Eternal Inflation and Sinks in the Landscape Andrei Linde Andrei Linde

New Inflation V

Chaotic Inflation Eternal Inflation

Hybrid Inflation Hybrid Inflation

Inflation in string theory KKLMMT brane-anti-brane inflation Racetrack modular inflation D3/D7 brane inflation DBI inflation (non-minimal kinetic terms)

Initial conditions for inflation: In the simplest chaotic inflation model, eternal inflation begins at the Planck density, if the potential energy is greater than the kinetic and gradient energy in a smallest possible domain of a Planckian size. In the models where inflation is possible only at a small energy density (new inflation, hybrid inflation) the probability of inflation is not suppressed if the universe is flat or open but compact, e.g. like a torus. A.L Zeldovich and Starobinsky 1984; A.L. 2004

Recent probability measure proposed by Gibbons and Turok disfavors the probability of inflation. It requires that at the Planck time the potential energy density must be 12 orders of magnitude smaller than the kinetic energy. I am unaware of any physical mechanism that would enforce this requirement. A general problem with any measure on a set of classical trajectories is that it cannot describe the transition from the initial quantum regime at super-Planckian densities, so it can describe anything but… initial conditions. In the string landscape scenario, with many dS vacua stabilized by KKLT mechanism, one either directly rolls down to the state where life of our type is impossible (AdS, 10D Minkowski space), or enters the state of eternal inflation. This essentially eliminates the problem of initial conditions, replacing it by the problem of the probability measure for eternal inflation.

Let flowers blossom Let flowers blossom  < 0  = 0  > 0

Predictions of Inflation: 1) The universe should be homogeneous, isotropic and flat,  = 1 + O(10 -4 ) [    Observations: the universe is homogeneous, isotropic and flat,  = 1 + O(10 -2 ) 2)Inflationary perturbations should be gaussian and adiabatic, with flat spectrum, n s = 1+ O(10 -1 ) Observations: perturbations are gaussian and adiabatic, with flat spectrum, n s = 1 + O(10 -2 )

Any alternatives? 1) Pre-big bang and ekpyrotic/cyclic scenario 1) Pre-big bang and ekpyrotic/cyclic scenario Lots of problems. It is difficult to say what these models actually predict since they involve a transition through the singularity. Lots of problems. It is difficult to say what these models actually predict since they involve a transition through the singularity. 2) String gas cosmology 2) String gas cosmology Does not solve flatness problem, but was claimed to produce perturbations with n = 1. Does not solve flatness problem, but was claimed to produce perturbations with n = 1. Nayeri, Vafa, Brandenberger, Patil, A more detailed investigation demonstrated that the perturbations produced in this model have spectrum with n = 5, which is ruled out by observations. A more detailed investigation demonstrated that the perturbations produced in this model have spectrum with n = 5, which is ruled out by observations. Kaloper, Kofman, A.L. and Mukhanov, Kaloper, Kofman, A.L. and Mukhanov, hep-th/ Brandenberger et al, Brandenberger et al, hep-th/

Inflationary Multiverse For a long time, people believed in the cosmological principle, which asserted that the universe is everywhere the same. This principle is no longer required. Inflationary universe may consist of many parts with different properties depending on the local values of the scalar fields, compactifications, etc.

Example: SUSY landscape V SU(5)SU(3)xSU(2)xU(1)SU(4)xU(1) Weinberg 1982: Supersymmetry forbids tunneling from SU(5) to SU(3)xSU(2)XU(1). This implied that we cannot break SU(5) symmetry. A.L. 1983: Inflation solves this problem. Inflationary fluctuations bring us to each of the three minima. Inflation make each of the parts of the universe exponentially big. We can live only in the SU(3)xSU(2)xU(1) minimum. Supersymmetric SU(5)

