Download Effective Theories and Modified Gravity

Document related concepts
no text concepts found
Transcript
Effective Theories and
Modified Gravity
A Bad Cop wrapped in a Wet Blanket
standing in a Cold Shower
Cliff Burgess
Outline
• Effective Field Theory and Gravity
• EFT and theoretical error
• Decoupling
• EFT in time-dependent settings
• Decoupling and when it fails
• Gravitational exceptionalism
• Learning from UV completions
• The case for modifying gravity
• What microphysics can teach us
Ole Miss 2014
Outline
• Effective Field Theory and Gravity
• EFT and theoretical error
• Decoupling
• EFT in time-dependent settings
• Decoupling and when it fails
• Gravitational exceptionalism
• Learning from UV completions
• The case for modifying gravity
• What microphysics can teach us
Ole Miss 2014
Outline
• Effective Field Theory and Gravity
• EFT and theoretical error
• Decoupling
• EFT in time-dependent settings
• Decoupling and when it fails
• Gravitational exceptionalism
• Learning from UV completions
• The case for modifying gravity
• What microphysics can teach us
Ole Miss 2014
Part I
Ole Miss 2014
EFT and theoretical error
Decoupling
Ole Miss 2014
EFT and theoretical error
Quantum field theory is a precision science:
e.g. QED:
𝑎𝜇 = 1159652188.4 4.3 10−12 (exp)
𝑎𝜇 = 1159652140 27.1 10−12
(th)
QED’s renormalizability is an important part of its
calculability, and so also underpins the theory error
Ole Miss 2014
EFT and theoretical error
• General Relativity is also a precision science:
e.g. solar system tests, binary pulsar, ...
Ole Miss 2014
EFT and theoretical error
• General Relativity is also a precision science:
e.g. solar system tests, binary pulsar, ...
𝑑𝑃/𝑑𝑡 = −2.408(10) 10−12 (exp)
𝑑𝑃/𝑑𝑡 = −2.40243 5 10−12 (th)
This comparison meaningless if size of quantum
effects is unknown
Ole Miss 2014
EFT and theoretical error
Cannot afford a ‘split-brain’ mentality:
ℏ=1
𝐺𝑁 = 0
ℏ=0
𝐺𝑁 = 1
Quantum corrections can be calculable within an
effective field theory framework, even for gravity
Ole Miss 2014
EFT and theoretical error
e.g. for graviton-graviton scattering about a static
background in GR:
1
1 2
2
2
2 +. .
𝐿 = 𝜕ℎ +
ℎ 𝜕ℎ +
ℎ
𝜕ℎ
2
𝑀𝑝
𝑀𝑝
+
gives
𝐴𝑛𝑜 𝑙𝑜𝑜𝑝𝑠 ≅
𝑄2
𝑀𝑝
2
About flat space Q
is the CM energy
(de Witt)
Ole Miss 2014
EFT and theoretical error
• The loop integral diverges, and higher-order loops
diverge more and more (because the coupling has
dimensions of negative powers of mass)
𝐴1𝑙𝑜𝑜𝑝 =
𝐴2𝑙𝑜𝑜𝑝 =
𝑄2
𝑀𝑝
4
𝑑4 𝑝
𝑝6
2𝜋 4 𝑝2 + 𝑄 2
𝑄2
𝑑4 𝑝
𝑀𝑝 6
2𝜋
4
2
4
𝑝10
𝑝2 + 𝑄 2
7
Ole Miss 2014
EFT and theoretical error
All divergences cannot be absorbed into Newton’s
constant
2
M
L
c3 3
p
2


R  c1R  c2 R R  2 R  
2
m
g
Divergences can be absorbed if GR is just first term in
a derivative expansion that includes all possible local
interactions allowed by symmetries
Ole Miss 2014
EFT and theoretical error
Predictive provided we regard calculations as expansions
in 𝑄 2 /𝑚2 since no negative powers of Q arise in the
loop expansion.
e.g. L-loop contribution to graviton scattering at energy Q
involving E external lines (in dimensional regularization) and
𝑉𝑖𝑘 vertices involving 𝑖 fields and 𝑘 derivatives:
Ole Miss 2014
EFT and theoretical error
• Leading order, 𝑄2 /𝑀𝑝 𝐸−2 , corresponding to:
• L = 0 and Vik = 0 unless k = 2 ;
• i.e. classical GR: tree graphs using only interactions with
two derivatives.
