A comment on gravitational waves and the scale of supersymmetry breaking

Shamit Kachru (KITP & Stanford)

Based on work with T. He and A. Westphal

(arXiv : 1003.4265)
Friday, May 21, 2010

I. Introduction
There are many reasons to be particularly excited about our prospects for learning new facts about fundamental physics in the next few years. Two of the biggest (and most expensive!) such reasons are:

Friday, May 21, 2010

One will teach us about physics at the weak scale, and the other about the inationary cosmology which preceeded the big bang (reheating).

It is of course quite clear that lessons coming from cosmological distance scales have signicant impact on our thinking about short distance physics, and vice versa.
Friday, May 21, 2010

So: what are the possible new lessons, or interactions, between cosmology and particle theory that could intertwine the LHC and Planck results? * The leading candidate for stabilizing the physics of the weak scale, in many peoples minds, is supersymmetry. And where there is supersymmetry, there is (probably!) supergravity. The SUGRA potential takes the form:
V =e
Friday, May 21, 2010

K i

|W | |Di W | 3 2 MP
2

And in a nearly at-space vacuum, of the sort which we inhabit, this implies that the dominant F-component (the scale of SUSY-breaking) is related to the expectation value of the superpotential in vacuum. Probably the single most important observable in a realistic supersymmetric theory is the scale of SUSY breaking. Equivalently, it is the gravitino mass:
2

2 m3/2

|W | e 4 MP
K

Friday, May 21, 2010

detecting the gravitational wave contribution, and a proximately the same result the other hand, the zeroth should hold for the averag * In inationary cosmology, on even iforder question one would like to answer is: there is some variation. At present observation provides only a weak upp bound on r,What was the scale of quantied properly bu which has not been ination? is something like r < 1 [12]. The COBE observations giv In good normalization, [13] controls the tensor to scalar particular, the scale directly H 1.9(1 + r)1/2 105 , an a ratio r by the formula: using it one nds [11]
Lyth

1/4

(r/.07)

1/4

1.8 10

16

GeV
1/4

(3

Friday, May 21, 2010

Thus a detection of r would give a value V = 2 t Planck will get down to roughly .05, so observation of 16 4 10 GeV. gravity waves in the near future would x the energy scale The slow-roll paradigm also gives of ination at the GUT scale. 1 1 d V r 2 = mP l = (4 mP l dN V 6.9

It is often more natural to talk in terms of the Hubble amental observable in ination is the scale of ination parameter
V =
2 2 3MP H

y, the Hubble observation of primordial H). It directly contro Near term constant during ination, gravitational waves and is a major factor would pin this scale 14 GeV. perturba in setting the at 10 of density le in any realistic supergravity model is the scale of supersymm red by the gravitino mass m3/2 . Quantitatively, one has:

So to be maximally reductive, the next few years may K 2 teach us two fundamental numbers: e |W | 2 m3/2 the scale of supersymmetry breaking, and the scale of 4 MP ination. supersymmetry breaking is going to be:

expectation value of the superpotential and K is that of the K In common combinationwould think that the scale ahler tr models, we which is invariant under K of ppear above in a
Friday, May 21, 2010

m3/2 TeV gravity mediation m3/2 << TeV gauge mediation

This is astronomically lower than the scales we normally associate with ination -- at least 11 orders of magnitude lower than the Hubble scale of ination in models with detectable primordial gravitational waves. I bring this up because recently, two prominent theorists (Renata Kallosh and Andrei Linde) have proposed, based on studies of certain classes of ination models in string theory, that there is a rough bound:

Hination < m3/2

Friday, May 21, 2010

arXiv : hep-th/0411011, arXiv : 0704.0647

If this so-called KL bound is correct, then detection of primordial gravitational waves would rule out SUSY at the LHC, and vice versa. My purpose in this talk is to explain the rough logic that led Kallosh and Linde to propose this bound, and to explain why we think that despite their arguments it is very clearly NOT a universal (or even generic) bound on the behavior of ination models in string theory. Although string theory is an intricate subject, in fact the basic point that leads to the new constraint is just the fact that there are extra dimensions of space-time.
Friday, May 21, 2010

II. String Theory, Extra Dimensions, and all that...

The supersymmetric string theories famously live in ten dimensions. Realistic models are made by curling up six of these on, say, a Calabi-Yau manifold:

By Yaus theorem, there is a moduli space of Ricci-at metrics on the Calabi-Yau space. At the very least, this is a one parameter family of metrics, controlled by the volume of the Calabi-Yau.
Friday, May 21, 2010

In terms of the 4D effective eld theory, this means that in addition to whatever elds one has to represent the Standard Model particles, dark matter, and the inaton, one has a (complexied) volume modulus:
T =

J J +i

C4 = + ia

The complexication is expected because 4D chiral superelds contain complex scalars, and in string theory, the imaginary component is a period of a Ramond-Ramond gauge eld. In the simplest Calabi-Yau compactications, such elds enjoy shift symmetries and give rise to axions in the lowenergy theory. The partner of the volume modulus is then such an axion.
Friday, May 21, 2010

r with a at inaton potential; it comes instead largely from a actication of physics can we expect of a theory with extra What new the extra dimensions of string theory. It is this us of our investigation. and a volume modulus? dimensions

