.

S330 IRE
TR4NSACTIONS ON ANTENNAS AND PROPAGATION December

ElectromagneticScattering by High-DensityMeteor Trails*

H. BRYSKt

Summary-The discussion &-st concerns itself with establishing The phaseerror is then
limits on the validity of the low-density approximation. These de-
pend not only on the electron line densitg (as has sometimes been
loosely stated) but also on the wavelength of observation and on the 4 = csc e L m ( l - ?z)z)dr, (3)
altitude of thetrail. Next, a model is developed for scattering by a
supercritical density distribution of electrons, based on the idea that
the process can still be viewed as a superposition of individual where 0 is the angle between the trail axis and the ray
Compton effects, but with the wave incident on the electron attenu- (under the stipulated conditions, the bending of the ray
ated because of refraction ( inanalogy to the skin effect). Results of can be neglected). From ( 1 ) and ( 2 )
this model are compared with those obtained by the usual approach
of replacing the electron distribution by a metallic scatterer whose tz2(r) = 1 - (4aro/k2a2)exp ( - r z / u 2 ) (4)
surface is the critical density contour. Some calculations with non-
Gaussian electron distributionshelp to clarifythe physical interpreta- and the minimum value ofn is that at r = 0;
tion.
'?Z2(0)= 1 - ( 4 W o / k 2 U 2 ) , + . ' (5)
The expression in (3) can be reasonably ~ $ 1 bounded
I. PHASEERROR
IN INDIVIDUALSCATTERER
MODEL

s
by noting that I'
CATTERING by a meteortrail of low electron
densityisusuallytreatedbyconsideringeach 1 - ?Z = ( 1 - n " / ( l + n). 3
- ..._

9i..
electron to be subject to the incident field and to Since
scatter independently (according to the Thomson cross 'l.:
section),thecontributions from the electronsbeing 1 2 2 401, (7)
added up coherently.' 2 2 1 + I t 2 1 + n(0); (8)
For a moderately higher density (still below the crit-
and
ical density everywhere), with an electron collision fre-
quency well below the frequency of the radiation, the
model is disturbed by the deviation from unity of the [I + n(0)]-1k csc e So*( 1 - d ) d r
index of refraction of the inner region (which alters the
phase relations between electrons). As an indication of
when this effect can become significant, the phase error
242 2-1~ csc e S,
Fo

( 1 - fz2)dr. (9) j
made in neglecting the variation of the index of refrac-
tion is computedfor a ray coming in from infinityto the The integralis simple,
axis of a cylindrical Gaussian distribution. When this
phase error is small, the individual scatterer model is
good; as i t becomes large, the model breaks down.
som(l - d)dr
Fo

The index of refraction fz is givenby = (4aro/K2a2) Jodr exp (-."la2) = 2 ~ ~ / ~ a r ~(10)

/k~a,
I22 = 1 - (47rly7,/k2), (1) so that
where 2n*/~at-ocsc e / k a { 1 + [I - (4aro/k2~2)]1/2j
AT=electron density, El
2 +2 T1/2aro csc e/ka. (11)
70 = e2/mc2=classical electron radius, li
-.
k = 2n/wavelength. Y' I t is convenient to refer t o the characteristic dimen-
sionless constants
The electron distribution in a meteor tra.: is approxi-
mated locally by a cylindrical Gauss,.... B = 2(aro)1/2 (12)
.. 2.::. .
N ( Y ) = (a/na2)exp ( - r 2 'a2). ' (2)
K = ka. (13)
c7 ,* In terms of these, the last equation reads

* The research
reported in this paper i jored bythe
~ /e{K1)
~ - ~ T ' / ~ ( Bcsc + [l - ( B / K ) 2 ] ' / 2 ] - 1
Electronics Research Directorate, AF Car.:
Air Res. and Dev. Command, USAF, und
~ r Center,
:
r c c .IF 19(604)-
2 + >3&- w ~ ( B ~ / Kcsc
) e. (14)
4993.
f Physics Dept., Adelphi College, Garden City, N. Y . Formerly For B<<K, the phase error 4 i -xactly determined. At
at Radiation Laboratory, Universityof Michigan, Ann Arbor, Mich. worst, for B = K (critical den51 -hed onaxis),there
H,. Brysk, "Electromagnetic scattering by low density meteor
trails, J . Geophys. Res., vol. 63, pp. 693-716; December, 1958. is a factor of 2 uncertainty.
 1959 Brysk: Electromagnetic
Scattering by High-Density Metecur Trails S331

