PUPT-2105

hep-th/0312321

String creation and cosmology1

arXiv:hep-th/0312321v1 31 Dec 2003

Steven S. Gubser

Joseph Henry Laboratories, Princeton University, Princeton, NJ 08544

Abstract

I argue that string creation may have played a role in reheating the universe after inflation.
For strings in four dimensions that arise from branes wrapping cycles in the extra dimensions,
estimates from effective field theory show that the string tension need only fall a couple
of orders of magnitude below the Planck scale in order for string creation to extract a
significant fraction of the energy in coherent motion of the inflaton field. I also comment on
a special four-dimensional background which involves only Neveu-Schwarz fields and offers
the possibility of studying closed string creation on the worldsheet.

December, 2003

1
Contribution to the proceedings of QTS3, held at the University of Cincinnatti from September 10 to
14, 2003.
1 Introduction
As the experimental data on the large-scale structure and early history of the universe
improve, there has been increased interest in seeing how ideas from string theory may fit in to
understanding cosmology at the earliest of times. On the whole, efforts in this direction have
focused on either finding a home for inflation in string theory (for instance, [1, 2]) or replacing
it with something quite different but string-inspired (for instance, [3, 4, 5]). Important as it
is to establish in a fundamental theory of gravity either a sound underpinning for inflation
or a solid alternative to it, my focus here will be instead on the possibility of string physics
playing an important role just after inflation has ended. The characteristic scale of string
theory is usually closely to the Planck mass, considerably higher than a typical inflaton
mass scale. However, it turns out that objects known as “tensionless strings” [6, 7, 8, 9]
can be produced copiously from coherent motion of the inflaton field. Without significant
fine-tuning, it is possible that creation of such strings can extract a significant fraction of
energy from the inflaton, thus providing an alternative to preheating [10, 11, 12].
String creation is interesting in its own right, but on the whole it is ill-understood. The
best studied examples of it are decaying D-brane solutions [13, 14, 15, 16], where boundary
CFT methods are available. We will rely instead on estimates from effective field theory,
employing a method of steepest descent to extract Bogliubov coefficients from approximate
wave equations for massive string modes. This method was developed in [17, 18], and its
application to preheating was considered in [18]. It would be highly desirable to improve our
ability to compute string creation amplitudes: this is an important feature of time-dependent
backgrounds in string theory, and it is relevant to almost all the stringy proposals for the
physics of the early universe.
The organization of this paper is as follows. In section 2 we will briefly summarize the
results of the steepest descent analysis of string creation. In section 3 we will explain in
some detail how this analysis may be applied to provide an alternative to preheating. In
section 4 we will remark on a special background in which a worldsheet treatment of closed
string creation may be possible.
The results summarized here are based mainly on [18], except that section 4 is based on
ongoing work with J. Friess and I. Mitra.

2 Estimating string creation

The density of states in tree-level string theory is exponential:
dN 1
∝ E γ eE/TH TH = q , (1)
dE 2π α′ c⊥ /6

where TH is the Hagedorn temperature and γ depends on details of the string theory in
question. The quantity c⊥ is the central charge for transverse modes: c⊥ = 12 for the type II

1
superstring. The exponential density of states (1) is in contrast with the power-law behavior
one encounters for point-particle theories. The states that dominate (1) are highly excited
strings. String interactions are not taken into account in (1). TH appears to be a temperature
at which some phase transition occurs. Above the phase transition strings themselves may
not be a good description of the physics. There is an extensive literature on this so-called
Hagedorn transition (see for example [19, 20, 21]), and string interactions do enter into the
discussion in important ways. For the purposes of the present discussion, we will not need
to consider interactions explicitly.
The on-shell constraints for quantum states of a string include L0 |ψphys i = 0, which leads
to an equation roughly of the form

χ̈ + ω(t)2 χ = 0 , (2)

where ω(t)2 depends on the excitation level and the momentum: in a flat background,
ω 2 = k 2 + N/α′ , where N is the level (up to factors of 2 that depend somewhat on details).
In a cosmological background, a reasonable first guess is

