48th AIAA Aerospace Sciences Meeting Including the New Horizons Forum and Aerospace Exposition AIAA 2010-911

4 - 7 January 2010, Orlando, Florida

The Limits of Two-Temperature Model

Chul Park∗
Korea Advanced Institute of Science and Technology, Daejeon, Korea
Summary M : Second moment of transition, Eq. (19).
This paper reviews the issues surrounding the two- [M ] : Matrix in quasi-steady steady state master
: equation, Eq. (13).
temperature model used commonly in describing the flows Ni : Number density of species i, mol/m3 .
in which its thermodynamic state departs far from equilib- Nx : Number density of unspecified colliding species,
rium. The review begins with what the two-temperature : mol/m3 .
model is, and why it was developed. Then the review ex- Rji : Rate of energy transfer from mode j to mode i,
plains why the model is the way it is, and what it cannot : W/kg.
do. Seemingly slow rotational relaxation at high tempera- t : Time, sec.
tures and radiation-ionization coupling are cited as being T : Heavy particle translational temperature, K.
the phenomena that the existing two-temperature model Ta : Geometrically-averaged temperature, K.
cannot cope with. It ends with a suggestion as to what Te : Electron translational temperature, K.
one could do to improve on it. A three-temperature model Tel : Electronic excitation temperature, K.
recognizing the rotational temperature or a radiation tem- Tr : Rotational temperature, K.
perature different from heavy particle translational tem- TR : Radiation excittion temperature, K.
perature is suggested. Tvi : Vibrational temperature of species i, K.
α : Rossland mean absorption coefficient, cm−1 .
Nomenclature
i : Energy level of state i, J/kg.
ai : symbolic form of a factor in quasi-steady state κ : Absorption coefficient, cm−1 .
: distribution, Eq. (13). τv : Vibrational relaxation time, sec.
A, B : Hypothetical atoms.
bi : symbolic form of a factor in quasi-steady state Subscripts
: distribution, Eq. (13).
ci : symbolic form of a factor in quasi-steady state E : Equilibrium.
: distribution, Eq. (14). R : Radiation.
di : symbolic form of a factor in quasi-steady state s : Shock wave.
: distribution, Eq. (14.
D : Dissociation energy, J/kg. 1. What is Two-Temperature Model
Ee : Electron translational energy, J/kg. 1.1. Different Temperatures
Eel : Electronic energy, J/kg.
Eev : Electron-electronic-vibrational energy, J/kg. Imagine an object flying in air at a high speed, i.e., at
Er : Rotational energy, J/kg. a speed in excess of 2000 m/s. A shock wave, or a system
Ev : Vibrational energy of all molecules, J/kg. of shock waves, will form around the object. The flows
Evi : Vibrational energy species i, J/kg. behind these shock waves are hot. One wants to know the
Et : Heavy particle translational energy, J/kg. pressure and heat transfer rate to various different areas of
Etr : Translational-rotational energy, J/kg. the object’s surface. The high temperature also produces
Ei : Energy in mode i per unit mass, J/kg. radiation emission, which will heat the flying object. In
f : Energy feedback factor. order to determine these, one needs to know the spatial
kf : Forward reaction rate coefficient, m3 /(mol-s).
distribution of thermodynamic state in the flow.
kr : Reverse reaction rate coefficient, m3 /(mol-s) orOne can also imagine a flow inside a rocket engine. The
: m6 /(mol2 -s). distribution of pressure will determine the thrust of the
K (c, i) : Rate coefficient for transition from free state to
engine. Pressure is affected by temperature. Temperature
: state i, m6 /(mol2 -s). will also determine the heat transfer rate to the nozzle
K (i,c) : Rate coefficient for transition from state i to
wall. Here again, thermodynamic state must be known to
: the free state, m3 /(mol-s). determine the performance of the rocket engine.
K (i,j ) : Rate coefficient for transition from state i to When the flow moves relatively slowly, or, conversely,
: state j , m3 /(mol-s). if the object or the engine we are considering is very large,
m : Hypothetical molecular. then there will be sufficient number of molecular collisions
within any small spatial volume to bring the thermody-
∗ Visiting Professor, Department of Aerospace Engi- namic state to the so-called equilibrium state. In a high
neering; Fellow, AIAA; cpark216@kaist.ac.kr. speed flow, the flow does not dwell at one location long