The definition of what constitutesa small phase error

is somewhat arbitrary. The simplest reasonable choice
from (14) is
(B2,'KI I 1; (15)
(for 8 = O , this leads to 4525' from ti..: right-hand esti-
mate). At the same time, if the trail is
tu $e of subcritical
?
density everywhere
B 2 R. (16)
Hence, the roughcriterion of applicability of the under- '5
Y

dense model is that

B S K , B < l (1'7)
B2 5 K , B > 1. (18)
IO" -
Fig. 1 exhibits the limiting value of the electronline
density as a function of the wavelength, using for a the
iriitiai widLh computed by Opik2 with densities at four
altitudesobtainedfromWatanabe13 forameteorveloc- 1010 -
ity of 40 I',m/second. This set of curves is to be con-
trasted wrth the frequently madeunqualified statement4
".-- . ransition from the underdense the to over- X (cm)
occurs a t a = electrons/cm. Fig. 1-Maximum electron
line densitywhich
for
independent
scatterer model assumptions are valid.
11. AsTENIJATION FACTOR DUETO PHASE V-4RIA-
CRITICALDENSITYCONTOUR
TION ON The index of refraction is givenby
Consider a plane perpendicular to the trail axis. If nz = ~ [-r2/a2]
1 - ( B / K )exp
there is a critical density region, the intersectionof this
= 1- ( ~ / ~ ) 2 e x
[-(z2+
p Psin2p)/a2]. (22)
plane with the critical density contour will be a circle.
The phase variation of rays reaching thiscircle willnext Since the index of refraction is zero at the critical den-
be investigated. For simplicity, the rayswill be assumed sitycontour,and R a r e related, so that (22) can
parallel. (This is fully justified, comparing the radius of be rewritten more conciselyas
the critical density region x
with the source size and
nz = 1 - exp [( X 2 cos2 p - z*)/a2].(23)
range, although i t would be incorrect over the lengthof
the trail.) To evaluate the integrals approximately, an inequal-
The phase of a ray coming in from infinity and reach- ity like (7) is again resorted to, after expressing
ing the critical density contouris, in general,
n = +
1 - (1 - n2)/(l fz) (24)
Q =K n@)dz, m o - n = (no2 - n2)/(no+ n), (25)
CSC8JW (19)
R 008 B
by replacing thedenominatorsbyestimated mean
where p is the azimuthal angle. values. Thus,
For a ray whose extension would hit the axis of the
trail,
Q - +o = K csc e{ X(1 - COS p) - (1 + n1-1
W
exp f T2 cos2 @/d)2 - 1 7 F a[+( Z / a ) - a(x cos /!?/a)]
Qo = K csc e
k
nodz. (20) + (G. -i-'[exp (Rz/a2) - exp (I? cos2 p / a 2 ) ] 2 - ~ 2 a
11.' .[I - +(Z/U)] 1. (26)
The phase difference between the rays is
I n the a s y q limit for cos P>>a, the last two
4 - +,, = K csc e { J-'
R cos0
lZdz - - .Idz} * (21)
terms in (26)ar;g small to order ( U / R )compared
the firsL . +1.
~
.26)canbereducedapproximatelyto
with

?sc r . .&d
!. = KT csc O(l - cos p). (27)
E. J. h i k , 'Physics of ..r-teor Flight in the Atmosphere," Inter- ; ; ~;;ii
As ( a / " ~ ~ ' ~,-,;;j : a~ ;less tidy situation ensues. F~~a
science Publishers,In:,. New York,X. Y . ;1958.
a K. iatanabe,Ulka\.alet absorptionprocessesin &e upper reasonable h&ation of the effectofphase variation
atmosphere," Admnces in J-.Bbhp., vol. 5, pp. 153-221; 1958. (accepting a possible error in phase of as much as a
4 L. A. Manning a-.' . . rR. Eshleman, UMeteors in the Iono-
sphere," PROC. I R E , VO;. '- -:>.186-199; February, 1959.
_' factor of 2), (27) will beused.
..-

s332 IRE TRANSACTIONS Oh' ANTENNAS AND PROPAGATIOhT December

Within the critical density contour the wave suffers magnetic radiation is sufficiently higher than the elec-
no further phase change. Hence,if the electron distribu- tron collision frequencythat absorption canbe neglected.
tionhascylindricalsymmetry,contributionsfor all The results will be expressed as the ratio of the scat-
values of /3 are equiprobable. Due to these phasediffer- tering amplitude obtained from the overdense region to
ences, there is an attenuation factorf, at a given point that which would be obtained from the same numberof
along the axis, givenfor a two-way traversal by electrons treated as individual scatterers without phase
differences; i e . , in effect, the reduction in effectiveness
j p= (27r)-1
S-: dp exp [ - 2 i k X csc e cos p] of scattering due to the denseness.
In particular, for a cylindrically symmetrical distribu-
= k Ke).
~ ~ ( 2 csc (28) tion of electrons the attenuation factorf'will be givenby
-
For Iarge argument, the asymptotic
fp = (&X csc e)-1'2 COS
expressionyields
[%X csc e - (./4)]. (29)
f' = { S,
R
rduljT(r)