* In dimensional reduction, to go to 4D Einstein frame, one allosh-Linde Problem by a Weyl transformation that must re-scale the metric depends on the volume. This rescales energies in the 4D Linde (KL) problem was originally described as follows (for a m effective potential by a factor:
3

e the original paper [10]). For concreteness, we imagine working 1 V denote its volume modulus el on a Calabi-Yau orientifold, and 3 d the imaginary part being comprised of an axion). In the scenari ialOf course this must arise W (in the in the 4D effective th K and the superpotential naturally eective theory below supergravity, and it does: ture moduli are stabilized by uxes) take the form K = 3 ln(T + T )
Friday, May 21, 2010

aT

Now, you might think other dependence on the volume in specic terms in V in 10D Einstein frame could complicate matters, but locality in the extra dimensions implies that no source of energy can grow faster than the volume itself as one expands the extra dimensions. The Weyl re-scaling, on the other hand, falls off as inverse volume squared. Therefore, all terms in the potential in 4D Einstein frame vanish if one allows the extra dimensions to decompactify. This has some obvious implications:
Friday, May 21, 2010

* Any positive energy conguration has a possibility of relaxing away its energy by rolling to innity, unless it is meta-stabilized.
V
4 3

with DW = 0 before the uplifting is given by 3m2 = |VAdS |. Uplifting of the AdS minimum 3/2 to the present nearly Minkowski vacuum is achieved by adding to the potential a term of the type of C/ n , where is the volume modulus and n = 3 for generic compactication and n = 2 for the highly warped throat geometry. Since the uplifting is less signicant at large , the barrier created by the uplifting generically is a bit smaller than |VAdS |. Adding the energy of the inaton eld leads to an additional uplifting. Since it is also proportional to an inverse power of the volume modulus, it is greater at the minimum of the KKLT potential than at the top of the barrier. Therefore adding a large vacuum energy density to the KKLT potential, which is required for ination, may uplift the minimum to the height greater than the height of the barrier, and destabilize it, see Fig. 2. This leads to the bound (4.1).

100

150

200

250

* While it has been argued that metastable supersymmetry breaking congurations with positive vacuum energy exist One should note exact form of this bound is complicated in this containing additionalany additional energy on topthanKKLT metacontext, that an which depend logarithmically aonbit moreparameters of the (4.1), of the factors certain potential. However, unless these parameters are exponentially large or exponentially small, one stable vacuum (e.g. bound, H m . with ination) creates a < can use the simple form of this associated < Therefore if one believes in the standard SUSY phenomenology with m O(1) TeV, one danger of destabilizing the vacuum. should nd a realistic particle physics model where the nonperturbative string theory dynamics
3/2 3/2

Figure 2: The lowest curve with dS minimum is the potential of the KKLT model. The second one shows what happens to the volume modulus potential when the inaton potential () Vin = V3 added to the KKLT potential. The top curve shows that when the inaton potential becomes too large, the barrier disappears, and the internal space decompacties. This explains < the constraint H m3/2 .

Friday, May 21, 2010

occurs at the LHC scale (the mass of the volume modulus is not much greater than the gravitino

Let us put some more detail into this discussion, and see what kind of bounds we get.

III. Bounds on ination in stabilized IIB orientifolds?

A popular class of models work in the following way. One compacties IIB string theory on a Calabi-Yau (orientifold), and stabilizes almost all of the moduli guaranteed by Yaus theorem using magnetic ux in the extra dimensions. Below the mass scale set by this ux-induced potential, the effective theory has a single modulus, T. The shift symmetry of the axion in T, prevents it from receiving a ux-induced mass.
Friday, May 21, 2010

structure moduli are stabilized by uxes) take the form ntial K and the superpotential W (in the eective theory below the The axion are stabilized by is broken) the form ucture moduli shift symmetry uxes) take(as in the case of the K = 3 ln(T + T

QCD axion) by strong dynamics, or by D-brane instanton K = 3 ln(T + T ) effects. The W = W effective theory has: resulting + AeaT .
0

he value of the ux superpotential at + Ae W = W0 the minimum for complex structure . exponential term in W arises from non-perturbative eects. The resultin alue of the ux superpotential at the minimum for complex structu l has an AdS minimum, which is supersymmetric. The F -terms vanish, and d onential term supersymmetric non-perturbative eects. W 0 is: e of the superpotential arises supersymmetric AdS vacuum bydepth(see eq There is a in W in the from AdS vacuum, whose The resu s an AdS minimum, which is supersymmetric. The F -terms vanish, an e potential has a depth of
aT 0

the superpotential in the supersymmetric AdS vacuum by W |VAdS | = 3eK | W 0 |2 . tential has a depth of

(see

then further incorporates some eects of supersymmetry breaking to lift K 2 |VAdS | = 3e | W 0 |are many ways that one can . occurring at m to a metastable de Sitter minimum. There Supersymmetry-breaking dynamics, ating supersymmetry low scales, these constructions; for a vacuum of in cancels this negative discussion exponentially breaking eects of supersymmetry breaking to li n further incorporates some ee the reviews [20].and yields a metastable false vacuum. The upshot in many cases is that one obtains a co energy a metastable de Sitter minimum. There are many ways thatsmall c one i C potential of the form V 2 , where C can be parametrically g supersymmetry breaking the correction is obviously a a discussion ck units. This additive form of in these constructions; for crude model o
Friday, May 21, 2010

years.

