
⎤ n ⎛ ε ⎛ − ∆ncV ⎞
⎛ ML ⎞ −ln i ⎟ ∆V
⎞ ⎛ n
⎞ ⎜ e
− 1 ⎟⎟
−
⎝ nc <ε> ⎠
So, ⎜ ⎟ = π ⎢e − εi ⎥ = π ⎜ i
e nc
− εi ⎟ = ⎜ π ε i ⎟ ⎜
M
⎝ I ⎠OPT i=1
⎢ ⎥ i=1 < ε >
⎝ ⎠ ⎝ i=1
⎠ ⎜ < ε > ⎟
⎣ ⎦ ⎝ ⎠
So, less ∆Vi when stage is less structurally efficient.
n
⎛ ML ⎞ ⎛ − ∆V ⎞
⎜ ⎟ = ⎜ e nc − < ε > ⎟
⎝ MI ⎠OPT ⎝ ⎠
Note:
⎛ M ⎞
∂ ⎜ ln L ⎟ Sensitivity of payload ratio to overall
M0 ⎠
Meaning of α = ⎝ <0 ∆V changes (after re-optimizing)
∂∆V
Generally: Max f(xi ) given gj (xi ) = Gj φ = f − ∑ λ j gj

i = 1 to n j = 1 to m < n j
∂ Gj ∂ gi ∂f ∂gj
dGj = ∑ i ∂xi
dxi = ∑ ∂x
i
dxi and
∂xi
= ∑ λj
j ∂xi
i
∂f ⎛ ∂ gi ⎞ ∂ gi
∂f = ∑ ∂x dxi = ∑ ⎜∑ λ j
∂xi ⎠
⎟ dxi = ∑ λ ∑ ∂xj dxi = ∑ λ dG j j
i i i ⎝ j j i i j
⎛ ∂f ⎞
So, λ j = ⎜
⎜ ∂ G ⎟⎟
⎝ j ⎠at optimum

Review of Orbital Dynamics (Single center)
p
r= ( θ from perigee)
1 + e cos θ
“true anomaly”
2
c ⎛b⎞
e = = 1−⎜ ⎟ ;
a ⎝ a⎠
p ⎛ 1 1 ⎞
Apoapse (apogee, aphelion): θ = π → ra = ra + rp = p ⎜ + = 2a
1−e ⎝ 1 − e 1 + e ⎟⎠
p 2
Periapse (perigee, perihelion): θ = 0 → rp = p = 2a
1+e 1 − e2
→ p = a (1 − e2 )
→ rp = a (1 − e ) ,ra = a (1 + e)

1 2 µ
Energy Conservation: v − =E ( µ = GM)
2 r
1 2 µ
At perigee vp − =E
2 a (1 − e )
µ
E+
a (1 − e )
2
1 2 µ ⎛ vp ⎞
At apogee va − =E ⎜ ⎟ = *
2 a (1 + e ) ⎝ va ⎠ µ
E+
a (1 + e )
Angular momentum conservation: r v θ = h (or r 2 θ = h )
at perigee: h = a (1 − e ) vp vp 1+ e
=
at apogee: h = a (1 + e ) v a va 1−e **
µ
E+
a (1 − e )
2 2
equate (*) = (**) ⎛1 + e ⎞ ⎛1 + e ⎞ µ 1+e µ
⎜ ⎟ = ⎜1 − e ⎟ E + a =E+
⎝1 − e ⎠ E+
µ ⎝ ⎠ (1 − e )
2
a (1 − e)
a (1 + e )
⎡ (1 + e )2 ⎤ µ ⎛ 1+ e⎞
E⎢ − 1⎥ = ⎜ 1−
⎢⎣ (1 − e )
2
⎥⎦ a (1 − e ) ⎝ 1 − e ⎟⎠
4e µ −2 e µ
E = E=− indep. of e (given a)
(1 − e )
2
a (1 − e ) (1 − e ) 2a
and then 2µ µ µ 1+e µ 1+e µ 2ra

υp2 = − = vp = =
a (1 − e ) a a 1 − e a 1−e rp ra + rp
and µ 1−e µ 2rp

va = =
a 1+e rp ra + rp
⎡ µ 1+e ⎤
and ⎢h = a (1 − e )
a 1−e
= ( )
µ a 1 − e2 ⎥ or h= µp
⎣⎢ ⎦⎥

Period r2θ = h
1 dA h h 2A
dA = r (r dθ ) = A= T T =
2 dt 2 2 h
A = πab = πa2 1 − e2 a3/2 indep. of e

T = 2π
µ
h= (
µa 1 − e 2
)
Velocity: From energy conservation 1 v2 − µ = − µ

2 r 2a
2µ µ
v= −
r a
h (
µa 1 − e2 ) µrarp / (ra + rp )
2µ µ µa 1 − e
2
( )
vθ = vθ = = vr = − − =r
r r r r a r2
=rθ
Time in orbit: dθ h µa 1 − e2 ( ) t = t (θ)

(1 + e cos θ )
2
= 2 = 2 →
dt r a 1 − e2 ( )
not easy – Lambert’s prob.
(except for full orbit)

Path angle:
vr dr e ( + sin θ )
tan γ = = =+
vθ r dθ 1 + e cos θ
µ ⎡ 2πv r3 ⎤
Circular orbits r=a → v= ⎢T = = 2π ⎥ check,
r v µ ⎥⎦
⎣⎢
Time in orbit (elliptic case)
dθ µ
(1 + e cos θ )
2
=
( )
3
dt 3
a 1−e 2
1 + cos θ θ 1
= cos2 =
2 2 1 + t2
µ dθ θ 2 1 − t2 2dt
dt = tan =t cos θ = − 1 = dθ =
( ) (1 + e cos θ ) 2 1 + t2 1 + t2 1 + t2
3 2
a3 1 − e2
=
(
2 1 + t2 dt ) =
2 1 + t2
dt
(1 + t ) (1 + e )
2 2 2
2
+ e − et2 ⎛ 1−e 2⎞
⎜1 + 1 + e t ⎟
⎝ ⎠

