Aspects of QCD Dynamics From String Theory: NSF-ITP-98-089 CERN-TH/98-298 Hep-Th/9809106
Aspects of QCD Dynamics From String Theory: NSF-ITP-98-089 CERN-TH/98-298 Hep-Th/9809106
Aspects of QCD Dynamics From String Theory: NSF-ITP-98-089 CERN-TH/98-298 Hep-Th/9809106
CERN-TH/98-298
hep-th/9809106
a
Institute for Theoretical Physics
University of California
Santa Barbara, CA 93106
b
Theory Division, CERN
CH-1211, Geneva 23, Switzerland
Abstract
September 1998
1 Introduction
1
In section 6 we make a detailed comparison with lattice and field theory results.
Section 7 is devoted to a discussion and conclusions where we also present analogous
results for QCD3 .
The starting point for the construction of QCD4 is the superconformal theory in six
dimensions realized on N parallel coinciding M5-branes. Compactification of the theory on
a circle of radius R1 gives a five-dimensional theory whose low-energy effective theory is the
maximally supersymmetric SU(N) gauge theory with gauge coupling constant g52 = 2πR1 .
In order to describe QCD4 , one compactifies this theory further on another S1 of radius R2 .
The gauge coupling constant gY M in four dimensions is given by gY2 M = g52 /R2 = 2πR1 /R2 .
In order to break supersymmetry, one imposes the anti-periodic boundary condition on
the fermions around the second S1 .
As conjectured in [1], the large N limit of the six-dimensional theory is M theory on
AdS7 × S4 . Upon compactification on S1 × S1 and imposing the anti-periodic boundary
conditions around the second S1 , we find M theory to be on the black hole geometry
[2]. Taking the large N limit while keeping the ’t Hooft parameter 2πλ = gY2 M N finite
requires taking R1 R2 . In this limit, M theory reduces to type IIA string theory and
the M5 brane wrapping on S1 of radius R1 becomes a D4 brane, and the large N limit of
QCD4 becomes string theory on a black hole geometry.
2
The near horizon limit of the black M5-brane metric is given by [5, 6]
" 5
#
2y2 y06 2 X 2 4L2 y06 −1 2
ds = (1 − )dt + dx + (1 − ) dy + L2 dΩ24 , (2.4)
L2 y6 i=1
i
y 2 y 6
√ 3
H = 3 6L dΩ4 (2.5)
where
κ211 4πL2 T
L9 = N 3
, y 0 = . (2.6)
27 π 5 3
and dΩ4 is the volume form on the 4-sphere. Compactifying x5 on a circle and using the
M theory/Type IIA duality one gets the Type IIA background corresponding to a non
extremal D4 brane with worldvolume coordinates x1 , .., x4 , t
" 4
# 3
y3 y6 X 4L y6 y 2
ds = 3 (1 − 06 )dt2 +
2
dx2i + (1 − 06 )−1 dy 2 + Ly dΩ24 , eφ = , (2.7)
L y i=1 y y L
where φ is the dilaton field.
2.2 Fluctuations
As will be discussed in the next subsection, in order to compute the correlation functions
of QCD operators using supergravity/gauge theory correspondence, we need to examine
the dynamics of small fluctuations of the supergravity fields. We will be interested in
scalar operators of QCD4 . The relevant fields of Type IIA supergravity are the dilaton φ,
the volume factor, and the time component of the RR 1-form C0 . There will be a mixing
between the type IIA dilaton and the volume factor which must be disentangled. A useful
parameterization of fluctuations f (x1 , y), g(x1 , y), and h(x1 , y), which in the language of
eleven dimensional supergravity is appropriately diagonal, is
X 1 X4
2 4 2 2
ds = 1 − g(x1 , y) ġii dx dx + 1 − f (x1 , y) − g(x1 , y)
i i
ġii dxi dxi
5 i=0 5 5 i=2
2
6 4
+ 1 + f (x1 , y) − g(x1 , y) ġ55 dx5 + Ci dxi
5 5
4
+ 1 − g(x1 , y) ġyy dy 2 + (1 + g(x1 , y))ġΩΩ dΩ2 (2.8)
5
where ġ is the background metric. We consider the case C0 = h(x1 , y) and Ci6=0 = 0.
