arXiv:hep-th/0106048v4 18 Oct 2001 Contents ∗ Theory Field Noncommutative lcrncades [email protected] address: Electronic I.SLTN N INSTANTONS AND SOLITONS III. V OCMUAIEQATMFIELD QUANTUM NONCOMMUTATIVE IV. I KINEMATICS II. .APIAIN OTEQATMHALL QUANTUM THE TO APPLICATIONS V. .INTRODUCTION I. .Ohrresults Other G. EFFECT THEORY .Guetheories Gauge D. .Mnplsadmnpl tig 21 strings monopole and Monopoles D. .Guetheory Gauge D. .Telws adulvl30 level Landau lowest The A. .Large A. .Fra osdrtos3 considerations Formal A. .Fymnrlsadpaaiy22 planarity and rules Feynman A. .Instantons C. .Pyiso VI iig25 mixing UV/IR of Physics C. .Fedter cin n ymtis6 symmetries and actions theory Field C. .Tefatoa unu aleet30 effect Hall quantum fractional The B. .Vre ouin nguetere 18 theories gauge in solutions Vortex B. .Cluaino opaa igas23 diagrams nonplanar of Calculation B. .Nnomttv a pc-ie4 space-time flat Noncommutative B. .Fnt eprtr 28 temperature Finite E. .Bssadpyia itrs11 pictures physical and Bases E. .Cnnclfruain29 formulation Canonical F. umte oRveso oenPhysics. level. q Modern quantum many of and and Reviews classical intensively, des to the studied to Submitted on and been both theory, have discovered, string they been and years theory M few of of last directions limits active from the of emerge most to covering wi and space-time basics to the theory with field of generalization the review We 3 2 1 ihe .Douglas R. Michael nttt o hoeia n xeietlPyis 117259 Physics, Experimental and Theoretical for Institute ntttdsHue tdsSiniqe,3 ot e Chart des route 35 Scientifiques, Etudes Hautes des Institut eateto hsc n srnm,RtesUiest,Pis University, Rutgers Astronomy, and Physics of Department .Tedrvtv n nerl4 IR of Symmetries 1. integral and derivative The 2. algebra The 1. .Mti ersnain 15 14 12 13 15 10 10 10 representations Matrix limit 6. 12 membrane/hydrodynamic The functions 5. Greens 8 Scalar bases 4. between Translations formalism 3. space Fock uncertainty position-space 2. and Gaussians 1. map Seiberg-Witten The 5. matter Fundamental 4. tensor Stress-energy 3. Observables space-time of 2. emergence The 1. .Tennomttv ou 6 5 5 torus noncommutative The symbols and 4. picture operators dipole Deformation, and basis 3. wave Plane 2. θ oiosi clrtere 16 theories scalar in solitons θ d ,2, 1, ∗ n iiaA Nekrasov A. Nikita and ,3 2, 29 30 27 20 16 22 3 8 9 3 5 1 hnnomtn oriae,starting coordinates, noncommuting th eerh uhtere r o known now are theories Such research. esro ieso (length) dimension of tensor omttvt n a otlt stecmuainre- commutation using the is lations this postulate non- describe can simplest one The and commutativity coordinates. commute, the to of noncommutativity fail might ments even commute. field, mutually magnetic to a fail of presence momenta the position In as such momentum. variables, it and conjugate of where pair any mechanics, to quantum applies us. concept in in mathematical uncertainty ingrained central expressing deeply the is is dimensions Noncommutativity more rotations and spatial three of noncommutativity in The physics. and INTRODUCTION I. CONCLUSIONS VIII. ihaparameter a with I.RLTOST TIGADMTHEORY M AND STRING TO RELATIONS VII. ocwRussia Moscow aiaieynwpeoeahave phenomena new ualitatively e,BrssrYet rne91440 France Bures-sur-Yvette res, rb unu alsae.I the In states. Hall quantum cribe I AHMTCLASPECTS MATHEMATICAL VI. n a uta aiyiaieta oiinmeasure- position that imagine easily as just can One mathematics in theme age-old an is Noncommutativity aaa J085U.S.A. 08855 NJ cataway References Acknowledgements .Dfrainqatzto 38 quantization Deformation G. .Time-like G. .Gueter n oooy35 topology and theory Gauge D. .Srnyepaain ftesltncsltos46 solutions solitonic the of explanations Stringy D. .Oeao leri set 31 aspects algebraic Operator A. .Lgtigoeve fMter 39 theory M of overview Lightning A. .Gopagba n ocmuaieqoins34 quotients noncommutative and algebras Group C. .Nnomttvt nsrn hoy42 theory string in Noncommutativity C. TPT-10,IE//12,RUNHETC-2001-18 IHES/P/01/27, ITEP-TH-31/01, .Ohrnnomttv pcs33 spaces noncommutative Other B. .Nnomttvt nMarx hoy41 theory M(atrix) in Noncommutativity B. .Tennomttv ou 36 torus noncommutative The E. .Qatmeet n lsdsrns47 strings closed and effects Quantum E. .Mrt qiaec 37 equivalence Morita F. .ASdaso ocmuaietere 48 theories noncommutative of duals AdS F. .Tedculn ii 45 45 map SW the and invariance Gauge limit 4. decoupling The 3. 43 open topological from quantization world-sheet Deformation the from 2. quantization Deformation 1. tigtheory string θ θ hc sa niymti (constant) antisymmetric an is which n xtctere 48 theories exotic and [ x i x , j = ] 2 . iθ ij , (1) 51 51 49 31 44 39 2
As has been realized independently many times, at pler than the original string theory yet keep some of this least as early as (Snyder, 1947), there is a simple modifi- nonlocality. cation to quantum field theory obtained by taking the po- One might also study noncommutative theories as in- sition coordinates to be noncommuting variables. Start- teresting analogs of theories of more direct interest, such ing with a conventional field theory Lagrangian and in- as Yang-Mills theory. An important point in this regard terpreting the fields as depending on coordinates satis- is that many theories of interest in particle physics are so fying Eq. (1), one can follow the usual development of highly constrained that they are difficult to study. For perturbative quantum field theory with surprisingly few example, pure Yang-Mills theory with a definite simple changes, to define a large class of “noncommutative field gauge group has no dimensionless parameters with which theories.” to make a perturbative expansion or otherwise simplify It is this class of theories which our review will focus on. the analysis. From this point of view it is quite interest- Until recently, such theories had not been studied very ing to find any sensible and non-trivial variants of these seriously. Perhaps the main reason for this is that pos- theories. tulating an uncertainty relation between position mea- Now, physicists have constructed many, many varia- surements will a priori lead to a nonlocal theory, with all tions of Yang-Mills theory in the search for regulated (UV of the attendant difficulties. A secondary reason is that finite) versions as well as more tractable analogs of the noncommutativity of the space-time coordinates gener- theory. A particularly interesting example in the present ally conflicts with Lorentz invariance, as is apparent in context is the twisted Eguchi-Kawai model (Eguchi and Eq. (1). Although it is not implausible that a theory Nakayama, 1983; Gonzalez-Arroyo and Okawa, 1983), defined using such coordinates could be effectively local which in some of its forms, especially that of Gonzalez- on length scales longer than that of θ, it is harder to be- Arroyo and Korthals-Altes (1983), is a noncommutative lieve that the breaking of Lorentz invariance would be gauge theory. This model was developed in the study of unobservable at these scales. the large N limit of Yang-Mills theory (’t Hooft, 1974) Nevertheless, one might postulate noncommutativity and we will see that noncommutative gauge theories show for a number of reasons. Perhaps the simplest is that many analogies to this limit (Filk, 1996; Minwalla, Van it might improve the renormalizability properties of a Raamsdonk, and Seiberg, 2000), suggesting that they theory at short distances or even render it finite. Without should play an important role in the circle of ideas relat- giving away too much of our story, we should say that this ing large N gauge theory and string theory (Polyakov, is of course not obvious a priori and a noncommutative 1987; Aharony, Gubser, Maldacena, Ooguri, and Oz, theory might turn out to have the same or even worse 2000). short distance behavior than a conventional theory. Noncommutative field theory is also known to appear Another motivation is the long-held belief that in quan- naturally in condensed matter theory. The classic exam- tum theories including gravity, space-time must change ple (though not always discussed using this language) is its nature at distances comparable to the Planck scale. the theory of electrons in a magnetic field projected to Quantum gravity has an uncertainty principle which pre- the lowest Landau level, which is naturally thought of vents one from measuring positions to better accuracies as a noncommutative field theory. Thus these ideas are than the Planck length: the momentum and energy re- relevant to the theory of the quantum Hall effect (Girvin quired to make such a measurement will itself modify and Prange, 1987), and indeed, noncommutative geom- the geometry at these scales (DeWitt, 1962). One might etry has been found very useful in this context (Bellis- wonder if these effects could be modeled by a commuta- sard, van Elst, and Schulz-Baldes, 1993). Most of this tion relation such as Eq. (1). work has treated noninteracting electrons, and it seems A related motivation is that there are reasons to believe likely that introducing field theoretic ideas could lead to that any theory of quantum gravity will not be local in further progress. the conventional sense. Nonlocality brings with it deep It is interesting to note that despite the many physi- conceptual and practical issues which have not been well cal motivations and partial discoveries we just recalled, understood, and one might want to understand them in noncommutative field theory and gauge theory was first the simplest examples first, before proceeding to a more clearly formulated by mathematicians (Connes and Rief- realistic theory of quantum gravity. fel, 1987). This is rather unusual for a theory of signif- This is one of the main motivations for the intense cur- icant interest to physicists; usually, as with Yang-Mills rent activity in this area among string theorists. String theory, the flow goes in the other direction. theory is not local in any sense we now understand, and An explanation for this course of events might be found indeed has more than one parameter characterizing this in the deep reluctance of physicists to regard a nonlocal nonlocality: in general, it is controlled by the larger of theory as having any useful space-time interpretation. the Planck length and the “string length,” the average Thus, even when these theories arose naturally in phys- size of a string. It was discovered in (Connes, Douglas, ical considerations, they tended to be regarded only as and Schwarz, 1998; Douglas and Hull, 1998) that sim- approximations to more conventional local theories, and ple limits of M theory and string theory lead directly to not as ends in themselves. Of course such sociological noncommutative gauge theories, which appear far sim- questions rarely have such pat answers and we will not 3 pursue this one further except to remark that, in our II. KINEMATICS opinion, the mathematical study of these theories and their connection to noncommutative geometry has played A. Formal considerations an essential role in convincing physicists that these are not arbitrary variations on conventional field theory but Let us start by defining noncommutative field theory in indeed a new universality class of theory deserving study a somewhat pedestrian way, by proposing a configuration in its own right. Of course this mathematical work has space and action functional from which we could either also been an important aid to the more prosaic task of derive equations of motion, or define a functional integral. sorting out the possibilities, and is the source for many We will discuss this material from a more mathematical useful techniques and constructions which we will discuss point of view in Sec. VI. in detail. Conventions. Throughout the review we use the follow- ing notations: Latin indices i,j,k,... denote space-time Having said this, it seems that the present trend is indices, Latin indices from the begining of the alphabet that the mathematical aspects appear less and less cen- a,b,... denote commutative dimensions, Greek indices tral to the physical considerations as time goes on. While µ,ν,... enumerate particles, vertex operators, etc, while it is too early to judge the outcome of this trend and it Greek indices from the begining of the alphabet α,β,... seems certain that the aspects which traditionally have denote noncommutative directions. benefited most from mathematical influence will continue to do so (especially, the topology of gauge field con- In contexts where we simultaneously discuss a non- figurations, and techniques for finding exact solutions), commuting variable or field and its commuting analog, we have to some extent deemphasized the connections we will use the “hat” notation: x is the commuting ana- with noncommutative geometry in this review. This is log tox ˆ. However, in other contexts, we will not use the partly to make the material accessible to a wider class hat. of physicists, and partly because many excellent books and reviews, starting with Connes (1994), and includ- ing (Nekrasov, 2000) focusing on classical solutions of 1. The algebra noncommutative gauge theory, (Konechny and Schwarz, 2000b) focusing on duality properties of gauge theory on The primary ingredient in the definition is an asso- a torus, as well as (Varilly, 1997; Gracia-Bondia, Varilly, ciative but not necessarily commutative algebra, to be and Figueroa, 2001), cover the material starting from this denoted . The product of elements a and b of will A A point of view. We maintain one section which attempts be denoted ab, a b, or a⋆b. This last notation (the · to give an overview of aspects for which a more mathe- “star product”) has a special connotation, to be discussed matical point of view is clearly essential. shortly. An element of this algebra will correspond to a config- Although many of the topics we will discuss were mo- uration of a classical complex scalar field on a “space” tivated by and discovered in the context of string theory, M. Suppose first that is commutative. The pri- we have also taken the rather unconventional approach of mary example of a commutativeA associative algebra is separating the discussion of noncommutative field theory the algebra of complex valued functions on a manifold from that of its relation to string theory, to the extent M, with addition and multiplication defined pointwise: that this was possible. An argument against this ap- (f + g)(x) = f(x)+ g(x) and (f g)(x) = f(x)g(x). In proach is that the relation clarifies many aspects of the this case, our definitions will reduce· to the standard ones theory, as we hope will become abundantly clear upon for field theory on M. reading section VII. However it is also true that string Although the mathematical literature is usually quite theory is not a logical prerequisite for studying the the- precise about the class of functions (continuous, smooth, ory and we feel the approach we took better illustrates etc.) to be considered, in this review we follow standard its internal self-consistency (and the points where this is physical practice and simply consider all functions which still lacking). Furthermore, if we hope to use noncommu- arise in reasonable physical considerations, referring to tative field theory as a source of new insights into string this algebra as (M) or (for reasons to be explained theory, we need to be able to understand its physics with- shortly) as “M ”.A If more precision is wanted, for most out relying too heavily on the analogy. We also hope this 0 purposes one can think of this as C(M), the bounded approach will have the virtue of broader accessibility, and continuous functions on the topological manifold M. perhaps help in finding interesting applications outside of The most elementary example of a noncommutative string theory. Reviews with a more string-theoretic em- algebra is Mat , the algebra of complex n n matrices. phasis include (Harvey, 2001a) which discusses solitonic n Generalizations of this which are almost as× elementary solutions and their relations to string theory. are the algebras Matn(C(M)) of n n matrices whose Finally, we must apologize to the many whose work matrix elements are elements of C(M×), and with addition we were not able to treat in the depth it deserved in this and multiplication defined according to the usual rules for review, a sin we have tried to atone for by including an matrices in terms of the addition and multiplication on extensive bibliography. C(M). This algebra contains C(M) as its center (take 4 functions times the identity matrix in Matn). uses either the single symbol Tr (as is done in mathe- Clearly elements of Matn(C(M)) correspond to con- matics) or to denote this combination; we do not follow figurations of a matrix field theory. Just as one can gain this convention here only to aid the uninitiated. some intuition about operators in quantum mechanics by We noteR that just as condition (b) can be violated in thinking of them as matrices, this example already serves conventional field theory for functions which do not fall to illustrate many of the formal features of noncommuta- off at infinity, leading to boundary terms, condition (c) tive field theory. In the remainder of this subsection we can be violated for general operators, leading to physical introduce the other ingredients we need to define non- consequences in noncommutative theory which we will commutative field theory in this familiar context. discuss. To define a real-valued scalar field, it is best to start with Matn(C(M)) and then impose a reality condition analogous to reality of functions in C(M). The most use- B. Noncommutative flat space-time ful in practice is to take the hermitian matrices a = a†, whose eigenvalues will be real (given suitable additional After Matn(C(M)), the next simplest example of a hypotheses). To do this for general , we would need an noncommutative space is the one associated to the alge- A bra IRd of all complex linear combinations of products of operation a a† satisfying (a†)† = a and (for c C) θ → ∈ d variablesx ˆi satisfying (ca)† = c∗a†, in other words an antiholomorphic involu- tion. [xi, xj ]= iθij . (2) The algebra Matn(C(M)) could also be defined as the The i is present as the commutator of hermitian oper- tensor product Matn(C) C(M). This construction gen- ⊗ ators is antihermitian. As in quantum mechanics, this eralizes to an arbitrary algebra to define Matn(C) , which is just Mat ( ) or n nAmatrices with elements⊗A expression is the natural operator analog of the Poisson n ij in . This algebraA admits× the automorphism group bracket determined by the tensor θ , the “Poisson ten- A 1 sor” or noncommutativity parameter. GL(n,C), acting as a g− ag (of course the center acts trivially). Its subgroup→U(n) preserves hermitian conju- By applying a linear transformation to the coordinates, the Poisson tensor can be brought to canonical form. gation and the reality condition a = a†. One sometimes refers to these as “U(n) noncommutative theories,” a bit This form depends only on its rank, which we denote as confusingly. We will refer to them as rank n theories. 2r. We keep this general as one often discusses partially In the rest of the review, we will mostly consider non- noncommutative spaces, with 2r < d. commutative associative algebras which are related to the A simple set of derivatives ∂i can be defined by the algebras (M) by deformation with respect to a param- relations eter θ, asA we will define shortly. Such a deformed algebra j j ∂ix δi (3) will be denoted by Mθ, so that M0 = (M). ≡ A [∂i, ∂j] = 0 (4) and the Leibnitz rule. This choice also determines the 2. The derivative and integral integral uniquely (up to overall normalization), by re- quiring that ∂ f = 0 for any f such that ∂ f = 0. i i 6 A noncommutative field theory will be defined by an We will occasionally generalize Eq. (4) to R action functional of fields Φ, φ, ϕ, . . . defined in terms of [∂i, ∂j ]= iΦij, (5) the associative algebra (it could be elements of , or − vectors in some representationA thereof). Besides theA alge- to incorporate an additional background magnetic field. bra structure, to write an action we will need an integral Finally, we will require a metric, which we will take to Tr and derivatives ∂i. These are linear operations sat- be a constant symmetric tensor gij , satisfying ∂igjk = 0. isfying certain formal properties: In many examples we will take this to be gij = δij , but R ij (a) The derivative is a derivation on , ∂ (AB) = note that one cannot bring both gij and θ to canonical A i (∂iA)B + A(∂iB). With linearity, this implies that the form simultaneously, as the symmetry groups preserved derivative of a constant is zero. by the two structures, O(n) and Sp(2r), are different. At (b) The integral of the trace of a total derivative is best one can bring the metric and the Poisson tensor to zero, Tr ∂iA =0. the following form: (c) The integral of the trace of a commutator is zero, r 2 R g = α=1 dzαdz¯α + b dyb ; Tr [A, B] Tr (A B B A)=0. 1 ≡ · − · θ = 2 α θα ∂z¯α ∂zα ; θa > 0. (6) A candidate derivative ∂i can be written using an ele- P ∧ P R R ment di ; let ∂iA = [di, A]. Derivations which can be Here zα = qα + ipα areP convenient complex coordinates. written in∈ A this way are referred to as inner derivations, In terms of p,q,y the metric and the commutation rela- while those which cannot are outer derivations. tions Eq. (6) read as We denote the integral as Tr as it turns out that for [y ,y ] = [y , q ] = [y ,p ]=0, [q ,p ]= iθ δ general noncommutative algebras, one cannot separate a b b α b α α β α αβ R 2 2 2 2 the notations of trace and integral. Indeed, one normally ds = dqα + dpα + dyb . (7) 5
d 1. Symmetries of IRθ The integral can be defined in this basis as
d An infinitesimal translation xi xi + ai on IR acts ikx θ Tr e = δk,0 (13) on functions as δφ = ai∂ φ. For the→ noncommuting co- i Z ordinates xi, these are formally inner derivations, as where we interpret the delta function in the usual phys- 1 j ical way (for example, its value at zero represents the ∂if = [ i(θ− )ij x ,f]. (8) − volume of physical space). One obtains global translations by exponentiating these, More interesting is the interpretation of the multipli- cation law in this basis. This is easy to compute in the i j i j i i iθij ε x iθij ε x f(x + ε )= e− f(x)e . (9) plane wave basis, by operator reordering:
′ i ij ′ ′ ikx ik x θ kik i(k+k ) x In commutative field theory, one draws a sharp dis- e e = e− 2 j e · (14) · tinction between translation symmetries (involving the ij derivatives) and internal symmetries, such as δφ = [A, φ]. The combination θ kikj′ appearing in the exponent We see that in noncommutative field theory, there is no comes up very frequently and a standard and convenient such clear distinction, and this is why one cannot sepa- notation for it is rately define integral and trace. ij k k′ θ kik′ = k θ k′, One often uses only [∂i,f] and if so, Eq. (8) can × ≡ j × be simplified further to the operator substitution ∂ i the latter notation being used to stress the choice of Pois- i(θ 1) xj . This leads to derivatives satisfying Eq. (5)→ − ij son structure. with− Φ = (θ 1) . ij − ij We can also consider The Sp(2−r) subgroup of the rotational symmetry xi ′ ′ → ij i j i j ikx ikx θ ki∂j i ij Rj x which preserves θ, Ri Rj θi′j′ = θij can be obtained e f(x) e− = e− f(x)= f(x θ kj ) (15) similarly, as · · − Multiplication by a plane wave translates a general func- i j i j i j iAij x x iAij x x i i ij f(Rj x )= e− f(x) e (10) tion by x x θ kj . This exhibits the nonlocality of the theory→ in a− particularly simple way, and gives rise iL i ik to the principle that large momenta will lead to large where R = e , Lj = Akj θ and Aij = Aji. Of course only the U(r) subgroup of this will preserve the Euclidean nonlocality. metric. A simple picture can be made of this nonlocality After considering these symmetries, we might be (Sheikh-Jabbari, 1999; Bigatti and Susskind, 2000) by tempted to go on and conjecture that imagining that a plane wave corresponds not to a parti- cle (as in commutative quantum field theory) but instead δφ = i[φ, ǫ] (11) a “dipole,” a rigid oriented rod whose extent is propor- tional to its momentum: for any ǫ is a symmetry of IRd. However, although these θ i ij transformations preserve the algebra structure and the ∆x = θ pj . (16) trace,1 they do not preserve the derivatives. Nevertheless they are important and will be discussed in detail below. If we postulate that dipoles interact by joining at their ends, and grant the usual quantum field theory relation p = ~k between wave number and momentum, the rule 2. Plane wave basis and dipole picture Eq. (15) follows immediately.
