16.

50 Lecture 1

Subjects: Rocket Equation; Gravity Loss; Optimum Acceleration.

1) Rocket Equation

A rocket is a propulsive device that produces a thrust force F on a vehicle by ejecting

mass a high relative velocity c. This force is simply equal to the rate of momentum
outflow from a control volume that encloses the vehicle. If m! is the mass expulsion rate,
we then find
.
F= mc
.
This is a particularly useful result because c and m are insensitive to the operating
environment of the rocket. In more detail, c is not necessarily the velocity of the expelled
mass as it leaves the rocket’s exit plane, because the pressure field may continue
accelerating this mass for some distance. In that case, c represents the limiting “far field”
velocity of the ejected mass.

For a rocket-powered vehicle in a vertical launch from the Earth,

dv dm dm
m = !c ! mg " dv = !c ! gdt
dt dt m

1
 t dm t t
v = ! "0 c(h) ! "0 g(h)dt; h = !0vdt
m

As indicated, both c and g are functions of h. In the case of c, the dependence is through
the atmospheric pressure, p0, an effect we will explore in detail later. The atmospheric
pressure varies with weather and with temperature, but a rule of thumb is:
r!r
p0 ! E
= e 6.68 r - rE = km
p0 (rE )
In fact the pressure variation is much more important in the context of aircraft engines,
whose thrust is essentially proportional to the atmospheric density, other factors being
equal.

The gravitational attraction varies as g(r) = gE (rE/r)2 , where rE is the Earth's radius, r =
rE + h and gE is the gravitational acceleration at the Earth’s surface. Since rockets
typically operate for a short time only, g may be taken here as a constant to a good
approximation.

If as a first approximation we take c and g constant we can integrate the rocket equation
to get:
m(t)
v ! v 0 = !c ln ! gt
m0
and this is the usual form of the Rocket Equation. Remember that it applies to a vertical
launch. For other flight paths the last term changes. Also remember that we have
neglected atmospheric drag forces (this last approximation is valid for large rockets, but
may not be for smaller vehicles, which have more area per unit volume).

2) Gravity Loss & Optimum Acceleration

As we will see, it is helpful to express this relation as one for the mass ratio in
terms of the velocity change, since it is the latter that is set by the mission, e.g. launch to
.
LEO. But this leaves gt to be determined. If we make the additional assumption that m
= constant, then
. F ma
m0 ! m(t) = m t = t = 0 0 t
c c
where a0 is the “initial acceleration”, defined as the thrust divided by the initial mass.
Note that a0>g for achieving takeoff.

So now we can eliminate t, to find

v ! v0 m(t) g " m(t) %

= !!n ! 1!
c m0 a0 #$ m0 '&

2
 m(t) v ! v0
This is an (implicit) equation for in terms of , with the initial acceleration a0
m0 c
as a parameter.

The last term is often termed the"gravity loss", since it is the reduction in velocity
increment that the vehicle suffers due to the (downward) acceleration of gravity. As a0
increases, the gravity loss decreases, since the time over which gravity acts is reduced. So
v ! v0 m(t)
why not use a very large a0 to get the largest possible for a given ? To see
c m0
why we must examine the effect of the acceleration on the mass of the propulsion system.
We divide the mass of the rocket into parts:

m0 = mstruct + meng + mpay + mprop

When all the propellant has been expended, at the time tb, m(t b ) = mstruct + meng + mpay , or
as a fraction of initial mass
m(t b ) m0 ! m prop m
= = 1 ! prop " 1! m' prop
m0 m0 m0
Here the prime on the propellant fraction indicates that it is divided by m0, i.e. is the
fraction of the initial mass represented by the propellant.

m(t) v ! v0
Clearly, the smaller or the closer m' prop is to 1, the larger the that can be
m0 c
achieved. But the limit is set by:

m(t b )
= 1 ! m' prop= m' struct +m' eng +m' pay
m0
We want to design the components of our vehicle so that m' struct and m' eng are as small as
possible, given the loads and available technology.

The engine mass is proportional to the thrust it develops; define

Engine Weight meng g meng m"

!= = = = eng
Thrust F m0 (a0 / g) n

where we call for short a0/g=n. Notice that this is the number of “g’s” due to thrust only,
while the more commonly used g-number (the specific downwards force felt by a pilot)
includes gravity and would be n+1.

We now have m!(tb ) = m!pay + ms!tr + n" , and the rocket equation becomes
1! m"pay ! m"str
vb ! v0 = !c ln(m"pay + ms"tr + n# ) ! ( !#)
n

3
We notice here that the logarithmic term (the ideal velocity increment) decreases with n
(heavy engine), but the last parenthesis (the gravity loss) decreases with n (reduced burn
time). We can optimize the choice of initial acceleration by differentiating wrt. to n:

" 1! m#pay ! ms#tr

! + =0
m#pay + ms#tr + n" n2
which is a quadratic equation with the solution

1! m"pay ! ms"tr 1! m"pay ! ms"tr 2 (m"pay + ms"tr )(1! m"pay ! mstr

" )
nopt = + ( ) +
2 2 #

Some reasonable estimates are:

m' struct ! 0.1 ; ! = 0.02

For these values, Fig. 1 shows how the actual velocity increment varies with n=a0/g for a
range of payload fractions. The optima are clearly visible, and can be obtained from the
graph or calculated from our quadratic solution. These nopt values, and the associated
maximum velocity increments, are shown in Fig. 2. Interestingly, nopt varies little, and
stays in the range from 3 to 4, familiar to astronauts. The velocity increment is usually set
by the mission requirement, e.g. to get to LEO, and Fig. 2 illustrates how small the
payload fraction becomes if this required velocity increment exceeds the jet velocity c (in
fact, for this example, m0! = 0 for (v ! v0 ) ! 1.5 )

This optimization has considered only the effects of the "gravity loss" and the engine
a
weight on the choice of 0 . If we include other effects, such as drag, we may get a
g
different answer. A large value of n implies reaching high speeds at low altitudes, where
air density is still high, and for small rockets this would generate large drag losses.

4
Fig. 1

5
Fig. 2

6
16.50 Lecture 2

Subjects: Rocket staging; Range of aircraft; Climb & Aceleration

1) Rocket Staging

The reason for staging is to avoid having to accelerate empty tanks. Assume for
simplicity only two stages; one does not want to stage either too early (and then carry a
heavy second stage tank) or too late (and carry a heavy first stage tank for a long time). It
can be shown that for ideal symmetrical stages (same specific impulse, same structural
fractions), the velocity increment should be divided equally between the stages, and this
is a good first cut for more general cases.

In its simplest form, staging consists of just replacing the payload of the first stage by a
complete second stage, which in turn has its own payload, as shown in the sketch.

Payload

Stage 2

Stage 1

Image by MIT OpenCourseWare.

We go back to the rocket equation, and neglect gravity losses. Write for short ! = ms"tr ,
and assume this structural fraction is the same for both stages (in each case the structural
mass is normalized by the initial mass of that stage); assume also that the jet velocity c is
the same for both stage engines (although typically the first stage engine will be larger
and have a bigger thrust). The final payload is given by

(m pay )2 = (m 'pay )2 (m0 )2

and we also have
"V1
!
( m0 )2 = (m f )1 ! (mstr )1 = (m0 )1 (e c
! #)

1
 #V2
"
and (m!pay )2 = (m!f )2 " (ms!tr )2 = e c
"$

But the total velocity increment !V = !V1 + !V2 is prescribed, and so we obtain

"V "V !"V1

(m pay )2 ! 1 !
= (e c
! # )(e c ! # )
(m0 )1
"V "V1 "V !"V1
! ! !
=e c
! # (e c
+e c
)+ #2

It is easy to se that this is maximum when !V1 = !V " !V1 , i.e., when !V2 = !V1 . If this
staging is selected, the overall payload fraction is

"V
(m pay )2 !
( )opt = (e 2 c ! # )2
( m0 )1
This derivation can easily be extended to N>2 stages, but things are more complicated if
the structural fractions or the jet velocities are different among stages.

Range of Aircraft

For aircraft, the simplest measure of performance is cruise fuel consumption and
the resulting range. At one time this was also the critical performance measure for
transports, bombers and fighters. This is less true now for transports because the ranges
accessible with modern engines and airframes are in the order of 8,000 miles. For
bombers aerial refueling extends the range to the extent that again range is no longer such
a challenge (although refueling is quite expensive). Range is still important for fighters
because the requirements for high speed and maneuverability conflict with those for long
range.

Consider an aircraft in straight, level flight

L mg
F= D= =
L/D L/D

2
We define the Specific Impulse by:
.
F = gI m

. dm .
where m is the fuel mass flow, so that m = - . Note that the inverse of I is the
dt
‘Specific Fuel Consumption, SFC, in appropriate units. From this definition and the force
balances we obtain

mg dm
= !gI
L/D dt

dm D
= ! dt
m LI

If we assume D/L and I are constant (Here we are implying models for both the
propulsion system and the aircraft), then

m(t) D
!n =! t
m0 LI

The range R = u0t where u0 is the flight velocity, so

L m
R = u0 I !n 0
D m(R)

where m(R) is the aircraft mass at the range R. This is the Breguet Range Equation.

As in the case of the rocket, it is useful to divide the mass m0 into structural,
payload, engines and fuel:
m0 = mstruct + meng + mfuel + mpay

1 = mstruct + meng + mfuel + mpay

' ' ' '
or

If the fuel is expended at R,

m(R)
= mstruct + meng + mpay = 1! m fuel
' ' ' '

m0

For a fixed m'struct and meng , we can trade off between m pay and m fuel , hence between
' ' '

Range and Payload. If we define:

mempty = mstruct + meng

' ' '

3
we can write
m(R)
= mempty + m pay
' '

m0
and
L 1
R = u0 I !n '
D mempty + mpay
'

from which
R
!
=e ! mempty
' U 0 I(L/ D ) '
m pay

(quite similar to what we obtained for a rocket).

From this we can construct a Range vs. Payload chart. As an example, suppose

L m
mempty = 0.7; = (300 )(4000s)(15) = 1.8x10 7 m
'
u0 I
D s
R( km )
!
=e ! 0.7
' 4
1.8x1 0
m pay

.3

'
mpay

0
0 .2 .3 .357 .4
.1

R(km)
1.8 x 104
Note this is not a straight line, although it is close.

Climb & Acceleration

Sometimes we are more interested in climb and maneuver rather than cruise, as
for fighter aircraft. Then an Energy Approach is most helpful.

4
Suppose the aircraft is climbing at angle ! from the horizontal. The equation of motion
along the path is
du
m 0 = F ! D ! mgSin"
dt

du0 (F ! D)u0
u0 = ! gu0 Sin"
dt m

dh
but u0 Sin! = , so it follows that
dt

dh du (F ! D)u0 dE
g + u0 0 = "
dt dt m dt

u0 2
Were we define a Total Energy E = gh + per unit mass.
2
If we had a simple model for F and D as a function of u0 and h we could integrate this as
we did for the rocket. But, as we shall see later in the semester these dependencies are
much more complex for the aircraft engine than for the rocket. Thus for the two classes
of engine we have these quite different thrust characteristics:

Rocket Engine
- Nearly independent of its environment, atmosphere and speed.
Aircraft Engine
- Strongly dependent on flight speed and atmosphere.

5
16.50 Lecture 3

Subjects: Orbital mechanics; Single force center

The most usual application of rocket engines is to propel vehicles under conditions where
the behavior of the vehicle is largely determined by the gravitational attractions of one or
more bodies of the solar system, and where aerodynamic drag is not very important. So it
is essential to understand the behavior of an orbiting body in order to appreciate the
requirements which must be met by the rocket engine. For this reason we shall spend a
short time discussing orbital mechanics. It should be noted, however, that the intent is to
present only the aspects that define the requirements for propulsion systems, not to
discuss the details of orbital computations. For these the student should refer to a text on
celestial mechanics or spacecraft guidance.

Forces between bodies. The planetary Sphere of Influence

The gravitational attraction between two bodies of masses m and M is given by

1
Q = GmM 2
r
where G = 6.67x10-11m3 kg-1 s-2, and r is the distance between the bodies. The
gravitational acceleration is then f = GM / r 2 ! µ / r 2 , where µ=GM is the gravity
constant of the body with mass M.

We are interested in the motion of a small body (spacecraft) in the force fields of one or
more larger bodies. If we consider the Sun and two or more planets, the problem is
extremely difficult mathematically. For our purposes, it can be simplified immensely by
considering the motion of the spacecraft under the influence of only the dominant
attractor at any given time. On the interplanetary scale, this means the Sun most of the
time, but near enough one of the planets (inside its “Sphere of Influence”, SOI), the
planet will dominate. To estimate correctly the radius of the SOI, one should work in the
frame of reference of the planet. The gravitational acceleration of the planet alone on the
S/C is fp,SC, and the perturbation due to the Sun of the spacecraft acceleration with respect
to the planet is fS,SC-fS,p. Conversely, the gravitational acceleration of the Sun alone on the
S/C is fS,SC, and the perturbation due to the planet of the spacecraft acceleration with
respect to the Sun is fp,SC-fp,S, which, since the distance to the Sun is much greater that that
to the SC, is almost equal to fp,SC alone. We now state that the relative errors in ignoring
either of the two perturbations are the same on the SOI:

fS,SC ! fS, p f p,SC f p,SC

! !
f p,SC fS,SC fS, p

The difference on the left hand side numerator is approximated to first order as

" fS r
fS,SC ! fS, p ! rp,SC = 2 µS p ,SC
"r rS,3 p
Substituting this and the other accelerations into the SOI definition, and rearranging, one
finds rSOI (= rp,SC) to be given by

rSOI 1 M
! 1/5 ( p )2/5
rS, p 2 M S

1
Since Mp/MS is a small number, rSOI/rS,p<<1. We can therefore model the relative body
motion near the planets (r<rSOI) as being under the influence of a single force center (the
planet), while the body and the planet experience a common acceleration toward the Sun.
Far from the planet (r>rSOI) we ignore its field and consider the body motion as influenced
only by the Sun. For preliminary calculations, these limiting models are simply “patched”
at the SOI.

To see how far from a planet is far, we refer to Table 1-1, which gives the ratios of the
planet's mass to that of the Sun and the size of hteir spheres of influence.

Table 1-1

Relative mass Orbital Radius, km rSOI (Km)

Mercury .167 x 10-6 .578 x 108 98,000
Venus .245 x 10-5 1.08 x 108 536,000
Earth .300 x 10-5 1.49 x 108 791,000
Mars .324 x 10-6 2.27 x 108 501,000
Jupiter .956 x 10-3 7.77 x 108 41.6x106
Saturn .286 x 10-3 1.42 x 109 47.3x106
Uranus .437 x 10-4 2.86 x 109 44.9x106
Neptune .518 x 10-4 4.49 x 109 75.5x106
Pluto .28 x 10-5 5.89 x 109 30.8x106
Earth's Moon, MM/ME=.368 x 10-7 rE,M=.384 x 106 57,600

We see for example that the sphere of influence of the Earth is just about twice the
distance to the Moon. On one hand this means we can just barely ignore solar
perturbations in considering cis-lunar operations not very near the Moon, but also that the
Moon’s own motion about the Earth must be appreciably affected by the Sun’s attraction.
Notice also the relatively large size of the SOI of the Moon in the Earth-Moon system
(15% of the Earth/Moon distance).

We can therefore reduce the complex many-force center situation to a set of simpler ones,
namely
a) Motion in the single force field of a planet
b) Motion in the single force field of the Sun
c) Transition from a Sun-dominated to a planet dominated situation, or vice
versa. As we shall see this transition can be thought of as a change of coordinate system.

Motion under a Single Force Center

The subject of motion under a single gravitational attractor is covered in all Dynamics
textbooks, and so we only give here a summary of the main conclusions and working
equations.

2
 r
p
π-θ

a(1-ε) c=aε

p
With reference to the figure, the polar equation of the trajectory is r= ,
1+ ! cos"
where p = a(1! " 2 ) , ε is the eccentricity and a is the semi-major axis. If ε<1, all the
quantities listed are positive and the trajectory is closed (an ellipse with a focus at r=0).
The case when ε>0 will be discussed later. The quantity p is called “the parameter”, and
has the significance shown in the figure. The minimum and maximum radii are called the
“periapsis” and “apoaxis”, respectively (perigee and apogee for orbits around Earth); they
are given by
p p
rp = = a(1" ! ); ra = = a(1+ ! )
1+ ! 1" !
The potential energy per unit mass, with zero at infinity, is E p = ! µ / r .

The two constants of the motion are:

1 2 µ µ
(a) The total energy per unit mass, E= v ! =!
2 r 2a
(b) The angular momentum per unit mass, h = r 2!! = µ p = a(1" # 2 )

a 3/2
The time to complete one orbit is T = 2!
µ
Using the conservation of energy, the velocity magnitude is given by the so-called “vis-
2µ µ
viva” equation: v= !
r a
µ µ
The circular orbital velocity (r=a) is therefore vc = =
r a
Notice that the kinetic energy in a circular orbit is EK = µ / 2a = !E = !1 / 2 E p . More
generally, the average kinetic and potential energies satisfy these same curious
relationships for any elliptic orbit.

The trajectory first becomes open when ε=1, ra tends to infinity, and the total energy
becomes zero. When this happens, the body is on a trajectory to barely “escape” the
attractor, arriving at infinity with zero velocity. The velocity on this trajectory depends on
2µ
distance as vesc = .
r
As a final note, the apogee and perigee velocities occur frequently in orbital calculations.
They are related to the apogee and perigee radii by

3
 µ 2 ra µ 2 rp
vp = > vc, p ; va = < vc,a
rp ra + rp ra ra + rp
Although the formulation in terms of the classical “orbital elements” a, ε, p, etc, is
standard, most simple problems can be solved quickly by using the energy and angular
momentum conservation laws between judiciously selected points, often the apogee and
perigee themselves.

4
16.50 Lecture 4

Subjects: Hyperbolic orbits. Interplanetary transfer.

(1) Hyperbolic orbits

p
The trajectory is still described by r = , but now we have ε>1, so that the
1 + ! cos "
radius tends to infinity at the asymptotic angle ! " = # $ cos $1 (1 / % ).

-aε
-a

p
r
δ θ θ∞

The “parameter” p still has the geometrical significance indicated in the figure, and is
therefore a positive number. It is still related to a and ε through p = a(1 ! " 2 ) , but now a
is a negative number, so it is (-a) that has a geometrical significance, as indicated in the
figure. Note also tat ε is still defined as the ratio of the distance from periapsis to center
to the distance from focus to center.

1 2 µ µ
The energy is still given by E = v ! =! , and is now positive. The angular
2 r 2a
momentum is still given by h = r 2!! = µ p = a(1" # 2 ) .

There are a few new parameters of interest in this case:

1 1
The trajectory deflection, ! = " # 2(" # $ % ) = " # 2 cos #1 = 2 sin #1
& &

1
 p
The miss distance ! = " a# sin($ " % & ) = " a# sin % & =
# 2 "1

µ
The excess hyperbolic velocity, v! = 2 E =
(" a)

In the specialized technical literature, the term “c3” is often used, meaning simply v!2 .

2) Interplanetary transfer

We assume for the moment that our craft has “escaped the field of planet 1” (meaning it
is outside its sphere of influence), and so may be considered to be in orbit about the Sun.
In order for it to reach planet 2, its orbit about the Sun must intersect that of planet 2.

P2 at P1 at
launch launch
r2

r1

Sun

r2

P1 and P2
at arrival
Assume that the planetary orbits are circular. Then it is clear that the trajectory of least
energy which will allow the transfer is that which is just tangent to the orbits of the home
and target planets; this is called the Hohmann transfer orbit, which is the half-ellipse that
is sketched in the figure. The Heliocentric velocity at the start of this Hohman arc is the
periapsis (perihelion in this case) velocity, as described at the end of the last lecture:

µ S 2 r2
vp1 =
r1 r1 + r2

2
so that if we launch the ship in the direction of motion of the planet, it must have a
relative velocity
µS 2 r2
vrel ,1 = ( !1)
r1 r1 + r2

with respect to planet 1 after escape from the planet. By definition, this is the “excess
hyperbolic velocity” relative to the planet, v∞1, and the total energy relative to planet 1 at
he edge of the sphere of influence is simply ½( v∞1)2.

Suppose the launch was for the surface of planet 1 (radius R1), and ignore its rotation.
Just after launch, during which the rocket has imparted an instantaneous velocity
1 µ
increment ΔV1, the energy per unit mass (relative to the planet) is (!V1 ) 2 " 1 , and this
2 R1
2
must be the same as ½( v∞1) , by energy conservation with respect to planet 1 inside the
sphere of influence. We then have
1 µ 1 2 1 µS 2 r2
(!V1 ) " 1 = v#1= " 1)
2 2
(
2 R1 2 2 r1 r1 + r2
from which the first delta-V delivered by the rockets must be

2 µ1 µ S 2r2
!V1 = + ( " 1) 2
R1 r1 r1 + r2

The procedure is similar when considering the approach to planet 2. The spacecraft will
have then a heliocentric velocity equal to the apoaxis (apohelion) velocity
µ S 2 r1
va 2 =
r2 r1 + r2
and a relative velocity with respect to the planet

µS 2 r1
vrel ,2 = (1 ! )
r2 r1 + r2
which is also the excess hyperbolic velocity with respect to planet 2. It is worth noting a
this point that the spacecraft heliocentric velocity is less than that of the planet itself, so
that, as seen from the planet, the spacecraft will be approaching from its advancing side.
For capture into a circular orbit of radius Rc2, the geometry is shown below:

To Sun
ΔV2

Vrel,2
rc,2

3
Just before the insertion rocket firing, the energy per unit mass relative to planet 2 is
equal to ½ (vrel,2)2, and it is also equal to the sum of the kinetic energy at that point of
1 µ
closest approach, plus the potential energy: (vclosest app. )2 ! 2 . Thus we must have
2 rc2

µ2 µS 2r1 2
vclosest app. = 2
+ (1! )
rc2 r2 r1 + r2
and the insertion velocity increment must be this, minus the orbital velocity around planet
2:
µ2 µS 2r1 2 µ2
!V2 = 2 + (1 " ) "
rc2 r2 r1 + r2 rc2

Comparison to simple Escape+Transfer+Capture.

A simple-minded approach to the same mission would be to first apply an impulse at the
surface of planet 1 to achieve escape ( !Vesc,1 = 2 µ1 / R1 ), then, after slowing down to
zero velocity with respect to planet 1, apply a second impulse to enter the elliptic transfer
orbit towards planet 2 (this would be our vrel ,1 ), then, in the vicinity (but still outside the
SOI of ) planet 2, apply a third impulse to match the heliocentric velocity of planet 2 (this
would be our vrel ,2 ), and finally, starting from zero relative velocity “far” from planet 2,
apply a fourth impulse to capture the craft into orbit about planet 2 (this is equal to the
escape velocity from a distance R2 to the planet, !Vesc,2 = 2 µ2 / R2 ). You can easily
check that the two impulses we derived before are, respectively,

!V1 = (!Vesc,1 )2 + (vrel,1 )2

!V2 = (!Vesc,2 )2 + (vrel,2 )2

and so our previous scheme is definitely more effective. These two strategies are called
sometimes the Hohmann (simple, four impulses) and the Oberth (combined, two
impulses) maneuvers.

4
16.50 Lecture 5

Subjects: Non-Chemical rockets; Optimum exhaust velocity

1) Non-chemical rockets

A shared characteristic of all non-chemical propulsion systems is that the energy and
propellant mass are separate initially

Chemical

Chemical .
Energy mc
mass

Non-chemical
. 2
m c
Energy 2
Source
.
mc
Propulsive
Mass .
m

There are several possible energy sources:

1) Solar
a) Photovoltaic
b) Solar thermal
c) Solar pressure

2) Nuclear
a) Fission
b) Radioisotope
c) Fusion?

There are also many ways to bring the mass and energy together to produce thrust, but all
behave according to the rocket equation.
"V
m final !
=e c
mtot
or breaking the final mass into its constituent parts,

1
 m pay !
"V m propsys + mstruct
=e c !
mtot mtot
There are 2 general categories of systems, Thermal and Electrical, separable according to
whether the energy is available in electrical or mechanical form, or only as thermal
energy at some limiting temperature.

A. Thermal

Here the energy is used directly to heat the propellant, which is then expanded
through a nozzle to produce thrust.

Energy Tc

Propellant C

Now there is a chamber temperature Tc, limited by the energy source, and the exhaust
velocity is given approximately by:
c2 2" R
c pTc = or c ! 2 c pTc ! T
2 " #1 M c

What limits Tc?

1) Source Temperature e.g. Tsun = 6000°K

2) Materials

So for these we generally want low M, e.g., H2 ! 2H. For a nuclear thermal rocket
Tc ~
< 3000°K

c < 8500 m/s

H2

graphite
F
Figures of merit are: a) Specific impulse, c; b) Thrust per unit mass,
mengg

2
B. Electrical

If the energy is available in electrical form, then there is no limit in principle on c (other
than the speed of light) and in practice we can achieve very high c with good efficiency
by using any of a number of electrical accelerators.

The system requirement is to produce a ΔV on a payload, mpay. In the absence of gravity

"V
m final !
loss, = e c , so why not make it very close to 1 by increasing c? To see the answer
mtot
we must analyze the whole system, and take account of the mass of the energy source.

P
m elect
.
mpay m accel m, c
m
propellant .
m
The total mass of the system can be broken out as:
mtot = mpay + melect + mprop + meng + mstruct
so that ratio of final mass to initial mass is
m final !
"V
m + melect + meng + mstruct
= e c = pay
mtot mtot
The Figure of Merit for such a system is
m pay !
"V
m + meng + mstruct
= e c ! elect
mtot mtot
Let us neglect meng + mstruct for the moment, compared to melect (or simply redefine mpay to
include meng + mstruct, which makes sense for some missions).

We know that:
F = m˙ c
and the power P is
c 2 Fc
P ! m˙ =
2 2
melect F
Define a specific weight ! e " , and an initial acceleration ao = . Then
P mtot
"V "V
m pay ! #P ! # Fc
=e c ! e =e c ! e
mtot mtot 2mtot
or finally in terms of the minimum number of dimensionless parameters,
"V
m pay ! # a "V c
= e c ! ( e o )( ) (1)
mtot 2 "V

3
 ! eao "V
Here the group ( ) is determined by technology level (! e ) , the mission
2
requirement (ΔV) and how fast we want to achieve it (ao). So we should consider this
relation a way to find copt to maximize mpay/mtot , given ΔV, ao and αe.

Differentiating,

m pay
!( ) "V
mtot # $ a "V c 2
= # e c + ( o )( ) =0
"V 2 "V
!( )
c

!V 2 "( !Vc )opt # a !V

( )opt e =( o ) (2)
c 2
which we must solve for the optimum c/ΔV). For graphical presentation, let us eliminate the
! a "V
group ( o ) between (2) and (1):
2
"V
m pay ! "V
= e c (1 ! ) (3)
mtot c
Equations (3) and (2) are represented below over a broad range of ΔV/c:

4
 !V c
So we see that it only makes sense to choose < 1 or > 1 for such systems. This is
c !V
!V
because for > 1 the exponential is so small it outweighs the term representing melect.
c
!V
Expanding the range 0 < ( )opt < 1,
c

Let us take a look at the meaning of these results:

!V
1) If we choose ao and have given αe and ΔV, this gives us the ( )opt , and in turn the
c
m pay
maximum .
mtot
!V
2) For given ΔV and αe, increasing ao (for a faster mission) increases ( )opt , which reduces
c
m pay
.
mtot
m pay
3) For given ΔV and a0 , reducing αe (lighter power plant) increases .
mtot
Take an example:

Suppose that the mission gives as a requirement ΔV = 104 m/s and technology enables αe =
melect
= 20 kg/kW = 0.020 kg/W
P

5
 ! eao "V
Then = 100 ao where ao is in m/s2.
2

We can still choose how fast we want to do the mission, within limits. We know that the
! a "V
upper limit of e o = 1/e= .368. So for the assumed mission and technology,
2
! a "V
100ao= e o !.368. This implies ao !.00368 m/s2 or 3.8x10-4 g's and for this maximum
2
m pay m pay
available acceleration, = 0, not a very useful result! Suppose we insist on = 0.5.
mtot mtot
!v ! a "v
This gives ( )opt = .3, which in turn implies e o = .07. The acceleration is then ao =
c 2
.07
= 7x10-4 m/s2 = 7x10-5 g's. This is only about 1/5 the maximum acceleration, but now
100
we have lots of payload.