Copyright © 2010 by the author. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
2
enough to bring itself to the equilibrium state. That is,
the flow is in a nonequilibrium state.
In chemical industry, rate of change of chemical com-
position in a chemically reacting medium is calculated by
integrating the so-called chemical rate equations. This
procedure is well established. In the calculation, the gas
mixture under consideration has only one temperature.
At a high temperature, energy in a gas is contained in
the translational, rotational, vibrational, and electronic
modes. If the gas is ionized, the translational energy of
electron gas can be very different from that of heavy par-
ticle gas. For this reason, the translational energies of the
heavy particle gas and the electron gas must be recognized
as being different. Also, the vibrational energies can be
different among different molecules.
The energy contained in any mode in any atom or
molecule is usually designated by its temperature. In a
thermochemically nonequilibrium gas mixture, the tem-
peratures that characterize these different modes of en-
ergy can be different from each other. That is, there can
be several characteristic temperatures.
In the so-called two-temperature model, one approxi-
mates this situation with two main assumptions. First,
one assumes that there are only two different temper-
atures in this situation. The rotational temperature of
molecules is assumed to be the same as the translational
temperature of heavy particles. Vibrational temperatures
of all molecules are assumed to be the same as electron
temperature and electronic temperature.
Tr = T ≡ Ttr , Tv 1 = Tv2 = ... = Te = Tel ≡ Tev
Second, the forward and reverse rate coefficients kf and
kr for the chemical reactions involving molecules are as-
sumed to be a function of a geometrically averaged tem-
perature Ta
Ta = Ttr Tev , (2)
kf = kf Ta , kr = kr (Ta ). (3)
There is an extention of Eq. (2) in the form
Ta = Ttrs Tev1 s
−
where the parameter s is taken to be somewhere between
0.5 and 0.7. No clear advantage of 0.7 over 0.5 has yet
been demonstrated.
Because temperature of a gas mixture is split into
two, calculation of thermochemical nonequilibrium bears
a burden of describing how these two temperatures inter-
act with the flow motion and between them. It is usually
easier to work with the energy contained in each modes,
Ei, than to work with temperature. Symbolically, one can
write
dEi = R (4)
dt ji
j bation period, dissociation rises slowly. In this period,
By summing these up over the appropriate modes, one
obtains the rate of change of translational-rotational en-
ergy Etr and that for the electron-electronic-vibrational
energy Eev as
dEtr = dEt + dEr
dt dt dt
dEev = dEe + dEel + dEvi
dt dt dt dt
i
There are some other theoretical treatments that are,
strictly speaking, not in an exclusive domain of the two-
temperature model, but are loosely associated with it.
They are: 1) quasi-steady state distribution of internal
states, 2) the preferential dissociation (or ionization) phe-
nomenon, 3) the diffusion model for vibrational (or rota-
tional) relaxation, and 4) collision-limiting correction for
vibrational relaxation times.
1.2. Master Equation
Quasi-steady state is a state in which the rate of col-
lisional transitions into an internal state and the rate of
the transitions out of the state are nearly equal. That
is, the difference between the two rates is much smaller
than either of the two rates. This condition can exist to
all states except the ground state. The condition exists
after a short time has past after the flow passes through
a shock wave. This is illustrated in Figure 1.
(1)
Figure 1. A schematic explaining the occurrence
of incubation period and quasi-steady state pe-
riod.
In Figure 1, the variation in the fraction of atoms pro-
duced by a dissociation process is shown schematically
along the distance behind the shock wave. Immediately
behind the shock wave, in the region identified as incu-

collisions cause the ground-state molecules to be excited
to higher states. After this period, when the high states
are sufficiently populated to cause steady dissociation, is
the quasi-steady state period as indicated.
What happens in these two periods can be found by a
master equation calculation. In a master equation calcu-
lation, the number density, or population, of all internal
states is calculated by time-integration by accounting for
all incoming and outgoing transitions. In order to carry
out such a calculation, one must first know all state-to-
state transition rate coefficients. Such state-to-state tran-
sition rates are known only for hydrogen molecules collid-
ing with hydrogen atoms, helium atoms, or another hy-
drogen molecules. In Figure 2, a result of such calculation
[1] is shown.
Figure 2. Master equation calculation for H2 col-
liding with H2 [1].
1.3. Preferential Dissociation and Diffusion Model
Figure 2 shows a flow of pure H2 that passed through
a strong normal shock wave. Its translational tempera-
ture is 20,000 K, and is held constant at that temper-
ature by a magical means. The figure shows results of
calculations made by several slightly different methods.
All these calculations show that the energy contained in
the rotational and the vibrational modes are in a plateau
over a certain period of time. Obviously, in this period,
the quasi-steady state assumption would be valid. Af-
ter the plateau period, the energy content increases to its
equilibrium value asymptotically. Detailed examination
reveals that the difference between the incoming and out-
going rates is indeed very small compared with either of
them even in this asymptoting period. Thus, the quasi-
steady state assumption is valid everywhere except in the
incubation period.
3
Preferential dissociation refers to the fact that dissoci-
ation is likely to occur from high vibrational, rotational,
or electronic states. Because of this, vibrational (or rota-
tional energy larger than the average vibrational (or ro-
tational) energy is lost when a molecule dissociates. The
energy taken away is called feedback energy, and is ex-
pressed as a fraction of the dissociation energy of the
molecule. This energy feedback factor, called f here, is
calculated to be some 0.3 to 0.8 depending on the case
and the assumptions used.
The diffusion model for vibrational relaxation is de-
vised to correct the so-called Landau-Teller model at high
temperatures. Landau-Teller model describes the ladder-
climbing excitation process of vibrational mode rather
well at relatively low high temperatures. This is because,
at such a relatively low high temperature, only the first
few vibrational states are excited. Landau-Teller model
is exact when the energy gap between neighboring states
is constant and the neighbor-to-neighbor transition rate
increases linearly with vibrational quantum number. For
the lowest few vibrational states, these two conditions are
met. However, at very high temperatures, high vibra-
tional states are excited. Between these high vibrational
states, collisional transition rates do not increase linearly
with vibrational quantum numbers. Nor are these states
separated by a constant energy gap.
For these high vibrational states, the assumption of
a constant collisional transition rates would be more ap-
propriate. The vibrational states can be considered con-
tinuously distributed at high temperatures. In this case,
vibrational relaxation phenomenon occurs by satisfying a
diffusion equation in energy space. By solving a diffusion
equation under simplifying assumptions, a correction fac-
tor is devised to the Landau-Teller equation in the form
dEv = EvE − Ev | Ts − Tv |s 1 −
(5)
dt τv Ts − Tvs
s = 3.5exp(−5000/Ts). (6)
This is a preferred equation when the post-shock temper-
ature is very high.
The collision-limiting correction for vibrational relax-
ation times refers to the need to impose the condition
that vibrational relaxation time cannot be smaller than
the collision time. This is done by adding a collision time
to the relaxation time i.e.,
τv = τv,MW + τcol (6)
where τv,MW is the vibrational relaxation time given by
the formula of Millikan and White [2] and τcol is the elastic
collision time.
At a high hypersonic flight Mach number, i.e., Mach
number higher than 15, the incubation period and the
quasi-steady state period intermingle and become indis-
tinguishable. This is because dissociation can occur even
4

