hep-th/9711165

IHES/P/97/83
RU-97-95
QMW-PH-97-34
LPTENS 97/56
arXiv:hep-th/9711165v3 28 Nov 1997

D-branes and the Noncommutative Torus

Michael R. Douglas1,2 and Chris Hull3,4

1
Institut des Hautes Études Scientifiques
Le Bois-Marie, Bures-sur-Yvette, 91440 France
2
Department of Physics and Astronomy
Rutgers University
Piscataway, NJ 08855–0849 USA
3
Physics Department, QMW
Mile End Road, London E1 4NS, U.K.
4
Laboratoire de Physique Théorique, Ecole Normale Supérieure,
24 Rue Lhomond, 75231 Paris Cedex 05, France.

[email protected], [email protected].

We show that in certain superstring compactifications, gauge theories on noncom-

mutative tori will naturally appear as D-brane world-volume theories. This gives strong
evidence that they are well-defined quantum theories. It also gives a physical derivation
of the identification proposed by Connes, Douglas and Schwarz of Matrix theory com-
pactification on the noncommutative torus with M theory compactification with constant
background three-form tensor field.

November 1997
 It is often stated that in toroidal compactification in superstring theory, there is a
minimum radius for the torus, the string length. Tori with smaller radii can always be
related to tori with larger radii by using T-duality.
However, this picture is not correct in the presence of other background fields, as is
clear from the following simple example. Consider compactification on T 2 with a constant
Neveu-Schwarz two-form field B. We quote the standard result [1]: under simultaneous
T-duality of all coordinates, the combination (G + B)ij is inverted, where G is the metric
expressed in string units. Consider a square torus and take the limit

R1 = R2 → 0; B 6= 0 fixed. (1)

The T-dual torus has  

1 R22 −B
G+B = 2 2
R1 R2 + B 2 B R12

and the T-dual radii go to zero as R̃1 = R̃2 ∼ R1 /B.

Of course the full T-duality group is larger and includes an SL(2, Z) acting on the
upper half plane B + iR1 R2 in the usual way. For any point there exists an SL(2, Z) trans-
√
formation which brings it to the fundamental region with R1 R2 ≥ 3/2, so that it would
appear that there is again a minimal length. However, the particular SL(2, Z) transfor-
mation that maps back to the standard fundamental region depends on R1 R2 and varies
wildly as we take the limit (1). In any situation involving objects which transform under
SL(2, Z), the physics of the limit is not well described by making these transformations.
A better way to get at the limit is to T-dualize a single dimension. This exchanges the
complexified Kähler form B +iR1 R2 and the complex structure of the torus τ = R2 /R1 , to
produce a theory with fixed volume and B = 0, while the torus becomes highly anisotropic:
τ ′ = B + iR1 R2 . Let us think of this torus as R2 quotiented by the two translations
x → x + 1 and x → x + τ ′ . In the limit R1 R2 → 0, the original T 2 appears to degenerate
to R ×S 1 with coordinate 0 ≤ x1 < 1 quotiented by an additional translation x1 → x1 +B.
The nature of such a quotient depends radically on whether B is rational or irrational; if
we interpret it in the latter case naively as a pointwise identification, it will not lead to a
conventional Hausdorff manifold.
This is exactly the situation which was the original motivation for noncommutative
geometry [2], leading us to try to interpret the resulting space as a noncommutative space.
Not only is this a sensible thing to do, we will now give a physical argument which leads
to gauge theory on the noncommutative torus as defined in [3].

1
 Consider a theory of N D0-branes on the original torus. After T-dualizing one dimen-
sion, this becomes a theory of N D1-branes stretching along the T-dualed dimension; let
them be extended in x1 , and placed at x2 = 0. The compactness of x2 will be implemented
by placing images of the D1-branes at x + nR2 for all integers n; in other words extended
in x1 and at x2 = nR2 .*
This is not a conventional U (N ) 1 + 1-dimensional gauge theory, because it contains
additional light states. A string winding about x2 will take the shortest path between a
D1-brane and its image consistent with the boundary conditions on the D1-brane, Dirichlet
in x2 and Neumann in x1 . This means that the string will end at a right angle and follow
a path (x1 , s) for 0 ≤ s < R2 . Its length is R2 and it is light, with mass m ∝ R2 , so it
must be kept in the low energy theory.
A string winding w2 times will have mass m ∝ w2 R2 , so we must consider fields
with arbitrary dependence both on x1 and integral w2 to get the complete low energy
spectrum. By the usual Fourier transform of w2 to x2 , we could rewrite this as a U (N )
2 + 1-dimensional gauge theory, in terms of fields Ai (x1 , x2 ).
However, this is not a conventional U (N ) 2 + 1-dimensional gauge theory either,
because the open string which starts at (x1 , 0) will end at (x1 , w2 R2 ) which is identified
with (x1 − w2 B/R1 , 0). The degrees of freedom do not live at points in 2 + 1-dimensional
space and thus the theory is not local in the conventional sense.
Still, it is not hard to guess what modification of the gauge theory is induced by
non-zero B. When two open strings interact, the end of the first must coincide with the
beginning of the second. We can express this in the (x1 , w2 ) basis by an interaction of the
form
X Z
Sint = dx1 φ1 (x1 , −w2 − w2′ )φ2 (x1 , w2 )φ3 (x1 − Bw2 /R1 , w2′ )
w2 ,w ′ 2
 
X Z
1 1 1 Bw2 ∂
= ′
dx φ1 (x , −w2 − w 2 )φ2 (x , w2 ) exp − φ3 (x1 , w′ 2 ).
R1 ∂x1
w2 ,w ′ 2

We then Fourier transform, and go to coordinates of unit radius σi ≡ Ri xi . The interaction

becomes
  
B ∂ ∂
Z
2′ 1′
1 2 1 2 1
 2
Sint = dσ dσ φ1 (σ , σ ) exp φ2 (σ , σ )φ3 (σ , σ ) . (2)
2πi ∂σ 1 ∂σ 2 ′
′
σ i ′ =σ i
p √
* The periodicities after T duality are V ′ /ℑτ ′ = 1/R1 and V ′ ℑτ ′ = R2 .