The time required to achieve the ΔV is

"V
! "V
m propellant m0 (1! e c ) c !
t= = = (1 ! e c )
m! F /c a0
and for our example,
10 4 / 0.3
t= (1" e"0.3 ) ! 1.23 ! 10 7 s. = 142 days
7 ! 10 "4
This type of propulsion requires patience! Note that for a H2, 02 rocket,
#V 1 04
m pay " m "
!e c " struct
!e 4500
".1 !.0084
mtot mtot
So we would probably use 2 stages. But the main point is the very much smaller payload to
total mass ratio of the chemical system. In addition, if the coasting period for a chemical
rocket is very long, as in an interplanetary transfer, a continuous low thrust can in many cases
accelerate the mission.

6
16.50 Lecture 6

Subject: Modeling of Thermal Rocket Engines; Nozzle flow; Control of mass flow

Though conceptually simple, a rocket engine is in fact physically a very complex device and
difficult to represent quantitatively by mathematical models. But the difficulty and high cost
of arriving at a successful design makes it essential that the performance-limiting
phenomena be described as accurately as possible. It is no longer feasible to approach the
development of a new rocket engine by a purely cut-and-try method.

The range of phemomena that must be dealt with is suggested by the sketch. They range
from the fluid dynamics of the pumping system through the very complex phenomena of
combustion to structures with heat transfer rates beyond those experienced anywhere other
than in reentry from space.

Hot gas, T c p c 3D Supersonic Flow

Pump p
e
u
e
Fuel Oxidizer

q
Heat Transfer
Thrust Chamber
mixing and combustion

To deal with this complex situation, we use models that represent adequately the phenomena
we want to deal with, while suppressing other phenomena. Some models we will find useful
are those for:

1) Turbopumps
2) Injectors
3) Combus ion
4) Nozzle Flow
5) Heat Transfer
6) Structure of combustion chambers and nozzles

1
Nozzle Flow

Let us focus on the Nozzle Flow first. For this purpose we regard the combustor as just a
source of hot gas of known chemical composition, stagnation temperature Tc and stagnation
pressure pc.

We have already used a very simple model to represent the result of this flow process.
Assuming that the expansion was to zero pressure and therefore zero temperature we used
conservation of energy to show that
!
c = 2cp Tc = 2 RT (model 1)
! "1 c
where c was defined as the "exhaust velocity" meaning the velocity which, multiplied by the
mass flow rate, gives the thrust. Now we want to explore the fluid mechanical processes in
more detail, and also represent the characteristics of the gas more accurately.

The gas is in fact a mixture of gases resulting from the combustion process Its composition is
specified in terms of mole fractions yi of species i or mass fractions xi of species i.

For example if the rocket burns H2 and O2 in a mixture ratio of 3 moles of H2 to 1 mole of
O2 to produce H20 and H2, (by mass O/F=32/(3x2)=5.3, comparable to O/F=6 in the Shuttle
SSME engine),

3H2 + 0 2 ! 2H2 0 + H2

then the mole fractions and mass fractions of H20 and H2 are as shown in the table.
H2 H2 0
x 1/19 18/19
y 1/3 2/3
With these mole or mass fractions we can compute the properties of the gas mixture, such as
the specific heats, the gas constant and γ.

The fluid-mechanical and heat transfer phenomena that are of importance to describe are:

1) The effects of nozzle shape on flow

2) The effects of chemical reactions during the flow
3) The heat transfer from the flow to the nozzle

2
Effect of Nozzle Shape

The phenomena that influence the coupling between nozzle shape and the flow can be
divided into two general classes, first those associated with the gas dynamics and second
those associated with the properties of the gas. Depending on how accurately we attempt to
model each of these, our model of the flow can vary enormously in complexity. The range of
possibilities is shown schematically in the figure, where sophistication of the modeling
increases away from the origin along two axes, one measuring gas dynamics, the other fluid
properties.

MOC with
ideal gas CFD+Combustion

Dusty flow
Frozen, but cp(T)

One-D reacting gas

Along the gas dynamic axis the description can range from the simple channel flow model
that we will discuss here to very sophisticated three-dimensional descriptions of the transonic
flow. Similarly along the properties axis the range is from the thermally and calorically
perfect gas model we will use (often called the "ideal gas") to full kinetic description of the
effects of chemical reactions that occur as the gas flows through the nozzle. The simplest
model of all (Model 1) equates the kinetic energy of the exhaust flow to the thermal energy
in the chamber

Here we will limit the discussion of the coupling of flow to the geometry of the nozzle to a
simple model (Model 2) at the origin. The implications of the more complex phenomena will
then be discussed separately through two more models. Model 5 will deal with the effects of
chemical reactions, and Model 3 with the effects of nozzle shape other than the simple area
variation. Model 4 will represent some special chemical and fluid mechanical aspects of solid
rocket motors.

3
Model 2: Channel Flow of Perfect Gas

The channel flow model (often erroneously referred to as "one-dimensional flow") is one in
which the entire flow is considered to be in a single stream tube having a mass flow that is
constant along the axis of the stream tube. The stream tube is considered to be narrow
enough that the angle of the walls to the axis may be considered small, the gas properties
nearly constant across the stream tube, and the components of velocity normal to the axis
small enough that their squares are negligible compared to that of the streamwise velocity.

If the cross-sectional area of the streamtube is denoted A(x), the conservation of mass
becomes
.
ρuA(x) = m = constant (1)

The conservation of energy becomes

u2 v2
h+ = hc (Note: 2 <<1)
2 u
or if the specific heat is assumed constant.
u2
c pT + = cp Tc (2)
2
The final relationship defining the flow is a connection between two of the thermodynamic
properties of the fluid. In general this is a description of the variation of the entropy of the
fluid along the flow direction. For our simple model we assume that the flow is Isentropic
(Adiabatic and Reversible). The entropy is referred to a reference state which we take to be
the chamber condition. Conventionally, we assume that the velocity in the chamber is low
enough that this is also the stagnation condition in the chamber.

T dT p
s ! sc = " c p ! R !n = 0
Tc T pc
so if s is constant, and cp is as well,
T p
c p !n ! R !n = 0
Tc pc
and from this it follows that
R ! "1
T p C p !
= ( ) p =( )
Tc pc pc (3)

where
cp
! " ; cp-cv=R
cv
We want to manipulate these three relationships to arrive at an expression giving the
variation of Mach number with A, hence with x. We begin by expressing equation (2 )in the
form:
Tc u2 ! "1 2 u2
=1+ =1+ M , M2 !
T 2c p T 2 " RT

4
and then from (3)
!
pc ! " 1 2 ! "1
= (1 + M )
p 2
With these we can write (1) in the form,

. p p Tc 1 T pc
m = !uA = uA = M "RTc
RT pc T R Tc Tc

! 1
! " 1 2 " ! "1+ 2 pc
= (1 + M ) M ! Tc / RA
2 Tc

.
m ! M
= pc ! +1
A RTc ! " 1 2 2( ! "1)
(1 + M )
2

For given (and constant) pc and Tc, this expression gives the mass flow per unit area as a
function of the Mach number M. Since the mass flow is constant, we have the desired
expression for M as a function of A(x).

We see that the mass flow per unit area, ρu, is proportional to the stagnation pressure and
inversely proportional to the square root of the stagnation temperature.

It is both conventional and useful to plot the function of Mach number in ratio to its value at
the point in the flow where M=1. The value of ρu at this point we call ρu*. It is given by
" +1
" # " + 1 % ' 2 ( " '1)
(!u)* = pc
RT C $ 2 &
Then the ratio is
" +1
2( " #1)
$ " +1 '
!u A* & 2 )
= = M&
!u * A " #1 2 )
&1 + M )
% 2 (
This function of M is shown in the figure.

5
Some of the dominant features are:

1) There is a maximum in the mass flow per unit area at M=1.

2) For M<1, ρu (and also u) increases with M, so M increases with decreasing A.

3 For M>1, ρu (and also u) decreases with M, so M increases with increasing A.

We note however that p always decreases with increasing M. It follows that if we want to
increase M from a small value to one in excess of 1, as we do in a rocket nozzle, the nozzle
should be shaped as in the sketch at the beginning of the lecture, with a convergence to a
minimum area "throat", followed by a divergence.

For given pc and Tc, the area of the throat controls the mass flow.

To get an idea of the required throat size we can evaluate the (dimensional) factor in front of
the function of M for a set of typical values.

Say pc = 100 atm = 107 N/m2

Tc = 3000°K
! = 1.2
8.32x10 3
R= = 640 J/kg/K
13

! 1.2
pc = 10 7 = 7.91x10 3 kg /m 2 /s
RTc (640)(3000)

We define the throat of the nozzle as the point of minimum area, where M=1 and call the
area there At. Then

6
 . ! +1
m ! 2 2( ! "1)
= ( )
pc At RTc ! + 1)

The conclusion that the point of minimum area in the nozzle controls the mass flow is valid
beyond the limits of our simple ideal gas mode.. More generally, we define a quantity c* by
. pA
m ! c t that is a measure of the mass flow per unit area at the throat. For the ideal gas
c*
model,
! +1
RTc ! + 1 2(! "1)
c* = ( )
! 2
For the example above,
(640)3000 2.2 5.5
c* = ( ) ! 2138m / s
1.2 2
In general c* depends only on the propellant combination and pc.

One further condensation of these formulas that is sometimes useful is to define the
" +1
2 2(" #1)
quantity ! = " ( ) , which is close to 0.65 for most values of γ, and then write
" +1
RTc pA
c* = ; m˙ = c * t
! c

7
16.50 Lecture 7

Subject: Modeling of rocket nozzles; effects of nozzle area ratio.

In the last lecture we saw how the throat area of the nozzle controls the mass flow rate. Now
we will explore the effects of the shape of the nozzle downstream of the throat.

The Mach number and hence velocity at any point in the nozzle is determined by the ratio of
the area of the stream tube to the area of the throat (the area ratio), so if we assume that

a) The nozzle flows full, i.e. the streamtube shape matches the
shape of the nozzle
b) The flow is supersonic to the point x downstream of the throat,

then we can find M(x) from A(x)/At and it in turn determines p(x), T(x) and u(x).

As we shall see, it is the pressure that determines whether the nozzle flows full, so it is
convenient to relate the exit velocity directly to the pressure. We do this from the Energy
Equation,
u2
Tc ! T +
2c p
T
u 2 = 2c p (Tc ! T ) = 2c pTc [1 ! ]
Tc
So at any point in the nozzle where s=sc
" !1
p " 12
u = 2c pTc [1! ( ) ]
pc
and this is independent of whether the nozzle is full at this pressure, because the area is not
referred to. In particular, if we apply this at the end of the nozzle where the pressure is pe,
the velocity at that point is
" !1
pe " 12
ue = 2c pTc [1! ( ) ]
pc
You know from your previous work that if the nozzle flows full the thrust of the rocket can
be written
.
F = m ue + Ae ( pe ! po )

substituting the expression for ue we then have the following expression for F in terms of the
pressure ratio pe/p0.
" !1
. pe " 12
F = m! 2c pTc [1! ( ) ] + Ae ( pe ! po )
pc
The mass flow rate is given in terms of c* by
. pA
m! c t
c*

1
and for ideal gases, we have the estimate of c* given by

" +1
RTc 2 2(" #1)
c =
*
; != "( )
! " +1
F
Substituting these into the expression for F and non-dimensionalizing it:

! +1
! "1
F 2 ! 2 #% 2 & ! "1 p A # p " po &(
= [1 " ( e ) ! ] + e % e
pc At ! " 1 $ ! + 1' pc At $ pc '
where the area ratio is itself related to the pressure ratio through continuity:

" +1
Ae ! t ut " #1 2 2(" #1) 1
= = ( )
At ! e ue 2 " +1 p p
" #1

( e )1/ " 1# ( e ) "

pc pc
More generally, we define a Thrust Coefficient by

F ! pc At cF

and the above expression then gives us an estimate of cF for ideal gases.

The dependence of cF on nozzle area ratio and pressure ratio is conventionally

summarized as in the figure below, which is drawn for γ=1.2 and 1.3. Notice:

-γ is replaced by k in these graphs

-pc is replaced by p1
-pe is replaced by p2
-pa, or p0, the ambient pressure, is replaced by p3

2
 2.0
k = 1.20 lue=2.246
um va
Maxim
=

8
p 1lp 3 1000
1.8
Line of optimum
thrust coefficient 500
333
P2=P3 p1 lp
3 = 20
100 0
1.6
50.
0
33.3

1.4 20
.0
CF

10.0
1.2
Region of flow separation
for conical and bell
5.
0 shaped nozzles
1.0
3.
3
2.
5
0.8
2.
0

0.6
1 2 4 6 8 10 20 40 60 80 100

Area ratio ε = A2 lAt

Thrust coefficient CF versus nozzle area ratio for k = 1.20.

Image by MIT OpenCourseWare. After Sutton and Biblarz 2001.

2.0
k = 1.30
lue=1.96 4
Line of optimum Maximum va
1.8 p1lp3 = 8
thrust coefficient 1000
P2=P3 500
p1 lp
200 3 =
1.5 333
100
50
33.3
1.4 20.
0
CF

10.
0
1.2

5.0 Region of flow separation

for conical and bell shaped
1.0 nozzles
3.
3

2.
5
0.8
2.0

0.6
1 2 4 6 8 10 20 40 60 80 100

Area ratio ε = A2 lAt

Thrust coefficient CF versus nozzle area ratio for k = 1.30.

Image by MIT OpenCourseWare. After Sutton and Biblarz 2001.

3
The thrust is also expressible in terms of the effective exhaust velocity c, as we discussed in
Lecture 1. Then we have
pA
F ! m! c = c t = pc At cF
c*
and we see that
c = c * cF

The "Characteristic Velocity" c* depends mainly on propellant properties. The Thrust

Coefficient cF depends on propellant properties through γ (or equivalent), but mainly on the
pressure ratio and nozzle geometry. So by the definitions of cF and c* we have managed to
separate the effects of propellant properties and the effects of nozzle geometry into the two
factors of c. This separation, though demonstrated only for ideal gases, holds also for the
more general situation of complex chemically reacting propellants.

Effects of Non-ideal Expansion

So far we have assumed in the discussion of cF that the flow fills the nozzle and is supersonic
to the exit. If this is the case, the thrust is given by the above expression for cF . But in fact
the flow can be somewhat more complex than this, depending the ratio pc/p0 compared to
that which leads to ideal expansion. This behavior is summarized in the figure below.

Kerrebrock, Jack L. (1992). Aircraft Engines and Gas Turbines (2nd Edition).
MIT Press, © Massachusetts Institute of Technology. Used with permission.

Here An is the throat area, Ae the exit area. The ideally expanded situation is at the upper
right, and it is the condition at which thrust is maximum for a given external pressure p0. This
is easier to visualize than to prove analytically: if the divergent nozzle were extended a little
by adding a section at its exit, this section would see an internal pressure lower than p0, and
would therefore generate suction, or negative thrust. If on the contrary, the nozzle were
shortened a little, one would lose the positive thrust that was being produced by the removed
portion.

Returning to consideration of a given nozzle, if p0 is lowered below its ideal matching

pressure (for example by the rocket ascending in the atmosphere) the nozzle becomes
underexpanded, as at the lower right. In this case the flow fills the nozzle and our formula
for F works fine.
4
If on the other hand p0 is larger than corresponds to ideal expansion, the situation can
become more complex. For pressure ratios p0/ pe< 2 to 2.5, the nozzle is likely to remain
full, and again the formula holds. But for larger pressure ratios the oblique shocks that form
at the exit of the nozzle are strong enough to separate the boundary layer, and the point of
separation moves into the nozzle so that its effective area decreases, as shown at the upper
left. In this case the separation occurs approximately at a pressure ps such that ps/p0 = ½ to
1/(2.5).

Again we can use the formula for thrust by replacing pe by ps and Ae by As, the area at which
the pressure is ps. So now
.
F = m us + As (ps ! po )

With this understanding, we can write the thrust coefficient for the separated nozzle as

! +1
! "1
F 2 ! 2 %# 2 & ! "1 p ! As %# ps " po &(
= [1 " ( s ) ]+
pc At ! " 1 $ ! + 1' pc At $ pc '

where now it is understood that ps is the pressure at which the separation occurs and As is
the corresponding area.

5
16.50 Lecture 8

Subjects: Types of Nozzles; Connection of flow to nozzle shape.

Types of Nozzles

The axisymmetric convergent-divergent "bell" nozzle that has been used as the example to
this point is the standard for rocket nozzles, for several reasons:

1) Structural - It has essentially only "hoop" or tangential

stresses which are the easiest to design for.
2) Cooling - It can be constructed with walls of simple tubular
construction that enables the cooling in a straightforward way.
3) Matching to combustor - It is easy to match to the
combustor, which is most naturally a simple cylinder.

But it has some disadvantages, which stem from the need to operate at overexpanded
conditions at low altitude (specifically, the nozzle is overexpanded, not the flow). The flow
situation shown as the left diagrams in the last lecture implies:

4) An overexpansion thrust loss, whether or not separation occurs.

5) Flow instability when overexpanded with separating flow, which may lead to
uncertainty or unsteadiness of the thrust direction.

These have led to variations on the basic bell nozzle design, and some radically different
designs. Amongst the variations are:

a) Extendable nozzles. Here the idea is to use the short inner nozzle at low altitudes, and
deploy the outer extension at high altitudes, so that the expansion ratio is more nearly ideal at
all operating conditions.

1
b) The External Expansion nozzle, termed by Rocketdyne the "Aerospike". Schematically, it
operates with the flow configuration shown in the top figure at design, with the outer
streamtube essentially parallel to the axis and at atmospheric pressure.

M0

(Me)design

L
C

M0

Me< (Me)design

L
C

Image by MIT OpenCourseWare.

At higher back pressures, that is at lower altitudes than the design value the bounding
streamline adjusts by moving inward as shown in the lower figure, and there is no problem
with suction, and no separation as encountered in the bell nozzle at high external pressure
conditions. In detail, the flow now has a complex structure, with repeated oblique shocks and
expansions, but this has little effect on performance.

Advantages cited for this type of nozzle are:

1) Better off-design performance

2) Ability to use modular combustors around a nozzle spike, as suggested by the view
below, looking upstream at the spike.

Disadvantages are:

3) It is difficult to cool the spike.

4) It tends to be heavy, because the structure is not based on simple cylindrical
pressure vessels as in the bell nozzle.

2
Note - Such a nozzle was originally planned for the DC-X, but was not used because of
weight and development cost. It was included in the Rockwell concept for the X-33, which
was abandoned for unrelated structural reasons.

Connection of flow to nozzle shape

Our channel flow model connects the flow Mach number to the area of the flow passage, but
it gives no information as to how the area should vary along the axis. We will now consider
another model for the flow, called in its more fully developed form the Method of
Characteristics, or MOC. This is our Model 3, that provides a basis for quantifying the
connection between the nozzle shape and the flow. For this purpose we will consider the
flow to be in two dimensions, supersonic and isentropic. The same procedures can be
extended to axisymmetric flow, but the mathematics are somewhat more involved. For
references see the following:

Kuethe & Chow, pp 219-228

Hill & Petersen, pp 523-530

The expansion of the flow in the supersonic nozzle occurs primarily through weak,
isentropic waves. Such a wave in a supersonic flow stands at an angle to the flow, and
turns that flow through an angle d as shown in the diagram.
Mach wave

 a
V V du
V

d dv
dV

a 1 1
  sin 1  sin 1  tan 1
M 1
2
V M
1
tan  
M2  1

Associated with this turning is a change in total velocity dV and changes in velocity du and
dv in the original direction and perpendicular to it. For d small,

dv  Vd , du  dV

From the geometry of the diagram,

du
 tan 
dv
and substituting the above values for du and dv,

3
 du dV 1
= =
dv Vd! M2 " 1
It follows that we can relate V to θ and M
dV d!
=
V M2 " 1

From the energy equation we can relate V to M.

T 1 dT "1
= # = (! " 1)MdM
T0 1 + ! " 1 !
T 1 + " 1 M2
M2
2 2
2
V dM dV dT
M =
2
#2 =2 "
! RT M V T
" !1 2
M
dV dM 1 dT dM 2 dM
= + = ! " ! 1
V M 2 T M 1+ M2 M
2
dM 1 d#
[ ! 1 ] =
M 1+ " M2 M2 " 1
2
Rearranging, and counting angles from the sonic point, we have an expression for θ in terms
of M (the so-called Prandtl-Meyer function):
M M 2 ! 1 dM
θ(M) = "1 r !1 2 M
1+ M
2
! +1 "1 ! " 1 "1
(M " 1) " Tan M " 1
2 2
= Tan
! "1 ! +1

This function is tabulated in Kuethe & Chow and Liepmann & Roshko. It is plotted below:

4
There is a maximum angle (equal to 130.5 degrees for γ=1.4, and 208.5 degrees for γ = 1.2).

Nozzle Shaping

How do we shape a nozzle using this information? Let us assume γ=1.4 so we can use the
tables, and suppose we want to design a nozzle for (pc/pe)design = 100.
1) Start with Channel Flow to get the exit Mach number and area ratio. For γ = 1.4,
Me is approximately 3.7, and Ae/At is 8.2. Suppose we want a nozzle that produces this
Mach number at the exit, and that the flow should be approximately parallel to the axis. The
nozzle should therefore look something like:

CL
M0
Me
M1

θ1

Image by MIT OpenCourseWare.

For simplicity, let us assume the flow turns around a corner downstream of the throat, where
the Mach number is M0 > 1, to an angle θ1, and let us indicate just three of an infinite number
of waves making up the "expansion fan". Since the centerline is an axis of symmetry, the
waves must reflect from it while canceling the flow deviation, so that the flow downstream of
the reflection is axial. When the flow passes through these reflected waves it is deflected
toward the axis. If the wall turns to match this deflection where the reflected wave hits it, the
wave is just cancelled. The net result is that the flow arrives at the exit with an axial
direction and a uniform Mach number. Because of the crossing of waves from two
expansion fans, the detailed flow properties at most points need to be computed step-by-step.
For a good discussion of this see Liepmann & Roshko. Unfortunately time does not permit it
here, except for the uniform region just downstream of the sharp turn, where things are
simple.

Since the flow is isentropically turned by the corner to the angle θ1, M1 is determined by this
angle, from ! (M1 ) " ! (M 0 ) = !1. Also, since the flow is again turned by the second set of
waves through the same (but opposite) angle, ! (M e ) " ! (M1 ) = !1. By addition,
2!1 = ! (M e ) " ! (M 0 )

5
which determines θ1. After this, the Mach number M1 follows directly from either of the
above formulas. Of course, if M0=1, θ( M0)=0, and the results simplify further.

To get the geometry correct we have to take account of the actual shape of the nozzle throat.
More important, this whole argument has been in two dimensions and real rocket nozzles are
almost always axisymmetric, but again time does not permit discussing these matters.

This sort of calculation results in a nozzle that will produce an axial exit flow of uniform
Mach number for a given expansion ratio. The length can be excessive, however, and non-
idealities tend to cancel the small thrust contributions of the far downstream sections.
Methods have been developed to generate shorter nozzles while accepting some losses due to
non-uniformities. The so-called “Rao nozzle” is the nozzle with the optimal contour for a
given (less than ideal length).

These same ideas can by used to understand the external expansion, or "plug" nozzle. It looks
like:

M0

Me

L
C

Image by MIT OpenCourseWare.

In this case each of the expansion waves from the outer "lip" is cancelled by a turn in the
contour of the center body, so that there is just one family of expansion waves. This can be
either a two-dimensional nozzle or an axi-symmetric one. It is the basis for the Aerospike
nozzle projected for use in the X-33 by Lockheed Martin. It has also been used for aircraft,
eg. the F-4, in which the underside of the aft fuselage provides the expansion surface. Such a
nozzle is also a key ingredient of the SCRAMJET idea, as exemplified by the X-30 and later
implementations.

6
16.50 Lecture 9

Subject: Solid Propellant Gas Generators; Stability; Grain designs

We have thus far discussed two models for the nozzle flow in rocket engines, the Channel
Flow Model and the Two Dimensional Isentropic Model. Now we will introduce a model for
the source of the hot gases in Solid Propellant Rockets.

Gas Generators

The distinguishing feature of a solid propellant rocket is that the fuel and oxidizer are pre-
mixed. Typically, a modern propellant consists of:
polybutadiene 14%
aluminum 16%
ammonium perchlorate 70%

It is prepared as a physical mixture of NH4 ClO4 grains and aluminum powder in the plastic
matrix

The mixture reacts at the surface of the propellant grain to produce hot gas. The surface
regresses( i.e. decomposes and evaporates) at a rate:
.
r = apc
n

where a and n are empirical coefficients. This is an empirical rule. There is no

straightforward way to derive it from first principles. The physical processes involved are
quite complex, including heat transfer to the grain by conduction, convection and radiation,
decomposition of the solids, mixing and finally combustion in the gas phase.

Typical values of the empirical constants are:

r = 6.8x10-3 @ 6.9x106 N/m2
n = 0.15
r = 1,775 kg/m3
and as we shall discuss, such a propellant gives a specific impulse

Is = 260 - 265 sec (at ground)

Is =280 – 295 sec (in vacuum)

1
Once the rocket is built and ignited, it has a mind of its own. That is pc, Tc, r and the mass
flow are all set by the propellant properties and geometryof the grain. For design purposes
we want to find the relationship between these performance parameters and the geometry.

First assume the rocket is operating in steady state, then

pA
m˙ p = rAb ! p = c t
c*
and using the above expression for the regression rate,
pA
apc Ab ! p = c t
n

c*
where Ab is the "burning area", ie the surface area of the grain that is regressing. Solving for
pc gives
A
pc = ac * " p b
1!n
(1)
At
So given the propellant properties we can compute the chamber pressure, and the mass flow
of the propellant. Combined with a nozzle model, this enables us to compute the thrust of the
rocket and its specific impulse

Stability

This relationship applies if the operating point is stable, that is if the pressure and mass flow
are steady in time. If a, n, c*, ρp are all well defined, we can get pc from it. Note however
that pc depends on Ab and that it can be sensitive to unintended variations in the grain
configuration..
Crack
Ab

Tc
r pc C

Vc
At
L

Image by MIT OpenCourseWare.

Cracking of the grain or its debonding from the case can lead to increases in the burning area,
very high pc and an "explosion". This is probably the most frequent cause of solid rocket
failures.

We have assumed that there is a steady pc. To determine whether this is a stable operating
point, we first construct a model that allows unsteady behavior of pc. In first approximation,
unsteadiness implies a variation in the mass stored in the chamber, which modifies the mass
balance we used to determine pc.

d . pA
(!cV c ) = r Ab ! p " c t
dt c*

2
 pc At
= apc Ab! p "
n

c*
d dVc 1 dpc
( !cVc ) = !c + Vc ; Tc " const
dt dt RTc dt
. 1 dpc
= ! c r Ab + V c
RTc dt
V dpc
= ! capc Ab + c
n

RTc dt
Substituting above, and noting that ! p >> ! c ,
Vc dpc pA pA
= ( ! p " !c )Ab apcn " c * t # ! p Ab apcn " c * t
RTc dt c c
Now suppose the chamber pressure consists of a steady part plus one that varies in time,

pc = pco + δpc(t) where δpc<<pco

Vc d!pc ( p + !pc )At

= a( pco + !pc ) n Ab " p # co
RTc dt c*

!p c n p A !p
= apco (1 + ) Ab " p # co t (1 + c )
n

pco c* pco

"p c p A "p
! apco (1 + n )Ab # p $ co t (1 + c )
n

pco c* pco
The zero’th order terms on the right hand side cancel out, leaving

Vc d!pc !p p A !p
= " p apco Ab n( c ) # co t ( c )
n

RTc dt pco c * pco

= # apco
At %
Ab n" p ! 'p
n!1

$ c *& c

1 d!pc RTc A
= [apco Ab n# p " t ]
n"1

! pc dt Vc c*

So we have divergence if the quantity in brackets >0, stability if it is <0. Simplifing

the expression by use of Eq. (1):
A aAb n A A
apcon!1 Ab n " p ! t = " p ! t = (n ! 1) t
c* A
ac * " p b c * c*
At

1 d!pc RT A
= " c t (1" n)
!pc dt Vc c *

3
So we have:
stability for n < 1
instability for n > 1
Vc c *
The time scale for growth is , and using ! c = pc /(RTc ) and m! = pc At / c* ,
(1 ! n ) RTc At
1 # cVc t residence
t growth ! !
1" n m˙ 1" n

For typical values, this time is about 1 millisecond, which is of course short compared to
most rocket burning times, so we conclude that if n<1 the chamber pressure will relax from a
perturbation in a time of order of 1 millisecond. If n>1 the rocket will explode in about the
same time.