from a relatively low vibrational state: even when the 2.2. Freezing in Expanding Flow
low vibrational states are being populated, dissociation
occurs already from those relatively low states. The two- On the other hand, the flows in an expanding noz-
temperature model is intended to describe the thermody- zle brought an equally intriguing problem. Starting in
namic state even under this circumstance. 1950s, several hypersonic shock tunnels were built world-
wide. The temperature in the settling chamber region of
2. Why was the Two-Temperature Model Born these tunnels were between about 2000 K and 8000 K. In
the test sections of these tunnels, calculated temperatures
2.1. Nonequilibrium Radiation Overshoot were below 300 K. Yet, one observed radiation emitted
Two unexpected phenomena led to the ushering in of from this supposedly cold flows. The best known method
the two-temperature model. The first was the radiation of measuring temperature, which was developed in the
over-shoot phenomenon behind a sock wave. According combustion community and which was based on the elec-
to the traditional gasdynamics, which recognizes only one tronic excitation temperature of sodium (sodium D-line
temperature, the temperature would be the highest im- reversal method), was utilized to determine the tempera-
mediately behind the shock wave. Intensity of radiation ture in the test section. The temperatures so determine
emitted by a hot gas is a very steep function of temper- were much higher than the calculated temperatures. This
ature. Therefore, the one-temperature model would pre- phenomenon is explained schematically in Figure 4.
dict that radiation would be the strongest immediately The top left figure in Figure 4, below the sketch of the
behind the shock wave and decay thereafter, as shown nozzle, schematically shows the calculated temperature.
schematically in the top two figures in the left of Fig- The second from the top on the left shows the calculated
ure 3. The intensity immediately behind the shock wave radiation intensity in a logarithmic scale. The observed
would be very large. radiation intensity is shown in the left bottom figure.
In order to determine the radiative heating in the shock The top figure on the right of Figure 4 shows the trans-
layer over the Apollo entry vehicle, experiments were lational temperature and vibrational temperature. As
made in a shock tube to study this phenomenon. The shown, vibrational temperature is frozen at a fairly high
experiments found that there is no radiation immediately temperature. As a consequence, radiation emission also
behind the shock wave, as shown in the bottom left figure is frozen. The sodium D-line method measures essentially
in Figure 3. Radiation intensity rose from zero to a peak the vibrational temperature of the flow.
and subsided to what was believed to be the equilibrium
value. The fact that radiation would reach a peak in be-
tween was referred to as radiation overshoot phenomenon.

Figure 4. Temperature and radiation observed in

a hypersonic shock tunnel.
Figure 3. Schematic of the radiation intensity vari- A similar behavior is seen also in the exhaust plumes
ation behind a shock wave. of rocket engines. The temperature at the exit of a rocket
 5

engine is typically of the order of 1000 K. At high altitudes be related by the equilibrium constant unconditionally at
where ambient pressure is very low, the flow expands fur- all times; these one-way rate coefficients kf and kr do not.
ther in the plume. Temperature steadily decreases. Ulti- The phenomenological expression for dissociation rate
mately, in vacuum, they reach a very cold temperature of is
the order of 10 K. Even at such low temperatures, radia- dNA = dNB = − dNm = k N N − k N N N (10)
tion is being emitted. f M x r A B x
These two phenomena represent the most severe de- dt dt dt
parture of the flow behavior from the conventional one- A question existed as to whether the coefficients kf and
temperature description. The two-temperature model de- kr are uniquely functions only of temperature T .
veloped over the years were able to explain these two phe- Extensive experiments on a flow behind a shock wave
nomena successfully. revealed that, except in the incubation period mentioned
earlier, the kf and kr are indeed uniquely functions only
3. Why Is The Two-Temperature Model of T . This then proves that Eq. (9) is incorrect.
The Way It Is One then searches for the condition under which the
3.1. Rate Coefficients rate coefficients kf and kr are uniquely functions only
of temperature, at least in this one-temperature environ-
The most frequent objection to the two-temperature ment. One notices that, outside of the incubation region,
model concerns the fact that an irrational expression of one can write the master equation, except for the ground
the geometrically-averaged temperature Ta is assumed to state i = 0, as
control reaction rates. One wishes that a rational ex-
pression, though complicated, could be possible for the dNi = K (j,i)N N − K (i, j )N N
dt j x i x
reaction rates as a function of temperature. At relatively j j (11)
low high temperatures, i.e., in a one-temperature envi- +K (c, i)NA NB Nx − K (i, c)Ni Nx ≈ 0.
ronment, the rate of dissociation of a molecule m into an
atom A and atom B But there is mass conservation condition
m