Without loss of generality, f (x1 , y), g(x1 , y), and h(x1 , y) are assumed to depend only on
x1 and y. In the language of Type IIA supergravity, these fluctuations are parameterized
according to
X1 X4
2 3 6 i i 1 6
ds = 1+ f (x1 , y)− g(x1 , y) ġii dx dx + 1+ f (x1 , y)− g(x1 , y) ġii dxi dxi
5 5 i=0 5 5 i=2
3
3 6 3 3
+ 1 + f (x1 , y) − g(x1 , y) ġyy dy 2 + 1 + f (x1 , y) + g(x1 , y) ġΩΩ dΩ2
5 5 5 5
9 3
eφ = 1 + f (x1 , y) − g(x1 , y) eφ̇ ,
10 5
C0 = h(x1 , y). (2.9)
4 −2 00 3 −3 0 9M 2
(r − r )f (r) + (7r − r )f (r) + 2 2 f (r) = 0,
4π T !
2
4 −2 00 3 −3 0 9M 2
(r − r )g (r) + (7r − r )g (r) + g(r) − 72r = 0,
4π 2 T 2
9M 2
(r 4 − r −2 )h00 (r) + (7r 3 − 7r −3)h0 (r) + h(r) = 0. (2.11)
4πT 2
The string/gauge theory correspondence asserts that the string modes ϕ couple at the
boundary to gauge invariant operators O of the gauge theory. The generating functional
for the correlation function of O is the string partition function evaluated with specified
boundary values ϕ0 of the string fields. When the supergravity description is applicable
we have R 4
he− d xϕ0 (x)O(x) i = e−ISG (ϕ0 ), (2.12)
where ISG is the supergravity action.
The supergravity fluctuations f, g, h couple to operators on the D-brane world volume
in flat background. For small ’t Hooft parameter, this coupling can be inferred from the
form of the Born-Infeld action. These same operators, in the strong coupling limit, are
described as “sources at infinity” in the dual AdS description.
The Born-Infeld Action describing the D4 brane worldvolume theory is given by :
Z q
1
IBI = TrT4 d5 x e−φ det(Gµν + Fµν + Gij ∂µ Xi ∂ν Xj ) + µνσρκ Cµ Fνσ Fρκ , (2.13)
8
where i = 5 . . . 9 and T4 is the D4 brane tension.
Dimensionally reducing along the t direction on a circle with period 1/T leads to
Z q q
T4 1
IBI = Tr d4 x e−φ G00 det(Gµν + Fµν + Gij ∂µ Xi ∂ν Xj ) + 0µνσρ C0 Fµν Fσρ ,
T 8
(2.14)
4
where now, i = 0, 5 . . . 9. In order to restore the canonical dimensions of Yang-Mills
fields, we rescale F and X:
F = T −1 F, X = T −1 Φ, (2.15)
where T −1 = 2πls2 .
Expanding the action (2.14) to quadratic order gives
Z
4 T4 −φ q q T4 0µνλσ
I = Tr dx 2
e G 00 detGµν F 2 + C0 Fµν Fλσ , (2.16)
4T T 8T 2 T
where we dropped the scalars since they are expected to decouple from the dynamics in
the far infra-red. Using [7]
we get
T4 1 N
= = , (2.18)
T 2T (2π)2 gls T 2πλ
where in the last equality, we set 2πgs ls = (λ/N)T −1 which follows from M theory/Type
IIA duality. Before substituting these fluctuations into (2.16), note that the metric fluctu-
P
ation due to field f (x1 , y) is polarized along (3(dx1 )2 + 4i=2 (dxi )2 ) which is not isotropic
in the 1234 plane. One could however think of this as contribution from two pieces
P
( 4i=1 (dxi )2 ) and (dx1 )2 . The second term must then couple to T11 (p1 ) component of the
stress-energy tensor which vanishes due to conservation law pµ Tµν = 0. We will therefore
think of fluctuation f (x1 , y) as giving rise to a fluctuation of the metric of the form
X4
1 6
ds2 = 1 + f (x1 , y) − g(x1 , y) ġii dxi dxi
5 5 i=1
along the 1234 plane. Substituting these fluctuations into (2.16), we find that f (x1 , y),
g(x1 , y), and h(x1 , y) couple to the world volume operators via
Z
1 N 3 1 N
I = Tr d4 x 1 − f (x1 , y) F 2 + h(x1 , y)Fµν Fλσ µνλσ . (2.19)
4 2πλ 5 8 2πλ
Note that all dependence on g(x1 , y) canceled out to this order. Indeed, the AdS/SCFT
correspondence in six dimensions imply in view of (2.11) that g(x1 , y) couples in six
dimensions to an operator of dimension 12 in the (0,2) theory. This operator is H 4 where
H = dB, which reduces to a dimension 8 operator in the four dimensional theory in the
low energy limit. Comparing (2.19) to the Yang-Mills action
Z !