One can introduce several useful bases for the algebra d 3. Deformation, operators and symbols IRθ. as we discuss in subsection E. For discussions of perturbation theory and scattering, the most useful basis d There is a sense in which IRθ and the commutative al- is the plane wave basis, which consists of eigenfunctions d of the derivatives: gebra of functions C(IR ) have the same topology and the same “size,” notions we will keep at an intuitive level. In ikx ikx ∂ie = ikie . (12) the physical applications, it will turn out that θ is typ- ically a controllable parameter, which one can imagine The solution eikx of this linear differential equation is the increasing from zero to go from commutative to noncom- exponential of the operator ik x in the usual operator mutative (this does not imply that the physics is contin- sense. · uous in this parameter, however). These are all reasons to study the relation between these two algebras more systematically. There are a number of ways to think about this rela- 1 Assuming certain conditions on ǫ and φ; see Sec. VI.A. tion. If θ is a physical parameter, it is natural to think 6
We will freely assume the usual Fourier relation between ikx e position and momentum space for the symbols, while be- ing careful to say (or denote by standard letters such as x and k) which we are using. f(x) The star product for these symbols is
′ i ij ′ ′ ikx ik x θ kik i(k+k ) x e ⋆e = e− 2 j e · . (21)
Of course all of the discussion in B.2 above still applies, as this is only a different notation for the same product Eq. (14). -ikx e Another special case which often comes up is
FIG. 1 The interaction of two dipoles. Tr f⋆g = Tr fg. (22) Z Z
d d of IRθ as a deformation of IR . A deformation Mθ of C(M) is an algebra with the same elements and addi- 4. The noncommutative torus tion law (it is the same considered as a vector space) but Much of this discussion applies with only minor a different multiplication law, which reduces to that of Td C(M) as a (multi-)parameter θ goes to zero. This notion changes to define θ, the algebra of functions on a non- commutative torus. was introduced in (Bayen, Flato, Fronsdal, Lichnerowicz, d and Sternheimer, 1978) as an approach to quantization, To obtain functions on a torus from functions on IR , we would need to impose a periodicity condition, say and has been much studied since, as we will discuss in i i i Sec. VI. Such a deformed multiplication law is often de- f(x ) = f(x +2πn ). A nice algebraic way to phrase Td noted f⋆g or “star product” to distinguish it from the this is to instead define θ as the algebra of all sums of original point-wise multiplication of functions. products of arbitrary integer powers of a set of d variables This notation has a second virtue, which is that it al- Ui, satisfying lows us to work with M in a way which is somewhat ij θ iθ more forgiving of ordering questions. Namely, we can UiUj = e− Uj Ui. (23) choose a linear map S from Mθ to C(M), fˆ S[fˆ], ixi called the “symbol” of the operator. We then represent7→ The variable Ui takes the place of e in our previous the original operator multiplication in terms of the star notation, and the derivation of the Weyl algebra from product of symbols as Eq. (1) is familiar from quantum mechanics. Similarly, we take 1 fˆgˆ = S− [ S[fˆ] ⋆S[ˆg] ]. (17) [∂ ,U ]= iδ U , One should recall that the symbol is not “natural” in i j ij j the mathematical sense: there could be many valid defini- and tions of S, corresponding to different choices of operator ordering prescription for S 1. − Tr U n1 ...U nd = δ . A convenient and standard choice is the Weyl ordered 1 d ~n,0 symbol. The map S, defined as a map taking elements Z d d of IRθ to (IR ) (functions on momentum space), and its There is much more to say in this case about the topo- inverse, areA logical aspects, but we postpone this to Sec. VI.
1 ikxˆ f(k) S[fˆ](k)= Tr e− fˆ(ˆx) (18) ≡ (2π)n/2 Z C. Field theory actions and symmetries 1 1 n ikxˆ fˆ(ˆx)= S− [f]= d k e f(k) (19) (2π)n/2 Field theories of matrix scalar fields are very familiar Z and are treated in most textbooks on quantum field the- Formally these are inverse Fourier transforms, but the ory. The matrix generalization is essential in discussing first expression involves the integral Eq. (13) on (IR ), θ Yang-Mills theory. In a formal sense we will now make while the second is an ordinary momentum spaceA inte- explicit, any field theory Lagrangian which is written in gral. terms of matrix fields, matrix addition and multiplica- One can get the symbol in position space by performing tion, and the derivative and integral, can be equally well a second Fourier transform; e.g. regarded as a noncommutative field theory Lagrangian, 1 n ik(x xˆ) with the same equations of motion and (classical) sym- S[fˆ](x)= d k Tr e − fˆ(ˆx). (20) (2π)n metry properties as the matrix field theory. Z Z 7
Let us consider a generic matrix scalar field theory used to define momentum and angular momentum oper- with a hermitian matrix valued field φ(x) = φ(x)† and ators, for example (Euclidean) action 1 i 0 1 Pi = i(θ− )ij Tr x T , (26) S = ddx √g gij Tr ∂ φ∂ φ + Tr V (φ) (24) − 2 i j Z Z Thus we refer to it as the “restricted stress-energy ten- where V (z) is a polynomial in the variable z, ∂ = ∂ i ∂xi sor.” ij are the partial derivatives, and g is the metric. The One can also apply the Noether definition to the gen- constraint we require in order to generalize a matrix ac- eral variation xi xi + vi(x), to define a more con- tion to a noncommutative action is that it be written 7→ ventional stress-energy tensor Tij, discussed in (Gerhold, only using the combination Tr appearing in Eq. (24); Grimstrup, Grosse, Popp, Schweda, and Wulkenhaar, we do not allow either the integral or the trace Tr 2000; Abou-Zeid and Dorn, 2001a). In general, the action to appear separately. In particular,R the rank of the ma- R of this stress tensor changes θ and the underlying alge- trix N can not appear explicitly, only in the form Tr 1 bra, and its interpretation has not been fully elucidated combined with the integral. at present (see Secs. VI.G,VII.C.2 for related issues). Under this assumption, it is an easy exercise to check Finally, there is a stress-energy tensor for noncommu- that if we replace the algebra Mat (C(M)) by a general N tative gauge theory which naturally appears in the rela- associative algebra with integral and derivative satis- tion to string theory, which we will discuss in Sec. II.D.3 fying the requirementsA above, the standard discussion of and Sec. VII.E. equations of motion, classical symmetries and Noether’s We could just as well consider theories containing an theorem, all go through without change. The point is arbitrary number of matrix fields with arbitrary Lorentz that formal manipulations which work for arbitrary ma- transformation properties (scalar, spinor, vector and so trices of functions can always be made without commut- on). However, at this point we will only consider a gen- ing the matrices. Another way to think about this result eralization directly analogous to the treatment of higher is to imagine defining the theory in terms of an explicit spin fields in Euclidean and Minkowski space. We will matrix representation of the algebra . discuss issues related to curved backgrounds later; at Thus the noncommutative theory withA action Eq. (24) present the noncommutative analogs of manifolds with has the standard equation of motion general metrics are not well understood. ij g ∂i∂j φ = V ′(φ) Although one can be more general, let us now assume that the derivatives ∂ are linearly independent and sat- i i and conservation laws ∂iJ = 0 with the conserved cur- isfy the usual flat space relations [∂ , ∂ ]=0. i i j rent J associated to a symmetry δφ(ǫ, φ) determined by Given Poincar´esymmetry or its subgroup preserving the usual variational procedures, θ, one can use the conventional definitions for the action
i of the rotation group on tensors and spinors, which we δS = Tr J ∂iǫ. do not repeat here. In particular, the standard Dirac d Td Z equation also makes sense over IRθ and θ, so spin 1/2 For example, let us consider the transformations Eq. particles can be treated without difficulty. (11). In matrix field theory, these would be the infinites- The discussion of supersymmetry is entirely parallel to imal form of a U(N) internal symmetry φ U †φU. Al- that for conventional matrix field theory or Yang-Mills though in more general noncommutative theories→ [∂ ,ǫ] = i 6 theory, with the same formal transformation laws. Con- 0 and these are not in general symmetries, we can still straints between the dimension of space-time and the consider their action, and by exponentiation define an number of possible supersymmetries enter at the point analogous “U(N)” action. We will refer to this group as we assume that the derivatives ∂i are linearly indepen- U( ), the group of unitary operators acting on a Hilbert H dent. With care, one can also use the conventional super- space admitting a representation of the algebra . In field formalism, treating the anticommuting coordinates more mathematicalH terms, discussed in VI, willA be a H as formal variables which commute with elements of module for . (Ferrara and Lledo, 2000). A Of course,A if we do not try to gauge U( ), it could H Finally, as long as time is taken as commutative, also be broken by other terms in the action, for exam- the standard discussion of Hamiltonian mechanics and ple source terms Tr Jφ, position-dependent potentials canonical quantization goes through without conceptual Tr fV (φ) and so forth. difficulty. On the other hand, noncommutative time im- Application of theR Noether procedure to Eq. (11) leads R i plies nonlocality in time, and the Hamiltonian formal- to a conserved current T , which for Eq. (24) would be ism becomes rather complicated (Gomis, Kamimura, and i ij Llosa, 2001); it is not clear that it has any operator in- T = ig [φ, ∂j φ]. (25) terpretation. Although functional integral quantization As we discussed in Sec. II.B.1, Eq. (11) includes transla- is formally sensible, the resulting perturbation theory is tions and the rotations which preserve θij , so T i can be problematic as discussed in Sec. IV. It is believed that 8 sensible string theories with time-like noncommutativity 1. The emergence of space-time exist, discussed in Sec. VII.G. The first point to realize is that the gauge group in noncommutative theory contains space-time translations. D. Gauge theory This is already clear from the expression Eq. (8), which j allows us to express a translation δAi = v ∂j Ai in terms j 1 k The only unitary quantum field theories including vec- of a gauge transformation Eq. (28) with ǫ = v (θ− )jkx . tor fields are gauge theories, and the standard defini- Actually, this produces tions also apply in this context. However there is a great j j 1 deal more to say about the kinematics and observables δAi = v ∂j Ai + v (θ− )ji, of gauge theory. A gauge connection will be a one-form Ai, each com- but an overall constant shift of the vector potential drops out of the field strengths and has no physical effect in ponent of which takes values in and satisfies Ai = Ai†. (See Sec. VI.D for a more generalA definition.) The asso- infinite flat space. ciated field strength is Taking more general functions for ǫ will produce more general space-time transformations. As position- Fij = ∂iAj ∂j Ai + i[Ai, Aj ], (27) dependent translations, one might compare these with − which under the gauge transformation coordinate definitions or diffeomorphisms. To do this, we consider the products as star products and expand δAi = ∂iǫ + i[Ai,ǫ] (28) Eq. (21) in θ, to obtain transforms as δF = i[F ,ǫ], allowing us to write the ij ij δφ = i[φ, ǫ] = θij ∂ φ∂ ǫ + (∂2φ∂2ǫ). (32) gauge invariant Yang-Mills action i j O φ, ǫ , (33) 1 →{ } S = Tr F 2. (29) −4g2 ( , is the Poisson bracket), so at leading order the gauge Z { } All this works for the reasons already discussed in group is the group of canonical transformations preserv- Sec. II.C. ing θ (we discuss this further in E.5). Of course, the Gauge invariant couplings to charged matter fields can higher derivative terms modify this result. In fact the full gauge group U( ) is simpler, as we will see in E.2. be written in the standard way using the covariant deriva- H tive Another aspect of this unification of space-time and gauge symmetry is that if the derivative is an inner Diφ ∂iφ + i[Ai, φ]. (30) derivation, we can absorb it into the vector potential it- ≡ self. In other words, we can replace the covariant deriva- Finite gauge transformations act as d tives Di = ∂i + iAi with “connection operators” in IRθ, (∂i + iAi, F, φ) U † (∂i + iAi, F, φ) U 1 j → C ( iθ− ) x + iA (34) and these definitions gauge the entire U( ) symmetry. i ≡ − ij i H One can also use MatN ( ) to get the noncommutative such that analog of U(N) gauge theory,A though (at this point) not the other Lie groups. Di f [Ci,f]. (35) As an example, we quote the maximally supersymmet- → ric Yang-Mills (MSYM) Lagrangian in ten dimensions, We also introduce the “covariant coordinates,” from which = 4 SYM in d = 4 and many of the sim- i i ij pler theoriesN can be deduced by dimensional reduction Y = x + θ Aj (x). (36) and truncation: i ij If θ is invertible, then Y = iθ Cj and this is just another S = d10x Tr F 2 + iχ¯ Dχ/ I , (31) notation, but the definition makes sense more generally. ij I In terms of the connection operators, the Yang-Mills Z where χ is a 16-component adjoint Majorana-Weyl field strength is fermion. This action satisfies all of our requirements and 1 F = i[D ,D ] i[C , C ] (θ− ) , (37) thus leads to a wide variety of supersymmetric noncom- ij i j → i j − ij mutative theories. Indeed, noncommutativity is the only and the Yang-Mills action becomes a simple “matrix known generalization (apart from adding irrelevant oper- model” action, ators and taking limits of this) which preserves maximal 1 2 supersymmetry. S = Tr (i[C , C ] (θ− ) ) . (38) i j − ij Because one cannot separately define integral and i,j trace, the local gauge invariant observables of conven- X tional gauge theory do not carry over straightforwardly: Now, although we motivated this from Eq. (29), we only Tr O is gauge invariant. We now discuss this could look at this the other way around, starting with point. the action Eq. (38) as a function of matrices Ci and R 9 postulating Eq. (34), to derive noncommutative gauge A step forward is to define the Wilson loop operator. theory (and Yang-Mills theory in the limit θ 0) from Given a path L, we write the holonomy operator using a matrix model. This observation is at the heart→ of most exactly the same formal expression as in conventional of the common ways that noncommutative gauge theory gauge theory, arises in particle physics, as the action Eq. (38) and its supersymmetrization is simple enough to arise in a wide WL = P exp i dσA(x(σ)), variety of contexts. For example, it can be obtained as ZL a limit of the twisted Eguchi-Kawai model (Eguchi and Nakayama, 1983; Gonzalez-Arroyo and Okawa, 1983), but where the products in the expansion of the path or- which was argued to reproduce the physics of large N dered exponential are star products. This undergoes the Yang-Mills theory. The maximally supersymmetric ver- gauge transformation sion obtained in the same way from Eq. (31), often re- W U †(x )W U(x ) ferred to as the “IKKT model” (Ishibashi, Kawai, Ki- L → 1 L(x1,x2) 2 tazawa, and Tsuchiya, 1997), plays an important role in where x and x are the start and end points of the path M theory, to be discussed in Sec. VII. 1 2 L1,2. Having so effectively hidden it, we might well wonder We can form a Wilson loop by taking for L a closed how d-dimensional space-time is going to emerge again loop with x1 = x2, but again we face the problem that from Eq. (38). Despite appearances, we do not want to we can only cyclically permute operators, and thus cancel claim that noncommutative gauge theory is the same in U 1(x ) with U(x ), if we take the trace over , which all dimensions d. Now in the classical theory, to the ex- − 1 1 includes the integral over noncommutative space.H tent we work with explicit expressions for A in Eq. (34), i We can at least formulate multilocal observables with this is generally not a problem. However, in the quantum this construction, such as theory, we need to integrate over field configurations Ci. We will need to argue that this functional integral can be restricted to configurations which are “similar to” Eq. Tr O1(x1)W [L1,2]O2(x2)W [L2,3] ...O(xn)W [Ln,1] (34) in some sense. Z This point is related to what at first appears only to be with arbitrary gauge covariant operators Oi at arbitrary 1 a technical subtlety involving the θ− terms in Eq. (38). points xi, joined by Wilson loops. This allows us to con- 1 1 They are there to cancel an extra term [(θ− x)i, (θ− x)j ], trol the distance between operators within a single trace, which would have led to an infinite constant shift of the but not to control the distance between operators in dif- action. The subtlety is that one could have made a mis- ferent traces. take at this point by assuming that Tr [Ci, Cj ] = 0, as Actually one can do better than this, using what are for finite dimensional operators. called open Wilson loops (Ishibashi, Iso, Kawai, and Ki- Of course the possibility that Tr [Ci, Cj ] = 0 probably tazawa, 2000). The simplest example is comes as no surprise, but the point we want6 to make is that Tr [C , C ] should be considered a topological aspect i j W (k) Tr W [L ]eikxˆ2 . of the configuration. We will argue in Sec. VI.A that it L ≡ 1,2 is invariant under any variation of the fields which (in a Z sense) preserve the asymptotics at infinity. This invariant If the distance between the end points of L and the mo- mentum k satisfy the relation Eq. (16), θij k = (x x ), detects the presence of the derivative operators in Ci and i 1 − 2 this is the underlying reason one expects to consistently this operator will be gauge invariant, as one can see by identify sectors with a higher dimensional interpretation using Eq. (15). in what is naively a zero dimensional theory. This provides an operator which carries a definite mo- mentum and which can be used to define a version of local correlation functions. There is a pleasing correspondence 2. Observables between its construction and the dipole picture of sub- section B.2; not only can we think of a plane wave as All this is intriguing, but it comes with conceptual having a dipole extent, we should think of the two ends problems. The most important of these is that it is of the dipole as carrying opposite electric charges which difficult to define local observables. This is because, as for gauge invariance must be attached to a Wilson line. noted from the start, there is no way to separate the trace The straight line has a preferred role in this construc- over (required for gauge invariance) from the integral tion, and the open Wilson loop associated to the straight over noncommutativeH space. We can easily enough write line with length determined by Eq. (16) can be written gauge invariant observables, such as k C ik Y W (k) Tr e × = Tr e · ≡ Tr F (x)n, Z Z Z where Y i are the covariant coordinates of Eq. (36). This but they are not local. construction can be used to covariantize local operators 10 as follows: given an operator (x), transforming in the Bilinears such as ψ†ψ, ψ†Diψ and so on will be gauge adjoint, and momentum k, weO define invariant and can be used in the action and to define
m new observables, either by enforcing an equation of mo- k C ikmY (x) W [ ](k)= Tr e × (x)= Tr e (x) O O O tion on ψ or doing a functional integral over ψ (Amb- d ik (Y y) [yR]= d k Tr e R· − (x) (39) jorn, Makeenko, Nishimura, and Szabo, 2000; Gross and O O Nekrasov, 2001; Rajaraman and Rozali, 2000). R R Although in a strict sense this is also a global ob- 3. Stress-energy tensor servable, an important point (which will be central to Sec. VI.D) is that one can also postulate an independent As we discussed above, the simplest analog of the rule for multiplication by (and the unitaries in ) on stress-energy tensor in noncommutative field theory is the right, A A Eq. (25), which generates the noncommutative analog ψ ψa. of canonical transformations on space-time. However, in → noncommutative gauge theory, this operator is the gen- Indeed, if we take ψ , we will clearly get a nontrivial erator of gauge transformations, so it must be set to zero second action of this∈ type, A since left and right multipli- on physical states. This leads to a subtlety analogous cation are different. In this case, we can think of ψ†ψ to one known in general relativity: one cannot define a as a function on a second, dual noncommutative space. gauge invariant local conserved momentum density. This For each f one obtains a gauge invariant observable is compatible with the difficulties we just encountered in ∈ A Tr ψ†ψf, which is local on the dual space in the same defining local gauge invariant observables. sense that an noncommutative field is a local observable One can nevertheless regard Eq. (26) as a nontrivial in an ungauged theory. global conserved momentum. This is because it corre- Taking ψ is a choice. One could also have taken sponds to a formal gauge transformation with a param- ψ , which∈ does A not lead to such a second multiplica- eter ǫ xi which does not fall off at infinity (on the ∈ H ∼ tion law. The general theory of this choice is discussed torus, it is not even single valued), and as such can be in Sec. VI.D. consistently excluded from the gauge group. This type The two definitions lead to different physics. Let us of consideration will be made more precise in Sec. VI.A. compare the spectral density. If we take the Dirac oper- One can make a different definition of stress-energy i r 1 ator γ Di acting on ψ r, this has dρ(E) dEE − , tensor, motivated by the relation Eq. (34) between the as for a field in r dimensions.∈ H ∼ connection and the noncommutative space-time coordi- If we take the same Dirac operator with ψ , we nates, as the Noether current associated to the variation would get infinite spectral density. A more useful∈ defini-A ik Y C C + a (k)e · tion is i → i i Dψ/ = γi(D ψ ψ∂ ) which for the action Eq. (38) can easily be seen to pro- i − i duce which fixes this by postulating an ungauged right action 1 of translations, leading to a spectral density appropriate isk Y i(1 s)k Y Tij (k)= ds Tr e · [Ci, Cl]e − · [Cj , Cl], to 2r dimensions. l Z0 Z X (40) mn which is conserved in the sense that kmθ Tnl(k)=0for 5. The Seiberg-Witten map a solution of the equations of motion. This appears to be the natural definition in string theory, as we discuss Having discovered an apparent generalization of gauge in Sec. VII.E. theory, we should ask ourselves to what extent this the- ory is truly novel and to what extent we can understand it as a conventional gauge theory. This question will be- 4. Fundamental matter come particularly crucial once we find noncommutative gauge theory arising from open string theory, as general Another type of gauge invariant observable can be ob- arguments imply that open string theory can always be tained by introducing new fields (bosons or fermions) thought of as giving rise to a conventional gauge theory. which transform in the fundamental of the noncommu- Is there an inherent contradiction in these claims? tative gauge group. In other words, we consider a field Seiberg and Witten (1999) proposed that not only is ψ which is operator valued just as before, but instead of there no contradiction, but that one should be able to transforming under U( ) as ψ UψU †, we impose the H → write an explicit map from the noncommutative vec- transformation law tor potential to a conventional Yang-Mills vector poten- ψ Uψ. tial, explicitly exhibiting the equivalence between the two → classes of theories. More generally, we need to define multiplication a ψ by One might object that the gauge groups of noncommu- any element of , but this can be inferred using linearity.