Inflation and Cosmological Constant 1) Anthropic solutions of the CC problem using inflation and fluxes of antisymmetric tensor fields (A.L. 1984), multiplicity of KK vacua (Sakharov 1984), and slowly evolving scalar field (Banks 1984, A.L. 1986). All of us considered it obvious that we cannot live in the universe with 2) Derivation of the anthropic constraint Weinberg 1987; Martel, Shapiro, Weinberg 1997; Vilenkin, Garriga 1999, 2001, 2005 Three crucial steps in finding the anthropic solution of the CC problem:

Inflation and Cosmological Constant 3) String theory landscape Multiplicity of (unstable) vacua: Lerche, Lust and Schellekens 1987: vacuum states Bousso, Polchinski 2000; Feng, March-Russell, Sethi, Wilczek 2000 Vacuum stabilization: KKLT 2003, Susskind 2003, Douglas 2003,… perhaps metastable dS vacuum states - still counting…

Latest anthropic constraints on  Aguirre, Rees, Tegmark, and Wilczek, astro-ph/ observed value

Dark Energy (Cosmological Constant) is about 73% of the cosmic pie What’s about Dark Matter, another 23% of the pie? Why there is 5 times more dark matter than ordinary matter?

Example: Dark matter in the axion field Standard lore: If the axion mass is smaller than eV, the amount of dark matter in the axion field contradicts observations, for a typical initial value of the axion field. Anthropic argument: Due to inflationary fluctuations, the amount of the axion dark matter is a CONTINUOUS RANDOM PARAMETER. We can live only in those parts of the universe where the initial value of the axion field was sufficiently small ( A.L. 1988). Recently this possibility was analyzed by Aguirre, Rees, Tegmark, and Wilczek.

Latest anthropic constraints on Dark Matter  Aguirre, Rees, Tegmark, and Wilczek, astro-ph/ The situation with Dark Matter is even better than with the CC ! The situation with Dark Matter is even better than with the CC ! observed value

Two types of questions: Two types of questions: What is a typical history of a given point in an eternally inflating universe? (That is what we studied so far.) What is a typical population ? (Volume- weighted distributions.) Let us apply the same logic to inflation

Example: Example: 1) What is a typical age when any particular scientist makes his best discoveries? 2) What is the average age of scientists making greatest discoveries? Both questions make sense, but the answers are significantly different because of the exponentially growing population of scientists.

Chaotic inflation requires  > M p. This condition can be easily satisfied in the scalar field theory plus gravity, but it is difficult to construct inflationary models with  > M p in supergravity and string theory. It took almost 20 years to construct a natural version of chaotic inflation in SUGRA. In string theory we started doing it only 3 years ago, with limited success. Let us imagine that just one out of of string theory vacua allows existence of the field  ~ 100, in units of M p. The volume produced during chaotic inflation in this universe is proportional to e   ~ Therefore a part of the universe in such an improbable and “unnatural” state will have more volume (and contain more observers like us) than all other string theory vacua combined. HOWEVER…HOWEVER… What will change if we study physical volume?

Consider a more complicated potential : Naively, one could expect that each coefficient in this sum is O(1). However, |V 0 | < , otherwise we would not be around. A.L. 1984, S. Weinberg 1987 For a quadratic potential, one should have m ~ to account for the smallness of the amplitude of density perturbations  H ~ Is there any reason for all other parameters to be small? Specifically, we must have

A simple argument: Suppose that the upper bound on the inflaton field is given by the condition that the potential energy is smaller than Planckian, V < 1. In addition, the effective gravitational constant should not blow up. In this case In these models the total growth of volume of the universe during inflation (ignoring eternal inflation, which will not affect the final conclusion) is For a purely quadratic model, the volume is proportional to But for the theory the volume is much smaller:

But in this case at the end of inflation, when one has The greatest growth by a factor of occurs for Thus, if we can have a choice of inflationary parameters (e.g. in string landscape scenario), then the simplest chaotic inflation scenario is the best. (See also the talk by Kachru.)