4 1 θ
IY M = Tr dx 2
F2 + F F̃ , (2.20)
4gY M 16π 2
5
JPC 1 2 3 4 5 6
f (x1 , y) 0++ 9.85 15.6 21.2 26.7 32.2 37.7
g(x1 , y) 0++ 22.5 28.8 34.9 40.7 46.5 52.1
h(x1 , y) 0−+ 11.8 17.8 23.5 34.6 34.6 40.1
Table 1: Mass spectrum in units of T for 0++ and 0−+ glueballs corresponding to nor-
malizable modes of fluctuations f (x1 , y), g(x1 , y), and h(x1 , y).
3 Mass Spectrum
The glueball masses in QCD can be obtained by computing correlation functions of gauge
invariant local operators or the Wilson loops, and looking for particle poles. Following
[8, 9] correlation functions of local operators are related at large N and large gY2 M N to
tree level amplitudes of supergravity. We will consider the two point functions of the
operators (2.22). The correlation function of O takes the form
X
hO(x)O(y)i = ci e−Mi |x−y| , (3.1)
i
where O refers to (2.22), and Mi are the corresponding glueball masses. Therefore the
spectrum of scalar glueballs2 0++ and 0−+ can be obtained by finding the normalizable
solutions to (2.11) for f (x1 , y) and h(x1 , y) respectively. The results are summarized
in table 1, where the masses are given in units of T . We also included the spectrum
associated with the field g(x1 , y).
We can compare the ratio of masses of the lowest glueball states 0−+ and 0++ with
the lattice results [10]
M0−+
= 1.20
M0++ supergravity
M0−+
= 1.36 ± 0.32, (3.2)
M0++ lattice
2
In the following we will use the notation J P C for the glueballs, where J is the glueball spin, and P ,
C refer to the parity and charge conjugation quantum numbers respectively.
6
and the results are in good agreement.
The spectrum of masses related to g(x1 , y) is higher than the spectrum of masses
related to f (x1 , y) since it corresponds to a higher dimensional operator. Unlike QCD
where different operators with the same quantum numbers give the same mass spectrum,
in the supergravity limit the corresponding mass spectrum is different. Since a TrF 4
operator couples on the worldvolume theory to both f and g modes we expect that in the
QCD limit the spectrum of masses associated with g will decouple in order for TrF 2 and
TrF 4 to give the same spectrum.
Let us make some comments on the results. In comparison to [4] the masses of 0−+
are lower. This is due to the fact that in [4] the computation of the spectrum of 0−+ was
done using the RR 3-form which couples to an operator with 0−+ quantum numbers but
of higher dimension than TrF F̃ . An example of that is the 0++ spectrum derived above
using the f and g modes. Also, the masses of 0++ are not in agreement with [4]. This
is because the supergravity equation used in [4] to compute the 0++ mass spectrum does
not precisely correspond to TrF 2 which is the lowest dimension glueball operator with
0++ quantum numbers.
Finally, in order to compute the coefficients ci in (3.1) one has to evaluate the action for
the supergravity fields requiring that they approach a boundary configuration at infinity.
This is a procedure that can be done numerically, but will not be reported here.
4 Topological Susceptibility
In this section we will compute the topological susceptibility in the supergravity de-
scription. The discussion is closely related to the study of the θ dependence of the vacuum
energy in [13]. Working in Type IIA supergravity, the action of RR 1-form Cµ is given by
Z
1 √ 1 0 0
I= d10 x g (∂µ Cν − ∂ν Cµ )(∂µ0 Cν 0 − ∂ν 0 Cµ0 )g µµ g νν . (4.3)
2κ210 4
7
For the fluctuation Ct = h(x1 , y), the action simplifies to
Z
1 √ 1 0
I= 2 d10 x g (∂µ h(x1 , y))(∂µ0 h(x1 , y))g µµ g tt . (4.4)
2κ10 2
The solution to the corresponding equation of motion (2.11) which has dependence only
in the y coordinate is
y6
h(y) = h∞ (1 − 06 ). (4.5)
y
Plugging this into the action gives
Z
1 √ 1 2π 2 y 6
I= d10 x g g yy g tt (∂y h(y))2 = h∞2 2 30 dx4 . (4.6)
2κ210 2 κ10 L T
Using (2.6), (2.3), dropping the volume factor and differentiating twice with respect to
h∞ we get
Z
128λN 2 π 3 T 4
dx4 hÕ4 (x)Õ4 (0)i = . (4.7)
729
Recalling (2.22) and (4.1) we get the topological susceptibility
Z !2
4 λ 32λ3πT 4
χt = dx hÕ4 (x)Õ4 (0)i = . (4.8)
2πN 729
5 Gluon condensate
The gluon condensate h 4g21 F 2 i plays an important role in the applications of Wilson
YM
OPE to the study of strong coupling phenomena in QCD via the condensate expansion.