This yields

V
2 1

3m2 3/2

1 2 D + |F |2 . 2

100

150

200

250

Now let us discuss the height of the barrier V stabilizes dS state after the uplifting. Since the u is achieved by adding a slowly decreasing functio to a potential which rapidly approaches zero at the height of the barrier VB is approximately eq to a factor O(1)) to the depth of the AdS minimu see Fig. 1: VB |VAdS | m2 . 3/2

-1

-2

The supersymmetry breaking dynamics, in terms of this potential, can be thought of as adding an exponentially small term which depends on some inverse power of the volume (like all energies do, in 4D Einstein frame).
Since Di W = 0 in the AdS minimum, its depth is given by VAdS = 3eK |W |2 . (1)

FIG. 1: Thin green line corresponds to AdS stabilized potential for W0 = 104 , A = 1, a = 0.1. Dashed line shows C the additional term 2 , which appears either due to the contribution of a D3 brane or of a D7 brane. Thick black line C shows the resulting potential including the 2 correction with C = 2.61011 , which uplifts the AdS minimum to a dS minimum. All potentials are shown multiplied by 1015 .

To complete the list of important features model, let us remember what should be done t for the description of ination.

The simplest possibility would be to use the ex of the potential of the height VB as an initial p ination. A particular realization of this scena proposed in [17]. (In order to do it, it was nece consider a racetrack superpotential with two exp In this case one has an interesting relation betw ious parameters of our model and the Hubble c during ination: H 2 VB /3 |VAdS |/3 m2 . 3/2

The important point is that when the energy of the metastable SUSY-breaking can be treated as a small perturbation of the AdS vacuum, the barrier height preventing decay of the new de Sitter vacuum is:
Friday, May 21, 2010

Here all functions are calculated at = cr , where cr is the position of the minimum of the potential prior to the 2 uplifting. We use the units where MP = (8GN )1 = 1.

Before the uplifting, the potential has only one extremum, at = cr , and its absolute value exponentially C decreases at cr . When we add the term 2 , the minimum shifts upward in such a way that the new dS minimum is positioned at 0 cr . This means that the values of the function eK() |W ()|2 in the minimum of the eective potential remain almost unchanged during the uplifting. Meanwhile, the value of Di W () in the minimum after the uplifting is no longer equal to zero, but it still remains relatively small, Di W (0 ) W (0 ). At the dS minimum, the total eective potential must vanish, with the accuracy of 10120 . Therefore one has C V = 3eK |W |2 .

One may also achieve ination by considering ics of branes in the compactied space. This in second uplifting, which corresponds to a nearly d tionary) potential added to the KKLT potential for example in D3/D7 case [18]. The added p should be at in the inaton direction, and, acco [18], it has a 3 dependence on the volume mo
in Vtot VKKLT () +

V () . 3

Here is an inaton eld. The resulting potent function of is schematically shown in Fig. 2 for d values of the function V (). It is apparent fr gure that the vacuum stabilization is possible model only for suciently small values of the potential,
in

<

B . We assume that T modulus stabilization works in the same way nation, when vanishes. It follows from this that the nal VB is s VB |VAdS |.

theBut since the SUSY-breaking F-termconstant, the po present-day cosmological Thad to precisely2 K 2 T V = e (G |D W | + G |DT W |2 3|W | ) + (almost) cancel the AdS vacuum energy, we get a relation y vanish (giving rise to a vacuum energy dens We note that if the new terms due to the inaton in eq. (2.10) are between the barrier height and the gravitino mass! inimum. The can expect problem with decompacticatio barrier height VB , we gravitinoa mass is given by eq. ( Finally, let us add() inaton. The|chiral inas a power ofZ, our eK Gsome 2energy multiplet 1/ natonby the F -term of |D W vanishes the inaton candidates V oken 3 K
3 where c 3 for the original version of the KKLT model.1 KKLT minimum after the tunneling with bubble formation following some previous stage of ination, because such tunneling would create an open universe. After such tunneling, we will still need to have a long stage of ination, which should make the universe at, form the large scale structure of the observable part of the universe, and end by a slow roll to the KKLT minimum. Our results imply that the Hubble constant H at this last and most important stage of ination should be smaller than the present value of the gravitino mass.

Now, let us consider the eects of the inaton contributions t nation, when

G = 1). Hence the F -terms, which measu K/2 5 me order as e | W | = m , and so the grav
1

also depends is 1/ from the volume, and In other words, value of the power inversely on the prefactor e ). to prevent a distortion of the potential 2 that creates form to the D3 cont eectively an uplifting term, similar in functionalan2instability to ZZ K V = it has been |argued that|Wprevent the |F|2 te G |FZ we need: | , 3e to of [12]. Empirically, decompactify, V he potential and destroying the minimum in the volume modulus 2 K/2 2 K 2 es that |FZ | |e DZKW | = 3e | W 0 | (he 2 e |D W | O(10)VB . ZZ
4 3 2

III. PROBLEMS WITH SUSY BREAKING AND INFLATION IN THE SIMPLEST KKLT MODEL

100

150

200

250

Friday, May 21, 2010 KKLT model. The second one describes, e.g., the D3/D7 the

FIG. 2: The lowest curve with dS minimum is the one from

Now we are ready to formulate a list of unusual features of this scenario.