· tative gauge theory and conventional gauge theory are A 11 different, as is particularly clear in the rank 1 case. How- the action. This however ignores the possibility that the ever, this is not an obstacle to the proposal, as only the map might take nonsingular field configurations in the physical configuration space – namely the set of orbits one description, to singular field configurations in the under gauge transformation – must be equivalent in the other. Indeed, Eq. (44) gives an explicit example. When 1 two descriptions. It does imply that the map between F = θ− , the noncommutative description appears to the two gauge transformation laws must depend on the break− down, as Fˆ would have a pole. Conversely, F is 1 vector potential, not just the parameter. singular when Fˆ = θ− . Thus, the proposal is that there exists a relation be- As we continue, we will find many examples in which tween a conventional vector potential Ai with the stan- noncommutative gauge theory has different singular so- dard Yang-Mills gauge transformation law with parame- lutions and short distance properties from conventional ter ǫ, Eq. (28), and a noncommutative vector potential gauge theory, and despite this formal relation between Aˆi(Ai) and gauge transformation parameterǫ ˆ(A, ǫ) with the theories it will become clear that their physics is in general rather different. noncommmutative gauge invariance δˆAˆi = ∂iǫˆ+ iAˆi ˆǫ ∗ − A solution to the Seiberg-Witten equation was recently iǫˆ Aˆi, such that ∗ found (Liu, 2000; Liu and Michelson, 2001a; Okawa Aˆ(A)+ δˆǫˆAˆ(A)= Aˆ(A + δǫA). (41) and Ooguri, 2001b; Mukhi and Suryanarayana, 2001). Namely, the following inhomogeneous even degree form d d d This equation can be solved to first order in θ without on IR , defined as the integral over the superspace IR | , difficulty. Writing θ = δθ, we have is closed: Aˆ (A) A = 1 δθkl A , ∂ A + F + (δθ2)(42) i − i − 4 { k l i li}+ O ddkddϑ Tr exp F ρ(k) ˆǫ(A, ǫ) ǫ = 1 δθkl ∂ ǫ, A + (δθ2) (43) H 2πi 4 k l + Z − { } O i i i i j = ki(Y y ) ϑidy + ϑiϑj [Y , Y ] where A, B + AB + BA. The corresponding first F − − order relation{ } between≡ the field strengths is (45) 1 where y’s are the coordinates on IRd, and ρ(k) can be any Fˆ F = δθkl (2 F , F A ,D F + ∂ F ) . ij − ij 4 { ik jl}+ −{ k l ij l ij }+ smooth function such that ρ(0) = 1 (this slightly gener- alizes the references, which take ρ = 1). This expression This result even admits a reinterpretation which de- has an expansion in differential forms on IRd, whose two- fines the map to all finite orders in θ. Consider the prob- 1 ˆ(θ) form part is the conventional F + θ− . Deeper aspects lem of mapping a noncommutative gauge field A de- of this rather suggestive superspace expression will be fined with respect to the star product for θ, to a noncom- discussed in Nekrasov (2001). mutative gauge field Aˆ(θ+δθ) defined for a nearby choice of θ. To first order in δθ, it turns out that the solu- tion to the corresponding relation Eq. (41) is again Eqs. E. Bases and physical pictures (42), (43), now with the right hand side evaluated us- ing the star product for θ. Thus these equations can d d The algebra IRθ of “functions on noncommutative IR ,” be interpreted as differential equations (Seiberg-Witten considered as a linear space, admits several useful bases. equations) determining the map to all orders. Since it is just a product of Heisenberg algebras and com- Eq. (42) can be solved explicitly for the case of a rank muting algebras, all of this formalism can be traced back one gauge field with constant F . In this case, it reduces to the early days of quantum mechanics, as can much of to its physical interpretation. In quantum mechanics, it ap- pears when one considers density matrices (the Wigner δFˆ = Fδθˆ Fˆ − functional) and free fermi fluids (in one dimension, this leads to bosonization and W algebra). where Lorentz indices are contracted as in matrix multi- ∞ plication. It has the solution (with boundary condition However we stress that the noncommutativity under F at θ = 0) discussion here is not inherently quantum mechanical. Rather, it is a formal device used to represent a partic- 1 Fˆ =(1+ Fθ)− F. (44) ular class of interactions between fields, which can exist in either classical or quantum field theory. In particu- This result can be used to relate the conventional and lar, an essential difference with the standard quantum noncommutative gauge theory actions at leading order mechanical applications is that these involve linear equa- in a derivative expansion, as we will discuss in detail in tions, while we are going to encounter general nonlinear Sec. VII.C. equations.2 All this might suggest that the noncommutative frame- work is merely a simpler way to describe theories which could have been formulated as conventional gauge the- ories, by just applying the transformation Fˆ F to 2 The special case of the equation φ2 = φ defining a projection can → 12
1. Gaussians and position-space uncertainty This result illustrates the sense in which interactions in noncommutative theory obey a position-space uncer- While the plane wave basis is particularly good for tainty principle. Formally, we can construct a Gaussian perturbation theory, nonperturbative studies tend to be configuration of arbitrarily small width in the noncom- simpler in position space. However, in noncommutative mutative theory; its limit is the delta function we just theory the standard position space basis tends not to be discussed. Unlike commutative theory, however, multi- the most convenient, because of the nonlocal nature of plication by a Gaussian of width b2 <θ does not concen- the interactions. One is usually better off using a basis trate a field configuration but instead tends to disperse which simplifies the product. it. This is particularly clear for the special case a = b of One expects the noncommutativity Eq. (1) to lead to Eq. (49). More generally, the operation of multiplication a position-space uncertainty principle, which will exclude by a Gaussian ψb1,0 (for any b) will cause the width to the possibility of localized field configurations. Although approach θ, decreasing (∆
1 −1 δ(z) ⋆f(y)= ddy eiyθ zg(y) (47) 2. Fock space formalism (2π)d det θ Z is a highly nonlocal operation: it is the composition of A nice formal context which provides a basis including a Fourier transform with the linear transformation z the minimal Gaussian is to use as noncommutative coor- 1 → θ− z. dinates creation and annihilation operators acting on a As in quantum mechanics, one might expect the Gaus- Fock space. These are defined in terms of the canonical sian to be a particularly nice basis state, since it is simul- coordinates of Eq. (7) in the usual way, taneously Gaussian in both conjugate coordinates. Let M a a M qα+ipα qα ipα ψ , be a Gaussian with center , covariance , and aα = ; a†α = − maximum 1: √2θα √2θα z = √2θ a ;z ¯ = √2θ a† i i j j α α α α α α ψM,a = exp (x a )Mij (x a ). a a − − − [ α, †β]= δαβ. (50) 1 d 2 2M 2 which satisfies d x ψ = (det π )− . The star product of| two| Gaussians can be easily worked We can now identify elements of IRd with functions of R θ out using Eq. (46). In particular, for concentric Gaus- the ya valued in the space of operators acting in the Fock sians of width a and b, we have space of r creation and annihilation operators. The Hr Fock space r is the Hilbert space of our previous ψ 1 1,0 ⋆ ψ 1 1,0 = C(a,b) ψ 1 1,0 (48) H H a2 b2 ∆(a,b)2 discussion, and this basis makes the nature of U( ) par- H with ticularly apparent: it is the group of all unitary operators 2 2 2 on Hilbert space. Explicitly, 2 a b + θ ∆(a,b) = 2 2 a + b r = C n1,...,nr d H | i 2 − 2 a ...,n ,... = n ...,n 1,... θ α α L√ α α C(a,b)= 1+ 2 (49) | i | − i (ab) a† ...,n ,... = √n +1 ...,n +1,... α| α i α | α i nˆα = a†αaα. (51)
arise as a normalization condition in quantum mechanics. Also, We have also introduced the number operatorn ˆα. Real somewhat similar nonlinear equations appear in the approxima- functions of the original real coordinates correspond to tion methods of quantum statistical mechanics. the hermitian operators. 13
In this language, the simplest basis we can use con- (¯z ξ√2θ) (z η√2θ)= (¯z ξ 2θ )(z η 2θ ) − · − − α α − α α sists of the elementary operators ~k ~l . These can be α a a | ih | X p p expressed in terms of †α and α as and a formula which follows from this for matrix ele- k +n α α lα+nα ments between Fock basis states with vectors of occupa- a† a ~k ~l = ( 1)nα α α tion numbers ~k and ~l: | ih | − n !√k !l ! ~n α α α α X Y 1 kα lα ~k fˆ~l = ∂ ∂ ξ=η=0(ξ fˆ η). (58) h | | i √ ξα ηα | | | Using Eq. (8), the derivatives can all be written as com- α kα!lα! mutators with the operators aα, a†α. The integral Eq. Y (13) becomes the standard trace in this basis, For example, let us consider the projection operator fˆ 0 0 . We have (η fˆ ξ) = 1, and using Eq. (58) we 0 ≡ | ih | | 0| 2r r 1 ˆ d xTr (2πθ )Tr . (52) can reproduce this with f0(z, z¯)=2 exp( z θ z¯), so f0 → α − · Z α is precisely the minimal Gaussian we encountered earlier. Y One could extend this to use ~ ~ The operators k k , their sums and unitary rotations ξ˜ η˜ r (¯z ξ˜√2θ) θ−1(z η˜√2θ) | ih | fˆ ˜ = e− · η˜)(ξ˜ f ˜ =2 e− − · − of these, provide a large set of projections, operators P η,˜ ξ | | ↔ η,˜ ξ satisfying P 2 = P . This is in stark contrast with the (59) d algebra C(M) (where M is a connected space) which as an overcomplete basis for IRθ, consisting of minimal would have had only two projections, 0 and 1, and this is Gaussians with centers x = (˜z, z˜∗) (e.g. see Eq. (15)) a key difference between noncommutative and commuta- multiplied by plane waves with momentum k = (˜κ, κ˜∗), tive algebras. Physically, this will lead to the existence with of new solitonic solutions in noncommutative theories, as 1 θ we discuss in Sec. III. κ˜ = (ξ˜ η˜∗), z˜ = (ξ˜+˜η∗) A variation on the projection which is also useful in 2θ − r2 generating solutions is the partial isometry, which by def- (where ∗ denotes complex conjugation). As another ex- inition is any operator R satisfying ample, the delta function δ(d)(x) has matrix elements (from Eq. (57)) (η δˆ ξ) = e ηξ, leading to the expres- RR†R = R. (53) − sions | |
Such an operator can be written as a product R = PU of ~k ~nˆ ~k δˆ ~l = δ ( 1)| |, δˆ = ( 1)| | (60) a projection P and a unitary U. The simplest example h | | i ~k,~l − − is the shift operator S† with matrix elements α with ~k = k , ~nˆ = nˆ . | | α α | | α α Sα† ...,nα,... = ...,nα +1,... . (54) Let us now express the general radially symmetric | i | i function inP two noncommutativeP dimensions in the two It satisfies 1 2 bases (we take θ = 2 ). These are functions of r = 2 2 S S =1, and p + q zz¯ =n ˆ, so the general such function in the α α† Fock basis∼ is S S =1 n , 0 n , 0 (55) α† α nβ ;β=α β α β α − 6 | ih | c fˆ = c n n . (61) n n n| ih | P n 0 n 0 X≥ X≥ 3. Translations between bases The corresponding symbols fn can be found as solutions of the equations zz¯ f = (n + 1 )f (Fairlie, 1964; Cur- In this subsection we assume for simplicity that d =2r. ∗ n 2 n The translation between this basis and our previous de- tright, Uematsu, and Zachos, 2000). A short route to scriptions involving commutative functions and the star the result is to form a generating function from these product, as we commented above, can be done using the operators, plane wave basis and Fourier transform. fˆ = un n n , (62) A standard tool from quantum mechanics which fa- | ih | n 0 cilitates such calculations is the coherent state basis X≥ (Klauder and Sudarshan, 1968). Note that this is not whose matrix elements are (η fˆ ξ) = exp uηξ. This can a basis of IRd but rather is a basis of . We recall their | | θ Hr be obtained from Eq. (57) by taking for f the generating definition function a a† (ξ = 0 eξα α , η)= eηα α 0 , (56) 1 | h | | | i f(zz¯; u) = (λ )exp λzz¯ − 2 a useful formula for matrix elements of fˆ in this basis, with 1 u = 1/(λ 1 ). Substituting u for λ in this r r − − 2 d z d z¯ 1 √ √ expression leads to a generating function for the Laguerre ˆ ξ η θ (¯z ξ 2θ) (z η 2θ) (ξ f η)= f (z, z¯) e · − − · − (57) 2 | | πθ polynomials Ln(4r ) (Bateman, 1953). Z α α Q 14
The final result for the symbol of Eq. (62) is Greens function on a Gaussian centered at x = a. The IR (long distance) behavior is controlled by the limit t n 2 2r2 → ∞ f =2 u Ln(4r )e− . of the proper time integral, and we see that it has the n 0 same dependence on a as on y. Thus, in this sense, the X≥ IR behavior is the same as for the commutative Greens function. 4. Scalar Greens functions The UV behavior is controlled by short times t 0, → We now discuss the Greens function of the free non- and at first sight looks rather different from that of the commutative scalar field. This is very simple in the plane commutative Greens function, due to the shifts t t + 1 . However, this difference is only apparent and→ comes wave basis, in which the Klein-Gordon operator is diag- 2 onal: because the noncommutative Greens function in effect contains a factor of the delta function Eq. (60), which is i 2 ikx 2 2 ikx ( ∂i∂ + m )e = (k + m )e . hiding the UV divergence. To see this, we can compute − Tr δˆGˆ, which is the correct way to take the coincidence The Greens function satisfying limit. ∂2 This can be done in the coherent state basis, but we ( + m2)G(x, y)= δ(x, y) (63) instead make a detour to expand the Greens function in − ∂x2 i i X the Fock basis. This can be done using Eq. (58); one finds that the matrix elements are nonzero only on the is then just diagonal and are a function only of the sum of the occupa-
δk,k′ tion numbers a na – this reflects rotational invariance. G(k, k′)= . (64) 1 k2 + m2 Changing variables from t to λ =1/(t + 2 ), setting y =0 and going to theP Fock basis, we obtain This result can be easily transformed to other bases 1 using coherent states. We set θα = and start from 2 2 d d ˆ 2 2 r 2 n ikz¯+ikz¯ 1 kk¯+ikη¯ +ikξ+ηξ n G n = dλ λ − (2 λ) − (1 λ) . (68) (η e ξ)= e− 2 (65) h | | i − − | | Z0 which is derived from Eq. (57) by Gaussian integration. This allows us to derive the matrix elements of the This result makes it easy to answer the previous question: ˆ ˆ ˆ ˆ Greens function G by Fourier transform: the sum over modes Tr δG = n n G n produces (2 r h | | i − λ)− , which is exactly the UV divergent factor which was eikx missing from Eq. (68). P (η ddk ξ)= | k2 + iǫ| Z In effect, the coincidence limit of the Greens function diverges in the same way in noncommutative theory as in r r ¯ d 2r conventional theory. This was clear in the original plane d k d k d − p 1 kk¯+ikη¯ +ikξ+ηξ+ipy e− 2 . wave basis, and will imply that many loop amplitudes kk¯ + p2 + iǫ Z have exactly the same UV divergence structure as they It is convenient to express this as a proper time inte- would have had in conventional theory, as we will discuss gral. We can then easily include commuting dimensions in Sec. IV. as well; let there be 2r noncommuting and d 2r com- − We will find ourselves discussing more general Greens muting dimensions, with momenta k and p respectively. functions in the interacting theory. Given a two-point We include a separation y in the commuting directions function G(2)(k) in momentum space, the same procedure for purposes of comparison. We then have can be followed to convert it to the coherent state basis
r r d 2r tp2+ipy and thus interpret it in noncommutative position space. (η Gˆ ξ)= ∞ dt d k d k¯ d − p e− | | 0 We will not try to give general results but instead simply tkk¯ 1 kk¯+ikη¯ +ikξ+ηξ R e− R− 2 . (66) assert that this allows one to verify that the standard × relations between asymptotic behavior in momentum and The integrals over momenta are again Gaussian: position space are valid in noncommutative space.
2 d ∞ dt 2ηξ y First, the long distance behavior is controlled by the 2 ηξ 2t+1 4t = (2π) e d−2r e− − . (67) analytic behavior in the upper half plane; if the closest 0 (2t) 2 (2t + 1)r Z pole to the real axis is located at Im k = m, the Euclidean mr This is the result. Let us try to compare its behavior in Greens function will fall off as e− in position space. noncommutative position space, with that in commuta- Similarly, the short distance behavior is controlled by (2) α tive space. This comparison can be based on Eq. (59), the large k asymptotics, G (k) k− implies G(r) a2 α d ∼ ∼ which tells us that e− (η = a Gˆ ξ = a) evaluates the r − . | | 15
5. The membrane/hydrodynamic limit d + k-dimensional Yang-Mills theory, dimensionally re- d duces to d dimensions, takes these to be IRθ and takes d d d+k The formal relation between IRθ and the Heisenberg the limit, one obtains M = IR and N = IR . algebra of quantum mechanics suggests that it should be If some of the coordinates are commuting, one gets a interesting to consider the analog of the ~ 0 limit. gauged sigma model. The case of a single commuting → Since θ is not ~, this is not a classical limit, but rather a time-like dimension is particularly simple in the canoni- limit in which the noncommutative fields can be treated cal formulation: one has canonically conjugate variables as functions rather than operators. This limit helps i 2 (pi,y ), a Hamiltonian H = i pi + Sh, and a constraint provide some intuition for the noncommutative gauge i 0 = J = i pi,y . In this form, the construction symmetry, and has diverse interpretations in the various we just described{ essentially} P appears in Hoppe (1982), physical realizations of the theory. while its maximallyP supersymmetric counterpart is pre- d We consider noncommutative gauge theory on IRθ and cisely the light-cone gauge fixed supermembrane action of take the following scaling limit: let de Wit, Hoppe, and Nicolai (1988). Thus one interpreta- i 2 2 2 2 4 tion of the fields y is as the embedding of a d-dimensional θ = ℓ θ0; A = ℓ− A0; g = g0ℓ− (69) membrane into d + k-dimensional space. Although this picture is a bit degenerate if k = 0, the and take ℓ 0, keeping all quantities with the 0 sub- configuration space of such maps is still nontrivial. The script fixed.→ This corresponds to weak noncommutativ- related time dependent theory describes flows of a “fluid” ity with strong gauge coupling, fixing the dimensionful satisfying Eq. (72). This is particularly natural in two combination λ2 = gθ. dimensions, where θ dxidxj is the area form, and these Assume that θij is invertible, and define the functions ij are allowed flows of an incompressible fluid. This “hy- i i ij drodynamic picture” has also appeared in many works, y = x + θ Aj (x) (70) for example (Bordemann and Hoppe, 1993), and is also In the limit, the Yang-Mills action Eq. (29) becomes known in condensed matter theory. We shall use this hydrodynamic picture in what fol- 1 2 lows to illustrate solutions and constructions of the non- S = ddx yi,yj θij (71) h λ4 { }− commutative gauge theory. As an example, the hydro- Z i,j X dynamic limit of the Seiberg-Witten map Eq. (45) is where (Cornalba, 1999)
ij 1 k l 1 f,g = θ ∂ f∂ g Fij + θ− = [ y ,y ]− (73) { } 0 i j ij { } ij is the ordinary Poisson bracket on functions. As Susskind (2001) points out, for d = 2 this is just the Infinitesimal gauge transformations take the form Eq. translation between the Euler and Lagrange descriptions (32) in the limit. The corresponding finite gauge trans- of fluid dynamics. formations are general canonical transformations, also Similarly, the hydrodynamic analogue of the straight called symplectomorphisms, which are diffeomorphisms Wilson line is the Fourier transform of fluid density, x x˜(x) which preserve the symplectic form: 7→ i d 1 ikiy (x) 1 i j 1 i j W [k]= d x Pf(θ− )e (74) θij− dx˜ dx˜ = θij− dx dx . (72) ∧ ∧ Z In general, not all symplectomorphisms are generated by Hamiltonians as in Eq. (6); those which are not are the analogs of “large” gauge transformations. 6. Matrix representations The model Eq. (71) is a sigma model, in the sense that the fields yi can be thought of as maps from a “base In a formal sense, any explicit operator representation space” IRd to a “target space,” also IRd in this example. is a “matrix representation.” In this subsection we dis- Td One can easily generalize this to let y be a map from cuss θ as a large N limit of a finite dimensional matrix one Poisson manifold (M,θ) to another Poisson manifold algebra, and how this might be used to formulate regu- (N, π), which must also have a volume form µ and a lated noncommutative field theory. We take d = 2 for metric 2 on the space of bi-vectors. The action Eq. definiteness, but the ideas generalize. (71) willk·k read: One cannot of course realize Eq. (1) using finite dimen- sional matrices. One can realize Eq. (23) for the special 12 M 2 case with θ = 2πM/N, for example by U1 = Γ1 and S = µ y θ π n U2 = Γ2, where (Γ1)m,n δm,n exp2πi is the “clock” N k ∗ − k ≡ N Z matrix, and (Γ2)m,n δm n,1(modN) is the shift matrix. ≡ − and will again have the group Diffθ(M) of symplecto- Products of these matrices form a basis for MatN (C) morphisms of M as a gauge group. This generalization (we assume gcd(M,N) = 1) and in a certain sense this T2 also appears very naturally; for example if one starts with allows us to regard MatN (C) as an approximation to θ: 16 namely, if we write unitary matrices U U(N), and the action i ∈ ij i m1 m2 πiθ mimj 2πimix 1 Φ[x] U1 U2 e − , † † ≡ S = 2 Zij Tr UiUjUi Uj . 0 m ,m 2 E = Tr [z, φ][φ, z¯]+ θTr V (φ). V ′(φ)= gφ + φ =0. (77) The kinetic term breaks U( r) symmetry and might be While in commutative theory these equations would ad- expected to destabilize mostH of the infinite dimensional mit only constant solutions, in noncommutative theory space of solutions Eq. (78). On the other hand, for low the story is rather more interesting. It is simplest in the modes ( n n with n 1), the kinetic energy will be Fock space basis, in which the field can be taken to be an O(1), so| forih sufficiently| ∼ large θ a stable solution should arbitrary (bounded) operator on r, and for which the et al. H survive. This was checked by Gopakumar (2000c) multiplication is just operator multiplication. Since φ is by an analysis of linearized stability, with the result (for self-adjoint it can be diagonalized, so we can immediately the n = 1 solutions in d = 2 + 1) that only the minimal write the general solution of Eq. (77): Gaussian 0 0 and its translates are stable. The solution| ih | cannot exist at θ = 0 and it is interesting 1 φ = U †PU to ask what controls the critical value θc at which it dis- −g appears (Zhou, 2000; Durhuus, Jonsson, and Nest, 2001; Jackson, 2001). One can easily see the rather surpris- where U is a unitary and P 2 = P is a projection op- ing fact that this does not depend directly on the barrier erator, characterized up to unitary equivalence by Tr P height. This follows because we can obtain a family of (which must be finite for a finite energy configuration) or equivalent problems with very different barrier heights by equivalently the number of unit eigenvalues. As discussed the rescaling φ aφ and E E/a2, all with the same in Sec. II.E.3, the diagonal operators correspond to ra- θ . → → dially symmetric solutions, from which unitary rotations c Rather, the condition for noncommutative solitons to produce all solutions. exist is that noncommutativity be important at the scale The simplest solution of this type uses the operator set by the mass of the φ particle in the asymptotic vac- P = 0 0 , of energy 2πθV ( 1/g) = πθ/3g2. It is re- 0 uum, i.e. θV >> 1. Consider the symmetric φ4 poten- markable| ih that| this energy depends− only on the value of ′′ tial. By the above argument, θ can only depend on V the potential at the critical point, and nothing else. c ′′ at the minimum; numerical study leads to the result As we discussed in Sec. II.E.2, this is a Gaussian of width √θ which squares to itself under star product. θcV ′′(0) = 13.92. (79) We see that the scaling argument for instability is vio- lated because of the position-space uncertainty principle, There is some theoretical understanding of this result, which causes the energy of smaller Gaussians to increase. which suggests that this critical value is roughly inde- 1 pendent of the shape of the potential. It is plausible The general solution in this sector is φ = U † 0 0 U. − g | ih | This includes Gaussians with arbitrary centers, the that only radially symmetric configurations are relevant higher modes discussed in Sec. II.E.2, and various for stability and if one restricts attention to this sector, “squeezed states.” Neglecting the kinetic term, they the equation of motion reduces to a simple three-term are all degenerate and are parameterized by an infinite recurrence relation for the coefficients cn in Eq. (61), dimensional moduli space “limN U(N)/U(N 1).” θ →∞ − (n + 1)cn+1 (2n + 1)cn + ncn 1 = V ′(cn); n 0. This infinite degeneracy will however be lifted by the − − 2 ≥ kinetic term. The story is similar for solutions with (80) 18 Suppose that V (φ) is bounded below, the vacuum is φ = Of course, it is solved by the vacuum configuration C =z ¯ 0, and we seek a solution which approximates the θ = and C¯ = z, and gauge transformations of this, one soliton solution (c = λ; c =0,n> 0). Finiteness∞ of 0 n ¯ the energy requires limn cn = 0, and we can get the C = U †zU¯ ; C = U †zU. (83) large n asymptotics