Burgess NordicSchool2019

INTRODUCTION  
TO EFFECTIVE FIELD
THEORY
Nordic Winter School 2019
C.P. Burgess
OUTLINE
• Motivation and Overview
• Decoupling and quantifying theoretical error
• Effective field theories & a toy model
• Particle physics
• Known unknowns and unknown unknowns
• Technical naturalness?
• Gravity and cosmology
• Time-dependence and EFTs
• Relevance to present puzzles
• A brief cold shower
• Inflation, Black holes, Dark matter and Dark energy
MOTIVATION & OVERVIEW

DECOUPLING
• Decoupling: most details of small-distance physics are not needed when

understanding long-distance physics (this is why science is possible)
• It turns out that QFT very generally shares this property: how to see it
explicitly? Can it be exploited to simplify calculations?
ACCURACY OF CALCULATIONS
• Some predictions work too well
• eg superconductivity relies on
electrons pairing into Cooper pairs:
• neglect Coulomb interaction
• keep interactions via lattice ions
• predictions work at 10% level

RENORMALIZATION
• Renormalization: Why does it
make sense to subtract aμ = 1159652188.4(4.3) 10−12 (exp)
infinities and then compare aμ = 1159652140(27.1) 10−12 (th)
with experiment to 10 decimal
places? μ
ℒint = ieAμψγ ψ
4
d p 1
∫ (2π)4 (p + q + m)2(p + q)2
3
e
( 4π ) [ ∫ p 4
2
]
e d4p d4p
∫ p5
∼e +q +⋯

RENORMALIZATION
• Renormalizable: don’t need

new parameters beyond “e” ℒint = ieAμψγ μψ
and “m” for other observables
4 2
d p 1
∫ (2π) (p + q + m) ( 4π )
e2 d4p
∫
4
e 4 4
∼ q 4
+⋯
p 8
( 4π ) m
2
e q4
∼ 4
+⋯

RENORMALIZATION
• Renormalizable: higher orders

do not make things worse ℒint = ieAμψγ μψ
∫ ( (2π)4 ) (p + q + m)6(p + q)2

4
( 4π )
4
d p 1 e d 4
p
∫ p8
6
e ∼e 2
q 4
+⋯
( 4π ) m 4
4 4
e q
∼ e2 +⋯

RENORMALIZATION
• Do nonrenormalizable theories ·
(eg GR) make sense? If not P = − 2.408(10) 10−12 (exp)
why is comparison with ·
P = − 2.40243(5) 10−12 (th)
observations meaningful?

RENORMALIZATION
• Nonrenormalizable theories
hμν
have couplings with dimension gμν = ημν +
inverse power of mass Mp
2 1 1 2
ℒGR = (∂h) + h(∂h) + 2 h (∂h)2 + ⋯
2
Mp Mp
Mp ( tu )
q2 s3
𝒜tree ≃ 2 = 8πiG
De Witt

RENORMALIZATION
• Higher orders diverge worse
and worse 2 1 1 2
• New types of divergences ℒ = (∂h) + h(∂h) + 2 h (∂h)2 + ⋯
2
Mp Mp
q2 d4p (p + q)6
Mp ∫ (2π)4 (p + q)8
𝒜loop ≃ 4
( 4πMp2 ) [ ∫ p 2 ]
q d4p d 4
p
∫ p4
2
≃ + q + ⋯

TOY MODEL
• Effective field theories (EFTs) are
the formalism for addressing ℒ = − (∂ϕ)*(∂ϕ) − V(ϕ*ϕ)
these questions.
• Designed to exploit λ
hierarchies of scale m/M as V(ϕ*ϕ) = (ϕ*ϕ − v 2)2
efficiently as possible. 4
1
( I)
ϕ=v+ ϕ ̂ + iϕ̂
R
• Toy model: concrete example
that illustrates the construction
2
• λ ≪ 1 is semiclassical limit
mR2 = λv 2 mI2 = 0
TOY MODEL
• Can calculate eg tree level
scattering of massless particles ℒ = − (∂ϕ)*(∂ϕ) − V(ϕ*ϕ)
λ
V(ϕ*ϕ) = (ϕ*ϕ − v 2)2
4
1 2
ϕ*ϕ − v = 2 vϕR + (ϕR + ϕI2)
2
2
[ ] ( 4π )
2
3iλ i(λv)2 1 1 1 λ
𝒜=− + + + +𝒪
2 2mR 1 + 2p ⋅ q/mR 1 − 2q ⋅ q′ /mR 1 − 2p ⋅ q′ /mR
2 2 2 2

