Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

INSTRWENTS

&METHoos
IN PHVSICS

Nuclear Instruments and Methods in Physics Research A 399 (1997) 285-300

Penetration of cosmic ray muons into the Earth


Jack L. Uretsky”
Argonne National Laboratory, 9700 S. Cass Avenue, Argonne, IL 604392 USA

Received 18 February 1997; received in revised form 18 July 1997

Abstract
I present a new analytic solution to the integro-differentialequation that describes the underground propagation of cosmic
ray muons. The exact solution is given in the form of an infinite series in inverse powers of the muon energy. Convergence
is proved for sufficiently high energies.
The series is shown to be summable in closed form, in certain approximations. The closed forms provides analytic
continuations to low energies of the series solution. One approximation resembles a well-known solution that ignores discrete
energy loss, but this approximation introduces additional constants. I apply the approximate solution using an expression for
the surface muon flux, derived from the primary flux, as a boundary condition. The result predicts the underground muon
vertical intensity over seven orders of magnitude (10 km depth), compares favorably with published Monte Carlo calculations,
and can be performed in seconds on a personal computer.
As an application, the same approximation predicts the “catastrophic” energy-loss event rate at Soudan II.

1. Introduction Earth [4]. Hayakawa et al. [4] credit Greisen [5] with
the observation that muons constituted the penetrating
Physicists were attempting to describe the penetra- component. Nishimura [6] applied shower theory to
tion of cosmic rays through matter before muons were the problem of muon penetration into the earth and
known to exist and before relativistic quantum electro- exhibited a formal solution to the resulting diffusion
dynamics had been formulated. Cosmic ray physicists equation.
were originally interested in shower development; The dill&ion equation is easily solved in the ap-
Rossi and Greisen’s review article [l] credits Carl- proximation that hard collision losses are neglected.
son and Oppenheimer [2] and Bhabha and Heitler [3] The Nishimura solution, which includes the effect
with the seminal contributions to the subject. These of the hard collisions, gives the integrated intensity,
early investigations led to the description of cosmic but not the spectrum. The solution also appears to
ray penetration into matter as a diffision process. be numerically tractable only for a power law spec-
Underground observations during World War II trum with integral spectral index, an inconvenient
showed that cosmic rays penetrate deeply into the limitation.
Gurentsov et al. [7] have attempted to include
*Tel.: +l 630 252 8570; fax: +l 630 252 5047; e-mail: the effect of collision losses through an ansatz that
jlu@hep.anl.gov. makes a simple modification to the no-collision-loss

0168-9002/971$17.00 0 1997 Elsevier Science B.V. All rights reserved


PIZSO168-9002(97)00996-O
286 J.L. Uretskyl Nucl. Instr. and h4eth. in Phys. Rex A 399 (1997) 285-300

solution. It is not clear, however, to what extent with


the modified solution is in fact a solution of

I
I
the diffusion equation with hard collision losses XP
Hip = dxp.
included. B-’ x + yiE
The conventional modern-day approach to the
The symbols in the equation have the following
diffusion problem is the Monte Carlo calculation,
meanings:
first applied to the study of underground muons
i stands for either rc or K
by Bollinger [8]. See Ref. [9] for a more recent
C,(E/EO)-P estimated primary flux, with C,
discussion.
= 1.6 f 0.2 cm-’ s-’ sr-t GeV-’
The purpose of the present work is to present an
analytic solution to the l-dimensional diffusion prob- Eo 1 GeV
P primary index, usually taken to be
lem, including the effect of the hard collisions. The
about 2.7
results provide a simple, inexpensive way to calculate
= 1 - (Mp/A4i)’
the spectrum of cosmic ray muons propagating under- Zi the mass of the indicated particle
ground down to depths of 10 km (water-equivalent).
Bi branching ratio of the indicated
The calculations may be done handily on a personal
particle to muons
computer, and the computational time is measured
gni and gnn respective fractions of mesons
in seconds. The method is currently in use for cal-
and nucleons per collision
culating background for searching for high energy
of a high energy nucleon
neutrinos that may be emitted by active galactic
with a nucleus, assuming scaling
nuclei [lo].
Pi = (MiI”p )*
The next section briefly reviews the derivation of
Yi =&co4
the surface muon spectrum from the primary cos-
zi lifetime of meson i
mic ray flux. Section 3 states the diffision problem
ho scale height of the upper atmosphere
to be solved, and recalls the well-known solution
(6.43 km)
that ignores discrete collisions. Section 4 gives an
C as usual, the speed of light
exact solution in the form of an infinite series. Sec-
I3 the zenith angle of the incident
tion 5 presents a proof of convergence of the series.
primary; for zenith angles very close
Section 6 describes two approximations to the exact
to 90” this will be appreciably
result, one for high energies and the other for the
different from the zenith angle of a
entire energy range. Section 7 compares the latter
muon at or below the surface
approximation with Monte Carlo results and with
The scaling parameters have the following values
experiment. Section 8 presents, as an application,
(see [ll]):
an estimate of the frequency of large energy-loss
events from cosmic-ray muons at Soudan II. The
gnn = 0.146, g,,k = 0.0162, g,,,, = 0.203.
conclusions and suggestions for further work are in
Section 9. Dar makes the important observation that there is a
ridiculously simple, and quite accurate, approximation
for the integrals Hip. First observe that for values of a
2. Surface muon flux less than unity