 A+B (7) Ni = Nm (12)
i
can be described microscopically as which governs the ground state population.
dNA = dNB = − dNm = Eqs. (11) and (12) form a set of linear algebraic equa-
dt dt dt K (i, c)Ni Nx tions of the form
i (8)
− K (c,i)NA NB Nx [Mi,j ]Nj = ai Nm + bi NA NB (13)
i
The solution of this algebraic equation is in the form
By defining the forward rate coefficient to be the sum Ni = ci Nm + di NA NB (14)
of rates of all such transitions divided by the total number
of initial states



kf = i K (i,c)Ni , kr = K (c, i) (9)
where ci and di
ci = [M ] 1 ai , di = [M ] 1 bi
− −

Ni i
are unique functions of T only. That is, number density
one should be able to derive a rational expression for the of an arbitrary internal state is a linear combination of
reaction rate coefficient kf , the argument goes. All re- the number density of the molecule and the product of
search effort intending to produce a rational expression the number densities of atoms A and B . By substituting
for reaction rates start from here. But, if indeed kf and Eq. (14) into Eq. (8), one obtains the dissociation rate in
kr are as given by Eq. (9), they should be functions not the form
only of temperature T but also of time. That is because dNA = dNB = − dNm = [ K (i,c)c ]N N
Ni changes with time and the change pattern will differ dt dt dt i m x
from case to case. (15)
The expression, Eq. (9), is called one-way rate coef- − [−K (i,c)di + K (c,i)]NA NB Nx
ficient. Unfortunately, this reaction rate is not what we
observe experimentally. Moreover, these kf and kr do not One can immediately identify kf and kr to be
generally satisfy the condition that the net rate will be
zero under equilibrium. The coefficients kf and kr must kf = K (i,c)ci , kr = [K (i,c)di + K (c,i)] (16)
6

The functions ci and di were named homogeneous solution which can be named the second moment of transition rate
and particular solution of the quasi-steady state master
equation, Eq. (13).
To distinguish these kf and kr from the one-way rate
coefficient,
M = Nx
 K(, )( − )2δ (19)
coefficients in Eq. (9), they will be called quasi-steady
state rate coefficients. It cannot not be proven analyti- is bound. One first expresses K ( , ) in terms of K (, )
cally, but numerical calculation shows that these kf and by invoking the detailed-balance relationship. One then
kr are always related by the equilibrium constant. expands N and K (, ) by a Taylor series in δ around
Thus, the quasi-steady state rate coefficients kf and kr N and K (,). Then these Taylor series forms are substi-
contain ci and di, which are obtained by inverting an alge- tuted into Eq. (18). Neglecting the terms higher than the
braic equation of the form Eq. (13). The inversion process second power in δ, and neglecting the bound-free transi-
destroys all rational forms existing in K (i, c), K (c,i) and tions, one obtains
Ni . That is, rational expression is impossible for kf and
kr . ∂ N ∂ ∂ N
Question arises as to how large is the difference be- ∂t ( N ,E ) ≈ ∂ [M ∂ ( N ,E )] (20)
tween the one-way rates and the quasi-stead state rates.
For vibrational and rotational problems, the difference is Eq. (20) is in the form of a diffusion equation. Thus,
typically a factor of 3: the one-way rates are three times one can say that, at high temperatures, internal energy
larger. For electron-impact ionization processes, the dif- excitation spreads like diffusion. By assuming M = con-
ference is many orders of magnitude: the one-way rates stant, an approximate solution to Eq. (20) was obtained.
are that much larger. Some simplification and approximation made to that re-
sult lead to Eq. (5).
3.2. Energy Feedback Factor
This issue of the one-way rate vs quasi-stead state rate
enters also in calculating the energy feedback factor f ,



f = D1 i K (i, c)(DK−(i,c



i )Ni − i K (c,i)(D − i )NA NB
i )Ni − i K (c,i)NA NB
(17)
By substituting Eq. (14) for Ni , one sees that the f is also
determined by the functions ci and di . Thus, a rational
expression is impossible for f also.
In existing CVDV model [3] and in other similar mod-
els [4,5], this procedure was not followed. Therefore, the
feedback factors derived in those models are wrong. These
erroneous feedback factors are generally too high.
3.3. Diffusion Model
In many areas of statistical mechanics, quantum-
mechanical results approach the results of classical me-
chanics calculations. The vibrational and rotational en- Figure 5. Shock stand-off distance over a sphere
ergy levels can be considered to be continuously dis- in air [6].
tributed. Instead of discrete energy levels i and j , we
shall consider a continuously distributed energies  and  . 4. What The Two-Temperature Model Cannot Do
The summation signs in Eq. (11) can be replaced by an 4.1. Shock Stand-off Distance
integral sign in the form
dN =  [K ( ,)N − K (,  )N ]N d
The two-temperature model is not able to correctly pre-
dict the shock stand-off distance for a sphere at interme-
dt x diate hypersonic speeds, i.e. between Mach numbers of
(18) 10 to 15 in air. In this Mach number range, atoms and
−K (,c)N Nx + K (c,)NA NB Nx molecules exist simultaneously. Figure 5 shows the com-
parison of the measured and calculated shock stand-off
Eq. (18) is an integral-differential equation. It can be distances [6]. The measurement was made in a ballistic
approximated by a differential equation if the quantity, range. The calculations were made using the extended
 7

two-temperature model in which the vibrational temper-
ature of O2 , N2 , and NO were considered different from
each other. The solid squares and triangles are calculated
values. The open squares and triangles are the measured
values. As seen, measurement tends to underestimate the
shock stand-off distance.
In Figure 6, comparison is made for a test in a CO2
flow [7]. The measurement was made in a shock tun-
nel. The calculation was made using the two-temperature
model. As seen, the calculation severely underestimates
shock stand-off distance.
The spectrum of the radiation emitted in the flow be-
hind a shock wave was measured and analyzed by several
researchers. In air and in N2 , the radiation overshoot phe-
nomenon mentioned earlier was prominently seen. At the
point of peak radiation intensity, the spectrum was ana-
lyzed in detail. From such analyses, the vibrational and
rotational temperatures were determined. In Figures 7 to
9, these results are shown.