Grain Designs

The burning time is set by mission requirements, in particular by the acceleration level of the
vehicle. For example for an acceleration of 2 "g's" and a propellant fraction of 0.7,

c m(t) 2500
t= [1 ! ]" [.7] " 89sec
ao m(o) 2(9.8)
The regression rate as quoted above is about 0.7 cm/s, so in the direction normal to the
burning the thickness of the grain must be about 56 cm.

Grain Designs

The geometry of the grain must be chosen so as to meet the above requirements. The
simplest grain configuration is what is termed "end burning".

a) End burning:

Rubber insulation
Ab

r C

Image by MIT OpenCourseWare.

Here the grain regresses axially across the end facing the nozzle. The outer (cylindrical)
surfaces of the grain are generally bonded to the case with an intermediate layer of insulating
material such a rubber and coated with a substance that inhibits decomposition. As the grain
recesses it leaves the layer of rubber, which protects the case from the hot combustion gases.
This type of grain works well if the size of the grain needed for the application results in a
length that matches the desired burning time. But if the two do not match then a design must

4
be found that gives the right burning length for the particular grain mass that is required for
the mission.

b) Radial burning:

A configuration that shortens the burning time for a given grain mass is the "radial burning"
grain:

This configuration has the disadvantage that the burning area increases as the grain regresses,
resulting in an increase of chamber pressure with time, a situation that is undesirable for three
reasons. First, the case must be designed structurally for the maximum pressure and it is used
inefficiently for much of the burn time. Second, at least for launch systems, we want a high
chamber pressure at the beginning of the burn to provide a high nozzle pressure ratio in the
presence of the high atmospheric pressure. And finally, if the thrust increases with time it
exacerbates the increase in acceleration due to the decreasing mass of the vehicle that results
from the propellant consumption, and results in very high g loads at the end of burning.

To get around these difficulties, many large rocket engines use a "star" grain
configuration that has the objective of maintaining the burning area constant as the grain
regresses, or even decreasing it.

c) "Star" grain

These grains have a series of points protruding inward, as shown in the sketch, such that as
the points burn off, they keep the area roughly constant. In a first approximation one can see
that the periphery of the "star" should be equal to the outer(circular) periphery of the grain, so
that the burning area is equal at beginning and end. Detailed geometric constructions have
been developed that keep the area very nearly constant throughout the burn.

5
c) Segmented Grains

The normal process for manufacturing solid propellant grains consists of mixing the
ingredients in a batch process (essentially a big food mixer), then pouring it into the case,
where the rubber matrix cures. Because of limits on the size of the mixer and for safety and
transportation reasons, the amount that can be poured into a single case is limited. So for
very large rocket motors such as the Space Shuttle Solid Rocket Boosters, the grain is made
up of several axial "segments" that are poured separately then assembled to make the
complete motor. It was a failure of one of the case joints that led to the Challenger accident.

6
16.50 Lecture 10

Subjects: Models for rocket engines; Flow of reacting gases

Models for Rocket Engines

In Lecture 6 we described in general terms a set of models we might use to describe the
various features of rocket engines, making the point that no one model incorporates all the
features. The models we have discussed thus far are:

Model 1;
c2
= c p Tc
2
which simply equates the kinetic energy of the exhaust to the thermal energy in the
chamber, ignoring all questions of efficiency and gas dynamics.

Model 2;, which describes the nozzle flow for cp and γ constant, enabling an
approximate description of the effects of the pressure expansion ratio and of the gas
properties. This resulted in the definitions of the Thrust Coefficient cF, and
Characteristic Velocity c*, which are applicable for more general situations than this
model represents. The model in turn gave us approximate formulae for estimating cF
and c* for particular nozzle geometries and propellants.

Model 3, which describes the behavior of the flow in terms of the actual geometry of
the nozzle, but within the simplification of ideal gas behavior.

Model 4, which dealt with Solid Propellant Internal Ballistics. This models the gas
formation process for solid rockets, combined with the most important fluid
mechanical phenomena governing the burning rate and stability.

In all of these models, the properties of the gases in the combustion chamber were taken as
given. It should be clear by now that this is a key issue in understanding rocket engines. So
we need a model that connects the properties of the gases in the chamber to the propellant
characteristics, i.e. a model that describes the combustion process. This set of processes we
term Combustion Thermochemistry. We will treat it as Model 5.

Model 5 - Combustion Thermochemistry

In this model we want to represent what happens in the combustor, to convert liquid
propellants (one or two) to hot gases at Tc and pc. Qualitatively, what happens is that the
difference between the chemical energies of the reactants (the chemicals injected into the
combustion chamber) and that of the reaction products ( the hot gases we seek to describe)
shows up as thermal energy of the latter. To describe this quantitatively it is helpful to show
it as a sequence of two steady flow processes. The first is the formation by chemical
reactions of the products from the reactants at standard temperature and pressure. Of course
such a process cannot happen in an isolated reactor, but we can think of it as a process in

1
which the reactions occur at standard temperature and pressure while heat is added or
removed in the amount necessary to keep the temperature constant, and the volume is
expanded or contracted as needed to maintain the pressure constant as well. The heat added
in such an ideal process is defined as the Standard Heat of Formation !H f of the products if
0

the reactants are the elements in their standard states.

The second step involves using the heat added in the first step to change the temperature of
the products. The temperature change is then given by the Steady Flow Energy Equation as

h( Tp , p0 ) ! h(T0 , p0 ) = ! "H f
0

where the products are indicated by the subscript p and the reactants by the subscript 0, but
note that the pressure is the same in both states.

Every chemical substance has a Standard Heat of Formation defined in this way. For
example for water,
1
H2 + O2 ! H2 O(l)
2
"H f , H2 O (l ) = #68.3174kcal / mole
0

The rather cumbersome notation is necessary to indicate that the reaction product in this case
is liquid water. The heat of formation of gaseous water would differ from this by the heat of
vaporization.

Note that the heat of formation of water is negative, so that heat is released in its formation.
The convention that heat added is positive is purely arbitrary.

From the definition of the heat of formation, it is zero for the elements in their standard
states.

In general one should remember that substances that are stable at normal temperature and
pressure have negative heats of formation, those that are unstable (explosives) have positive
heats of formation.

As a simple example of the application of these ideas to rocket engines, consider a

monopropellant rocket engine using Hydrazine, N2H4:

2
This substance has physical properties much like those of water, but has a positive Standard
Heat of Formation of 50.6 kJ/ mole. If we assume for now that it decomposes to produce N2
and H2, and the enthalpy of each can be approximated by cpT, then the chamber temperature
is given by
[c p N2 + 2c pH2 ](Tc ! T0 ) = "H 0f , N2 H4

where the specific heats are per mole if the heat of formation is as well. In this argument we
have assumed that the products of combustion are nitrogen and hydrogen. In general there
might be some ammonia and perhaps other compounds in the decomposition products of
hydrazine. We will see how to deal with this in the following discussion. For a first-cut
estimate, say both N2and H2 behave as ideal diatomic molecules with γ=1.4, so that, for each
of them, cp=γ/(γ-1)ℜ=1.4/0.4*8.314= 29.1 J/mole/K. We then obtain Tc-T0=580K, and if
T0=298K, Tc=878K.

Now let us extend the argument to the somewhat more complex situation of a H2 - O2 liquid
bipropellant rocket

To avoid some complexity, we assume that the hydrogen and oxygen are gases, at 298.16°K.
They react to form products according to

H2 + nO2 ! "H 2 O + #H 2 + $OH + %H + &O2 + ' O

Here n is set by the mix of propellants injected, but the coefficients on the right side are to be
determined. We regard pc as given for this discussion. Actually it is set by the nozzle throat
area for specified mass flows of hydrogen and oxygen, which in turn are determined by the
pressure differential across their injector orifices, and the total orifice areas. So we have to
find Tc and the values of ! ," , # , $ , % , & for given n.

3
To find ! ," , # , $ , % , & (6 unknowns), we first note that the total moles of hydrogen and
oxygen atoms must be equal on the right and left. In general we must have Conservation of
Atomic Species

H : 2! + 2" + # + $ = 2 (1)
O : ! + " + 2# + $ = 2n (2)

If the reaction products are in Chemical Equilibrium (as we shall assume) the other four
relations needed to find the six unknowns are provided by the Law of Mass Action, which
provides a set of relationships between the pressure of each of the compounds which is not an
element and the pressures of its elementary constituents. These expressions take the form of
Equilibrium Constants defined through the stoichiometric relationship for formation of each
of the compounds:

pH2 O
H2 + 12 O2 ! H2 O K pH2 O (Tc ) = 1 (3)
pH2 (pO2 )2

p2H
H2 ! 2H K pH (Tc ) = (4)
pH 2

2
pOH
H2 + O2 ! 2OH K pOH (Tc ) = (5)
pH 2 pO2

pO2
O2 ! 2O K pO(Tc ) = (6)
pO2

The actual choice of reactions to be “equilibrated” is arbitrary, as long as the chosen set can
be used to generate by linear combinations any of the included species out of the “major”
expected product species. In our case we mainly expect to see H2O and, since the mixture is
purposely fuel-rich (n<1/2), excess hydrogen in the form of H2. It is a good practice to select
reactions in which only one of the “minor” species appears, so as to facilitate iteration
starting with the majors only. In the set above, this is true of all but reaction (5), in which two
“minors” (O2, OH) appear. We can replace (5) by the result of eliminating O2 between (3)
and (5), namely
K 2 p H 2O p 2 H 2O
H 2 + 2OH ! 2H 2O = (5’)
K pOH p H 2 pOH

As indicated by the functional expression, the K's are dependent only on the temperature.
This can be shown fairly readily by general thermodynamic arguments. Because of the wide
latitude in selecting the set of reactions to equilibrate, Standard thermodynamic tables do not
list all possible Kp(T), but they do list the basic quantities from which they can be calculated,
namely, the “Standard Chemical Potentials”, µi0 (T ) (molar Gibbs Free Energies at 1 atm.

4
partial pressure), or the Standard Entropies, si0 (T ) , also at 1 atm. pressure. For example, the
equilibrium constant for reaction (3) above is calculated as
1
µ 0H 2O ! µ 0H 2 ! µ O0 2
K pH 2O = exp[ ! 2 ]
!T
where each of the chemical potentials is read from the tables for the given T. If the standard
molar entropies are listed instead, we can calculate first each of the standard chemical
potentials as µ i0(T ) = hi (T ) ! T s i0(T ) , where hi is the molar enthalpy.

Returning to Eqs. (3)-(6), each of the p's can be written in terms of α, β, etc, if pc is set.
Thus if ! " # + $ + % + & + ' + ( then by Dalton's Law of Partial Pressures,
! ! ! ! ! !
pH 2 O = pc pH 2 = pc pOH = pc pH = pc pO2 = pc pO = pc
" " " " " "
We will continue this discussion in the next lecture.

5
16.50 Lecture 11

Subject: Reacting Gases (continued); Temperature dependence of specific heats.

Reacting gases (continued)

We were at the point in the last lecture of solving for the composition in the combustion
chamber.

If we set Tc and Pc, we can solve for the set of six relations outlined for ! ," , # , $ .... , and the
gas composition is known. But how do we find Tc and pc? As indicated earlier the pressure
is determined by the mass flow and the nozzle throat area:
pA
m! = * c t
c (Tc )
We regard the pressure as prescribed for now.

To find Tc we go back to the definition of !H f and write conservation of energy as

0

h H 2 (To ) + nh0 2 (To ) = ! ["h H 2 0 (Tc ) + "H 0f H 0 ]

2

+! [h H 2 (Tc )] + "[#h0 H (Tc ) + #H 0f 0 H ]

+! ["h H (Tc ) + "H 0f H ] + #[h0 2 (Tc )]
+!["h0 (Tc ) + "H 0f 0 ] (7)

where the symbol !h H 2O (Tc ) means the excess specific enthalpy of H2O at Tc, over and above
that at the reference temperature Tf=298.16K. For species like H2 and O2 the Δ is omitted,
because the enthalpy at Tf is zero, by convention. These excess enthalpies are tabulated vs. T
in standard data sets.

If the α, β ... are known we can solve this for Tc.

In general, we have to represent the Kp(Tc)'s and the h(Tc)'s and solve this system
numerically. This is no problem but requires iteration and some care to ensure that the
numerical procedures converge.

The primary constituents in the products are H20 and H2, so as a first estimate, assume
δ=0, υ=0, ε=0, η=0. Then ! " 2n, ! + # " 1, # " 1$ 2n and we can now estimate the
magnitudes of δ and υ. The quantities ε and η will be henceforth neglected, because of the
fuel-rich nature of the mixture. For example from (5’) of the previous lecture,
!
( pc )2
2
K pH2 O " !2 "
= = 2( )
# $
K pOH ( pc )( pc )2 #$ pc
" "

1
 " 2 $ K pOH
! =
2
( )
# pc K 2pH 2O
and for σ=α+β=1 and the above estimates for α and β,

(2n)2 1 K pOH
!2 =
1" 2n pc K 2pH 2O

We now need an estimate of Tc. From (7), and using constant specific heats,

h H 2 (To ) + nhO2 (To ) ! 2n["h H 2 0 (Tc ) + "H 0f H 0 ]

2

+(1-2n)[hH (Tc)]
2
where we read from tables
kcal

!H f H2 0 = "57.8
0

gmole

and we approximate for now

cpH2= 7.5 cal/mole/K cpH2O = 10.6 cal/mole/K
Let h(To)=0 (as noted above this is arbitrary: we assume we are injecting gaseous H2 and O2
at t0=298K). The energy balance then becomes

0=2n[(10.6)(Tc-298)-57,800]+(1-2n)[7.5(Tc-298)]

and we can solve this for (Tc-298):

2n(57,800) 1.15×105 n
Tc − 298 = =
2n(10.6) + (1− 2n+)7.5 7.5 6.20n

To proceed further we must specify n. Let us take n=0.25, or half the stoichiometric value of
0.5. This corresponds to an Oxygen/Hydrogen mass ratio of 4. A more usual value is about
5. Then we find Tc = 3480 K as a first estimate.

Now we have to check to see what the actual composition is for this temperature. Using the
methods explained before, the values of Kp0H and KpH 0 at the estimated temperature are:
2

n Tc Kp,0H Kp,H20
atm1/2
0.25 3480 1.70 4.93

c ! = 0.19 , so now if pc=100 atm, δ=.019 as a second

And from our previous result for δ, p1/2
estimate, compared to the original estimate of 0.

What about ν? From Eq. (4) of the previous lecture,

2
 !
( ) 2 pc 2 ! 2 p
pH2
KpH = = " = c

pH 2 # # "
( ) pc
"
$
! " #K pH (Tc )( )
2

pc
For n = .25, Tc = 3480°K, we calculate from the standard chemical potentials KpH = .35 atm,
1 .42
so ν 2 = (0.5)(.35) ; ν = 1/2
pc pc
For pc = 100 atm, and the estimated temperature, we now have ν = .042 as a second estimate.

So, using Eqs. (2) for α and (1) for β, we find α=2x0.5-0.019=0.481, and
β=1-0.481-(0.019+0.042)/2=0.489:

H2 + .25 02 ! (.481)H2 O + (. 489)H2 + (.019)OH + (.042)H

With this composition we can now go back and calculate Tc more accurately:

0 = hH2(To) + .25 h02(To) = .481 [10.6(Tc-298)-57,800]

+.489 [7.5(Tc-298)] + .019 [7.8(Tc-298)+10.06]

+ .042 [5(Tc-298)+ 52.09]

.481(57, 800)!.019(10.06)!.042(52.09)
Tc-298 = =3047
.481(10.6)+.489(7.5)+.019(7.8)+.042(5)

Tc=3345 K

We stop the iteration here. A useful observation is that, just as the first approximation for Tc
was too high because it ignored the endothermic dissociations, the second is likely to be too
low, because it must have over-estimated these dissociations, since the equilibrium shifts in
favor of dissociation when the temperature is high. It is a good idea to under-relax the
temperature, as for example using for the next estimate the average of the previous two:
1
Tc ! (3480 + 3345) = 3413 K
2

Temperature dependence of specific heats

In working through the above example, we assumed for simplicity that the specific heats of
the gases were constant. In fact they vary with temperature, generally increasing with
increasing temperature. Generally we account for this by taking the specific heat or the
enthalpy from tables of thermodynamic properties. But it is desirable for you to understand
qualitatively why the specific heat varies, hence the following discussion.

3
For a full understanding of this subject one must have recourse to Statistical Mechanics,
which deals with the thermal behavior of matter from a microscopic viewpoint, drawing on
quantum mechanics. There is not time for such a discussion here, so we must be content
with a summary of some of the results.

A key concept is that of active degrees of freedom. This is the number of terms in the
classical expression for total energy that are quadratic in a coordinate or a velocity
component. According to the Law of Equipartition of Energy each of these states if fully
excited will contain an energy per mol of ℜT/2 where ℜ is the ideal molar gas constant.

Real molecules behave quantically, so at low enough temperatures, where the quantum of
energy is greater than the available thermal energy per molecule, a quantum degree of
freedom may not be “active”. Generally, translational and rotational quanta are small enough
that translational degrees of freedom are regarded as active in all cases of interest to us, but
vibrational freedoms become active only at temperatures of the order of 2,000-3,000K.
Taking the simplest case first:

a) Monoatomic gases

The molecule of a monoatomic gas can be thought of as a point mass, and as such it has three
translational degrees of freedom. So the Internal Energy (per mol) is 3ℜT/2, so we find a
molar Specific Heat at Constant Volume
cv=12.47 Joule/mole/K ( Monatomic)
The ideal gas constant being ℜ= 8.32J/mole/K, the Specific Heat at Constant Pressure is
cp= cv+ℜ= 20.79 Joule/kg mole/K
and the ratio of specific heats is γ=5/3= 1.667.

All of these values are independent of temperature because the translational degrees of
freedom are fully excited at normal temperatures.

b) Diatomic molecules.

The simplest non-trivial case, the diatomic molecule, may be thought of as a "dumbbell" with
in general unequal weights on the two ends, connected by a spring that can extend along the
line of their common centers.

Like the monoatomic molecule, the diatomic molecule as a whole has 3 translational degrees
of freedom, each containing energy ℜT/2 per mol. In addition it has 2 rotational degrees of
freedom (note that the moment of inertia about the line of centers is very small, so this degree
of freedom does not count). Each of these rotational degrees of freedom is fully excited at
the temperatures of interest here (though not at very low T) and contains ℜT/2. So at "low"
temperatures (near room temperature) the Internal Energy per mole is 5ℜT/2 and the Specific
Heat at Constant Volume is
cv=20.79 Joule/mole/K ( Diatomic at low T)

4
It follows that
cp= cv+ℜ= 29.11 Joule/mole/K (Diatomic at low T)
and γ=7/5= 1.4.

At higher temperatures, the vibration of the molecule along its line of centers is excited,
introducing two more degrees of freedom. This happens gradually between room
temperature and the maximum temperatures encountered in propulsion devices. It is this
variation that causes the variations of cv of interest to us here. If the vibrational degrees of
freedom are fully excited, the Internal energy per mole is 7ℜT/2 and

cv=37.41 Joule/mole/K ( Diatomic at high T)

Correspondingly,
cp= cv+ℜ= 45.73 Joule/mole/K (Diatomic at high T)
and γ=9/7= 1.286.

c) Polyatomic Molecule

Most polyatomic molecules are non-linear, so they have an additional rotational degree of
freedom as compared to diatomic molecules. Thus at low temperatures, the Internal Energy
is 6ℜT/2=3ℜT. It follows that

cv=24.94 Joule/mole/K (Polyatomic at low T)

cp= cv+ℜ= 33.260 Joule/mole /K (Polyatomic at low T)

and the ratio of specific heats is γ=4/3= 1.333.

The behavior of the vibrational degrees of freedom of polyatomic molecules can be very
complex if the number of atoms is large. A useful rule for estimating their energy content is
to compute the number of vibrational degrees of freedom as the difference of the total for the
molecule minus the translational and rotational degrees of freedom. Thus if N is the number
of atoms, Vibrational Degrees of Freedom = 3N-6, and the Internal Energy per molecule if
they are all excited is 3kT+(3N-6)kT=(3N-3)kT. The ratio of specific heats then becomes

3N " 2
! = (Polyatomic at high T)
3N " 3

and it is clear that this can be close to 1 for N large.

5
16.50 Lecture 12

Subject: Nozzle flow of reacting gases

In the last two lectures we discussed the phenomena that occur in the combustor, and how to
estimate the properties of the gas in the (near) stagnation state there. Suppose now that we
have determined the composition of the gases in the rocket chamber, and we wish to compute
the flow through the nozzle, taking account of chemical reactions.

There are three things that we must account for, that were not included in the simple model
based on ideal gas behavior. The first is that the composition of the gas is not necessarily
constant in the flow, so that all the properties that are composition dependent must be treated
as variables along the flow direction. These include the specific heats, the gas constant R and
the ratio of specific heats γ. Second, the sum of the thermal enthalpy and the kinetic energy is
no longer constant because there can be exchange of chemical energy and thermal energy.
But if the enthalpy is defined as in the previous two lectures, i.e., including in it the chemical
enthalpy of formation, then the sum of it and the kinetic energy is indeed conserved.
Finally, because of this energy exchange between chemical and kinetic, there can be a change
in the entropy of the flow. Fortunately we can neglect this entropy change in some important
special cases, as will be explained.

To take these effects into account quantitatively we may proceed as follows:

The chamber condition is specified by Tc, pc and the mole (yi), or mass (xi), fractions of the
various chemical constituents. As explained in the last two lectures we can compute all the
thermodynamics properties of the gas per unit mass at chamber conditions:

Hc = ! x i Hi ; H i = (!hi (Tc ) + H !f i ) / M i

Sc = ! xi Si
dT p
Si = " c pi ! Ri ln i (Si, cpi per unit mass)
Tr
T pr

where it will be recalled that xi is the mass fraction of species i and therefore
! xi = 1
Notice also that each enthalpy Hi is now per unit mass, and so is Hc.

To deal with the flow we note first that energy conservation for the gas flow gives

u2
Hc = H(T,p) + 2 (1)
which replaces the ideal gas energy equation
u2
(cpTc = cpT + 2 )

1
We must have some statement about the variation of the entropy, since in general the transfer
of energy from chemical to thermal takes place at a finite rate and there is therefore an
entropy increase.
But there are two limiting cases for which the entropy change is very small

a) chemical equilibrium
b) frozen flow

The case of Chemical Equilibrium will be approached if the reactions occur fast enough to
keep up with the temperature and pressure changes caused by the expansion (flow time >>
reaction time), so the chemical energy is transferred through an infinitesimal !T , and

S = ! xi ( p, T)Si ( p, T ) = Sc (2e)

and the xi are determined by Equilibrium at the local T & p. Re-calculating all these mass
fractions is clearly a tedious task, but conceptually not very different from what was done for
the chamber conditions, with the major difference that Entropy per unit mass, rather than
Enthalpy per unit mass is now constrained.

In the case of Frozen Flow the reactions occur so slowly that the xi are fixed at their chamber
values (flow time << reaction time). In this case there is no chemical energy release and
again
S = ! x i (pc ,Tc )Si (p, T ) = Sc (2f)

These are two limiting cases, which provide upper and lower limits for the velocity u, at a
given p, since the case of Equilibrium Flow gives the maximum thermal energy availability
for conversion to kinetic energy, while the Frozen Flow case gives the minimum.

In these calculations, the entropy per unit mass Si of one species at (p,T) is related to that per
mol, Si ( p,T) = S˜ i ( p,T) / M i , and this can be calculated from tabulated “standard molar
entropies”, which are at p=1atm, as

S˜ i ( p,T) = S˜ i (T) ! "ln( pi (atm)); pi = py i

0
(2g)

It is important to note that the conservation laws for both energy and entropy are per unit
mass. Thus it is the total energy and the entropy of a fixed mass of gas that is conserved. At
times it will be convenient to write the relations in terms of moles of the constituents (using
yi rather than xi) but a fixed mass of gas may contain different numbers of moles at different
points in the flow, so if we use the yi we must be careful to keep track of the changes in the
number of moles.

To go from mole fractions to mass fractions or vice-versa,

2
 yi Mi xi / Mi
xi = ; yi =
! yi Mi ! xi / Mi
For example for a mixture of H2 + H20
1 1
y H 2 = , y H 2O =
2 2
1
(2 ) 1
x H2 = 1 2 1 =
2 (2) + 2 (18) 10
1
(18) 9
x H2 O = 2
=
1
2 (2) + 2 (18) 10
1

The above are all the relationships we need, so let us see how we proceed, for given chamber
conditions, pc and Tc.

a) Compute xi (pc,Tc)
Hc
Sc

b) Select a value of p<pc. We can treat p as the independent variable, finding all the
other properties as functions of it. If we wish to find the conditions at the nozzle exit
then the pressure is the exit pressure, p=pe.

c) From the Entropy equation (2e or 2f), find T

r !1
T p "
(This replaces T = (p ) of the Ideal Gas model)
c c

1) For Frozen Flow, since xi (p,T) = xi (pc,Tc) are known, Eqs. (2f) and (2g)
contain only T as an unknown, and can be solved iteratively at each p (this
replaces the enthalpy conservation iteration we did in the chamber).

2) For Equilibrium Flow, we still have s = sc, but now we don't know the xi,
so we have to solve for them at each p, as noted above.
A. Assume a T
B. Compute the yi (and xi) from Kpi(T)
C. Compute S(T,p) and Iterate on T until it equals
Sc(Tc,pc).

d) In either Fozen or Equilibrium cases, having T and xi(T,p) compute H(T,p), then
the velocity is given by
u2
2 = Hc - H(T,p)

3
These procedures enable you to describe the flow in the nozzle with p as the independent
variable. Having u, p, T we can find ρ and hence ρu and the variation of the nozzle area
with p. In particular, to find the throat area we can compute ρu for a set of values of p
near (1/2)pc and by plotting them determine the p that maximizes ρu. This is the throat
pressure, and the maximum value of ρu defines c* = pc / ( !u)t

Normally, of course, such calculations are handled by standardized computer programs.

They all follow the logic outline above although the numerical procedures may become
quite complex because of considerations of stability etc. You don't need to know all the
details, but it is important that you understand the logic they follow. As we say, Garbage
In-Garbage Out. Be sure you know what the computer is doing.

A very widely used suite of codes for thermochemical problems in aerospace is the CEA
set, available at http://cearun.grc.nasa.gov.

4
16.50   Lecture  13  

Subject:    Rocket  casing  design;  Structural  modeling  

 
Thus  far  all  our  modeling  has  dealt  with  the  fluid  mechanics  and  thermodynamics  of  
rockets.  This  is  appropriate  because  it  is  these  features  that  set  rockets  apart  from  most  
other  devices.    On  the  other  hand  it  is  not  possible  to  understand  the  characteristics  and  
limitations  of  rockets  as  systems  without  at  least  a  rudimentary  understanding  of  their  
structural  characteristics  which  determine  their  mass,  durability  etc.    To  aid  such  
understanding  we  must  develop  some  simple  Structural  Models.    The  first  step  is  to  
understand  the  Loads  that  the  rocket's  structure  must  withstand.  Let  us  begin  with  a  
Solid  Propellant  Rocket  
 
1) Loads  
 
We  model  the  rocket  case  as  a  sphere-­‐cylinder  full  of  fuel,  acted  on  by  the  thrust  F  and  
the  payload  reaction  (a+g)(Mpay)  where  a  is  the  acceleration  and  g  the  gravitational  
acceleration.  

mpay

Propellent
D

A A
Casing

Nozzle

Image by MIT OpenCourseWare.

               
The  casing  is  subjected  to  the  following:  
a)   internal  pressure,  pc  
  b)   shear  loads  from  the  propellant  grain,  which  is  bonded  to  the  case  

1
   c)   compressive  loads  
 
To  find  the  forces  in  the  casing  due  to  the  pressure,  we  make  free-­‐body  diagrams.    First  
consider  section  A-­‐A.  