of such→∞ a solution by ignoring the What is amusing, is that U does not need to be unitary nonlinear terms in V ′(λ); this leads to in order for this transformation to produce a solution 1/4 √nθV ′′(0) cn n e− . (81) (Witten, 2000; Harvey, Kraus, and Larsen, 2000a). It ∼ needs to satisfy UU † = 1, but U †U need not be the This shows that c varies smoothly when θV ′′(0) 1, n ≪ identity. This implies the partial isometry condition Eq. and can be approximated by a solution of the differential (53) and is a bit stronger. equation analog of Eq. (80). However, corresponding to The simplest examples use the shift operator Eq. (54): the nonexistence of a solution in the commutative theory, U = Sm. To decide whether these are vortex solutions, one can show on very general grounds that no solution we should compute the magnetic flux. This is of this differential equation can have the boundary value n 1 c(0) = λ, even with the nonlinear terms in V ′ included. m m m m − A nontrivial soliton is possible only if this continuous F = (S†) [¯z,z]S +1=1 (S†) S = Pn − n=0 approximation breaks down, which requires the control X parameter θV ′′(0) to be large. In particular, a nontrivial one soliton solution must where Pn n n . The total flux is Tr F = m. Thus noncommutative≡ | ih Maxwell| theory allows nonsingular vor- have a discontinuity c0 c1 >> 0, and matching this on to Eq. (81) for c provides| − | a lower bound for θV (0). tex solutions, sometimes called “fluxons,” without need- 1 ′′ ing a scalar field or the Higgs mechanism. This analysis leads to an estimate which is quite close to Eq. (79). Physically, we might interpret this as an noncommu- tative analog of the commutative gauge theory vortex Multi-soliton solutions have been studied recently in 1 imθ (Lindstrom, Rocek, and von Unge, 2000; Gopakumar, Ai = g− ∂ig with g = e . This is pure gauge except at Headrick, and Spradlin, 2001; Hadasz, Lindstrom, Rocek, the origin where it is singular, so we might regard this as Td an example of noncommutative geometry smoothing out and von Unge, 2001). Solutions on θ have been studied in (Bars, Kajiura, Matsuo, and Takayanagi, 2001a). a singularity. The soliton mass M is proportional to Tr F 2 = m. To restore the dependence on the coupling constants, we B. Vortex solutions in gauge theories first note that (as for any classical soliton) the mass is proportional to 1/g2. This quantity has dimensions of Derrick’s theorem does not hold in gauge theories and length in 2 + 1 dimensions, so on dimensional grounds as is well known, the abelian Higgs model (Maxwell the- the mass must be proportional to 1/θ, consistent with ory coupled to a complex scalar field) has vortex solu- the nonexistence of the fluxon in the conventional limit. tions, which (among other applications) describe the flux In the conventions of Eq. (82), tubes in superconductors (Nielsen and Olesen, 1973). πm We proceed to discuss analogous solutions in the M = . (84) g2θ noncommutative gauge theory (Nekrasov and Schwarz, 1998; Polychronakos, 2000; Gross and Nekrasov, 2000a; The most general solution is slightly more general than Aganagic, Gopakumar, Minwalla, and Strominger, 2001; this; it is Bak, 2000). We work in 2 + 1 dimensions with θxy = 1 m m m 1 0 and look for time-independent solutions in A = 0 gauge. C = (S†) zS¯ + − c (x ) n n (85) 0 n=0 n | ih | Using Eq. (35) and the Fock basis, the energy is m m m 1 0 C¯ = (S†) zS + − c¯ (x ) n n . (86) Pn=0 n | ih | 2π 1 2 E = dtTr F + Diφ Diφ† + V (φ). (82) P 2 2 g2 2 The 2m functions cn,¯cn must satisfy ∂0 cn = ∂0 c¯n = 0 i=1,2 Z X and can be seen to parameterize the world-lines of the m Here fluxons. This is particularly clear for m = 1 by recalling Eq. (34). F = [C, C¯]+1 A peculiar feature to note is that fluxons exist with and φ is a complex scalar field (satisfying no hermiticity only one sign of magnetic charge, F aligned with θ. It condition). is also rather peculiar that they exert no force on one We first note that unlike Maxwell theory, even the pure another; the energy of the configuration is independent of their locations. rank one noncommutative gauge theory admits finite en- 2 2 ergy solitonic solutions. The pure gauge static equation Even more peculiar, the equations ∂0 cn = ∂0 c¯n = 0 of motion is admit as solutions cn = xn + vnt andc ¯n =x ¯n +v ¯nt with no upper bound on v. In other words, the fluxons 0 = [C, [C, C¯]] can move faster than light (Hashimoto and Itzhaki, 2000; 19 Bak, Lee, and Park, 2000). Of course θ defines a preferred used to generate many exact solutions (Hamanaka and rest frame, and there is no immediate contradiction with Terashima, 2001; Harvey et al., 2000a; Hashimoto, 2000; causality in this frame. Schnabl, 2000; Tseng, 2000; Bergman, Ganor, and Kar- These peculiarities may be made more palatable by czmarek, 2001). the realization that all of these solutions (with no scalar Of course the properties of these solutions, including field) are unstable, even to linearized fluctuations. For stability, depend on the specific form of the scalar poten- example, the m = 1 solution admits the fluctuation tial and the choice of matter representation; both “ad- joint” matter with Diφ = [Ci, φ], and fundamental mat- C + T = S†zS¯ + tS†P0 (87) ter as defined in Sec. II.D.3 have been studied. A particularly nice choice of potential (Jatkar, Mandal, where t is a complex scalar parameterizing the fluctuation and Wadia, 2000) is and S†P = 1 0 . One can straightforwardly compute 0 | ih | 1 2 2 [C + T, C¯ + T¯]+1= P + t 2(P P ). V = (φφ† m ) , 0 | | 1 − 0 2 − The total flux Tr F is constant under this variation, while as in this case the energy can be written as the sum of the energy is proportional to squares and a total derivative as in (Bogomolny, 1976), leading to a lower bound E Tr F . Solutions saturat- Tr F 2 = (1 t 2)2 + t 4 ing this bound are called BPS≥ | and are| clearly stable. − | | | | For adjoint matter and θm2 = 2, the exact solution dis- which exhibits the instability. This is of course just like cussed above is BPS. For fundamental matter, the bound the commutative case; the flux will tend to spread out can be attained in two ways.3 One can have over all of space in the absence of any other effect to 2 confine it. F = m φφ†; Dφ¯ =0, − There is no stable minimum in this topological sector. This follows if we grant that Tr F is “topological” and which has positive flux solutions. For θm2 = 1 the exact cannot change under any allowed variation of the fields. solution as above with φ = S† is BPS. Bak et al. (2000) This of course depends on one’s definitions, but in con- show that BPS solutions only exist for θm2 1, and 2 ≤ ventional gauge theory one can make a definition which argue that for θm > 1 the exact solution is a stable does not allow flux to disappear. Starting with a local- non-BPS solution. ized configuration, the flux may disperse or go to infinity, One can also have but one can always enlarge one’s region to include all of 2 F = φφ† m ; Dφ =0. it, because of causality. Similarly, energy cannot be lost − at infinity. Since noncommutative theory is not causal and indeed This has been shown to have a negative flux solution for 2 et al. a fluxon can move faster than light, this argument must any value of θm (Jatkar , 2000; Lozano, Moreno, and Schaposnik, 2001), which in the θ 0 limit reduces be reexamined. The belief at present is that flux and → energy are also conserved in noncommutative gauge the- to the conventional Nielsen-Olesen solution. ory; they are conserved locally and cannot not run off The vortex solutions admit a number of direct general- to infinity in finite time. In the case of the fluxon, a izations to higher dimensional gauge theory. If one has 2r given solution will have some finite velocity, and as it noncommuting coordinates, one can make simple direct disperses, it would be expected to slow down, so that products of the above structure to obtain solutions local- ized to any 2(r n)-dimensional hyperplane. One can once it has spread over length scales large compared to − √θ conventional causality will be restored. This point also introduce additional commuting coordinates, and it could certainly use more careful examination however. is not hard to check that the parameters cn,c ¯n in the solution Eq. (85) must then obey the wave equation This physical statement underlies the conventional def- ij inition which leads to topological sectors characterized 0= η ∂i∂j c (resp.c ¯) in these coordinates. This is as ex- by total flux; one only considers variations of the fields pected on general grounds (they are Goldstone modes for which preserve a specified falloff at infinity. An analo- space-time translations and for the symmetries δφ = ǫ) gous definition can be made in noncommutative theory, and fits in with the general philosophy that a soliton in as we will discuss in Sec. VI.A. d + 1-dimensional field theory which is localized in d p dimensions, should be regarded as a “p-brane,” a dynam-− One can use the same idea to generate exact solutions ical object with a p + 1-dimensional world-volume which to the noncommutative abelian Higgs theory, even with a general scalar potential V (φ). Now one starts with the scalar in a vacuum configuration V ′(φ) = 0 and Diφ =0 (so φ 1), and again applies an almost gauge trans- ∝ 3 In the references, the two types of solution are sometimes re- formation with UU † = 1, but U †U = 1. The same ar- 6 ferred to as “self-dual” and “anti-self-dual,” but not consistently. gument as above shows that this will be a solution for It seems preferable to speak of positive flux (F aligned with θ) any U. This “solution generating technique” has been and negative flux. 20 can be described by fields and a local effective action on It turns out that the ADHM construction can be the world-volume. While the vortex in 3 + 1-dimensional adapted very readily to the noncommutative case gauge theory, which is a string with p = 1, may be the (Nekrasov and Schwarz, 1998). Let us quote the re- most familiar case, the story in more dimensions is en- sult for anti-self-dual gauge fields, P +F = 0, referring tirely parallel. to Nekrasov (2000) for proofs, further explanations and One can get a nontrivial fluxon solution in the hydro- generalizations. See also Furuuchi (2000). dynamic limit (as in Sec. II.E.5 with k = 0) by rescaling To construct charge N instantons in the U(k) gauge d the magnetic charge as m ℓ− . The solution becomes theory we must solve the following auxiliary problem in- ∼ volving the following finite-dimensional matrix data. Let Ld Xi,i = 1, 2, 3, 4, be a set of N N Hermitian matrices, yi = xi 1 , r>L, × d transforming as a vector under SO(4) space rotations r − r yi =0, r L and in the adjoint of a ‘dual’ gauge group U(N). Let ≤ 2 i 2 λα, α =1, 2 be a Weyl spinor of SO(4), transforming in r = (x ) (88) (N, k) of U(N) U(k). Instanton solutions will then be i × X in correspondence with solutions of the following set of equations: The vortex charge m (L/ℓ)d is no longer quantized, but it is still conserved.∝ + kl i j ¯ i j ij 1 0 = (P )ij [X ,X ]+ λσ σ¯ λ θ N N . (92) − × i These equations admit a U(N) symmetry (X , λα) C. Instantons i 1 1 7→ (gN X gN− , λαgN− ), and two solutions related by this symmetry lead to the same instanton solution. Thus To obtain qualitatively new solutions of gauge theory, the moduli space of instantons is the non-linear space we must move on to four Euclidean dimensions. As is well of solutions of Eq. (92) modulo U(N) transformations. known, minima of the Euclidean action will be self-dual By counting parameters, one finds that it has dimension and anti-self-dual configurations, 4Nk. mn The only difference with the conventional case is the (P ∓)kl Fmn = 0 (89) shift by θ. This eliminates solutions with λ = λ¯ = 0, mn 1 m n 1 mn which would have led to a singularity in the moduli (P ±) δ δ ε , (90) kl ≡ 2 k l ±4 kl space. In conventional Yang-Mills theory, instantons + have a scale size parameter which can be arbitrary (the where P and P − are the projectors on self-dual and classical theory has conformal invariance), and the sin- anti-self-dual tensors. These solutions are classified topo- gularity is associated with the zero size limit. The scale logically by the instanton charge: size is essentially ρ √λλ¯ , and in the noncommuta- ∼ 1 tive theory is bounded below at √θ, again illustrating N = Tr F F (91) −8π2 ∧ that position-space uncertainty leads to a minimal size Z for classical solutions. Note that θ breaks parity symmetry, and the two types To find the explicit instanton configuration corre- of solution will have different properties. Let θ be self- sponding to a solution of Eq. (92), one must solve an dual; then one can obtain self-dual solutions by the direct auxiliary linear problem: find a pair (ψα, ξ), with the product construction mentioned in the previous subsec- Weyl spinor ψα taking values in N k matrices over θ, tion, while the anti-self-dual solutions turn out to be the and ξ being a k k -valued matrix,× such that A × Aθ noncommutative versions of the Yang-Mills instantons. i Instantons play a central role in the nonperturbative X σiψ + λξ =0 physics of Yang-Mills theory (Schafer and Shuryak, 1998) ψ†ψ + ξ†ξ = 1k k (93) and have been studied from many points of view. The × most powerful approach to constructing explicit solutions Then the instanton gauge field is given by and their moduli space is the so-called ADHM construc- Ai = ψ†∂iψ + ξ†∂iξ (94) tion (Atiyah, Drinfeld, Hitchin, and Manin, 1978), which j j C = iθ ψ†x ψ + ξ†x ξ . (95) reduces this problem to auxiliary problems involving sim- i − ij ple algebraic equations. Although when it was first pro- One can make these formulae more explicit in the k = posed, this construction was considered rather recherch´e N = 1 case. We follow the conventions of Eq. (6). Let by physicists (Coleman, 1985), modern developments in θ (ˆn + 1) string theory starting with (Witten, 1996) have placed Λ= α α α ; D = C + iC ; D = C + iC . it in a more physical context, and in recent years it has θ nˆ 1 1 2 2 3 4 P α α α formed the basis for many practical computations in non- perturbative gauge theory, e.g. see (Dorey, Hollowood, Then P and Khoze, 2000). We will explain the stringy origins of 1 1 1 Dα = SΛ− 2 z¯αΛ 2 S†, the construction in Sec. VII.D. √θα 21 where the operator S is the shift operator in Eq. (54). We start with the fact that in three spatial dimen- This solution is nonsingular with size √θ. One cantake sions (we assume only spatial noncommutativity), since the limit θ 0 to return to conventional Maxwell theory, the Poisson tensor has at most rank two, there will be obtaining a→ configuration which is pure gauge except for a commutative direction. Call its coordinate x3. This a singularity at the origin. Whether or not this counts theory will admit the vortex solution of III.B, a string as an instanton depends on the underlying definition of extending to infinity along x3, say the theory; in string theory we will see later that it is. B = 0 0 . 3 | ih | This can be used to find a simple rank 1 solution of Eq. D. Monopoles and monopole strings (99) by postulating A3 = 0 and One might at this point suspect that in general non- φ = x3 0 0 . commutative gauge theory solitons look quite similar to | ih | their Yang-Mills counterparts, if perhaps less singular. Now, we might not be tempted to call this solution a The monopole will prove an exception to this rule. monopole. However it is only one point in a moduli space We consider static field configurations in 3 + 1 gauge of solutions. The linearized equation of motions around theory with an adjoint scalar field φ, and the energy func- the solution take the form tional i i D Diδφ =0, D DiδAj =0, DjδAj + [φ, δφ]=0. 1 3 E = Tr (F )2 + ( φ)2. (96) 4g2 ij ∇i Besides variations of the location of the vortex in the YM 1 i Just as in the conventional theory, one can rewrite this (x3)2/θ δφ e− 0 1 as a sum of a total square and a total derivative: ∝ | ih | 1 Turning on this mode corresponds to splitting the string E = d3xTr ( φ B )2 Z (97) in two, as one can see by looking at the eigenvalues of 4g2 ∇i ± i ∓ YM Z the operator φ + δφ. 1 3 This linearized variation extends to a finite variation Z = d xTr Di (Bi ⋆ φ + φ⋆Bi) (98) 4g2 of the solution which can be used to send one half YM Z of the string off to infinity. The remaining half is a The total derivative term Z depends only on the bound- rank 1 monopole attached to a physical analog of the ary conditions and in the conventional case would have Dirac string, the flux tube of III.B stretching from the been proportional to the magnetic charge of the soliton. monopole to infinity carrying magnetic flux and energy, Minimizing the energy with fixed Z leads to E = Z and | | and cancelling the monopole magnetic flux at infinity. Its the equations (Bogomolny, 1976) energy diverges, but precisely as the string tension Eq. (84) times the length of the string. Thus the noncommu- φ = B . (99) ∇i ± i tative gauge theory has found a clever way to produce a We are only going to consider nonsingular configura- solution despite the absence of U(1) magnetic charge in three dimensions. tions, and the first observation to make is that this im- One can express the solution in closed form in terms plies that the diagonal part of the total magnetic charge of error functions (Gross and Nekrasov, 2000b). Self- is zero, dual solutions to gauge theory generically admit closed 3 i form expressions; the deep reason for this is the inte- 0= Q = d xTr [Di,B ], grable structure of these equations (Ablowitz and Clark- Z son, 1991). The noncommutative deformation respects by the Bianchi identity, analogous to the conventional this; for example the noncommutative Bogomolny equa- theory. tions for axially symmetric monopoles are equivalent to This would seem to rule out the possibility of a non- nonabelian Toda lattice equations. singular rank 1 monopole solution. Nevertheless, it turns Multimonopoles also exist, and as for the vortices, this out that such a solution exists, as was discovered by ex- moduli space will have limits in which it breaks up into plicit construction (Gross and Nekrasov, 2000b) making U(1) monopoles. In particular, the noncommutative ana- use of the Nahm equations (Nahm, 1980). These equa- log of the ’t Hooft-Polyakov monopole of the U(2) theory tions are very analogous to the ADHM equations and is better thought of as a multimonopole solution, with admit an equally direct noncommutative generalization. two centers associated to the two U(1) subgroups, con- They are ordinary differential equations and their anal- nected by a string. ysis is somewhat intricate; we refer to (Nekrasov, 2000) A simple observation following from the form of the for this and approach the solution in a more elementary noncommutative Nahm equations is that (in contrast to way. the instantons) this multimonopole moduli space is the 22 same as that of the conventional theory. Since the con- ventional solutions and moduli space were already non- singular, this fits with the general picture of desingu- larization of solitons by noncommutativity, but is still non-trivial (it was not forced by symmetry). There are pretty string theory explanations for all of this, discussed in Sec. VII.D. In string theory, monopoles turn out naturally to be associated with “D-strings” ex- tending in a higher dimension. In commutative gauge FIG. 2 Double line notation and phase factors. theory, the string extends perpendicular to the origi- nal dimensions, and its projection on these is point-like. Noncommutativity tilts the string in the extra dimen- ij where (as in Eq. (21)) k k′ θ kikj′ . The same holds sions, leading to a projection onto the original dimensions in the presence of derivative× interactions,≡ multiple fields which is itself a string. and so on. The factor Eq. (100) is not permutation symmetric but is only cyclically symmetric. This is just as for a matrix IV. NONCOMMUTATIVE QUANTUM FIELD THEORY field theory, and the same double line notation can be used to represent the choice of ordering used in a specific In this section we discuss the properties of quantized diagram (’t Hooft, 1974; Coleman, 1985). In this nota- noncommutative field theories. We start by developing tion, propagators are represented as double lines, while the Feynman rules and explain the role of planarity in the ordering of fields in a vertex is represented by con- organizing the perturbation theory (Filk, 1996). necting pairs of lines from successive propagators. We then give examples illustrating the basic structure This additional structure allows defining “faces” in a of renormalization: the UV properties are controlled by diagram as closed single lines and thus the Euler charac- the planar diagrams, while nonplanar diagrams generally ter of a graph can be defined as χ = V E + F , with lead through what is called “UV/IR mixing” (Minwalla V , E and F the numbers of vertices, edges− and faces et al., 2000) to new IR phenomena. The limit θ 0 respectively. For connected diagrams, χ 2 with the in these theories is non-analytic. We cite a range of→ ex- maximum attained for planar diagrams, those≤ which can amples in scalar and gauge theory which illustrate the be drawn in a plane without crossing lines. A diagram possibilities; there is some physical understanding of its with χ = 2 2g for g 1 is nonplanar of genus g and consequences. We also discuss the question of gauge in- can be drawn− on a surface≥ of genus g without crossings. variance of the effective action. In matrix field theory, summing over indices provides a We then discuss a variety of results for physical ob- factor N F , while including appropriate N dependence in servables, including finite temperature behavior and a the action (an overall prefactor of N will do it) completes universal high energy behavior of the Wilson line opera- this to a factor N χ. This is the basis for the famous ’t tor. We also outline the Hamiltonian treatment of non- Hooft limit in which the dynamics of certain field theories commutative gauge theory on a torus, and conclude with is believed to reduce to that of a free string theory: as miscellaneous other results. N , planar diagrams dominate. → ∞ In noncommutative field theory, planarity plays an im- A. Feynman rules and planarity portant role in organizing the phase factors Eq. (100). The basic result is that a planar diagram in noncommu- We start by considering the theory of a single scalar tative theory has the same amplitude as that of conven- field φ with the action tional theory multiplied by an overall phase factor Eq. (100), where the momenta ki are the ordered external 2 1 m gn momenta. This can be seen by checking that the dia- S = Tr (∂φ)2 V (φ); V (φ) φ2 + φn. 2 − ≡ 2 n! grams in figures 3 and 4 can each be replaced by a single n>2 Z X vertex while preserving the phase factor, and that these Functional integral quantization can be done in the stan- operations can be used to reduce any planar diagram dard way and leads to Feynman rules which are almost with n external legs to the single vertex Tr φn. the same as for conventional scalar field theory. In par- Since this additional phase factor is completely inde- ticular the propagator is the same thanks to Eq. (22). pendent of the internal structure of the diagram, the con- The only difference is an additional momentum depen- tribution of a planar diagram to the effective action will dence in the interaction vertices following from Eq. (21): be the same in noncommutative theory and in the corre- n a φ vertex with successive incoming momenta kµ has sponding θ = 0 theory (the noncommutative phase factor the phase factor will be reproduced in the course of evaluating the vertex in the effective action). In particular, divergences and i exp k k (100) renormalization will be the same for the planar subsec- −2 µ × ν 1 µ<ν FIG. 5 A non-planar diagram. FIG. 3 The contraction (Tr φm)(Tr φn) ↔ Tr φm+n−2. In practice, this principle might be obscured by the presence of other divergences. A more precise statement one can make is that in the limit θij , with fixed external momenta, UV and IR cutoffs,→ integrating ∞ over this phase factor would cause the general nonplanar dia- gram to vanish. Thus this limit would keep only planar diagrams. This observation goes back to the early works on the twisted Eguchi-Kawai models, as mentioned in the introduction, and indeed was the main focus of interest in these works. B. Calculation of nonplanar diagrams The most important features are already visible in the one-loop renormalization of the scalar field theory propa- gator. We consider two examples, the φ4 theory and the 3 FIG. 4 Tadpoles come with no phase factor. φ theory in d Euclidean noncommutative dimensions. We will not be careful about (1) numerical factors in our discussion and often substituteO generic positive real We can obtain the phase factor for a nonplanar dia- constants c, c′, etc., to better make the qualitative points. gram by comparison with the planar case. Given a spe- We will try to keep track of signs. 