Figure 6. Shock stand-off distance over a sphere

in CO2 [7]. Figure 7. Comparison of vibrational and rota-
tional temperatures in N2 behind a shock wave ob-
Since the publication of this photograph, a more de- served in emission spectrum by Sharma and Gille-
tailed calculation [8] was made. The calculation ac- spie [9]. Flow velocity = 6.2 km/s.
counted thoroughly for the nonequilibrium processes oc-
curring in the nozzle of the shock tunnel. The result was
that the calculated shock stand-off distance was slightly
larger than the value seen in Figure 6, but still much
smaller than the measured value.
Not being able to correctly predict the shock stand-off
distance over a sphere will have an impact on the aero-
dynamic characteristics of a flying body in general. If
one cannot correctly predict the shock stand-off distance
over a sphere, one is not able to correctly describe the
shock shape over a flying object. Because shock shape
determines pressure distribution, this means that pres-
sure distribution, and hence lift, drag, and moments, are
not correctly predicted.
The only way to explain the large shock stand-off dis-
tance is that the rotational temperature in the shock
layer is lower than the translational temperature. The
molecules are thus behaving not exactly as a diatomic
molecule but somewhat as a monatomic atom. This vio-
lates the first assertion in the two-temperature model that
the rotational temperature is the same as the translational Figure 8. Comparison of vibrational and rota-
temperature. tional temperatures in air behind a shock wave
observed in emission spectrum by Fujita et al [10].
4.2. Rotational Temperature in Shock Tube Tests Flow velocity = 11.9 km/s.
8
In both Figures 7 and 8, one sees that the rotational
temperature is no higher than the vibrational tempera-
ture at the peak-intensity point. In fact, in Figure 8,
made at flow velocity in excess of 11.9 km/s [10], rota-
tional temperature appears much lower than vibrational
temperature. That is, rotational relaxation is slower than
vibrational relaxation at such high temperatures. This
result agrees with the results of shock stand-off distance
measurements sown in Figures 5 and 6, that rotational re-
laxation is slow, and so that the two-temperature model is
wrong. Though not shown, another set of data [11] exists
which is nearly identical to that shown in Figure 8.
One may doubt the results of emission measurements
on account that they represent what is happening to the
electronically excited states. One would prefer to see the
rotational temperature of the ground electronic state. In
Figure 9, such result is shown. In this case, coherent
anti-Stokes Raman spectroscopy measurement was made
of the ground electronic state of N2 [12]. As shown, the
measured rotational temperature is lower than the vibra-
tional temperature, in agreement with the results shown
in Figure 8.
Figure 9. Comparison of the measured and calcu-
lated vibrational and rotational temperatures in
air in the nonequilibrium region behind a shock
wave, observed by coherent anti-Stokes Raman
spectroscopy [12]. Flow velocity = 5.3 km/s.
4.3. Rotational Relaxation in an Expanding Flow
Rotational temperature of NO was measured also in
an expanding flow. In Figure 10, a result of such an
experiment [13] is shown. In this experiment, the fluo-
rescence spectrum was obtained from a highly expanding
flow. The figure shows the distribution of population of
each rotational states up to the rotational quantum num-
ber of about 15. As seen, the higher rotational states
are populated more in comparison with the lower states.
This also means that rotational relaxation is slow at high
temperatures: the higher states are excited only at high
temperatures.
This phenomenon will impact the performance of in-
struments aboard an Earth satellite: the substance issu-
ing from the attitude-control rocket engines will condense
on these instruments if the rotational temperature is high.
This is because the high rotational temperature will cause
spreading of the rocket effluent over a larger volume.
Figure 10. Rotational temperature in an expand-
ing NO flow [13].
4.4. Double-Cone Flow
A flow over a double-cone geometry is perhaps the most
complicated flow but which is relatively easy to experi-
ment in a shock tunnel. For this reason, a concerted effort
has been made to experimentally study this flow. Flow
shapes have been studied optically, and pressure and heat
transfer rate distributions were measured. Computational
fluid dynamics calculations were performed to reproduce
these experimental data. In the enthalpy range where
dissociation is insignificant, a satisfactory agreement was
seen between the measurement and calculation. However,
in a highly dissociating regime, there were significant dis-
crepancies. The calculations were made using different
versions of the existing two-temperature model. In Fig-
ure 11, one latest such result [14] is shown.
The reason for the discrepancy is presently unknown.
There are two possibilities: rotational noneqilibrium and
breakdown of Navier-Stokes equation. In this flow, a rapid
spatial change occurs in translational temperature. Per-
haps the rotational mode cannot follow this fast change.
Secondly, at least in the region where flow is laminar, the
Navier-Stokes equation may be breaking down. Navier-
Stokes equation with the traditional transport properties