TOY MODEL
• Low-energy limit: E << mR
2mR [ 1 + 2p ⋅ q/mR 1 − 2q ⋅ q′ /mR 1 − 2p ⋅ q′ /mR2 ]

3iλ i(λv)2 1 1 1
𝒜=− + + +
2 2 2 2
( ) ( 4π )
6 2
[ ] [ mR ]
(p ⋅ q)2 + (p ⋅ q′)2 + (q ⋅ q′)2 q λ
≃ 2iλ +𝒪 λ +𝒪
mR4
• Amplitude suppressed by (E/mR)4 in low-energy limit. What if E/mR smaller

than coupling? Is dominant low-energy contribution at one loop?
TOY MODEL
• This low-energy amplitude
[ ]
(p ⋅ q)2 + (p ⋅ q′)2 + (q ⋅ q′)2
( R )
−6
𝒜 ≃ 2iλ + 𝒪 m
mR4
is what would have been obtained at lowest order from the following
lagrangian
1 μ λ μ ̂ ν ̂
ℒ = − ∂μφ̂ ∂ φ̂ + (∂ μφ̂ ∂ φ)(∂ νφ̂ ∂ φ)
2 4mR4
TOY MODEL
• Turns out 2 to 2 amplitude is proportional to (E/mR)4 at each loop. N to
N’ scattering is proportional to (E/mR)N+N’. Easy way to see why?
• Obvious once recognized at low energies all massless scattering (also

N to N’) is governed by èffective lagrangian’
1
ℒ = − ∂μφ̂ ∂ μφ̂ + Geff (∂μφ̂ ∂ μφ)(∂
̂ νφ̂ ∂νφ)̂
2
λ
Geff = [ 1 + 𝒪(λ) ]
4mR4
• All other interactions are suppressed by more powers of E/mR.
(Suppression of low-energy scattering better than a tree level result.)
TOY MODEL
Two steps to see why:
• Expand in powers of E/M as early as possible.

• Do so by integrating out heavy particle to obtain effective
lagrangian (applicable to all low-energy observables), rather than
observable by observable.
• Make symmetries manifest in this effective lagrangian

SYMMETRIES
• Toy model has U(1) symmetry
ℒ = − (∂ϕ)*(∂ϕ) − V(ϕ*ϕ)
λ ϕ → e iθϕ
V(ϕ*ϕ) = (ϕ*ϕ − v 2)2
4
1
( R I)
ϕ=v+ ϕ ̂ + iϕ̂
2
2
How to realize symmetry
2 2 mI = 0
mR = λv using only light field?
SYMMETRIES
• Redefine variables
iξ/v Symmetry as realized on
ϕ = (v + χ) e low-energy field is not linear
• Symmetry transformation
iθ
ϕ→e ϕ
implies ξ → ξ + θv χ → χ
• Toy model lagrangian becomes
( v)
2
2 χ
ℒ = (∂χ) + 1 + (∂ξ)2 − V( χ)

EFFECTIVE FIELD THEORIES
• To integrate out heavy particle: split fields into high- and low-energy
parts (won’t be unique way to do this)
ϕ̂ ↔ {H,̂ L}
̂ where E(L) < Λ and E(H) > Λ
• Low-energy observables can be computed from
∫ ( )
̂ 1)⋯L(x
⟨L(x ̂ n)⟩ = 𝒟L𝒟
̂ Ĥ L(x
̂ 1)⋯L(x
̂ n) exp iS[L,̂ H]̂
• Want to efficiently identify effects of heavy physics in powers of 1/M.

• Want:
∫ ( )
̂ 1)⋯L(x
⟨L(x ̂ n)⟩ = 𝒟L𝒟
̂ Ĥ L(x
̂ 1)⋯L(x
̂ n) exp iS[L,̂ H]̂
• Define Wilson action:

exp (iSW[L]̂ ) = 𝒟Ĥ exp (iS[L,̂ H]̂ )
∫
• Then
∫ ( )
̂ 1)⋯L(x
⟨L(x ̂ n)⟩ = 𝒟L̂ L(x
̂ 1)⋯L(x
̂ n) exp iSW[L]̂
• Any Λ dependence of SW cancels in < L(x1)…L(xn) >.