But that is precisely the energy density contained in the inaton eld! Therefore, V 2 Thus, as H = during ination,
3

V H = eK |D W |2 O(10)VB O(10)|VAdS |, (2.12) 3 where the last approximate equality is from eq. (2.7). So, under this set of assumptions, VB is related to both the gravitino mass, and the maximal possible scale of ination.
2

which is a relation between the Hubble scale of ination, It is now easy to formulate the KL problem. If we assume that the Linde to at its and the gravitino mass. This led Kallosh & eld remains minimum during ination, then the scale of ination is given by (using eq. (2.6)) postulate a2 bound on generic string theory ination: K 2 2
H O(10)VB O(10)|VAdS | e | W 0| m3/2 , (2.13) which is just eq. (1.3) from the introduction (with MP = 1). This equation leads to the H m3/2 statement in [10] that, due to the need to maintain stability of the volume modulus during ination, inationary models in string theory should generically be expected to satisfy

This bound is not related to ination itself, but rather to a Hination mtoday . (2.14) 3/2 desire to avoid decompactication during ination in This ties the scale of SUSY breaking to the scale of ination. For many high-scale inationary models this yields m3/2 1010 GeV in the simplest scenario of [12], many orders of magnitude theories with extra dimensions. Stringy and other extragreater than the 1 TeV value predicted by typical supersymmetric models. The KL problem dimensionaldicultcompletionsmodels introduce extra both a UV to nd ination can that can accommodate suggests that it may be potential future observationconstraints from ination. a light gravitino. of tensor modes on ination and
Friday, May 21, 2010

Kallosh and Linde did propose a way to circumvent this problem. They noted that

IV. Evading the KL bound

While the logic that has been described is clearly correct given the assumptions, there is an obvious assumption that one can relax to evade this bound. Our goal will be to exhibit models that have all of the same generic SUGRA structure that was assumed in deriving the bound, but which give
H >> m3/2

Friday, May 21, 2010

Issues I will sweep under the rug: * The problem is one which is evident in a wide class of string-inspired supergravities. We will work in that framework without deriving the model from string theory. * The problem as stated has nothing to do with observational facts about the precise model of ination which governed our world. We will try to exhibit easy violations of the bound without requiring that the ination model be realistic in its observables.
I believe that because the essence of the problem is as I described it, omitting realism of the inationary perturbation spectrum & a string derivation is actually ne. Our mechanism could presumably be generalized to realistic models; we worked instead in the most obvious setting that could display gross violations of the proposed bound, with models of largeeld chaotic ination.

Friday, May 21, 2010

The clear strategy to pursue is the following: * The depth of the AdS vacuum that is being lifted by supersymmetry breaking & inationary energy density, determines the barrier height. * For models where SUSY is relevant to the hierarchy problem, this gives very low barrier height, because the depth of the AdS vacuum (the expectation value of W) is xed by the gravitino mass. This is easily xed. Intuitively, the ux-induced parameter W0 is setting the scales. But who says one cannot generalize this structure, so that one has:
Friday, May 21, 2010

W = W0 ( ) + AeaT
varies adiabatically as inaton rolls

For instance, one can design a model where:

1 0.001 10
6

W
10 10 10
9

12

15

10

20

50

Figure 2: The vev of the superpotential | W | = | W (, X(), T ()) | plotted as a function of the inaton with X and T adiabatically tracking their instantaneous minima. exit . To prevent this from happening, F in eq. (3.29) needs to satisfy eq. (3.27) for all . This is easy to check; the function xn1 /(xn + c) has one global maximum for x > c > 0. Thus, we need |F | O(1). (3.30) max

Friday, May 21, 2010

With this obvious idea in mind, lets try to build a simple model. Well see that simple models exist, but there are generic supergravity issues that arise with large-eld ination which make them less simple than you might have thought. For instance, ignoring the additional volume modulus, we note that a large-eld model in the spirit of Lindes chaotic ination, involving just the inaton chiral multiplet, is very hard to get in N=1 supergravity. This is because with a polynomial superpotential (and a shift-symmetric K), at large elds
>> MP
Friday, May 21, 2010

|W | |D W | << 2 MP
2

They include one additional eld X beyond the minimal content one might expect (th modulus and the inaton = + i) in any discussion of the KL problem. This additio eld X is needed to avoid very general constraints on large-eld ination in supergrav discussed in the insightful paper of Kawasaki, Yamaguchi and Yanagida [22]. The same allows us to overcome other detailed problems with keeping the T modulus stable dur ination, which would also pose obstacles in a large-eld model with only T and the ina eld . We explain these general constraints in detail in section 3.2, using our toy mode section 3.1 as an illustration. In section 3.3, we then scan over the range of parameters t Kawasaki, Yamaguchi, are accessible in this class of models, exhibiting many models that have H Yanagida . m3/2

So, the supergravity potential slopes downward at large elds, and goes negative. If we want to avoid futzing around with the eld range (which is against the spirit of large-eld ination), we should probably include a second eld.

A simple class of large-eld models with a 2-eld A toy model inationarywriting down the Khler potential and superpotential of ourand model. sector, whicha circumvents this problem toy also We begin by before, we allows = 1. take MP large violations of the KL bound, is:
3.1 1 K = ( + )2 + X X (X X)2 3 log(T + T ) 2 Any function of real part
is ne here

O(1), sign important

W = W0 g(X) + f (X) n + eaT with : g(X) = 1 + O(X) and f (X) = b + X + O(X 2 )

generic functions, we just wrote out the terms that matter
Friday, May 21, 2010

(3.

Here, = + i is the inaton, X is a chiral multiplet, and T is the modulus eld. T

Let us explain how and why anyone would write down such an ugly model.