4 cific diagram, we choose a way to draw it in the plane, The φ theory has two one-loop self-energy diagrams, now with crossing propagators. Let Cµν be the intersec- one planar and one nonplanar. They contribute tion matrix, whose µν matrix element counts the num- d Γ(2) = g4 d k (102) ber of times the µ’th (internal or external) line crosses 1 planar 3(2π)d k2+m2 d the ν’th line (with sign; a crossing is positive as in the (2) g4 d k ik p Γ = d R2 2 e × . (103) figure). Comparing this diagram with the correspond- 1 nonplanar 6(2π) k +m ing planar diagram obtained by replacing each crossing The planar contribution is proportionalR to the one-loop with a vertex, we find that it carries the additional phase diagram of the θ = 0 theory, and is divergent for d 2. factor If we introduce a momentum space cutoff Λ, we will≥ find i µν (2) g4 d 2 exp C kµ kν . (101) Γ1 planar = d Λ − + , −2 × 2 2 · · · µ,ν 3(4π ) X a mass renormalization which can be treated by standard Although the matrix Cµν is not uniquely determined by renormalization theory. the diagram, the result Eq. (101) is. Compared to this, the nonplanar contribution has an Since the phase factor Eq. (101) depends on the in- oscillatory factor eik p, arising from the noncommutative ternal structure of the diagram, nonplanar diagrams can × nature of space-time, which will play the role of a cutoff. have very different behavior from their θ = 0 counter- On general grounds, this cutoff will come in for momenta parts. Since the additional factor is a phase, one would k such that d/dk(k p) (1), i.e. ki Λ θij p 1, expect it to improve convergence of integrals, leading to p j − and produce × ∼ O ∼ ≡ | | better UV behavior. This expectation will be borne out below, leading to the principle that the leading UV di- (2) d 2 Γ g4(cΛ − + ). vergences (in particular the beta function) come from the 1 nonplanar ∼ p · · · planar diagrams (no matter what the rank) and are thus This cutoff goes to infinity as we take the limit p 0 the same as for the large N limit of a matrix field theory. of the external momentum. To get a finite result in→ this 24 limit, we will need to keep the original cutoff Λ in the f(Λ2), will obtain an IR divergence f(1/p p) in the calculation as well. This can be done in many ways; noncommutative theory. One might further expect◦ that, 1/Λ2t the usual approach is a proper time cutoff e− in the if the conventional amplitude had no IR divergences, integrand of Eq. (67) which leads to the result the leading IR divergence of the noncommutative the- ory would take precisely this form. We will refer to this 1 Λ2 ; (104) as “the standard UV/IR relation.” It is typical but not p ≡ 1/Λ2 + p p universal and in particular is violated in gauge theory. ◦ ik jk Another example is the one-loop self-energy diagram p p piθ θ pj (105) ◦ ≡ in φ3 theory (figure 5). The expectations we just stated (in fact this is the simplest function with the qualitative are indeed valid (Van Raamsdonk and Seiberg, 2000) and behavior we want). The notation p p for the quantity we find controlling the cutoff is standard; it◦ is positive definite cg2M 2 if θ0i = 0 (purely spatial noncommutativity). If θij has Γ(2)(p)= p2 + M 2 + 3 + . . . (109) (p p +1/Λ2)d 4 maximal rank and all its eigenvalues are θ, then p p = ◦ − θ2p2. If there are commutative directions± as well,◦ they where M 2 is again the renormalized mass. At d = 4 this will not enter into this quantity. becomes a logarithmic divergence. Adding the classical and one-loop contributions to the The effects we have seen so far drastically change the two-point function leads to a 1PI effective action IR behavior and we refer to this case as “strong UV/IR Γ[φ]= ddp 1 φ(p)φ( p)Γ(2)(p)+ . . . (106) mixing.” There are also theories, particularly supersym- 2 − metric theories, in which the IR behavior is not quali- Γ(2)(p)= p2 + M 2 + cg4 + (107) R (p p+1/Λ2 )d−2 · · · tatively different from conventional theory, and we can ◦ speak of “weak UV/IR mixing.” This would include 2 2 d 2 where M = m +c′g4Λ − is the renormalized mass. At models where renormalization is logarithmic and only af- d = 2 the powerlike divergences become logarithmic in fects the kinetic terms (assuming the standard UV/IR re- the usual way. lation), such as the d = 4 Wess-Zumino model (Girotti, The novel feature of this result is that the limit Λ Gomes, Rivelles, and da Silva, 2000). → ∞ is finite, and the UV divergence of the nonplanar diagram In proceeding to higher loop orders, one faces the dan- has been eliminated, but only if we stay away from the ger that the new IR divergences will mix with other di- IR regime p 0. The limiting theory has a new IR vergences in an uncontrollable way. In fact the IR di- → divergence, arising from the UV region of the momentum vergences under discussion are rather similar to those integration. This type of phenomenon is referred to as appearing in thermal field theory, in that they give a “UV/IR mixing” and is very typical of string theory, but large effective mass to low momentum modes (we will dis- is not possible in local field theory. However, it is possible cuss this point further below), and can be addressed by in noncommutative theory, thanks to the nonlocality. similar techniques (Fischler, Gomis, Gorbatov, Kashani- One way to see that this is not possible in local field Poor, Paban, and Pouliot, 2000a; Gubser and Sondhi, theory is to observe that it contradicts the standard 2000; Griguolo and Pietroni, 2001a). dogma of the renormalization group. Let us phrase this One easy way to do this is to add and subtract a coun- in terms of a Wilsonian effective action, defined by inte- terterm to the bare action which represents the leading grating out all modes at momentum scales above a cutoff divergence, so that it can be taken into account in the Λ. The result is an effective action bare propagator. This leads to the next correction to Eq. (107) being SW ilson[φ;Λ] = Z(Λ) g2(Λ)Z2(Λ) d ik p 4 2 2 2 4 (2,1) d k 1+ e × cg4 d x 2 (∂φ) + m (Λ)φ + 4! φ(108). Γ (p)= c′g4 2 2 2 d d 2 . k + M + cg4(p p) − − (p p) − RenormalizabilityR in this framework means that one can Z ◦ ◦ choose the functions Z(Λ), m(Λ) and g2(Λ) in such a way with the IR divergence explicitly subtracted. that correlation functions computed with this Lagrangian The main feature of this result is that the IR divergent term in the propagator (for d 2) causes it to vanish at have a limit as Λ , and correlation functions com- ≥ puted at finite Λ differ→ ∞ from their limiting values by terms low momenta, and thus the resummed loop corrections of order 1 for all values of momenta. are controlled in this region. Λ (2) However, one sees from Eq. (107) that this is man- In fact, at weak coupling these corrections to Γ (p) ifestly untrue of noncommutative scalar φ4 theory, and are small for all momenta. They do exhibit nonanalyt- indeed the generic noncommutative field theory with UV icity in the coupling similar to that in finite tempera- divergences. The two point function at any finite value ture field theory (Griguolo and Pietroni, 2001a), begin- 3 3/2 of Λ differs drastically from its Λ value, for small ning at (g4 log g4) in the massive theory and (g4 ) enough momenta ((pθ)2 << 1/Λ2).→ ∞ in the masslessO theory. This brings new difficultiesO into The arguments above lead to the general expectation the perturbation theory; one could try to work at finite that a nonplanar diagram with UV cutoff dependence coupling by using a self-consistent Hartree equation, or 25 any of the many other approximation methods available Once one observes new IR effects at the quantum level, for field theory. the first question one must ask is whether the original In any case, this discussion appears to justify consider- perturbative vacuum φ = 0 (the disordered phase, in the ing Eqs. (107) and (109) as valid approximations to the language of statistical field theory) is stable, or whether Greens function at weak coupling, and we discuss physics these effects are a signal that we are expanding around following from this idea in the next subsection. It falls the wrong vacuum. This question could be answered if short of a proof, which would require discussing (at the we knew the exact quantum effective action; in particu- least) all Greens functions which obtain large UV renor- lar the perturbative vacuum will be at least metastable malizations in conventional theory, such as the four-point if the inverse propagator Γ(2)(p) in Eq. (106) is positive function in φ4 theory in d 4. The argument we just at all momenta. This includes the usual condition on the ≤ gave generalizes to some extent to these diagrams as their effective potential V ′′(0) > 0 but since the effects we are IR corrections will be controlled by the propagator in the discussing modify the dispersion relation so drastically, same way. This might fail for special values of external we need to entertain the possibility that a phase transi- momenta however, and a real proof of these ideas has tion could be driven by modes with non-zero momentum, not yet appeared. Relevant work includes (Chepelev and perhaps leading to a “stripe” phase as one finds in certain Roiban, 2000). condensed matter systems. See also (Chen and Wu, 2001; Li, 2000; Kinar, Lif- This is a real possibility here, as is clear from Eqs. schytz, and Sonnenschein, 2001). (107) and (109). One expects and it might be possible to prove that with certain hypotheses, the standard UV/IR relation will hold for the exact (cutoff) quantum effective C. Physics of UV/IR mixing action. This would mean that Γ(2)(p) ∆M 2(1/pθ) in theories whose planar limit has a UV divergent∼ mass The UV/IR mixing observed by Minwalla et al. (2000) renormalization ∆M 2(Λ). appears to be the main qualitative difference between Then the relevant question is whether this mass renor- conventional and noncommutative perturbation theory. malization is positive or negative (for large Λ). If it is One cannot say that its full significance, and the issue negative, the perturbative vacuum is clearly unstable to of whether or not it spoils renormalizability or leads to condensing p 0 modes. On the other hand, if it is pos- other inconsistencies, is well understood at present. On itive, the resulting→ dispersion relation will make the low the other hand, there is an emerging picture which we p modes “stiff” and there might or might not be a phase will outline. transition, but if there is, it will be driven by a mode A reason for caution at this point is that, for the rea- with p = 0. Gubser and Sondhi (2000) have argued that sons we just discussed, we cannot blindly assume that the such a phase6 transition is indeed expected (in scalar field general framework of the renormalization group, which theory in d> 2) and will be first order. underlies most of our understanding in the conventional case, is applicable. This is not to say that it is obviously If we grant that the perturbative vacuum is stable, we inapplicable, but that to justify it one must show how can go on to ask about the meaning of the propagator (2) 1 to clearly separate UV and IR phenomena, taking into Γ (p)− . This discussion will depend on whether the account the higher loop subtleties mentioned above. noncommutativity is purely spatial (θ0i = 0) or has a If this can be done, justifying the idea that the high time-like component. energy behavior is controlled by the planar diagrams, It is not hard to see that timelike noncommutativity then the question of whether a noncommutative theory is combined with UV/IR mixing leads to violations of per- renormalizable will have the same answer as for the cor- turbative unitarity. This shows up in unphysical branch responding conventional theory in the planar limit (nor- cut singularities in loop diagrams (Gomis and Mehen, mally this limit has the same UV behavior as the finite 2000) and can also be seen from the behavior of the prop- N theory). It has recently been argued by Griguolo and agator and commutators of fields (Alvarez-Gaume, Bar- Pietroni (2001b) that this can be done in the Wilsonian bon, and Zwicky, 2001); see also (Seiberg, Susskind, and RG approach to proving renormalizability (Polchinski, Toumbas, 2000a). We now give a simple argument along 1984). There are also arguments against this in certain these lines. cases (Akhmedov, DeBoer, and Semenoff, 2000), which A theory with timelike noncommutativity will be in- we return to below. variant under boosts along the conjugate spatial momen- We assume for sake of discussion that this is indeed tum (e.g. pi θ01) and therefore, if we set the other true and now discuss the physics of strong UV/IR mix- momenta to zero,∝ the propagator will be a function of ing. Before we start, we note that the supersymmetric 2 2 2 the 1 + 1 Lorentz invariant p = E + (pi) . Thus we theories of primary interest to string theorists have weak can write a spectral representation− or no UV/IR mixing, and the considerations we are about to discuss, except for those regarding unitarity, have not 2 played a role in this context so far. On the other hand 1 ∞ dm = ρ(m2) they are likely to be important in other applications. Γ(2)(p) p2 + m2 Z0 26 with ρ(m2) 0 in a unitary theory. Unitarity requires can be understood by considering wave propagation in ≥ a medium with an index of refraction n(ω) = k(ω)/ω 2 4 1 ∞ dm 2 which grows as ω 0 faster than 1/ω. In our exam- lim = ρ(m ) > 0 (110) → p2 0 Γ(2)(p) m2 ple, ω g/θk. This leads to a negative group velocity → Z0 ∼ vg = ∂ω(k)/∂k. One can still proceed formally to derive assuming this limit exists (it might also diverge). On the a position space Greens function, which exhibits super- other hand, the general behavior produced by UV/IR liminal propagation. In a non Lorentz invariant theory, mixing is this might not be considered a major surprise; however its short time propagation is dominated by long wave- lim Γ(2)(p)=+ , p2 0 ∞ lengths and it does not satisfy the usual defining property → limt 0 G(x, t) = δ(x). This certainly looks unphysical which is incompatible with Eq. (110). and→ at present further interpretation would seem to be We turn to consider purely spatial noncommutativity. a purely academic exercise. On the other hand, perhaps This will lead to a dispersion relation of the form the usual position space interpretation is inappropriate and there is something deep here yet to be discovered. 1 E2 = ~p2 + m2 + ∆M 2( ). pθ Finally, we return to our original assumption of the validity of the Wilsonian RG, and discuss the work of If ∆M 2(Λ) grows at infinity, we again find low momen- (Akhmedov et al., 2000; Akhmedov, DeBoer, and Se- tum modes are stiff. menoff, 2001; Girotti, Gomes, Rivelles, and da Silva, We first discuss the Euclidean theory. The Greens 2001) on the noncommutative Gross-Neveu model. This (2) (2) 1 function G (p)=Γ (p)− can be transformed to posi- is a model of N flavors of interacting massless fermions tion space following our discussion in Sec. II.E.4. As dis- whose perturbative vacuum is unstable. In the 1+1 con- cussed there, the long distance behavior is controlled by ventional theory a nonperturbative condensate forms, as the pole in the upper half plane nearest the real axis, as in can be shown exactly in the large N limit by solving a 2 1 2 2 gap equation, leading to a vacuum with massive fermionic conventional field theory. Let us consider ∆M = 4 g Λ for definiteness, then the poles in G(2)(p) can be found excitations. This theory exhibits dimensional transmu- tation; the bare coupling g can be eliminated in favor of by solving a quadratic equation, leading to the follow- 2 2 the mass gap M Λexp 1/g . ing two limiting behaviors. If g << θm (weak coupling ∼ − or strong noncommutativity), the closer root will be at Following the procedure we just outlined for the non- p = ig/4θm. This can be interpreted as a new mode with commutative theory leads to Γ(2)(p) < 0 at low p (for the mass m2 = g/4θm, and the precondition is equivalent to condensate) and instability. However, Akhmedov et al. m2 << m, i.e. this is the limit in which the new mode argue that instead of following the standard Wilsonian dominates the long distance behavior. In the other limit, approach and defining the continuum theory as a flow out there are two poles at equal distance (g/8θ)1/2, leading of a UV fixed point, one should enforce the gap equation. to oscillatory-exponential behavior. This leads to a stable vacuum but spoils dimensional Thus, in either case, the IR effects appear to have a transmutation and results in a trivial continuum limit. sensible description at finite g, in terms of a new mode. Akhmedov et al. also point out the interesting possibility One can go on to ask whether the new mode can be of a nontrivial “double scaling limit” with fixed ΛMθ. described by adding an additional field to the effective One cannot a priori say that one of these definitions Lagrangian. This question is best discussed in the con- is “right;” this depends on the underlying microscopic text of the connection to string theory and we postpone physics and the application. What we would insist on is it for Sec. VII.E. that one can choose to define the theory as a flow from We now discuss the quantum noncommutative field a UV fixed point and that in this sense the theory is theory in Minkowski space. The primary question is renormalizable; one however does not know (at present) whether it is unitary. Formally, the main thing to check whether it has a stable vacuum. 2 is that the (∂0φ) term in the effective action is positive Furthermore, we feel that even the alternate definition for all spatial momenta (we also assume that higher time can be fit within the conventional RG picture, allowing derivative terms are not significant). This is necessary for for IR physics to determine a condensate which will then perturbative unitarity of the effective action but almost react back on the UV physics, determining the couplings certainly will be true if the analog conventional theory through dynamics at various scales. In itself, this phe- was renormalizable and unitary. The cutting rules can nomenon is not new but has fairly direct analogs in con- also be checked and appear compatible with unitarity in ventional field theory, particularly supersymmetric field this case (Gomis and Mehen, 2000). Since these theories admit a (cutoff) Hamiltonian formulation, any problem with unitarity would almost certainly be tied to instabil- ity as well. 4 Strictly speaking, we are discussing spatial dispersion (Landau, However, the position space Greens function in this Lifshitz, and Pitaevskii, 1960), but the difference is not relevant case is quite bizarre at low momenta. Its main features for us. 27 theory where IR physics can lift a flat direction, leading has also been argued that spontaneously broken noncom- to large scalar vevs and large masses for other degrees of mutative gauge theories are renormalizable (Petriello, freedom. 2001).5 Unitarity has recently been discussed in (Bas- However, one has opened Pandora’s box as the larger setto, Griguolo, Nardelli, and Vian, 2001). point seems to be that strong UV/IR mixing can lead Thus we turn our attention to the structure of the non- to very different physics for condensates, which is not commutative effective action. It turns out to be some- understood. For example, if the noncommutative Gross- what subtle to write this in a gauge invariant form (Liu Neveu model defined as a UV fixed point has a stable and Michelson, 2000; Zanon, 2001b) (this was also seen vacuum, it might not be translationally invariant and in related string computations by Garousi (2000)). One perhaps not even be describable by a classical condensate. cannot use the usual rule for Yang-Mills theory, according The tentative conclusion we will draw from all of this to which one can infer higher-point functions by complet- is that quantum field theories with spatial noncommu- ing an operator ∂iAj ∂j Ai to the gauge-invariant Fij . tativity and strong UV/IR mixing can be consistent, of- The basic problem is− apparent from the analogy with ten with rather different physics from their conventional conventional large N field theory. The rules of IV.A tell analogs. us that nonplanar diagrams will produce contributions of the form D. Gauge theories ddk Tr O (k ) . . . Tr O (k )G(k , , k ). i 1 1 n n 1 · · · n i The usual discussion of perturbative gauge fixing and Z Y the Fadeev-Popov procedure go through essentially un- changed. The Feynman rules for noncommutative gauge Such terms do not follow the rule from Sec. II.C of as- theory are the standard ones for Yang-Mills theory, with sociating a single trace with each integral and thus we the Lie algebra structure constants augmented by the cannot apply the arguments made there. phase factors Eq. (100). Clearly this problem is related to the problems with Gauge theories can have IR divergences which are defining local gauge invariant operators discussed in stronger than the standard UV/IR relation would sug- Sec. II.D.2, and it is believed that the effective action will gest (Hayakawa, 1999; Matusis, Susskind, and Toumbas, be gauge invariant if expressed in terms of the the open 2000). Consider for example the one-loop contribution Wilson loop operator (Liu, 2000; Mehen and Wise, 2000). to the NCQED photon self-energy due to a massless This has been checked in examples (Pernici, Santambro- fermion. Standard considerations lead to the amplitude gio, and Zanon, 2001) and one can also make strong ar- guments for it from string theory (see Sec. VII.E). 2 2 4g 4 (2kikj ηij k ) ik p iΠ (p)= d k − 1 e × . The basic gauge invariant local Greens function then ij −(2π)4 k4 − Z is the two-point function of two open Wilson loops, W (k)W ( k) . A particularly interesting question is the The nonplanar part of this is highh energy− behaviori of this quantity, which was found 2 by Gross, Hashimoto, and Itzhaki (2000). Our general iΠ(np)(p) = 4ig (∂ ∂ η ∂2)log p p +1/Λ2(111) ij (2π)4 i j − ij ◦ arguments that high energy behavior reduced to planar 2 k l 16ig θikp θ jlp diagrams applied to naive local observables, but do not = (2π)4 (p p+1/Λ2)2 . (112) − ◦ apply to the open Wilson line, because its extent grows with energy. If we remove the cutoff, this diverges as 1/θ2p2 as p 0. Physically, this term would lead to different dispersion→ At high energy, the open Wilson loops will become very relations for the two photon polarizations. Peculiar as long and the computation becomes identical to that for this is, it is consistent with gauge invariance in this non- the expectation value of a single rectangular Wilson loop, Lorentz invariant theory. A similar one-loop contribution a well known result in the conventional theory. Since the can be found to the three-point function. It behaves as leading large momentum behavior is given by the sum 2 of planar ladder diagrams, the result can be applied di- (θp)i(θp)j (θp)k/(p p) at low momenta and diverges as 1/θp as p 0. It is◦ not clear at present what the signif- rectly. For a rectangular Wilson loop of sides L and r icance of these→ effects might be. (with L >> r), it is exp (E + M)L with the Coulomb interaction energy E = −g2N/4πr and M a nonuniver- In any case, it has been observed that these effects − vanish in supersymmetric theories. The logarithmic ef- sal mass renormalization. This result can then be Fourier fects expected from our previous discussion will still be present. A detailed discussion at one loop is given by Khoze and Travaglini (2001); see also (Zanon, 2001a; 5 Ruiz, 2001). There is some controversy about this and about the renormal- izability of the related nonlinear sigma model, possibly related All of the UV and IR divergences found in any com- to issues discussed at the end of Sec. IV.C; see (Campbell and putation done to date are gauge invariant, and it is gen- Kaminsky, 2001; Sarkar and Sathiapalan, 2001; Girotti et al., erally believed that these theories are renormalizable. It 2001). 