• The Wilson action defined by:
exp (SW[L]̂ ) = 𝒟Ĥ exp (iS[L,̂ H]̂ )

∫
behaves “as if” it is the classical action for the low-energy theory.
• SW defined this way would be nonlocal, BUT becomes local once
expanded in powers of 1/M (consequence of uncertainty principle):
[M ]
d4p 1 1 □
∫ (2π) p + M
4
G(x − y) = = + + ⋯ δ (x − y)
4 2 2 2 M 4

• Expanding propagator in 1/M gives local result:
g
G g2
G= 2
M
g
• New coupling is generically nonrenormalizable, but underlying theory
could be renormalizable so must be predictive
• Predictivity comes from compulsory low-energy approximation

• Not quite so simple as contracting a line once loops
included (eg integrating out muons in QED)
+ +
• Circling back: renormalization. Recall definition of SW depends on cutoff
exp (iSW[L,̂ Λ]) = 𝒟Ĥ exp (iS[L,̂ H]̂ )

∫Λ
• But cutoff also enters into its use:
Λ
∫ ( )
̂ 1)⋯L(x
⟨L(x ̂ n)⟩ = 𝒟L̂ L(x
̂ 1)⋯L(x
̂ n) exp iSW[L,̂ Λ]
• Any Λ dependence of SW cancels in < L(x1)…L(xn) >: nothing depends

on the cutoff!

INTRODUCTION TO EFT
(LECTURE 2)
C.P. Burgess
TOY MODEL RECAP
• Scattering of massless states is
suppressed at each order in ℒ = − (∂ϕ)*(∂ϕ) − V(ϕ*ϕ)
the loop expansion by powers
of E/mR λ
V(ϕ*ϕ) = (ϕ*ϕ − v 2)2
• This can be understood by
4
building an EFT for the light
(Goldstone) particle alone. ϕ = (v + χ) e iξ/v
• Low energy field nonlinearly
realizes symmetry
ϕ → e iθϕ implies ξ → ξ + θv χ → χ
TOY MODEL EFT
• Wilson action for the Toy Model must be invariant under shift
symmetry, so built from only derivatives of the light field
SW[ξ + θ] = SW[ξ] ⇒ SW = SW[∂ξ]

1
ℒW = − ∂μξ ∂ μξ + Geff (∂μξ ∂ μξ)(∂νξ ∂νξ) + ⋯
2
• Could get G by integrating out heavy field, but λ
better simply to “match”: choose it to reproduce a Geff =
result of the full theory (eg a scattering amplitude). 4mR4
TOY MODEL EFT
• To drive home how SW describes all low energy physics (and that EFT need
not be restricted to expansions about vacuum configuration) consider time-
dependent solution to full theory’s field equation
λ
□ ϕ = (ϕ*ϕ − v 2)ϕ
2
• Consider the time-dependent solution: Energy gain
2 because field
ϕ= ϱ0 eiωt
where 2ω
ϱ0 = v 2 + climbs the
λ potential to
( λ )
2 balance
· · λ 3ω
with energy ε = ϕ*ϕ + (ϕ*ϕ − v 2)2 = ω 2 v 2 + centrifugal
4 force
TOY MODEL EFT
• How does the EFT know about the radial field climbing the potential
given there is no radial field in the EFT?
1
ℒW = − ∂μξ ∂ μξ + Geff (∂μξ ∂ μξ)(∂νξ ∂νξ) + ⋯
2
• This gives the field equation
∂μ {∂ μξ [1 − 4Geff (∂ξ)2 + ⋯]} = 0

ξ
with solution = ωt
2v
TOY MODEL EFT
• Compute the energy of this solution within the EFT:
· 1 ·2 ·4
ε = ℋ = Πξ − ℒ = ξ + 3Geff ξ
2
which with the matched value for G becomes:
4
Geff =
λ 3ω
4mR4
ε = v 2ω 2 +
λ
in agreement with the full theory
REDUNDANT INTERACTIONS
• What about other terms with same dimension (or less) and so
same (or lower) power of 1/mR in its coefficient?
e.g. ℒ = G1 (∂μξ) □ ∂ μξ or G2 ∂μ∂νξ ∂ μ∂νξ
• These differ only by a total derivative (so are not independent)