   
    Section  A-­‐A  
From  this  free  body  diagram,  the  circumferential  stress σy  is    
pD
          !y =          
2t
There  is  also  an  axial  load  generated  by  the  combination  of  the  internal  pressure,  the  
compressive  load  from  the  payload  and  the  compressive  stress  in  the  propellant.    
Constructing  another  free-­‐body  diagram,  
 
 
 

[MPay+(MCase+Mprop  x/L]a  
 

Ap   x  

A  

pc   σp   σx  

x
[Mpay  +  (Mcase  +  Mprop)  L  ](a+g)  +  σx  πDt  =  pcA  +  σpAp  
1 x
    σx  =  πDt    {pcA  +  σpAp  -­‐  [Mpay  +  (Mcase  +  Mprop)  L    ]  (a+g)}  
 
This  stress  is  most  positive  (tension)  at  x=0:  

2
 ( pc A + ! p Ap ) pc Acase pc D
      (! x )Max = ! =  
" Dt " Dt 4t
where  we  have  assumed  hydrostatic  grain  equilibrium,  i.e.,   ! p = pc .  
At  the  other  end,  x=L,  we  have  either  least  tension  or  possibly  compression  of  the  casing  
wall;  this  last  possibility  would  imply  buckling  problems.  Putting  x=L,  
p A " ( M 0 " M nozzle )(a + g)
(! x )min = c case  
# Dt
and  recalling  that   a + g = F / M 0 = cF pc At / M 0 ,  
 
p M
      (! x )min = c [Acase # cF (1 # nozzle )At ]  
" Dt M0
So,  unless  At  is  less  than  about  Acase/cF,  there  will  still  be  tension  at  the  casing’s  base.  
 
2)   State  of  Stress  
 
With  this  representation  of  the  loads  we  can  now  deduce  the  State  of  Stress  in  the  
casing  wall,  subject  to  some  simplifying  assumptions.  We  assume  t  <<  D,  so  the  stress  
may  be  described  as  a  state  of  plane  stress  in  x-­‐y  coordinates,  where  x  is  axial  and  y  
tangential.  In  these  coordinates  imagine  a  small  triangular  element.    The  stresses  on  the  
element  in  general  are  tensions  σx,  σy  and  shear  τxy  =  τyx  :  :  
 
x
σx
τxy
α

τxy

σy

σx'
τx', y'

y
       
  Image by MIT OpenCourseWare.
Now  consider  the  stresses  on  a  plane  perpendicular  to  a  line  making  the  angle  α  with  
the  x  axis.    In  treatments  of  stress  and  strain  it  is  shown  that  
      σx'=  σx  cos2α  +  σy  sin2α + 2τxy  sinα  cosα  
      τx'y'  =  -­‐(σx-­‐σy)  sinα  cosα  +  τxy  (cos2α  -­‐  sin2α)  
It  is  then  shown  that  the  maximum  (or  minimum)  of  σx'  occurs    at  an  angle  α such  that    
2τxy
        tan2α  =      
σx-­‐σy
and  that  these  stresses  are:  

3
 σx  +  σy !x "!y 2
      σ1,2  =      ±   ) + # xy  
2
(
2 2
In  our  case  τxy  =  0,  so  σmax  is  simply  the  larger  of  σx  or  σy.  and  usually  σy>σx:  
pc D pcD 1
σy  =   2t  ;     (σx  )Max  =   4t    =    σy  
2
 
So  the  casing  must  be  designed  to  withstand  the  hoop  stresses:    
p
        t = D c " (a safety factor)  
! ult
3)    Case  Mass  
 
Again  assuming  constant  t<<D  and  hemispherical  caps,  
L pc
    Mcase  =   (!D2 + !DL)t    ρcase  =  πD3(1+D  )        ρ  
2σy case
The  mass  of  the  propellant  is  
πD3 πD2L 1 1L
    Mprop  =  ( 6    +   4  )  ε  ρprop  =  πD3(6    +  4  D  )  ε  ρprop  
where  ε  is  the  fraction  of  case  volume  filled  by  propellant.  
 
L pc
M case 3(1+ D ) ! "case
      =  
M prop (1+ 3 L ) # "
prop
2D
 
Typically,  for  pc  =  50  atm=  750  psi  ,  σy  ≈  150,000  psi  (about  100kg/mm2  in  European  
lb lb
notation),  and    ρcase  =  .3   3  (steel)  ,  ρprop  =  .06   3    
ln ln
L
1  +  D
M case
        ≈  .093  
M prop 3 L    
1  +  2  D
L M case
  and  for  D    =  3,   ≈  .068.  
M prop
 
Is  this  a  reasonable  estimate?  For  the  Minuteman,  from  Sutton  &  Ross  
 
    Mcase  =  2557  lb       Mprop  =  45,831  
M case
        = 0.056  
M prop
It  seems  we  have  been  somewhat  conservative  in  our  design  parameters.  
 
4)   What  is  left  out  of  our  estimate?  

4
  
  a)   Attachments  
  b)   End  bells  have  1/2  the  stress  -­‐  make  1/2  as  thick?  
  c)   Shear  loads  on  propellant  grain  
 
 
 
 
 

5
Is  this  a  reasonable  estimate?  For  the  Minuteman,  from  Sutton  &  Ross  
 
    mcase  =  2557  lb  
          }  =  .056  
    mprop  =  45,831  
 
4)   What  is  left  out  of  our  estimate?  
 
  a)   Attachments  
  b)   End  bells  have  1/2  the  stress  -­‐  make  1/2  as  thick?  
  c)   Shear  loads  on  propellant  grain  
 
It  seems  we  have  been  somewhat  conservative  in  our  design  parameters.  
 
 
 
 
 
 
 
 
 

6
16.50 Lecture 14

Subjects: Heat Transfer and Cooling

Because the combustion temperatures in most rocket engines are far beyond the
levels tolerable by most common structural metals, the walls of the combustion
chambers and nozzles must be cooled. The high pressures in the combustion
chambers exacerbate the cooling problem by causing very high heat transfer rates, as
we will see. For these reasons cooling is one of the critical technologies of rocket
engines.

There are four ways of cooling rocket chambers and nozzles:

a) Convective cooling by fuel or oxidizer
b) Ablative cooling
c) Heat sink cooling
d) Radiation cooling

We will discuss each of these briefly, then return to a discussion of the physical
mechanisms of importance in each, with the major emphasis on the first two because
they are of the greatest practical importance.

a) Convectively Cooled Chambers and Nozzles

There are 3 rather distinct heat transfer processes involved in such cooling:

1) Boundary layer heat transfer to the wall from the nozzle flow.
2) Conduction through the wall.
3) Convective heat transfer to the fluid in the cooling passages.
b) Ablatively Cooled Chambers and Nozzles

1
In this type of cooling, the important mechanisms are

1) Boundary layer heat transfer to the wall from the nozzle flow.
2) Decomposition and evaporation of the material of the wall, which
generates a buffer gas layer, reducing the heat transfer rate.

c) Heat Sink Cooling

Here the wall temperature is maintained at a sufficiently low level by conduction of

the heat into the material of the wall, so the mechanisms are

1) Boundary layer heat transfer to the wall from the nozzle flow.
2) Conduction of the heat into the wall, resulting in a temperature rise
with time.

d) Radiative Cooling

For small rocket engines used in space applications, for example for station keeping,
it may be possible to cool the nozzle by thermal radiation, eliminating the
requirement for cooling passages, or for erosion of the nozzle as in ablative cooling.

Since all of these cooling schemes have in common the heat transfer through a
boundary layer to the wall, this aspect of them will be discussed first.

Heat transfer from a flowing fluid to a surface.

2
For this process it is conventional to define a film coefficient h by:

q = h(Tg - Tw)

where Tg is the gas' stagnation temperature and Tw is the wall temperature.

A more detailed way to represent this is:

q = ρucp (Tg - Tw) St

where St is the Stanton Number, with the physical meaning

heat flux to wall

St = energy flux in flow relative to wall

Equating the two relations for the heat flux,

h= ρucp St
and in dimensionless form we can write
hL ρuL cpµ
kf = µ ( kf ) St
where the group on the left is called the Nusselt number, and the group involving
just fluid properties is the Prandtl number:

Nu = Re Pr St
Nusselt Reynolds Prandtl Stanton

Here L is the length scale of the flow, for which we have differing interpretations
depending on the flow geometry.

Heat Transfer to Passage Flows and Frictional Pressure Drop

Consider first the heat transfer to fluid flowing through a long narrow passage,
which may represent the cooling passages in the nozzle.

3
 L
If the length L of the passage is such the d > 20, the flow becomes Fully Developed,
meaning that it has velocity and temperature profiles independent of x. The heat
balance on a fluid element of length dx is

( !uA)dTg = S q dx

where A is the cross sectional area of the passage and S is the surface area per unit
πD2 A D
length (perimeter). For a circular passage, S = πD, A = 4 , S = 4 ; more generally,
D here is defined as the “hydraulic diameter”, D=4A/S.

Writing qw = !uc p St (Tw " Tg ) , we then have

dTg 4St
= (Tw ! Tg )
dx D
If Tw and St can be taken as constant, this integrates to
Tg
!4 St
x
Tw
Tw ! Tg = (Tw ! Tg0 )e D

Tw0

x/D
The total heat transferred to the fluid in a length L is then
L
!4 St
Q = m! c p [Tg (L) ! Tg0 ] = m! c p (Tw ! Tg0 )(1! e D
)

But any time there is boundary layer heat conduction to a wall, there is also wall
friction, and in the case of a duct, pressure drop. To see this quantitatively, take now
a force balance on the same fluid element:
dp ρu2
A dx dx = -τwSdx = -cf 2 Sdx
dp S ρu2 4 ρu2
dx = -cf A 2 = -D cf 2
and if compressibility is ignored, the pressure drop Δp=p0-p in a pipe length L is
4c "u 2
!p = f L
D 2

4
It is interesting to calculate the ratio of the non-dimensional pressure drop and the
heat added. In particular, for small values of 4StL/D, so that the expression for Q can
be expanded to first order, we find
(!p / "u 2 ) c /2
" f
(Q / m! c p (Tw # Tg0 ) St
which is a quantity we will shortly find to be of order unity.

Sometimes we define a "friction factor" f ! 4Cf . We then find that f or cf can be

ρuDh
represented as a function of only the Reynolds Number, defined as Re !
µ
where ρu is the mass flow per unit area of the passage and Dh is the hydraulic
diameter, defined as above. All fluid mechanics texts present the variation of Cf or f,
and a sample is shown at the end of this lecture. Note that in that graph the quantity
hf on the left scale is Δp/ρ, and the scale on the left is es/d, where es is the
“equivalent sand roughness” (for a hydraulically smooth surface, es<<1).

Although there is some dependence of this relationship on Mach number for Mach
numbers approaching 1, for most of our purposes we can take this relationship for
the friction coefficient to be valid. For incompressible flow, the pressure drop is
then directly calculable from the definition of the friction coefficient. For
compressible flow we must account for the acceleration of the flow that occurs
because of decreasing density as the fluid is heated.

Reynolds Analogy

To transfer our understanding of the viscous pressure drop to heat transfer, we

observe that both in turbulent flow and in laminar flow the same mechanism
transports momentum and energy to the wall (flow eddies in turbulent flow,
molecular motion in laminar flow). Thus we would expect the ratio of heat flux into
the wall to heat flux along the wall to equal the ratio of momentum flux into the wall
divided by momentum flux along the wall.
c f !2u
2
q #w
= =
!uc p (T " Tw ) !u 2 !u 2
This results in
cf
St =
2
More detailed analysis shows that the thermal conductivity matters slightly, and we
find (Hill/Peterson, pg. 129)
c
St Pr.67 = f
2
cp µ
where Pr = k is the Prandtl Number.

Now we can use the empirical information on pressure drop to estimate heat
transfer. Thus for turbulent flow in a smooth pipe,

5
 c f ".67 "1
St ! Pr = .023Re 5 Pr ".67
2

Heat Transfer Through Boundary Layers

On the gas side of the rocket nozzle wall we have a nearly inviscid flow, with a thin
viscous layer near the wall. In rocket nozzles, the flow is strongly influenced by a
streamwise pressure gradient. The "favorable" pressure gradient produced by the
pressure drop in the flow direction accelerates the flow in the boundary layer, so the
Boundary Layer tends to be thinner than it would be on a flat plate or on a pipe with
fully developed flow. It is thinnest at the nozzle throat, where the pressure gradient
is the strongest. For this reason and because ρu is largest there, the heat flux is
largest at the nozzle throat.

Much of the early progress in quantifying the heat transfer rate at the throat came
from using data from passage flows, with Boundary Layer Thickness ≈
Tube diameter Dh
2 (δ ≈ 2 ). One then applies various techniques to estimate δ, and
applies the passage flow data in the form

St = .023 Reδ-1/5 Pr-.67

(ρu) ! 2δ
where Reδ ! . But it is hard to accurately predict δ, so in practice one resorts
µ
to empirical results. Bartz (See HP pg. 549) recommended for the nozzle throat

h
St = Gc = .023 (ReD) -1/5 Pr-.67
p

where D is the throat diameter.

With these elements you are now in a position to analyze the most critical
processes in the cooling system for a convectively cooled rocket engine.

6
Image: Wikipedia. © Wikipedia User: Donebythesecondlaw. License CC BY-SA.
This content is excluded from our Creative Commons license.
For more information, see http://ocw.mit.edu/fairuse.

7
16.50 Lecture 15

Subject: Ablative cooling

By ablation we mean the recession of a surface due to heating, usually by a hot gas.
It is the key process for

a) Re-entry heat shields

b) Solid propellant nozzles
c) Rocket case insulation
d) Fire-proofing skyscrapers' structures

Consider a hot gas flowing over a surface which can

1) evaporate, and
2) whose vapor can react with the external flow

A heat balance at the surface gives

qs = qw - (ρv)w Δhw (1)

where the heat of ablation, Δhw=hw – hs, in J/kg, can include a heat of vaporization
and decomposition and qw is the heat flux from the fluid boundary layer.

Let us assume that it is useful to write

ue 2
qw = ρeueSt (He + 2 - Hw)

where Hw is the total enthalpy for the wall material at the wall temperature, and
He=cpgTge is the gas specific enthalpy at the temperature Tge just outside the
boundary layer . Notice that this generalizes our previous expression for heat
transfer by replacing the total enthalpy for the static enthalpy; this is of general
validity for high-speed flows. By energy conservation in the core flow,
u2
H e + e = c pg Tc .
2

1
Then the heat transfer to the solid is

ue2
qs = ρeueSt (He + 2 - Hw) - (ρv)w (hw - hs)

Since qs is the heat flux available to heat the wall, it is clear that the evaporation
reduces the wall heat flux, i.e. the heat flux into the solid.

Now what determines (ρv)w? First note that (ρv)w = rρs where r is our "recession
rate". Consider a "thermal wave" propagating into the solid. We will now work in
the receding frame, in which we can assume a steady situation. If the recession rate
is r(m/s), we see solid material moving at ! y = "r , and convecting a heat flux . Then,
if there is no local heat generation,
d dT
(!ks ! r " s cs T ) = 0
dy dy
dT
which integrates to ks + r! sc sT = const. = r! sc sTw" , or
dy
d (T ! Tw" ) r k
+ (T ! Tw" ) = o (! = s , heat diffusivity, in m2 /s)
dy # " sc s

r
! y
T ! Tw" = (Tw ! Tw" )e #

dT " r
The heat flux at at y = o is qs = !k s $ = k s (Tw ! Tw& ) , giving
dy # y= o %

qs
qs = ! sc sr(Tw " Tw# ) , or ! sr =
c s (Tw " Tw# )

So returning to our expression for the heat flux to the wall,

qs = ! e ue St c pg (Tc " Tw ) " ! sr#hw

! src s (Tw " Tw# )

! e ue c pg (Tc " Tw ) ! sr c pg (Tc " Tw )

! ! sr = St , = %&
c s (Tw " Tw# ) + $hw ! e ue St c s (Tw " Tw# ) + $hw

You may compare this to Sutton pg. 510 Eq. 15-6

Now how does (ρv)w influence St? Physically, we know that the vapor comes off the
wall with u=0, so it will tend to retard the flow near the wall.

2
Quantitatively, Lees says, for small surface blowoff effect,
St B 1.2 8, 320 (ρv)w
= c ! ! 2500 J / Kg / K , B =
Sto eB ! 1
pg
0.2 20 ρeueSto
while Sutton quotes Lees as giving for the larger blowoffs
St -.77
Sto = 1.27 B 5 < B < 100
These give the same trends, but somewhat different numbers. We adopt here Lees’
! sr S B
formulation: We had = ", so B = ! t = ! B
! e ue S t Sto e "1

or e B !1 = " # B = !n (1+ " )

+- % c pg (Tc " Tw ) ( /-
so finally, ! s r = ,!n '1 + * 0 !eueSto (1)
-. & cs (Tw " Tw# ) + $hw ) -1
and then
% c pg (Tc " Tw ) (
qs = ( ! s r)cs (Tw " Tw# ) = !eueSt 0 cs (Tw " Tw# )!n '1+ * (1’)
& cs (Tw " Tw# ) + $hw )
m! pc
At the throat, ( !eue )t = = . This is where qw, and hence r, is maximum.
At c*
Numerically, one often sees !hw >> c s (Tw " Tw# ) , and also ! << 1. This would leave
us with the approximations

c pg (Tc # Tw )
! sr " ! e ue Sto (2)
$hw
St ! Sto (3)

and going back to the surface heat balance, qs ! o , i.e., to the first order, the heat
does not penetrate below the ablating layer, which is as intended.

3
Equations (1) or (2) can be used to select the proper thickness of the ablative
(sacrificial) layer, once the burn time is given: !Abl = r t b .
Now let us look at some numerical examples, to show the magnitudes of the
terms:

Joule
A. For carbon, ∆Hfg = 172 kcal/gmole=60 x 106 kg
172000 cal
≈ hw - hs = 12 = 15,000 g
1.2 8, 320
c pg ! ! 2500 J / Kg / K , ρs ≈ 2000 kg / m 3 kg / m 3
0.2 20

and assume Tc-Tw=1,000K.

p p
Now ρeue ≈ = ! " 0.65
RT RT

105 (pc atm) kg

≈ ≈ 58 (pc atm)
8.32x10 3 m2s
(3000)
20
Taking pc=100 atm and St=0.001, we then calculate

2,500 *1,000
r! " 58 " 100 " 0.001 = 1.2 " 10 #4 m / s = 0.12 mm / s
2,000 " 60 " 10 6

So in 100 sec., the change in the surface is ≈ 12 mm = 1.2 cm.

This is of the right magnitude, but most important is to see what it depends on.

B. For rubber hw-hs ≈ 10,000 cal/gm mole. For an approximate empirical

formula C10 H20 , this gives per unit mass

10000 10000 cal Joule

120 + 20 = 140 ≈ 70 g ≈ 280,000 kg

Also cp ≈ 1 cal/g ≈ 4,000 Joule/kg K

So now we would get about 100 times the regression rate for carbon. In this
case though, if the rubber is used in the casing of a solid rocket engine then ρeue <<
1
(ρeue)throat, maybe by 100 , so the net regression may end up being similar.

4
16.50 Lecture 16

Subjects: Thrust Vectoring ; Engine cycles; Mass estimates

Thrust Vectoring

Liquid Bipropellant rockets are usually "gimballed" to change the thrust vector

Fuel Tank

Flex Line Actuator

(in two
planes)
Pumps
Ball Joint

Here all the high pressure pumps and pipes move together, and the flexible joints
are in the low pressure feed pipes

Since the case is pressurized in the solid rocket we either have to have a high
pressure joint between the case and the nozzle, or a way to deflect the jet from a
stationary nozzle. Because of the design difficulty of the joint, the early motors used
jet deflection either by vanes or by gas injection.

1) Jet Deflection
Vanes Gas Injection

Gas injection

Vanes deflect jet

Image by MIT OpenCourseWare.

More modern solid rockets use movable nozzles with joints that allow rotation by
means of shearing motion in elastomer layers between spherical shells.

1
These are difficult to implement. The design issues include
a) Large pressure differences
b) Sealing against leaks of hot gases
Schematically, the joints look like:

Center of rotation

Rubber in shear

Image by MIT OpenCourseWare.

Pressurization Systems

The systems for pressurizing the propellant for injection into the combustion
chamber of liquid bi-propellant rockets range from very simple gas bottle
pressurization, to complex engine cycles. We have time only to outline them here,
and indicate some of their advantages and disadvantages.

He
a) Gas bottle pressurization of fuel tanks: Valve

Advantage – Simplicity
Propellant
Disadvantage - Weight of both gas and
Propellant tanks
Valve

2
 b) Monopropellant or Bipropellant Gas Generator

In these systems a hot gas is generated either by decomposition of a monopropellant

such as H2O2 or N2H4 or by combustion of a small fraction of the main propellants.
The hot gas is expanded through a turbine and dumped overboard, the turbine
driving the propellant pumps.

Advantage - Simple start-up even in space, and straightforward

development process
Disadvantage - Overboard dump of exhaust reduces the effective Isp
Examples- V-2 (H2O2), Atlas, Thor, Delta, Saturn 5, Titan

3) Topping Cycle (Expander)

In this system the entire fuel flow is pumped above the injection pressure, heated in
the cooling jackets of the nozzle and combustion chamber, expanded through a
turbine that drives the pump, then injected into the combustion chamber.

3
 Advantage - High Isp and simplicity relative to the preburner
systems to be described next.
Disadvantages - Complex startup dependent on stored heat in
system, and limit on the chamber pressure that can be
achieved due to temperature limit on turbine drive gas.
Example - RL-10 for Centaur

4) Preburner (Staged Combustion Cycle)

This system predominates in modern high-performance engines. In US practice

most if not all of the fuel is pumped to a high pressure and burned with a part of the
oxidizer (to keep the temperature within the limits of the turbine), expanded
through a turbine that drives the fuel and oxidizer pumps then injected into the
main combustor. The central feature is that all the propellant goes through the main
combustor and nozzle.

In some Russian engines that use Kerosene as the fuel, the preburner operates
oxidizer rich rather than fuel rich.

Advantages - Ability to operate at very high chamber pressures,

high Isp, and flexibility of a cycle design.
Disadvantages - Complexity of design construction and startup,
high cost, high pump delivery pressures.

6) Tank Weights (connected to pressure)

How thick do the tank wals need to be? Balancing the pressure forces against the
hoop stress in a cylindrical chamber,
t p
pD = 2σt ! D =
2σ
Assume the cylinder, of length L is terminated with two hemispherical caps:

4
 L
Mass = ρt[πDL + πD2] = πρD2t[1 + D ]

D2 D3 1 1L
Vol = π [ 4 L + 6 ] = πD3[ 6 + 4 D ]

L L L
Mass t 1+D t 1+D p 1+D
Vol = ρ D 1 1 L = 4ρ D 2 L = 2ρt σ 2 L
6+4D 3+D 3+D

L
Tank mass ρt p 1 + D
Prop mass = 2 ρp σ 2 L
3+D

Example: Pressure-fed Hydrazine (N2H4), Nitrogen Tetroxide (N2O4 )

3
pc = 20 atm = 2.03x106Pa, ρHydrazine =995 kg/m

Titanium alloy ρt = 4,100 kg/m3,

σ ≈ 8x108 Pa
L L
Tank mass 4100 2.03 1+ D 1+
D
Prop mass ~ 2 62995 800 2 L ~ .021 2 L
+ +
3 D 3 D

or about 0.03-0.04 when attachments are added.

Example: Including thrust load

During engine operation, the pressure at the bottom of the tank (of length L) is
p = p0 + n! p gL , and this should replace p0 in the calculation. Here n is the sum of
gravity and acceleration, divided by gravity (i.e., the effective gravity, in g’s). If the
hydrostatic second term dominates, as it may in long tanks, then we get
MTank ! gL 1+ L /D
= 2n t
M Pr op. " 2 /3 + L /D
Suppose acceleration is 4 g's, the tank length is 30 m, and other data are as before
(except that propellant mass does not matter now). We calculate
Tank mass 1+ L /D
Prop mass = 0.012 2 /3 + L /D
so the static part of the pressure still dominates, but thrust loading does matter.

Note the scaling - worse for large vehicles in this case.

5
16.50 Lecture 18

Subject: Aircraft Engine Modeling; Turbojet engine.

All aircraft engines are Heat Engines, in that they use the thermal energy derived
from combustion of fossil fuels to produce mechanical energy in the form of kinetic
energy of an exhaust jet. The excess of momentum of the exhaust jet over that of the
incoming airflow produces thrust.

In studying these devices we thus employ two types of modeling.

a) Thermodynamic, in which the production of mechanical energy from

thermal is studied by the approaches of Thermodynamics. Here the change in
thermodynamic state of the air as it passes through the engine is studied. The
physical configuration of the engine is not identified. Rather the processes
are specified, by pressure and temperature ratios.

b) Fluid mechanical, in which we relate the changes in pressure,

temperature and velocity of the air, to the physical characteristics of the
engine.

With these ideas in mind, let's first outline a general approach to the modeling of
aircraft propulsion systems. Our general expression for thrust , in which we have a
main interest, is
. .
F = m eue - m ouo + Ae(pe - po) (1)
.
. . mf
where m e = (1 + f) m o ; f = .
mo
We write this more conveniently in dimensionless form as

F ue Aepo pe
. = (1 + f) uo - 1 + . (p - 1) (2)
o
mouo mouo

In our modeling of the aircraft engine we will often assume pe = po, and usually take
f<<1, so, using the flight Mach number M0=u0/a0, this expression becomes simply
F ue
. = M o [uo - 1] (3)
moao
but it should be recalled that just as for the rocket engine, the behavior of the nozzle
can be somewhat more complex. In practice the deviation from ideal expansion
becomes important for supersonic flight. In particular, there can be sonic or
supersonic underexpansion, with an exhaust pressure pe>p0. This happens in
particular with purely convergent nozzles, that are commonly used in subsonic
engines, when they operate off-design. It can also happen in a variable-geometry
supersonic nozzle that is not correctly adapted to the ambient pressure.

1
Our tasks in estimating F are then
.
a) to estimate m o
ue
b) to estimate u .
o

Many of the engines we deal with will have 2 exhaust streams. In this case we apply
(3) separately to each stream.

Let us begin with a Turbojet Engine, shown schematically below, and break the
engine into a set of Components with functions as follows.

Station Numbers
mf ma
0 1 2 3 4 5 6 7
e

uo ue

Nozzle
Diffuser πd Compressor Turbine
πc τc πt τt
Burner
πb Afterburner π
a

Image by MIT OpenCourseWare.

Adapted from Figure 1.4 Kerrebrock, Jack L. (1992).
Aircraft Engines and Gas Turbines (2nd Edition). MIT Press.

Diffuser (d) - Brings airflow from the flight Mach number Mo, to the axial
Mach number M2, required by the compressor.

Compressor (c) Raises temperature and pressure of airflow, as nearly

isentropically as possible.

Combustor (b) Raises temperature, nearly at constant pressure.

Turbine (t) Drops temperature and pressure, as nearly isentropically as

possible.

Afterburner (a) Heats air again, at nearly constant pressure ( to a higher

temperature than the turbine can tolerate without cooling).

Nozzle (n) Expands hot gases to produce a high-velocity jet. Station 7

denotes the sonic throat.

We first note that for this engine,

2
 ue M e Te
=
uo M o To

It is most efficient to find the exit Mach number and temperature by keeping track of
the stagnation temperatures and pressures through the several components. The
following procedure works for all aircraft engines, so it's worth your paying some
attention to the procedure itself, as well as the result. We employ the defining
relations for the stagnation properties:

# ! "1 2%
Tt = T 1+ M
$ 2 &
!
# ! " 1 2 % ! "1
pt = p$ 1 + M &
2

It is very helpful to define a set of symbols that represent explicitly ratios of these
stagnation properties and distinguish them from the static or thermodynamic
properties of the gas, because in general it is the stagnation properties that most
conveniently represent the effect of the components on the fluid as it flows through
the engine. Thus,

A ratio of pt's will be denoted by the symbol π;

A ratio of Tt's will be denoted by the symbol τ.
A ratio of a stagnation temperature to the ambient static temperature To
will be denoted θ and
A ratio of a stagnation pressure to the ambient static pressure po will be
denoted δ.
So for the flow upstream of the engine,
Tto ! "1 2
= 1+ Mo = # o
To 2

pto ! " 1 2 ! !"1

= (1 + Mo ) = # o
po 2
The turbine-inlet temperature is represented by
Tt 4
= !t
To
or, alternatively, by

Tt 4 Tt 4 ! t
!= = =
Tt 2 Tt 0 ! 0
which is more convenient for scaling purposes, since it relates two engine total
temperatures, a ratio that is often independent of ambient conditions.
For the compressor and for the turbine (both ideal by assumption),

3
 pt 3 Tt 3 #
pt 5 Tt 5 #

= ! c, = " c ; ! c = " c# $1 = !t, = " t ; ! t = " t# $1

pt 2 Tt 2 pt 4 Tt 4

Now let us use this system of notation to develop expressions for the Thrust and
Specific Impulse of the Turbojet Engine. We begin by tracking the changes of
stagnation temperature and pressure through the engine.