28 transformed, leading to a growing exponential in momen- diagram obtained by contracting the external legs in fig- g2N k L ure 5. This contributes (for d = 4) tum space, exp+ 4π| | . The main adaptationq required to compute the open ∞ 3 3 2 2 d p d k ip k Wilson loop Greens function is just to use Eq. (16) and g T e × − (2π)3 (2π)3 take L = θk . This produces n,l= Z | | X−∞ g2N kθ k W (k)W ( k) exp+ | || | (113) 1 h − i∼ 4π 2 2 . r 4π 2 2 2 4π 2 2 2 ( β2 n + p + m )( β2 l + k + m ) This exponential growth with momentum is universal and applies to correlators of Wilson loops with operator If we neglect the mass (appropriate if m< However, the more general situation is that the cor- Taking into account these considerations, one can find responding classical integrals are still UV divergent. In the quantum Hamiltonian and its perturbative spectrum particular, this is true in the cases summarized in Eq. along standard lines. We quote as an example the ground (114). It must also be remembered that the planar con- state energy in 2 + 1 noncommutative Yang-Mills on T2 tributions are still present and dominate the contribu- Matp( θ) at leading order in perturbation theory (this tions we are discussing. Thus the full import of these can be shown to be the exact result given enough super- rather striking results is not clear. symmetry). Sectors of this theory are labeled by con- See also Landsteiner, Lopez, and Tytgat (2001), who served integral quantum numbers ni (related to electric argue for a finite temperature phase transition in non- flux as in Eq. (116)), q (magnetic flux), and mi (momen- commutative gauge theory. tum). Finally, we write N =Tr 1= p θq; as we explain in Sec. VI, the naive expression Tr 1 =− p is not valid on Tθ. F. Canonical formulation One then has (Connes et al., 1998; Brace and Morariu, 1999; Hofman and Verlinde, 1999; Konechny In general terms, canonical quantization of noncom- and Schwarz, 1999) mutative field theory and gauge theory with purely spa- 2 tial noncommutativity can be done by following standard gY M i ik j jl E = gij (n + θ mk)gij (n + θ ml) procedures, which we assume are familiar. This leads to 2√gN commutation relations between free fields, say φ and a 2 conjugate momentum π, which in momentum space take π 2 2π j + 2 q + mip qǫij n . (117) the same form as conventional theory. One furthermore 2gY M √gN N | − | has standard expressions for energy and momentum op- erators associated to commutative dimensions. Possible The first two terms are the energies associated to zero modes, while the last term is the energy E = p asso- quantum corrections to their Ward identities have not | | been much studied. ciated to a state with massless excitations carrying mo- mentum mi as well as a contribution from E B. One can define momentum operators Pi for the non- × commutative dimensions by using the restricted energy- momentum tensor T 0 = i[φ, π], as in Eq. (26). In non- commutative gauge theory, this makes sense for the rea- G. Other results sons discussed in Sec. II.D.3. A point where interesting differences from the conven- Unfortunately, space does not permit us to discuss all tional case appear is in discussing the action of gauge of the interesting results obtained in quantum noncom- transformations and charge quantization. Consider 2 + 1 mutative field theory, but let us mention a few more. 2 The Seiberg-Witten solution of = 2 supersymmetric Maxwell theory on a square torus T defined by the iden- N i i Yang-Mills (Seiberg and Witten, 1994), giving a prepo- tifications x ∼= x +2π. The total electric flux has two components d2x Ei, each conjugate to a Wilson loop tential which encodes the dependence of BPS masses and i low energy couplings on the choice of vacuum, is a bench- Wi = exp i dx Ai: R mark for nonperturbative studies. It turns out that the R i i solution is the same for noncommutative as for conven- [E , Wj ]= δ Wi. j tional theory; this can be seen from instanton computa- This theory admits large gauge transformations, acting tions (Hollowood, Khoze, and Travaglini, 2001) and M i inix theory considerations (Armoni, Minasian, and Theisen, on a charge 1 matter field ψ as ψ e ψ, with ni ZZ to make the transformation single→ valued. Thus the zero∈ 2001). See also (Bellisai, Isidro, and Matone, 2000). i One may wonder what happened to the UV/IR mixing. mode of Ai is a periodic variable, and eigenvalues of E must be integrally quantized. The noncommutative theory necessarily includes a “U(1) A similar argument can be made in the rank 1 noncom- sector,” and one can show that only this sector is affected. mutative theory on Td, but now the corresponding gauge It is not visible in the prepotential, which is independent θ of the diagonal component of the scalar vev. This sector transformations ψ U ψU † (notation as in II.B.2) act i i (and the rank 1 theory) is nontrivial, and does not reduce on the nonzero modes→ of ψ as well as in Eq. (15), taking to Maxwell theory as θ 0 (Armoni, 2001). It is not ψ(xj ) ψ(xj θji). Thus the wave function must be in- understood at present. → variant→ under a− simultaneous shift of the zero mode of A i Noncommutative sigma models are discussed by Lee, and an overall translation (Hofman and Verlinde, 1999), Lee, and Yang (2001). The noncommutative WZW leading instead to a quantization law for the quantity model is discussed in (Moreno and Schaposnik, 2001, 2000; Lugo, 2000). Tr Ei θij P . (116) − j Anomalies have been studied in (Ardalan and Z Sadooghi, 2000; Gracia-Bondia and Martin, 2000; Mar- One also finds modified quantization laws for magnetic tin, 2001; Bonora, Schnabl, and Tomasiello, 2000b). charges, which we will discuss in Sec. VI.D. Gauge anomalies appear to be directly analogous to those 30 of conventional theory, consistent with their topologi- put me = 0. The resulting Lagrangian is first order in cal origin, and can be described using fairly straight- time derivatives, turning the original coordinate space forward noncommutative generalizations of the Wess- into an effective phase space defined by Eq. (120). A Zumino consistency conditions, descent equations, etc. more precise argument can be made by taking coordi- Somewhat surprisingly, however, these formulas appear nates ζ, z¯ as in Sec. II.B and showing that to lead to more restrictive conditions for anomaly can- cellation, because more invariants appear. For example, P zP¯ = θ∂/∂z + z (121) while the conventional d = 4 triangle anomaly only in- volves the invariant dabc = tr ta tb,tc , which can can- acting on LLL states { } cel between different representations, its noncommuta- ∗ a b c 1 k zz /2θ tive generalization involves tr t t t , which cannot. This k = z e− . (122) | i √ apparent contradiction to the general similarity we have k! seen between noncommutative and conventional quan- The resulting single particle Hamiltonian has been much tum effects as well as to the topological nature of the studied and we refer to (Bellissard et al., 1993) for a anomaly deserves further study. discussion of the uses of noncommutative geometry in this context. To obtain a field theory, one introduces electron cre- V. APPLICATIONS TO THE QUANTUM HALL EFFECT ation and annihilation operators ψ†(x) and ψ(x), in terms of which the electron density is A physical context leading directly to noncommutativ- ity in the Fock space basis is the dynamics of electrons in 1 2 ρ(~x) = µ δ (~xµ ~x) a constant magnetic field B~ , projected to the lowest Lan- √Ne − ψP†(x)ψ(x). (123) dau level (LLL). This is well known in the theory of the ≡ quantum Hall effect (QHE) (Girvin and Prange, 1987; Girvin, 1999), and we summarize the basic idea here. The single particle and pairwise interaction Hamiltonians More recently, it has been proposed that a good de- then become scription of the fractional quantum Hall effect (FQHE) H = dx V (x) ρ(x); (124) can be obtained using noncommutative rank 1 Chern- V H = dx dx′ U(x x′) ρ(x)ρ(x′). (125) Simons theory (Susskind, 2001; Polychronakos, 2001), U R − and we give a brief introduction to these ideas as well. The effects of truncationR to the LLL are now expressed It seems fair to say that so far, the noncommutative by noncommutativity, and more specifically enter when framework has mainly provided a new language for pre- we use the star product to compute the commutators of viously known results. However, it also seems fair to density operators. In momentum space, say that this formulation connects these problems to a very large body of field theory results whose potential iqx relevance had not been realized, and one can hope that ρ(~q)= e ρ(x), these connections will prove fruitful. Z we have A. The lowest Landau level [ρ(~q),ρ(~q′)] = sin i~q ~q′ρ(~q + ~q′) ×θ The Lagrangian of a system of interacting electrons which leads to deformed equations of motion, etc. in two dimensions, subject to a perpendicular magnetic This type of description has been used by many au- field, is thors; we cite (Sinova, Meden, and Girvin, 2000) on lo- calization in quantum Hall states, and (Gurarie and Zee, Ne 1 2 ieB 2000) on its classical limit, as the tip of a large iceberg. L = m ~x˙ ε xi x˙ j + V (~x ) (118) 2 e µ − 2c ij µ µ µ µ=1 X + U(~x ~x ) (119) B. The fractional quantum Hall effect µ − ν µ<ν X The generally accepted explanation of the FQHE is Defining a projection operator P to the first Landau level that interactions lead to a state similar to the filled LLL for each electron, one finds that the projected coordinates but allowing fractionally charged quasi-particle excita- i P xµP do not commute, but instead satisfy tions. A good microscopic description of such a state ~c is provided by the N-electron wavefunction (Laughlin, [xi , xj ]= iδ εij iδ θij . (120) 1983) µ ν µν eB ≡ µν ∗ m zµz /2θ Heuristically, this is because in the limit of strong mag- Ψ= (z z ) e− µ µ . (126) µ − ν netic field one can neglect the kinetic term, i.e. formally µ<ν P Y 31 where m is an odd integer. This state has charge den- where Xi are Hermitian matrices with i = 1, 2 and ψ sity 1/m that of a filled Landau level and was argued by is a complex vector. The gauge field in this action is Laughlin to be a ground state of Eq. (118), at least for nondynamical but enforces a constraint which selects the small m. sector with ψ charge m, for which the Hamiltonian is Eq. States with quasiparticles are obtained by simple mod- (128). ifications of this. An operator creating a quasi-hole at z0 Reversing the procedure which led to Eq. (38), we can acts as multiplication by (z z ), while the conjugate regard Xi as covariant derivative operators, to obtain an i i − 0 operator (which can create quasi-particles) is (¯zi z∗) noncommutative gauge theory action, whose kinetic term Q i − 0 withz ¯i as in Eq. (121). Acting m times with one of these (coming from the first term of Eq. (129)) is precisely operators adds or subtracts a particle, so theQ quasiparti- Eq. (127). The other terms are secondary: the X2 term cles have charge 1/m. localizes the state in space, while ψ, although required A more subtle property of the quasiparticles is that for consistency at finite N, plays no dynamical role. they satisfy fractional statistics; a 2π rotation of a state In this sense, one has a precise noncommutative field of two quasiparticles produces a phase exp 2πi/m, as can theory description of the fractional Hall state. In partic- be seen by a Berry phase argument (Arovas et al., 1984). ular, the quasiparticles are well-defined excitations of the The low energy excitations of this ground state can also noncommutative gauge field; for example the quasihole be described by a Landau-Ginzburg theory of a superfluid is rather similar to the fluxon Eq. (83). We refer to the density φ and a fictitious vector potential A, such as Eq. cited references for more details. (82) (Kallin and Halperin, 1984; Laughlin, 1985; Girvin, MacDonald, and Platzman, 1985; Fradkin, 1991). In this picture, the original quasi-particles are vortex solutions, VI. MATHEMATICAL ASPECTS and their fractional statistics is reproduced by an abelian Chern-Simons term in the action, As we mentioned in the introduction, noncommutative gauge theory was first clearly formulated by mathemati- im ijk cians to address questions in noncommutative geometry. SCS = ǫ Ai∂j Ak. (127) 2π Limitations on length would not permit more than the Z most cursory introduction to this subject here, and since With this term, a vortex with unit magnetic charge will so many introductions are already available, starting with also carry electric charge 1/m, so the Aharonov-Bohm the excellent (Connes, 1994), much of which is quite read- effect will lead to fractional statistics (Wilczek, 1982). able by physicists, and including (Connes, 1995, 2000b,a; Recently Susskind (2001) has proposed that noncom- Gracia-Bondia et al., 2001; Douglas, 1999) as well as the mutative Chern-Simons theory is a better description of reviews cited in the introduction, we will content our- fractional quantum Hall states, which can reproduce the selves with a definition: detailed properties of these quasiparticles. Perhaps the Noncommutative geometry is a branch of mathemat- simplest argument one could give for this is simply to ics which attempts to generalize the notions of geometry, combine the arguments leading to LG theory with the broadly defined, from spaces M whose function algebras arguments leading to noncommutativity in a magnetic C(M) are commutative, to “spaces” associated to general field. algebras. The word space is in quotes here to emphasize One way to make this claim precise has been proposed that there is no a priori assumption that these spaces by Polychronakos (2001). One first observes that Eq. are similar to manifolds; all of their attributes emerge (126) (considered as a function of one dimensional po- through the course of formulating and studying these ge- sitions xµ = Re zµ) is the ground state for a Calogero ometric notions. model, defined by the quantum mechanical Hamiltonian The remainder of this section provides an introduction to other examples of noncommutative spaces, and other 1 1 1 m(m + 1) 2 2 topics for which a more mathematical point of view is ad- H = pµ + 2 xµ + 2 . (128) 2 2θ 2 (xµ xλ) µ µ<λ − vantageous. We will only be able to discuss a few aspects X X of noncommutative geometry with direct relevance for One can continue and identify all the excited Calogero the physics we described; there are many others for which states with excited Laughlin wavefunctions (Hellerman such a role may await, or for which we simply lack the and Van Raamsdonk, 2001). expertise to do them justice. Among them are cyclic co- It is furthermore known (Olshanetsky and Perelomov, homology and the related index theorems (Connes, 1982), 1976) that this model can be obtained from a matrix- and the concept of spectral triple (Connes, 1996). vector U(N) gauged quantum mechanics, with action (Polychronakos, 1991) A. Operator algebraic aspects i j 1 2 S = dt Tr ǫ X D X X +2mA + ψ†D ψ ij 0 − 2θ2 i 0 0 The origins of noncommutative geometry are in the Z (129) theory of operator algebras, which grew out of functional 32 analysis. Certain issues in noncommutative field theory, which A p is finite. This is more or less the direct analog especially the analogs of topological questions in conven- of the| conventional| condition that a function (or some tional field theory, cannot be understood without these power of it) be integrable. A related condition which ideas. The following is loosely inspired by a discussion of also expresses “falloff at infinity” is for A to be compact, the meaning of the instanton charge in (Schwarz, 2001a); meaning that the sequence of eigenvalues of A†A has the we also discuss a proposed definition of the noncommu- limit zero. tative gauge group (Harvey, 2001b). Although these are important conditions, they do not A good example of a topological quantity is the to- solve the problem at hand, because they are not pre- tal magnetic flux in two spatial dimensions, Tr F . In served by the derivatives Eq. (3), so do not give strong commutative theory, this is the integral of a total deriva- constraints on Eq. (130). In particular, one can have tive F = dA. In a pure gauge background, thisR integral Tr [z,K] = 0 (and even ill-defined) for K in any of the will be quantized; furthermore, it cannot change under classes above6 (e.g. try Tr [O, z¯]). variations A + δA by functions δA which are continuous, Another approach is to adapt a falloff condition on single-valued and fall off faster than 1/r at infinity. Thus n n functions on IR to IRθ by placing the same falloff con- one can argue with hardly any dynamical input that the dition on the symbol Eq. (18). An important point total flux in a sufficiently large region must be conserved. to realize about the noncommutative case is that one Let us return to the question raised in Sec. III.B, of cannot separately define “growth at large radius” from whether this argument can be generalized to noncommu- “growth with large momentum.” The only obvious crite- tative theory. The magnetic flux has the same formal rion one can use is the asymptotics of matrix elements at expression. In operator language, it is the trace of a large mode number, which does not distinguish the two. commutator, Physically, as we saw in our considerations of Gaussians in II.2.C, interactions easily convert one into the other. ¯ Tr F = Tr [C, C]+1. (130) Thus questions about whether configurations disperse or go off to large distance, are hard to distinguish from ques- We need to understand in what sense this is a “boundary tions about the possibility of forming singularities. term,” which is preserved under continuous variation of Thus, useful falloff conditions must apply to both po- the fields with a suitable falloff condition. sition and momenta. Such conditions are standard in the If one does not try to define these conditions, one can theory of pseudodifferential operators, and can be used easily exhibit counterexamples, such as the path to define various algebras which are closed under the star ¯ product (Schwarz, 2001a). Perhaps the simplest is the C = λz + (1 λ)S†zS (131) d − class of operators S(IRθ), which can be obtained as the which purports to continuously interpolate between the transform Eq. (18) of smooth functions with rapid de- fluxon and the vacuum. Is this a continuous variation? crease, i.e. which fall off both in position and momentum We have to decide; there is no single best definition of space faster than any power. This condition is very much continuous in this context. stronger than boundedness (roughly, it requires matrix The minimal criterion we could use for a continu- elements in the Fock basis to fall off exponentially) and ous variation of the fields is to only allow variations by is preserved under differentiation, so Tr F will certainly bounded operators. A bounded operator has finite oper- be preserved by this type of perturbation. 1/2 ator norm A (this is the largest eigenvalue of (A†A) ). As we discussed in III.B, the question of physical flux | | A small variation C′ of the connection C then has small conservation is whether time evolution allows flux to get C′ C , so will be a bounded operator times a small co- to infinity at finite time. This is a dynamical question, efficient.| − | Boundedness is mathematically a very natural but might be addressed by finding conditions on the fields condition as it is the weakest condition we can impose on which are preserved under time evolution. For example, a class of operators which guarantees that the product of in the conventional case one can take fields which are pure any two will exist, and thus the definition of an operator gauge outside a radius r = R, and then causality guaran- algebra normally includes this condition. tees that they will stay pure gauge outside of r = R + t. d However, boundedness is not a strong enough condi- Since S(IRθ) is preserved by both the product and the tion to force traces of commutators to vanish: one can Laplacian, it is a good candidate for an analogous class have Tr [A, B] = 0 even if both A and B are bounded of noncommutative fields which is preserved under time 6 operators (e.g. consider Tr [S†,S] where S is the shift evolution. The fluxon and its perturbations by finitely operator). The operator O = z S†zS appearing in Eq. many modes fall into this class, so an argument along (131) is bounded. − these lines should show that flux is conserved, and thus A condition which does guarantee Tr [A, B]=0 is for the decay of the fluxon does not lead in finite time to a A to be bounded and B to be trace class, roughly mean- stable ground state. Of course conservation of Tr F (and ing that its eigenvalues form an absolutely convergent energy Tr F 2) is surely true for a larger class of initial series. More precisely, A is trace class if A 1 is finite, conditions. p/2 1/p | | where A p = (Tr (A†A) ) is the Schatten p-norm. A related question, discussed by Harvey (2001b), is the More generally,| | the p-summable operators are those for precise definition of the gauge group U( ). In conven- H 33 tional gauge theory, the topology of the group 0 of gauge noncommutative quantum theory; perhaps other classes transformations which approach the identityG at infinity, of fields such as discussed here will turn out to be equally is directly related to that of the configuration space of or more useful. gauge fields (connections modulo gauge transformations) by a standard argument: since the space of connections B. Other noncommutative spaces is contractible, we have πn( / 0) ∼= πn 1( 0). One mightA expect the topology of UA( G) to play an− analogousG role and in fairly direct correspondenceH to the consid- Our discussion so far was limited to noncommutative d Td erations above, this would imply that we cannot iden- field theory on IRθ, mostly because it and θ are the tify U( ) with the group of unitary bounded operators only examples for which field theoretic physics has been on HilbertH space; this group is contractable (Kuiper, explored sufficiently at present to make it worth writing 1965). One can get the expected nontrivial topology, a review. There are many other examples which allow πk(G) = limN πk(U(N)) = ZZ for k = 2n +1, by us- defining noncommutative field theories, whose physics ing the unitaries→∞ 1 + K with K compact or p-summable has been less well explored. (Palais, 1965). Harvey (2001b) suggests using K com- For example, let us try to define a noncommutative d pact, as the largest of these groups. Alternatively, one sphere Sθ . This should be a space associated to an alge- might try to use a smaller group defined by imposing bra of d + 1 operators xi, 1 i d + 1, satisfying the ≤ ≤ conditions involving the derivatives. relation A similar open question about which less is known at d+1 present is that of what class of fields to integrate over (xi)2 = R2 1. (132) in the functional integral (the functional measure). As i=1 we commented in Sec. II.D.1, formally all pure bosonic X noncommutative gauge theories have the same action; We next need to postulate some analog of Eq. (1). however we expect that different quantum theories exist, Although one can define algebras without imposing com- mutation relations for each pair of variables, these will be distinguished by the choice of measure. In particular, d this subsumes the choice of the dimension of space-time. much “larger” than the algebra of functions on S . For d = 2, there is a natural choice to make, Again, the minimal proposal is that one uses an action such as Eq. (38), expands around a pure gauge config- [xi, xj ]= iθǫijkxk, i i uration Ci =z ¯ and C¯i = z for 1 i r, and inte- grates over all variations of this by bounded≤ ≤ operators. which preserves SO(3) symmetry. This type of prescription does make sense in the context In fact there is a very simple way to define such an al- of the perturbation theory of Sec. IV with IR and UV gebra, called the “fuzzy two-sphere,” discussed in depth cutoffs, and would specify the dimension of space-time by Madore (1992), in (Madore, 1999) and many other there. This dimension can also be inferred from the na- works. It is to consider the 2j +1-dimensional irreducible ture of one-loop divergences; see (Connes, 1994; Varilly representation of the SU(2) Lie algebra, defined by her- and Gracia-Bondia, 1999). mitian generators ti satisfying the commutation relations For the reasons we discussed above, this prescription [ti,tj ]= iǫijktk. One can then set probably does not make sense beyond perturbation the- R ory. One might address this by proposing a smaller class xi = ti of variations to integrate over. At our present level of j(j + 1) understanding, however, it may be better to work with from which Eq. (132) followsp easily, and one finds θ = an explicit cutoff, such as that provided by the large N limit of matrix approximations as discussed in Sec. II.E.6. R/ j(j + 1). This algebra can serve as another starting point for We described there the proposal of (Ambjorn et al., 1999) p along these lines; while concrete, this