• The first can be removed to this order in 1/M with field redefinition
δξ = G1 □ ξ
REDUNDANT INTERACTIONS
• Field redefinitions can be used to remove any term in the effective
action that vanishes when evaluated at the solution to the lowest
order field equations - for the Toy Model: □ ξ = 0
if S[ξ] = S0[ξ] + ϵS1[ξ] + ⋯ then when δξ = ϵ F[ξ]
δS0
∫
4
δS[ξ] = ϵ d x F(ξ) + ⋯
δξ(x)
which can be used to remove any term in S1 that vanishes using the
e.o.m. of S0.
DIMENSIONAL REGULARIZATION
• Dimensional regularization does not introduce a new scale
(apart from logs)
∫ (2π)D [ (p 2 + q 2)B ] (4π)D/2 [ ]

dDp p 2A 1 Γ(A + D/2)Γ(B − A − D/2) 2 A−B+D/2
= (q )
Γ(B)Γ(D/2)
• Divergences arise as poles as D goes to 4

• Convenient because it preserves symmetries (eg gauge
invariance) broken by cutoffs, and simplifies dimensional reasoning

EFTS IN DIM REG
• Dimensionally regularize both the full theory and EFT
ℒfull(χ, ξ) ℒEFT(ξ)
• Renormalize in any convenient way (eg minimal subtraction)

• Match the couplings in the EFT by demanding they give same
observables as for the full theory
• Any error introduced by keeping very high energy modes of
light field is absorbed into the effective couplings.

POWER COUNTING
• The real power with EFTs comes beyond leading order in E/M.
• Need an algorithm to systematically identify which interactions and

which Feynman graphs must be included to any specific order.
• When all the low-energy scales are similar in size this algorithm
amounts to dimensional analysis of Feynman graphs
• much simplest to do this using dimensional regularization.

POWER COUNTING
(M v)
• Consider lagrangian of form 4 ∂ ξ
∑
ℒ=f cn 𝒪n ,
n
• Consider (amputated)
Feynman graph with
• E external lines,
• I internal lines
• Vn vertices involving dn
derivatives and fn fields
∑ ∑
• All such diagrams satisfy: L =1+I− Vn 2I + E = fnVn
n n
POWER COUNTING
I
[ f 4 ∫ (2π)4 p 2 ]
2 2 4
M v dp 1
• Internal lines bring: Line Factor =
Vn
∏ [ v fn ( M )
dn
]
• Vertices bring f4 p 4 4
Vertex Factor = (2π) δ (p)
n
∑
• Number of independent integrals I− Vn + 1 = L
n
POWER COUNTING
I Vn
∏ [ v fn ( M )
dn
[ f 4 ∫ (2π)4 p 2 ]
2 2 4
]
4
M v dp 1 f p 4 4
Line Factor = Vertex Factor = (2π) δ (p)
n
∑
• Net power of f4: −I + Vn = 1 − L
∑
• Net power off 1/v: −2I + fnVn = E
• In dimensional reg: p
( M)
∑
N N = 4L − 2I + dnVn
becomes q so net 4L q
power of q, M is M
∑
= 2 + 2L + (dn − 2)Vn
POWER COUNTING
• Combining terms
dn−2 Vn
( ) ( ) n( )
E 2L
[ ]
q2 f 4 1 Mq q
∏
𝒜E(q) ∼ c
M 2 v 4πf 2
n
M
• For Toy Model f 2 = Mv M = mR
dn−2 Vn
( v ) ( 4πv ) ∏ [ ( mR )
E 2L
]
1 2 2 q q
𝒜E(q) ∼ q v cn
n
POWER COUNTING
dn−2 Vn
( v ) ( 4πv ) ∏ [ ( mR )
E 2L
]
2 2 1 q q
𝒜E(q) ∼ q v cn
n
• Notice always positive powers of q since dn is 2 or larger (actually 4 or
larger in the toy model)
• Interactions with no derivatives (ie scalar potential) are potentially

dangerous at low energies

POWER COUNTING FOR GRAVITY
• Aside: a similar expression can be derived for gravity:
Mp2 c2 μν
μνλρ λραβ
ℒ=− R + c1RμνλρR + 2 RμνλρR Rαβ +⋯
2 m
• Coefficient of curvature cubed term is set by mass of particle
integrated out, and smallest m wins in the denominator
Vn
E 2L 2
(m)
dn−4
( Mp ) ( 4πMp ) d∏ ( )
2 2 1 q q q
𝒜E(q) ∼ q Mp
≥4
Mp
n