1. In the nal vacuum when X and the inaton vanish, this reduces to the low-energy effective theory that governs many of the string models of moduli stabilization. 2. In the Kahler potential, we have assumed that there is a shift symmetry for :
= + i

It is broken by alpha, which can naturally be very small.

Friday, May 21, 2010

lude one additional eld X beyond the minimal content one might expect (the T and the inaton = + i) in any discussion of the KL problem. This additional s needed to avoid very general constraints on large-eld ination in supergravity, in the insightful paper of Kawasaki, Yamaguchi and Yanagida [22]. The same eld to overcome other detailed problems with keeping the T modulus stable during which would also pose obstacles in a large-eld model with only T and the inaton We explain these general constraints in detail in section 3.2, using our toy model of 1 as an illustration. In section 3.3, we then scan over the range of parameters that sible in this class of models, exhibiting many models that have H m3/2 .

3. The superpotential can be natural if there is an approximate R-symmetry under which X is neutral and the inaton carries charge 2/n. The unknown functions of X can be arbitrary, only the terms we explicitly wrote out ever matter in the dynamics. Why do the single correction in K and the few terms in f(X) and g(X) that we wrote out, matter?
1 K = ( + )2 + X X (X X)2 3 log(T + T ) 2 W = W0 g(X) + f (X) n + eaT with : g(X) = 1 + O(X) and f (X) = b + X + O(X 2 ) (3.15)

toy model

by writing down the Khler potential and superpotential of our toy model. As a e take MP = 1.

=Friday,+ i 2010 the inaton, X is a chiral multiplet, and T is the modulus eld. The May 21, is

1. The constant in g is necessary to get the correct structure after ination. 2. The linear term in f guarantees that the X F-term dominates the energy during ination. This avoids the problem that we noted with large-eld ination in supergravity (negativity of the potential at large elds!). 3. The constant term in f guarantees that at large elds, the value of the effective W0 seen by the T-modulus does indeed vary adiabatically with the inaton. 4. Given the large F-term for X during ination, the quartic correction in K keeps X stabilized at the origin.
Friday, May 21, 2010

eep the rst term in the Taylor expansion of g(X) and the rst two These models are close of f (X). Inclusion of further terms (with generic coecients) would to being ruled out for n>2... 2n V in f are sions. The reason the rst two terms () relevant will become ion 3.2.

So, with reasonable choices for all of the parameters, lly nd that Xclass both during and afterchaotic in our models, 1 of models gives ination ination with this

e modulus T during ination

V. Dynamics of these models

itulate the dynamics of uplifting the AdS vacuum for T as in [12], to A. Dynamics of the volume modulus T during ination constraint eq. (3.20) is sucient to guarantee the continued existence m for T . We start by choosing the parameters of the setup such that The dominant energy source is for T at large occurs at = Re T 1. Within that regime, the tionary uplifting comes from |FX |2 = eK |DX W |2 1/ 3 acting like a to the T -dynamics,5 shows us empirically that the actual T -minimum
0

This is one of the canonical forms assumed also for the terms () T0 : DT W ()|T = 0 Tmin induced by SUSY-breaking starting in the nal AdS (3.23) vacuum. So the dynamics of T a indeed inection m of T is not very close to disappearing into is barrier-lessto adiabatically

to estimate the size of the terms in the superpotential at this eventual Friday, May 21, 2010

fact that fact that the main inationary uplifting comes from |FX | e e |DWW | 1/ acting like aa the main inationary uplifting comes from |FX | = = |DX X | 1/ acting like D3-braneD3-brane with respect T -dynamics,5 shows us empirically that the actual T -minimum with respect to the T the T -dynamics,5 close us empirically that the actual T a barrier-less ine ong as this minimum of to is not very shows to disappearing into -minimum producedproduced sits at sits at nt. We can use this to estimate min ()size0 ofDT W ()|T = 0in the superpotential at this even the T the : TD : W ()| terms (3.23) Tmin () T0 =0 0 (3.23) T T0 point. We can use for T to estimate the size of the terms ()the0superpotential at this eventual minimum this . Given that as argued above DT W in at 1 = Tmin () we have that T aTmin () DT T Given that as 0 e minimum forW.() argued above DT W () 0 at2T = Tmin ()W0,ef f. () we have that
DT W () 0

eventual minimum for T atTlarge occurs at = Re T 1. Within that regime Dynamics of the volume modulus during ination Dynamics of the volume modulus T during ination that the main inationary uplifting comes from |FX |2 = eK |DX W |2 1/ 3 acting l We pause here to recapitulate the dynamics of uplifting the AdS vacuum for T as in [12], to We pause here track an instantaneous vacuum: brane with respect to recapitulate the dynamics of uplifting the AdS vacuum for T asthe actual T -minim to the T -dynamics,5 shows us empirically that in [12], to see why satisfying the constraint eq. (3.20) is sucient to guarantee the continued existence see why satisfying the constraint eq. (3.20) is sucient to guarantee the continued existence of an sits at uplifted minimum We . We by choosing the parameters of the setup such that duced uplifted minimum for T . for T start start by choosing the parameters ofthe setup such that of an the eventual eventual minimum atT T at() T at: at Re T()| 1.= 0 that regime, the occurs = W the minimum for T forlarge large occurs D = Re T 1. Within that regime, the Within ( min 0 T 2 2 K KT0 2 2 33