2
But this is exactly the characteristic interaction term in gauge theory on the noncom-
mutative torus! Without repeating all the details (given in [4] and other works on this
theory), the general connection on the noncommutative torus is a sum ∇i + Ai where ∇i
is a constant curvature connection and Ai are are elements of the algebra defined by

Z 2 Z 1 = e2πiB Z 1 Z 2 .

i
One can be more concrete by choosing an identification Z i ∼ e2πiσ ; then ∇i = ∂/∂σ i and
the noncommutative multiplication is exactly realized by the differential operator which
appears in (2). Since the interactions are all associated with the end of one open string
joining the beginning of another (as is clear by the Chan-Paton rules and the fact that the
resulting action is a single trace), every time two fields appear multiplied in the D-brane
action, such an operator will appear. This is exactly the action of gauge theory on the
noncommutative torus.
The essential phenomenon can already be seen in the D0-brane theory without T-
duality. Consider a state containing an open string starting on D0-brane i, winding w1
times around x1 and ending on brane j. If one carries it about the loop x2 , it will pick
up a phase e2πiB . Thus its interactions with the strings starting on j, winding w2 times
around x2 and ending on i must contain a relative phase e2πiBw1 w2 between the case of
attaching at i and attaching at j.
To come back to the original issue of minimal length, we have found that after dualizing
to a description with B = 0, the radii of the dual torus on which the final 2 +1-dimensional
theory is defined are 1/R1 and 1/R2 , and do become large in the limit.

A context in which the nature of the limit (1) is important is the recent proposal by
Sen and by Seiberg [5] for a general definition of Matrix theory [6,7]. They propose quite
generally that Matrix theory definitions of M theory and DLCQ M theory (with a compact
null direction) can be obtained from type IIa superstring backgrounds by considering a
sector with D0-branes and taking a scaling limit, which includes the limit R1 , R2 → 0.
Recently in [4] it was proposed that DLCQ M theory backgrounds with non-zero back-
ground three-form C−ij along the null direction have Matrix theory definitions obtained
by replacing conventional gauge theory with gauge theory on the noncommuting torus. By
the standards of superstring duality, quite strong evidence (equivalence of the BPS mass
formulas on both sides) was given. On the other hand gauge theory on the noncommut-
ing torus has been unjustly neglected by physicists and basic questions such as whether

3
it is a well-defined quantum theory have not yet been answered, leaving some room for
scepticism.
Applying the idea of [5], we can define Matrix theory in these backgrounds by starting
with type IIa superstring theory with non-zero Bij field and taking a scaling limit. This
R
corresponds to M theory d11 x C11,ij which under a boost turns directly into the back-
ground of interest. Thus the result we just described leads directly to matrix theory on
the noncommutative torus as proposed in [4].
We see that gauge theory on the noncommutative torus is a well-defined quantum
theory, considered as a subsector of string theory. This argument might well prove that
its low energy limit decoupled from string theory makes sense as a renormalizable quan-
tum theory. The only caveat is that the non-locality of the theory might invalidate the
standard effective field theory arguments. Since these are concrete theories with an action
and a conventional perturbative expansion, this question can be studied using standard
techniques, and we look forward to a detailed investigation of this point.

4
 References

[1] A. Giveon, M. Porrati and E. Rabinovici, Phys.Rept. 244 (1994) 77-202, hep-
th/9401139; and references there.
[2] A. Connes, Noncommutative Geometry, Academic Press, 1994.
[3] A. Connes, C. R. Acad. Sci. Paris Sér. A-B 290 (1980) A599-A604;
M. Pimsner and D. Voiculescu, J. Operator Theory 4 (1980) 93–118;
A. Connes and M. Rieffel, “Yang-Mills for noncommutative two-tori,” in Operator
Algebras and Mathematical Physics (Iowa City, Iowa, 1985), pp. 237–266, Contemp.
Math. Oper. Algebra. Math. Phys. 62, AMS 1987;
M. Rieffel, “Projective modules over higher-dimensional non-commutative tori,” Can.
J. Math, 40 (1988) 257–338.
[4] A. Connes, M. R. Douglas and A. Schwarz, “Noncommutative Geometry and Matrix
Theory: Compactification on Tori,” hep-th/9711162.
[5] A. Sen, “D0 Branes on T n and Matrix Theory,” hep-th/9709220;
N. Seiberg, “Why is the Matrix Model Correct?” hep-th/9710009.
[6] T. Banks, W. Fischler, S. H. Shenker and L. Susskind, Phys. Rev. D55 (1997) 5112-
5128; hep-th/9610043.
[7] L. Susskind, hep-th/9704080.