Professional Documents
Culture Documents
You may find the metric for AdSd+1 in Poincaré coordinates useful:
L2 L2
ds2 = 2 µ ν
ηM N dxM dxN ,
dz + η µν dx dx = (1)
z2 (xd )2
where η may be either the Minkowski metric or Kronecker delta, depending on the signature you
prefer. In the latter equation z = xd . The boundary is at z = 0 (with a single “point at infinity”
z = ∞). Notationally, we will use Greek indices µ to run over boundary coordinates and capital
Latin indices M to run over all AdS coordinates, unless otherwise specified.
Solution: 2 2
p The key is that the mass of the AdS5 scalars is given by L m = H(k/2, k/2, k) +
16 − 8 H(k/2, k/2, k) + 4, where H(k/2, k/2, k) = 6[2(k/2)(k/2 + 1) − k 2 /8] = 9k 2 /4 + 6k.
Then we get
r
2 2 9 2 9 2 9 2 9 2 3 3
L m = k +6k+16−8 k + 6k + 4 = k +6k+16−12k−16 = k −6k = k k−4 .
4 4 4 4 2 2
The dimension of the dual operator, as we know from our notes, is (in a d-dimensional field
theory/AdSd+1 )
r r
d d2 2 2
9 2 3
∆+ = + +L m =2+ k − 6k + 4 = k .
2 4 4 2
That’s just what we want.
λ → γλT γ −1 → γ(γ −1 )T λγ T γ −1 .
Since this must = λ, we find γ T = γ. We already know that γ must be unitary to respect
the CP factor inner product.
(b) [3 points] As it turns out, string oscillators are odd under the orientifold reflection, so the
allowed massless states, including the gauge bosons, must satisfy γλT γ −1 = −λ. In a basis
where γ is the identity, show that exp[iθλ] for some parameter θ is a 2N × 2N orthogonal
matrix, so the gauge bosons are proportional to generators of the orthogonal group.
Solution: In this basis, we have λ = −λT . Therefore, if we say O = exp[iθλ], we know
OT = exp[iθλT ] = exp[−iθλ] = O−1 . Hence O is orthogonal. That means λ is a generator
of SO(2N ), so that’s the gauge group.
Page 2
where the overall sign depends on signature (Lorentzian vs Euclidean). This theory is used
widely in holographic superconductor models.
Write the correlator hŌ(x1 )O(x2 )Ō(x3 )O(x4 )i to order g 2 in terms of the scalar bulk-boundary
and bulk-bulk propagators K(z, x; y) and G(z, x; z 0 , x0 ) respectively and the vector bulk-boundary
and bulk-bulk propagators KM (z, x; y) and GM N (z, x; z 0 , x0 ). Do not substitute forms for the
propagators or attempt to evaluate any bulk integrals. You may also ignore boundary terms
in the action.
Solution: The correlator will be the functional derivative of the action evaluated on-shell with
respect to sources ψ̄0 (x1 ), ψ0 (x2 ), ψ̄0 (x3 ), ψ0 (x4 ). To find this, we start by finding the equations
of motion. We take a small shortcut by defining the covariant derivative Dψ = ∇ψ − igAψ,
since this integrates by parts (in all gauge invariant terms) and follows the product rule, etc,
just like a derivative. Then the equations of motion are
∇N F N M + ig(DM ψ̄)ψ − ig ψ̄DM ψ = 0
D 2 ψ − m2 ψ = 0 .
Integrating by parts and dropping boundary terms gives us (taking the Lorentzian sign for
definiteness)
Z
p 1
S = · · · + dd+1 x |g| AM ∇N F N M + ψ̄D2 ψ − m2 |ψ|2
2
Z
d+1
p ig M M
= · · · + d x |g| AM ψ̄D ψ − (D ψ̄)ψ .
2
Next, we recognize that a perturbative solution to the AM equation of motion in the presence
of boundary sources for ψ, ψ̄ (and no vector source) is
Z
AM (z, x) = 0+ig dd+1 x0 |g| GM N (z, x; z 0 , x0 ) ψ̄(z 0 , x0 )DN ψ(z 0 , x0 ) − (DN ψ̄(z 0 , x0 ))ψ(z 0 , x0 ) +· · · ,
p
where the scalar fields are the classical solutions in the presence of sources
Z
ψ(z , x ) = dd y K(z 0 , x0 ; y)ψ0 (y) + · · ·
0 0
and conjugate. Now we can assemble everything together into the part of the action propor-
tional to ψ̄0 ψ0 ψ̄0 ψ0 ; note that only terms with partial derivatives on ψ in the on-shell action
contribute to this order. Then there are two ways to carry out the functional derivatives, so
we end up with
g2 dzdd x dz 0 dd x0 M N
Z Z
hŌ(x1 )O(x2 )Ō(x3 )O(x4 )i = − G (z, x; z 0 , x0 )
2 z d+1 z 0d+1
× K̄(z, x; x1 )∂M K(z, x; x2 ) − K(z, x; x2 )∂M K̄(z, x; x1 )
× K̄(z 0 , x0 ; x3 )∂N K(z 0 , x0 ; x4 ) − K(z 0 , x0 ; x4 )∂N K̄(z 0 , x0 ; x3 )
+ K̄(z, x; x3 )∂M K(z, x; x2 ) − K(z, x; x2 )∂M K̄(z, x; x3 )
0 0 0 0 0 0 0 0
× K̄(z , x ; x1 )∂N K(z , x ; x4 ) − K(z , x ; x4 )∂N K̄(z , x ; x1 )
Page 3
4. The Bulk-Boundary Vector Propagator adapted from McGreevy
Witten (see our reading from hep-th/9802150) gave a clever derivation of the scalar bulk-
boundary propagator in the AdS Poincaré coordinates. In this problem, you will find the
bulk-boundary propagator for a vector field AM in AdS. You may set the AdS scale L = 1 and
work in either signature.