J J (
I
z* 1 z* * z*
The differential muon flux dF(E)/dE at the
dz -= dz --a~. (3)
surface of the earth may be deduced from the (I z+b 0 z+b z + bJa >
cosmic-ray primary flux following, e.g. Dar [l l] or
Gaisser [ 121: Dar’s approximation is

J
I
dW ) Z* 1
- = Cp(E/Eo)-* C CriBixHip (1) dz (4)
dE i 1 - Snn 0 zfb= p+(p+ 1)b’
J. L. Uretsky/ Nucl. Instr. and Meth. in Phys. Rex A 399 (1997) 285-300 287

Surface Vertical Muon Spectrum vs. Muon Energy


Predicted and Experimental Differential Flux times E *I

100 E(GeV)

Fig. 1. Predicted and experimental vertical muon spectrum (multiplied by E2.‘) at the surface of the Earth vs. muon energy. only data taken
near the Earth’s surface are included in this comparison. Data taken deep underground are included in Fig. 4. The data are from Rastin
[23] (filled circles), Nandi and Simha [24] (filled squares) and Allhofer et al. [25] (open squares). The curves show the predictions for
the three indicated spectral indices. The flux falls off at low energies because of muon decay, an effect not included in the present theory.

The approximation is obviously exact for b either 3. Underground muons - general


zero or much larger than unity. Direct evalu-
ation by numerical integration reveals that for a Rossi’s [13] discussion of cascade showers is a
p-value of 2.7 th e approximation differs from the starting point for a discussion of muon propaga-
integral by less than 3% for any positive value tion through the ground. The object of discussion
ofb. is the differential spectrum f&E) dE of muons
Evaluation of Eq. (1) is now easy. Fig. 1 shows a with energy E at depth t underground. Define [14]
comparison with data, of the vertical muon spectrum &E, v) du dt to be the probability density for a muon
at the surface predicted from Eq. (1). ’ Also illus- of energy E to lose a fraction between v and v + dv
trated in Fig. 1 is the effect of making small changes of its energy in traversing an increment dt of ground.
in the primary index from the value 2.70. An index in The probability C$is the sum of the four probabilities
the range 2.70-2.75 accords well with the data above
4b, ~~011~&air, 4ion which, respectively, correspond
a few hundred GeV. to the four modes of energy loss, bremsstrahlung,
nuclear collisions, pair production and ionization.
’ Gaissers’s expression for the vertical muon fhtx (p. 70) is
The ionization loss is continuous. The contributions
formally different from my Eq. (1) evaluated by Dar’s trick, but
the predictions are not distinguishable from those in Gaisser’s log of the other loss mechanisms are shown in Fig. 2,
plot, Fig. 6.1. where it may be seen that hard bremsstrahhmg is the
288 J.L. Uretsky/ Nucl. Instr. and Meth. in Phys. Res. A 399 (1997) 285-300

with a and b taken as constants despite their logarith-


mic energy dependence. This is, in fact, not a bad as-
sumption because the high energy end of the muon
spectrum is strongly suppressed. Evaluating the log-
arithms at an energy of 1 TeV, the variation of the
logarithmic terms over the important energy range,
about 100 GeV to tens of TeV, is small. The error
may be expected to influence mostly the distribution
of low energy muons which are not of interest in many
applications.
Eq. (5) then becomes

af(xA
~ -@+bE)v
ax

Fig. 2. Energy loss probability versus fractional loss of energy by where &E,u) includes only “discrete” losses from
energetic muons. The Figure is reproduced from Adair and Kasha hard collisions. If the hard collisions are neglected,
[14]. The curves show the probability of fractional energy loss u then Eq. (7) becomes
from bremsstrahhmg (&,). pair production (~$a) and nuclear colli-
sions (+a). The product 4(u) du gives the probability of fractional
afb9
~ -(a+bE)~xO.
energy loss between 11and v + dr.
ax (8)

The solution of Eq. (8) may be written down by


dominant loss mechanisms for large fractional energy
inspection once the spectrum at the surface f(0, E) is
losses.
given:
The dilksion equation for the muons passing
through the earth is [ 131 2

where f (0, E) = dF(E)/dE. A solution of the equation


that omits the hard-collision term will be referred to
as “the soft” solution.

