Volume 120B, number 1,2,3 PHYSICS LETTERS 6 January 1983

THE NOT-SO-HARMLESS AXION

Michael DINE
The Institute for Advanced Study, Princeton, NJ 08540, USA
and
Willy FISCHLER
Department of Physics, University of Pennsylvania, Philadelphia, PA 19104, USA

Received 17 September 1982

Received manuscript received 14 October 1982

Cosmological aspects of a very weakly interacting axion are discussed. A solution to the problem of domain walls dis-
cussed by Sikivie is mentioned. Demanding that axions do not dominate the present energy density of the universe is shown
to give an upper bound on the axion decay constant of at most 1012 GeV.

It has been suggested that the strong CP problem will occur, and 0 will take some random value on the
may be solved by extending the Peccei-Quinn idea interval [0, 2n] (actually, on a somewhat smaller inter-
[1] to grand unified theories [ 2 - 4 ] . In other words, val; see below).
one should require of a grand unified theory that it At temperatures well below the QCD scale, we can
possesses a U(1) symmetry broken explicitly only by calculate the axion potential using current algebra ar-
anomalies. Moreover, this symmetry should be broken guments. For definiteness, we focus on the model of
spontaneously at the unification scale. In such a theory, ref. [2], in which the Peccei-Quinn symmetry is
the CP violating angle, 0, becomes dynamical; it is the broken by the expectation value of a singlet field (with
would-be Goldstone boson (axion) of this spontaneous- Peccei-Quinn charge 1)
ly broken symmetry. QCD gives rise to a potential for
(~0) = 2 - 1 / 2 f A e i° .
this angle with a minimum very near the origin. The
axion itself receives a tiny mass as a result of this sym- Following ref. [6], it is natural to work with the
metry-violating potential. Its interactions are extreme- anomaly-free, partially conserved current (specializing
ly weak, having a strength inversely proportional to for simplicity to the case of two light quarks; inclusion
)cA, the axion decay constant. In particular, this ax- of the strange quark is straightforward)
ion would not appear in experiments nor would it
7** -]u-
"PQ -- [N/(1 + Z ) ] (fiVu75 u + Z d T u V s d ) ,
play any role in astrophysical environments such as
stars. where Z = mu/m d. We must also include the current
However, this lack of interaction raises concerns of axial isospin,
of a cosmological nature. For temperatures much ].3 1 -
u A = ~ (UTuYs u -- a Yu 3'5 d)
higher than QCD scales, we expect 0-dependent ef-
fects to fall as some large power o f T [5]. Thus, at the Following a method discussed by Sikivie, we write the
grand-unified scale, the Peceei-Quinn symmetry will expectation values of fermion bilinears as
be an essentially exact symmetry. At temperatures
(fiLUR) = ](fiLUR)I exp{i{a + IN/(1 + Z)] 0}} ,
above fA, this symmetry will be unbroken; as the tem-
perature is lowered, spontaneous symmetry breaking
(dLd R) = I(CtLdR)I exp {i{---a + [NZ/(1 +Z)] 0}} .