appears to share the defining noncommutative field theory. One cannot use problems of conventional lattice definitions of breaking Eq. (3) to define the derivatives, however, as this is the symmetries of the continuum theory, and of not ad- inconsistent with Eq. (132). A simple choice which works is to define a linearly dependent set of derivatives mitting supersymmetric generalizations. Thus the prob- i lem of finding the best regulated form of noncommutative ∂if = [X ,f] and use gij = δij in this basis as a met- field theory remains open. ric. Since these derivatives do not commute, the natural definition of curvature becomes Given such a regulated theory, one then wants to study the continuum limit. Of course the analogous ques- Fij = i[∂i + Ai, ∂j + Aj ] i[∂i, ∂j ], tions in conventional quantum field theory are not trivial, − and their proper understanding requires the ideas of the (as in Eq. (37)), in terms of which one can again use the renormalization group. We already made such a discus- Yang-Mills action. This and related theories have been sion in IV, following the usual paradigm in which the discussed by Grosse, Klimcik, and Presnajder (1996). d Td momentum space behavior of Greens functions is cen- Unlike IRθ and θ, this algebra is finite dimensional. tral. It is not yet clear that this is the best paradigm for If we base our theory on it directly, it will have a finite 34 2 number of degrees of freedom, and one might question satisfying g = 1. The general element of G is a + bg, the use of the term “field theory” to describe it. and (a + bg)(c + dg) = (ac + bd) + (bc + adA)g. One can certainly take the limit j , but if we take It is well known that for a finite group G, G is a direct R j as well to keep θ finite, we→ lose ∞ the curvature sum of matrix algebras, one for each irreducibleA represen- ∝ → ∞ 2 of the sphere, and end up with IRθ. tation of G. One goal of representation theory is to try Finally, if we keep R fixed and take j , it is a theo- to make analogous statements for infinite groups. This rem that (with suitable definitions) this→ algebra ∞ goes over requires being more precise about the particular linear in the limit to the algebra of continuous functions on or- combinations allowed, and leads one deep into the the- dinary S2, so we obtain a conventional field theory. This ory of operator algebras. We will not go into details, but feature of restoring commutativity in the limit applies to these algebras are clearly a good source of noncommu- a wide class of constructions, as we will discuss further tative field theories, as the original definitions of Connes in Sec. VI.G. It should be said that this theorem is clas- and Rieffel (1987) can be applied directly to this case. sical, and there might be a way to quantize the theory One of the main problems in trying to define noncommu- which does not commute with the large j limit, leading tative field theories on more general spaces is to either to a nontrivial noncommutative quantum field theory. define a concept of metric, or get away without one. Since One can still argue that for large but finite j, a theory group spaces are homogeneous spaces, this problem be- based on this algebra deserves the name “noncommuta- comes very much simpler. tive field theory.” We would suggest that the nomen- A variation on this construction is the twisted group 2 clature be based, first, on the extent to which a theory algebra G,ǫ, which can be defined if H (G, U(1)) = 0. A 6 displays physical characteristics similar to those we have This allows for nontrivial projective representations γ d seen for theories on IRθ, and second on the extent to characterized by a two-cocycle ǫ, which it shows universality (e.g. has finitely many pa- rameters) analogous to field theory; this is not usually γ(g1)γ(g2)= ǫ(g1,g2)γ(g1g2), (133) the case for constructions with a finite number of de- and is just the group algebra with this multiplica- grees of freedom. These questions have not been settled AG,ǫ and we will not take a position on this here. tion law. Since the phases Eq. (21) are a two-cocycle for ZZd Td We move on and discuss other noncommutative alge- , θ itself is an example. bras comparable to function algebras which can clearly A very important source of noncommutative algebras serve to define field theories. The simplest possibility was is the “crossed product” construction. One starts with an algebra , with a group G acting on it, say on the of course Matn(C(M)). Interesting variations on this can A be obtained by imposing further conditions which respect left: the product. For example, one could consider an algebra a U(g)a. and a ZZ2 automorphism preserving the metric, call it g. → A simple example would be to take IRd and let g be the re- θ One then chooses a representation R of G acting on a flection about some hyperplane. One could then impose Hilbert space H, g γ(g) (H), considers the tensor a = g(a) where g acts on each matrix element. This idea † product (H),7→ and imposes∈B the condition has been used to propose noncommutative gauge theories A⊗B with analogy to the other classical (SO and Sp) Lie alge- 1 γ− (g) a γ(g)= U(g)a. (134) bras (Bonora, Schnabl, Sheikh-Jabbari, and Tomasiello, 2000a; Bars, Sheikh-Jabbari, and Vasiliev, 2001b). The simplest example of this construction would be One can go on in this vein, using non-simple finite alge- to take finite dimensional. If G is abelian, we can bras and more complicated automorphisms. Indeed, one even takeA = C, and the general solution for a will be can obtain the complete action for the standard model some particularA solution added to an arbitrary element by choosing the appropriate algebra (Connes and Lott, of G. As we discuss below, the particular solution has 1991). Obviously the significance of this observation is theA interpretation of a connection, and this construction for the future to judge, though any example of a formal- leads directly to gauge theory on G. ism which only describes a subset of all possible gauge Suppose we had started with A= C(M). If we take theories but can lead to the standard model probably γ(g) to be the trivial representation,A Eq. (134) defines has something important to teach us. the algebra of functions on the quotient space C(M/G). More general choices of γ(g) thus lead to a generalized concept of quotient. The striking feature of this is that C. Group algebras and noncommutative quotients it provides a way to define quotients by “bad” group ac- tions, those for which the quotient space M/G is patho- A large class of “more noncommutative” algebras are logical, as is discussed in Connes (1994). This definition provided by the group algebras. Given a group G, we of quotient also follows from the standard string theory define G to be the algebra of all linear combinations definition of orbifolds, as discussed in (Douglas, 1999; of elementsA of G, with multiplication law inherited from Konechny and Schwarz, 2000a; Martinec and Moore, G. For example, consider G = ZZ2 with elements 1 and g 2001). 35 A natural generalization of this construction is the mits a right action of MatN ( ) which commutes with twisted cross product, whose definition is precisely the the action Eq. (135). ThereA is a general term for the same except that we take γ to be a projective represen- linear maps acting on a module E which commute with tation as in Eq. (133). This leads to gauge theory on Eq. (135); they are the endomorphisms of the module G,ǫ; we will discuss the toroidal case below. E, and the space of these is denoted End E. In fact A N A Another construction with a related geometric picture, End ( ) ∼= MatN ( ) so we know all endomorphisms the foliation algebra, is discussed in (Connes, 1994). of theA A free module.A For more general E, End E will always be an algebra, but need not be a matrixA algebra. We can now obtain more examples by starting with D. Gauge theory and topology free modules and applying a projection. We use the fact that the left module N admits a right action of A A good understanding of the topology of conventional MatN ( ). Given a projection P MatN ( ), the space gauge field configurations requires introducing the no- of solutionsA of v P = v is a module∈ with multiplicationA tions of principal bundle and vector bundle. We recall law a v = a v·. These examples are known as finitely that conventional gauge fields are connections in some generated× projective· modules and in fact it is only these principal G-bundle, while matter fields are sections of modules which are natural generalizations of vector bun- vector bundles with structure group G. On a compact dles, so one normally restricts attention to them. The space such as a torus, the topological classication of these endomorphisms of these modules also admit a simple de- bundles has direct physical implications. scription: they are the elements in MatN ( ) of the form The noncommutative analog of these ideas is a central P aP . A part of noncommutative geometry. We refer to Konechny We are now prepared to make a more general defini- and Schwarz (2000b) for a detailed discussion focusing on tion of connection (Connes, 1980). So far we have been the example of the noncommutative torus, but we give taking a connection to be a set of operators Di = ∂i + Ai the basic definitions here. See also (Harvey, 2001a) for a where the components Ai are taken to be elements of related discussion. MatN ( ), in direct analogy to conventional Yang-Mills One aspect of a conventional vector bundle is that one theory.A We used this in several ways; as Eq. (30) act- can multiply a section by a function to get another sec- ing on fields in MatN ( ) (the “adjoint action”), in the tion. This will be taken as the defining feature of a non- curvatures Eq. (37), andA finally acting on “fundamen- commutative vector bundle associated to the algebra ; tal” matter in Sec. II.D.3. The last of these is the point A one considers the (typically infinite dimensional) linear where vector bundles enter the conventional discussion, space of sections and requires it to be a module over the and where our generalization will apply most directly. algebra . More generally, a connection on the module E could A A module E over is a linear space admitting a mul- be any set of linear operators D which act on E and A i tiplicative action of which is bilinear and satisfies the satisfy the Leibniz rule, rule A D (a v)= ∂ (a) v + a D (v). (136) i · i · · i a (b v) = (ab) v; (a,b ; v E), (135) · · · ∈ A ∈ To see the relation to our previous definition, we first note that the difference between two connections A and carrying whatever other structure has (e.g. conti- i ≡ A D D′ will commute with the action of , and is thus nuity, smoothness, etc.) The simplest example is just a i − i A N-component vector with elements in , which is called a an endomorphism of E. Thus we can also describe the A general connection of this class by choosing a fiducial free module of rank N. Sections of a trivial rank N con- (0) N ventional vector bundle form the free module C(M) , connection Di and writing and this is the obvious generalization. (0) Di = Di + Ai; Ai End E. We chose the multiplication in Eq. (135) to act on the ∈ A left, defining a left module, but one can equally well let Let us consider E = N so we can compare with it act on the right, defining a right module, or postulate A (0) Sec. II.D.3. One can clearly take D = ∂ for the free independent multiplication laws on both sides, defining i i module, and as we discussed, the endomorphisms are just a bimodule. One can also speak of a ( , )-bimodule, Mat ( ), so we see that in this case the new and old def- which admits an action of on the leftA B and another N initionsA of connection agree. algebra on the right. A All of the modules constructed from projections as A simpleB example of a bimodule would be the space of above also admit a natural candidate for D(0), namely M N matrices of elements of , with 1M acting on i the× left and 1 on the right.A TheseA⊗ sit inside left and we apply ∂i and project back: A⊗ N right actions of MatM ( ) and MatN ( ), respectively. D(0)(v) = (∂ v) P. An obvious but importantA point is that theseA two actions i i · commute, i.e. (a v) b = a (v b). Thus connections on E can be identified with elements of We can regard· the· free module· · N as a bimodule (the End E, which as we discussed will be some subalgebra 1 N matrices) and this commentA shows that it ad- of matricesA Mat ( ), but not itself a matrix algebra. × N A 36 Finally, the gauge theory action uses one more ingre- Naively P = δ(σ1); this is too naive as this operator is dient, the trace. This can also be defined in terms of the not bounded. A correct projector can be found using the projection P ; an endomorphism which can be written ansatz P aP as above has trace Tr Mat ( )P aP . In particular, | N A we can define the dimension of the module as P = U2†g(U1)† + f(U1)+ g(U1)U2 2 Tr E1 = Tr MatN ( )P. for functions f and g chosen to satisfy P = P (Connes, A 1980; Rieffel, 1981; Bars et al., 2001a), and this can be These definitions can be used to obtain all the “non- used to compute dimE = θ/2π. commutative vector bundles,” and largely reduces the Rather than continue by following the general theory, classification problem to classifying the projections in in this case it is easier to write explicit results. For ex- MatN ( ). The next step in such a classification is to find ample, the endomorphisms of E are invariantsA which tell us when two projections are related 2πiσ/θ by a continuous deformation. A natural guess is that one (Z1f)(σ)= f(σ +2π); (Z2f)(σ)= e f(σ). wants to generalize the Chern classes of the conventional (139) theory, and indeed we have been implicitly doing this These operators also satisfy the defining relations of a 2 in claiming that quantities like Eq. (130) are topological noncommutative torus, but one with θ′ =4π /θ. quantum numbers. This turns out to be true for the non- The fact that two dual tori are involved may seem commutative torus and indeed there is a well-developed counterintuitive. However, there is a sense in which the formalism generalizing this to arbitrary algebras, based T2 T2 tori θ and θ′ are equivalent, called Morita equivalence, on cyclic cohomology (Connes, 1982). T2 based on the observation that E is also a module for θ′ (since we have an action of the Zi). We return to this below. E. The noncommutative torus (0) A reference gauge connection Di on E satisfying D(0)(U )=2πiδ U can now be defined as The noncommutative torus and its associated modules i j i,j j can be obtained using almost any of the constructions (0) 2πiσ (0) ∂f we cited, so besides its physical relevance it serves as a D1 f = f; D2 f =2π (140) Td − θ ∂σ good illustration. We follow the definition of θ made in (Connes, 1980), as the algebra of linear combinations (0) The general connection is a sum D + Ai with Ai of products U ni of generators with the relations Eq. i ∈ i i EndT2 E. In other words, the vector potential naturally θ (23), with coefficients decreasing faster than any power 2 lives on a “dual” noncommutative torus T ′ . Note that of n . Our discussionQ will mostly stick to T2. θ θ 1 Tr [D ,D ] = 1 is integrally quantized, even though One| | can regard commutative C(T2) as the group al- 2πi 1 2 dimE was not. gebra G 2 , i.e. all linear combinations of products of ZZ This construction can be generalized to produce all two commuting generators U = γ(g ) and U = γ(g ). 1 1 2 2 modules over T2. These modules E are characterized We can think of γ as the regular representation; it decom- θ p,q by two integers p and q, and can be produced by tensoring poses into the direct sum of all irreducible representations a representation with θ = p/q constructed in Sec. II.E.6, of ZZ2, which can be written U = exp iσ , parameterized 1 i i with Eq. (138) with θ chosen to satisfy θ = θ +θ . The by σ coordinates on T2. Each such representation is a 2 1 2 i computations above then generalize to one dimensional module over GZZ2 . Taking instead a projective representation Eq. (133) dimEp,q = p qθ/2π ; (141) with ǫ(g ,g ) = e iθ/2 leads directly to Eq. (23). The | − | 1 2 − Tr F =2πiq. (142) regular representation can be written explicitly as Ep,q iσ1 iσ2 The example E of Eq. (138) is E , while regarded as a U1 = e Sσ , θ; U2 = e (137) 0,1 2 − T2 module on θ′ it turns out to be E1,0. where Sσ,af(σ) f(σ + a)Sσ,a. This can again be de- The same construction generalizes to produce all mod- ≡ Td composed into irreducibles, each labelled by a fixed value ules on θ (at least for irrational θ) (Connes, 1980; Rief- of σ1. All of these are equivalent however (if we take this fel, 1988) and thus we can get the topological classifica- 2 iασ2 Td to act on functions on IR ), under conjugation by e . tion of gauge field configurations on θ . This turns out This gives us an example of an projective module E to be the same, in a sense, as for the commutative torus. T2 over θ: to repeat, we take for E the smooth functions In both cases, the topological class of a connection is de- on a real line (IR), and define the action of the operators termined by its Chern classes (Connes, 1980), which can S Ui on them as be defined as Tr 1, Tr F Tr F F , etc. and i1i2 i1i2 ∧ i3i4 iσ which obey additive quantization rules corresponding to (U1f)(σ)= f(σ θ); (U2f)(σ)= e f(σ). (138) dR−1 R R − the group ZZ2 . This is reasonable as a continuous vari- In terms of the general construction of the previ- ation of θ should not change a topological property. ous subsection, we obtained E from 1 by a projec- Many properties of these modules are easier to see from A tion, which should allow us to compute dimE = Tr E1. other constructions. In particular, one can also regard 37 2 2 2 T as the quotient IR /ZZ , and then define this quotient An example is that ˆ = MatN ( ) is Morita equivalent using the crossed product. Let ni ZZ2 act on IR2 as to , by taking P toA be N . ThusA we do not want to j j i j i ∈2 A A x x + n (ei) , and take X IR B( ), then Eq. think of Morita equivalent algebras as literally equivalent; (134)→ becomes ∈ ⊗ H however their K theory and many other properties are the same. A more striking example of a Morita equivalence is that 1 j j j U − X Ui = X + (ei) . (143) T2 T2 i θ is Morita equivalent to θ′ as above. The module E we constructed in Eqs. (138), (139) is the bimodule which If we take the regular representation as above, and T2 j j provides this equivalence. Thus every module Ep,q on θ (ei) = δ for simplicity, these equations are solved by 2 i is associated by Eq. (147) to a module E q,p on T ′ . − θ There is a second, simpler equivalence of this type ob- j ∂ X = i + Aj (144) tained by taking θ θ +2π, which manifestly produces − ∂σj the same algebra. Considering→ Eq. (141) shows that this where A commute with the U ; i.e. are general functions acts on the modules as E E . j i p,q → p+1,q on T2. It is no coincidence that these solutions look like These two transformations generate the group covariant derivatives; we can rewrite Eq. (143) as a b SL(2, ZZ) of 2 2 matrices with integer en- × d d j j j X Ui = UiX + (ei) Ui (145) tries and determinant 1, acting on τ θ/2π as τ (aτ + b)/(cτ + d) and on (p, q) as a≡ vector. This→ is which is precisely Eq. (136) with Xj = D . j an example of a “duality group” which is a candidate To get T2, we instead take the twisted crossed product, θ equivalence between gauge theories based on the pair of which amounts to solving Eq. (143) with U satisfying i Morita equivalent modules. This equivalence can be fur- Eq. (23). Using the representation Eq. (137) for this, ther strengthened by introducing a stronger notion of the solutions become “gauge Morita equivalence” (Schwarz, 1998), which pro- Xi = D(0) + A (146) duces a map between the spaces of connections on the i i two modules which preserves the Yang-Mills action. (0) Although somewhat abstract, in this example these with Di as in Eq. (140), and Ai general elements of T2 equivalences have a simple geometric origin, which can θ′ . This generalizes to a procedure for deriving connec- be understood in terms of the constructions of the previ- tions on Td, which by taking more general U produces all ous subsection (Connes, 1994; Douglas, 1999). It is that θ i T2 constant curvature connections. It corresponds directly θ can be obtained as a quotient of IR by the identifi- cations x x +2π x + θ, in other words by a two to the string and M(atrix) theory definition of quotient ∼ ∼ space, and thus we will find in Sec. VII.B that T appears dimensional lattice, and as such its moduli space will ad- θ ZZ naturally in this context. mit the SL(2, ) symmetry of redefinitions of the lattice, T2 2 ZZ Although we started with the continuum definition Eq. just as in the construction of as IR / 2. Indeed, as T2 this suggests, θ can be regarded is a zero volume limit (137), one could also obtain this as an explicit limit of T2 matrix representations. Using this in the constructions of , a picture which will reappear in the string theory we just discussed leads to the regulated gauge theory discussion. discussed in Sec. II.E.6. Morita equivalence between higher dimensional non- commutative tori is also understood (for irrational θ at least). The basic result is the following (Rieffel and Td Td F. Morita equivalence Schwarz, 1998): two tori θ and θ′ are Morita equiva- lent if they are related as Two algebras and ˆ are Morita equivalent if there is 1 a natural one-to-oneA mapA carrying all modules and their θ′ = (Aθ + B)(Cθ + D)− . associated structures from either algebra into the other. A B On the face of it, this is quite a strong relation, which for where is a 2d 2d matrix belonging to the commutative algebras C(M) and C(Mˆ ) would certainly C D × ˆ group SO(d, d; ZZ). The simplest example is A = D = 0, imply that M ∼= M. We first note that if such a map exists, we can derive B = 1, C = 1 which corresponds to the Morita equiv- 1 − T2 the general map just knowing the counterpart in ˆ of , alence on θ we described above. The corresponding the free module of rank 1. Call this P ; it is anA ( ˆ, A)- transformation on the modules is also given by a linear bimodule, because we know that 1 also admits aA rightA action of SO(d, d) on the Chern class data, which can be action of . There is then a generalA construction (the regarded as a spinor of this group. mathematicalA notion of tensor product) which produces These transformations agree precisely with the action the map: of T-duality in toroidal compactifications of string the- ory (in the limit of a zero volume torus) and this is how E Eˆ = P E. (147) they were first conjectured. Conversely, the mathemati- → ⊗A 38 cal proof of these equivalences is part of a new argument a special class of coordinate transformations which pre- for these dualities in M theory. serve θ, the canonical transformations, or symplectomor- phisms as they are called in mathematics. In infinitesimal form, these are defined by δx = S, x for some generat- { } G. Deformation quantization ing function S (if π1(M) = 0, one must allow multivalued generating functions). One6 can regard these transforma- Deformation quantization (Bayen et al., 1978; Stern- tions as generating an infinite dimensional Lie algebra, heimer, 1998) is a reformulation of the problem of quan- for which the symplectomorphisms are the correspond- tizing a classical mechanical system as follows. One first ing Lie group. considers the algebra of observables of the classical Having defined the deformations of interest, we can A problem; if one starts with a phase space M (perhaps now discuss the question of whether they exist. Con- the cotangent space to some configuration space, but of structing one involves postulating additional terms in Eq. course it can be more general), this will be the algebra (148) at all higher orders in h to make the star product of functions C(M). One then finds a deformation of this associative, (f g) h = f (g h). One might wonder if algebra in the sense of Sec. II.B.3, i.e. a family of alge- this procedure∗ requires∗ more∗ data∗ than just the Poisson bras h depending on a parameter h which reduces to bracket, and a little reflection shows that it surely will. A C(M) as h 0, and for which the leading term in the After all, we know of a valid star product for constant → star commutator in a power series expansion in h is the θij , the Moyal product Eq. (21), and if this were the out- Poisson bracket, put of a general prescription depending only on θij , we would conclude that the Moyal product intertwines with f g g f = ih f,g + . . . (148) ∗ − ∗ { } canonical transformations; i.e. if f T (f) is a canonical → One can then reinterpret Heisenberg picture equations of transformation, we would have motion such as T (f g)= T (f) T (g). ∗ ∗ ∂f i = [H,f] However, a little experimentation with nonlinear canoni- ∂t h cal transformations should convince the reader that this as equations for observables in h involving star commu- is false. Thus the question arises of what is the additional tators. A data required to define a star product, and what is the As discussed in Sec. II.B.3, the Moyal product is pre- relation between these different products. cisely such a deformation of the multiplication law of Deformation quantization has been fairly well under- functions, and deformation quantization would appear stood by mathematicians and we briefly summarize the to be a very direct way to generalize the construction of main results, referring to (Fedosov, 1996; Kontsevich, noncommutative field theory on IRθ and Tθ to general 1997) for more information. First, deformation quantiza- noncommutative spaces. Not only is it more general, it tion always exists. In the symplectic case this was shown can be formulated geometrically, without recourse to a by DeWilde and Lecomte (1983) and by Fedosov (1994), specific coordinate system. The primary input, the Pois- who also constructed a trace. For more general Poisson son bracket, can be specified by the choice of an antisym- manifolds, it was shown by Kontsevich (1997). As we metric bi-vector field θij , will discuss in Sec. VII.C, Kontsevich’s construction is in terms of a topological string theory, and has been rather ij f,g = θ ∂if∂j g influential in the physical developments already. { } In the symplectic case, it is known that the additional such that f,g satisfies the Jacobi identity. data required is a choice of symplectic connection, analo- For the{ further} discussion, it will be useful to recall gous to the connection of Riemannian geometry but pre- a bit of the canonical formulation of classical mechanics serving the symplectic form, ω = 0. Unlike Riemannian and its underlying mathematics. A simplifying assump- geometry, this compatibility∇ condition does not deter- tion which often holds is that θij is everywhere nondegen- mine the connection uniquely, and acting with a canon- erate, in which case one speaks of a manifold with sym- ical transformation will change the connection. In this plectic structure. The symplectic form is ω = (θij ) 1 ij − language, the special role of the canonical coordinates is (the matrix inverse at each point), and the Jacobi iden- analogous to that of normal coordinates; they parame- tity for the Poisson bracket is equivalent to dω = 0. More terize geodesics of the symplectic connection, xi = 0. generally, θij can degenerate at points or be of less than Although the precise form of the star product∇ depends full rank, in which case one speaks of a manifold with a on the connection, different choices lead to “gauge equiv- Poisson structure. alent” star products, in the sense that one can make a The simplest symplectic structure is constant θij , in transformation which case by a linear transformation we can go to canon- ical coordinates as in Eq. (6). Indeed, one can make G(f g)= G(f) ′ G(g) a coordinate transformation to canonical coordinates in ∗ ∗ of the form any contractable region in which θij is everywhere non- i 2 ij degenerate (Darboux’s theorem). One then distinguishes G(f)= f + hg1∂if + h g2 ∂i∂j f + . . . 