Ole Miss 2014
EFT and theoretical error
• Next-to-leading order, 𝑄4 /𝑀𝑝 𝐸 , corresponding to one of two
cases:
• L = 1 and Vik = 0 unless k = 2 ;
• L = 0 and Vik = 0 unless k = 2 or 4 and Vi4 = 1 ;
• i.e. one-loop GR, plus tree graphs including one
‘counterterm’ taken from the R2 interactions .
Ole Miss 2014
EFT and theoretical error
• Points to notice:
• Must do Q/M expansion: if not even the semi-classical
approximation breaks down!
• The reference scale M is a physical (renormalized) mass
and never a cutoff
• Cutoffs are guaranteed never to appear in a low-energy
theory, since they always cancel in physical results.
Ole Miss 2014
EFT and theoretical error
about cutoffs, :
• PointsMore
to notice:
• Must do Q/M expansion: if not even the semi-classical
EFT canbreaks
depend
on  since it is
approximation
down!
generated by integrating out physics
E >
• Thewith
reference
scale M is a physical (renormalized) mass
and never a cutoff
onceareused
in physical
• Cutoffs
guaranteed
neverprediction
to appear inone
a low-energy
theory,
since they E
always
in physical
results.
also integrates
< ,cancel
at which
point 
drops out of physical quantities.
Ole Miss 2014
EFT and theoretical error
•
Can calculate coeff a
about cutoffs, : purely within LE
PointsMore
to notice:
theory
by tracking
• Must do Q/M expansion: if not even
the semi-classical
logfor
cutoff dependence.
 is often
a useful
approximation
breaks
down!proxy
following how physics depends on a real
• Thephysical
reference mass
scale MMis a physical (renormalized) mass
and never a cutoff
LE:are guaranteed
A = a lnnever
(/m)
• Cutoffs
to appear in a low-energy
theory,
always
HE:since they
A=
a ln cancel
(M/)in physical results.
phys:
A = a ln (M/m)
Ole Miss 2014
EFT and theoretical error
•
Calculation of coeff a
about cutoffs, : purely within LE
PointsMore
to notice:
theory
needn’t track
• Must do Q/M expansion: if not even
the semi-classical
Tracking
powers
of  iscoeff b or c of physical
approximation
breaks
down!
result.
is often less useful
• The reference scale M is a physical (renormalized) mass
LE:a cutoff A = a 2  b m2
and never
2  a 2
HE:are guaranteed
A = c Mnever
• Cutoffs
to appear in a low-energy
theory,
since they
always
phys:
A=
c M2cancel
 b min2 physical results.
Ole Miss 2014
Decoupling
But how to interpret the non-GR terms in the action?
2
M
L
c3 3
p
2


R  c1R  c2 R R  2 R  
2
m
g
As would be obtained if we ‘integrate out’ a collection of
particles with 𝑚2 ≫ 𝑄 2
Ole Miss 2014
Decoupling
Mass in loop
scale,
so:are there?
Whatsets
other
scales
lightest mass dominates for dim > 4 terms
eg: inmass
4D inflation,
given
and<M4p
heaviest
dominates
for Hdim
there is also v2 = H Mp plus mass
𝑐3 coupling to𝑎𝑘
of any other particles
= 𝑘
2
2
inflaton
2
𝑚
16𝜋 𝑚𝑘
In extra dimensional inflation there is
Notice in particular 𝑀𝑝 is the least important scale when it
mKK < Ms < Mp , and so on…
appears also
in a denominator
Ole Miss 2014
Decoupling
Decoupling: at low energies heavy particles always
contribute suppressed by their mass (once the dim < 5
couplings are appropriately renormalized)
mass M
𝛿𝐿~𝑀 𝑅 + 𝑅2 ln 𝑀 + 𝑅3 𝑀2 + ⋯
Ole Miss 2014
EFT summary
• Quantum effects in gravity are calculable
• Must recognize the implicit low-energy expansions:
Q/𝑀𝑝 and Q/𝑚
• In particular this justifies domain of classical
approximation
Ole Miss 2014
EFT summary
• Quantum effects in gravity are calculable
• Must recognize the implicit low-energy expansions:
Q/𝑀𝑝 and Q/𝑚
• Never get anything but a series in local powers of
fields and derivatives in this way: f ( R) ?