is based on the assumption that the velocity distribution
in the flow departs only slightly from the Maxwellian dis-
tribution. In the flow under consideration, this condition
is not met. A general hydrodynamic equation, which ac-
counts for this aspect [15] may have to be used.
Figure 11. Measured and calculated heat transfer
rate over a 25 -55 double cone in air [14].
◦ ◦
4.5. Radiation-Ionization Interaction
When radiation intensity is strong, it will cause ioniza-
tion of the flow. Two such examples are shown in Figures
12 and 13.
Figure 12. Measured and calculated electron den-
sity behind a shock wave in a hydrogen-helium 4.5. The Consequences of Failure of the Two-
mixture [16]. Temperature Model
9
In Figure 12, the electron density is measured behind
the shock wave in a shock tube [16]. The shock tube is
driven by an arc-heated helium. The flow pressure behind
the shock wave is sufficiently high to produce equilibrium.
Yet, the measured electron density is much higher than
the calculated equilibrium value. This is caused by the
absorption of radiation emanating from the driver of the
shock tube.
In Figure 13, electron density is measured ahead of the
shock wave [17]. As is seen, electron density is observed 6
m ahead of the shock wave. Such is possible only by the
absorption of radiation emitted by the downstream flow
or the driver gas.
Figure 13. Electron density in air ahead of the
shock wave measured in a shock tube [17].
Park [18,19] and Furudate [20] examined this in-
teraction of radiation and ionization phenomenon for
hydrogen-helium mixtures. For hydrogen at least, absorp-
tion of radiation is causing an early ionization of the flow:
the Lyman-α line emits and absorbs strongly. This radi-
ation emitted in the downstream region shines upstream,
gets absorbed there, and causes ionization. In Figure 14,
the rate coefficient for electron-impact ionization of H
H + e → H+ + e + e
is shown as a function of the power absorbed by the
Lyman-α line. The absorption power is normalized by
the emission power of the line. As seen, the ionization
rate increases linearly with the absorbed power. The two-
temperature model is not able to cope with this situation.
10
It seems obvious that rotational relaxation is quite slow
at high temperatures. But then what ill effect will it
bring?
In air, for practical flying machines, probably the im-
pact on aerodynamic characteristics is minor. The prac-
tical hypersonic flying machines fly in a density regime
where the nonequilibrium region is limited to a small
portion of the flow surrounding it. In fact, even two-
temperature phenomenon had little impact on Apollo or
Space Shuttle. It could possibly affect the aerodynamic
characteristics of small flying machines, e.g. Stardust or
Hayabusa. The rotational nonequilibrium phenomenon
will have no effect in the atmosphere of Venus because its
atmosphere is denser than that of Earth.
-17
10
Forward rate coefficient, cm /s
-18
10
3
10
-19
Adiabatic
-20
Rate coefficient at Te = 10,000 K for
10
H + e -> H + e + e
as a function of L-α absorption power
-21
10 2 3 4 5 67
1
2 3
Absorption power/emission power
Figure 14. Rate coefficient for electron-impact
ionization of H at Te = 10,000 K as a function
of power absorbed by L-α line [16].
As was seen in Figure 6, the failure of the two-
temperature model for Mars will be a problem worthy of
examination. Until now, the Mars entry vehicles stayed
in the atmosphere for relatively short time. If, in the fu-
ture, the vehicles are made to fly over a long time, then the
aerodynamic characteristics will be a serious subject. The
same can be said of the vehicles flying in the atmosphere
of Titan. Even though Titan has a thick atmosphere, the
flight speed in Titan will be of the order of 4 to 5 km/s. N2
will be the dominant species, and therefore, its rotational
temperature will affect the aerodynamic characteristics.
The phenomenon associated with ionization and radi-
ation remain beyond the reach of the two-temperature
model. In RAM-C series flight experiments made in the
U.S. and the OREX flight experiment made in Japan,
electron density and temperature have been measured.
The calculated values did not completely agree with the
measured values. As a consequence, the radio blackout
phenomenon associated with ionization is not fully char-
acterized.
The contamination of mirror surfaces and solar panels
on Earth satellites caused by condensation of the rocket
effluent has not yet been fully characterized using the two-
temperature model. The experimental data obtained in
the high-vacuum facility in DLR Goettingen, which could
illuminate this problem, are not yet available.
Not being able to correctly predict heat transfer rates
over a double-cone remains as a nagging problem. As
mentioned, possibilities are that rotational nonequilib-
rium is significant in the flow and the Navier-Stokes equa-
tion may be breaking down. More work is needed in the
future.
The problem of radiation-ionization coupling is a prob-
lem for outer-planet exploration. Outer-planet explo-
ration is not yet finished. Not being able to correctly de-
scribe the flow field over an entry body in the atmosphere
of the outer planets will impact the planning process for
those outer-planet missions.
5. What We Could Do To Improve On The
Two-Temperature Model?
5.1. Rotational Temperature
The foregoing study leads to the one inevitable con-
clusion that rotational temperature must be recognized
as a separate temperature. This will lead to a three-
+
temperature model. In order to do so, one must first
know the rotational relaxation time.
Rotational relaxation times were measured in 1960s
4 5 67
10
2 3 4 5 67
100 and 1970s up to a temperature of about 1000 K in a
shock tube and a wind tunnel. From the measurements,
the number of collisions required to bring equilibrium be-
tween the rotational and translational modes was calcu-
lated. This collision number was found to be below about
10. In Figure 13, the results of these early works are
shown. A tendency for the collision number to rise with
translational temperature was recognized by early inves-
tigators.
Recently, the rotational state-to-state transition rates
were measured for rotational quantum numbers of about
15 to a temperature of 1500 K. An empirical formula was
devised by the investigators to express the measured state-
to-state rate coefficients. By using this formula to higher
temperatures, rotational equilibration collision numbers
were calculated by Park [21]. In Figure 15, the results of
this calculation are shown. As seen, rotational collision
number increases steadily with temperature.
The Landau-Teller type relaxation time value is calcu-
lated for the rotational relaxation phenomenon. In Figure
16, this rotational relaxation time is compared with the
vibrational relaxation time for N2 . As seen, at low tem-
peratures, rotation relaxes faster than vibration. At the
temperature of about 12,000 K the two relaxation times
 11