In this section we will compute the gluon condensate in the supergravity description.
According to the supergravity/gauge theory correspondence (2.12), the one point func-
tion of an operator corresponds to the first variation of the supergravity action. This
quantity usually vanishes identically by the equation of motion. In asymptotically anti de
Sitter spaces, however, there is a subtlety. The first variation is only required to vanish
up to a total derivative term. Since asymptotically anti de Sitter space has a time-like
boundary at infinity, one must carefully examine the possible boundary contribution. In-
deed, the one point function of F 2 operator, corresponding to first variation of the action
with respect to the field f (x1 , y), will turn out to give a non-vanishing contribution.
This result can be anticipated without any detailed computation. The non-extremal
M5-branes has non-vanishing action density I (which is related to free energy F by I =
T −1 F )
64 3 3
I = 26 3−7 π 3 N 3 V5 T 5 = π N V5 T 5 . (5.1)
2187
8
When the M5-brane is wrapped on a circle of period (λ/N)T −1 , we expect the corre-
sponding free energy to be given simply by making the substitution V5 → V4 T −1 λ/N
which leads to an expression for the free energy of non-extremal D4-branes:
64 3 2
I = 26 3−7 π 3 λN 2 V4 T 4 = π λN V4 T 4 . (5.2)
2187
The gluon condensate operator can be gotten by acting on the Born-Infeld action with
∂
λ ∂λ . This suggests that one can act on I with the same operator to infer the value of
h 4g21 F 2 i in the dual formulation:
YM
1 ∂ 64 3
h 2
F 2i = λ I = π λN 2 V4 T 4 . (5.3)
4gY M ∂λ 2187
6 Comparison
M0++ = 9.86T
M0−+ = 11.8T
16π 2 2
QCD string tension = λT
27
32π 3 4
χt = λT
729
1 64π 3 2 4
h 2 TrF 2 (0)i = N λT . (6.1)
4gY M 2187
These results are the leading in the strong coupling expansion of the gauge theory.
The lattice QCD results are [10]4
9
The entry for the gluon condensate is derived by scaling the N = 3 result of [16] by a
factor of N/3.
N αs 1
0.1GeV4 = h G2 i = h 2 F 2 i. (6.3)
3 π 2π
Note that plugging the phenomenological values Nf = 3, N = 3, mη0 ∼ 1 GeV, fπ ∼
0.1 GeV in (4.2) leads to a prediction χt ∼ (180 MeV)4 , in agreement with the lattice
simulation.
The supergravity results depend on two parameters, λ and T . This should be thought
of as the leading 1/λ asymptotic behavior of the full string theory expression (F (λ)T )d
where d is the dimension of the observable. Ideally, the comparison with the lattice QCD
result should be made in the limit where λ is taken to be small and T is taken to be large.
In this range of parameters, one can trust perturbative QCD, which tells us that F (λ)
goes as
12π
F (λ) = Ce− 11λ . (6.4)
where C is some numerical coefficient. The genuine QCD data is encoded in the ratio of
this numerical coefficient C between different observables. Unfortunately, this is precisely
the information that is not accessible via the supergravity approximation. In the absence
of fully string theoretic formulation of this system that allows one to probe the small
λ region, one must contend oneself with qualitative comparisons. To facilitate such a
comparison at a qualitative level, it is useful to go to the so-called “correspondence point”
in the λ-T parameter space. This is the point where dual descriptions of the theory
crosses over from one to the other. As such, one can not trust neither description at a
quantitative level, but one expects the results to match up to numerical factors of order
one. One can estimate the cross-over point based on natural scale of our problem: λ ≈ 12π 11
and T ≈ ΛQCD ≈ 200 MeV. With this choice of λ and T , the ratio of observables in units
of energy comes out as follows:
M0SUGRA
++
Lattice
= 1.2
M0++
M0SUGRA
−+
= 1.1
M0Lattice
−+
!1/2
(QCD string tension)SUGRA
= 2.0
(QCD string tension)Lattice
!1/4
χSUGRA
t
= 1.7
χLattice
t
!1/4
(Gluon condensate)SUGRA
= 0.9
(Gluon condensate)Lattice
10
It is encouraging to find that these ratio are indeed of order unity 5 .