The assumption that muon energy losses may be di-


vided into “continuous” and “discrete” losses then per- 4. Including hard collisions
mits me to rewrite the let? hand side of Eq. (5) so that
The integro-differential equation, Eq. (7), is some-
-=-
df af afdE what simplified by the assumption of high-energy scal-
ax+dxdE. (6)
dx ing for the cross-sections involved. Then 4 becomes
The dE/dx term in the last equation represents the a function only of u. The resulting equation is then or-
continuous energy change of a muon with depth. Tra- dinarily Mellin transformed with respect to the energy
dition [ 141 dictates that we write dE/dx = -(a + bE) E [ 151, but this procedure is overkill in the present
instance, as the next section will demonstrate. Merely
2 My discussion differs slightly from that in Rossi; note the total expand Eq. (1) in inverse powers of the energy and
derivative on the left. then let the coefficient of each power be a function
J.L. Uretsky /Nucl. Instr. und Meth. in Phys. Rex A 399 (1997) 285-300 289

J
a,(x) of x that is to be determined. Letting E = E/Eo, Tk-I

then, rewrite Eq. (1) in an obvious notation: dzk exp{[b - n(p + I - k)


0

f(% s) = c,c NXx, s) (10) +qp + 1+ 1 - k)lu}]


. (1%
with the subscript indicating the parent meson, 7c or
K, and This form of solution, which appears to be new, ex-
hibits the Q(X) as generated from unity by a random
walk of up to 1 steps, denoted by the k-fold integrals
fi(x,s)=yil s’ Zp dZE (-Z/vi)‘af(x)&-‘-P-‘.
0
(see, generally [ 16]), times a decaying exponential.
8-l
The soft solution is readily obtained from Eq.
(11)
(15) by setting all the I’s equal to zero. The k-fold
In what follows, I shall drop the superscript “i”, on multiple integral is then just (ebx - l)k/k!bk. The
the ~(1. soft solution follows after inserting the resulting ~(1
We may think of this expansion in inverse powers of into Eq. (1 1), inverting the order of k and I sum-
energy as a sort of “mode” expansion with Q(X) giving mations and introducing a new summation index
the intensity of the lth mode at depth t. Scattering defined by 1 =j + k (with j ranging from 0 to oe).
events will have the effect of decreasing the intensity The k-sum is then recognizable as a binomial ex-
of lower modes (lower values of I) and increasing the pansion, and the j-sum is just an expansion like that
intensity of higher modes. Muons are thus scattered of l/(1 -tz).
from lower to higher modes in this expansion. The
expansion must give Eq. ( 1) at the surface (x = 0), so
the ~1’s must all satisfy: 5. Convergence and such niceties
@1(O)= 1, 1=0,1,... . (12)
It is not hard to show that the expansion in in-
Substitution of the expansion, Eq. (11) into Eq. (7) verse powers of energy, Eq. (1 1 ), with the coefficients,
then yields a set of coupled first order differential equa- Eq. ( 15), converges. The demonstration applies to any
tions for the al’s: surface spectrum, such as that given by Eq. ( 1 ), which
is expandable in a convergent series of inverse pow-
$ + [(I + p + l)b + n(p + 1)]cq(x) ers of the energy. I will, however, simplify the dis-
cussion by taking the surface spectrum as given by
= (p + I) (F) U/-l(X) a simple power law, s-p. Extension of the proof to
(13)
more general cases is straightforward (see Appendix),

I
with the definitions although one must pay attention to the radius of con-
vergence (about the point at infinite energy) of the

J(Y) =
J 0
dt+(o)[l - (1 - a)“] (14)
expansion.
The differential
given by
spectrum is now assumed to be
and a-i(x) - 0. The solution of Eq. ( 13) that satisfies
the initial condition is
f(x, E) = E Q&-P-C (16)
Q(X) = exp{-[(p + 1 + I)b + n(p + Z)]x} I=0

The tx coefficients must satisfy

[J
x
dar
X dri exp{ [b - A( p + 1 - 1) -jy + [(I + P)b + 4p + 1- l)]ar(x) =
0
-a(p + I - l)q_l(X) (17)
+qp + 1)]2,}. ..
290 J.L. Uretskyl Nucl. Ins@. and Meth. in Phys. Rex A 399 (1997) 285-300

with the conditions ~(0) = 1; al(O) = 0, I > 1. The the spectrum where the inverse energy expansion con-
solution is verges well. When the low energy end of the spectrum
is also important, as when one wants to integrate over
‘q@) = (_& + 1- l)! the spectrum to obtain a total rate, a different strategy
(P - l>! is called for. In that case one needs to sum the in-
x exp{ -[(p + I)b + l(p + I - 1)1x} verse energy series approximately to obtain a closed

[I
x
expression that can be analytically continued to low
X dTle[b--6(P+~-2)h . .. energies. The soft solution is one such expression, but
0
the proof of convergence suggests that better approx-
(18) imations are available that take into account some of
the effects of discrete energy losses. The next section
where contains the derivation of one such approximation.