(a) [5 points] Assuming that the vector field in question has the usual Lagrange density
−(1/4)FM N F M N , write the equation of motion for Aµ that is a function of z only. Then
find the solution that vanishes as z → 0 for all values of xµ . Assume that Az = 0 and that
Aµ is proportional to some constant vector µ .
Solution: We saw on assignment 1 that this action can be written as
Z Z
1 d+1
p MN 1
S=− d x |g| FM N F =± F ∧ ?F ,
4 2
with the sign depending on the signature. Either way, the equation of motion in form
language is d ? F = 0. Now, with these conditions, the only non-vanishing compo-
nent is Fzµ = −Fµz = ∂z Aµ . Therefore, the nontrivial component of the EOM is the
(zµ1 · · · µd−1 ) component, which is (up to overall sign)
∂z z −(d+1) z 4 νµ1 ···µd−1 ∂z Aν = 0 ,
(b) [10 points] The bulk-boundary propagator for a vector field satisfies the vector equation
of motion at every point xM in the bulk and is ∝ ν δ d (x − y) for z → 0, where y µ is a
specified boundary point and ν is a fixed polarization vector. Your solution from part
(a) is therefore the bulk-boundary propagator when the boundary point is the point at
infinity. Use Witten’s inversion trick and translation invariance in the y µ to find the general
bulk-boundary propagator up to an overall normalization constant for a given polarization
vector ν . Hint: remember that AM is a vector and must transform as such under the
inversion diffeomorphism.
Page 4
Therefore, the vector bulk-boundary propagator is (dropping the bar for convenience)
z d−2 1 2xν xM
ν
KM (z, x; 0) = C δ − ν ,
x·x x·x M x·x
z d−1
Kz (z, x; y) = −2C (x − y)λ λ
(z 2 + ηµν (x − y)µ (x − y)ν )d
z d−2 2(x − y)µ (x − y)ν
ν
Kµ (z, x; y) = C 2 δµ − 2 ν .
(z + ηµν xµ (x − y)ν )d−1 (z + ηµν (x − y)µ (x − y)ν )
1−z 4 /z04
Page 5
and is conserved. Now, at the maximum value z(x) = z̄, we know ∂x z = 0, so −2πα0 H/T =
1/z̄ 2 . Solving, we find (∂x z)2 = (z̄ 4 /z 4 − 1)(1 − z 4 /z04 ).
(b) [8 points] Assume r z0 . Argue in this case that z̄ ≈ z0 and that z(x) ≈ z̄ except in a
small region of length δx ∼ z0 near x = ±r/2.
Solution: If we integrate the differential equation for z, we get the solution
z(x) z(x)/z̄
y2
Z Z
r dz
x+ = p = z̄ dy p
2 0 (z̄ 4 /z 4 − 1)(1 − z 4 /z04 ) 0 (1 − y 4 )(1 − y 4 z̄ 4 /z04 )
near the left endpoint. (The LHS is r/2 − x with the same RHS near the right endpoint.)
By symmetry, z(x = 0) = z̄, so the above gives us an implicit defintion of z̄ in terms
of r. In particular, when r z0 ≥ z̄ (since z > z0 does not exist), the y integral with
upper limit of 1 must become very large. We can see in fact that the integral diverges
like − ln(1 − y) near y = 1 when z̄ = z0 , so we see that the integral gets large for z̄ ≈ z0 .
In fact, if we substitute z̄ = z0 in the integral, we can solve it to find
δx 1 z(x) 1 z(x) 1 −1 z(x)
= − ln 1 − + ln 1 + − tan .
z̄ 4 z̄ 4 z̄ 2 z̄
δx 1 1 π
≈ ln(10) + ln(2) − ≈ 0.93 ,
z̄ 2 4 8
which means z(x) ≈ z̄ ≈ z0 except in a small region near the endpoints. (I will accept
many variations on that argument, as long as they are sound.)
(c) [8 points] Using the previous two parts of the problem, show that the quark/anti-quark
potential is given approximately by r/2πα0 z02 (once it is appropriately renormalized). This
is what we expect in a confining phase of a gauge theory. Hint: Having a good physical
idea of the shape of the string worldsheet is important, so sketch it if you need to.
Solution: The conserved quantity we found lets us write
s
(∂x z)2 z̄ 2
1+ = ,
1 − z 4 /z04 z2
so we can write
z̄ 2
Z
1
V (r) = S/T = dx .
2πα0 z4
To renormalize this potential, we are supposed to subtract off the action of the worldsheets
with x = ±r/2 constant for all values of z. On the other hand, we have learned from
the second part of the problem that the extremal worldsheet hanging from the Wilson
line goes from z = 0 to ∼ z0 very quickly at the endpoints and is basically constant
in between. Therefore, to a good approximation, the Wilson line is composed of the
two “constant x” worldsheet segments connected by a segment at constant z = z0 for
−r/2 < x < r/2 (ie, two vertical parts and a horizontal part). The renormalization just
subtracts the divergent “vertical” parts. This leaves V (r) ∼ r/2πα0 z02 .
Page 6