We can check validity of SUGRA and the single instanton imum for T . Given minimum ofarguedvery close DT W () intoatbarrier-less inection have tha as long as this that as T is not above to disappearing 0 a T = Tmin () we approximation:superpotential at this eventual as long as this minimum of T is not very close to disappearing into a barrier-less inection point. We can use this to estimate the size of the terms in the
DT W () 0
e
aTmin ()

eaTmin ()

s demanding

Thus demanding

where : W0,ef f. () W0 + (b + X)n

where: : W0,ef f. ()f. () + (b +0X)(b + X) . W0,ef W0 W + n . (3.24) where

.
60

13 aTmin () W0,ef f. () 1+ 2 W0,ef f. () 2 1 + 3 aTmin ()

1 + 3 aTmin () 1

(3.24) n

Thus demanding

at Re T does a these conditions,that theofinstantaneous minimum will 1 and guarantee at all.existence 1 if Under 1 if it occurs the all. ThisaguaranteesT willvaliditysupergravity minimum for the be determined by the T 1 and a it occurs2 at This guarantees the validity of both the of both the supergr approximation and the one-instantonwhich is trying to that we use. close to the location of the competition between |DT W | , approximation relax to zero roximation AdS conditions, the X |2 = eK |DX Wa|2 minimum for that we positive contribution old and the one-instanton approximation is adding a use. minimum, and |F existence of 1/ 3 , which T will be determined by the Under these
Friday, May 21, 2010

and Re T 1 and a 1 if it occurs at all. This guarantees the validity of both the supergravity and 1 together are the one-instanton approximation that we uplifted -minimum approximation andsucient to ensure that the use. a < a1<together are sucient to ensure that the uplifted T -minimumToccurs at = occurs

|W0,ef f. ()| |W0 +0(b to X)nX)n | || = |ImT| <= 60 occurs< (3.25) 1 -minimum a < 10,ef f. ()| are sucient (b + | that the uplifted || |Im | at60 = |W together |W + + ensure 1

|W0,ef f. ()| |W0 + (b + X)n |

So demanding= |Im | < 1 ||

(3.25)

eld remains stabilized. Naturally, not all choices of parameter sets taneous)where the for T , and we must deriveinstanton action is occur minimum volume is large and the conditions on the a W0 . The constraints on these parameters arise from arguments give healthy. The F -terms in V act eectively as an uplifting term. In orde cation, we must have, by equations (2.6), (2.7), and (2.11) with th Then, as long as one satises an avoidance of m, overuplifting with the given instantaneous effective value |F |2 + |FXnice O(10)3eK | for |2. This gives |2 minimum W T. of W0 , there will be a

the constraint: ed the FT term as it is dominated by either F or FX . Rewritin we have 2 2 |F | + |FX | O(1), 3eK/2 | W | eplaced O(10) with O(1) to be conservative in our estimates. We is generic relation to our specic modelfor the instantaneous o which is indeed just the KL bound will produce a constraint value of W.
8
Friday, May 21, 2010

Avoiding decompactication during ination We will now evaluate the constraint eq. (3.20), which must be satised to avoid decompactication. For this purpose, let us rst focus on the region where is large. Note that we want to produce a hierarchy in W , so we want W0 to be many orders of magnitude smaller than the polynomial eld terms (at least until the end of ination, when the polynomial terms disappear). Combined with the fact |X| 1, this allows us to approximate W bn . Furthermore, since FX has the highest power of among the F -terms (FX eK/2 n ) FX dominates among the F -terms for large . Hence, eq. (3.20) becomes,

B. Decompactication constraints

i. At large inaton vevs

When the inaton vev is large, we can approximate:

2 2 |F | + |FX | 1 |FX | eK/2 n . K/2 | W | K/2 | W | K/2 bn 3e 3e 3e 3b

(3.26)

b was just some parameter of O(1) in the theory. This shows behaviorthere for an exemplary choice ofat large inelds. that in Fig. 1 is no problem parameters eq. (3.15) We depict this If you do not trust me, here is a plot:
given by: A = 1, a = 2 , W0 = 1015 , = 5 1019 , b = 2/5, n = 10, and = 2. This 10 choice of parameters gives an eective inationary potential V () 20 for 60 50MP with the choice of giving us / 1.6 105 at 60 . Here we have approximated the functions f, g in eq. (3.15) by f (X) = b + X + X 2 /2 and g(X) = 1 + X for deniteness, 3.0 to check explicitly that the higher-order terms do not spoil the behaviour of the model, as 2.5 FX 2 expected from the smallness of X during ination. Fig. 2 shows us | W (, X(), T ()) | as W 2 2.0 a function of the ination , where X(), T () denote the elds X, T adiabatically tracking 1.5 their instantaneous minima at every given value of . Now, let us examine the region when is 20 small,30i.e. sub-Planckian. In this region, the 0 10 40 50 F term dominates as it has one smaller power of than FX . This means in the small
Friday, May 21, 2010

Therefore, at large , the ratio between the F -terms and W is constant. Thus, as long as we pick an O(1) value for b such that the ratio is O(1), there is no danger of decompactication for large .