ω(t)2 = k 2 + a(t)2 N/α′ , (3)

where a(t) is the scale factor and t is conformal time. One would of course like to improve
upon (2): this is one place where input from a worldsheet treatment is needed. The form (3)
is predicated on working with the string frame metric, where the parameter α′ is a constant
length squared.
The steepest descent method, developed in this context in [17, 18], allows us to crudely
estimate the Bogliubov coefficients for (2): if |β| ≪ 1, then
iπ 4 t∗
 Z   Z 
β≈ exp −2i dt ω(t) exp −2i dt ω(t) , (4)
3 −∞ r

where t∗ = r − iµ is the location of the zero of ω(t)2 in the lower half plane which is closest
to the real axis, and r and µ are real. The last of the three factors in (4) is the most
important: it leads to exponential suppression of |β|2 when ω is large. Assuming that ω(t)
is well-approximated by an elliptical arc over the contour from r to t∗ , one has
 2
2 π
|β| ≈ e−πµω(r) . (5)
3
From this one can estimate the total number of strings created:
Z ∞ dN 2 Z ∞
Ntot = dω |β| ∼ dω eω/TH e−πµω . (6)
dω
Evidently, this total number converges precisely if

πµTH > 1 . (7)

2
If this bound is violated, it’s somewhat like passing to a temperature above the Hagedorn
temperature: clearly, interactions and/or back-reaction must then be involved in rendering
the total energy passing into strings finite.
The punch-line of the next section will be that, for non-perturbative strings, it is possible
without appreciable fine-tuning of parameters to violate (7) just after inflation. If this
happens, then strings must indeed play an important role in reheating. Moreover, it would
be highly excited strings that would make the dominant contribution, so the effects of strings
would be quite unlike anything in a point-particle theory—hence characteristic of string
theory.
Needless to say, it would be very desirable to learn what happens in time-dependent
backgrounds when the number or energy of strings created diverges. Decaying D-branes
may be the simplest laboratory in which to attack this question: see for example [22].

3 Extracting energy from inflaton oscillations

Let us start with a brief overview of the theory of preheating [10, 11, 12], which will be a
useful point of comparison when considering post-inflationary string creation.

• At the end of inflation, the universe is cold and empty.

• The inflaton φ is oscillating around its minimum with an amplitude assumed to be

comparable to the Planck scale.

• These coherent oscillations may die off via perturbative production of particles. This
is ordinary reheating.

• Preheating can occur if another boson has a coupling gφ2 χ2 to the inflaton.

• The effective mass of χ varies: m2eff = m20 + gφ2. This means parametric resonance can
occur and give exponentially growing occupation numbers for χ.

• Preheating is “efficient” iff m2eff varies by a large factor from its minimum to its maxi-
mum: m20 needs to be a small contribution to the mass.

The last point appears contrived, especially when it is stated more precisely, as we will
do shortly. “Efficient” in this context means that a significant fraction of the energy in the
coherent inflaton oscillations should be extracted via the parametric resonance. We may then
ask, if string states could be produced in a similar way, could they play a role in reheating?
We have already noted in the introduction that the string scale is usually only slightly
smaller than the four-dimensional Planck scale, MPl,4 = (8πGN )−1/2 ∼ 1018 GeV, while
a typical value for the inflaton mass (in power-law inflation) is minflaton ∼ 1013 GeV. It
will not come as a surprise that this prevents copious production of fundamental strings,
unless we make somewhat unnatural assumptions. Although our eventual aim is to consider

3
production of non-perturbatively constructed strings which are tensionless at some point in
moduli space, it will be instructive first to consider the production of fundamental strings in
slightly more detail; also, this discussion will tie in with the special background discussed in
section 4.
The first goal in assessing a model of preheating is to find the regime of parameters
where preheating is efficient. The analogous goal here is to find where the bound (7) is
first violated. When this happens, the particle occupation numbers of the individual string
modes should still be small: if they are not, then according to (6), the number of strings
created will diverge exponentially. Thus we have no need of parametric resonance and can
continue to use the steepest descent method.
The onshell condition in string metric is, approximately, ω 2 = k 2 + m2 |gttstr |, where t is
conformal time. The string metric is related to the Einstein metric by

ds24E = eγϕ/MPl,4 ds2str = a(t)2 (−dt2 + d~x2 ) , (8)