Averaging out theoscillatory factor (;.e., averaging over

time during the expansionof the trail, or simply recog-
nizing that the critical density contour is less than per- where fi denotes the absolute valueof n .
fectly sharp), ( 2 9 ) reduces to For the Gaussian distribution, with the further nota-
f,,= ( 2 d R csc e)-1'2. ( 3 0 ) tion
REGION
IN SUPERCRITICAL
111. ATTENUATION x = r/a, X = R/a, (32)
Scattering by anoverdense region is usually handled the exponent in (31) becomes .. ., . . .
.?.
: :4' :i
by replacing the region by a perfect conductor whose X
surface is defined by the critical density contour. Aside 2 k l F t i ( r ' ) C f = 2 K [ [(B/K)2exp(-P) - 1I1I2dt
from problems of computation of the scattering by the
conductor, there are intrinsic objections to this model. X
An ionizedregion can maintain a transverse current, = Z B J [exp (-PI - exp ( - X * ) ] ~ W . ( 3 3 )
while a conductor cannot. Within a metal, the real and
imaginary parts of the index of refraction are equal in The denominator is simply
magnitude,whereaswithintheoverdense

entirelydifferent).Furthermore,thecriticaldensity
ionizedre-
gion the real part is zero (so t h a t phase properties are
-
RrdrR so
contour is not a perfect reflector. There is penetration
into the overdense region with an attenuation factor, = (a)/.)
X
S,
~ d x e x p(-x21 = ( a / 2 n ) [ 1 - exp(-~2)1. (34)
analogous to the skin effect for a metal. The electron
density usually increases toward the center of the over- Thus, the attenuation factor is
dense region, so that there are the counteracting trends
of the skin effect tending to reduce the scattering from f' = 2[1 - exp (-F)]-I
deep inside per electron while the density distribution
indicates a n increase inthe numberof scattering centers.
xdx exp (-GI
SoX
The approach attempted hereis a mongrel model.
The scattering is considered to consist of three distinct
regimes: the incoming wave, the Compton process it- The indicated integrations cannot be done analytically.
self, and the outgoing wave. The Compton process is Two special cases will be studied below.
treated as an individual particle effect, just as for the
underdensecase. The waves,ontheotherhand,are A . Very High Density
handled from a ray-tracing viewpoint. The amplitudeof
If the electron line density parameter B is much larger
the wave being scattered by the electron is taken to be
thanthe
critical-density-region size parameter K ,
the incident amplitude reduced by the skin-effect at-
(35) simplifies a bit on going to the limit X+ C O .(This
tenuation, the latter being computed along the shortest
limit can be used provided X is greater than about 3 . )
opticalpathfromthecriticaldensitycontourtothe
The integral in the exponentbecomes
electron. The amplitude of the scattered wave is re-
duced by the same factor in coming out of the over-
dense region. I t should be noted that theskin effect rep-
resents a reduction of amplitude due to refraction away
from a region rather than absorption in theregion. The
discussion assumes that the frequency of the electro-
1959 Brysk: Electrmagnetic Scattering bg High-Density
Meteor Trails 5333

so that (35) reduces t o

J = 2 S, ~ d x
m
exp ( - exp [ - ( 2 7 r ) l / 2 ~erfc (x/21/2)I. (37)

In this last form, f' depends only on B ; Le., only on a.

Thus for a sufficiently dense region, the attenuation de-
pends only on the line density of electrons, and not on
the wavelength of the radiation or the widthof the dis-
tribution. These quantities do enter(in the combination
k a ) in thedetermination of whatconstitutes a "suf-
ficiently dense" region. Fig. 2 exhibitsf', from (37), as a
function of a ;on a log-log plot, the curve is very nearly
a straight line.
The scattering amplitude of the meteor trail is pro-
portional to af' and hence to B Y . In Fig. 3, BY' has
been plotted againstB. The curveis fitted by
BY = 0.91 In B (38)
Fig. 2-Attenuation factor for very high density.
apart from the small-B end[where (37) is nota good ap-
proximation anyhow].
For an intuitive grasp of the characteristics of the
scattering, it is instructivetocompare (38) withthe
scattering amplitude obtained for two simpler electron
distributions-a uniform cylinder and two coaxial uni-
form cylinders-with the same line density and in the
high density limit.
For the uniform cylinder, N is a constant, hence +Z
is also a constant, and( 3 1 ) reduces to