2
M
L
c3 3
p
2


R  c1R  c2 R R  2 R  
2
m
g
Ole Miss 2014
EFT summary
• Quantum effects
in gravity
are calculable
These
are usually
useful for cosmology
and particle
physics,
and when
used
• Must recognize
the implicit
low-energy
expansions:
make these unusually sensitive to UV
Q/𝑀𝑝 and Q/𝑚
physics.but a series in local powers of
• Never get anything
fields and derivatives in this way: f ( R) ?
• Heavy particles generically decouple. Yet theories with lowdimension interactions – like scalars with 𝑚2 𝜑 2 terms – can
be sensitive to very heavy particles.
Ole Miss 2014
EFT summary
• Quantum effects
inare
gravity
area calculable
These
usually
problem for theories
that break
lorentzlow-energy
invarianceexpansions:
in the
• Must recognize
the implicit
UV, since they spread to all particles
Q/𝑀𝑝 and Q/𝑚
throughbut
loops
• Never get anything
a series in local powers of
fields and derivatives in this way: f ( R) ?
• Heavy particles generically decouple. Yet theories
with low-dimension interactions – like scalars with
𝑚2 𝜑 2 terms – can be sensitive to very heavy
particles.
• Ditto for lorentz-violating diffs btwn 𝜑 2 and 𝛻𝜑 2
Ole Miss 2014
Part II
Ole Miss 2014
Time dependent situations
Gravitational exceptionalism
Ole Miss 2014
Decoupling and Horizons
• Effective field theory as applied to geometries having
horizons (eg: black holes, inflation) introduces some
new issues
• Adiabatic time slicing is not the static one
• These are similar to those which arise for timedependent background fields, F(t)
• Slow evolution
• No level crossing
1 dF

F dt
Ole Miss 2014
Time dependent situations
• Can effective theories be used for time-dependent
problems?
• Q1: when do effective theories capture the time dependence
of evolution in the full theory?
• Since EFTs have higher-derivative interactions, why
aren’t there new runaway solutions?
Ole Miss 2014
Time dependent situations
• Can effective theories be used for time-dependent
problems?
• Q1: when do effective theories capture the time dependence
of evolution in the full theory?
• Since EFTs have higher-derivative interactions, why
aren’t there new runaway solutions?
• Q2: Can an effective theory be set up to describe the
fluctuations about a time-dependent background?
Ole Miss 2014
Time dependent situations
• Time dependent solutions to EFT only need agree with
adiabatic evolution in the full theory
Ole Miss 2014
Time dependent situations
• Time dependent solutions to EFT only need agree with
adiabatic evolution in the full theory
For example, for the theory:
𝐿 = 𝜑2 + 𝜓 2 − 𝑀2 𝜓 2 − 𝑔 𝜑2 𝜓
Feynman tree graphs give
𝐿𝑒𝑓𝑓
2
𝑔
4+⋯
= 𝜑2 −
𝜑
2𝑀2
Ole Miss 2014
Time dependent situations
• Time dependent solutions to EFT only need agree with
adiabatic evolution in the full theory
𝑀2 𝜓
𝑔 2
= − 𝜑
2
But the equation
𝜓+
has solutions
𝑔
2
𝜓= −
𝜑
+⋯
2
2𝑀
and so equivalent way to get 𝐿𝑒𝑓𝑓 (at tree level) is
𝐿𝑒𝑓𝑓 (𝜑) = 𝐿 𝜑, 𝜓(𝜑)
Ole Miss 2014
Time dependent situations
• Time dependent solutions to EFT only need agree with
adiabatic evolution in the full theory
After all
𝑔 2
2
𝜓+ 𝑀 𝜓= − 𝜑
2
also has non-adiabatic solutions
𝜓=
𝑐𝑘
𝑘=∓
𝑒 𝑖𝑘𝑀𝑡
𝑔
2+⋯
−
𝜑
2𝑀2
and so equations of motion of the EFT only agree with
the adiabatic solutions
Ole Miss 2014
Time dependent situations
• This is also why runaway solutions do not arise
• Higher-order equations of motion normally acquire new,
often runaway, solutions:
• e.g. if
𝐿 = 𝑞 2 + 𝑞 2 /𝑀2 then 𝑞 + 𝑞 /𝑀2 = 0
.
and
so 𝑞 𝑡 = 𝑎 + 𝑏𝑡 + 𝑐 𝑒 𝑀𝑡 + 𝑑𝑒 −𝑀𝑡
• Key point: must perturb in the effective interactions, since
they are only required to reproduce the full physics to fixed
order in 1/𝑀
Ole Miss 2014
Time dependent situations
• This resembles models that inflate based on highercurvature interactions:
L  M R  z R 
2
p
2
• inflates with 𝑅 = 𝐻 2 = 𝑀𝑝 2 /𝜁
• gives the right primordial density fluctuations provided z ~
108.