6(z) z J(z) - L(z + 1). (19)


6. Approximate inclusion of hard Bremsstrahlungs
It is clear from the definition that 6(z) is always
negative and decreases in magnitude as l/z with Adair and Kasha [ 141 invoke scaling to approximate
increasing z. the bremsstrahhmg energy loss probability for a TeV
One sees that the inverse energy expansion con- muon passing through standard rock by
verges by use of the comparison theorem [ 171. To this
end, first bound the ]LY~]‘sby noting that the multiple @b(f)= $ [I - ;(I - f) + (1 - n2] . (22)
integrals in Eq. (18) are bounded from above if each
6 is replaced by 6( p - 1). The multiple integrals may Then
then be evaluated to obtain

(
JarJ5
p+l-1
(b-BPp-i))’
1’
I >
1
x exp{-[(p + Z)b + l(p + I - 1)1x}
+5 - 113
(23)
6 p+l+l -p+Z+2
X(elb--G(P-i)lx _ l,‘, (20) where $(z) is the logarithmic derivative of the Gamma
where (i) is the binomial coefficient. Denoting the function and y is Euler’s constant (~0.577). Asymp-
upper bounds by /?I, the ratio of successive terms of totically,
the series BIE -I is , for large 1, less than unity provided
L(p-+Z)-in(l). (24)
that
a -&p-1)x Also, the relation [ 181
E’b-6(p- l)e .
(21)
ti(z + 1) = tie> + ;,
The conclusion is that for energies satisfying Eq. (2 1)
the series fila-’ is absolutely convergent and, by the leads to the useful equation
comparison theorem, so is the series in Eq. (16). The
bound in the inequality (2 1) can probably be substan- - 6(z) = n(z + 1) - L(z)
tially improved.
Recovery of the soft solution from Eq. (15), dis- =pb[&+&s]. (25)
cussed in the preceding section, suggests a system-
atic method for improving the soft solution, namely, When p = 2.7, l(p) = 1.795& and i’s for other
replacing only the first few modes by their exact ex- relevant arguments may be quickly calculated
pressions. This procedure makes mathematical sense from Eq. (25). It is also useful to note that the
when one is interested only in the high energy end of bremsstrahlung contribution to b is just /$,.
J.L. betsky/ Nucl. Instr. and Meth. in Phys. Res. A 399 (1997) 285-300 291

Now return to Eq. (15) and rewrite the multiple 7. Comparison with Monte Carlo calculations and
integral at the end of that equation as experimental data

[S
x

0
dz, e[b--6(P+l- 1)I71. . .
s Tk-I

0
dzk eV--s(p+l--k)ln

(26)
7.1. Monte Carlo - intensity us depth

Lipari and Stanev [9] have published the results of a


Monte Carlo calculation of the penetration of a power
The 6’s are slowly varying functions of their argu- spectrum of muons into standard earth. These authors
ment, so it is not a bad approximation to a replace parameterize their results 3 by noting that in the soft
each 6 by some average value 2. The expression (26) approximation the surface energy ~0 of a muon with
then becomes energy E(X) at depth x is just (in the present notation)
(b _ s)-k.$,(b--*_ 1 )k. (27) ebx E(X) + %(1 - e-“)I .
EO = (31)
[
Eq. (15) becomes A power law surface spectrum

dN
- =KcOp (32)
d%
then becomes at depth x (omitting the argument of E)

(28) dhf dN d.s,,


F&=dEgxdF
A further simplification is suggested by writ-
ing the argument of the first exponential on the = Ke-(P- 1)bx [E(X)
+ %(1 - eebX)] -‘. (33)
right of Eq. (28) as (p+ 1+ l)[b + {A(p + l)/
( p + 1 + 1 )}]. The second term in brackets is a slowly The authors parameterize their Monte-Carlo results
decreasing function of I that can also be replaced by by retaining the functional form of Eq. (33) but treat-
an average value 1. These modifications permit me ing a and b as “simple parameters without a direct
to sum the inverse energy series to obtain a result physical meaning”. The authors present results for
that is reminiscent of the soft approximation, namely 3 values of p: 3.7, 2.7 and 2.0. Their value for a
(dropping, for the moment, the “i” subscript) is fairly stable, about 0.25 TeV/km (km means km
water-equivalent). The b parameter varies from about
e--p(b+hx 1
ZP 0.383 km-’ ( p = 3.7) to 0.465 km-‘.
f(x7E)= [E + y(x)]P .Ip 2 + Y[E+ y(x)]e(b+l@' To compare the Monte Carlo calculation with the
present work, first observe that the power-law spec-
(29)
trum result is quickly derived from the proof of con-
where vergence in Section 5. The spectrum at depth t then
(b-&,e(b-6)x
follows from:
y(x)= A-e - 11.
b-6 dlv
_ = Ke-(P-l)(b+b
The arguments of the 6’s range from p to co; cor- ds
responding 6 values range from -0.303/.$, to zero. -P
a
-_,-(b+nh(,(b-% _ 1) , (34)
A value of -0.1 ~/IF, is a reasonable choice for 3. The b-6 I
choice can be justified by demonstrating that the final
results are insensitive to the exact value. where one should recognize that d&c/d&is just ecb+‘b.
1 obtain an estimate for 1 by averaging the first 10 Integrating Eq. (34) over E gives a result that is of
values of A/(p + I+ 1). The result is 0.41/&. Justifi-
cation for this choice again follows from a demonstra- 3 I am grateful to Professor Stanev for comments related to
tion of insensitivity to the precise value. Eq. (17) of Lipari and Stanev [9].
292 J.L. Uretsky/ Nucl. Instr. and Meth. in Phys. Res. A 399 (1997) 285-300