For a largeelectrondensity,the second term in ( 1 ) I-

predominates, so that / '-31 In6
/
0' I I I

The electron line densitya is IO0 IO' IO2 lo3

B
Fig. 3-Scattering amplitude for very high density.

so that kfizreduces to
~ / ~B.
kBX = 2 ( ~ r r O ) = (42)
For large B , the attenuation factoris then simply
f' = B-1 (43)
and the scattering amplitude
is
B y = B. (44)
For two coaxial uniform cylinders, the procedure is
analogous except that ther-integral is split into two por-
tions (0 to X' and X' to E) for each of which N a n d t? are
.~ ~.
s334 IRE TRANXACTIONX ON ANTEXNAS
AND PROPAGATION Uecember

In the limit of large density and extent for both regions regioncase, and as its logarithm(still slower)in the
[;.e., for (40) valid and the negative exponentials on the Gaussian case.
right-hand side of (45) all small], there results The implication of these observations is that, for a
region of radially decreasing high electron density, the
BY = (4nro.~72) ll2T?. (46)
scattering characteristics are predominantly determined
Note that by the outer portionsof the supercritical densityregion

a = 2*[ N l s,
i-1

+
rdr x 2 s,, rdr]
-
R (the core not being sufficiently penetrated by the radia-
tion). The result from increasing the line density of elec-
trons is primarily to push the effective scattering region
= 7r[YlX’2 + l Y 2 ( X 2 - RZ)] outward, rather than to increase its density. This does
+
= nhTzX2 n(M1 - LY2)i?‘2. (4T) result in an increase in the scattering because the sur-
face area is increased (hence more electrons are acces-
If the inner region has thehigher electron density, there sible).
is thus a lower f’ (for a given a) than in the uniform
cylinder case. B. Critical Demity Contour of Xaximunz Width
For definiteness, the coaxialcylindercase will be T h e electron distribution spreads out in time due to
specialized by making the proviso that each of the two diffusion, and a increases. From (4) i t can be deduced
regions contain the number of electrons that would be that the radiusof the critical density region is given by
present in the corresponding part of the Gaussian. Then -
R2 = u2 In (4crro/k2a2). (55)
Jx:rdrN = N z ( X 2 - p2)/2 The maximum value that i t can attain (as a function of
= (cr/27r) [exp (- X’2/u2) - exp (- X Z / U ~ ) ] (48)
. a-i.e., of time) is given by
a P / a a 2 = In (4aro/k2a2) - 1 = 0 (56)
I t was shown abovethat
exp (- X2/u2) = (R/B)2. (49) which means that
x = 1. (5 7)
For convenience, write
Heuristically,thewidestcriticaldensitycontour
pa = pX2 (504 might be expected to yield the largest scattering return
where from the supercritical region, because i t corresponds to
0<p<l, (job) the greatest number of electrons being exposed t o a n
unattenuatedincident field. This argument would be
so that much weakened if i t should turn out thatf’(viewed as a
exp (- X‘2/u2) = ( R / B ) 2 ~ . (51) function of X) has resonance-type oscillations.
Ideally, the maximum value off’ should be obtained
The value of N2 can then be expressed simply by by setting df’/aX= 0, where f’ is given by (35). Un-
( n / ~ ~ ) N 2 (-
l p ) X 2 = ( R / B ) 2-~ ( K / B ) 2 (52) fortunately, this is impractical becauseaf’/dX includes
a term inf’ and also a term involving a n integral likef’
so that (46) becomes with an additional integral asa factor in the integrand,
B y = B(l - p>-‘/”(R/Bp - (R/B)2]’/2. (53) as well as a term independent of the f’-integral; the
equation af’/dX = 0 cannot be solved unless both f’ and
As the discussion applies to high densities (B>>R),only the somewhat more complicated companion integral are
the first term in the bracket need be retained since p < 1 known as a function of X.Hence, there is no direct way
(withsomecare that p notbetoo close to 1). Thus of determining the maximum value off’ short of actu-
B2f‘ = (1 - p)-l/ZKPBl-P. (54) ally computing f’ as a function of X.Since this must
then be repeated for each B of interest, the computa-
Comparing now the high-density limits of the three tional effort required is quite large.
cases, we find that: In what follows, i t will be assumed that the maximum
I) for a given line density, the scattering amplitude value off’ is indeed attained for X = 1. ThisIis also the
is less for the two-region case (with inner region more assumption in the metallic scatterer approximation, so
dense) than for the uniform cylinder case, and still less there will be a direct comparison of results for the same
for the Gaussian; configuration. The resultant special case of (35) is
2) the scattering amplitude for thetwo-region case is
1
in fact, according to (46), just that which would occur
f = 2 [ 1 - e-11-1J d x exp ( - x 2 1
d
if both regions had the density of the outer one; 0
k
e 3) the scattering amplitude varies as the square root
of the linedensityintheuniform regioncase, as a
smaller positive power of the line density in the two- t.
.exp (--ZBJol[exp ( - t 2 ) - e - l ] l W t (58)
1959 Brysk: Electromagnetic
Scattering bg High-Dewiiy Neteor Trails s335