This may explain why chaotic inflation is so simple: A power-law fine-tuning of the parameters gives us an exponential growth of volume, which is maximal for a purely quadratic potential

Probabilities in the Landscape We must find all possible vacua (statistics), and all possible continuous parameters (out-of- equilibrium cosmological dynamics). Douglas 2003 Douglas 2003 We must also find a way to compare the probability to live in each of these states. A.Linde, D. Linde, Mezhlumian, Bellido 1994; Vilenkin 1995; Garriga, Schwarz-Perlov, Vilenkin, Winitzki, 2005

Example: Two dS vacua is dS entropy This is the square of the Hartle-Hawking wave function, which tells that the fraction of the comoving volume of the universe with the cosmological constant V i =  is proportional to P0P0P0P0 P1P1P1P1 P i is the probability to find a given point in the vacuum dS i In this context the HH wave function describes the ground state of the universe, it has no relation to creation of the universe “from nothing”, and it does not require any modifications recently discussed in the literature.

This does not mean that the probability to live in the state with a cosmological constant  is given by. Life appears soon after the new bubble formation, which is proportional to the probability current. And the currents in both directions are equal to each other, independently of the vacuum energy: P0P0P0P0 P1P1P1P1 P i is the probability to find a given point in the vacuum dS i That is why one may expect a flat prior probability distribution to live in the bubbles with different cosmological constants. But in string theory all vacua are metastable. What are the implications of this metastability?

Metastable vacua in string theory (string theory landscape) Based on the KKLT scenario AdS minimum Metastable dS minimum Kachru, Kallosh, A.L., Trivedi ) 1) Start with a theory with the runaway potential 2) 2) Bend this potential down due to (nonperturbative) quantum effects 3) 3) Uplift the minimum to the state with positive vacuum energy by adding a positive energy of an anti-D3 brane in warped Calabi-Yau space

KKLT potential always has a Minkowski minimum (Dine-Seiberg vacuum) KKLT potential always has a Minkowski minimum (Dine-Seiberg vacuum) Now we will study decay from dS to the collapsing vacua with negative vacuum energy. Probability of tunneling from dS to Minkowski space typically is somewhat greater than e -S ~ e - 1/  ~ e e -S ~ e - 1/  ~ e After the tunneling, the field does not jump back - no recycling.

  Tunneling from an uplifted supersymmetric AdS vacuum

Tunneling to the sink Tunneling to the sink For the tunneling from an uplifted AdS vacuum, the final result depends not on V dS but on |V AdS | prior to the uplifting. In the class of the KKLT models that we explored, |V AdS | is related to SUSY breaking (to the square of the gravitino mass) after the AdS uplifting, and is of the order In the simplest SUSY models, the rate of a decay to a sink is e times greater than the probability to jump from our vacuum to a higher dS vacuum. Ceresole, Dall’Agata, Giryavets, Kallosh, A.L., hep-th/

Two dS vacua and AdS sink Parts of dS space tunneling to space with negative V rapidly collapse and drop out of equilibrium (one-way road to hell). Therefore instead of the detailed balance equations, one has flow equations: Ceresole, Dall’Agata, Giryavets, Kallosh, A.L., hep-th/ P0P0P0P0 P1P1P1P1

Narrow and Wide Sinks Narrow and Wide Sinks If the decay to the sink is slower than the decay of the upper dS vacuum to the lower dS vacuum, then the probability distribution is given by the Hartle-Hawking expression, despite the vacuum instability and the general probability flow down. In this case the probability currents between different dS states remain equal, which suggests flat prior probability distribution for the cosmological constant. On the other hand, if the decay to the sink is very fast, one will have an inverted probability distribution, and the CC will take its smallest value.

In the string landscape scenario we do not study the ground state of the universe, as we did before. Instead of that, we study the universe with many holes in the ground. Conclusion: Conclusion: Instead of studying static probabilities, like in a pond with still water, we study probability currents, like in a river dividing into many streams. In other words, in addition to exploring vacuum statistics, we also explore vacuum dynamics, including irreversible vacuum decay and the growth of the volume of different parts of the universe.

The main conclusion: It is unusual, it is complicated, it has many unsolved problems, one may like it or hate it, but we must learn how to live in a world where many different possibilities are available IT IS SCIENCE