precisely the same form as Eq. (33) after the same in- Eq. (34), the result of the present analysis, does not
tegration over E. The dependence upon b + 1 cancels, predict such a “hardening” at all depths. The mean
and the combination b - 3 replaces the b in Eq. (33). energy increases with depth down to a depth
A numerical comparison between the two results
follows from taking b to have its customary value
[19] of 0.36kn-’ (the parameter a, which does not x*=&ln(l+$$). (36)
affect the t-dependence of the integrated spectrum is
about 0.26TeV/km at 1 TeV). One finds a value of
The values for b and bt, of the preceding subsection
0.192 km-’ for /$, in Adair and Kasha [ 141. The result
then predicts a “crossover depth” x* for hardening of
is that b - 2 ranges from 0.389 to 0.4 10 as the spectral
about 5.5 km. The prediction that there is any such
index ranges from 3.7 to 2.0. The difference between
crossover depth is in clear disagreement with the
the intensities calculated from the two sets of parame-
Monte Carlo parameterization, Eq. (33), if not with
ters are within a couple of per cent over a range of5 km
the Monte Carlo results.
for an index of 3.7. The difference is less than 30%
Eq. (34) gives mean energies of 228 GeV at 3 km
over the same range at the (rather unphysical) index
depth, and 2 11 GeV at 10 km depth. The 3 km predic-
of 2.0.
tion agrees well with the reported Monte Carlo result
I conclude that the intensity-depth relationship
of 225 GeV. The 10 km predicted energy contrasts
for an integrated muon spectrum is essentially
with the reported Monte Carlo result of 325 GeV.
the same in the present approach and the Monte
Fig. 3(a) shows the two distributions at 1Okm depth.
Carlo approach. Neither one is very different from
Each distribution is normalized to have a value of
the soft approximation (b = 0.36 km-’ compared
unity when integrated over all energies and is, ac-
with b - 2 = 0.38 km-’ ), qualitatively or quan-
cordingly, a probability distribution. It is seen that
titatively, despite the authors’ of [9] contention
the distributions are, in fact, very close except for
that their parameters are “without a direct physical
energies less than about 10 GeV. Such low energy
meaning”. muons have a range of only of the order of a me-
ter in standard rock and are not of interest in the
present study (But see Ref. [14], p. 338, et seq.). I
7.2. Energy distributions
conclude that the Monte Carlo calculations and those
of this work give quite similar energy distributions
The normalized energy distribution from the Monte for muons of more than a few GeV down to at least
Carlo power spectrum (index 3.7) calculations are 10 km depth.
shown in Fig. 4 of Ref. [9]. These authors report mean The normalized energy distribution at 3 km depth
energies (E,) of 225 and 325 GeV at respective depths is plotted in Fig. 3, for comparison with the Monte
of 3 and 1Okm. Carlo generated distributions in Ref. [9]. The present
The mean energy for the spectrum of Eq. (34) is prediction of the 1Okm normalized distribution
just is nearly identical to the 3 km distribution (with-
in 2%).
-(b+Tqe(b-& _
(35)
1)
W = (p-2;cb&?je
7.3. Comparison with data
with a similar expression for the Monte Carlo pa-
rameterization Eq. (33). The parameterization for the I have shown that the present analytical approach
Monte Carlo expressions gives respective mean en- to muon penetration into the earth and Monte Carlo
ergies of 248 and 356GeV at 3 and 1Okm depths. calculations give essentially identical predictions of
These values are somewhat higher than the reported vertical intensity-depth variation for a muon spectral
means, but the parameterization implies that there is index 3.7. My analysis goes further, however, be-
monotic increase of mean energy with depth over its cause it predicts the normalization of the vertical in-
entire range of validity. tensity on the basis of the known primary cosmic ray
J.L. Vretsky/Nucl. Instr. and Meth. in Phys. Res. A 399 (1997) 285-300 293

i-

I-

Fig. 3. Normalized energy distribution of muons at a depth of 3km for a power spectrum with index 3.7 as predicted by the present
analysis. The 1Okm normalized distribution differs by less than 2% at any energy and would be indistinguishable at the scale shown. (a)
Comparison of the energy distributions of muons at 1Okm depth as predicted from the Monte Carlo calculation reported by Lipari and
Stanev [9] and from the present work. Each distribution is normalized to unity when integrated over all energies. Note that the vertical
scale is linear, in contrast to that in Fig. 3.

flux, scaling assumptions, and some pion and kaon de- -fi,lp[p + (p + l)~C~ie-‘b+‘N&~]-‘}
cay properties. The results are contained in Eqs. (I),
cme2 s-l sr-’ GeV-’ (37)
(2), (4), and (29). Putting these together and remem-
bering the factor d&g/d& gives a result that may be with ~1 defined by
written as
dF(x, E) = 0.67e(‘-P)(h+‘)re;P
ds
The constants K, and KK are found to be K, = 1 and
x c Ki{[p + (p + l)yie-(b+“)XE1]-’
KK N 0.038. The index p is the index that appears
294 XL. Uretsky 1 Nucl. Instr. and Metk. in Pkyx Rex A 399 (1997) 285-300