POWER COUNTING FOR GRAVITY
Mp2 c2 μν
μνλρ λραβ
ℒ=− R + c1RμνλρR + 2 RμνλρR Rαβ + ⋯
2 m V n
E 2L 2
(m)
dn−4
( Mp ) ( 4πMp ) d∏ ( )
2 2 1 q q q
𝒜E(q) ∼ q Mp
≥4
Mp
n
• Dominant contribution: L=0 and V_n = 0 for dn > 2 (ie classical GR)
• Next-to-leading contributions: L = 1 and V_n =0 for dn >2 (ie 1-loop GR);
or L = 0 and V_n = 1 for d_n = 4 term (tree level with one insertion of R2 term)
2
( 4πMp )
• Size of quantum corrections: q

INTRODUCTION TO EFT
(LECTURES 3 & 4)
C.P. Burgess
UNITARITY BOUNDS
• For toy model power counting:
dn−2 Vn
( v ) ( 4πv ) ∏ [ ( mR )
E 2L
]
2 2 1 q q
𝒜E(q) ∼ q v cn
n
• Could gauge the U(1) symmetry to get a gauge boson with mass:
MA2 = 2g 2v 2
• Massive gauge boson is in the low energy theory if g 2 ≪ λ

UNITARITY BOUNDS
dn−2 Vn
( v ) ( 4πv ) ∏ [ ( mR )
E 2L
]
2 2 1 q q
𝒜E(q) ∼ q v cn
n
• Theory breaks down when
4πMA
q ∼ 4πv ∼
g
• Often quoted as a “unitarity bound”: when low-energy cross section
exceeds unitarity limit

ZERO DERIVATIVE INTERACTIONS
dn−2 Vn
( v ) ( 4πv ) ∏ [ ( mR )
E 2L
]
2 2 1 q q
𝒜E(q) ∼ q v cn
n
• What about mass term for light particle
V = m 2ϕ 2 ⇒ cn = m 2v 2 /f 4 = m 2 /mR2
• Mass insertions have dn = 0 and come with factors
(m 2 /mR2)(mR2 /q 2) = m 2 /q 2
ZERO DERIVATIVE INTERACTIONS
• For relativistic particles q >> m so perturbing in m/q is OK.
• For q ~ m the kinematics becomes non-relativistic and so path

integral becomes dominated by Schrodinger action, which scales t
and x differently
• Leads to different form of low energy theory (eg NRQED or NRQCD or

HQET) and it is the relevant interaction that signals the instability
towards this transition at low energies.

RELEVANT INTERACTIONS
• Normally a relevant interaction signals a transition to a
new scaling regime
μ μ μ
x → x̃ = sx ϕ → ϕ̃ = ϕ/s
[ ]
2
m
S = d4 x[(∂ϕ)2 + m 2ϕ 2] = d4 x̃ (∂˜ ϕ̃)2 + 2 ϕ̃2
∫ ∫ s
So mass term becomes more important as s tends to
zero: although can perturb in the mass for relativistic
problems, once q ~ m nonrelativistic scaling takes over
EFTS IN PARTICLE PHYSICS
Known unknowns

RAYLEIGH SCATTERING
• As a first practical example of EFT methods consider photons scattering from
a neutral body much smaller than wavelength.
a≪λ
• Lowest dimension interaction between photon and neutral field is
g
ℒ = cΨ*Ψ ∇ ⋅ E + Ψ*Ψ E2 + ⋯
2
where first term is redundant and g has dimensions (length)3 and is called the object’s
polarizability. Amplitude and cross section for photon scattering from neutral body then is
dσ g 2k 4 g 2k 4
𝒜 = igkk′ϵ ⋅ ϵ′ = (1 + cos2
θ) σ=
dΩ 32π 2 6π
QED (BELOW ELECTRON MASS)
• Another illustrative example of EFTs at work is QED
1
ℒ = − FμνF μν − ψ(γ μDμ + m)ψ − eAμJ μ
4
• Most general possible low-energy interactions of these kinds of fields
• Integrate out the electron
m [ ]
1 b1 7
ℒeff = − eAμJ − Z FμνF + 4 (FμνF ) + (FμνF̃μν)2 + ⋯
μ μν μν 2
4 4