1
1−e 2 E 1+e E dE
Define E by t = tan2 t= tan 1+e 2
dt =
1+e 2 1−e 2 1−e E
cos2
2
1+e E
1+ tan2
µ 2 1−e 2 1 + e dE / 2
dt =
( ) (1 + e )
3 2
3
a 1−e 2
⎛ 2 E⎞
2
1−e E
cos2
⎜ 1 + tan 2 ⎟ 2
⎝ ⎠
1+e E
1+ tan2
µ 1 − e2 1−e 2 dE = 1 − e
2
⎛ 2 E 1+e 2 E⎞ 1 − e2 ⎛ 1 + cos E 1 + e 1 − cos E ⎞
3
dt = ⎜ cos 2 + 1 − e sin 2 ⎟ dE = 1 + e ⎜ + ⎟ dE
a 1+e ⎛ 2 E⎞
1+e ⎝ ⎠ ⎝ 2 1−e 2 ⎠
⎜1 + tan 2 ⎟
⎝ ⎠
µ
dt =
1 − e2 ⎛ 1
−
e ⎞
cos E ⎟ dE =
(1 − e cos E) dE
a3 ⎜
1 + e ⎝1 − e 1 − e ⎠ 1 − e2
µ (t from perigee passage)

dt = E − e sinE
a3
⎛ 1−e θ⎞
with E = 2 tan−1 ⎜ tan ⎟
⎜ 1+e 2 ⎟
⎝ ⎠
E 1−e θ θ
1 − tan2 1− tan2 1 + e − (1 − e) tan2
from which cos E = 2 = 1 + e 2= 2
E 1−e θ θ
1 + tan2 1+ tan2 1 + e + (1 − e) tan2
2 1+e 2 2
1−e
1 + cos θ − (1 − cos θ)
1 − cos θ 1+e
t2 cos θ + cos θ = 1 − t2 t2 = cos E =
1 + cos θ 1−e
1 + cos θ + (1 − cos θ)
1+ e
2e + 2 cos θ e + cos θ cos E − e 1 − e2

cos E = cos E = ⇒ cos θ = 1 + e cos θ = *
2 + 2e cos θ 1 + e cos θ 1 − e cos E 1 − e cos E
So, directly
µ (1 − e2 )3 2 (1 − e cos E)2 − (cos E − e)2 1 − e2 sinE

dt = dθ (1 − e cos E)2 sin θ = =
a3
(1 − e2 )2 1 − e cos E 1 − e cos E

µ 1 1 − e2 1 − e2 sinE sinE (1 − e cos E ) + ( cos E − e sinE)
dt = (1 − e cos E) dE = (1 − e cos E) dE dθ =
a3 1−e 2
1−e 2
1 − e cos E (1 − e cos E)2
µ 1 − e2 dE
t = E − e sinE 1 − e2 dθ =
a3 1 − e cos E
P a(1 − e2 )
From (**) r= = (1 − e cos θ) r = a(1 − e cos E)
1 + e cos θ (1 − e2 )
ae − a cos E = r(− cos θ)
a (cos E − e) = a (1 − e cos E) cos θ
cos E − e
cos θ =
1 − e cos E

Lecture 33: Performance to LEO
∆V Calculations for Launches to Low Earth Orbit (LEO)
Ideal Earth-to-orbit launch
∆V1 cos α RE = v2' R
∆V12 µ V '2 µ
− = 2 −
2 RE 2 R
2
1⎛ RE ⎞ µ
= ⎜ ∆V1 cosα ⎟ −
2⎝ R ⎠ R
µ µ
−
R∈ R
∆V12 = 2 2
⎛R ⎞
1 − ⎜ ∈ ⎟ cos2α
⎝R⎠
RE
1−
µ R
∆V1 = 2 2
RE ⎛ R ⎞
1 − ⎜ E ⎟ cos2α
⎝ R ⎠
µ
vc =
R
⎡ ⎤
RE ⎢ R ⎥
1− 1− E
µ R∈ µ R µ ⎢ RE RE R ⎥
∆V2 = v c − v '2 = − cos α 2 = − cos α 2
R R R∈ ⎛ RE ⎞
2
RE ⎢ R R ⎛ R ⎞
2 ⎥
α ⎢ ⎥
⎟ cos α
2 2
1−⎜ ⎟ cos 1−⎜ E
⎝ R ⎠ ⎢⎣ ⎝ R ⎠ ⎥⎦
∆V 2 (1 − η ) 1 −η
∆V = ∆V1 + ∆V2 = + η − η cos α 2 '
µ / R∈ 1 − η cos α
2 2
1 − η 2 cos2α
R∈ ∆V 1 − η cos α
=η = 2 (1 − η ) + η (increasing f.of α )
R µ / R∈ 1 + η cos α
(1 − η )
2
∆VMIN
For α = 0 = 2 + η
µ / RE 1+η

Note: Max at η = 0.064178 → R = 99,260 km (worst altitude)
⎛ ∆V ⎞
⎜ MIN
⎟ = 1.5363
⎜ µ /R ⎟
⎝ F ⎠MAX
⎛ µ ⎞
⎜⎜ = 7910 m / s ⎟
⎟
⎝ R∈ ⎠
1 1 ε
−1 = ε η = 1 −η =
η 1+ε 1+ε
2+ε
1+η =
1+ε
ε2 1+ε 1 ε 3 3 ε 3
2 + = 1 − + ε 2... + ε − ε 2 = 1 + − ε 2...
(1 + ε ) 1 + ε /2 1+ε 2 8 4 2 8
2
⎛ 3 ⎞
ε ⎜1 − ε ...⎟ α =0 η = 0.9 → 1.05128
⎝ 4 ⎠
(approx. 1.05093)
∆V / µ / R∈ η = 0.15095 η = 0.23951 η = 0.87620 η = 0.91392 η = 0.95502