0.a

Fig. 3. Continued

in Eq. (1 ), it is that of the primary spectrum. The Note that the dependence on the 2 parameter in
integrated flux is Eq. (39) cancels, just as it does in Eq. (33).
Eq. (39) is compared with experimental data in

1
1-P Fig. 4. Also shown are data points from a compila-
0.67 ,Cb-b _ 1)
F(x) = --k( tion by Crouch [20], a fit by the LVD Collaborators
P(P- 1) i (b-6)
[Z l] to their own data taken at a depth of 3 km. The
-1
l+P+l Q yi(e(b-zb - 1) Monte Carlo results for a pure power-law spectrum
xC&
i ii Pm I 1 with spectral index 3.7, normalized to the prediction
-1 of Eq. (39) at a depth of 3 km we, would not be dis-
-p,?J 1+Z-?JA biyj(e(bJZr - 1) tinguishable fi-om the illustrated Eq. (39) result. The
(i p - 1 (b - 3) I I Eq. (39) normalization predicts a flux that is about
cme2 s- ’ sr-’ . (39) 50% greater than that of the LVD parameterization.
J.L. Uretsky / Nucl. Ins~. and Meth. in Phys. Res. A 399 (1997) 285-300 295

Fig. 4. Comparison of muon intensity vs. depth as predicted by Eq. (39) with data points from the Crouch [20] compilation, a fit to data
by the LVD Collaboration [21]. A Monte Carlo pure power-law spectrum with spectral index 3.7, normalized to the prediction of Eq. (39)
at a depth of 3 km we would be indistinguishable from the illustrated prediction.

The data points at depths greater than 9 km in the 0.26 TeV/km and 0.39 km-‘. The parameter 2 is about
Crouch compilation are all from the Kolar gold fields. -0.30 km-‘, as before. Fig. 5 shows the resulting pre-
When one includes a point at 10.06 km (not shown), dictions of Eq. (39) for the zenith angle distribution
these data seem to be systematically lower than the of the Soudan I events compared with Ruddick’s data.
fluxes measured by LVD. The LVD collaboration adds It may be helpful to note, for purpose of compari-
a further complicating factor by noting that for slant son with Fig. 4, that 1Okm slant depth corresponds
depths less than 14 km, the LVD results cannot be di- at Soudan to a cosine(zenith angle) of about 0.18.
rectly compared with any other experiment that is not I emphasize that the flux prediction is absolute. The
physically located at Gran Sasso. normalization is given by the the estimated primary
Ruddick [22] has collected a sample of 293 K muon flux.
events that passed through a 2.5 m diameter spherical A Monte-Carlo prediction for the same data can be
region surrounding the center of the Soudan I detec- obtained by using the parameterization of Ref. [9] for
tor during 172 live days of running time. He esti- a power spectrum with index 3.7, divided by the co-
mates the specific gravity of the rock to be 2.95 and sine of the zenith angle. The resulting curve would
the respective values of the a and b coefficients to be be indistinguishable from the curve shown in Fig. 5.
296 J. L. Uretsky /Nucl. Instr. and Meth. in Phys. Res. A 399 (1997) 285-300

Fig. 5. Zenith angle distribution of 293 K muon events that passed through a 2.5 m spherical region within the Soudan 1 detector during
172 days of ruuning time. The data were collected by Ruddick [22]. Also shown is prediction of Eq. (39), the normalization determined
by the cosmic-ray primary flux. A Monte Carlo index 3.7 power-law spectrum divided by the cosine of the zenith angle, normalized to
the vertical value would be indistinguishable from the curve shown.

Both curves somewhat overestimate the muon rate for constitute background to a search for high energy neu-
values of cosine( zenith angle) less than about 0.3. trinos from active galactic nuclei or other esoteric
I conclude that the intensity-depth relations from extra-terrestrial sources.
Eq. (39) are substantially the same as Monte Carlo pre- The collision probability for a TeV muon passing
dictions, except that Eq. (39) provides an absolute nor- through iron is, in the scaling approximation [ 141
malization for the flux. Both are in reasonable accord
with experiment, given experimental uncertainties.
4C.f1df= ;lO@1 - $1 -f)+(l
3-88 _f)2
[

+0.072(1-f+if2)1n(y)
8. High energy neutrinos, an application

Eq. (37) provides means for estimating the rate at +0.0087( f + 0.007 1)-2 d f cm2/g (40)
I
large zenith angles for discrete large energy losses
(“catastrophic” events) in the Soudan II detector by $(f)df in Eq. (40) is the probability that a muon
cosmic-ray muons having energies in the TeV range. undergoes a collision in which it loses fractional en-
This estimate is of interest because such events would ergy between f and f + df from brehmsstrahlung,
J.L. Uretsky/ Nucl. Instr. and Meth. in Phys. Res. A 399 (1997) 285-300 297