are nearly the same. At higher temperatures, rotation re- the rate coefficient appropriately, Kim et al have shown
laxes slower than vibration. A three-temperature model that the results of a master equation calculation can be
that accommodates this behavior can explain the shock reproduced by a Landau-Teller type relaxation rate cal-
tube data shown in Figures 7 to 9. culation. In Figure 17, the dissociation phenomenon cal-
Any such three-temperature model has a burden of culated in this way is compared with that calculated by
modeling the rate of energy transfer between the vibra- master equation. As seen, the model can closely repro-
tional and the rotational modes. Presumably, at very high duce the master equation calculation.
temperatures, the energy transfer rate will be so large that
the rotational temperature could be taken to be the same
as the vibrational temperature.
1000

6
5 Present work
4 (Z2, e-folding)
3

2
Prangsma et al (1973) The point where
expeirment vibrational relaxation
Effective collision number Z

100
becomes faster
than rotational
6
5 relaxation
4
3 Parker (1959) as ineterpreted
2
by Wysong and Watdsworth (1998)

Lordi and Mates (1970)

10

6
5
4
3
Carnevale et al (1967) Temkin (1991) as interpreted
2 experiment by Wysong and Wadsworth(1998)

1
Figure 17. Vibrational and rotational tempera-
3 4 5 6 7 8 2 3 4 5 6 7 8 2 3 4 5 6 7 8

tures of H2 behind a shock wave calculated by in-

3 4 5
10 10 10

Figure 15. The number of collisions needed to tegrating the master equation [1].
Translational bath temperature, K

bring rotational temperature to the translational Kim et al’s work on H2 is based on being able to calcu-
temperature [21]. late the state-to-state transition rate coefficient among all
ro-vibrational states. The rate coefficients were calculated
10
-1
using quasi-classical molecular dynamics calculation. So
far, this was done thoroughly only for H2 -H, H2 -He, and
Relaxation times for N2
-2 Rotational (e-folding)
H2 -H2 collisions. In order to carry out such a molecu-
10
Vibrational
-14 1.692

lar dynamics calculation for any given pairs of colliders,

-3 2.47x10 T
10
Relaxation parameter pτ, atm-s

10
-4
Collision-limiting cross section
-17
=10 cm
2 one needs to know the potential energy surfaces involved.
Such potential energy surfaces must first be generated be-
fore the three-temperature model can be implemented us-
-5
10 -16 2
10 cm

10
-6
ing quantum chemistry.
5.2. Radiation Temperature
-7
10
-15 2
10 cm
In the work of Furudate [20], the radiative transfer
-8
10

10
-9 without collision limiting problem was painstakingly in detail. This procedure could
be simplified if we devise a method of solving the radiative
transfer problem for the important emitting/absorbing
-10
10 2 3 4 5 6 7 2 3 4 5 67 2 3 4 5 67 2
3 4 5

spectral features. For this strongly emitting and absorb-

10 10 10
Translational temperature, K

Figure 16. Comparison of calculated rotational ing lines or continua, the so-called Rossland approxima-
relaxation time with vibrational relaxation time tion [22] is valid. One first determines the Rossland mean
for N2 [21]. absorption coefficient for the emission/absorption mecha-
In the work of Kim et al [1], the vibrational and rota-
tional modes were lumped into one mode, and a rate equa-
nism α

1 = 15 x2 1 exp(x) x4 dx
tion was written for this ro-vibrational mode. By choosing α 4π4 x1 κ (exp(x) − 1)2
12