( ∆V ) (GEO, ⎛1 ⎞ (Z = 900 Km) (Z = 600 Km) (Z = 300 Km)
R=42,200Km) ⎜ 2 day, η = 26,580 Km ⎟
⎝ ⎠
α = 00 1.50775 1.45534 1.06387 1.04399 1.02275
(11,918 m/s) (11,504 m/s) (8,409 m/s) (8252 m/s) (8,084 m/s)
α = 150 1.51366 1.46373 1.07960 1.05952 1.03748
(11,965 m/s) (11,570 m/s) (8,534 m/s) (8375 m/s) (8201 m/s)
α = 300 1.53109 1.48854 1.12032 1.09755 1.06953
(12,102 m/s) (11,766 m/s) (8,856 m/s) (8676 m/s) (8454 m/s)
∆V1 ε 5ε 2 ∆V2 ε 7ε 2
NOTE: 1+ − −
µ / RE 4 32 µ / RE 4 32
So, for LEO, mainly ∆V1 (apogee kick)
Sticking to α = 0 , variation with R

1 R 1.1 1.05 2 4 6 10 15.6
= = 1.05
η R∈
∆V 1.02041 1.04651 1.18164 1.28446 1.44868 1.49934 1.52978 1.53626
µ / R∈
Effects of Earth’s Rotation
(a) ∆V reduction
Ω R∈ = 463 m / sec
β is launch azimuth w.r.t

East
( α = 0 , near-horizontal
launch)
∆V1 = rocket-imparted ∆V
Starting velocity (abs.) is now
( ∆V1 ) + ( Ω R∈ cos L ) + 2∆V1 ( Ω R∈ cos L ) cos β

2 2
v1 =
So, v1 replaces ∆V1 in previous formulation

RE µ
1− 2
µ R R µ R / RE
v1 = 2 ⎯⎯⎯ ⎯
→ E
=
RE ⎛ RE ⎞
2 α =0
R ⎛ RE + R ⎞
1 + E
⎟ cos α
2
1−⎜ R ⎜ 2 ⎟
⎝ R ⎠ ⎝ ⎠
R
2µ
RE
= ( ∆V1 ) + ( Ω R∈ cos L ) + 2∆V1 ( ΩRE cos L ) cos β
2 2
R + RE
R
2µ
RE
∆V1 = − ( Ω RE cos L ) cos β + ( ΩR∈ cos L cos β ) − ( ΩRE cos L )
2 2
+
R + RE
R
2µ
R∈
− ( Ω RE cos L sin β )
2
∆V1 = − Ω RE cos L cos β +
R + R∈
Notice rotation reduces ∆V1 even for β = 900 . The benefit is low for some larger β
(Westwards launch). For v1 = ∆V1 , need
Ω R∈ cos L
cos β = − −0.056 cos L (for ∆V1 = 8200 m / s )
2∆V1
(for L = 28.50 , β = 92.80 )
Example: For L = 28.50 , R = 6370+500 = 6870 Km,
∆V1 = −407 cos β + 6.48399× 107 − ( 407 sin β )

2
(8052.8)
2
β (0) 0 ±300 ±600 ±900 ±1200 ±1500 ±1800

∆V1 (m/s)’ 7645 m/s 7694 7841 8042 8248 8402 8459
∆V1 reduction 407 m/s 355 m/s 211 m/s 10 m/s -195 m/s -350 m/s -407 m/s

(b) Orbit inclination
For β = 0 (launch due East), i = L. For any other azimuth, higher inclination.
vR = ΩR∈ cos L
vR ∆V1
=
sin ( β − γ ) sin γ
Sin γ ⎛ sin γ ⎞
vR = ∆V1 ⎜ sin β cos γ − cos β ⎟
Cos γ ⎝ cos γ ⎠
∆V1 sin β
tan γ =
∆V1 cos β + v R
tan β
tan γ =
⎛ vR ⎞
1+ ⎜ ⎟
∆
⎝ 1V cos β ⎠
Given two angles and the side included, find opposite angle
( )
cos i = −cos 900 − γ cos 900 + sin ( 90 − γ ) sin 90 cos L = cos γ cos L

Example: Continuing from previous example,
L = 28.50 → vR = 407 m / s, ' R = 500 Km → RE
tan β
tan γ = , cos i = 0.87882 cos γ
407
1+
∆V1 cos β
∆V1 from previous table
β 0 300 600 900 1200 1500 1800

α 0 28.540 57.490 87.100 117.490 148.550 1800
i 28.50 39.50 61.80 87.50 113.90 ( 41.4 )
0
(28.5 )0
(66.1 )0
retro-orbits (westwards)
−300 −600 −900

−28.500 −57.390
39.40 61.70
slightly different inclination
In reality, we probably require the orbit altitude, the orbit inclination and then launch
azimuth β must be calculated
cos i ⎛ cos i ⎞ 1 cos2 L

cos γ = γ = cos −1 ⎜ ⎟ 1 + tan2 γ = =
cos L ⎝ cos L ⎠ cos γ
2
cos2 i
cos2 L ∆V1 sin β

tan γ = −1 =
2
cos i ∆V cos β + v R
cos2 i
sin γ = 1 −
cos2 L
R
2µ
R∈
with ∆V1 = − vR2 sin2 β − v R cos β ∆V1 cos β tan γ + vR tan γ = ∆V1 sin β
R + R∈

R
2µ
RE v R tan γ vR
or ( ∆V1 )
2
+ vR2 + 2vR ∆V1 cos β = ∆V1 = =
R + RE sin β − cos β tan γ sin β
− cos β
tan γ
R
2µ
v R2 2 v 2
cos β R
+ v R2 + R
= E
⎛ sin β ⎞
2
sin β R + R
− cos β E
⎜ tan γ − cos β ⎟ tan γ
⎝ ⎠
R
2µ
⎡ ⎛ sin β
2
⎛ sin β ⎤ 2
⎞ ⎞ RE ⎛ sin β ⎞
vR2 ⎢1 + ⎜ − cos β ⎟ + 2 cos β ⎜ − cos β ⎟ ⎥ = ⎜ − cos β ⎟
⎢⎣ ⎝ tan γ ⎠ ⎝ tan γ ⎠ ⎥⎦ R + RE ⎝ tan γ ⎠
sin2 β 2sin β cos β 2sin β cos β