• Integrate out the electron
m [ ]
1 b1 7
ℒeff = − eAμJ − Z FμνF + 4 (FμνF ) + (FμνF̃μν)2 + ⋯
μ μν μν 2
4 4
• Compute Z using vacuum polarization graph
( μ2 )
α 1 m2
Z =1− − γk + ln (D = 4 − 2ϵ)
3π ϵ
• Compute b1 using box graph
α2
b1 =
90
• The four-photon term provides the simplest way to compute low-energy
photon-photon scattering cross section
2
6
4π ( 90 ) ( m )
2
Ecm
dσ 139 α
(3 + cos θ)
2 2
≃ 2
dΩ 8
• There are also redundant operators
∂μF μν∂ λFλν ∝ e 2J ν Jν Fμν □ F μν ∝ eF μν∂μJν ≃ − e 2J μJμ

• In this case redundant interactions generate current-current
interactions, whose presence can be ignored only at places where
there are no currents
• The Maxwell term is also interesting, must rescale Aμ = Z −1/2 A′μ
1
ℒeff = − ephys A′μJ − F′μνF μν′ + ⋯
μ
4
• Upshot: low-energy influence of electron is suppressed by m only after
appropriate redefinition of e. (Precise statement of decoupling.)
( μ2 )
α 1 m2
e = Z 1/2ephys = ephys 1− − γk + ln
6π ϵ
• Corrections toα macroscopic classical E+M given by powers of E/m

rather than . This is at root of why Rutherford scattering is same
in classical and quantum calculation.
QED (ABOVE ELECTRON MASS)
• For E bigger than m, the renormalization Z gives information about
large logs
( μ2 )
α 1 m2
Z =1− − γk + ln
3π ϵ
• In modified minimal subtraction remove just first two terms
( )
−1/2 α 1
Aμ = ZMS A′μ Z MS = 1 − − γk
3π ϵ
• Corresponding charge is not itself physical
( Z ) phys 3π ( μ 2 )
Z MS α m2
αMS = α = αphys 1 + ln
• Since physical charge cannot depend on mu, must have
2
∂α α
μ 2 MS = − MS
∂μ 2 3π
• Because MSbar is mass-independent its RG evolution is easy to solve
αMS(μ) αMS(μ0) 3π ( μ02 )

1 1 1 μ2 with αMS(μ = m) = αphys
= − ln
This is RG improved in that it holds even when both terms on RHS are similar size

• Why do we care? Consider E >> m limit of scattering
(E )
1 me
σ(E, me, αphys) = 2 F , αphys, f, θk
E
where there is a sum over soft photons up to energies
Eγ = fE with 1 > f ≫ m /E
• Cannot Taylor expand F due to log(m/E) singularities, but these are not
present when using MSbar couplings. Identify log(E/m) by setting mu=E in
E [ ( ) ]
1 E
σ(E, me, αphys) = 2 F0 , αMS(μ), f, θk + 𝒪(m /E)
μ
QED (INCLUDING MUONS)
• Next consider QED at energies above the muon mass
1
ℒ = − FμνF μν − ψ(γ μDμ + m)ψ − χ(γ μDμ + M)χ
4
• Most general possible low-energy interactions of these kinds of fields
• Integrate out the muon gives
M [ ]
1 b1 7
ℒeff = − Z FμνF − Ze ψ(γ Dμ + Zmm)ψ + 4 (FμνF ) + (FμνF̃μν)2 + ⋯
μν μ μν 2
4 4

• Integrating out muon gives Fmn4 term with coefficient M-4. Integrating
out the electron gives m-4. At lower energies smallest mass wins.
(M m )[ ]
1 1 μν 2 7 μν 2
ℒeff ⊃ b1 + (Fμν F ) + (Fμν F̃ ) +⋯
4 4 4
Barring selection rules should expect smallest mass to dominate in
denominators, but largest mass wins in numerators. From that point
of view the large size of the Planck mass makes sense
1 2
ℒ ⊃ − (m + M 2 + Mp2)R + ⋯
2
while the cosmological constant is a puzzle…
• In minimal subtraction both muons and electrons contribute to the
running of the EM coupling
( Z ) phys 3π ( μ ) 3π ( M 2 )
Z MS α m2 α m2
αMS = α = αphys 1 + ln 2
+ ln
and so 2
∂α 2α
μ 2 MS = − MS
∂μ 2 3π
αMS(μ) αMS(μ0) 3π ( μ02 )