0 031-9163/83/0000-0000/$ 03.00 © 1983 North-Holland 137

Volume 120B, number 1,2,3 PHYSICS LETTERS 6 January 1983

Then the effective potential for a and 0 is obtained by This equation.has the solution, for a radiation-dominat-
replacing fields by their expectation values (using (flu) ed universe (R /R = 1/2t)
<dd>) 0 = t - 1/4 [ajl/4(mt) + bJ_ 1/4( mr)]
V = I(fiu)l [m u cos(a + [N/(I + Z ) ] 0)
t~'m-1 O0(mt)- 3/4 sin mt ' (4)
+ mdcos(ZNO[(1 +Z) - O ] • (1)
Here afar is the pion field and OfA is the axion field. For a matter-dominated universe (R/R = 2/3t)
Expanding to second order in the n 0 and axion fields,
0 = t - 1/2 [a,Jl/z(mt) + b'J l/2(mt)]
the cross terms cancel (this is the reason for the choice
of Z) we find the usual results:
t~,m -100(mt)- 1 sin mt (5)
2 2
f~rm~r = (m u + md)(fiu),
2 2 2 2 2 2 The coefficients of the Bessel functions in these equa-
f~m A = [Z/(1 +Z) I N mJCr tions depend on the initial conditions which we im-
As Sikivie has emphasized, this potential has a discrete pose. The asymptotic behavior, however, is not par-
Z N symmetry. It corresponds to the transformation ticularly sensitive to these conditions. The question
of appropriate initial conditions will be described be-
O-*O+2nJN, 0~-+a + 2rr - 2hi(1 + Z)
low, when we describe the axion potential at high tem-
Modulo this discrete transformation, the mini- peratures.
mum of the potential lies at 0 = 0 (it will, of course, The axion mass, of course, does not simply turn on
acquire a tiny non-zero value once we include CP vio- at the phase transition, but presumably falls off, at
lating effects). However, the question arises, how quick- high temperatures, as some power of the temperature.
ly will 0 approach its minimal value? In particular, The 0-dependence of the free energy arising from in-
since the axion is so weakly coupled to matter, there is stantons can, in fact, be calculated at high temperature
virtually no damping term in the equation of motion and will fall as a high power of T [5]. For example,
for 0. If we neglect, for a moment, the expansion of for three light quarks, instanton calculations give a re-
the universe, 0 obeys the equation (keeping, for sim- sult for the free energy
plicity, only the quadratic term in the potential),
F ~ -C(rrT)- 8 A9V mumd ms [47r2/g2(T)l 6 cos NO ,
g + ['0 + m20 = 0 (2)
(6)
The only damping comes from the decay of the ax- Here the constant, C, is obtained from a numerical inte
ion to two photons. The axion lifetime in vacuum is gration of an expression for the instanton density due
1058 yr. In the early universe (essentially for any tem- to Gross et al. [5]. We have made the estimate
perature greater than about 10 eV) in fact, it is not
C = 2.4 × 10- 2 (7)
clear that the axion can decay at all! The reason is that
for these temperatures the Debye frequency of the Apv is the QCD scale factor appropriate to the Pauli-
plasma is larger than the axion mass. In any case, P is Villars scheme. From ref. [7], we know Apv ~ A ~ .
extremely tiny. Thus if we neglect the expansion of Also, in obtaining the expression (6), it should be
the universe, 0 simply rolls back and forth in the po- noted that we have neglected the scale dependence of
tential, eq. (1), essentially forever. the (l/g2) 6 factor in the integration over scale sizes.
Of course, the universe does expand, and this will Of course, at low temperatures, many effects (apart
induce a significant decrease in 0 with time. At tem- from instantons) presumably conspire to give the 0-de-
peratures small compared to QCD scales, the generally pendence of the vacuum energy, and these presumably
covariant version of eq. (2), in a Robertson-Walker also fall as some power of T. It is possible that even at
metric, is large temperatures these effects are the dominant ones.
Thus to obtain an idea of what 0 may do in the early
0"+ 3(t~[R)O + m20 = O . (3)
universe not too deeply tied to our prejudices from in-