1+ + cos2 β − + − 2cos2 β
tan γ
2
tan γ tan γ
⎛ 1 ⎞
sin2 β ⎜1 + ⎟
⎝ tan2 γ ⎠
R
2µ 2
sin2 β R ⎛ 1 1 ⎞
vR2 = E
sin2γ ⎜ − ⎟
sin2γ R + RE ⎝ tan γ tan β ⎠
1
tan β =
1 v R + RE
− R
tan γ sin γ R
2µ
RE
tan γ
tan β =
vR R + RE
1−
cos γ R
2µ
RE
L = 28.50 1
Check: ) → γ = 57.470 → tan β = 0
= 1.7308
0
i = 61.8 ⎛ 407 ⎞
0.6377 − ⎜ ⎟
⎝ 0.8431 × 8053 ⎠
β = 59.980 ( 600 )

cos2 L
−1
tan β = cos2 i
cos L R + RE
1 − vR
cos i R
2µ
RE
cos2 L − cos2 i
tan β =
v R cos L
cos i −
⎛ ⎞
⎜ ⎟
⎜ R + R E ⎟
−1
⎜ R ⎟
⎜ 2 µ
⎝ RE ⎟⎠
V1 vR
Directly: =
sin β sin ( β − α )
vR sin γ
= cos γ −
V1 tan β
sin γ 1 − cos2γ
tan β = =
vR v
cos γ − cos γ − R
V1 V1
cos i
and cos γ =
cos L
v R Cos L
Cos i −
Cos2 β =
1
=
1
=
( )
1 + tan2 β Cos2 L − Cos2 i ⎛ ⎞
2
1+ 2 ⎜ v R Cos L ⎟ − 2 cos i v R cos L + cos2 L
⎛ ⎞
⎜ Cos i − v R Cos L ⎟ (
⎜⎜
⎝ ) ⎟⎟
⎠ ( )
⎜⎜
⎝ ( ) ⎟⎟
⎠
tan α 1 tan2 γ
tan β = = 1 + tan2 β = 1 +
vR R + RE cos β ⎛ vR R + R∈ ⎞
2
1− ⎜⎜1 − ⎟⎟
cos γ 2µ R / RE cos γ 2 µ R / R∈
⎝ ⎠

vR 1
∆V1 =
cos β tan β
−1
tan γ
2
vR cos γ ⎛ vR R + RE ⎞
∆V1 = ⎜1 − ⎟ + tan2 γ
R + R∈ ⎜ cos γ 2µ R / RE ⎟
vR ⎝ ⎠
2µ R / RE
v R cos γ 1 v 2R ⎛ R + RE ⎞ vR ⎛ R + RE ⎞
= + 2 ⎜ ⎟−2 ⎜ ⎟
R + RE cos γ cos γ ⎝ 2µ R / RE ⎠
2
cos γ ⎝ 2µ R / RE ⎠
vR
2µ

Lecture 34: Performance to GEO
∆V Calculations for Launch to Geostationary Orbit (GEO)
Idealized Direct GTO Injection

(GTO = Geosynchronous Transfer Orbit)
Assumptions:
- Ignore drag and "gravity" losses

- Assume impulsive burns (instantaneous impulse delivery)
- Assume all elevations α>0 at launch are acceptable
Launch is from a latitude L, directed due East for maximum use of Earth's
rotation. The Eastward added velocity due to rotation is then
vR = ΩE RE cosL = 463 cosL (m/s) (1)
If the launch elevation is α, and the desired velocity after the first burn is V1,
the rocket must supply a velocity increment
∆V1 = V12 + vR2 − 2 V1 vR cos α (2)
vR
∆V1 v1
The trajectory will then lie in a plane LOI through the Earth's center which
contains the local E-W line. In order to be able to perform the plane change to the
equatorial plane at GEO, we select the elevation α such as to place the apogee of the
1/3
⎛ 2
⎞
transfer orbit (GTO) at the GEO radius R GEO = ⎜ µ T 2 ⎟ = 42, 200 km
⎝ 4π ⎠
(T = 24 hr, µ = 3.986 × 1014 m3/s2)

North
ΩE
L
V1 O
α
EQ
UA
TO R
R
GE
O
GTO
EQUATORIAL GEO ORBIT
Fig. 1
Since OL is perpendicular to OI, the view in the plane of the orbit is:
V1
α
L
r RE
θ
P I
o
RGEO
Fig. 2
The polar equation of the trajectory is r = p ,>0

1 + e cos θ
π
In our case p = R E (corresponding to θ = ). The elevation is given by
2

⎛ dr ⎞ ⎛ e sin θ ⎞
tan α = ⎜ ⎟ = ⎜ ⎟ = e
⎝ r dθ ⎠ θ=π / 2 ⎜ (1 + e cos θ )2 ⎟
⎝ ⎠θ=π / 2
and, in turn, the eccentricity follows from (at θ = π )
RE RE
R GEO = e =1−
1−e RGEO
RE 0
and so tan α = 1 − = 0.849 ; α = 40.3 (3)
R GEO
The angular momentum (per unit mass) is h = µp = µR E .
Equating this to R E V1 cosα ,
µ
V1 cos α = (4)
RE
(i.e., the horizontal projection of the launch velocity is the local orbital speed, for any
apogee radius, RGEO in this case)
µ ⎡ RE ⎞ ⎤
2
⎛
Combining (3) and (4), V1 = ⎢1 + ⎜1 − ⎟ ⎥ (5)
RE ⎢ ⎝ R GEO ⎠ ⎥
⎣ ⎦
and this can now be substituted in (2):
µ ⎡ R ⎞ ⎤
2
⎛ µ
∆V1 = ⎢1 + ⎜1 − E ⎟ ⎥ + vR 2 − 2vR
RE ⎢ ⎝ R GEO ⎠ ⎥ R
⎣ ⎦ E
2 2
⎛ µ ⎞ µ ⎛ RE ⎞
∆V1 = ⎜⎜ − vR ⎟ + ⎜1 − ⎟ (6)
⎟
⎝ RE ⎠ RE ⎝ R GEO ⎠
Upon arrival at I, there will have to be a second burn that will simultaneous
Δ
i
=
L
µ
accelerate the rocket to vGEO = , and rotate the plane to equatorial ( ).
R GEO