with the choiceF giving us / of 2 1.6 105 at 60 . Here we have approximated the 6 functions f, g in eq. (3.15) by f (X) = b + X + X 2 /2 and g(X) = 1 + X for deniteness, W that the vevs toii. At explicitly 2 check small inaton higher-order terms do not spoil the behaviour of the model, as 4 expected from the smallness of X during ination. Fig. 2 shows us | W (, X(), T ()) | as a function of the ination , where X(), T () denote the elds X, T adiabatically tracking 2 their instantaneous minima at every given value of .
X

limit eq. (3.20) can be written as

In this region, instead, the F-term of the inaton itself dominates. The region when is small,avoid decompactication Now, let us examine the constraint to i.e. sub-Planckian. In this region, the 0 F term dominates as it has one 10 becomes: 30 F . This means in50 small smaller power of than the 0 20 40

|F | functions O(1). (3.27) igure 1: The |F/W |2-ratios plotted as K/2 | W | of the inaton with X and T adiabatical 3e F . Hence, we are allowed to apply the global SUSY approximation, so

acking their instantaneousignore the Khler covariantization of the derivative in evaluating Because vanishes, we can minima. a

the K/2 (due n1 the (X X)2 term We Ondropped the Xn1K/2 evaluating n1X functional forms gives: in th also the other hand, in n1 term (3.28) since 0 to F e (nb + nX ) e nb . (3.28) hler potential). It follows that a
11 nbn1 1 |F | 3eK/2 | W | 3(bn + W0 )

(3.2

We have again dropped the exponential in T from W since after ination ends (and hen so So thereend of ination), the non-perturbative term in the superpotential exp(aT close to the is indeed a danger of losing the minimum during the minimum for Tthe exit from large-eld by a factor 1/(aT ) is again smaller than W0,ef f. () ination. 1 (and an be neglected when evaluating the ratio eq. (3.29)). Furthermore, we again dropped th Friday, May 21, for n term2010 the same reason as above ( X 0 during ination).

To avoid disaster, the function of the inaton that appears there needs to satisfy: Calculating the maximum of eq. (3.29), our constraint becomes
|F | max 3eK/2 | W | n1 = 3 b W0 (n 1)
1/n

This is easy to check; the function xn1 /(xn + c) has one global maximum for x > c > 0. Thus, we need |F | O(1). (3.30) max K/2 | W | 3e

O(1).

(3.31)

If we take b O(1), then we need W0 . This can be satised in principle, even with the correct size of inaton-generated density perturbations at the 60th e-folding, by choosing n > 1 suciently large. We will now make this statement precise by discussing the hierarchies and perturbations that result for various values of n and W0 . W0 .

This requires that alpha should not be much larger than

3.2 is theoretically F -terms and a no-go result This The structure of the ne, but will lead to unrealistic models.

Since this is a constraint arising from the behavior at small Firstly, note that eq. (3.29) is universal for all ination models in which we would elds, while our ination occurs at large elds,F is the dominant contribution to V at small values of , i.e. when the inatons own F -term dominates at guessThis statement holds regardless of whether the model is small-eld hybrid that by slightly complicating the exit (as in or large-eld, small . ination), one could only at completely realistic theories. since the analysis uses eq. (3.29) make small close to the post-inationary minimum,
Friday, May 21, 2010

By inspecting the properties of the ratio eq. (3.29) we are led towards two related observations.

2]. Our setup here shares the property with their model that ination ination). One way to satisfy these requirements is to have a linear function of X multiplyi ad of F , with |FX /eK/2 Winside const. led 3 at large .eq. (3.15). it ge the inationary polynomial in | W , which to the choice of However, VI. the superpotential used are similar to the ones which led of coupling pling in that these latter considerations by thesehierarchies kind to the rst na Note Horse-trading : achievable authors; this vs realistic ural realization allm2 2 chaotic ination in 4D supergravity by yielding a model w W 0 at of (due to theination vev), thus Kawasaki, Yamaguchi small X Yanagida [22]. Our setup here shares the property with their model that ination is driv would decompactify. K/2

whether we can match data while getting parametrically l, V is dominated at large values by the FX term, as it has the larg large hierarchies. At required inaton power n and 3.3 Horse trading: Achievable hierarchies vslarge elds:
FX is the dominant term driving ination, and we may approximate

3 at large . However, it generaliz by FX instead of F , with |FX /e W | const. theSo, nally, let us see how used by these authors; this by computing its X coupling in the superpotential these models fare kind of coupling by would density 0 at all hierarchies vs spectral index,a model where ta trading: the give WAchievable (due to and theXrequiredyielding and seeing perturbations the small vev), thus inaton power n modulus T would decompactify.

In our model, V is dominated at large values by the FX term, as it has the largest pow V |Fdriving ination, and we may approximate |2 2 2n of . Hence, FX is the dominant term X

Now, the magnitude of the

and the density uctuations are given by:

1 V = V 2 =1501
150 2

2 V |FX |perturbation 2 2n density

(3.3 at 60 e-folds is given

for large . Now, the magnitude of the density perturbation at 60 e-folds is given by

,
=60
=60

(3.3

1 V 2 is 1 V where = 2 the 2 V V
Friday, May 21, 2010

slow is the

roll parameter and 60 is the value of of at 6 slow roll parameter and 60 is the value at 60 e-fold

W
10
9

In reality this10 For this reason, welets pretend its a free and is measured, but leave as a free parameter nting the KL problem. of n in the inaton potential V () 2n necessary t Figure 3: Minimum value parameterour 10 at xed is = 1015betweenthe[2] and in our achieve a given and derive W0 easy to calculate 60observables. case no-decompactication es it can take in /, model. It relations , satisfying
12 15