138
Volume 120B, number 1,2,3 PHYSICS LETTERS 6 January 1983

stanton physics, we take I f 0 ~ 1 at T = 100MeV, then 0 ~ 10 -18 today. Cer-

tainly, then, we do not need to worry about the neu-
m2(T) = ~2(T/T)-a , (8)
tron electric dipole moment.
and consider a range of plausible values for ~ , t and a, Also, in this picture, we can discuss a serious con-
including the values obtained from instanton computa- cern raised by Sikivie [8]. He points out that, since
tions. Note that eq. (8) is written in this form for pur- the Z N symmetry respected by QCD is spontaneous-
poses of motivation; obviously only the product ~ 2 ~ a ly broken by (~b), one expects the formation of domain
is physically meaningful. We expect T to be some char- walls. This leads to conflicts with standard cosmology.
acteristic QCD scale, and m to be some fraction of the One way out of this problem is to choose the fermion
zero-temperature axion mass. content so that the left-over discrete symmetry is tri-
The equation of motion for 0 follows from conser- vial* 1. This has been discussed by Sikivie and by
vation of energy at constant entropy. In a radiation- Georgi and Wise [9]. An alternative possibility is sug-
dominated universe one obtains gested by the picture above of the time evolution of 0.
In the early universe scenario discussed by Guth [ 10]
0" + (3/2t)~J + t~ 2 [T(t)/T] - a = 0 . (9)
and refined by Linde [11] and Albrecht and
Our previous remarks about the Peccei-Quinn phase Steinhardt [ 12], the universe goes through a period
transition suggest that at some very small time, to, cor- of exponential expansion at temperatures below the
responding to a temperature well above QCD scales grand unified mass. We live, essentially, in one of the
(the precise meaning of "well-above" will become causal regions of this early universe and hence in one
clear shortly) we should set 0 = 00, tJ0 = 0. The tem- domain. In this region, 0 will take a single value, which
perature in eq. (9) depends upon the time through the will evolve in time according to the equations of mo-
relation tion we have discussed earlier. [One can easily check,
T = (Mp/27t) 1/2 , (10) from eq. (3), that 0 is virtually constant during the
period of exponential expansion.1 Thus a period
where above QCD scales, for three light quarks, of exponential expansion can solve the domain wall
problem, provided that the Peccei-Quinn transition
7~13.
occurs at high enough temperatures, and that one does
This equation has the solution, with these initial condi- not reheat the universe above this temperature.
tions A more serious concern is the energy density as-
0 = O0(t[i ) - 1/4j~ [(t/}~)e] P(v + 1)2" sociated with the axion field, at the time of helium
synthesis and today (this point has been emphasized to
- - 2vp(u + 1)(2fir)l/2Oo(t/t)-(t~+l/2)/2
t>>t
us by L. Abbott and P. Sikivie). The pressure and en-
ergy density of the axion field obey an "equation of
× cos[(t/i)e - ~' Try - 'Tr]
a , (11) state" appropriate to non-relativistic matter, for t
>> r n - 1. This can be seen directly by substituting our
where solutions, eqs. (4) and (5), into the expressions for the
I + , I energy density, P- and pressure, p, and averaging over
t3 = a a 1 23v = E
a period, m - 1
"{ = [ 3 - 1 ( ~ 2 " f a ) l 1 2 ( M p / 2 7 ) - a / 4 ] -1/e (12)
P = "~SAW
',,'2,'A2+m202 )
Note that this time corresponds to a temperature
(rot)-3/2 , radiation domination,
which can be computed from eq. (10). t can be mt} l
thought of as a characteristic time for the motion of
0. In particular, we should choose t O ~ }'. m~>>l(mt)-2 , matter domination, (13)
We are now in a position to discuss the cosmologi-
cal development of 0. First, we can see that 0 today is
extremely small, no matter what we assume for a, T,
,1 This can also be achieved by choosing the Peccei-Quinn
~ , the location and nature of the phase transition, etc. charge of ~ to be N.

139
Volume 120B, number 1,2,3 PHYSICS LETTERS 6 January 1983

1 2 "2
p=~fA(O --m202) these to be simply g - 2 (1 GeV). Of course, the instan-
ton calculation cannot be trusted down to arbitrarily
0. (13 cont'd) low temperatures, since the free energy diverges. We
mt> l will make the ad hoc assumption that the temperature,
Te, is just the temperature where formula (6) gives the
Thus, if at some time O is large compared to the energy zero temperature result, eq. (1). One can check that
density of radiation, the universe will expand as if mat- the final results are not too sensitive to this choice. In any
ter dominated. Note that, in either case 0 m R - 3 for case, these assumptions yield (taking AM-S = 0.150
large times. GeV)
Suppose that the potential is given by the zero tem-
a = 8, ~2~8f2 = 5.8 X 10-12 (GeV) 12
perature expression, eq. (1), below some temperature
T c (presumably near the chiral symmetry violating
phase transition). If the universe has been expanding
T c = 0.089GeV, 7=5 X IO18GeV-I(fA/IO15GeV)I/3
adiabatically since that time, then the ratio of 0 to the Plugging this into our solution, we find that we need
entropy density is a constant. If we calculate the en- (assuming 00 ~ 1)
tropy at T c appropriate to a quark-gluon plasma* 2,
tje requirement that the universe be radiation dominat- fA ~ 1011 GeV (18)
ed at the time of helium synthesis gives in order to satisfy the bound, eq. (15).
0 2 ( r c ) / r 3 ~ 30 GeV - 3 (14) As a more optimistic extreme, we can take

From the limits on the present energy density of the 2=0.1m2, a : 2 ,

universe, and the present temperature of the cosmic
radiation background we obtain the stronger limit = 1 GeV, T c = 1 GeV. (19)