vGEO
∆i
∆Va
va,GTO
Fig. 3
The apogee velocity is va,GTO , given by
R GEO va,GTO = ( V1 cosα) R E = µR E (7)
2 2
and so ∆Va = vGEO + va,GTO − 2vGEO va,GTO cos ∆i
µ RE R
∆Va = 1+ - 2 E cos L (8)
R GEO R GEO R GEO
This second burn is probably provided by the spacecraft itself, or else by the
launcher's upper stage.
IDEALIZED TWO - BURN GTO INJECTION
One difficulty with the direct injection scheme is the fact that GEO insertion at I
must occur on the first pass, because the GTO perigee is actually below the Earth's
surface (see Fig. 2). Most operators prefer a temporary parking of the spacecraft in a
GTO orbit which has a perigee above the ground, so as to make functional tests and
adjustments prior to the final apogee burn (over a period of 2-4 weeks). A
modification of the launch sequence to accommodate this is:
(1) Fire Eastwards with α selected for a low apogee ( ∼ 200 km above ground) at the
equatorial crossing.
(2) Fire again at equatorial crossing to raise the apogee to RGEO (no plane change)
(3) At one of the apogee passes, perform the final (circularization + plane change
burn).

The formulation is very similar to the previous case.
The elevation α is now given by
RE
tan α = 1 − (9)
Rp
( Rp = perigee radius RE + 200 km ).
This gives a very shallow trajectory, which is unrealistic; but it is a fair

approximation to a real high-elevation launch, followed by a rapid rotation during the
rocket firing. For RP - R∈ = 200 km , α = 1.740 .
ACTUAL
IDEALIZED
α
x
Fig. 4
Eqs. (5) and (6) still hold, with the quality R GEO replaced by Rp , and so
2 2
⎛ µ ⎞ µ ⎛ RE ⎞
∆V1 = ⎜ - vR ⎟ + ⎜1 - ⎟ (10)
⎜ R ⎟ RE ⎝ RP ⎠
⎝ E ⎠
which is now smaller, since we are going to a much lower apogee (at rp ).
At this apogee (at the equatorial crossing), we have, as in Eq. (7),
µR E
va = (11)
Rp
and we next need to effect a second rocket firing that will increase velocity to that
for the GTO perigee:
µ 2R GEO
vPGTO = (12)
R p R p + R GEO

No plane change is involved yet, so
µ ⎡ 2R GEO RE ⎤
∆V2 = ⎢ − ⎥ (13)
Rp ⎢⎣ R p + R GEO R P ⎥⎦
This places the spacecraft on an elliptical GTO orbit, still in the original plane, with
apogee at R GEO . The speed at this apogee is:
µ 2R P
va,GTO = (14)
R GEO RP + R GEO
and so,
∆Va = vGEO2 + v2a,GTO − 2vGEO va,GTO cos L
µ µ 2R P µ 2R P
∆Va = + −2 cos L
R GEO R GEO R P + R GEO R GEO R P + R GEO
µ 2R P 2R P
∆Va = 1+ −2 cos L (15)
R GEO R P + R GEO R P + R GEO
vGEO
L
∆Va
va,GTO
Fig. 5

Some numerical comparisons
We will illustrate these ∆V 's by considering launches to GEO from two different
locations:
(1) Near the Equator, on at the French kouron complex, and

(2) From mid-latitude, as from Café Canoveral ( L = 28.50 ).
(1) Equatorial Launch
Option (a): Ground to LEO (300 km), plus LEO-GEO Hohman transfer. No plane
changes. Launch to the East.
∆V = ∆V1 + ∆V2 − VR + ∆V3 + ∆V4

To LEO, α= 0 GTOinjection GEO circularization
∆V = (8084 – 463) + (10,151 – 7725) + (3071 – 1573)
= 7,621 + 2,426 + 1,498 = 11,545 m/s
Notice this is more than to Escape from mean Earth ( ∆V 11,200 m / s )

Option (b): Direct injection into GTO from ground
∆V = ∆V1 + ∆V2
α= 0 launch to R = 42,200km GEO circularization
( −463 m / s for rotation)
= (10,420 – 463) + (3071 – 1573) = 9,957 + 1,498 = 11,455 m/s
(2) Launch from L = 28.5o. Launch to East, υR = 407 m / s
Option (a): Direct injection to GTO, circularization + plane change at GEO. 2 firings,
∆V = ∆V1 + ∆V2
Launch with α= 40.30 GEO circularization
andplane change
= 10,070 + 2,102 = 12,172 m/s
Note the two penalizations for latitude: the elevated launch increased ∆V1 , and the
plane change at GEO increases ∆V2 .
Option (b) Direct injection with 3 firings (LEO at 300km)
∆V = ∆V1 + ∆V2 + ∆V3

Launch to a300 km apogee Firing to raise apogee to GEO Circularization
+ Plane change
= 7,512 + 2,605 + 1,830 = 11,947 m/s

Is it true that plane change should be all done at end of GTO?
Actually, a small turning combined with initial ∆V1 (say, from LEO) costs very little
∆V loss, even though V is then large. Try splitting into a ∆i1 and ∆i2 = ∆i − ∆i1
∆V1 = v2c1 + vp2GTO − 2vc1 vpGTO cos ∆i1
∆V = ∆V1 + ∆V2
∆V2 = v2c2 + vp2GTO − 2vc2 vaGTO cos ( ∆i − ∆i1 )
d∆V +2vc1 VP sin ∆i1 +2 vc2 Va sin ( ∆i − ∆i1 )