Min n but to demonstrate a mecha alistic model compatible with experiment,

2 5 10 50 constraint eq. (3.20). Points1 are labelled in the format (n,20 O(|Wi |/|W0|)), where Wi = W (60 , X(60 ), (60 is the the 2 60(n at initial inaton (and O(|x|) here denotes the order ComputingT the))vev of 60 = superpotential1).the 60th e-folding and o magnitude of 2: The vev of the superpotential | W | = | W (, X(), T ()) | plotted as a function Figure |x|).

pluggingand can solve for as their instantaneousminima.n, givin in, we can nd the function of and relation: aboveof the inaton with X we T adiabatically tracking a three equations,

a fully realistic model compatible with experiment, but to demonstrate a mechanism fo exit . To prevent this from For this 10 3n leave as to satisfy eq. (3.27) for explor circumventing the KL problem. happening, F in eq. (3.29) needs a free parameter and all . reason, = n1 n+1we , This it can take in the function x /(xn to has one 60 [2] and in for case what values is easy to check;our model. It is easy + c)calculateglobal maximumour x > c > 0. 60
Thus, we need 3eK/2 | W |

where 60 is given by eq. Substituting took O(1) take2b forO(1),(3.34).we need purposes.expression into eq. (3.31), 3/2even O(1) = calculational W0this Now, in order for m we havewith If we then . This can be satised in principle, one correct size of as a function of n 1/n the density perturbations = 1015theSolving forinaton-generated103bn yields at 60th e-folding, by choosing . nds:
Friday, May 21, 2010

small elds: 1/n Using the above three equations, we can solve for as a function of and n, giving us Calculating the maximum of eq. (3.29), our constraint becomes 10 3bn 10 3n (n 1) 2 1/n3, n+1 = (3.35 |F | 1 60 W0 (nn 1) , b n+1 max O(1). (3.31) = 60 K/2 W (n 1)
3e |W | 3
0

60 60(n (3.34 decompactication is Pluggingeq. (3.34). Substituting |F| expression into eq. (3.31), we ha given by this into the max = 2 this 1). O(1). constraint at (3.30)

n > 1 suciently large. We will now make this statement 2 3,by discussing the hierarchies precise (n 1) n+1 (3.36 n (n and perturbations that result for 60 W0values1) n and W0 . various n+1 of 2 3
n1

60 W0 (n 1)

i.e.

2 3 n1

n+1 W0 (n 1) 60 . 10 3bn 15

(3.37)

Finally, putting in the value of W0 that is consistent with intermediate scale SUSY-breaking, we get:
0.01
10, 1013 8, 1010

10

10
8

5, 107

3, 105

10

11

2, 103

1.0

1.5

2.0

3.0

5.0

7.0

10.0

Min n
Figure 3: Minimum value of n in the inaton potential V () 2n necessary to achieve a given /, at xed W0 = 1015 , satisfying the no-decompacticationconstraint eq. (3.20). Points are labelled in the format (n, O(|Wi |/|W0|)), where Wi = W (60 , X(60 ), T (60 )) is the initial superpotential (and O(|x|) here denotes the order of magnitude of |x|).
Friday, May 21, 2010

If insteadthis sectiontheexploring theperturbations0 to come force = we x by density relation between W and n if we out We conclude 2 105 (which is of interest as it is the observed value!). We rearrange eq. (3.35) to get correctly, we obtain a relation between W0 and n:
W0 10 3nb
2 3 n1 n

out that to satisfy observational data on , we require n 7.

n+1 (n 1) 60

(3.38)

This equation gives us a lower limit for W0 and hence gives us a lower limit for m3/2 after ination, and we have plotted the results in Fig. 4. Observational constraints on the spectral 10 8 index and the tensor to scalar ratio (from e.g. [26]) require 2n 4. Therefore, m3/2 cannot 2,10
3

10

11

3,105

16

W0
10 10

5,107

14

7,1010
17

10,1013

1.0

1.5

2.0

3.0

5.0

7.0

10.0

Min n
Figure 4: Minimum value of n in the inaton potential V () 2n necessary to achieve a given post-inationary vacuum VEV of the superpotential W0 , at xed / = 2 105 , satisfying the no-decompactication-constraint eq. (3.20). Points are labelled in the format (n, O(|Wi |/|W0|)), where Wi = W (60 , X(60 ), T (60 )) is the initial superpotential (and O(|x|) here denotes the order of magnitude of |x|).
Friday, May 21, 2010

This equation gives us the minimum value of n needed to realize a given for a xed value

Since observation requires n < 2, this model cannot be fully realistic. Even with n < 2, it does have
H >> m3/2

but would certainly not have low enough supersymmetry breaking scale to explain the gauge hierarchy. Of course, our goal here was NOT to write a fully realistic theory, but to illustrate that the tension between high-scale ination and low-energy supersymmetry in string theory is not at all a generic feature. Our analysis was conned to supergravities containing the basic ingredients used in proving the KL bound.
Friday, May 21, 2010

The obvious things to do, to make the case more convincing, would be: * Write down models which are fully realistic, at the level of supergravity with a T-modulus. This can presumably be done by slightly complicating the exit from ination as in hybrid models, since our main constraints (that led to conict with data) actually came from the small-eld region. * Embed high-scale ination with low-energy supersymmetry (with or without a realistic set of inationary observables) in full string compactications.
Friday, May 21, 2010