This is plotted in Fig. 4. The curve is well fitted by

f’ = 1.05B-.”. (59)
So far, the “skin effect” attenuation and the phase re-
lations around the critical density contour have been
considered. There remains to examine the phase change
along the trail axis due to the deviation of the index of
refraction from unity.This consists of two contributions:
the phase error in reaching the critical density contour
(evaluated in Section I), plus the error incurred in in-
cluding the region inside the critical density region in
the optical path
4’ = ki? csc 8. (60)
During the growing phase (case A), the contribution of
(14)dominates over thatof (60)and leads to oscillations I I

superposed on the scattering integrand. Near the maxi- IO0 IO’ IO2
0
mum expansion of the critical density contour (case B),
on the other hand, the two contributions are comparable Fig. &Attenuation factor for critical density
contour of maximum extent.
and vary in opposite directions, so that the longitudinal
phase variation is very slow.
Accordingly, for the maximum critical density con-
tour the attenuation factor, instead of the exponential
of Bryskl is approximately the product of the phase
reduction factor of (30) and the “skin effect” attenua-
tion factor of ( 5 9 ) . In (30),the relations for this maxi-
mum lead to
kB = ka = K = e-1/2B (61)
so that
* B19)-~/~1.05B--.~~
f = ( 2 ~ e - ~ /csc = .52 sin112 (62)
The maximum return is now given by
S = R’) (63)
.076 CSCBPGG‘X~(~~~O).~~(~.~‘)~/~~~~RR‘(R +
or monostatically
S = .038PGG’X3(~ro).81(8.&‘)2/16~2R3.
(64)
This result is to be compared with that obtained by
assuming the critical density region to scatter the elec- a (cm-I)
tromagnetic radiation likea metallic cylinder, the return
Fig. 5-Comparison of “skin effect:’ model with “metallic
from the latter being computed by geometrical optics, cylinder” model.
;.e., in the limit i?>>L6 The correct geometrical “cross
section” for broadside backscattering from a metallic
model. The Lovell-Clegg low-density result is also shown
cylinder is
for orientation. For clarity of display exclusively, the
u = nXR. (65)
overdense and underdense trail curves have beenex-
This is a factor of two less than that quoted by Green- tended to meet. The temptation to bridge the transition
how,j and it leads to by fairinginfromonecurve to the other should be
resisted; the intermediateregion very probably does not
S = .048PGG’X3(~ro)1/2((8.6‘)2/16n2.R3. (66)
behavethatsimply.Thetwooverdensetrailcurves
Fig. 5 compares the maximum return from an over- cross for relatively low 4-7 X 1013 cm-I). On the low-a!
densetrailasobtainedbythepresent “skin effect” side, theyyield essentially undistinguishable predictions
model with t h a t obtainedbythe“metalliccylinder” (especially taking into account the pile-up of theoretical
errors in the transition region). On the high-a! side, the
6 J. S. Greenhow, “Characteristics of radio echoes from meteor
“skin effect” model yields an increasingly larger result,
Proc.
trails: 111, the behaviourof the electron trails after formation,”
23.
P h s . Sod.. vol. 65. DD. 169-181; March, 1952. as expected intuitively.
S336 IRE TRAN8ACTIONS ON ANTENNAS AATD PROPAGATION Decembtv

A feature deservingof comment is the specular nature ( L e . , before enough time has elapsed for some form of
of the return. In this respect, the present model leads turbulence to be invoked), it would appear that the
to the same variation as the underdensetrail model. basic assumption of a uniform line density of ionization
T h e metallic cylinder model, of course, s h o w a strong underlying both models must be abandoned; and that,
aspect dependence.On the other hand,a loss of specular- although a uniformlinedensitycan be satisfactorily
ityastheelectronlinedensity increases has been assumed for underdense trails, it is an essential feature
reported experimentally. In so far as thereis such a loss of scattering by overdense trails that theline density is
of specularity in the relatively early history of the trail markedly nonuniform.