= − =0
d∆i1 2 vc1 + VP − 2vc1 VP cos ∆i1 2 vc2 + Va2 − 2vc2 Va cos ( ∆i − ∆i1 )
2 2 2
µ µ µ 2R 2 µ 2R1
vc1 = , vc2 = , vp = , va =
R1 R2 R1 R1 + R 2 R 2 R1 + R 2
R2
Call ρ =
R1
ρ 1 1 2
2 sin ∆i1 sin ( ∆i − ∆i1 )
1+ρ ρ ρ1+ρ
=
2ρ 2ρ 1 1 2 2 1 2
1+ −2 cos ∆i1 + − cos ( ∆i − ∆i1 )
1+ρ 1+ρ ρ ρ1+ρ ρ ρ1+ ρ
2ρ 1 ⎡ 2 2 ⎤ 1 2 ⎡ 2ρ 2ρ ⎤
sin2 ∆i ⎢1 + −2 cos ( ∆i − ∆i1 ) ⎥ = 2 sin2 ( ∆i − ∆i1 ) ⎢1 + −2 cos∆i⎥
1+ρ ρ ⎢⎣ 1+ρ 1+ρ ⎥⎦ ρ 1 + ρ ⎢⎣ 1+ρ 1+ρ ⎥⎦
42200 2ρ
ρ= = 6.14265 = 1.31148
6370 + 500 1+ρ
0.52916
Sin (28.5 − ∆i1 )
1.31148 Sin ∆i1 6.14265
=
1 + 1.71999 − 2 × 1.31148 Cos ∆i1 1
1 + 0.28001 − 2 × 0.52916 Cos (28.5 − ∆i1 )
6.14265

Sin ∆i1 0.16280 Sin (28.5 − ∆i1 )
=
2.71999 − 2.62296 Cos ∆i1 1.28001 − 1.05832 Cos ( 28.5 − ∆i1 )
∆i1 = 2.260 optimum ∆i2 = 26.240
⎛ ∆V ⎞ 2ρ 2ρ 1 1 2 2 2
⎜ ⎟ = 1+ −2 cos∆i1 + + − cos ∆i2
⎜ vc ⎟
⎝ 1 ⎠op 1+ρ 1+ρ ρ ρ1+ρ ρ ρ (1 + ρ )
⎛ ∆V ⎞ 1
⎜ ⎟ = 2.71199 − 2.62296 cos ∆i1 + 1.21001 − 1.05832 cos ∆i2
⎜ vc ⎟ 6.14265
⎝ 1 ⎠op
= 0.30178 + 0.23227 = 0.53405 - small improvement
Compare to same with ∆i1 = 0
⎛ ∆V ⎞
⎜ ⎟ = 0.29838 + 0.23868 = 0.53706 - small improvement
⎜ vc ⎟
⎝ 1 ⎠ref

Example: Effects of doing a small plane change ∆i2 simultaneous with the second
(apogee-raising) firing in a 3-impulse direct GTO injection.
1.58
1.57
1.56
(R2-RE)/RE= 0.25
1.55
0.2
dVTot/vcE
1.54 0.15
0.1
1.53
1.52
(R2-RE)/RE=0.05
1.51
1.5
1 2 3 4 5 6 7 8 9 10
Di2(deg)
Total dV for three-impulse launch from L=28.5 deg to GEO. Here vcE =sqrt(mu/RE)

0.975
0.97 (R2-RE)/RE= 0.25
0.965
dV1/vcE
0.96
0.955
0.95
(R2-RE)/RE= 0.05
0.945
1 2 3 4 5 6 7 8 9 10
Di2(deg)
dV1 for three-impulse launch from L=28.5 deg to GEO. Here vcE=sqrt(mu/RE)

0.4
0.39
0.38
(R2-RE)/RE=0.25
0.37
dV2/vcE
0.36
(R2-RE)/RE=0.05
0.35
0.34
0.33
1 2 3 4 5 6 7 8 9 10
Di2(deg)
dV1 for three-impulse launch from L=28.5 deg to GEO. Here vcE=sqrt(mu/RE)

0.23
0.225
0.22
(R2-RE)/RE=0.05
dV3/vcE
0.215
0.21
(R2-RE)/RE=0.25
0.205
0.2
0.195
0.19
1 2 3 4 5 6 7 8 9 10
Di2(deg)
dV3 for three-impulse launch from L=28.5 deg.to GEO. Here, vcE=sqrt(mu/RE)

Lecture 35-36: Impulsive and Low-Thrust Maneuvers in Space
See Lectures 3-4 of 16.522 (Space Propulsion) for coverage of Low Thrust
Maneuvers and Re-positing within an orbit.
We add here material on Hyperbolic Orbits and Interplanetary Transfer.
Hyperbolic trajectories:
p
r= e>1
1 + e cos θ
1 ⎛1⎞
Asymptotes : e cos θ = -1 cos θ = - θ = θ∞ = π − cos−1 ⎜ ⎟
e ⎝e⎠
h= µP still valid
Instead of “semimajor
axis” a>0 distance from
perigee to “center” is (-a),
and we still have
µ
E=− >0
2a
Also
( ) (
p = a 1 − e2 = ( −a) e2 − 1 )
Similarly, distance |focus -
center| is (-a)e (it is ae in
ellipse)
⎛1⎞ ⎛π ⎛ 1 ⎞⎞
Turning angle: δ = π − 2(π − θ∞ ) = π − 2 cos−1 ⎜ ⎟ = 2 ⎜ − cos−1 ⎜ ⎟ ⎟
e
⎝ ⎠ ⎝ 2 ⎝ e ⎠⎠
⎛1⎞
δ = 2 sin−1 ⎜ ⎟
⎝e⎠

Miss distance or Impact Parameter
1
∆ = (−ae) sin(π − θ∞ ) = (−ae) sin θ∞ ∆ = (−ae) 1 −
e2
P e2 − 1 P
or ∆= 2
e ∆=
e −1 e 2 2
e −1
Excess hyperbolic velocity V∞ = 2E =

µ
=
(
µ e2 − 1 )
( −a) p
" C3 " = v2∞ used in many books, reports
Orbit characterized by any pair of parameters, like

(p, e) , (vp , θ∞ ) , (v∞ , θ∞ ) , (v∞ , ∆) , (γp , v ∞ ) , (γp , δ)
µ ∆ ∆v2∞
Example : Given (v∞ , ∆) , -a = e2 − 1 = =
v2∞ −a µ
1 ⎛ v2 ∆ ⎞
then δ = 2 sin−1 = 2 cot −1 e2 − 1 = 2 cot −1 ⎜ ∞ ⎟
e ⎝ µ ⎠
Planetary “Spheres of Influence” (SOI)
Locus of points for which the ratio of (the Sun’s disturbing acceleration of the
relative vehicle-planet motion) to (the planet-induced vehicle acceleration) equals
the ratio of (the planet-induced acceleration of the relative vehicle-Sun motion) (the
Sun-induced vehicle acceleration).
This attracting centers, one vehicle