Temperature accounting:

Tte = Te (1+ ! 2"1 M e ) = To# o$ c $ b $ t = To# t $ t

2 (4)

Pressure accounting:
!
pte = pe (1+ ! 2"1 M e2 ) ! "1 = po#o$ c $ b $ t (5)

From (5), if pe = po(ideally expanded nozzle) and if πb ≈ 1

! "1
(1+ ! 2"1 M e2 ) = (#o$ c $ b $ t ) ! = % o& c & t (5b)

where the second equality assumes that the compression and expansion processes
are reversible adiabatics. From this we find an expression for the exit Mach number,
2 (6)
M e2 = [# o$ c $ t "1]
! "1
It is very important to realize that although this expression for the exit Mach number
is written in terms of temperature ratios, it comes from the pressure changes in the
engine. This is a general result, namely that the exit Mach number depends on the
ratio of jet stagnation pressure to the ambient pressure, not at all on the temperature.

If the exhaust is to be choked ( M ! 1), we must have, from (6),

e
2 (6b)
(# $ $ " 1) % 1
! "1 0 c t
which may not be satisfied at low power and/or low Mach number.

From (4)
Te !t" t ! !
= = t =
# $1
To 1 + M e2 ! o" c " c
2
So far these are quite general expressions applicable to any gas stream.
substituting them in our expressions for the velocity ratio and the thrust we have
2 #
[# o $ c $ t " 1]( t )
ue ! "1 # o$ c (7)
=
u0 Mo

Finally, the thrust per unit of mass flow (times the speed of sound to make it
dimensionless) is

4
 F 2 # (8)
= [# o $ c $ t " 1]( t ) " M0
m˙ 0 a0 ! "1 # o$ c

So far we have not made this peculiar to the turbojet engine, because we have not
included the relationship between the compressor and turbine. The fact that
distinguishes the turbojet engine from other engines we may consider later is that
the turbine power equals the compressor power, so

Tto(τc - 1) = Tt4 (1 - τt)

θ
τt = 1 - o (τc - 1) (9)
θt

So finally for the Turbojet Engine

F 2 # (10)
= [# t " # o ($ c " 1) " t ] " Mo
! "1 # o$ c
.
m o ao

.
We are also interested in the fuel consumption. We get m f from a combustor heat
balance,
. . C T
m f = m o p o (θt - θoτc) (10b)
h
so that the fuel-specific impulse, I = F / (m! g) is
f

I = ( hao ) (F / m˙ o ao ) (11)
gC p To (! t " ! o # c )

Discussion on nozzle choking Eq. (6b) was the condition for the exhaust to be at

least sonic, with p ! p or for the throat to be sonic. It involves both the compressor
e 0
and the turbine temperature ratios, but we can eliminate the turbine ratio using the
shaft balance (Eq. (9)), so that the condition is now
" #1 % + 1 (12)
! 0" c (1# c ) $
! 2
When the equal sign applies, we have Me=1 while still pe =p0. This limit can be
rearranged into a quadratic equation for τc :
$ +1 #
! c 2 " (# + 1)! c + =0
2 #0
with the two solutions
# +1 # +1 2 $ +1 #
! c +," = ± ( ) "
2 2 2 #0

5
It can be verified that τc must be between these two roots to ensure M ! 1 . The τc+ is
e
normally very high, so the relevant condition is τ> τc- . Values of τc- are tabulated
below:

θ=4 θ=6 θ=8

M0=0 1.296 1.253 1.234
M0=0.85 1.066 1.059 1.056

These are fairly low compressor ratios, even for stationary engine conditions, so the
assumption of a choked nozzle is a good one in general. Whether or not the nozzle is
also matched is a different question, as noted before.

6
16.50 Lecture 19

Subject: Turbojet engines (continued); Design parameters; Effect of mass flow on

thrust.

In this chapter we examine the question of how to choose the key parameters of the engine to
obtain some specified performance at the design conditions, and how the performance varies
if these parameters are changed, still at the design conditions. Later we will look at a
complementary question, namely, how the performance of a particular design changes when
conditions are different from design conditions.

With the results that we worked out last time for the Turbojet engine, let us look at the
.
dependence of F/m ao on the principal parameters, Mo, τc and θt. We can view them this
way
τc ! design choice (compressor pressure ratio)

Mo ! flight speed

θτ or θ ! combustor outlet temperature, an operating variable, limited by turbine

materials to some maximum value.

Assuming the exhaust is matched, we can re-write Eq. (10) of the previous lecture as

F 2 1
= [# t (1 " ) " # o ($ c " 1)]" Mo
! "1 # 0 $c
.
m o ao

we can see that since θoτc>1, F always increases with θt. This relationship is displayed
m˙ ao
below for a subsonic case:

1
For a given θt, what is the variation with τc? By inspection we see there is a maximum at the
maximum of the bracketed quantity, so at the value of τc that satisfies
! 1 # 1
[#t (1 $ ) $ # o (" c $ 1)] = t 2 $ # o = 0
!" c #o " c #o "c
This value is
(τc)max = ! t
!o
This result can be seen to be equivalent to Tt 3 = T0Tt 4 , namely, the compressor exhaust
should be at the geometrical mean of the ambient and combustor exhaust temperatures. If it
were much lower or much higher, the T-S diagram of the equivalent Brayton cycle would be
too “skinny”, and enclose too little area (too little work per unit mass):

(a) Too little compression (b) Optimal compr. (c) Too much compr.

T T T
Tt4 Tt4 Tt4
Tt3
Tt3 Tt3
T0 T0 T0

S S S
Whether this power is utilized as jet kinetic energy, as in the turbojet, or as shaft power in a
turboprop, is immaterial. Also, as far as this argument, the compression Tt3/T0 can be
arbitrarily divided between ram compression (θ0) and mechanical compression (τc).
What is the meaning of this for the turbojet? Putting this value into (10) and the
corresponding expression for the specific impulse, we have the thrust and I for engines
optimized for thrust per unit of airflow:

2
 F 2
= ( # t " 1) + Mo " Mo
2 2
( )
m˙ ao max F ! "1
(11)
(Ι)maxF = ( ao h ) (F /m˙ a0 )
C p To g (! t " ! t )

As an example take: θt = 6.25 ! t = 2.5 , γ = 1.4

F
)max F = 5(2.25) + Mo ! Mo
2
(
m˙ ao

(Ι)maxF = ao h m˙Fa o
( )
Cp To g 3.75

ao h (283m /s)(4.3x10 7 J /kg)

= = 6178s
C p To g (1004J /kgK)(200K)(9.81m /s2 )

There is an upper limit on Mo reached when the compressor outlet temperature equals the
turbine inlet temperature, so no fuel can be added. That is we must have θoτc < θt Since θ
0
increases with Mach number, the theoretical limit is reached when τ =1 and θ =θ , i.e. when
c 0 t

there is no compressor and we have a ramjet. But for τc = ! t the real limit is reached
!o
when θo = ! , i.e., all the compression is due once again to the ram effect, but we are
t
allowing a good margin for heat addition in the burner, so this ramjet does produce maximum
thrust. For a currently practical value of θ = 9, θo < 3 or Mo < 3.9
t

Propulsive Efficiency

The Turbojet engine is attractive for its simplicity and its good thrust behavior at high Mach
numbers. Unfortunately it is not very efficient at low Mach numbers, because its jet velocity
is too high. To see this we consider the Propulsive Efficiency, defined as
power # to # airplane
!propulsive "
power # in # jet
Fu 0 2m˙ (ue # u0 )u0 2u0
= = =
$u m˙ (ue # u0 )(ue + u0 ) (ue + u0 )
2 2
m˙ & e # 0 '
u
%2 2(
From this we see that there is a direct conflict between the desire for high jet velocity to give
high thrust, and jet velocity near the flight velocity, to maximize the propulsive efficiency.

3
In terms of our expression for thrust, since

F u
= Mo ( e ! 1)
m˙ ao u0

ue
= (F / m˙ ao ) / Mo + 1
u0
and we can write the expression for the propulsive efficiency in terms of our expression for
thrust
2
!propulsive = F
+2
m˙ ao M0
.
Since F/m ao ~ 2 to 3 for low Mo, ηprop is not good for the turbojet at low Mach numbers.
We will see later how this deficiency is remedied by adding a fan to the engine to produce a
Turbofan.

Thermal and Overall Efficiencies

The Thermal Efficiency is define for the Turbojet Engine as

$ u 2 u2 '
m˙ & e # 0
power # in # jet %2 2(
!thermal " =
power # in # fuel # flow m˙ f h

Finally we can define an Overall Efficiency as

power # to # airplane Fu0
!overall " =
power # in # fuel # flow m˙ f h

We see that
!overall = !thermal! propulsive

It is also important that the overall efficiency is directly related to the specific impulse:

Fu0 F gu0 gu0

!overall = = = I
m˙ f h m˙ f g h h

4
16.50 Lecture 20

Subject: Introduction to Component Matching and Off-Design Operation

At this point it is well to reflect on which of the many parameters we have introduced
(like M2, τc, τt, ϑt, f, etc.) are free for the pilot to control, and what the inter-relationships
are that determine the others. This connectivity is in part mechanical, like the shaft power
balance (Eq. 9 of Lecture 18), but it also comes via flow continuity among components.
This topic is usually relegated to the very end of the study of engine components, where
it is introduced under the rubric of “Component Matching” (Lecture 31 in our NOTES).
We find it advantageous to move most of it forward to this point.

The price to pay for the insight to be gained is the need to introduce one assumption at
this point (to be justified later). This is the assumption that the stators leading to the
turbine (the “turbine nozzles” are choked. This means the mass flow rate can be written
as
# " +1 &
Pt 4 A4 %! = " # 2 & 2(" )1)
(
m˙ = m˙ 4 = !(" ) % ( (1)
R Tt 4 % $ " + 1 ' (
$ '
Where A4 is the effective flow area of these nozzles. But in addition, we have already
shown in Lect. 18 that the exhaust nozzle is also choked. Passing the same flow through
two choked apertures in series imposes very strong constraints on the flow conditions.
The flow can be expressed at the throat as

Pt 7 A7
m˙ = m˙ 7 = !(" ) (2)
R Tt 7

and equating (1) and (2) ,

Pt 7 Tt 4 A4
= (3)
Pt 4 Tt 7 A7

For a non-afterburning turbojet, Pt 7 ! Pt 5 and Tt 7 = Tt 5

!t A
= 4 (4)
" t A7

#
and if the turbine is ideal, ! t = " t # $1 , and we obtain

2(( )1)
"A % ( +1
!t = $ 4 ' (5)
# A7 &

1
 2(
" A % ( +1
and then !t = $ 4 ' (6)
# A7 &

This is a strong result: as long as both, the turbine nozzles and the exhaust throat remain
choked, the turbine maintains the same pressure and temperature ratios (same operating
point), regardless of fuel flow, Mach number, altitude, etc. We can now trace the
variability of other quantities:

(1) Compressor ratios. In terms of ϑ = Tt4/Tto, Eq. (9) of Lecture 18 gives

! c = 1+ " (1# ! t ) ; ! c = " #c / # $1 (7)

Thus τc and πc do vary, but only as a function of the single quantity ϑ:

τc = τc (ϑ) for a given engine.

(2) Mach number at compressor inlet (M2). The flow at compressor inlet is
generally subsonic, so we express the flow rate there as
" +1
$ " + 1 ' 2(" #1)
Pt 2 A2 & )
m˙ = m˙ 2 = ! m2 ( M 2 ) ; m2 ( M 2 ) = M 2 & 2 (8)
R Tt 2 " #1 2 )
&1 + M2 )
% 2 (

The dimensionless flow function m2 (M 2 ) increases to a maximum of 1 when

M2 = 1, then decreases again.
Equating (8) to (1), we see that

m2 (M 2 )
1

Pt 4 Tt 2 A4
m2 (M 2 ) =
Pt 2 Tt 4 A2

0 1 M2

For an ideal combustor, Pt 4 = Pt 3 , and so, using ! c = " c# / # $1, Tt 2 = Tto ,

! "c / " #1 A4
m2 ( M 2 ) = (9)
$ A2

2
Since ! c = ! c (" ) , we see now that M 2 = M 2 (! ) as well (the supersonic solution for M2
given m2 can be disregarded).

(3) Dimensionless air flow. Returning to (8), we see that the dimensionless mass flow

m˙
m2 ! (10)
# P A &
%%" to 2 ((
$ R Tto '

(flow rate as a fraction of what the compressor would pass if its inlet were choked), is
once more a unique function of ϑ. This is very useful for scaling from one operating
condition to another.

(4) Fuel/air ratio. The combustor heat balance is

m˙ f h = m˙ c p (Tt 4 ! Tt 3 ) = m˙ c p Tt 2 (" ! # c ) (11)

and using Tt 2 = Tto and f ! m˙ f /m˙ ,

fh
= ! " # c (! ) (12)
c p Tto

so the quantity f /Tto is another function of ϑ alone. But notice that f itself does depend
on Mo at a fixed To.

(5)Throat pressure (normalized)

P7 Pt 2 Pt 3 Pt 4 Pt 5 Pt1 P7
=
Pto Pto Pt 2 Pt 3 Pt 4 Pt 5 Pt 7

P7 1 $ 2 '! / ! "1 P P
and = ! / ! "1 = & ) . Also t 3 = ! c = " c# / # $1, t 5 = ! t = " t# / # $1;
Pt 7 $ ! "1 ' % ! + 1( Pt 2 Pt 4
&1 + #1)
% 2 (

!
P7 $ 2 ' ! *1
=& " t " c (# )) (13)
Pto % ! + 1 (

which is yet another function of ϑ alone.

(6) Thrust (matched nozzle). We already have Eq. (10) of Lecture 18, but it is sometimes
better to normalize thrust by the total free-stream pressure on the compressor inlet, PtoA2,

3
which is known from flight conditions. If Pe = Po (variable nozzle, or just design point for
a fixed nozzle),

F m˙ ( ue ! uo )
!2 " = (14)
Pto A2 Pto A2

Pto A2 ue # uo
! 2 = m2"
RTto Pto A2

For ue, we go back to Lecture 18, Eq. (7):

2 #
ue = ao (# o$ c$ t "1)
! "1 $c

ao ! R/ To !
and = = . All together then,
RTto R/ Tto "o

# ' 2$ ($ o& c & t %1) *

! 2 = "m2 ) % M o, (15)
$ o () # %1 &c ,+

Here the quantities m2 and ! c depend on ϑ only, but we can see that the Mach number
" #1 2
Mo appears explicitly (as Mo and as ! o = 1+ M o ), so the normalized thrust ! 2
2
depends on both ϑ and Mo.

(7) Thrust (truncated sonic nozzle). We now have me = m7 = 1, but Pe = P7 > Po, so

F m˙ ( ue " uo ) + ( Pe " Po ) Ae
!2 = =
Pto A2 Pto A2
(16)
u " uo $ Pe Po ' Ae
= #m2 e +& " )
R Tto % Pto Pto ( A2

2 2!
and this time Me = 1, so ue = !RTe = !R Tt 5 = RTto"# t
! +1 ! +1

ue 2! P
so that = "# t depends on ϑ alone. Since we also know that m2 and e are
R Tto ! +1 Pto
functions of ϑ alone, it makes sense to separate out Eq. (16) in the form

4
 # ue P A & uo P A
! 2 = %%"m2 + e e (( ) " m2 ) o e (17)
$ R Tt o Pto A2 ' R Tt o Pto A2
& # )
( 2# & 2 ) # ,1 Ae + - # 1 A 0
! 2 = m 2" $% t + ( % t% c + , /"m2 M 0 + # / # ,1 e 2 (18)
( # +1 '# +1 * A2 + . $o $o A2 1
' *
!######"######$
! *2 ($ )

Once again, the normalized thrust depends on both, ϑ and Mo, but the structure is fairly
simple, and in particular, the portion ! *2 of ! 2 (neglecting the incoming momentum and
the external pressure) is a function of ϑ alone. This portion can be very easily scaled
between conditions, and the rest can be subtracted separately.

A note on ϑ: the near-constancy of the engine operating point

Two important points in the flight envelope of an aircraft engine are (a) Take-off
conditions ( M o ! 0.25, To ! 290K ) , and (b) End-of-climb conditions (M0≈0.85,
T0≈220K. The total temperatures are Tto = 290(1+ 0.2 ! 0.25 2 ) = 294K (take-off) and
Tto = 220(1+ 0.2 ! 0.85 2 ) = 252K (end of climb). Suppose the engine is dimensioned for
end-of-climb, which is common, and that the peak temperature Tt4, which will have to be
maintained for many hours of cruise, is selected at a conservative Tt4 = 1600K. We then
1600
have ! = = 6.35 at this condition. If we now decided to maintain ϑ = 6.35 also for
252
take-off, we would need then Tt 4 = 6.35 ! 294 = 1868K. While this is too high for long-
term operation (creep, corrosion), it may be acceptable for the few minutes per cycle that
the engine will be at take-off maximum power.

As a second example, consider a commercial jet in a long cruise. As the fuel is consumed
and the weight decreases, so must the lift L = 1/2 ! 0 u0 Awc L . Now, the lift coefficient will
2

be kept close to that for optimum L/D, and the Mach number M0 is unlikely to change
much, as it will stay just below the transonic drag peak, and so u02 will be proportional to
T0 due to the speed of sound variation. Together with the density part of lift, we can see
that the ambient pressure p0 must be decreasing in proportion to the airplane’s weight,
i.e., the plane must be climbing gradually. Turning now to the forward force balance,
given a constant L/D, the drag, and hence the engine thrust, must also be decreasing in
time in the same proportion as the ambient pressure. Therefore, from Eq. (14), the
nondimensional thrust ! 2 (" , M 0 ) will remain constant, and since M0 does too, the peak
temperature ratio θ will also remain constant, and with it all the important ratios like τc,
M2, etc.

In other words, ϑ may not vary much among (important) flight conditions, and the engine
will be operating at a fixed nondimensional condition (constant compression ratio,
nondimensional flow, compressor inlet Mach number, etc.). But of course, the

5
dimensional quantities (flow rate, peak pressure, etc.) will be different, depending on po,
etc.

(8) The Operating Line in the compressor map. Compressor performance is

typically presented as a map of ! c vs. m2 , with lines of constant normalized
rotational speed ! and "c superimposed. The details are the subject of later
Lectures, but the general shape is as shown below. (The flow and speed variables
are renormalized by the “Design” values):

m2
=
(m2 )des

Kerrebrock, Jack L. (1992). Aircraft Engines and Gas Turbines (2nd Edition).
MIT Press, © Massachusetts Institute of Technology. Used with permission.

Actually the “nominal operating line” shown in the figure is not a property of the
compressor, but rather of the rest of the engine. We can calculate this line with the
information we have now, before deciding what particular compressor to use. From Eqs.
(11) and (9),

A4 ! c" / " #1
m2 = (19)
A2 $

and from the shaft power balance (Eq. 7),

" c #1
!= (20)
1# " t

where we recall that ! t is fixed for a fixed geometry. Eliminating ϑ,

6
 A4 " / " #1 1# ! t
m2 = !c (21)
A2 ! c #1

or, in terms of ! c ,
A4 1" # t
m2 = !c $ "1 (22)
A2
! c
$
"1

which is the equation for the operating line (written in reverse).

If the compressor is already available, we see from (22) that we can adjust the nozzle area
A4 to place this line in a “good” place on the map, i.e., below the stall line and through
the best efficiency points.
T
Since m2 depends on ! = t 4 , varying Tt4 moves the operating point along the operating
Tto
line, and this is what the pilot does with the throttle stick to power the engine up or down.
At each selected ϑ, the engine settles to a ! c , a M2, a (normalized) rotation rate, etc.

Effects of Mach number

If we look at operation of a given engine at different flight Mach numbers, we may try to
maintain the same non-dimensional conditions throughout, which, as we have seen, can
be done by maintaining for example a constant compressor inlet Mach, M2. This, in turn
T
guarantees a constant ! = t 4 , but since now we have a varying Mach number, so that
Tto
Tt0 increases with M0, we may find that the turbine inlet temperature Tt4 needs to become
too high at the higher Mach numbers. For example, Tt4 would have to be 1.8 times higher
at M0=2.0 than at static conditions, and 2.25 times at M0=2.5.

A more reasonable assumption is that the ratio θt=Tt4/T0 can be maintained the same at all
Mach numbers, since at least in the stratosphere, T0 is almost invariant. The compressor
" (1# ! t )
temperature ratio now follows from ! c = 1 + t , where the numerator is a
"0
constant; thus, τc will be lowered as the Mach number increases, but less strongly than
would be required to maintain maximum thrust per unit flow ( ! c = " t /" 0 ). The flow
parameter m2 is now determined by Eq. (9), i.e. compressor-turbine flow matching, and
then the compressor-inlet Mach number from Eq. (8). Once these parameters are known,
we can use Eq. (15) to calculate the normalized thrust; since we are interested in the
effect of Mach number, it makes sense to re-normalize thrust by p0A2, or
F #
= ! 2" 0 # $1 .
p0 A2

A numerical example

7
We take now θt=7, or Tt4=1540K in the stratosphere. The geometry of the engine must
have been specified in advance. This means that the turbine temperature ratio (Eq. 5) is a
known fixed number. For the example, we select τt such as to obtain maximum thrust at
M0=1. From the shaft balance equation,
#0
! t = 1"
# t (! c "1)
and we put now θ0=1.2 and τc=√7/1.2=2.2048 (at M0=1). This fixes τt=0.7935.
Similarly, the area ratio A4/A2 must have been fixed, and we select it here so as to obtain
at M0=1 a compressor-face Mach number M2=0.5, which, from Eq. (8) implies
m2 =0.7464. From Eq. (9) then,
! c 3.5 " 0
m2 = 0.7464( )
2.2048 1.2
and the rest of the steps are as described above. The table below summarizes the results:

M0 0 1 2 2.5
θ0 1 1.2 1.8 2.25
τc 2.4458 2.2048 1.8032 1.6426
m2 0.9796 0.7464 0.4523 0.3648
M2 0.8486 0.5 0.2737 0.2172
!2 2.9117 1.5531 0.5795 0.3503
F/(p0A2) 2.9117 2.9399 4.534 5.985

We find that at a fixed altitude the thrust is nearly constant up to Mach 1, then it increases
rapidly. Actually the increase is less rapid than this simple model predicts, because of
losses in the supersonic flow in the engine inlet.
Finally to this point, we should note that an aircraft normally flies at increasing altitude as
the Mach Number increases, so that dynamic pressure !p0 M 02 is roughly constant. In this
case the change in F between Mach 1 and Mach 2 is actually a thrust reduction.

8
16.50 Lecture 21

Subject: Turbofan Engines

In the lecture 19 we saw that for low Mo, η p is not very high for turbojets. In essence
there is too much kinetic energy in the exhaust jet (per unit mass). This is the main
reason for use of the Turbofan engine. If it is designed for subsonic cruise flight it
looks like:

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

If designed for both subsonic cruise and for supersonic flight with afterburning it
looks more like:

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

In both of these diagrams the inlet has been greatly simplified, of course.

1
Let us now see how we can model these engines thermodynamically. In the first
(cruise engine) we have 2 jets. Up until we evaluated τt, our argument for finding
the jet velocity applies to either jet. Thus from equation (8) of Lecture 18 we get for
the core jet:

Fc 2 #
= ( t )[# o $ c $ t "1] " Mo
m˙ ao ! " 1 # o$ c
For the fan stream, there is no turbine, so let us repeat the argument.
γ−1
Tt8 = T8(1 + Μ82) = Τοθοτf
2
and
! !
γ−1 2 ! "1 ! "1
pt8 = p8(1 + Μ8 ) = po δο πf = po (θoτf)
2
So for p8 = po
γ−1
Μ82 = θoτf
1+
2
2
M8 = (# $ "1)
! "1 o f
As in the turbojet, a fixed convergent nozzle is more likely to be sonic and under-
expanded, namely, to have M8=1 and p8>p0, but we will here ignore the differences
this entails.

We also have
T8 ! o " f
= =1
To ! o " f
so
, /
FBP # u8 & .
2
) "1
* 0+ f " 1 1 ( )
= M 0 % " 1( = M 0 . "11
! ma
! 0 $ u0 ' . M0 1
. 1
- 0
adding this thrust to the thrust of the core jet, we find the total thrust:

F #
= { ! 2"1 ( #o t$c )[# o $ c $ t "1] " Mo} + % { ! 2"1 (# o $ f " 1) " Mo } (14)
m˙ ao
Now we need τt. We will see later that most engines have two shafts, but for now we
lump their power together; at off-design conditions, this would have to be modified.
From a work balance between the compressor, fan and turbine, (both shafts),

m˙ c p (Tt4 ! Tt 5 ) = m˙ c p (Tt3 ! Tt2 ) + " m˙ cp (Tt 7 ! Tt 2 )

! t (1" # t ) = ! o (# c " 1) + $!o ( # f " 1)

2
 #o
! t = 1" [(! " 1) + $ (! f " 1)]
#t c (15)

Substituting this gives us our result for the thrust of the turbofan. It doesn't really
help to carry out the substitution at this point. Instead let us think about how to
simplify the expressions to make them more easily understandable.

For this engine there are more parameters than for the turbojet:
θt, τc - as before
α, τf - characterizing the fan flow.

We can relate some of the parameters to others by noting that the highest propulsive
momentum
efficiency is realized when u8 = u6, since this maximizes the ratio of energy in
the jets. For this situation, the two ' s in the thrust equation are equal, and this
requires that:
# !t &
% ( [! " " ) 1] = ! o " f ) 1 (16)
$ ! 0" c ' o c t
(u8 = u6)
Solving for τf as a function of α using (15),

!o " c ! o2" c
!t (! o " f # 1) + 1 = ! o " c # [( " c # 1) + $ ( " f # 1)]
!t
!t
! o " f # 1+ = !t # ! o [(" c # 1) + $( " f # 1)]
!o "c

1+ " t # ""o !t c # " o (! c # 1) + "o $

!f =
"o + " o $

1+ " t + " o (1+ # $ ! c ) $ ""o t!c

!f = (17)
" o (1+ # )
When this equation is satisfied, the thrust equation becomes simply
F
m˙ ao
[
= (1 + ! ) " 2#1 ($ o % f # 1) # M0 ]
Now as for the turbojet there is still the choice of the compression ratio. Generally
we want to choose it for maximum power, which in this case for given turbine inlet
temperature θt and bypass ratio α means maximum thrust. To maximize F, we
choose τc to maximize τf, since F increases monotonically with τf. From the
expression for τf

!" f $
= #$ o + t 2 = 0
!" c $o " c

3
 "t "t
!c2 = (! c ) F max =
"o 2
"0
Notice that this is precisely the same result as for the Turbojet. Substituting in the
expression for τf
1+ " t + " o (1+ # ) $ " t $ " t
(! f )max F =
" o (1+ # )

( " t # 1)2
(! f )max F = +1 (17b)
" o (1+ $ )
! F $ + 2 + ( *t ) 1)2 . .
# & = (1+ ' )- + * o ) 10 ) M o 0 (14b)
-
" m˙ a0 % max , ( ) 1 , 1+ ' / /

Now we have 3 parameters,

θt - which we set at the maximum feasible value
Mo - flight speed
α - the prime variable distinguishing the turbofan

As before
2 2
!propulsive = u = F / m˙ a
6
+1 0
+2
u0 M0 (1 + " )

The variation of F/ma0 and ηp with M0 and α is shown in the figures below for
θt=6.25. Of course for the higher bypass ratios we are really only interested in the
range of M0<1, but the lower bypass and higher M0 range may be of interest for a
supersonic transport.

4
5
16.50 Lecture 22

Subjects: Inlets or Diffusers

While the Gas Generator, composed of the compressor, combustor and turbine, is
the heart of any gas turbine engine, the overall performance of the propulsion
system is strongly influenced by the inlet and the nozzle. This is especially true for
high Mo flight, when a major portion of the overall temperature and pressure rise of
the cycle are in the inlet, and a correspondingly large part of the expansion in the
nozzle. So it is important to understand how these components function and how
they limit the performance of the propulsion system.