Ole Miss 2014
Time dependent situations
• What seems odd is that successive terms in the
curvature expansion are equal size (and not because
the first term was particularly small)
• Should the dimensionless part of the coefficient of the 𝑅3
term be order 1 or order 𝜁 2 ?
• Time dependence is 𝑎(𝑡)~𝑒 𝐻𝑡 with 𝐻~𝑀𝑝 /√𝜁 seems hard
to understand as agreeing with full theory order by order in
1/𝑀 ~ 𝜁/𝑀𝑝
Ole Miss 2014
Time dependent situations
• Can effective theories be used for expansions about
time-dependent backgrounds?
• At face value there is a problem: we distinguish light from
heavy states from one another using energy, but lose
energy conservation when expanding about a timedependent background.
• Can nonetheless work if background is adiabatically
evolving since in this case can define an adiabatic notion of
energy at a given time, 𝐸(𝑡)
Ole Miss 2014
Time dependent situations
• For EFT built around adiabatic evolution:
• Must check whether the low-energy criterion, 𝐸(𝑡) ≪
𝑀(𝑡) remains true as a function of time:
eg: must
avoid level crossing:
M
Ole Miss 2014
Gravitational exceptionalism
Gravitational exceptionalism:
Q: Isn’t gravity a unique situation for which insights from
EFTs in other areas do not apply?
In particular: doesn’t inflation (or Hawking radiation)
violate decoupling by stretching initially short-wavelength
modes out to longer distances?
Ole Miss 2014
Gravitational exceptionalism
• Descent of modes is not special to gravity:
• similar effects can also happen in other time-dependent
settings (like the collapse of Landau levels 𝐸 = 𝑛 𝜔 with
𝜔 ∝ 𝐵 when magnetic field 𝐵 is turned off).
B
Ole Miss 2014
Gravitational exceptionalism
• Descent of modes is not special to gravity:
• Normally the appearance of such states has no low-energy
effects, so long as the time dependence is adiabatic so states
descend in their adiabatic vacuum.
• All known examples of trans-Planckian phenomena rely on
breakdown of adiabaticity. (Some also violate lorentz
invariance in the far UV, but this is not required, as may be
seen from rolling scalar examples.)
Ole Miss 2014
Gravitational exceptionalism
Trans-planckianism
• Less precise version: Since we don’t understand
quantum gravity (at short distances), how do we
know that trans-Planckian physics decouples from
long-wavelength physics?
Ole Miss 2014
Gravitational exceptionalism
Trans-planckianism
• Less concrete version: Since we don’t understand quantum
gravity (at short distances), how do we know that transPlanckian physics decouples from long-wavelength physics?
Cannot answer for sure without knowing the physics
above the Planck scale. but:
• Trans-planckian physics appears to decouple in string theory
• If decoupling fails must explain why any understanding of
nature is possible at all (ie there is an uncontrolled
theoretical error)
Ole Miss 2014
Exceptionalism summary
• Time dependence can be captured within an EFT
• Must require the time dependence to be adiabatic
Ole Miss 2014
Exceptionalism summary
• Time dependence can be captured within an EFT
• Must require the time dependence to be adiabatic
• Must check that large mass hierarchies do not
become small over time
Ole Miss 2014
Exceptionalism summary
• Time dependence can be captured within an EFT
• Must require the time dependence to be adiabatic
• Must check that large mass hierarchies do not
become small over time
• Cosmology appears not to introduce qualitatively
new considerations beyond those associated with
time-dependence.
Ole Miss 2014
Part III
Ole Miss 2014
The case for modifying gravity
Microphysics as a crucial clue
Ole Miss 2014
Dark Matter/Energy or Modified GR?
• Evidence for Dark Matter
comes from many sources:
• Mass in galaxies
• Mass in clusters of galaxies
• Temperature fluctuations in
the CMB
• Start of galaxy formation.
• Evidence for Dark Energy
is less robust, but would
involve modifications to
gravity over large distances.