where ϕ is the dilaton and γ is a constant of order unity. Thus

N
ω(t)2 = k 2 + eγϕ(t)/MPl,4 a(t)2 . (9)
α′
The dilaton is probably not the same field as the inflaton; however, it seems reasonable to
assume some finite overlap, so when the inflaton is oscillating at the end of inflation, the
dilaton is too. Thus ω(t)2 oscillates between slowly varying extremes:
2 2
ωmax + ωmin ω 2 − ωmin 2
ω(t)2 = + max cos Ωt
2 2
2 N
ωmin = k 2 a(t)2 + eγϕmin /MPl,4 ′ (10)
α
2 N
ωmax = k 2 a(t)2 + eγϕmax /MPl,4 ′ .
α
The zeroes of ω(t)2 are
2πn i ωmax + ωmin
t±
∗,n = ± log , (11)
Ω Ω ωmax − ωmin
and it is plausible to assume that t∗ = t−
∗,0 = r − iµ makes the biggest contribution to string
production. String production is finite if

πµTH > 1 , (12)

which translates approximately to

√
τmin >
∼ minflaton , (13)

where τmin is the minimum string tension as measured in Einstein frame.

If the bound (12) is violated, then what must happen is that a significant fraction of
the energy of the inflaton oscillations passes into excited string states, and the back-reaction

4
slows down the inflaton. This would correspond to the idea of efficient preheating. Note
√
however that ωmin ∼ τmin and Ω ∼ minflaton , so for (13) to occur, the string tension as
measured in Einstein frame must dip temporarily below minflaton . This is possible, but it
seems like asking a lot in conventional string compactifications. The upshot, then, is that
dimensional analysis didn’t fail, and we have proposed a contrived scenario.
Let’s compare with conventional preheating, where it is useful to define “Mathieu pa-
rameters” A and q as follows:
2 2
ωmax ωmin
 
A + 2q = , A − 2q = . (14)
2Ω 2Ω
√
A “broad resonance” occurs in the standard theory when 2q < A < 2q + bq for some order
one constant b: this leads to efficient particle production. On the other hand, efficient string
production occurs when 2q < A < 2xq for some x > 1. So string production occurs in a
qualitatively bigger wedge of parameter space, albeit still a small one.
Another problem with proposing that fundamental strings might have been created af-
ter inflation is that they can wind around the extra dimensions. In phenomenologically
interesting string compactifications, there usually is a non-trivial first homotopy group on
the compactification manifold, so such windings can be topologically stable. Moreover, the
density of states of wound strings is smaller only by roughly a constant factor compared
to unwound strings, and this constant factor is at most a couple of orders of magnitude in
conventional setups. Wound strings would act as dark matter, or perhaps they could even
have visible sector effects like fractional charge. Just the dark matter consideration is dis-
couraging: a naive estimate suggests that wound strings would overclose the universe today
by something close to a factor of 1017 .
Considering non-perturbatively constructed strings in place of fundamental strings alle-
viates all the difficulties so far discussed. These strings would arise as branes wrapped on a
shrinking cycle in the extra dimensions. Consider for example D3-branes on a shrinking S 2 .
The S 2 may become small only at a point on the compactification manifold. This avoids the
difficulty of having wound strings, because there is a strong energy barrier against wrapped
branes wandering away from the location of the shrinking cycle. The parameter controlling
the size of the S 2 is in fact complex: it is
Z
ϕ= (J2 + iB2 ) , (15)
S2

where J2 is the Kahler form on the extra dimensions (assumed to form a complex manifold)
and B2 is the Neveu-Schwarz two-form potential of string theory. Starting from the DBI
action for D-branes, one arrives at
Z q
S = −τD3 d4 ξ Gµν + Bµν + . . . (16)