where κ is the absorption coefficient and x = hν/kTR . [6] Furudate, M., Nonaka, S., and Sawada, K.,
Here, TR is the temperature controlling the emission and “Calculation of Shock Shapes over Simple Geometries in
absorption phenomenon, and needs to be found. It satis- Intermediate Hypersonic Air Flow,” AIAA Paper 99-3686,
fies a diffusion equation of the form 1999.
[7] MacClean, M., and Holden, M., “Catalytic Effects
DER 16 σTR3
= ∇[ 3α ∇TR ] + S on Heat Transfer Measurements for Aerothermal Studies
Dt with CO2 ,” AIAA Paper 2006-0182, January 2006.
[8] Doraiswamy, S., Kelley, J. D., and Candler, G. V.,
ER is the energy contained in the excited state of con- “Vibrational Modeling of CO in High Enthalpy Nozzle
cern, σ is Stefan-Boltzmann constant, and S is the exter- Flows,” a paper to be published 2
in Journal of Thermo-
nal source of radiation energy. By solving this equation physics and Heat Transfer.
throughout the flow field, one could obtain TR . This TR
could be recognized as a radiation temperature in control and[9]Equilibrium
Sharma, S. P., and Gillespie, W., “Nonequilibrium
Shock Front Radiation Measurements,”
of ionization rates. Journal of Thermophysics and Heat Transfer, Vol. 5, No.
This radiation temperature could be a third tempera-
ture. This model becomes relevant only at very high flight 3, July-September 1991, pp. 257-265.
[10] Fujita, K., Sato, S., Abe, T., and Ebinuma, Y.,
speeds. For those high speed flights, rotational nonequi- “Experimental Investigation of Air Radiation Behind a
librium may be insignificant. Thus, a three-temperature Strong Shock Wave,”
model recognizing this TR as the third temperature may Transfer, Vol. 16, No.Journal 1,
of Thermophysics and Heat
January-March 2002, pp. 77-82.
be possible. [11] Koreeda, J., “An Expermental Study on Nonequi-
librium Radiation of Strong Shock Waves in Air,” Ph.D
6. Conclusions
Thesis, 1995, Chiba University, Chiba, Japan.
The existing two-temperature model is irrational be- [12] Miyazaki, K., Ando, H.,Takeda, K., Toyoda, K,
cause rational description is impossible. It does not rec- and Maeno, K. (2005), “Establishment of Coherent Anti-
ognize the fact that rotational temperature deviates from Stokes Raman Spectroscopy Measurement System for Ra-
heavy particle translational temperature and cannot de- diative Flow Behind Hyperveolocity Shock Waves. Proc.
scribe the effect of radiation absorption on ionization re- 25th Intern. Symp. on Shock Waves, edited by G. Ja-
actions. But the two-temperature model is acceptable gadeesh, E. Arunan and K. P. J. Reddy, Univerisities
for most applications. Exceptions are 1) Martian and Press(India). 877-882.
Titan entries, 2) radio blackout, and 3) rarefied rocket [13] Dankert, C., “Rarefied Flow Investigation in Gas-
plume flows. Further research is recommended to develop Surface Interacion and Supersonic Plums,” Colloquium
a three-temperature model which recognizes either a sep- on Nonequilibrium Phenomena of Low-Density Jets in
arate rotational temperature or a radiation temperature. Space,” February 1998, Kyoto University, Kyoto, Japan,
pp. 54-60.
References [14] Nompelis, I., and Candler, G. V., “Investigation of
[1] Kim, J. G., Kwon, O. J., and Park, C., “State- Hypersonic Double-Cone Flow Experiments at High En-
to-State Rate Coefficients and Master Equation Study for thalpy in the LENS Facility,” AIAA Paper 2007-203, 2007.
H2 +H2 ,” AIAA Paper 2009-1023; to be published in Jour- [15] Ahn, J. W., Kim, C., and Rho, O. H., “Numerical
nal of Thermophysics and Heat Transfer. Analysis of Hypersonic Flow Over a Double-Cone Geom-
[2] Millikan, R. C., and White, D. R., “Systematics etry Using Eu’s GH Equations,” AIAA Paper 2004-1062.
of Vibrational Relaxation,” Journal of Chemical Physics, [16] Bogdanoff, D. W. and Park, C., “Radiative Inter-
Vol. 39, No. 12, 1963, pp. 3209-3213. action Between Driver and Driven Gases in an Arc-Driven
[3] Marrone, P. V., Treanor, C. E., and Messerschmid, Shock Tube. The J. Shock Waves. Vol. 12, No. 3, 2002,
E. W., “Chemical Relaxation with Preferential Dissoci- pp. 205-214.
ation from Excited Vibrational Levels,” The Physics of [17] Omura, M., and Presley, L. L., “Electron Den-
Fluids, Vol. 6, No. 9, 1963, pp. 1215-1221. sity Measurements Ahead of Shock Waves, in Air,” AIAA
[4] Knab, O., Fruehauf, H.-H., Messeschmid, E. W., Journal, Vol. 7, No. 12, December 1969, pp. 2363-2365.
“Theory and Validation of the Physically Consistent Cou- [18] Park, C., “Effect of Lyman Radiation on Nonequi-
pled Vibration-Chemistry-Vibration Model,” Journal of librium Ionization of Atomic Hydrogen,” AIAA Paper
Thermophysics and Heat Transfer, Vol. 9, No. 2, April- 2004-2277, 2004.
June 1995, pp. 219-226. [19] Park, C., “Nonequilibrium Ionization and Radia-
[5] Macheret, S. O., Fridman, A. A., Adamovich, I. V., tion in Hydrogen-Helium Mixtures,” AIAA Paper 2009-
Rich, J. W., and Treanor, C. E., “Mechanisms of Nonequi- 0814, 2010.
librium Dissociation of Diatomic Molecules,” AIAA Paper [20] Furudate, M., “Nonequilibrium Calculation of
94-1984, June 1994. High-Temperature Radiating H2 -He Flowfield,” Journal
 13

of Thermophysics and Heat Transfer, Vol. 23, No. 4,

October-December 2009, pp. 651-659.
[21] Park, C., “Rotational Relaxation of N2 Behind a
Strong Shock Wave,” Journal of Thermophysics and Heat
Transfer, Vol. 18, No. 4, October-December 2004, pp.
527-533.
[22] Vincenti, W. G., and Kruger, C. H., Jr., Introduc-
tion to Physical Gas Dynamics, Krieger Publishing Co.,
1965, pp. p467.