JJJG
d2 rabs G MP m → G Ms m →
m 2
=− r− R
dt r3 R3
JJJJJG
d2 rabs,p G MP m → G Ms MP →
MP 2
=− r− ρ
dt r3 ρ3
→
⎛ → →
⎞
d2 r G(MP + m) → ⎜ −R ρ
Subt. =− r + GMs + 3⎟
dt 2
r 3
⎜R 2
ρ ⎟
⎝ ⎠
Accel. of vehicle Relative accel. Sun’s effect on d. Notice: difference

Relative to planet due to planet only between attractions vehicle and on planet
We could reverse the roles of planet and Sun, and get
→
⎛ → →
⎞
d2 R G(Ms + m) → ⎜ −ρ r ⎟
=− R + GMP 3 − 3
dt 2
R 3
⎜ρ r ⎟
⎝ ⎠
So, SOI defined by locus of
→ → →
ρ R −ρ r
Ms − Mp −
ρ3 R 3 ρ3 r3
=
Mp + m Ms + m
r 2
R2
Since m << Mp << Ms and r<<p, R,
→ →
Ms ρ R 2 Mp 1 2
− r R
Mp ρ3 R 3 Ms r2
r
− cos ϕ
ρ
−3 2
⎛ → →
⎞ ⎛ →→
⎞
→ → → → →
−3 ⎜ 2 ρ.r ⎟ ρr
Also, R = ρ+ r 3 2 2
R = (ρ + r + 2 ρ . r ) −3 2
ρ 1+ ρ 1−3 2 ⎟
−3⎜
⎜ ρ2 ⎟ ⎜ ρ ⎟
⎝ ⎠ ⎝ ⎠

→ → → →
ρ.r ρ.r G
→ → → 1−3 2 → → → 3 2 ρ
ρ R ρ ρ ⎛ ⎞ −r ρ r ⎛ → →
⎞
3
− 3
3
− 3 ⎜ ρ+ r ⎟ 3
+ 3
= 3 ⎜ − 1r − 3 cos ϕ 1ρ ⎟
ρ R ρ ρ ⎝ ⎠ ρ ρ ρ ⎝ ⎠
r 3r cos ϕ
Magn.:
ρ3 ρ3
− cos ϕ
→ →
ρ R r r
(
− 3 = 3 1 + 9 cos2 ϕ + 6 cos ϕ (1r − 1ρ ) )
12
3
1 + 3 cos2 ϕ
ρ R ρ ρ3
2
r3 2 ⎛ MP ⎞ R 2
1 + 3 cos ϕ = ⎜ ⎟ 2
ρ3 ⎝ Ms ⎠ r
(Mp Ms )
25 25
r r ⎛ Mp ⎞
= → ⎜ ⎟
( )
1 10
R 1 + 3 cos2 ϕ R ⎝ Ms ⎠
But (1 + 3 cos2 ϕ )
1 10
is between 1 and 1.15
Planet SOI (Km)

Mercury 113,000
Venus 617,000
Earth 924,000
Jupiter 48.3 × 106
Neptune 86.7 × 106
Moon 66,000 (in E-M system) (17.2% of FEM , not too small)
The Patched Conic Method for Interplanetary Transfers
Since the SOI is typically small compared to the interplanetary distances,

when dealing with trajectories that go from one planet to another we can make the
approximation that only one body is attracting the space craft at each time. This is
initially the first plane then the Sun, and finally the destination planet. Further, the
“hand-over” from one planet to the sun or vice-versa can be assumed to be at the
planet’s location, when viewed on the Solar System scale. Care must be taken to
converse momentum at these hand-over points.

Example: Hohmann transfer from Earth to Venus. Assume no plane changes are
involved (in reality the Elliptical plane of the planetary orbits does not coincide with
the equation plane of either, but we ignore this complication).
As a preliminary, notice that this type of (minimum ∆ v) maneuver requires a

specific Earth–Venus configuration at launch; this configuration occurs once every ∆t
days, given by
1 1
∆t = = = 584 days (1)
⎛ 1 ⎞ ⎛1⎞ 1 1
⎜ ⎟ − ⎜ ⎟ 224.7 − 365.3
⎝ Tv ⎠ ⎝ TE ⎠
The Earth head angle at launch ( d0 in the sketch) is calculated by stating that the
time in the ITO (half the ellipse’s period ) is the same as that taken by Venus
between Vo and Vi :
32
1 2π ⎛ R SE + R SV ⎞ R 3SV2
⎜ ⎟ = ( π + α0 )
2 µs ⎝ 2 ⎠ µs
(2)
or
⎡⎛ R + R ⎞32 ⎤
α0 = π ⎢⎜ SE SV
⎟ − 1⎥
⎢⎝ 2 R SV ⎠ ⎥
⎣ ⎦
(3)
Using Rsv = 0.7223 a.u , this gives α0 = 54.00 .
In the heliocentric part of the trajectory, the apohelion velocity (which the
craft must have as it leaves Earth) is
µs 2 R SV
va = (4)
R SE R SE + R SV

µs
while the Earth’s circular velocity is vE = . This is more than va , i.e., the
R SE
spacecraft must leave the Earth’s vicinity traveling backwards in the heliocentric
frame. Later on it will overtake Earth as it accelerates in the transfer free-fall
trajectory. The magnitude of this backwards velocity, which is the hyperbolic excess
velocity when viewed from Earth, is
⎛ 2 R SV ⎞
v ∞,E = vE − va = vE ⎜ 1 − ⎟ (5)
⎜ R SV + R SE ⎟
⎝ ⎠
The situation for time near launch, and when viewed in the Earth frame is as
shown in the following sketch:
Conversation of
energy in the
Earth’s frame gives
v2∞ ,E 2
vPE µ
= − E (6)
2 2 rPE
where υPE is the velocity after application of the escape firing at rPE . Before this
µE
firing, the space craft was in orbit, at vCE = , and so we find
rPE
2µE
vPE = v2∞,E + (7)
rPE
and therefore the single impulse needed to enter the ITO is