for a D3-brane wrapped on the S 2 . Using τD3 ∼ 1/(α′2 gs ), one obtains

MPl,4
τeff ≈ M|ϕ| with M∼ , (17)
gs

5
where ϕ has been rescaled to be a canonically normalized complex scalar field. We have
derived (17) using a particular construction of light strings in four dimensions, but it is in
fact typical for a variety of non-perturbative constructions, many of which are related to one
another by string dualities.
Let us now consider the situation where after inflation, various moduli of string theory
are rolling around, presumably before stabilizing in a preferred vacuum. The scalar ϕ in
particular rolls around the complex plane, and if it comes close to the origin, then the light
strings we have discussed become nearly tensionless. Based on our earlier discussion, we may
expect appreciable string creation at such a moment. To estimate it, suppose ϕ(t) ≈ ϕ0 + ϕ̇t,
where ϕ0 is the closest approach to the origin on the trajectory ϕ(t). Then
q
τeff ≈ M|ϕ0 + ϕ̇t| = M |ϕ0 |2 + (|ϕ̇|t)2 . (18)

To estimate the amount of string creation, we want to find a complex solution t∗ to ω(t)2 = 0.
For highly excited strings, it is equivalent to solve τeff = 0, which gives

t∗ ≡ r − iµ = −i|ϕ0 |/|ϕ̇| . (19)

Thus the convergence criterion on the total number of strings produced is

|ϕ0 | q
πµTH ∼ M|ϕ0 | >
∼ 1. (20)
|ϕ̇|

The minimum tension for the light strings is τmin ≈ M|ϕ0 |. We can estimate that ϕ̇ ∼
MPl,4 minflaton : the ϕ oscillations are of amplitude MPl,4 and of frequency minflaton . The
condition (20) becomes
q !3/2
3/2 3/2
Mϕ30 τmin τmin τmin
1<
∼ = ∼ ∼ 104 2
, (21)
ϕ̇ M ϕ̇ MMPl,4 minflaton MPl,4

or equivalently,
√ 1
τmin >
∼ 20 MPl,4 , (22)

where we have assumed M ∼ 10MPl,4 and minflaton ∼ 10−5MPl,4 .

The estimates leading to (22) are inevitably fairly crude as long as we have not specified
what the inflaton is and how much it overlaps with the scalar ϕ controlling the effective
string tension. The result (22) is rather different from what we had for fundamental strings,
and the reason is that we almost have a discontinuity in the first derivative of ω(t)2 ∝ τeff ∼
M|ϕ| ≈ M|ϕ̇t|.
Evidently, the non-perturbatively constructed strings do not have to dip very far below
the Planck scale in order to violate this convergence criterion! So their participation in
reheating is quite plausible. One would still have to answer the question of how dense in
moduli space the loci of tensionless strings are. Quantifying this is technically difficult, but
the fact that the loci are typically real codimension two is encouraging.

6
4 Backgrounds involving only Neveu-Schwarz fields
Let us now consider a background which might be more amenable to a worldsheet treatment.2
As noted earlier, it is desirable to explore the possibilities for such vacua, since one would
eventually want to go beyond the effective field theory method of estimating string creation
that we described in section 2. Essentially all string theory compactifications give rise to the
following terms in the effective action (written in four-dimensional string frame):
1 Z 4 √ −φ 1 2
 
2
S= d x Ge R + (∂φ) − H
2κ24 2 3
(23)
Vol M6 1
−φ
e = √ 2κ24 = 2πα′ .
(2π α′ )6 gs2

In four-dimensional Einstein frame, defined by ds2E = e−φ ds2str , we have instead

1 √ 1 1
Z  
S= d x g R − (∂φ)2 − e2φ (∂χ)2 ,
4
(24)
2πα′ 2 2
and the fundamental string tension as measured in Einstein frame is τeff = eφ /2πα′. Except
in type I string theory, H3 is the four-dimensional part of the field strength of the Neveu-
Schwarz two-form, and χ is defined via the equation dχ = e−φ ∗ H3 .
The scalar φ and the axion χ parametrize the upper half plane. As a first guess, we might
suppose that there are FRW solutions where χ+ie−φ approximately describes a semi-circular
geodesic in the upper half plane (with the usual metric), and that by making gs small, we can
avoid significant back-reaction of the scalars on the geometry. Solving the scalar equations
of motion in a background assumed to be flat Minkowski space, one finds

t t eφ0 cosh t/t0

 
−φ −φ0 (25)
χ + ie =e tanh + i sech τeff = ,
t0 t0 2πα′
where φ0 and t0 are integrationq constants. The usual convergence criterion for the total
number of strings created is t0 τeff (0) >
∼ 1, or, more explicitly,