Inlets or Diffusers

These two titles are used interchangeably for the component that captures the
oncoming propulsive streamtube and conditions it for entrance to the engine. The
function of the inlet is to adjust the flow from the ambient flight condition, to that
required for entry into the fan or compressor of the engine. It must do this over the
full flight speed range, from static (takeoff) to the highest Mo the vehicle can attain.
Comparing the simple diagrams of subsonic and supersonic inlets, we can
appreciate that the subsonic inlet has the simpler task:

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

For any inlet the requirements are two:

a) To bring the inlet flow to the engine with the highest possible stagnation pressure.
This is measured by the Inlet Pressure Recovery, πd .

b) To provide the required engine mass flow. As we shall see the mass flow can be
limited by choking of the inlet.

1
Subsonic Inlets:

The subsonic inlet must satisfy two basic requirements:

a) Diffusion of the free-stream flow to the compressor inlet condition at cruise.

b) Acceleration of static air to the compressor inlet condition at takeoff.

There is a compromise to be made because a relatively thin "lip" aligned with the
entering flow is best for the cruise condition. This is to avoid accelerating the flow,
already at a Mach number of the order of 0.8, to supersonic speeds that will lead to
shock losses. But a more rounded lip will better avoid separation for the takeoff
condition because the air must be captured from a wide range of angles:

The minimum area, A1 is

set to just avoid choking (M
= 1) in the inlet when M2
has its largest value (this is
set by the maximum blade
Mach number and the
blade shape).

A well designed subsonic

A1 inlet will produce a
stagnation pressure
recovery πd in the order of
0.97 at its design condition.

From Kerrebrock, Jack Aircraft

L. Engines and Gas Turbines
.
2nd edition. MIT Press, 1992. © Massachusetts Institute
Supersonic Inlets:
of Technology. Used with permission.
At supersonic flight speeds
the pressure and temperature rise in the inlet can be quite large. For the best
thermodynamic efficiency it is important that this compression be as nearly
reversible (isentropic) as possible. At a flight Mach number of 3 the ideal pressure
ratio is
p ! "1
!

δο ≡ to = (1+ 2 Mo ) ! "1 =36.7

2

po
while the temperature rise is
T ! "1
θo = to = 1+ 2 Mo =2.8
2

To

For the turbojet cycle the compression is partly in the inlet and partly in the
compressor:

2
 Note:
T3-T2=T4-T5

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

If the diffusion from point 0 to point 2 is not reversible, the entropy increases, and
this results in a lower value of Pt2. This has two effects:

a) The expansion ratio of the nozzle is decreased, so the jet velocity and thrust are
lower.

b) The lower pressure at point 2 limits the mass flow through the compressor to a
lower value than for isentropic diffusion. So there is a double penalty for losses in
the inlet.

Unfortunately some losses are inevitable, the more so the larger Mo. To see why we
will discuss the flow in supersonic diffusers, beginning with the simplest.

3
Normal-Shock diffuser:

All existing compressors and fans require subsonic flow at their inlet with 0.5 < M2 <

0.8 at high power conditions. So the inlet must reduce the flow Mach number from
Mo > 1 to M2 < 1. The simplest way to do this is with a Normal Shock.

Here M1 < 1 is entirely determined by Mo, according to the normal shock relation

1+ ! 2"1 Mo 2
M1 2 =
!Mo 2 " !2"1

The stagnation and static pressures are also determined by Mo:

1
pto 2! Mo 2 ! " 1 ! "1 ( ! " 1)Mo 2 + 2 ! !" 1 p1 2!
=( " = 1+ (M 0 "1)
2
) ( 2 )
pt1 ! +1 ! +1 (! +1)Mo p0 ! +1

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

For low supersonic speeds, such diffusers are adequate because the stagnation
pressure loss is small, but at Mo = 2, pt2/pto ≈ .71, a serious penalty, and at Mo = 3
p2t/pto ≈ .32. For example the F-16 fighter has a simple normal shock diffuser, while
the F-15 has an oblique shock diffuser such as will be discussed next.

4
Oblique - Shock diffusers:

The losses can be greatly reduced by decelerating the flow through one or more
oblique shocks, the deflection and the pressure rise of each being small enough to be
in the range where the stagnation pressure ratio is close to unity. It is very
important to understand that an Oblique Shock is in fact just a normal shock
standing at an angle to the flow.

All of the change across the shock takes place because the Mach number normal to
the shock is larger than unity. The velocity normal to the shock decreases with
consequent increases in temperature and pressure, but the velocity parallel to the
shock is unchanged:

M1n is given in terms of Mon by the same relation given for M1 as a function of Mo.
But Mon can be made close to 1. The condition for a weak or sound wave is just Mon
= 1, or
1
Sinθ = M (weak compression)
o
By choosing the wedge angle (or deflection angle) δ we can set the shock angle.

A series of weak oblique shocks, for each of which the Mn is near unity, hence all
lying in the range of small pt loss, can yield an efficient diffuser.

5
Diffusers with internal contraction

One might ask why we do not just use a convergent-divergent nozzle in reverse as
an inlet:

This would work at one design Mach number, the one for which the isentropic area
ratio between the incoming supersonic flow and the sonic throat is exactly the as-
built area ratio A1/Athroat . But during the acceleration to this Mach number the fully
supersonic flow cannot be established in the inlet without varying the geometry. To
see this, imagine the inlet flying at Mo , lower than the design Mach number. The
flow will look as depicted in the top right diagram below. This is because at the lower
M0 the flow area that would decelerate isentropically to sonic at the throat is smaller
than the built A1:
A/A*
A1/A*
A0/A*

0 1 M0<MD MD
So, if the flow did arrive undisturbed at the inlet, it could only occupy a fraction of
it, the rest of the flow into the frontal area A1 having to be somehow “disposed of”.
This “spillage” is accomplished by the detached normal shock; behind it the flow is
subsonic and it can turn around the inlet. Unfortunately, the shock at the full flight
Mach number is very lossy, and it is not practical to simply force the plane to
continue accelerating to the design condition (there may not even be enough thrust
left to do it). What can be done is to manipulate the geometry to swallow the shock
and reduce its strength. This is called “starting” the diffuser.

6
 From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

To "start" the diffuser, meaning to pass the shock through the convergent portion,
we must increase the throat area until the normal shock is just at the lip. At that
point, any further small increase in throat area causes the shock to jump rapidly to a
position in the divergent part of the nozzle where the area is again A1; this is because
the shock is unstable in the converging section, but stable in the divergent section.
So far nothing has been accomplished, because the shock is still as strong as it was.
But if the pilot now reduces fuel flow, the compressor moves down the operating
line to a lower value of m2 and since the actual flow rate ṁ does not change any
more, this means a higher total pressure pt2 into the compressor, and hence a weaker
shock. This is accomplished by the flow through a repositioning of the shock to a
location nearer the throat, on the supersonic side. The process can continue until the
shock is almost at the throat (if too near, it may accidentally pop all the way to the
front (an “unstart”, which is a violent transient event). The final condition is
depicted in the diagram at the bottom right.

This is an adequate solution for modest values of Mo, but for Mo > 1.5 the loss due to
the shock at the throat is still excessive.

7
16.50 Lecture 23

Subject: Exhaust nozzles

Like the inlet, the nozzle can range from very simple to quite complex. A simple
fixed-area convergent form usually suffices for subsonic aircraft except when jet
noise suppression is required, while a complex variable-area convergent-divergent
device is essential for adequate performance in supersonic aircraft. In either case,
the functions of the nozzle are two:

a) Provide the required throat area to match the mass flow and exit conditions
of the engine.

b) Efficiently expand the high-pressure, high-temperature gases at the engine

exhaust to atmospheric pressure, converting the available thermal energy to
kinetic energy.

Throat Area:

The required throat area is determined by conservation of mass. We had from

Lecture 20
p A
ṁ = !m2 t 2 2
RTt 2
and if the throat (station 7) is choked,

pt 7 A7 p A
ṁ = ! = ! t5 7
RTt 7 RTt 5

"
pt 5 T
Equating and using = (! c ! t ) " #1 and t 5 = !" t , we can calculate
pt 2 Tt 2
A7 m (M )
= " 2 2" +1 $
A2 ! " #1 ! 2(" #1)
c t

The ratios in the expression for A7/A2 are determined by cycle analysis, as outlined
previously, so from this expression we can find the nozzle area ratio as a function of
the engine parameters. As we have seen in our discussion of the matching of
components, once the nozzle area is set, the operating point of the engine depends
only on the turbine temperature ratio θ.

Exit Area:

The ratio of exit area to throat area required for ideal expansion can be found
from the usual compressible channel flow relations. Thus the exit Mach number is
set by the stagnation pressure of the jet and the ambient pressure

1
 pte pte !
= = (1+ ! 2"1 M e ) ! "1
2

po pe
pte
where = ! 0" c " t
p0
and the area ratio is then set by this Mach number,
1 1+ ! 2"1 M e 2(!!+1"1)
2
Ae
= [ ! +1 ]
A7 M e 2
Since pte/p0 involves θ0, Me for this matched condition does depend on flight Mach
number, and so does then Ae/A7. This is important for supersonic engines, as we
discuss below.

These relations taken together serve to define the geometry of an ideally expanded
nozzle. If the nozzle is not ideally expanded, the behavior is quite like that of the
rocket nozzle at off-design conditions, as discussed earlier.

Effects of nozzle mismatching: Subsonic vs. supersonic

We can use Eqs. (15) and (18) of Lecture 19(b) to calculate the thrust of engines
whose nozzles are respectively pressure-matched or truncated at the sonic point. We
illustrate this for two different engine designs, one subsonic and one supersonic:

Case 1: M0=0.85 (θ0=1.1445)

Case 2: M0=2 (θ0=1.8)

For both cases, we take θt=6.25 and M2=0.6 (or m2 = 0.8416 ). We also assume the
compressor ratio is that which gives maximum thrust in each case ( ! c = " t /" 0 ).
The results are shown below:

M0 τc τt AThroat/A2 Ae/A2 (! 2 ) Matched (! 2 )Truncated

0.85 2.1844 0.7831 0.2658 0.2700 0.9682 1.4991
2 1.3889 0.8880 0.7093 0.7190 0.9682 0.7666

There is little effect (8%) on thrust for the subsonic nozzle when the nozzle is
truncated at its throat, but the effect is a 26% thrust reduction in the supersonic case.
This is why such engines carry an adjustable convergent-divergent nozzle.

Additional Requirements:

The exhaust nozzle may have to meet a number of other requirements. Some are:

a) Variable area for afterburning, to increase throat area in proportion to the square
root of the temperature after afterburning:

2
 From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

Such nozzles have been built in a number of forms. At the top is the relatively
simple type used on the F-111 and F-15 engines, which have top flight Mach
numbers of the order of 2. At the bottom is the type used on the SR-71, which
reaches 3.5 or thereabouts.

b) Noise suppression

As we shall see in Lecture 37, the principal way to decrease jet noise is to lower the
jet velocity for a given thrust. The turbofan is the most efficient way to do this, but
for some applications it is not practical to use a fan. In this case there is the desire to
increase the mass flow rate of the jet, by mixing in additional air that lowers the
velocity for a given total momentum. One way to do this is the lobed mixer:

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

Lobed mixers are also used sometimes in turbofans to help equalize the core and
bypass stream velocities, and hence increase performance (by 3-4%). The
improvement is, however, less than if the velocity changes were isentropic.

c) Thrust-vectoring, as for advanced VTOL aircraft.

3
16.50 Lecture 24

Subject: Compressors and Fans

These are fluid dynamic devices, i.e., they depend on fluid accelerations to compress
and expand gases, in contrast to positive displacement devices such as the familiar
piston engine. The moving blades exert a force on the fluid by virtue of their motion,
resulting in an added velocity of the fluid, and an increase in its energy in stationary
coordinates. It is the fact that the force is dependent on the blades' velocity, roughly
as the square, that makes this a dynamic machine. In contrast, the energy per unit
mass added to the fluid in the positive-displacement piston-cylinder mechanism is
nearly independent of the speed of the piston, depending only on the volumetric
compression ratio.
Positive Displacement Fluid Dynamic

ωr

Image by MIT OpenCourseWare.

Large aircraft engine compressors or fans are mainly of the axial-flow type, with
rotating rows of rotor blades, separated by stator blades. A cross-sectional view
through the axis looks like

Meridional Stream Surface

R S R S R S

Image by MIT OpenCourseWare.

Both rows of blades function to deflect the flow in the tangential direction, the rotors
adding angular momentum in the direction of the rotation, the stators removing it.
Because of their motion, the rotating blades can do work on the flow, increasing its
energy. The stator blades in contrast only diffuse the flow, exchanging momentum
for pressure rise.

1
In the sketch the surface of rotation defined by the dashed line is termed a
meridional stream surface. On this stream surface, the cross section of the blades
looks like:

ωr

Rotor

Stator

Image by MIT OpenCourseWare.

To simulate the rotational symmetry of the actual blade rows we think of a cascade,
extending to infinity in both directions. In this limit the rotor blades then have a
constant velocity and the stators are stationary.

To describe the effect of the blades, Velocity Diagrams are used, which show how
the velocities change across the rotor and stator blade rows, and in particular the
effects of the relative motion of the two:
ωr

V’2 V1
V1 V’1
V’1 V2
(absolute)
(relative)
ωr

Rotor Composite Velocity Triangle

V2
V’2 (absolute)
(relative)

ωr

Stator

V3
(absolute
)

If the velocities at 3 are the same as at 1 then we can put another identical stage after
this one.

The energy exchange all takes place in the rotor blade rows. The force on a blade F is
equal to the rate of change of tangential momentum per blade:

2
 .
F = (m per blade) (v2 - v1)

where v is the tangential component of V at any point.

The power delivered to the fluid by a blade is the force times the blade velocity and
this must equal the change in fluid energy per blade, across the rotor, so

Power = F ωr
.
Power per blade = (ωr) (m per blade) (v2 - v1)
.
= (m per blade) cp (Tt2 - Tt1)

cp (Tt2 - Tt1) = ωr (v2 - v1)

This is the famous (and important) Euler Turbine Equation.

The pressure generally rises across both rotor and stator blade rows. If the flow
relative to the blades is subsonic we can see this easily from Bernoulli's equation:

A2 > A1 (because of the turn towards axial)

V 2 < V1
1 1
p1 + 2 ρV12 = p2 + 2 ρV22
p 2 > p1

The energy exchange across the rotor is most generally understood as follows. We
focus on a streamtube passing through the rotor, with a mass flow δm:

The torque on the rotor due to this stream tube is

.
δΤ = δm(r2v2 - r1v1)
and the power is
.
δPower = ωδΤ = δm ω(r2v2 - r1v1)
Equating this to the total enthalpy rise of the fluid
.
δPower = δm cp (Tt2 - Tt1)
cp (Tt2 - Tt1) = ω(r2v2 - r1v1)

3
This is the Euler Turbine Equation in a general form, although the left hand side is
even more generally the total enthalpy rise in the streamtube across the rotor,
whether or not the gas can be modeled as being ideal.

The Euler Turbine Equation applies to other geometries than the axial flow one, e.g.
the centrifugal or radial compressor

ωr

r2

Image by MIT OpenCourseWare.

In this case v1 ≈ 0 and if the vanes are radial, v2 ≈ ωr so we have

cp (Tt2 - Tt1) = (ωr2)2

For the axial compressor, r2 ≈ r1, and we have

cp (Tt2 - Tt1) ≈ ωr(v2 - v1)

Tt 2 (# r)2 $ v2 ! v1 '
!1 " & )
Tt1 c pTt1 % # r (

! R T1 (" r)2 $ v2 # v1 '

= & )
c p Tt1 ! RT1 % " r (

2$ v ! v
! 1 = (" ! 1) 1 MT % 2 1 &'
Tt 2 T
Tt1 Tt1 #r

4
16.50 Lecture 25

Subjects: Velocity triangles; Compressor performance maps

In the last lecture we discussed the basic mechanisms of energy exchange in

compressors and drew some simple velocity triangles to show how we go from the
stationary coordinate system to one in the moving blades. In more detail, the
velocity triangle is:

Now we can write

v1 = w1 tanβ1

v2 = w2 tanβ2 = ωr - w2 tanβ2'

where β2' is the flow angle in the rotating coordinate system of the rotor blades.
The advantage of the latter form for v2 is that it is β2' which is nearly set by the
shape of the blades, i.e. β2' is nearly constant. The difference between the flow angle
and the angle of the chord line at the trailing edge of the blades is called the
deviation, and is usually a small number of degrees.

Taking advantage of this small deviation, we can write the Euler equation as

Tt 2 (! " 1)MT 2 & w2 tan $ 2# w1 tan $ 1 (

=1 + 1" "
Tt1 1 + ! 2"1 M1 2 ' %r %r )

where M1 is the flow Mach number upstream and MT is the tangential Mach
number of the blade motion.

Since the two flow angles expressed this way are nearly constant, the temperature
ratio becomes a function primarily of two variables, the tangential Mach number MT
and the ratio of the axial flow velocity to the blade speed. For the usual case of β2' >
0, β1 > 0 we get a functional dependence something like:

1
 Tt2
Tt1 - 1
2
MT

w
"
!r

Actually, we usually plot this in terms of the pressure ratio rather than the
temperature ratio, and include superimposed lines of efficiency, so that the map
looks like:

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

Here the compressor efficiency is defined as:

# $1
p #
Tt1 ( t 2 ) $ Tt1 # $1

!c =
("Tt )Ideal
=
pt1
=
( p t 2 / p t1 ) # $ 1
("Tt )Actual Tt 2 $ Tt1 Tt 2 / Tt1 $ 1

2
Since ηc<1, the total pressure rise for a given total temperature rise is less than it
could be:
% %
% $1
! c = [1+ "c (# c $ 1)]% $1 < # c

By decreasing β1 and β2', or even making them negative, we can increase Tt2/Tt1.
What limits this increase? The answer is best given in terms of a Diffusion Factor,
which describes the tendencies for the boundary layer to separate under the
influence of the pressure rise in the blade passage.

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

The critical region is the suction surface of the blade. This region feels a pressure
rise due to the decrease of V’ from V’1 to V’2, (the prime is a reminder that these
velocities in the relative frame) and also due to the acceleration, followed by
deceleration on the suction surface. The Diffusion Factor, defined as

V2" v2 ! v1 c
D = 1! + ; #$
V1" 2# V " s
1
is a crude, but effective way to account for these two flow deceleration components.
Here c is the chord of the blades and s is their spacing (in the peripheral direction).

A value of about 0.5 for D is the upper limit for good efficiency. It is conventional
and useful to represent the loss in the blading in terms of a Loss Factor, defined as

pt 1 " pt 2
!1 =
pt 1 " p1
3
We then find that we can correlate the losses in the form

! cos # 2" cos # 2" 2

( )
2$ cos #1"

.06

.04

.02

0 .1 .2 .3 .4 .5 .6 .7

Diffusion Factor, D

Image by Manuel Martinez-Sanchez. Adapted from Fig. 5.12 in Kerrebrock,

Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT Press, 1992.

Compressor limitations. Rotating Stall. Surge.

As the incidence angles increase in the compressor a point is reached when the
flow begins to separate from the blades, with serious consequences. Large
incidence angles occur when the ratio (axial velocity)/(blade speed) is low, i.e.,
for a fixed rotational velocity, at reduced flow, and hence at increased pressure
ratio. All compressors have a fairly well defined “stall line” that runs diagonally
from low flow and low ! c , to high values of both. Operation must be restricted
to the region below and to the right of this line. The operational line for the
compressor runs roughly parallel to this stall line. As we saw in Lecture 20, The
reason for the compressor to be restricted to his operation line have to do with
flow continuity between it and the turbine and nozzle. It is a fortunate fact that
the line thus defined tends to correspond to a constant blade incidence, which,
when chosen properly, is close to that which optimizes blade performance, and
so the operating line is more or less the peak compressor efficiency line.

The phenomena that occur at and beyond stall are complex, dynamical and
highly nonlinear. A detailed understanding of these effects has only emerged in
the last two decades, and remedial measures based on this understanding are
still under evaluation. Much of the pioneering work in this field has been done at
MIT (see E. Greitzer, Engineering for Power, 98, 2, (1976) and J. of Fluids
Engineering, 103 (1981), Paduano et al., ASME Paper 91-GT-87, 1981). A more
complete exposition of the principles is given in Kerrebrock’s book, (Aircraft
4
Engines and Gas Turbines, (MIT Press, 1992, Sec. 5.7, pp. 3254-266), a short
summary of which is presented below.

In most axial compressors, incipient separation first leads to the Rotating Stall
phenomenon: sections of the stalling rotor then operate in deep stall, with almost
zero flow, while the rest carries normal or high flow per unit frontal area. These
regions move backwards in the rotating frame at ~0.4 – 0.6 ωr, so that, when
observed from rest they rotate forward, but only at a fraction of the rotor speed.
The reason for the bimodal flow distribution is that adjacent streamtubes become
unstable with respect to flow exchange: if the rotor as a whole is near stall
conditions, and one streamtube loses some flow and diverts it to its neighbors,
the streamtube with less flow goes into stall and loses even more flow, while the
neighbors gain flow and remain stable. The reason the “stall cells” move
backwards relative to the rotor is an elaboration of the same argument: if one
flow passage stalls, the swirling incoming flow is re-routed such that the passage
ahead of the one stalled sees a more axial flow, while the one behind sees a larger
incidence angle. The stall moves to this trailing passage, and the stalled passage
clears.

Rotating stall, since it moves rapidly about, tends to average out and, aside from
high-frequency excitation, may not be dynamically significant. On the other
hand, the net compressor performance drops strongly, and in addition, it is not
easy to reverse once started, except by stopping and re-starting the engine.
Detecting rotating stall and instituting the appropriate control reaction is very
important. If the engine controls simply detect a loss of pumping performance
(pressure loss) they may react by increasing fuel flow, which combined with the
reduced airflow, may lead to overheating and burnout.

Under some conditions having to do mainly with the ratio of flow inertia to flow
passage capacitance, the complete “pumping system” (compressor plus choked
turbine nozzles) can enter a global oscillation, called Surge. Unlike rotating stall,
surge involves deep oscillations or even reversals of the whole flow through the
compressor, and can be mechanically destructive (certainly quite noticeable, in
the form of loud, repeated bangs, accompanied by flame ejection from both ends
of the engine). When stall is reached, the engine may go into either Rotating Stall
or Surge. The detailed mechanisms that determine which of the phenomena will
prevail are encapsulated in Greitzer’s “B parameter”
!r V p
B=
2a Vc
where a is the speed of sound in the burner, Vp is the “plenum” volume (mainly
the burner volume) and Vc is the volume in the compressor flow passages. In a
single-stage compressor, values of B below ~ 0.8 lead to rotating stall, while
higher values lead to surge. For multistage (N) compressors, Bcrit. is lower by
somewhere between N and N. One favorable aspect of surge (as opposed to
rotating stall) is that it can usually be cleared by simply reducing fuel flow (or, in
a test stand, opening the downstream throttle).

5
16.50 Lecture 26

Subjects: Compressor Blading; Design; Multi-Staging

Compressor blading; Radial variations

In all of the discussion so far the blade speed has been taken as a parameter. In fact the
blade speed varies with the radial location in the compressor, so the “designs” that we
have discussed are applicable only at one radius in any real compressor. In practice it is
usual to begin a compressor design with such a treatment, called a “mean line design”
carried out at some mean radius. But the effects of the blade speed variation with radius
are important, so we must be aware of the constraints they impose on the design.

Normally it is desirable for the pressure (or temperature) ratio of a rotor blade row to be
approximately constant over the radial length of the blade, so that the outlet airflow has
uniform pressure. From the Euler equation
Tt 2 " ( r2# 2 ! r1#1 )
!1 =
Tt1 c pTt1
Supposing for simplicity that !1 = 0 (no inlet guide vane), we see this condition requires
const
r2! 2 = const.; ! 2 =
r2
This implies that the rotor blade row should generate a Free Vortex in the flow

1
Now let us draw the velocity triangles for tip and hub, assuming rH/rT = ½, and w1 = w2 =
const = !rT /2.

V'1
V1 V2 V'2 V1 V'2
V'1 V2 50% reaction at hub

ωrT ωrH

v2r = 1 v2H = 2ωrH v2H

2

Image by MIT OpenCourseWare.

Here it has been assumed that ! 2 ( rH ) = "rH , (that is that ! '2 = 0 ) and it follows that

! 2 ( rT ) = "rH /2 = "rT /4

The approximate blade shapes as sketched are determined by the condition that the
leading and trailing edges are aligned with the flow. We see that the blades are strongly
“twisted” from hub to tip.

Now let us see what these variations imply for the Diffusion Factor, D.
Taking v1 = 0,
V" # !#
D = 1! 2 + 2 1
V1" 2$V1"
Remember that these are relative to the rotor blade.

From the geometry,

TIP HUB

! 2 " !1 = #rT /4 ! 2 " !1 = #rH = #rT /2

V1! = "rT 1+ 1/4 = 1.12"rT V1! = "rH 2 = "rT / 2

# 1 & # 3 &2
V2! = "rT % (2 + % ( = .901"rT V1! = "rH = "rT /2
$2' $ 4 '

1
.901 1/4 1 2
DT = 1 ! + DH = 1! +
1.12 2(1.12)" T 2 2/ 2 "H ( )

2
 .116 353
DT = .195 + DH = .293 +
!T !H

The solidities at hub and tip, ! H and ! T are of course design choices, but since the
spacings SH and ST are related by
S H rH
= .
ST rT
! H C H ST C H rT
= =
! T CT S H CT rH

Let us choose ! T = 1, CH/CT =1, then ! H = 2

DT = .195 + .116 = .311

.353
DH = .293 + = .470
2
Both of these are acceptable from the viewpoint of losses, but DH barely so. It is
generally true that the hub section of the blade is limiting from the viewpoint of diffusion.

Mach Number Effects:

Typically the first stages of modern aircraft engine compressors operate with axial Mach
number ≈ 0.7 and blade tangential Mach number MT ≈1.3 giving
( M1 )T = M12 + MT2 = 1.48
so the tips of the blades see supersonic flow. For rH/rT = 0.5, ( M1!) H = .49 + .42 = .95 ,
so the hub is at a very high subsonic speed. Just as for inlets, hub and tip therefore require
quite different diffusion techniques. At the hub, blading that is the equivalent of a
subsonic inlet is called for

At the tip, the blades must be designed to minimize shock losses.

3
Multi-staging

Because ρ increases, the flow area must decrease as we go through the compressor. We
can choose to effect this area variation by decreasing the tip radius, by increasing the hub
radius, or something in between, like keeping a constant mean radius. All these solutions
have both advantages and disadvantages.

The constant outer diameter design maintains the blade velocity at the maximum for the
entire length of the compressor, hence gives the highest pressure ratio for a given number
of stages. But because the length of the blades becomes small at the exit end of the
compressor, leakage through the tip clearance is most serious for this design choice.

The constant inner diameter design minimizes the tip clearance problem and also yields
lower stresses in the discs of the last stages, but requires more stages for a given pressure
ratio.

The Polytropic Efficiency

What is most nearly constant among stages of a multi-stage compressor is the limiting
isentropic efficiency for small compression, which is called the Polytropic efficiency:
1
dp
(dht ) S # t t RTt d ln pt $ %1 d ln pt
!Poly " = = =
dht c p dTt c p dTt $ d lnTt
If we assume ηPoly remains constant along the compression path, this definition can be
integrated to
! "1
Tt,out p !#
= ( t,out ) Poly
Tt,in pt,in
# $1
#%Poly
or ! c = " c . Inserting this into the definition of the normal Isentropic Efficiency
allows one to compute it in terms of ηPoly and the finite pressure or temperature ratio:
$ %1 $ %1
# $ %1 # c $ %1
!c " c = $ %1
& c %1 # $! poly %1
c

As an example, take a multi-stage compressor with an overall pressure ratio πc=16 and
assume the polytropic efficiency is constant and equal to 0.9 in all stages. Its isentropic
efficiency (ideal work required divided by actual work) is then calculated to be ηc=0.856.
This is less than ηPoly, the small-increment efficiency, because the last stages of the
compressor receive gas that is hotter than it should, due to the inefficiencies in the
previous stages, and hence more work is needed to compress it. It could be mentioned
here that a similar argument can be made for turbines, and in that case one finds ηPoly>ηc,
because the extra thermal energy due to inefficiencies of the early stages is now available
for conversion to work by the latter stages.
One warning: the isentropic efficiency of a compressor with N stage, each with a stage
efficiency ηs (close to ηPoly) is not ηsN.