Ole Miss 2014
Dark Matter/Energy or Modified GR?
• Modifications of gravity
does a poor job describing
observations
• Mass in galaxies
• Mass in clusters of galaxies
• Temperature fluctuations in
the CMB
• Start of galaxy formation.
• Both require modification
at large distances: very hard
to do sensibly
Ole Miss 2014
Modifications in the UV
• Short-distance modification to
gravity is very likely, since
predictability breaks down at high
energies.
• Sensible UV modifications seem
difficult to get (but not impossible).
• String theory provides an example
UV modification.
• Difficult to test since their details
decouple from low-energy physics
Ole Miss 2014
Modifications in the IR
• Long-distance modifications to gravity are very
difficult to construct without violating fundamental
principles like unitarity, cluster decomposition, etc.
(Weinberg)
• Eg: gauge invariance emerges as a requirement for
massless spin-two particle coupled to stress energy
massless spin 2  C     h  
Lint  h T


Ole Miss 2014
Modifications in the IR
• Long-distance modifications to gravity are very
difficult to construct without violating fundamental
principles like unitarity, cluster decomposition, etc.
• Eg: gauge invariance emerges as a requirement for
massless spin-two particle coupled to stress energy
• After decades of investigation, only known
consistent IR modifications are scalar/vector/tensor
theories (possibly in higher dimensions).
Ole Miss 2014
Microphysics as a crucial clue
• The puzzle of ‘non-decoupling’ in cosmology
• Decoupling: usually small-distance physics is not required
to understand long-distance physics
Why should UV physics have anything to say about cosmology?
Ole Miss 2014
Microphysics as a crucial clue
• The puzzle of ‘non-decoupling’ in cosmology
• Most cosmological models rely on long-distance features
that do depend on the details of short-distance physics
• the existence of scalars with small masses, m ≤ H < M
• the existence of small vacuum energies, G ≤ H2
• the existence of exotic low-E states or dynamics
Ole Miss 2014
Microphysics as a crucial clue
• Non-decoupling as a double-edged sword
• If short-distance physics does not decouple, why are
cosmological predictions possible without first solving
quantum gravity?
i.e. are inflationary interpretations of the CMB robust?
• The dangers of ‘Planck slop’
• Even if heavy physics decouples, why can’t small Plancksuppressed interactions ruin cosmology?
𝐿𝑒𝑓𝑓 = 𝑉0 −
𝑉0
𝑀𝑝 2
𝜑2 + ⋯
Ole Miss 2014
Microphysics as a crucial clue
• Both questions require a UV completion to answer.
Seek:
• A mechanism for explaining why quantum corrections do
not ruin required light-scalar properties or vacuum energy
• Understanding of presence/absence of Planck slop and its
implications
Ole Miss 2014
Summary
• Quantum corrections are calculable in GR much like in
any other non-renormalizable theory.
• Quantum effects are typically negligibly small (with
inflationary cosmology a notable exception!)
• Allows efficient tracking of finite-size effects in, eg,
radiation during binary in-spiralling.
• UV modifications to gravity decouple from lowenergy observables
• IR modifications to gravity are difficult to make
consistently
• Scalar, vector, tensor theories (possibly in higher D)
Ole Miss 2014
Summary
• EFTs provide our only current method for comparing
gravitational predictions with data
• Must check the validity of low-energy, classical and
adiabatic approximations.
Ole Miss 2014
Summary
• EFTs provide our only current method for comparing
gravitational predictions with data
• Must check the validity of low-energy, classical and
adiabatic approximations.
• Leads to prejudice for local theories that include all
possible interactions consistent with symmetries in a
systematic derivative expansion
Ole Miss 2014
Summary
• EFTs provide our only current method for comparing
gravitational predictions with data
• Must check the validity of low-energy, classical and
adiabatic approximations.
• Leads to prejudice for local theories that include all
possible interactions consistent with symmetries in a
systematic derivative expansion
• Most heavy physics decouples: robust cosmology
Ole Miss 2014
Summary
• EFTs provide our only current method for comparing
gravitational predictions with data
• Must check the validity of low-energy, classical and
adiabatic approximations.
• Leads to prejudice for local theories that include all
possible interactions consistent with symmetries in a
systematic derivative expansion
• Most heavy physics decouples: robust cosmology
• Suggests some features (technical naturalness, etc)
are important criterion for most cosmologies.
Ole Miss 2014
Ole Miss 2014