2πα′
eφ0 >
∼ t2 . (26)
0

It would appear that by making φ0 very negative (corresponding to extremely weak string
√
coupling), with t0 large compared to α′ but fixed, we can get the bound in (26) to fail. This
is a first indication that this type of solution is an interesting laboratory for string creation.
It is nice that the wave equation for string modes is approximately a Mathieu equation, and
that in the early and late time limits it is possible to construct infinite order adiabatic in
and out vacua.
2
This section represents work in progress with J. Friess and I. Mitra.

7
 In fact, it is not difficult to promote the approximate solution described above to an
exact solution of the equations of motion following from the two-derivative action:

H3 = eφ ∗ dχ = hdx1 ∧ dx2 ∧ dx3 eφ = eφ0 sec ht

!√3 !1/√3
cos ht + sin ht −φ0 cos ht + sin ht (27)
ds24E φ0 2
= −e (sec ht) 2 2
dt2 + e 3 2 2
d~x2 .
cos ht
2
− sin ht
2
cos ht
2
− sin ht
2

Here t runs from −π/2h, where there is a curvature singularity, to π/2h, which is actually
a non-singular infinite future when expressed in terms of proper time. This solution is not
new: to my knowledge it first appeared in [23].
The main qualitative difference between the full solution (27) and the no-back-reaction
approximation (25) is that the in vacuum is more difficult to define in (27), due to the
singularity. Two possibly interesting approaches would be to replace the early time evolution
by an appropriate instanton, or to find a CFT formulation in which α′ corrections are under
control.
The solution (27) can be lifted to ten dimensions, and in the cases where it comes from
IIA strings or E8 × E8 heterotic strings, there is a further lift to eleven dimensions, where
the size of the eleventh dimension is controlled by the dilaton. When the lift is done to
Horava-Witten theory, it describes two E8 that start far apart at t = −π/2h, move swiftly
together to a minimum distance at t = 0, and then move apart again. This is reminiscent
of ekpyrosis, but actually it is quite different: in four-dimensional Einstein frame, there is
never any contraction of the scale factor. There is also no singularity when the branes are
at their distance of closest approach. The reason they don’t actually collide is that the four-
form field G4 is turned on in such a way as to make the E8 branes repel. It is notable that
in ekpyrosis as envisioned in [5], strings become tensionless at the instant that the branes
collide. It seems extremely likely that string creation is important for understanding how
the universe might pass through this singular event.
The point of involving only Neveu-Schwarz fields is that it is more likely that one can
perform a worldsheet analysis of such backgrounds. In fact, the background we have con-
sidered resembles a limit of NS5-branes, doubly Wick rotated so that radius is replaced by
time. The main stumbling block to finding a CFT description is that the geometry does not
factorize in the way that the NS5-brane geometry does, into a three-sphere part, a linear
dilaton background, and flat space.

5 Conclusions
Part of the appeal of the discussion in section 3 is that nearly tensionless strings arise in a
variety of contexts, on loci in moduli space of real codimension two. Thus, independent of
detailed ideas about how string theory connects to the Standard Model, it is plausible to
have non-perturbatively constructed strings become temporarily light enough to play a role
in post-inflationary cosmology. Because we remain ignorant of what string vacuum we might

8
really be living in, it’s highly desirable to have some ideas about characteristic features of
string theory which may become testable, which are difficult to contrive without invoking
strings, and which are common to many backgrounds. Creation of light strings seems to
qualify on all counts. On the other hand, the experimental constraints on precisely how
the universe was reheated seem rather weak, so to really make the idea of post-inflationary
string creation testable, one must gain a better idea of how such string creation would affect
quantities that will be measured in the near future, like properties of the primordial spectrum
of perturbations.
On the formal side, a better understanding of string creation from a worldsheet point of
view is needed, and it would also be useful to learn more about the decay of highly excited
strings. One reason that it is particularly important to have some examples of closed string
creation described on the string worldsheet is that for bounds like (7) to be violated, one is
usually forced to have string scale curvatures in the background.3 So a worldsheet treatment
is needed to be sure one has α′ corrections under control, even before quantum creation
processes are contemplated. I hope to report on some of these issues in the future.