LOSS PROBABILITY vs. LOSS


0.8 I I I I I I I

0 FOR MUONS OF DESIGNATED ENERGY

‘_
-_
+.. -..
‘\ -..
\. ‘.
. . .
x -_

---~‘=--~--m.-,?___.____
ii -- --
I I 1 I I 1 I
OO 0.02 0.04 0.06 0.08 0.1 0.12 0.14
LOSS IN TeV
.- 5 TeV
-^ 20TeV
- 45 TeV
-- 8OTeV
- 125TeV
Fig. 6. Probability that a muon having the designated energy will lose energy in excess of the designated loss value (TeV) when traversing
the Soudan II detector.

pair production or a nuclear collision as it passes Fig. 6 shows probabilities for muons of several des-
through iron. Multiplication of 4( f ) by the mean den- ignated energies to lose energy greater than “loss” TeV
sity p of the detector, about 1.6 g/cm3, then gives the in traversing the Soudan II detector.
inverse mean free path for a collision in the detector The probability per unit length of detector track for
with fractional energy loss between f and f + df. cosmic ray muons to undergo collisions with fractional
The probability that a muon with energy in the TeV energy loss f is obtained by folding Eq. (37) with
range will deposit at least a fraction fo of its en- Eq. (40). This probability is approximately constant
ergy while traversing length L of detector (500 cm for throughout the volume of the detector. Multiplying by
Soudan II) is then simply the detector length and the exposure then gives the
expected number of such events.
The Soudan II detector is approximately a rectan-
@Gfo)=l -exp [-,{$(f)df]. (41)
gular parallelepiped having sides measuring (in m)
298 J.L. Uretsky/Nucl. Instr. and Meth. in Phys. Rex A 399 (1997) 285-300

I- --_&----------!

--- I I I A’\ &_


J l
\\_ l*
0.1
___---
__---__ _-.
-.- I
\
- I I
\
_-.--_---__-. ---_- __. -_._A

O.O’J.35 0.3 1 0.27 0.23 0.19 1


0.15

c4miMoftithAa~e
- 1oGevL4ws
l *. 40 Gev1oss
- 70 GeVloss
- 1ooGevloss

Fig. 7. Differential zenith angle distribution for energy deposition of the indicated value by muons traversing the Soudan II detector at
large zenith angles in 372 days running time.

5 x 8 and 5 x 14. DeMuth and Goodman4 have done 9. Conclusions and suggestions for further work
a Monte Carlo study of the effective exposure area of
the detector for horizontal incoming muons. Their area I have shown that the one-dimensional diffusion
is about 8.24 x lo5 cm2, their estimated efficiency is equation for muon propagation through the earth,
0.77, and their solid-angle binning Eq. (5), in the scaling approximation with constant
a and b coefficients, can be solved by a convergent
d4 d cos( 0) = 2~ x 0.005.
series in inverse powers of the energy whenever the
The total exposure over a 372 day time span was then flux at the surface can be expressed by such a series.
6.41 x 1012 cm* ssr. The coefficients in the expansion are reminiscent of
Figure 7 shows the prediction for the differential an n-step random walk, where n is the nth term in the
and cumulative (in energy loss) distributions of catas- expansion.
trophic events during the 372 day running time at I compare the results with published Monte Carlo
Soudan II. calculations and data. The dependence of intensity
upon depth is nearly identical to the Monte Carlo
4 Private communication result for realistic spectral index (3.7), less so for an
J.L. Uretsky /Nucl. Ins@. and Meth. in Phys. Res. A 399 (1997) 285-300 299

unrealistically flat index (2.0). The mean muon en- to the definitions of the 3 and 1 parameters might
ergy at depth is the same for both calculations at be improved. In particular, one might seek a better
3 km depth, and the energy spectra at that depth look way to sum the series in powers in inverse energy.
very similar. The predicted spectra become different An approach should be found for restoring, at least
at deeper depths. Monte Carlo predicts a monotoni- partially, the suppressed energy dependence of the b
cally increasing mean energy with depth; the present parameter. And the problem of calculating
calculation predicts a decreasing mean energy below two-dimensional effects also presents an interesting
about 5.5 km. The differences are not profound, how- challenge.
ever, because the spectra are in fact quite flat (on a
linear plot).
The present calculation predicts the absolute muon Acknowledgements
intensity as a function of depth. The predicted fluxes
accord with experimental data down to about 4 i km I am indebted to Maury Goodman for motivating
depth, below that depth they are about 50% higher than this work, to Maury, Tom Fields and Keith Ruddick
a data fit from the LVD collaboration (which, we are and David DeMuth for helpful comments, and to Keith
warned, should not be compared with data not taken for supplying his Soudan I dataset.
at Gran Sasso). A power spectrum with index 3.7,
normalized to the present prediction at 3 km (using
either the present formalism or Monte Carlo), falls off Appendix
with depth a bit faster than either the present prediction
or the LVD fit. The most general surface spectrum to which the
Comparison with data from Soudan I gives a some- present analysis is applicable can be expressed as a
what similar result. The predictions accord well with convergent series in inverse powers of the energy,
data down to a slant depth of about 6 km, below that cc

depth the predictions appear to be about 50% high. j-(0, E) = c C/&-P--I. (A.11
I=0
A power spectrum of 3.7, normalized to the vertical
intensity, would be indistinguishable from the present The surface condition on the alpha coefficients now
predictions over the range of angles at which measure- becomes a,(O) = q.
ments were made. Define a new coefficient /?l(x) by
I have used the formalism developed in this paper
to estimate the annual rate of horizontal cosmic ray CII= e -[(P+~)b+r.(P+l-l)lx~~(x). (A.2)
muons with large energy losses.