4
Starting and Low-Speed Operation

Because the density variation through the compressor is much less at low speed
conditions, the compressor develops adverse blade loading situations in both the inlet and
the outlet stages at low operating speeds. As shown in the sketch, the axial velocity in the
inlet stages tends to be lower, and that in the outlet stages higher, relative to the blade
speed, than at design, so the front stages tend to be stalled and the rear ones to
“windmill”. This makes it difficult to design a compressor that will operate satisfactorily
over a wide range of speeds.

The solution to this problem has taken two forms: one is to divide the compressor into
two or even three “spools”, operating at different speeds, the other is to use stator blades
of variable angle, to adjust the flow direction at low speeds. Most modern high pressure
ratio engines use both.

5
16.50 Lecture 27

Subjects: Turbines; Stage characteristics; Degree of reaction

Turbines behave according to the same principles of dynamic energy exchange as

compressors. However they differ from compressors in some important ways. First
since they extract energy from the flow rather than adding it, the pressure drops
throughout a turbine. This leads to different fluid mechanical limitations than for
compressors. Second, because of the thermodynamic requirements, they operate in
gases at high temperatures, so that materials and cooling requirements are of central
importance. We will begin by exploring the fluid-mechanical energy exchange in
the turbine.

Euler Equation:

The exchange of energy is described by the Euler equation, just as for compressors.
We will discuss here only axial flow turbines such as are generally used in aircraft
engines, but it is important to realize that turbines in fact come in a wide variety of
types, from extremely large hydraulic turbines for hydroelectric power generation to
the familiar lawn sprinkler.

.
From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

Axial flow turbines use a stator blade row (usually called a nozzle row) followed by
a rotating blade row that extracts the energy (often called the buckets). The nozzles

1
expand the flow to a high velocity and turn it, imparting an angular momentum to
the flow.
The buckets generally further expand the flow, and turn it back toward the axial
direction. The combination of the two blade rows is called a Turbine Stage.
Here we use the alphabetical notation to avoid confusion with the numerical station
notation for the engine. Station a is equivalent to station 4 in the engine layout, and
for a single stage turbine, station c should be equivalent to station 5 in the engine. A
general stage layout is indicated at the top and two specific blade types at the
bottom.

From the Euler equation, for the general situation,

cp (Ttc - Ttb) = ω(vcrc - vbrb)

Ttc ωrc rb
1 - T = c T ( r vb - vc)
tb p tb c

Let us assume rb/rc = 1, and also that vc = 0 . The latter is desirable if the rotor is the
last stage of a turbine, in order that there not be swirl in the flow entering the jet
nozzle.
Ttc (ωrb)
1 - T = 1 - τt = c T vb
tb p tb

Writing this in terms of Mach numbers and the flow angles indicated in the
diagram, with vb = Vb sin !b = " RTb M b sin !b , and ! rb = " RTb M T ,

(! "1)MT M b Sin# b
1 - τt = (No exit swirl)
! "1
1+ Mb 2
2
Here MT is the tip Mach number of the blades and Mb is the flow Mach number.
From this we can see that we want large MT and large Mb Sinβ for large work per
stage.

Degree of Reaction

As for the compressor there are many possible design choices for the turbine. One
key choice is the relative amounts of pressure drop in the nozzles and buckets. This
is best characterized in terms of the Degree of Reaction, which we will define as the
ratio of static enthalpy change in the rotor to that in the rotor plus stator (stage):
hb ! hc
R=
(hb ! hc ) + (ha ! hb )
We now use energy conservation in the moving frame of each element:

1 2 1 2 1 2 1 2
hb + Vb! = hc + Vc! ; ha + Va = hb + Vb
2 2 2 2
to obtain the alternative interpretation

2
 (Vc')2 - (Vb')2 change in KE in rotor
R= 2 2 2 2 = Total change in KE
(Vc') - (Vb') + Vb - Va

Then, for constant axial velocity.

sec 2 ! c '"sec 2 ! b '

R=
(sec 2 ! c '"sec 2 ! 'b + sec 2 ! b "1)

For zero exit swirl, i.e. vc = 0.

ωr MT
vc = ωr – w tanβc' = 0; tanβc' = w =
Mb cosβb

and with this relation we can eliminate βc' from the expression for R.

Similarly,
ωr MT
w tanβb' = w tanβb - ωr; tanβb' = tan βb - w = tan βb -
Mb cosβb
Using sec2 = 1 + tan2 we can now eliminate βb' and find
Mb Sinβb
R=1-
2ΜΤ
and using this to eliminate Mb Sinβb in the expression for the temperature ratio

(γ-1)MT2
1 - τt = 2(1-R) ( zero exit swirl)
γ-1
1 + 2 Mb 2

From this result we see that for a value of MT limited by stresses and the
temperature and for a given value of Mb, the temperature drop in the turbine, hence
its work per unit of mass flow, is related to R. As R increases the work decreases.

An alternative often used representation is in terms of the so-called flow and power
coefficients, φ and ψ. The definitions are
w $ht v % vc
!= ; #= = b
"r ("r) 2
"r
For zero exit swirl, vc=0, we can also put ! = v b /("r) = M b sin # b / MT , so that
"
R = 1! or " = 2(1! R)
2
which very directly shows how the stage power decreases with the degree of
reaction. In terms of flow coefficient and stator exit angle,
1
R = 1! " tan # b
2

3
These coefficients can all be readily visualized from a scaled velocity triangle in
which the blade speed ωr is taken to be unity:

Vb βb
φ V’b

1
ψ/2 R

Drawing the velocity triangles for R= 1/2 (called 50% reaction) and for R=0 (called
Impulse),

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

In the 50% reaction turbine the pressure drop in the moving blades equals that in
the nozzles, while in the impulse turbine there is no pressure drop in the buckets.
It all takes place in the nozzles.

In designing the turbine we have some latitude in choosing the degree of

reaction R. The tangential velocity of the blades, hence MT, is limited by the
strength of materials at high temperature, so it may seem that we would like to
use small R to maximize the temperature drop per stage. But the efficiency of the
turbine decreases as R is decreased from R=1/2 toward R=0. The reason is that
the boundary layers have more tendency to separate in the moving blades for
R=0 than for R=1/2 because for R=0 there is no pressure drop as the flow is
deflected.

Radial variations:

Just as for the compressor, the requirement for a constant temperature drop from
hub to tip across the turbine flow path, places constraints on the degree of
reaction. Since MT is proportional to radius, it is larger at the tip than at the hub
of the blades, so if Mb is about constant, R must decrease from the tip to the hub.
Thus typically if the degree of reaction is near 1/2 at the tip it may be
considerably smaller at the hub.

4
16.50 Lecture 28

Subjects: Turbine solidity; Mass flow limits; Blade temperature

Solidity and aerodynamic loading:

Just as for the compressor, the blade spacing required to assure that the flow directions at the
exits from the nozzles and the buckets are as intended, is determined by the aerodynamic
loading limits of the blades. Instead of the Diffusion Factor that we used to characterize this
loading in the compressor, it is usual to define a Zweifel Coefficient

β1 v’1=0
V’1 V1

ωr

p s ωr
cx

β’2

V’2 V2 v’2=-ωr

ωr

x2

!Z "
$ x1
( p p # ps )d cxx
pt1% # p2

where x is the axial coordinate, and

pp = pressure on the pressure side
ps = pressure on the suction side
This is a measure of the pressure difference across the blade divided by the outlet dynamic
pressure. In the ensuing argument we will relate it to the quantities that define the velocity
triangle, just as we did for the Diffusion Factor.

From a tangential momentum balance, if s is the blade spacing and c is the axial chord,
!1w1 v1" # v "2 = c x $ ( p p # ps )d cxx
x2

x1
Substituting in the above definition of the Zweifel Coefficient and assuming that p’t2=p’t1,
and M2<<1,

1
 v1#
" 2 w 2 s v 2# 1 $
v 2# w2 v!
!z = = cos " 2! ; 2 = sin " 2!
1 V2! V2!
" 2V2#2c x
2

cx v" v"
!Z = 2sin # 2" cos # 2" 1 $ 1 = sin 2 # 2" 1 $ 1
s v2" v2"
which equals sin 2β’2 when v’1=0.

Zweifel's design rule is that loss is minimized for 0.8 < ψΖ < 1.0.

Mass flow per unit of annulus area and blade stress:

Since the pressure and hence density are higher at the inlet to the turbine than at the
compressor inlet one might think that the mass flow per unit area would not be an
important factor in the turbine. But it is because for a given blade speed the stress at the
root of the blade increases with blade length and hence with the annulus area.

This is readily seen by computing the stress at the blade root. Consider a blade rotating as
in the figure. The stress is given by

rT2 % rH2 #2
! ( rH ) = $r "# 2 rdr = "# 2
rT
=" A
H 2 2 & Flow

This is often quoted as the "AN2 limit" on turbines, N being the rpm. One can see that for
a given blade tip speed the stress increases as the radius of the root decreases, as is
necessary to increase the flow area.

Returning to the question of the mass flow per unit area, we can compare stations a & b
outside the nozzle as shown in the figure. The mass flow density is given by

ρaVa Ha = ρbVb Hb cosβb

2
where H denotes the blade height.
The temperatures and densities at the two stations are related by
γ-1 2 1
Tb 1 + 2 Ma ! b "$ Tb %' ( )1
Ta = , =
γ-1 2 ! # Ta &
1+ 2 Mb a

and substituting these we find

! +1

M a #1 + ! 2"1 M b2 & H cos )b (A* )b H b

2 ( ! "1)

$ ! "1 2 ' = b = = cos )b

M b %1+ 2 M a ( Ha (A* )a H a

Suppose Mb = 1, then we see that as βb increases and therefore cos βb decreases, Ma

decreases, so the mass flow per unit area decreases. Recalling that we wanted large βb in
order to get large work from the turbine we see that there is a tradeoff between high work and
low stresses.

Rotating blade temperature:

Another matter of interest is the temperature of the rotating blades, since the hotter they are
the less stress they can tolerate. As we shall see, the stagnation temperature relative to the
moving blades is not equal to that entering the nozzles.

Let Ttb' be the stagnation temperature relative to the rotating blade:

V! V ! " Vb
2 2 2
Ttb! = Tb + b = Ttb + b
2c p 2c p
from which
Ttb! Vb! " Vb
2 2
= 1+
Ttb # "1 2
2c p Tb (1 + Mb )
2
Now, from velocity triangles,

3
 Vb = w 2 (1+ tan 2 ! b ) ; Vb" = w 2 + (w tan ! b # $r) 2
2 2

so that, after some cancellations,

Vb! " Vb = "#r(#r " 2w tan $ b )

2 2

1 w
We now recall that the degree of reaction can be written as R = 1! tan # b , so we can
2 "r
eliminate the term w tanβb. The result is

Ttr' γ -1 3 ! 4R
= 1 - 2 MT 2
Tta " !1 2
1+ Mb
2
So we find that Ttb' is generally lower than Ttb, reflecting the fact that the blade is “running
away” from the stator flow, but also that it increases as R increases. Therefore for a given
blade speed the impulse turbine will have a lower blade temperature than the 50% reaction
turbine. This just introduces one more factor in the turbine design.

4
16.50 Lecture 29

Subjects: Turbine cooling; General trends and systems; Internal cooling

Turbine cooling trends:

As we have learned from our performance analyses for turbojets and turbofans, the thrust
per unit of airflow of the aircraft engines increases monotonically with turbine inlet
temperature ratio, which we have designated as ! t . Though it is less obvious, the thermal
efficiency and specific impulse also increase due to the increase in ! t . Thus, there is a
powerful incentive to increase the turbine inlet temperature.

The trend of temperature increase with time is shown in the figure:

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

In the early period, the increase was limited by metallurgical progress in developing
oxidation-resistant materials, primarily Nickel and Cobalt based alloys, that could operate
for long times at high temperatures. Beginning in the mid-60’s, the technology of air
cooling was introduced, and most advancement since then has been due to more refined
cooling techniques, although the introduction of directionally-solidified and single-crystal
blade materials has allowed some increase in metal temperature.

Cooling Systems: The general scheme for air cooling of the “hot section” of an engine is
shown in the figure below:

1
 From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

In most all cases the cooling air is drawn from the compressor discharge because it must
be at a higher pressure than that of the part of the flow path to be cooled. This leads to the
important condition that the cooling air temperature is the compressor discharge
temperature. As shown in the diagram, some of the air is used to cool the turbine nozzles,
by a combination of internal convection cooling and film cooling, processes that will be
discussed in detail below. Another portion of the cooling air is transferred onto the rotor
and used to cool the rotating blades and their supporting disc. Here again a combination
of internal convection and film cooling is used.

Internal Convection Cooling:

In this form of cooling, the blade temperature is maintained below that of the hot gas
passing through the turbine by a flow of compressor-discharge air through passages
internal to the blade, as suggested in the diagram. To model this process we must
therefore model first the heat transfer to the outside of the blade from the hot gas, then the
heat transfer on the inside to the coolant. The objective in cooling design usually is to
maintain the blade temperature nearly constant in the presence of a variation of the heat
flux to its surface from the hot gas.

External Heat Transfer:

Let us look first at the heat transfer to the outside of blade. We can write it as
qw = St [ !uc p (Tr " Tw )]
where ρu is the mass flow density in the hot gas, Tr is the “recovery temperature” and Tw
is the wall temperature. The definition of Tr is the temperature the wall would reach if
adiabatic, and it is close to the external stagnation temperature. St is defined by this
relation, but what is its physical significance? We have

2
 cqw
= #St = fraction of heat flowing thru the blade row which enters the blade.
!uc p (Tr " Tw ) s
Here as before c is the blade chord and s is the blade spacing.

Fortunately, St is a small number because the boundary layers on the blades are very
thick compared to the blade chord. In fact St depends on the details of the boundary layer
behavior, and of course, is not a constant, but for our purposes here we can think of St as
being an average value for the blade surface. To estimate its magnitude we can model the
blades as flat plates, in which case the local St is given approximately by:
0.66
2 St = C f = (laminar)
Rex

.0592
2 St = C f = (turbulent)
( Rex )
.2

where Rex is the Reynolds number based on distance from the leading edge of the plate.
One important difficulty for turbine cooling arises because transition from laminar to
turbulent flow occurs at some Reynolds number in the range of 3 !10 5 to 10 6 . Looking at
the expressions for St we see that for example if the transition occurs at Rex=106, then

St = 0.66 ×10-3 (laminar)

St = 0.037 (turbulent)

So the heat transfer jumps suddenly by a large factor at the point of transition.

To get an idea of the magnitude of Rex in an engine let us look at some typical numbers.

For pt4 = 20 atm, θt = 6, Tt = 6(298) = 1788 K, Rex = 3 × 107/m at M = 1, so for a blade

that has a chord of 4 cm, Rex based on chord is 1.2 × 106 and we would expect transition
on the blade surface. Actually external flow turbulence tends to promote the transition so
that it will probably occur at a lower value than this. But the exact location of the
transition is difficult to predict.

This gives rise to a major problem in scheduling the internal cooling system so that
internal cooling matches the external heat transfer.

Internal cooling rate:

For the internal cooling flow we model the flow as that through a long thin passage, for
which
1
2St int. = Cfint = 0.023
( Re D ) .2
One important question to be addressed is how much cooling air is required. Forming a
heat balance such that the heat flux into the blade equals that absorbed by the cooling
flow,

3
 2cqw = 2!St e [ "uc p (Tr # Tw ) s] $ ṁc c p (Tw # Tc )

ṁc T $T
" 2#St e r w
!us Tw $ Te

For Tr – Tw = 200*K, Tw – Tc = 400*K , and σ=1,

m! c $ 1'
" 2# (.005 ) & ) * .005
!us % 2(

Actually a ratio more nearly .01 is required, because the cooling air is not heated all the
way to the blade temperature. This is per blade row, not counting casing cooling. For a
two-stage turbine, including these effects we may need 2-4% of the air flow.

4
16.50 Lecture 30

Topics: Film cooling; Thermal stresses; Impingement

Cooling; How to do cooling design

Film cooling:

With internal cooling the heat flux through the blades’ surface increases as the external
gas temperature increases. That is large Tr – Tw implies large qw. This large heat flux
must be conducted through the blade material according to
!T
qw = k w
t
where t is the thickness of blade skin.

So for a given back-side T, a small t is needed if the surface facing the hot gas is to be
maintained at an acceptable level. As noted in the last lecture, it is also necessary to
match the internal cooling precisely to external heat load to prevent a large Tw variation.
All of these difficulties can be mitigated somewhat by the technique known as film
cooling, which is illustrated in the figure:

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

Essentially the idea is to introduce cooling air into the boundary layer over the surface of
the blade, reducing the effective temperature of the gas heating the surface. This lowers
the heat flux, making less cooling air necessary. In addition the cooling air cools the wall
in passing through the small holes.

The performance of film cooling is related to the “mass velocity ratio”

!fuf
m=
! sus
where f refers to the film coolant and s to the free stream flow:

1
 From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

As indicated in the figure, which shows heat transfer rate versus the (dimensionless)
magnitude of the cooling air flow, the cooling air actually has two effects on the heat
transfer to the wall, which can be separated by varying the cooling air temperature
relative to the wall temperature. If the cooling air is at the wall temperature its effect is to
increase the heat transfer, by stirring the boundary layer. But if it is cooler than the wall,
this effect is partially offset by the reduction of the effective hot gas temperature, so that
there is a “best” cooling air flow rate.

To make these trends quantitative for design purposes, we define a Cooling Effectiveness
by the relation
T "T
!ad = r rf
Tr " Tc

where Tr is the “adiabatic recovery temperature”, i.e. the temperature the wall would
reach if adiabatic in the absence of film cooling, Trf is the recovery temperature in the
presence of film cooling, and Tc is the coolant temperature, which is the lowest Trf. The
cooling effectiveness so defined applies to a specific geometrical arrangement of cooling
holes, and also on the cooling mass flow rate. A set of data for a single row of holes is
shown in the figure below. The effectiveness decreases downstream of the row of holes,
and in general increases with cooling flow, at least for the range shown.

2
 From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

Now with this cooling effectiveness, how do we estimate the heat transfer rate to the
wall? The usual approach is to assume that the conventional film coefficient h, or the
equivalent Stanton number St, can be applied, with an effective gas temperature given by
the cooling effectiveness, Thus,
qw
h=
Tr f ! Tw
where h is that for a normal boundary layer, and Trf ! Tw = Tr ! Tw ! "ad (Tr ! Tc ) . We then
calculate
qw = !uc p (Trf " Tw )St
where h or St are correlated in the usual way to the Reynold’s number and Prandtl
number.
Impingement cooling

Because the heat transfer rate is largest at the leading edge of the blade where the
boundary layer thickness is least, special measures are sometimes called for to cool this
region. One is Impingement Cooling, illustrated in the figure:

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

3
For the scheme shown the cooling rate can be correlated as

! d $ '0.85( dl )( dp )( dL )0.4
.8

Nustag = 0.44 Re # & e

.7
d
" p%

where the Reynolds’s number is that for the impinging jet

( !u) Jet d
Re d =
µ
and the Nusselt Number similarly is based on the jet diameter d
hd
Nu =
k fluid
Thermal stresses

For a thin skin supporting a temperature difference due to a heat flux,

a thermal stress is generated by the tendency of the hot side of the wall to expand relative
to the colder side. To see this we can note that in general the strain in the material due to
a combination of a normal stress σc and a temperature difference from some mean
temperature is
$
! = (1" # ) c + % (T " T )
E
If the plate extends to infinity in both directions, then the strains on both the hot and cold
sides will be the same (note however that this assumes the plate does not curve or bulge).
In this approximation the stress, called the Thermal Stress, is found by equating the two
strains:
# ! #C
(1! " ) H = !$ (TH ! TC )
E
If the plate is unloaded, force balance imposes ! H = "! C , and the temperature difference
is related to the heat flux through the plate by TH ! TC = t qw /k . Combining and solving
for the cold-side stress (a positive tension), we obtain

"E
!C =
2k(1# $ )
( qwt )

The quantity !E /k is termed the “thermal stress susceptibility”. Clearly we would like it
to be small. It is proportional to the thermal expansion coefficient, and to the elastic

4
modulus and inversely to the thermal conductivity. Unfortunately the oxidation-resistant
Nickel and Cobalt alloys favored for turbine blades have relatively low thermal
conductivity and high modulus, so thermal stress is a controlling problem for turbine
blades. It results in cracking an ultimately limits the blade life.

(1! " )#Ult.

A related material quantity is the “Figure of Merit”, Z = (in K) , which is a
E$
measure of the allowable temperature difference across the plate. A few values are:

Material Cu SS 302 Ti Alloy Steel

Z (K) 35 111 136 234

The Z value is low for copper, but, of course, even a small temperature difference can
drive a large heat flux through copper.

How to design cooled blades

It is presumptuous to engage this subject at the tail end of a one-hour lecture, because this
is one of the most difficult areas of gas turbine design. But the intent here is simply to
outline the way one might proceed, as follows:

a) Estimate the qw over the blade surface for specified Tw.

b) Find the thermal stresses in the skin.
c) Find the reduction in qw required to limit the stresses to acceptable values,
hence the !ad required of the film cooling.
d) Find the arrangement of cooling holes and cooling air flow for film cooling, to
give the required effectiveness.
e) Find the internal cooling airflow to absorb the residual qw.

5
16.50 Lecture 31

Subjects: Compressor -Turbine matching; Gas generators

Compressor-turbine matching:

We have described the characteristics of both compressors and turbines.

Assembling the two with a combustor between gives us a gas generator, which is
the heart of any gas turbine engine. Once we understand its behavior we can
appreciate most of the real characteristics of aircraft engines, and graduate from
thinking of them as a lot of abstract equations.

Actually, we have already done in Lecture 20 most of the work needed to

understand the performance of an ideal Gas Generator, including the important
concept of the “Compressor operating line”, which is set by the flow passing
characteristics of the rest of the engine. We summarize here the main findings of that
work:

(a) If the nozzle and the turbine stators are both choked, the turbine temperature
ratio is fixed once the flow area ratio A4/A2 is set.

(b) One additional single parameter is sufficient to specify all the other gas generator
parameters. This can be the temperature ratio θ=Tt4/Tt2, or the compressor
temperature ratio τc, or the engine-face Mach number M2, or the normalized air mass
fh
flow m2 , or the fuel flow ratio .
c p Tto
(c) A “Compressor Operating Line” in the plane of compressor pressure ratio vs.
normalized flow can be calculated once the ratios of all the flow areas are known.
For a given choice of one of the parameters listed above, a point is selected along this
operating line.

(d) All the above is independent of the compressor specifics. After the compressor
has been selected, its performance map (pressure ratio vs. normalized flow) contains
normalized rotational speed lines as additional information. This parameter is
therefore to be added to our list of possible parameters (as is done in the figure at
the end of this lecture), each of which uniquely specifies the state of the gas
generator.

It follows from the above that a single degree of freedom is left to the pilot (or to the
engine controller), unless geometry can be varied. It is probably most intuitive to
think of this unique freedom as the normalized fuel factor, or the peak temperature
ratio Tt4/Tt2, since these closely relate to the engine throttle control.

It is to be noted, however, that many idealizations have been made to obtain these
simple results. If the turbine or the nozzle un-choke, or if the engine inefficiencies
are rigorously accounted for, the overall detailed behavior is more complex, but its
main qualitative features are not too different.

1
Using these ideas, one can generate and plot a set of Gas Generator Characteristics,
such as those below. With this Gas Generator, we can do lots of things

a) Turbojet
b) Turbofan
c) Turboprop
d) Unducted fan
e) Helicopter

Notice that the single free variable chosen for this plot is the normalized rotational
speed (as a fraction of its design value). The quantity in the denominator is the non-
dimensional compressor-face temperature θ2= T2/T0, because the blade speed ωr is
made non-dimensional with the speed of sound at station 2.

2
From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

3
16.50 Lecture 32

Subjects discussed: Engine structures; Centrifugal stresses; Engine arrangements

Centrifugal stresses and design of discs

By disc we refer to the rotating structural members that carry the rotating blades in
the turbomachine. They are unusual as structural members in that they must
withstand very large tensile stresses generated by centrifugal forces. Their mass is a
large part of the total mass of an engine, and they set the limit on blade speed, so an
understanding of their characteristics is essential to appreciating the performance
limits of modern aircraft engines.

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

We begin by describing the state of stress of a small volume element of a disc. The
net outwards elastic force on the element is
d dz
(! zrd" )dr # ! z dr d" = ! r dr d"
dr dr
and, by design, in order to utilize the material most efficiently, we wish to keep the
stress level σ constant throughout the disc. The force balance is then
dz
! z" 2 r 2 d# dr = $% r( )drd#
dr
1 dz "(# 2 r)
=!
z dr $
"# 2 r 2
z = ct. x exp(! )
2$
Add a rim and consider the radial force balance on a slice of this rim, as in the
bottom left sketch. There are B blades, not necessarily at the same stress as the disc,
and their centrifugal pull is smoothly distributed around the periphery:

1
 $ bH Ab B
!oWoTo (rH d" )# 2 rH + d" = $ zH rH d" + $ WoTo d"
2%
Rim centrifugal Centrifugal Support from Radial force
force pull from blades rim stress from hoop stress

"0# 2 rH2 + ( 2B$ )! bH (Ab / WoTo )

!= (2)
z r
1+ H H
WoTo
So the disc supports the rim, and reduces its stress. Now we add an inner rim; it
supports the disc and there are no blade loads, so in this case

! IWI TI (rI d" )# 2 rI = $% zI rI d" + % WI TI d"

"# 2 rI2
!= (3)
zr
1$ I I
W I TI
For the blades themselves, the full blade mass is supported by the root stress:
" # 2 rT
! bH = b $ Ab (r)dr (4)
AbH rH
So how do we design a disc

a) From aerodynamics, choose ωrT, rH/rT, B, Ab(r) Wo

b) Set permissible stresses σ and σbH (ωrT may then be limited by (4))

c) Choose a To and get ZH from (2)

d) Choose an rI and get ZI from (1)

e) Get WITI from (3)

Engine arrangements

As noted, engines are unusual amongst engineering structures in that such a

large fraction of their total mass is rotating at high speeds. This large rotating
mass must be supported on bearings so as to maintain quite close clearances
between the blade tips and the stationary casings, on the order of 1 mm on a
rotor of 1 m diameter, or one part in 103. At the same time of course the
stationary structure must be as light as possible.

All existing engines use ball and roller bearings to support the rotating
assemblies, called "spools". These are shown schematically in the figure:

2
 From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

The function of the squeeze films will be discussed in the next lecture.

Ordinarily each rotating spool is supported by one ball bearing that positions it
axially and also absorbs radial loads, and one or more roller bearings that accept
radial loads but allow axial movement to accommodate thermal expansion and
structural deformations.

Two-spool support arrangements are shown schematically in the figure:

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

The top arrangement uses just two bearings, one ball and the other roller, to
support each of the two spools. It is compact and at least in principle relatively
light. The lower arrangement uses three bearings on the inner spool, one ball
and two roller, and four bearings on the outer (low speed) spool, one ball and
three roller. Both of these arrangements have advantages and both have been
used successfully in high-bypass commercial transport engines. This is just one
more illustration of the fact that design solutions are not unique.
3
16.50 Lecture 33

Subjects: Critical speeds and vibration

Critical Speeds

As noted in the last lecture, gas turbine engines are unusual structures in that a large
fraction of their mass is rotating at high speed. Thus, vibration and dynamics have
always been a serious issue with them. What is intended here is that the physical
phenomena of principal importance be conveyed by study of a set of simplified models.

Schematically, we can think of the rotating system as a mass on a flexible shaft,

mounted on flexible supports. If most of the mass resides in the rotor and the part of the
shat adjacent to it, we can combine the two flexibilities as two springs in series with a
combined stiffness which tends to restore the rotor’s center to the axis from whatever
direction it is deflected into. More complex models, with additional degrees of freedom,
would be needed if the bearings and the parts of the shaft near them had a significant
mass. We still retain the simpler formulation, but will allow the rotating mass to be
mounted eccentrically, to study the loads this can generate.