µE ⎛ r v2 ⎞
∆v1 = vPE − vCE = ⎜ 2 + PE ∞,E − 1 ⎟ (8)
rPE ⎜ µE ⎟
⎝ ⎠
The point within LEO where this firing must occur is given, as shown, by δ , where
2
⎛1⎞ rP,E r
δ = 2 sin−1 ⎜ ⎟ is the total hyperbolic turning angle. Since e = 1 + = 1 + P,E v2∞,E we
⎝e⎠ ( )
− a µE
have all the elements to calculate this angle.
After traveling in the ITO, the spacecraft will approach Venus with an
overtaking excess velocity
µs ⎛ 2 R SE ⎞
v ∞,ν = vp − v v = ⎜ − 1⎟ (9)
R s,v ⎜ R SE + R SV ⎟
⎝ ⎠
and, working now in

the Venus frame
2
v2∞.v vp.v µ
= − v (10)
2 2 rρv
µv
v c.v = (11)
rρv
So that the circularization ∆ v is
µ v ⎛⎜ rpv v2∞,v ⎞
∆v2 = vρ,v − vc,v = 2+ − 1⎟ (12)
rρv ⎜ µv ⎟
⎝ ⎠

Lecture 37: Future Developments
Some (dangerous) Forecasts on the Future of High Thrust Rockets (Chemical or

Nuclear)
Looking ahead in Rocketry
Liquid Rockets :
- No increase in Isp to be expected.
- Gradual reliability increase through:
o Better Health Monitoring, diagnostics, control, auto-reconfiguration

o Added redundancy
o Better materials, simplified designs
- Shorter, less expensive design cycle by
o Heavy use of computation

o Active stability enforcement
- Potential replacement of propellants Æ “green” propellants (H2O2 + HC ?

instead of Hydrazine + N2O4)
Solid Rockets :
- Stronger, lighter casings
- Active stability controls
- Less expensive manufacturing
- Basic combustion physics Æ “designer propellants”
Hybrid Rockets : Potential to replace solids (similar Isp, re-start, control, safety)
Need better sliver control techniques
Nuclear Rockets : Needed for interplanetary flight. Political issue. Phoebus 2A, 12 min
(1968) 119 Kg/s, H2 Pth = 4 GW, 2280 K, 47 atm, 9500 Kg mass
F 10
w
F 119 x 8000 ≈ 106 N 100 Ton

Open cycle, cannot (should not) operate from ground.
Can be launched chemically to orbit (inactive), then be used as upper stage.

Shielding only “shadow”, will require large exclusion zone about vehicle ( ∼ 10 Km,
except in protected cone)
New design by Carlo Rubbia claims scalability to small size. A thin films
surrounding H2 channels, says can build critical engine for 30 KW H.


Rocketry MIT

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Rocketry MIT

Uploaded by

Copyright:

Available Formats

16.

512, Rocket Propulsion

Types of Rockets (Engines)

- Depending on gas acceleration mechanism/force on vehicle mechanism.

“Thermal” Gas pushes directly on walls by P (pressure) forces

16.512 concentrates on Thermal

- Depending on energy source:

Chemical (always “thermal”) Liquid Propellant Monopropellant

16.512 deals mostly with Chemical.

- Depending on Thrust level (per unit mass)

- High thrust ( ≥ 1g) for launch, fast space maneuvering (16.512)

16.512, Rocket Propulsion Lecture 1

Dominant for chem. Rockets, range 200-500 sec

Thermal Efficiency ηth = Jet kinetic power

Very large ~ (20-100) for Chem.

Others (design selection factors)

- “Life”, most meaningful in total impulse capacity

Rocket Selection Guide (by mission)

16.512, Rocket Propulsion Lecture 1

2) Launch to space Solid, liquid or combinations,

3) Impulsive ∆V in space Small Solid Prop. (Apogee

16.512, Rocket Propulsion Lecture 1

Rocket Thrust (Thermal rockets)

16.512, Rocket Propulsion Lecture 2

In general then, define ue =

If things are nearly constant on spherical caps, modify control volume to

dAex = dA cos θ ux = ur cos θ

16.512, Rocket Propulsion Lecture 2

For ideal conical flow, ρ , ur , P are constant over Ae . Then

Also, since Pe = const on the exit surface, P e = Pe

16.512, Rocket Propulsion Lecture 2

At design, Pe = Pa (and parallel flow beyond). Also ue x

16.512, Rocket Propulsion Lecture 2

Start with a near-stagnant flow in the upstream plenum (“combustion chamber”, or

Since Pe ≅ uniform, so is ηN , even when htc is not. Also, υe is non-uniform if htc is

16.512, Rocket Propulsion Lecture 2

Consider a case where htc varies from streamtube ( dm ) to streamtube (but

and the input power is P = ∫∫ h tc dm (2)

⎧⎪P is minimum(For a given F, m)⎫⎪

FUNIF . = 2 ηN m htc ; PUNIF = m htc

Now, express in general F by (1) and P by (2)

16.512, Rocket Propulsion Lecture 2

Important in arcjets, less in film-cooled chemical rockets.

16.512, Rocket Propulsion Lecture 2

Ideal Nozzle Flow with No Separation (1-D)

- Quasi 1-D (slender) approximation

16.512, Rocket Propulsion Lecture 3

Compute m = ρ uA at sonic throat:

16.512, Rocket Propulsion Lecture 3

16.512, Rocket Propulsion Lecture 3

Choice of Optimum Expansion For a Rocket Flying Through an Atmosphere

The thrust-derived velocity increment is

16.512, Rocket Propulsion Lecture 3

so that (5) becomes

and we can now differentiate v w.r.t Me (holding p=const.)

16.512, Rocket Propulsion Lecture 3

Expand & simplify

16.512, Rocket Propulsion Lecture 3

Notice that the exit pressure is given by

and so the optimum exit pressure turn out to be

with a similar expression for Me :

The limiting condition in which the whole burn occurs at Pa o is simple.