x
All of the numerical calculations reported in this The PI coefficients satisfy the recursive condition:

J
paper were performed on a personal computer using
either a spreadsheet (Quattro Pro) or Mathcad 6.0. PI+1 = cl+1 - 4P + 0 du e[b-s(P+‘)l”/?,(u) (A.3)
0
Computation times (as when changing parameters)
were never more than a few seconds. with the PI vanishing for negative 1 and PO= co.
The results of the present investigation suggest an The a~ can be bounded, just as in Section 5 with the
improved parameterization of Monte Carlo results result that
over that contained in Ref. [9]. I assume that the
authors of Ref. [9] constrained themselves to a two-
parameter fit to their data in order to compare their
results with the “soft” analytic approximation. The
present results suggest that an improved fit should be (A.4)
possible by introducing a parameter equivalent to the
> of Section 6. where b stands for b - 6( p - 1).
It should be possible to extend the work reported in The next steps are to put the bound Eq. (A.4) into
this paper in several directions. The estimates leading the expansion Eq. (A.l), and exchange the k and 1
300 J.L. Uretskyl Nucl. Instr. and Meth. in Phys. Rex A 399 (1997) 285-300

summations. The result, which is made more transpar- [5] K. Greisen, Phys. Rev. 73 (1948) 521.
ent by use of the gamma function relation [6] J. Nishimum, Proc. Id. Conf. on Cosmic Rays at Jaiper
4 (1963) 224; Handbuch der Physik, Springer, Heidelberg,
T(z)T( 1 - z) = -d-- (A.51 1965, Bd., 4612.
sm( w) [7] V.I. Gurentsov, G.T. Zatsepin, E.D. Mikhal’chi, Sov. J. Nucl.
Phys. 23 (1976) 527.
is [8] P.H. Barrett, L.M. Bollinger, G. Cocconi, Y. Eisenberg,

1)]-*-’.
K. Greisen, Rev. Mod. Phys. 24 (1952) 133.
f(x,c) < e--d(p-l)r E cl [sebx + ff(e” - [9] P. Lipari, T. Stanev, Phys. Rev. D 44 (1991) 3543.
[lo] Private communication from D. DeMuth.
[II] A. Dar, Cosmic Ray Muons at Ground Level and Deep
(A-6)
Underground, Technion-Phys-84-4 1, unpublished.
The sum on the right converges for sufficiently large [12] T.K. Gaisser, Cosmic Rays and Particle Physics, Cambridge
University Press, New York, 1990, pp. 70 and 71.
E to $(x)-Pf(E(x)) where
[13] B. Rossi, High Energy Particles, Prentice-Hall, New York,
1952, Section 5.5.
E(x) = .sebx+ $e” - 1). (A.7) [ 141 R.K. Adair, H. Kasha, Cosmic Ray Muons, in: V.W. Hughes
and C.S. Wu (Eds.), Muon Physics, vol. I, Academic Press,
The convergence is guaranteed because it was as- New York, 1977, pp. 330-334.
[15] W.T. Scott, Phys. Rev. 80 (1950) 611.
sumed that the series in Eq. (A. 1) converged for some
[16] L.S. Schulman, Techniques and Applications of Path
sufficiently large E. Then, if a closed form is known Integration, Wiley, New York, 1981, Ch. 9.
for the function f,that closed form may be used to [17] E.T. Whittaker, G.M. Watson, Modern Analysis, Cambridge
analytically continue the solution to smaller values of University Press, Cambridge, 1950, p. 20.
the argument, as was done in this work. [ 181 I.S. Gradstein, I.M. Ryzhik, Table of Integrals, Series, and
Products, Academic Press, New York, 1965, Tables 8.36.
[19] G.L. Cassiday, Phys. Rev. D 3 (1971) 1109.
References [20] M. Crouch, 19th CRC, Paper HE 5.1-10.
[21] M. Aglietta et al., Astropart. Phys. 3 (1995) 311.
[l] B. Rossi, K. Greisen, Rev. Mod. Phys. 13 (1940) 240. [22] K. Ruddick, Soudan Collaboration note PDK- 435 of May 7,
[2] J.F. Carlson, J.R. Oppenheimer, Phys. Rev. 51 (1937) 220. 1990 (unpublished) and private communication.
[3] H.J. Bhabha, W. Heitler, Proc. Roy. Sot. 159 (1937) 432. [23] B.C. Rastin, J. Phys. G 10 (1984) 1609.
[4] S. Hayakawa, J. Nishimura, Y. Yamamoto, Supp. Prog. Theor. [24] B.C. Nandi, MS. Simha J. Phys. A 5 (1972) 384.
Phys. 32 (1964) 104. [25] O.C. Allkofer et al., Phys. Lett. B 36 (1971) 425.

You might also like