Acknowledgments
I thank the organizers of QTS3 for putting together a stimulating workshop. This work was
supported in part by the Department of Energy under Grant No. DE-FG02-91ER40671, and
also by a Sloan Fellowship.

References
[1] S. H. S. Alexander, “Inflation from D - anti-D brane annihilation,” Phys. Rev. D65
(2002) 023507, hep-th/0105032.

[2] S. Kachru, R. Kallosh, A. Linde, J. Maldacena, L. McAllister, and S. P. Trivedi,

“Towards inflation in string theory,” JCAP 0310 (2003) 013, hep-th/0308055.

[3] R. H. Brandenberger and C. Vafa, “Superstrings in the early universe,” Nucl. Phys.
B316 (1989) 391.

[4] G. Veneziano, “Scale factor duality for classical and quantum strings,” Phys. Lett.
B265 (1991) 287–294.

[5] J. Khoury, B. A. Ovrut, N. Seiberg, P. J. Steinhardt, and N. Turok, “From big crunch
to big bang,” Phys. Rev. D65 (2002) 086007, hep-th/0108187.

[6] E. Witten, “Some comments on string dynamics,” hep-th/9507121.

3
The treatment of nearly tensionless strings in section 3 may be an exception to this.

9
 [7] A. Strominger, “Open p-branes,” Phys. Lett. B383 (1996) 44–47, hep-th/9512059.

[8] O. J. Ganor and A. Hanany, “Small E8 Instantons and Tensionless Non-critical

Strings,” Nucl. Phys. B474 (1996) 122–140, hep-th/9602120.

[9] N. Seiberg and E. Witten, “Comments on String Dynamics in Six Dimensions,” Nucl.
Phys. B471 (1996) 121–134, hep-th/9603003.

[10] J. H. Traschen and R. H. Brandenberger, “Particle production during

out-of-equilibrium phase transitions,” Phys. Rev. D42 (1990) 2491–2504.

[11] L. Kofman, A. D. Linde, and A. A. Starobinsky, “Towards the theory of reheating

after inflation,” Phys. Rev. D56 (1997) 3258–3295, hep-ph/9704452.

[12] I. Zlatev, G. Huey, and P. J. Steinhardt, “Parametric resonance in an expanding

universe,” Phys. Rev. D57 (1998) 2152–2157, astro-ph/9709006.

[13] A. Sen, “Rolling tachyon,” JHEP 04 (2002) 048, hep-th/0203211.

[14] A. Strominger, “Open string creation by S-branes,” hep-th/0209090.

[15] A. Maloney, A. Strominger, and X. Yin, “S-brane thermodynamics,” hep-th/0302146.

[16] N. Lambert, H. Liu, and J. Maldacena, “Closed strings from decaying D-branes,”
hep-th/0303139.

[17] A. E. Lawrence and E. J. Martinec, “String field theory in curved spacetime and the
resolution of spacelike singularities,” Class. Quant. Grav. 13 (1996) 63–96,
hep-th/9509149.

[18] S. S. Gubser, “String production at the level of effective field theory,”

hep-th/0305099.

[19] Y. I. Kogan, “Vortices on the world sheet and string’s critical dynamics,” JETP Lett.
45 (1987) 709–712.

[20] J. J. Atick and E. Witten, “The Hagedorn transition and the number of degrees of
freedom of string theory,” Nucl. Phys. B310 (1988) 291–334.

[21] S. S. Gubser, S. Gukov, I. R. Klebanov, M. Rangamani, and E. Witten, “The

Hagedorn transition in non-commutative open string theory,” J. Math. Phys. 42
(2001) 2749–2764, hep-th/0009140.

[22] D. Gaiotto, N. Itzhaki, and L. Rastelli, “Closed strings as imaginary D-branes,”

hep-th/0304192.

[23] E. J. Copeland, A. Lahiri, and D. Wands, “String cosmology with a time dependent
antisymmetric tensor potential,” Phys. Rev. D51 (1995) 1569–1576, hep-th/9410136.

10