It is curious to note for comparison the known results in strong coupling lattice QCD
[17]:
They reflect the fact that the α0 expansion is not the same as the lattice strong coupling
expansion. In particular, while in the supergravity description the ratio of M0++ to the
QCD string tension goes to zero, it diverges in the strong coupling lattice QCD.
In this paper we computed various physical quantities which are of importance in QCD
dynamics. The computations were performed in the framework of supergravity. Since
the supergravity description is believed to provide the leading contribution in the strong
coupling expansion of the gauge theory, our results should be viewed as a first step in
understanding the QCD dynamics via the string/gauge theory correspondence. Clearly,
a complete solution requires a detailed knowledge of the string dynamics with R-R back-
ground.
The dual supergravity description corresponds to taking the bare coupling to infinity.
This is the opposite limit to continuum QCD where the bare coupling is taken to zero.
Therefore it is not possible, for instance, to exhibit asymptotic freedom via the super-
gravity description. In analogy with strong coupling lattice QCD, it is therefore far from
clear what properties of continuum QCD should be captured by supergravity. The su-
pergravity description includes also Kaluza-Klein modes that do not have corresponding
degrees of freedom in QCD. They reflect the physics of the higher dimensions and should
decouple in the small bare coupling limit. Since the background is singular in this limit,
their decoupling is a subtle issue.
We can repeat the analysis for QCD3 . Here the procedure is to compactify on a circle
the D3 branes of type IIB string theory and introduce anti periodic boundary conditions
5
The numerical value for the ratio of the QCD string tensions is modified compared to a previous
version of this paper in view of the change in the numerical value of the QCD string tension in (6.1).
11
for the fermions [2]. We find6
M0++ = 10.7T
s
1 2
QCD string tension = π g NT 3
2 Y M3
1 π2 2 3
h TrF 2 (0)i = N T . (7.1)
4gY2 M 3 8
The gluon condensate in QCD3 has not been computed on the lattice yet.
Acknowledgement
References
[1] J. Maldacena, “The Large N limit of superconformal field theories and
supergravity,” Adv. Theor. Math. Phys. 2 (1998) 231, hep-th/9711200.
[2] E. Witten, “Anti-de Sitter space, thermal phase transition, and confinement in
gauge theories,” Adv. Theor. Math. Phys. 2 (1998) 505, hep-th/9803131.
[3] D. J. Gross and H. Ooguri, “Aspects of large N gauge theory dynamics as seen by
string theory,” Phys. Rev. D58 (1998) 106002, hep-th/9805129.
[4] C. Csaki, H. Ooguri, Y. Oz, and J. Terning, “Glueball mass spectrum from
supergravity,” JHEP 01 (1999) 017, hep-th/9806021.
[5] M. J. Duff, H. Lu, and C. N. Pope, “The Black branes of M theory,” Phys. Lett.
B382 (1996) 73–80, hep-th/9604052.
12
[7] I. Chepelev and A. A. Tseytlin, “Long distance interactions of D-brane bound
states and longitudinal five-brane in M(atrix) theory,” Phys. Rev. D56 (1997)
3672–3685, hep-th/9704127.
[9] E. Witten, “Anti-de Sitter space and holography,” Adv. Theor. Math. Phys. 2
(1998) 253, hep-th/9802150.
[10] M. J. Teper, “Physics from the lattice: Glueballs in QCD: Topology: SU(N) for all
N,” hep-lat/9711011.
[11] E. Witten, “Current Algebra Theorems for the U(1) Goldstone Boson,” Nucl. Phys.
B156 (1979) 269.
[12] G. Veneziano, “U(1) Without Instantons,” Nucl. Phys. B159 (1979) 213–224.
[17] M. Creutz, Quarks, Gluons and Lattices. Cambridge University Press, Cambridge,
UK, 1983.
13