02(To)IT 3 ~ 10 - 4 GeV - 3 . (15) This yields the bound

If we take T c = 0.1 GeV, we obtain, from the latter fA <~ 1012 GeV. (20)
equation So it appears that the estimate from instantons is n o t
0 2 ( T c ) ~ 10 - 7 , (16) likely to be o f f b y more than a factor of 10.
So we see that the current mass density and photon
whereas if we take the much more optimistic value, temperature severely constrain the value which 0 can
T c = 1 GeV, the limit is take at the grand unified phase transition. The invisi-
02(Tc) ~ 10 - 4 . (lV) ble axion idea thus appears to be in serious trouble.
In particular, in conventional grand unified theories
We can calculate 02(Tc) from our solution of eq. (9). satisfying the bounds on f a described above requires
As a pessimistic estimate, we can use the formula for a fine-tuning of one part in 106-1010. Thus the strong
the axion mass implied by instantons. We will see later CP problem, or the problem of the fine-tuning of 0,
that more optimistic assumptions for m, T, T e, and a has simply been replaced by a fine-tuning of other
do not drastically alter the situation.We consider the parameters.
case of three light quarks (this will be seen to be self- A few solutions to this dilemma may be imagined
consistent), and we will, as stated earlier, neglect the but none, at present, look particularly attractive. The
logarithmic variation of the g - 2 ( T ) terms, taking
most obvious is a process occurring after T c which sig-
nificantly increases the entropy (by a factor of order
,2 This is an optimistic estimate. A pessimist would calculate 104). An example is the decay of the long-lived grav-
the entropy relevant to a gas of pions and nucleons. The itinos in certain supersymmetric theories [12]. How-
truth is probably somewhere in between, and depends on
ever, no particularly satisfactory model of this type
the details of the true QCD equation of state at and around
the chiral symmetry breaking transition. These considera- yet exists. Moreover, such a process would also dilute
tions, however, can change our results only by small fac- the baryons, which will be difficult to reconcile with
tors. current ideas concerning baryon number generation.

140
Volume 120B, number 1,2,3 PHYSICS LETTERS 6 January 1983

Finally, apart from fine-tuning, there is another [3] J. Kim, Phys. Rev. Lett. 43 (1979) 103.
disturbing feature of an axion with decay constant less [4] M. Wise, H. Georgi and S.L. Glashow, Phys. Rev. Lett.
47 (1981) 402.
than 1012 GeV. As mentioned earlier, the domain wall
[5] D.J. Gross, R.D. Pisarski and L.G. Yaffe, Rev. Mod. Phys.
problem might be solved by a period of exponential 53 (1981) 43.
expansion. However, in order to generate baryons, this [6] W.A. Bardeen and S.-H.H. Tye, Phys. Eett. 74B (1978)
period must be followed by significant reheating of the 229.
universe, almost certainly to temperatures above 1012 [7] R. Dashen and D.J. Gross, Phys. Rev. D23 (1981) 2340.
[8] P. Sikivie, Phys. Rev. Lett. 48 (1982) 1156.
GeV, i.e. above the temperature of the new Peccei-
[9] H. Georgi and M. Wise, Phys. Lett. l16B (1982) 123;
Quinn transition. So the domain wall problem discuss- G. Lazarides and Q. Shaft, Rockefeller preprint RU89-B027
ed by Sikivie will still be with us. B027 (1982);
All of this adds up to a troubling picture. For the B. Holdom, Stanford preprints ITP-712 (1982), 1TP-713
moment, at least, the strong CP puzzle remains a puz- (1982);
S. Dimopoulos, P. Frampton, H. Georgi and M. Wise,
zle.
Harvard preprint HUTP-82-A030 (1982);
S. Barr, D. Reiss and A. Zee, Phys. Lett. l16B (1982)
We thank L. Abbott, P. Sikivie, L. Susskind and 227;
E. Witten for instructive conversations. This work S. Bar, X.C. Gao and D. Reiss, University of Washington
was supported by the U.S. Department of Energy un- preprint 40048-22P2 (1982);
Y. Fujimoto, K. Shigemoto and Zhao Zhiyong, ICTP
der Contract Nos. EY-76-C-02-3071 0V.F.) and
preprint IC-82-410 (1982).
DE-ACO2-76ERO2220 (M.D.). [10] A. Guth, Phys. Rev. D23 (1981) 347.
[11] A. Linde, Phys. Lett. 108B (1982) 389;
References A. Albrecht and P. Steinhardt, Phys. Rev. Lett. 48 (1982)
1220.
[12] S. Weinberg, Phys. Rev. Lett. 48 (1982) 1303.
[I] R.D. Peccei and H. Quinn, Phys. Rev. Lett. Lett. 38
(1977) 1440.
[2] M. Dine, W. Fischler and M. Srednicki, Phys. Lett. 104B
(1981) 199.

141