The schematic below shows the situation. The center of mass C has coordinates (x,y),
and the geometrical shaft center, also assumed to be its elastic center, is at S. The center
of mass is offset by a distance e from S, and rotates with the shaft at the angular rate ω.
There is an elastic restoring force –k e directed from S to O, as well as a viscous
damping force, normally provided by an oil squeeze film damper between the shaft
and the bearing. This damping force is proportional to the rate of change of that part of
OS that comes from the bearing displacement, but since we ignore the mass of the
bearings, that distance is a fixed fraction of OS itself (depending on the ratio of shaft to
bearing stiffness), and we model it as –b (de/dt).

x C
e ωt θ

S
y

O
X

We can now write a pair of equations of motion, one for x, the other for y:

1
 M˙x˙ = !kx S ! bx˙ S
My˙˙ = !ky S ! by˙ S
where
xS = x ! ecos " t; yS = y ! esin " t

It is actually more convenient to work with a complex displacement z=x+iy, so that a

single complex equation is needed:

M!z! = !kzS ! bz!S ; zS = z ! e exp(i" t)

Introducing the “natural frequency” ! n = k / M and the friction parameter n = b / M ,

and eliminating zS, we obtain

˙z˙ + nz˙ + ! n 2 z = e(! n 2 + in! )e i!t

The general solution to this equation consists of the homogeneous part alone, with two
arbitrary constants determined by initial conditions, plus the forced, or “particular”
solution, determined by the right hand side. The homogeneous solution is a
superposition of damped sines and cosines at the natural frequency ωn. Because of the
damping, it will disappear some time after a transient event, leaving only the forced
solution, which is proportional to exp(iωt), like the right hand side of the equation.
We concentrate here on the forced solution only (the steady state). Putting z=B exp(iωt)
and substituting into the governing equation, we can solve for B, and find

! n 2 + n!i
z=e 2 e i!t
! n " ! + n!i
2

"2
zS = z ! e exp(i"t) = !e e i"t
" n ! " + n"i
2 2

The force on the supports is, in complex form,

F = Fx + iFy = !kzS ! bz!S = M!z! = !M " 2 z

where the last form can be directly rationalized as the result of “centrifugating” the
center of mass. It is, of course, a rotating force (around O), proportional to exp(iωt). The
amplitude of this force is obtained from the expression for z:

! n 4 + (n! n! ) 2
F = M! e 2

(! n " ! 2 ) 2 + (n! n! ) 2
2

2
and the phase can also be written down, although we omit it here.

The solutions for the amplitude of z (normalized by the eccentricity e) or for F

(normalized by Mω2e) are shown below, together with the phase information:

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

Some important points are:

1) With no damping, z/e → ∞ and F/e → ∞ for ω → ωn (resonance
condition).
2) Damping limits the resonant amplitudes to finite values.
3) The system can be run above ωn with a flexible shaft or support, in
which case the mass rotates about C for very high speed (x and y tend to zero).

Now let us ask how the bearing stiffness and damping influence the forces on the
structure.

a) At “subcritical” speeds, namely, ω2<<ωn2, F ! "M# 2e exp(i#t) , and the forces die
out as the square of the speed, as could be expected. Notice this condition might not
mean a very low rotational speed, if the shaft and the bearings are stiff enough, or
the rotor is light enough, to produce a very high natural frequency (this is unusual
in jet engines, though).

3
(b) At supercritical speeds, ω2>>ωn2, and with relatively weak damping,
# 2
z ! "e n2 ! 0 , i.e., the center of mass tends to remain stationary, with the shaft
#
rotating about it. The force then tends to F ! M" n e = ke, a constant elastic force
2

towards the center of mass (but a rotating force, that transmits to the supports and
causes vibrations). This is the normal condition for most engines at design speed.

Elaborate procedures are followed to reduce eccentricities to a minimum, but with

engine wear, they are difficult to maintain near zero. Their effects can be mitigated
by (a) Staying away from operation at a speed equal to the natural (“critical”)
frequency. (b) If normal operation is at supercritical speed, crossing the critical range
as quickly as possible, and (c) Providing as much damping as practical with
squeeze-film or other dampers.

4
16.50 Lecture 34

Subjects: Combustors; Afterburners

Combustors:

The combustor of an aircraft engine releases an enormous amount of heat per unit
volume, second only to high-pressure rocket engines. In designing the combustor
we must implement solutions to the following difficulties:

a) The gas temperature at the outlet is well above the limits of the materials from
which the combustor is built.

b) Kerosene-air mixtures will only burn in a narrow range of mixtures near

stoichiometric.

c) The overall mixture is very lean, and varies over a wide range with changes in the
power setting of the engine.

These requirements are met by introducing the fuel as a spray into a primary zone
where the combustion is near stoichiometric and is stabilized by a recirculating flow
generated by a set of jets from the periphery. The gas leaving the primary zone is
then cooled by mixing with additional air before it exits to the turbine nozzles.

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

There are also a number of “secondary” requirements to be met, some of which are
quite challenging too:

-The burner must be dimensioned to assure re-starting at high altitude, when

pressure, and hence chemical rates, are low. Designs that are optimized for cruise
may not meet this requirement.
-The combustor must have an operating life of over 20,000 h., or some 5,000 cycles.
-Pollutant formation must be below legally allowed limits (see next lecture).
-The total pressure drop must be small, despite the strong flow turbulence that is
purposely introduced for enhanced mixing.
-The temperature profile at combustor exit must be uniform (or sometimes must
meet some prescribed radial profile).

1
Now let us look at some of the fundamental phenomena involved in the combustion
process. The fuels used commonly in aircraft engines (kerosene, JP-4), react with air
only in gaseous form, so they must be atomized, vaporized, and mixed with the air,
starting from liquid form. Even then, the fuels burn only in nearly stoichiometric
proportions, so we have to first burn with such proportions, then mix with air to get
the desired temperature.

As an example, take a stoichiometric mixture of octane (C3H8) and oxygen:

C3H8 + 5 02 → 3CO2 + 4H20

! fuel $ 3(12) + 8
#" &% = (0.23) = 0.0633
air stoich 5(32)
whereas normal fuel/air ratios in jet engines are around 0.03. The ratio of fuel to
stoichiometric fuel is called the “fuel equivalence ratio”, ϕ, and is here seen to be as
as low as 0.3/0.66=0.45. Normal hydrocarbon fuels do not ignite this far from
stoichiometry, even at the relatively high compressor discharge temperture.

In the molecular mixture, the reaction proceeds by molecular collisions. There is an

activation energy, A, and the kinetic energy of impact between the reacting
1
molecules 2 mv2 must exceed A in order that a reaction to occur. Since the
molecules have a kinetic energy distribution according to Boltzmann
2

( )!e
" mv
v2 2 RgT
f 2

2 " A
the fraction of molecules with m v > A is ! e RT . The number of such collisions
2
per unit volume and time, called the reaction rate, is
reaction rate ∝ n02 nfuel n ... e-A/RT
so for a given fuel-air mixture we find

reaction rate ∝ pne-A/RT

where n=2 if 2 molecules are involved, 3 if 3 molecules are involved, etc.

Combustion in premixed gases is the simplest to visualize and in a sense the easiest
to experiment on, although is not often the actual mechanism in practical
combustion systems. We imagine a flame front propagating into the premixed fuel
and air

fuel & air

Reaction mixture
Products
vflame

2
The rates of propagation of such flames are determined by a balance between
reaction time and products diffusion time, and are roughly as shown in the figure:

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

This however is the laminar rate. Turbulence increases the rate by at least two
mechanisms.
a) stretching of the flame front
b) wrinkling of the flame front
Both of these mechanisms are promoted by a high level of fluid turbulence that is
generated by the jets of air entering the primary zone and the mixing zone.

Combustor sizing:

The chemical energy density in the burner is the mass of fuel per unit volume, times
the fuel heat value, E /V = ! a fh . If the time to full combustion, tburn is known, we can
then calculate the power density as
! fh
P /V = a
t burn
The burning time must be estimated differently for design (or cruise) conditions and
for low-pressure conditions.

(a) For design conditions the pressure and temperature are fast enough that the
rate-limiting step is not the reaction itself, but the turbulent mixing of the fuel
and air, or of the rich primary burnt gas with the dilution air. Let lt be a
caracteristic turbulent eddy size (of the order of the combustor diameter or
width), and ut some characteristic eddy velocity (some fraction of the mean
flow velocity in the burner). Then we estimate t burn ! lt /ut . For a numerical
estimate, take ρa=3 kg/m3, f=0.027, h=43MJ/kg, ut=ux=50 m/s and lt=0.1 m.
We then calculate a power density
P/V=1.7x109 W/m3
which is of the right order, and obviously very high.

3
 We can now calculate the combustor volume Vc=Pth/(P/V), where the total
thermal power is Pth = m˙ a fh , and the airflow can be expressed as m˙ a = ! a ux Ac .
Substituting and simplifying, the combustor length, Lc=Vc/Ac is found to be
u
Lc ! x lt
ut
which simply says that the combustor length scales with its width, or with
whatever determines the eddy size. We could call this “photographic
scaling”.

(b) At low pressure, chemical reactions limit the combustion time, and assuming
a bi-molecular reaction mechanism,
n n C(T)
t burn = products = products !
n˙ producs R(T)n1n 2 "a
The power per unit volume is now proportional to the air density, and
repeating the steps of case (a), we arrive at a combustor length estimate
u
Lc ! C(T) x
"a
which is independent of width or any other size parameter. In other words,
the chemically-limited combustor length is independent of engine size.

Roughly speaking, the larger of the two estimates must be chosen in order to satisfy
conditions both at design and at low pressure. For a given engine family, the smaller
engines will be chemically limited and will all tend to have equal length combustors
(so the combustor is most prominent in the smallest of them), but at some larger
size, length will need to be increased in proportion to width in order to obtain
sufficient mixing.

Afterburners:

The afterburner faces a different set of requirements

a) High outlet temperature

b) High flow Mach number
c) Low pressure drop required when non-afterburning
d) Accepts flow from both core and fan

In afterburners the flame needs to be stabilized by bluff bodies, usually triangular in

shape as indicated in the diagram. This is because the flame speed, even accounting
for turbulent effects, is not as large as the mean flow speed, and the flame would be
carried away without some recirculatory zone where combustion has time to occur.

4
 From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.
The requirement for stability is that the fuel-air mixture spend a certain time, τc, that
is required for the chemical reactions to initiate, in the mixing layer adjacent the
recirculating wake. So the flow time τf must be greater than this "chemical time" τc.
From the flow geometry,
L
τf = V
So we require
L
V > τc
The length/breadth of the wake region is nearly independent of the absolute
L
breadth, so we can say that L = (D )D, and we see that a certain size of flame holder
is required such that

D L
V (D) > τc
Since V and τc are independent of engine size, we find that D is as well. This is an
example of lack of geometric scaling, analogous to the low-pressure scaling for the
main combustor.

5
16.50 Lecture 35
Subjects discussed: Pollutant. Motivations for control. Formation. Strategies for
reduction

Motivations for control

The motivations for control of emissions from aircraft, other than purely esthetic or ethical,
stem from legislation regulating the amounts of emission in the neighborhood of airports, and
the possibility of regulation of emissions into the upper atmosphere. We will discuss each of
these in turn.

Emissions in airport neighborhoods;

In 1973 the EPA published standards, to go into effect in 1979. They are compared in the
table to the actual emissions from three engines in wide use at the time the standards were
issued.

1979 EPA Proposed Standards

1979 Actual Emissions

Pollutant Std JT8-D JT9D-7 CF6-50

CO 4.3 1.9 10.4 10.8

UHC 0.8 2.7 4.8 4.3
NOX 3 8.0 6.5 7.7
smoke* 19-20 visible 4 1.3

All values are grams per kg thrust, per hr. in standard approach-taxi-takeoff cycle.

Rs
* Smoke, SN = 100(1 - R ), the reduction in optical transmission through a glass witness plate exposed for a
w
prescribed time.

Since the FAA has regulatory jurisdiction over civil aviation, the implementation of these
standards fell to it. In July 1973 FAA issued SFAR27 to enforce only smoke regulations but
1979 regulations were never incorporated into regulating process.

Note: Manufacturers lobbied against them on the basis that the increasing pressure ratios
required for better fuel consumption would make their realization difficult.

In 1981 ICAO issued international emissions standards in terms of g/kg fuel:

Pollutant Standard
CO 118
UHC 19.6
NOx 40 + 2(π00)
Smoke 83.6 (F00)-.274

1
 π00 = rated pressure ratio F00 = rated thrust in Newtons

In 1990 FAR 34 was issued by the FAA, regulating only smoke and unburned hydrocarbons,
and setting the limits for these essentially at the values attained by the oldest engines in
operation.

The Committee for Aviation Environmental Protection (CAEP), an international organization

under ICAO, has periodically updated the emission limits on NOX. The various CAEP
regulations, including CAEP-6 , in effect as of 2008, are displayed in the figure below. The
units on the ordinate are in grams of pollutant per kN of thrust, averaged over a standard
airport turnaround cycle. The abscissa is the overall pressure ratio of the engine. For
reference, the CAEP-1 regulation is from 1986, and the CAEP-4 is from 1998. It is clear that
the trend is to tighten the emission rules over time, and also that commercial engines do
comply, as they must, with the rules in effect at the time they enter service (and usually
ahead of time, in anticipation of further strengthening of the rules).

CAEP1
90 CAEP2
1
EP
CA CAEP4

80 -12% CAEP6
CAEP6-10%
CAEP6-20%
E P2
CA CAEP6-30%
70
DP(NOx)/F00(g/kN)

CFM56-5A
CFM56-5B/P
EP4 EP6
CA
CFM56-5B/2P
CA
60 CFM56-5C

CFM CFM56-5C/P
CFM56-7B

50
DAC CFM56-7B/2
CF34

CF34(LEC)

40 Dem21

SM146

CFM56-TAPS

30 FUTURE ANTLE

PROJECTS CLEAN
RRD LNlll

20 AST(USA)

15 20 25 30 35 40 UEET(USA)

Overall pressure ratio

Image by MIT OpenCourseWare.

Mechanisms for formation of pollutants;

The principal pollutants are Unburned Hydrocarbons, Carbon Monoxide, and Nitrogen
Oxides. They are formed as follows.

UHC - mostly at idle and low power, due to poor atomization etc. Also from
venting of fuel systems.

2
 CO - Due to fuel-rich combustion in primary zone.
Smoke - Due to fuel-rich droplet burning by diffusion flames
NOx - Following an element of fuel and air through the combustor we would see the
following history of temperature and concentrations.

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

During this transport, NOx is formed by the following sequence of reactions.

N2 + O ! NO+ N; where the O is from (near-equilibrium) dissociation of

H20, 02 :
p0 2
K p (T ) =
p0 2

N + O2 ! NO+ O

N + OH ! NO + H where the H, OH are also from dissociation

69,460
d[NO] #
= 2k[N 2 ][0] ! [N 2 ][0 2 ]1/2 " 1.45 " 1017 T #1/2 e T
dt
(units are mol, cc, K)

The result of this set of reactions can be summarized as in the figure (for 1 atm. pressure,
starting at 700K).

3
 From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

Strategies for reducing N0x

a) Premix and burn lean

b) Burn rich and then dilute and cool quickly

Upper-Atmospheric Emissions:

Here the threat is to the 03 in the upper atmosphere. No regulations are in place as yet for
aircraft, although there is an international agreement limiting the manufacture and sale of
fluorocarbons.

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

4
 a) Formation of 03 (the Chapman cycle):

Molecular oxygen provides a reservoir of O atoms through photodissociation by UV

solar radiation:
O2 + h! " O + O
The atomic oxygen reacts rapidly in a three-body reactionwith molecular oxygen, to
form ozone:
O + O2 + M ! O3 + M
The population of ozone is mainly controlled by its destruction though
photodissociaation:
O3 + h! " O + O2

b) Effect of Nitrogen oxides:

The natural effect is the continuous upwards transport of NO2 from the troposphere and its
photodissociation to NO and O. This is followed by ozone destruction by

NO + O3 ! NO2 + O2

The natural transit time of NO2 from the troposphere is as long as 100 years, so even a small
amount of NO andNO2 released in-situ by high-flying aircraft could potentially lead to a
strong acceleration of the ozone destruction.

c) Other effects:

In addition to these NOx concerns, recent information suggests that the particulate content of
the exhaust, largely sulfuric acid drops formed from the sulfur in the fuel, is a major factor in
the overall effect on the atmosphere.

5
16.50 Lecture 36

Subjects discussed: Aircraft Engine Noise : Principles; Regulations

Noise generation in the neighborhoods of busy airports has been a serious problem
since the advent of the jet-powered transport, in the late 1950's. Although piston-
engined aircraft had caused some concerns prior to that, it was the turbojet-powered
first generation jets (707, Comet, DC-8) with their jet noise that led to wide public
concern, and to regulations by some airport authorities. With the continuing growth
of airline travel, the problem has continued to expand, leading to rules promulgated
by the FAA that limit the noise that any individual aircraft can make at each of 3
measuring stations.

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

Noise is the human ear's response to pressure fluctuations in time, at the ear
of the observer. As such it has both physical and psychological aspects. Thus, one
person's music can be another person's noise. There is general agreement for
example that the exhaust noise of motorcycles is annoying, but the biker will not
agree. We will from this point on avoid the psychological and concentrate on the
physical aspects of aircraft noise, without meaning to imply that the latter is more
important.

Noise is transmitted from the source (aircraft) to the observer by sound waves. So
let us begin with a brief review of sound propagation.

Conservation of mass

D!
+ !" • u = 0 (1)
Dt

Conservation of momentum

Du
! = "#p (2)
Dt

Suppose the velocity and pressure to be the sum of a large steady component and a
small time-varying one

1
 ! ! !
u = u0 + u' (t) p=po + p'(t) ρ=ρo + ρ'(t)
!
and further suppose that u0 = 0 . Then to first order in small quantities (1) and (2)
become
!"' !
+ " 0 #. u' = 0 (3)
!t
!
"u'
!0 = #$p' (4)
"t

In the absence of heat conduction and viscosity, ρ' and p' are related isentropically,
i.e.

p2 ! " dp dp p' "'

=( 2) # =" or =!
p1 !1 p0 !0 p0 p0
It follows that
p0
p' = γ !' = "RT0 !' # ao2 !' (5)
!0
where a0 is the speed of sound.

So in 1a
!"' 1 !p'
= 2
! t a0 ! t

!
Now taking (1a) - !.(2a) we have
!t

1 ! 2 p'
2 " # p' = 0
2
(6)
a0 !t
2

In one dimension (for a plane wave)

1 ! 2 p' ! 2 p
2 " =0
a0 !t !x
2

This is a Wave Equation, satisfied in general by solutions of the form p' = p'(x ± aot),
so that p' is constant for x = ± aot. Thus the solution for a plane wave would be of the
form
p ! = P cosk ( x " a0 t )
Excitation

We can now ask how such waves are generated in an engine. Some of the main
sound sources are schematized below:

2
 From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

Elementary sources. It is useful to examine the simplest acoustic sources, which are
all configurations with an imposed pressure fluctuation or an imposed wall
vibration. The first in a systematic series of such sources are shown below:

3
 From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

(a) Monopoles

Suppose now the perturbation is due to a pulsating sphere, whose radius

oscillates by some small !r0 at a frequency ω. This sets up spherically symmetric
pressure field that oscillates at ω as well, and propagates as a sound wave train.
In spherical coordinates, the wave equation is

1 ! 2 p" 1 ! 2 !p"
= (r ) (7)
a0 !t 2 r 2 !r !r
2

We expect acoustic energy to be conserved, and since at least the compression

part of this energy varies as p'2 , let us try a solution of the Monopole form:
r
p'= P( 0 )cos k(r ! ao t) (8)
r
It can be seen by direct substitution that this does satisfy the wave equation. The
2" a
quantity k=ω/a0 is called the wave number, and the wavelength is ! = = 2" 0 .
k #

The velocity field (purely radial) can be calculated from (4)

!u'r 1 !p'
=" (4a)
!t # o !r

and from (8), after integrating in time,

4
 P ro # sin( r " ao t ) &
u'r = %cos( r " ao t ) " ( (9)
! o ao r $ kr '

The acoustic power flux (energy crossing unit area per unit time, averaged over
one cycle) is the average of the work done by p!:

P 2 % ro (
2
1
# o p' u'r dt =
"
!= ' * (10)
" 2 $ o ao & r )

2" 2"
where ! = = .
# kao
P 2 ro2
The net acoustic radiated power is Pm = 4"r # = 2" 2
, which is seen to be
! o ao
independent of r, as it should. The monopole power radiated is independent of
wave number k = 2! / " , and only dependent on pressure amplitude. A possible
physical implementation of a monopole source is a pulsating jet, such as
produced by a pulse-jet (like that in the V-1 missile), or by the oscillations during
an engine surge.

(b) Dipoles

Consider next two monopoles of equal strength P operating in counter-phase to

each other, and spaced a small distance d along the x-direction. If an observer is
located at a distance r from one of them, and at an angle ϑ from the x direction,
its distance to the other will be (approximately r + d cos! . Then the pressure p'
at the observer’s location will vary as

Pro Pro
p' = cos k ( r ! aot ) ! cos k ( r + d cos " ! a0t ) (11)
r r + d cos "

Expanding the second term and assuming d to be much smaller than both, the
wavelength (≈1/k) and the observation distance r,

Pro d Pr
p' = ( k d cos ! ) sin k ( r " aot ) + ( cos# ) 0 cos k(r " a0t) (11b)
r r r
of the two terms in (11b), the first has the 1/r dependence that will ensure energy
flux conservation at all distances, while the second will decay faster and will be
negligible for distances r>>1/k; this is the “near-field” dipole sound, which can
be important near the source. We concentrate here on the “far-field” contribution
(first term in (11b)).

The calculation of the far-field power flux is as before, remembering that it is still
a radial flux, even though there is an angular dependence in ϑ. So we still have
the first equality in Eq. (10), and the calculation is straightforward. We obtain

5
 1 $ ro '2
!( r," ) = & P k d cos " ) (12)
2 # o ao % r (

cos2 !
This flux has now a distribution (no sound at 90° to the dipole axis,
r2
maximum along the dipole axis). The total radiated power is
2 (Pr k d )
2

Pd = 2! # "( r,$ ) r 2 sin $ d$ = ! o

!
(13)
o 3 % o ao
1
Aside from the factor, this differs from Pm, the monopole power, by the factor
3
(kd) = (2!d / ") , which is, by assumption a small number, and becomes smaller
2 2

the longer the wavelength is (or the lower the frequency ! = ao /k ).

An important observation is that a physical dipole requires application of an

external oscillatory force. A possible physical implementation of the radiating
dipole is any vibrating compact object, such as a fan or a turbine blade. The
dipole axis is then the direction of vibratory motion, and the surrounding air is
forced back and forth as it would between the hypothetical two monopoles in
counter-phase.

(c) Quadrupoles

Strongly turbulent flows, such as an engine exhaust jet, are known to be strong
sources of acoustic radiation. If the jet is steady and subsonic, there is no
possibility of macroscopic monopole (expansion/contraction) type of radiation,
and since there is no external force in the body of the fluid, no dipole sources
either. However, there are fluctuating pressures at different points (turbulent
eddies), exerting forces on each other with zero net on the larger scale. The
lowest order “multipole” with these features is the Quadrupole, which can be
built up from two dipoles with a common axis, and separated by 2d and with
opposing directions.

The detailed derivation is similar to that for a dipole. We calculate (in the far
field)
Pr
p' = !2 o ( k d cos" ) cos k ( r ! ao t )
2
(14)
r

and for the acoustic power flux,

" Pro % 2 ( kd cos ( )

4

! = 2$ ' (15)
# r & ) o ao

6
which integrates for all directions to a radiated power

8 (Pr ) ( kd )
2 4

Pq = ! o (16)
5 " o ao

The quadrupole radiator pattern is more sharply directional along the axis
(cos4 ! ) , and is an additional (kd)2 weaker than the dipole, with an even stronger
rate of increase with frequency.

It should be noted that the collinear-dipole type of quadrupole is not the only
one possible, but all of them share the (kd)4 feature (although with different
angular patterns).

7
16.50 Lecture 37

Subjects discussed: Jet noise, Turbomachinery noise

Having reviewed the mechanisms for noise propagation and the three types of
acoustic sources - Monopoles, Dipoles and Quadrupoles, we can now examine
aircraft engines as sources of noise.

Jet Noise

We begin with jet noise because historically it was the principal source of concern
with the introduction of jet transports. Jet noise is generated by the turbulent
mixing of the high speed exhaust jet with the ambient air. We can think of the
mixing region as a region of strong turbulent flow,

From Kerrebrock, Jack L. Aircraft Engines and Gas Turbines. 2nd edition. MIT
Press, 1992. © Massachusetts Institute of Technology. Used with permission.

that can be modeled as a random array of Quadrupoles, since there is no mass

source, and no force in the fluid.

In a turbulent jet, in order of magnitude P ~ ! o u 2 (u being the jet velocity) and

the “size” of the fluctuating regions is some fraction of the diameter D of the jet.
Similarly, the spacing d of these regions must also be some factor times D. The
!
wavenumber k = can be estimated by noting that ! ~ u /D. We then have,
ao
(! u D)
2
2
" u %4 u 8 D2
!
o
Pq~ $ D ' ~ (17)
! o ao # Dao &
o
ao5
The striking feature is the scaling of this radiated power with the 8th power of the
jet velocity (a result due to Lighthill (1963). To compare it to the jet kinetic power,
note that
P jet ~ ( ! o uD2 ) u 2 = ! o u 3 D2 (18)
so that
Pq ! u $
5

~ # & = M 5jet (19)

P jet " ao %

1
This very strong sensitivity to jet velocity (or Mach number) is one of the
significant disadvantages of the pure turbojet as compared to the high bypass
turbofan (the other being its lower propulsive efficiency, although that
disappears at a high enough flight Mach number).

Turbomachinery noise

Vibrations of lifting blades, or periodic passage of lifting blades past the

observer, produce dipole-type radiation, which emits a power given by Eq. (13).
! u
Estimating in this case ro ~ d ~ c (blade chord), k ~ ~ , where u is now the
ao a o c
blade speed, and P = ! o u 2 , the power radiated per blade is

Pbl ~ ! ou 6c 2 / ao3 (20)

The force on one blade is of the order of Fbl ~ ! o u 2cH ~ ! o u 2c 2 , and so the
“acoustic efficiency” is
Pbl " u %
3

!bl.noise = ~ $ ' = MT3 (21)

Fbl u # ao &

This is noise generated, but the duct, even with no special absorbing design,
attenuates the sound, and in fact, forbids propagation for the lower range of
frequencies. The frequency is here related to the blade Mach number MT and the
number B of blades:
! rT B! rot rT
= = BM T (22)
ao ao

and so MT must be higher than a certain threshold for blade noise to propagate.

In order to understand this phenomenon let us look at the simple case of sound
emitted by a moving source, meaning that the pressure perturbation on a given
surface x=0 is a sliding sinusoid, moving a velocity vs in the y-direction.

(-) vs

λs X
p’
(+)

2
From the Cartesian wave equation
! (Lecture 36), a plane wave propagating in the
direction of the “wave vector” k has the general form
"i(#t"kx x"ky y )
p! = Re[ pˆ e ] (23)

provided kx, ky and ω satisfy the “dispersion relation”, that is obtained by

substituting (23) into the wave equation itself:
! 2 = a0 2 (kx 2 + k y 2 ) (24)

Also, the moving boundary condition represented in the figure above is of the form
2"
(23), with x=0, ky=(2π/λs) and ! = v s . Substituting these ω and ky in (24) and
#s
solving for kx,
2! v s 2
kx = ± ( ) #1 (25)
" s a0

and we can see that kx is real only if vs>a0 , i.e., only if the source is moving at a
supersonic speed. The direction of the sound waves is given by

! = tan"1 (k y /k x ) = sin"1 (a0 /v s ) (26)

namely, the Mach angle at the wall Mach number.

For a source moving at subsonic speed, the sound is in fact exponentially attenuated
over a distance of the order of λs/2π. Physically, the reason is that for a supersonic
source the waves from the infinity of pressure peaks and troughs on the wall
interfere constructively with each other when their direction is as in (26). For a
subsonic source there is no such direction, and interference is always destructive.

In the case of turbomachinery, the supersonic relative velocity can come from two
sources:
a) Supersonic tip speed

b) Subsonic tip speed, but interacting blade rows

To understand the second situation, consider a rotor operating in the vicinity of a

stator. The stator blades produce a pressure variation

p' stator ! (incidence )eiV "

where V is the number of stationary vanes.

The disturbances from the rotor are periodic in rotor coordinates, so

3
 +iB(" # $ t)
incidence ! e R

where B is the number of moving blades. It follows that

i[(V ! B)" ! B# t]
p' stator = e R

So the Effective Speed of Rotation of the disturbance is

B!
! Eff = ( ddt" )const. phase = R
V #B

In early engines there was a tendency to make V ≈ B and they were very noisy. Now
we use the rule V = 2B + 1, so that the rotor-stator interaction is no more likely to
generate propagating disturbances than is the rotor rotation itself.