1 1 2 μ2
= − ln with αMS(μ = mM) = αphys

• Minimal subtraction makes it seem as if muons play a role in running
also at energies below the muon mass.
( Z ) phys 3π ( μ ) 3π ( M 2 )
Z MS α m2 α m2
αMS = α = αphys 1 + ln 2
+ ln
better is to make its decoupling manifest: decoupling subtraction.
• Can have decoupling and the convenience of MSbar running by

using MSbar for EFT with electrons and muons above muon mass;
MSbar for EFT with electrons only between m and M.
• Match the coupling constant across the thresholds as particle is
integrated out.
• Decoupling subtraction:
αMS(μ = m) = αphys
If m < mu < M:
2
∂α α
αMS(μ) αphys 3π ( μ02 )
2 1 1 1 μ2
μ MS
=− MS
= − ln
∂μ 2 3π
If mu > M
2
αMS(μ) αphys 3π ( m ) 3π ( m 2 )
∂α 2α 1 1 1 M2 2 μ2
μ 2 MS = − MS
= − ln 2
− ln
∂μ 2 3π

WEAK INTERACTIONS
• The weak interactions were a starting point for understanding EFTs.
Integrating out the W boson leads to the Fermi lagrangian
ℒsm ⊃ gWμψγ μγLψ+c.c. GF g2

=
2 8M 2
ℒF = 2GF ψγμγLψ ψγ μγLψ W
• Do not also expect weak interactions to get corrections proportional

to smaller masses as lighter fields are integrated out.
• Selection rules (parity, flavour transformations, etc) always require
at least one W propagator, so effective interactions need not
always be dominated by the lightest particle integrated out.
WEAK INTERACTIONS
• The possibility of having lighter masses can lead to surprises,
however. eg:
σ(νν → γγ) ∼ GF4E 6
( 4π ) me8
3
α GF2E 10
σ(νν → γγγ) ∼

WEAK INTERACTIONS
• Integrating out the W boson gives
(1) (2) ab μλ ν
ℒeff
ν1,2γ = Cab
M ab μν
μν F + Cab
Mμν F Fλ + ℒF
ab
Mμν := iν aγμγL∂νν b − i∂νν aγμγLν b
MW2
3 (μ )
(2)
2 2 α GF 4
Cab (μ) = 1 + ln δab
πMW2 2
• Symmetric derivative on neutrino leads to redundant operators

• Chirality requires odd number of gamma matrices
WEAK INTERACTIONS
• Evolving down to lower scales focus on graph involving only one
factor of Mw-2
90π me ( 2 )
e vab α GF
[ )]
eff ab μν λρ ab νλ ρμ
ℒν3γ = 4
5 (Nμν F )(Fλρ F ) − 14 (Nμν F Fλρ F
( 2 )
1
vab := vab ee(μ = me) = U*
ea Ueb + δab − + 2sw2
ab
Nαβ = ∂α(ν aγβγLν b) − (α ↔ β)
• Redundant for 1,2 photons since involves derivative of neutrino current

EFTS IN PARTICLE PHYSICS
Unknown unknowns

UNKOWN UV THEORY
• Low-energy degrees of freedom can be qualitatively
different from high-energy ones
• e.g. pions, or atoms, or planets can be ‘elementary’ fields at
low energies while their constituents are ‘elementary’ at high
energies
1
N N 𝐿𝑒𝑓𝑓 = 𝜕𝜇𝜋(𝑁𝛾 𝜇𝑁)
𝐹
π if E < 4π F ~ 1 GeV

TECHNICAL NATURALNESS
• The SM is most general renormalizable theory built from
given particle content and gauge symmetry
• smells like a low-energy EFT
• But SM also contains relevant interactions (those that get larger

at low energies) like
ℒSM ⊃ − ζ + w 2H †H
Is this a problem?

• Imagine embedding the SM into some UV theory, for
simplicity take it simply to be a singlet scalar, S, of mass M
• Compute the Higgs mass both in the EFT (the SM) below M and in the
UV theory above M
1 2 1 2 2 1 2 2 †
ℒ = ℒSM − (∂S) − M S − g S H H + ⋯
2 2 2
( μ2 )
2 2 2
g M M
mH2 = 2whe
2
(μ) + (SM loops) − ln
8π 2
= 2wle2 (μ) + (SM loops)

• The effective constant wle is order the weak scale always, while
whe is order M, everywhere except precisely at mu = M
• The same holds for higher thresholds: must adjust initial UV

coupling with high precision to arrive at low energies with the
SM value.
• This is not how hierarchies of scale usually work: normally if a

parameter is small, its small size can be understood at any scale
one chooses to ask: eg why are atoms larger than nuclei?
• “Technically natural”
understanding of why a
parameter is small:
α̂m̂ e ≪ ΛQCD
• why is it small in the M ~ ΛQCD
mμ
( me )
UV theory?
• why does it stay δme ∼ α ln
M ~ mµ
small as one
integrates out scales
between the UV and αme ≪ mp M ∼ me
measurement scale?
• The small size of the
Higgs mass relative to Mp ~1018 GeV
Planck scale is not w ∼ M̂
automatically
technically natural:
• add fermions w∼M M ~ 1011 GeV
• make Higgs
composite
• deny Mp is a scale w ∼ mH Mw ~ 102 GeV

• Threat to the
naturalness argument: mw ~1011 eV
cosmological
ζ ∼ MW4
constant is small, and
the scales where it is mµ ~ 108 eV
unnatural are well
understood. ζ∼ me4 me ~ 106 eV
gµν mν ∼ 10-2 eV
ζ ∼ (0.01eV )4
gµν
e e
γ

Burgess NordicSchool2019

Uploaded by

Copyright:

Available Formats

Burgess NordicSchool2019

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Burgess NordicSchool2019

Uploaded by

Copyright:

Available Formats

INTRODUCTION

Nordic Winter School 2019

• Decoupling: most details of small-distance physics are not needed when

Nordic Winter School 2019

Nordic Winter School 2019

• Renormalizable: don’t need

Nordic Winter School 2019

• Renormalizable: higher orders

∫ ( (2π)4 ) (p + q + m)6(p + q)2

Nordic Winter School 2019

Nordic Winter School 2019

Nordic Winter School 2019

Nordic Winter School 2019

Nordic Winter School 2019

2mR [ 1 + 2p ⋅ q/mR 1 − 2q ⋅ q′ /mR 1 − 2p ⋅ q′ /mR2 ]

• Amplitude suppressed by (E/mR)4 in low-energy limit. What if E/mR smaller

• Obvious once recognized at low energies all massless scattering (also

• Expand in powers of E/M as early as possible.

• Make symmetries manifest in this effective lagrangian

Nordic Winter School 2019

Nordic Winter School 2019

• Low-energy observables can be computed from

• Want to efficiently identify effects of heavy physics in powers of 1/M.

Nordic Winter School 2019

• Define Wilson action:

• Any Λ dependence of SW cancels in < L(x1)…L(xn) >.

exp (SW[L]̂ ) = 𝒟Ĥ exp (iS[L,̂ H]̂ )

Nordic Winter School 2019

Nordic Winter School 2019

exp (iSW[L,̂ Λ]) = 𝒟Ĥ exp (iS[L,̂ H]̂ )

• Any Λ dependence of SW cancels in < L(x1)…L(xn) >: nothing depends

Nordic Winter School 2019

SW[ξ + θ] = SW[ξ] ⇒ SW = SW[∂ξ]

∂μ {∂ μξ [1 − 4Geﬀ (∂ξ)2 + ⋯]} = 0

e.g. ℒ = G1 (∂μξ) □ ∂ μξ or G2 ∂μ∂νξ ∂ μ∂νξ

• These differ only by a total derivative (so are not independent)

if S[ξ] = S0[ξ] + ϵS1[ξ] + ⋯ then when δξ = ϵ F[ξ]

∫ (2π)D [ (p 2 + q 2)B ] (4π)D/2 [ ]

• Divergences arise as poles as D goes to 4

Nordic Winter School 2019

• Renormalize in any convenient way (eg minimal subtraction)

Nordic Winter School 2019

• Need an algorithm to systematically identify which interactions and

Nordic Winter School 2019

• Interactions with no derivatives (ie scalar potential) are potentially

Nordic Winter School 2019

Nordic Winter School 2019

Nordic Winter School 2019

Nordic Winter School 2019

Nordic Winter School 2019

• For q ~ m the kinematics becomes non-relativistic and so path

• Leads to different form of low energy theory (eg NRQED or NRQCD or

Nordic Winter School 2019

Nordic Winter School 2019

• Integrate out the electron

Nordic Winter School 2019

• There are also redundant operators

∂μF μν∂ λFλν ∝ e 2J ν Jν Fμν □ F μν ∝ eF μν∂μJν ≃ − e 2J μJμ

• Corrections toα macroscopic classical E+M given by powers of E/m

• In modified minimal subtraction remove just first two terms

INTRODUCTION  