39 which relates the two. Thus the two algebras are (for- ometry has a deep underlying significance in string the- mally) equivalent. ory, it was hard to guess just from this formalism what it We now ask what is the relation between deforma- might be. Further progress in this direction awaited the tion quantization and more conventional physical ideas discovery of the Dirichlet brane (Dai, Leigh, and Polchin- of quantization, or other mathematical approaches such ski, 1989; Polchinski, 1995), which gave open strings a as geometric quantization. Now the usual intuition is much more central place in the theory, and allowed mak- that the dimension of the quantum Hilbert space should ing geometric interpretations of much of their physics. be the volume of phase space in units of (2π~)d, so a fi- Now contact between string theory and conventional nite volume phase space should lead to a finite number geometry, as epitomized by the emergence of general rel- of states. ativity from string dynamics, relies to a large extent on This fits with our earlier discussion of fuzzy S2 but the curvatures and field strengths in the background be- leads us to wonder how we obtained field theory on Tθ. ing small compared to the string length ls. Conversely, 2 This appears to be connected to the fact that π1(T ) = when these quantities become large in string units, one ZZ2 = 0, and a nice string theory explanation of how this may (but is not guaranteed to) find some stringy gener- changes6 the problem can be found in Seiberg and Witten alization of geometry. (1999); because a string can wind about π1, the phase The simplest context in which noncommutative field space of an open string on T 2 is in fact noncompact. theory as we described it arises, and by far the best un- In any case, deformation quantization gets around all derstood, is in a limit in which a large background an- of these questions in a rather peculiar way: the series tisymmetric tensor potential dominates the background expansions in h one obtains are usually “formal” in the metric. In this limit, the world-volume theories of Dirich- sense that they do not converge, not just when applied let branes become noncommutative (Connes et al., 1998; to badly behaved functions but for any sufficiently large Douglas and Hull, 1998). This can be seen from many dif- class of functions. Thus they typically (i.e. for generic ferent formal starting points, as elucidated in subsequent values of h) do not define algebras of bounded operators, work, and it provides very concrete pictures for much of and do not even admit representations on Hilbert space the physics we discussed in sections III and IV. It will of the sort which explicitly or implicitly lay behind many also lead to new theories: noncommutative string theo- of our considerations (Fedosov, 1996; Rieffel, 1998). ries, and even more exotic theories such as open mem- Given some understanding of this point, in the general brane (OM) theory. case one also needs to define a noncommutative metric After reviewing a range of arguments which lead to gij (x) to make sense of Eq. (29); other field theory ac- noncommutative gauge theory, we focus on its origins tions require even more structure. from the string world-sheet, following Seiberg and Wit- At this writing, the question of whether and when de- ten (1999), who were the first to precisely state the limits formation quantization can be used to define noncommu- involved. We also describe related arguments in topo- tative field theory is completely open. Some suggestions logical string theory, originating in work of Kontsevich in this direction have been made in (Asakawa and Kishi- (1997). We then give the string theory pictures for the moto, 2000; Jurco, Moller, Schraml, Schupp, and Wess, solitonic physics of Sec. III, and other contacts with du- 2001). ality such as the AdS/CFT proposal (Maldacena, 1998). Finally, we discuss some of the limits which have been proposed to lead to new noncommutative theories. VII. RELATIONS TO STRING AND M THEORY We now discuss how noncommutative field theories A. Lightning overview of M theory arise from string theory and M theory, and how they fit into the framework of duality. Obviously space does not permit a real introduction to Historically, the first use of noncommutative geometry this subject, but it is possible to summarize the definition in string theory was in the formulation of open string of M theory so as to provide a definite starting point for field theory due to Witten (1986), which uses the Chern- our discussion. Details of the following arguments can be Simons action in a formal setup much like that of Sec. II, found in (Polchinski, 1996a,b). Throughout this section, with an algebra defined using conformal field theory we will try to state the central ideas at the start of each techniques, whoseA elements are string loop functionals. subsection, though as the discussion progresses we will Noncommutativity is natural in open string theory just reach a point where we must assume prior familiarity on because an open string has two ends, and an interac- the part of the reader. tion which involves two strings joining at their end points A unified way to arrive at the various theories which shares all the formal similarities to matrix multplication are now considered part of M theory is to start with one which we took advantage of in Sec. II.C. of the various supergravity actions with maximal super- Although these deceptively simple but deep observa- symmetry (32 supercharges in flat spacetime), compact- tions combined with the existence of the string field ify some dimensions consistent with this supersymmetry framework strongly suggested that noncommutative ge- (the simplest choice is to compactify n dimensions on the 40 torus T n), find the classical solutions preserving as much Considerations involving supersymmetry as well as many supersymmetry as possible (16 supercharges), make ar- nontrivial consistency checks have established this claim guments using supersymmetry that these are exact so- beyond reasonable doubt, as part of a much larger “web lutions of the quantum theory, and then claim that in of dualities” involving all of the known string theories a given background (say, with specified sizes and shapes and many of their compactifications. of the torus), the lightest object must be fundamental, The string theory limits are still much better under- so some fundamental formulation should exist based on stood than the others, because the string is by far the that object. most tractable fundamental object. One can use them Thanks to the remarkable uniqueness properties of ac- to make a microscopic definition of certain branes, the tions with maximal supersymmetry, in each case only one Dirichlet branes. A Dirichlet brane is simply an allowed candidate theory survives even the simplest consistency end point for open strings. The crucial generalization be- checks on these ideas, and thus one can be surprisingly yond the original definition of open string theory is that specific about these fundamental formulations and see one allows Dirichlet boundary conditions for some of the some rather non-trivial properties of the theory even in world-sheet coordinates and this fixes the endpoint to live the absence of detailed dynamical understanding. on a submanifold in space-time. For a simple choice of The simplest and most symmetric starting point is submanifold such as a hyperplane, the world-sheet the- eleven-dimensional supergravity (Cremmer, Julia, and ory is still free, so the physics can be worked out in great Scherk, 1978), a theory with no free parameters and a detail. single preferred scale of length, the Planck length lp. Its The central result in this direction and the starting fields are the metric, a spin 3/2 gravitino, and a third point for most further considerations is that the quanti- rank antisymmetric tensor potential, traditionally de- zation of open strings ending on a set of N Dp-branes, noted Cijk. Such a potential can minimally couple to occupying coincident hyperplanes in ten-dimensional a 2 + 1-dimensional extended object, and indeed a solu- Minkowski spacetime, leads to p + 1-dimensional U(N) tion exists corresponding to the background fields around MSYM. Its field content is a vector field, 9 p adjoint such an object and preserving 16 supersymmetries. One scalars and their supersymmetry partner fermions,− and can also find a supersymmetric solution with magnetic its action is the dimensional reduction of Eq. (31). charge emanating from a 5 + 1 dimensional hypersur- A crucial point is the interpretation of the adjoint face. We can thus define M theory as the well-defined scalars. Let us denote them as hermitian matrices Xi; quantum theory of gravity with the low energy spectrum the action contains a potential of this supergravity, containing solitonic “branes,” the 2-brane (or supermembrane) and 5-brane, whose long i j 2 range fields (at distances large compared to lp) agree with V = Tr [X ,X ] − the solutions just discussed. These branes have tensions i B. Noncommutativity in M(atrix) theory This construction admits a natural generalization, namely one can impose the relations As we commented in Sec. II.D.1, the simplest deriva- iθij tion of noncommutative gauge theory from a more fa- UiUj = e Uj Ui. miliar physical theory is to start with a dimensionally reduced gauge theory (or “matrix model”) action such Again as discussed in Sec. VI.C, Eq. (143) now defines a as Eq. (38), and find a situation in which the connection twisted crossed product, and its solutions Eq. (146) are operators Ci of (34) obey the defining relations of a con- connections on the noncommutative torus. Substituting nection on a module over a noncommutative algebra as these into Eq. (149) leads directly to noncommutative given in Sec. VI.D, perhaps after specifying appropriate gauge theory. This was how noncommutativity was first boundary conditions or background fields. introduced in M theory by Connes et al. (1998). A particularly significant theory of this type is max- Having seen this possibility, one next must find the imally supersymmetric quantum mechanics, the p = 0 physical interpretation of this noncommutativity. Since case of MSYM, with action: M theory has no dimensionless parameters, one is not allowed to make arbitrary modifications to its definition 9 2 i j 2 i but rather must identify all choices made in a particular S = dt Tr (D X) [X ,X ] +χ†(D +Γ X )χ. t − t i construction as the values of background fields. Although i=1 i SL(2, ZZ) duality of N = 4 SYM proposed by Montonen and Olive (1977). The generalized compactifications with C ij = 0 allow accessing more dualities, namely those which− can6 be seen in the limit of compactification on a T d+1 containing a light-like dimension. For example, the noncommutative 2 + 1 MSYM is predicted to have SL(2, ZZ) SL(2, ZZ) duality symmetry, the subgroup of SL(3, ZZ) × SL(2, ZZ) preserving the choice of light-like direction. Now× the sec- ond, nonclassical duality is manifest: it is the SL(2, ZZ) T2 Morita duality group of θ, and one can verify that the spectrum predicted in Eq. (117) is invariant under this change of parameters. This type of argument has been extended in many di- rections, and an in-depth treatment would require an- other full length review, which happily already exists (Konechny and Schwarz, 2000b). In the M(atrix) the- ory context, it has been argued (Verlinde, 2001) that by careful treatment of the limit, one can extend the du- ality to a general T d+1, not necessarily preserving the FIG. 6 T-duality to an anisotropic torus. light-like dimension. In the string theory context, as we discuss below, one can see that the nonclassical duality arises from T-duality, leading to the prediction that non- w2 becomes a new component of momentum in 2 + 1- commutative gauge theory on T d will have SO(d, d; ZZ) dimensional gauge theory. In this anisotropic geometry, duality, which was the motivation behind the theorem of the two ends of a winding open string will have differ- Rieffel and Schwarz (1998) discussed in Sec. VI.E. ent locations in x1, with separation θw2. Thus the fun- damental objects turn out to be dipoles in exactly the sense described in Sec. II.B.2, with the corresponding C. Noncommutativity in string theory noncommutative interactions. This construction directly 1 reproduces the quotient construction of Tθ as C(S )/ZZ. The arguments we just gave established that limits of A world-sheet argument treating D0-branes on the M theory and string theory compactified on a torus will original torus leads to the same conclusion (Cheung and lead to noncommutative gauge theory, realizing the in- Krogh, 1998). Now one must take the size R of both trinsic noncommutativity of open strings. However the axes of the torus small, and keep winding strings in both torus is not the simplest noncommutative space; one is directions. The point now is that since the term B in led to ask whether noncommutativity can also emerge d Eq. (150) is a total derivative, it contributes to the in- without compactification, leading to gauge theory on IRθ. R teraction of two strings with winding numbers (w1, w2) This question gained particular focus after the discov- ery of noncommutative instantons on IRd (Nekrasov and and (w1′ , w2′ ) about the two axes of the torus by a phase θ Schwarz, 1998). proportional to the product w1w2′ w2w1′ , directly pro- ducing Eq. (21). This argument can− be generalized to While the arguments of the previous section lead di- d other string theory situations involving similar phases, rectly to this result (to obtain IRθ, one just adjusts the such as discrete torsion on orbifolds (Vafa, 1986), and parameters to make the string modes light, which takes Td leads to the twisted crossed product with finite groups the volume of θ to infinity), many other string the- (Ho and Wu, 1998; Douglas, 1998). ory computations on IRd with general background B had These derivations arose naturally in the consideration been done previously and noncommutativity had not of M(atrix) theory, and in this context there is a rather been seen. Indeed, there are general arguments which striking test one can make. The original derivation led lead from open strings to conventional gauge theory, to an identification of M theory compactified on T d as making the new claim appear paradoxical. a large N limit of d + 1-dimensional MSYM, and the On the other hand, Kontsevich (1997) had argued that most basic prediction of this identification is that the two deformation quantization and the Moyal product in par- theories share the same duality properties. Connes et al. ticular could come from “open string theory,” at least (1998) discussed compactification on T 3, which according in a mathematical sense. This was turned into an argu- to M theory considerations must be invariant under an ment in topological string theory by Cattaneo and Felder action of the U-duality group SL(3, ZZ) SL(2, ZZ) on the (2000) and in physical string theory by Schomerus (1999), × d moduli and brane spectrum. The SL(3, ZZ) can easily be proving that IRθ could indeed emerge directly from world- identified with large diffeomorphisms of the T 3, but the sheet considerations. SL(2, ZZ) symmetry is a prediction: in fact it is just the The paradox was resolved by (Seiberg and Witten, 43 1999) who explained how both conventional and noncom- with arbitrarily large but slowly varying field strength, mutative descriptions could be correct, along the general the Nambu-Born-Infeld (NBI) action lines we already indicated in Sec. II.D.4. Their careful 1 treatment of the limit leading to noncommutative field S = dp+1x det(g +2πl2(B + F )). NBI g l (2πl )p s theory has spurred numerous further developments. s s s Z The first point to make is that on IRd and unlike T d,a p (155) constant background potential Bij is pure gauge and thus Here gs is the string coupling, and g is the induced metric cannot affect closed string physics. However, this is not on the brane world-volume. See Schwarz (2001b) for its true in the presence of a Dirichlet brane extending along supersymmetrization, and other references on this topic. the directions i, j, because the gauge transformation also The original string computations leading to the Born- acts on its world-volume gauge potential A as Infeld action were done by Fradkin and Tseytlin (1985) and Abouelsaood, Callan, Nappi, and Yost (1987) in the δB(2) = dλ(x); δA = λ(x). (151) context of type I open string theory, i.e. the case p = 9. For p < 9, the induced metric g contains the informa- This can be seen by rewriting the total derivative term tion about the embedding of the brane, and substituting in Eq. (150) as a boundary term, as is appropriate in the F = 0, Eq. (155) reduces to the Nambu action govern- presence of open strings, and adding the gauge coupling: ing its dynamics. Finally, the prefactor is the D-brane tension as computed from string theory. Besides detailed i j computation, there are simple physical arguments for the Sb = i Aj (x) Bij x ∂tx , (152) − Born-Infeld form (Bachas, 1996; Polchinski, 1998). Z∂Σ This action summarizes essentially all weakly coupled where ∂t and ∂n denote the tangential and normal deriva- and “weakly stringy” physics of a single D-brane and is tives along the boundary ∂Σ of the worldsheet Σ. The 2 even valid in the large ls (B + F ) limit, so one might transformation Eq. (151) can be undone by an integra- at first hypothesize that it is valid in the limit of large tion by parts. B we just discussed, without need of noncommutativity. Eq. (151) implies that the open string effective action However, it is not in general valid for rapidly varying field can only depend on the combination F + B, not F or B strengths ls∂F 1, nor is its nonabelian generalization separately. In particular, one can gauge B to zero, replac- understood. Thus∼ we can reconcile our earlier arguments ing it by a background magnetic field. However, because 2 2 for noncommutativity with the NBI action if θ dz V (z ) dz V (z ) . . . dz V (z ) . (154) or any operator obtained by multiplying this by confor- h 1 1 1 2 2 2 n n n i Z Z Z mal fields ∂x, fermions in the superstring, etc. Since the action Eq. (150) is quadratic, the world- on a world-sheet Σ with the disk topology, with operators sheet physics is entirely determined by the propagator V at successive points z on the boundary ∂Σ, integrated i i xi(z) xj (w) . The boundary conditions which follow over all z satisfying the same ordering as in Eq. (153). i hfrom varyingi the action Eq. (150) are: Taking only vertex operators for the massless fields, one finds that the leading ls 0 limit of the S-matrix is j 2 j gij ∂nx +2πiBij ls ∂tx ∂Σ =0 (157) reproduced by the MSYM effective→ action. It turns out | that these considerations also lead to a simple universal Now, taking Σ to be the upper half-plane with the coor- effective action which describes the physics of a D-brane dinate z = t+iy, y > 0 we find the boundary propagator 44 to be (162) will reduce to an associative algebra, is rather gen- eral and is best thought of in the framework of topolog- i j ij 2 i ij x (t)x (s) = α′G log(t s) + θ ǫ(t s) (158) ical string theory (Witten, 1988; Dijkgraaf, 1998). For h i − − 2 − present purposes, these are string theories for which cor- where ǫ(t)= 1, 0, +1 for t< 0,t =0,t> 0 respectively, relation functions depend not on the locations of oper- and − ators but only on their topological arrangement on the world-sheet, which is exactly the property we used of the ij ij 1 gij 0 limit of Eq. (162). G = g+2πl2B (159) → s S One can use topological string theory to construct a de- ij ij 2 1 formation quantization corresponding to a general Pois- θ =2πls +2 2 (160) g πlsB A son structure, generalizing the discussion we just gave. We summarize this, following the work of Baulieu, Lo- where S and A denote the symmetric and antisymmetric sev, and Nekrasov (2001). See also (Cattaneo and Felder, parts respectively. From this expression we deduce : 2001). [xi, xj ] := T (xi(t 0)xj(t) xi(t+0)xj(t)) = iθij . (161) We start with the action Eq. (150), not assuming that − − gij ,Bij are constant and rewrite it in the first order form: Thus the end-points of the open strings live on a non- 1 i a j 2 ab i j S = 2 (gij ∂ax ∂ x 2πilsBij ǫ ∂ax ∂bx )= commutative space with 4πls Σ − i 2 ij 1 ij piR dx πl G pi ⋆pj θ pi pj (163) [xi, xj ]= iθij ∧ − s ∧ − 2 ∧ where R with θij a constant antisymmetric matrix. Similarly, Eq. 1 (156) becomes 2 2 2 − 2πlsG + θ =2πls g +2πilsB 2 ij i ij 2l G piqj θ piqj Vp(t)Vq (s) = (t s) s e− 2 Vp+q(s) (162) 2 − Now take the ls 0 limit keeping θ and G fixed. The remaining part →p dx + 1 θp p of the action Eq. Indeed, in this free world-sheet theory, it is no harder 2 (163) exhibits an enhanced∧ gauge∧ symmetry (Cattaneo to compute the analogous phase factor for an n-point and Felder, 2000),R function: it is precisely Eq. (100). In the formal limit gij 0, one finds from Eq. (159) p p dλ ∂ θjkp λ , xi xi + θij λ (164) ij → ij 2 1 i i i i j k j and Eq. (160) that G = 0 and θ =2πls(Bij )− . Thus 7→ − − 7→ the dependence on the world-sheet coordinates s and t To quantize, this symmetry must be gauge fixed, which drops out (the vertex operators have dimension zero), can be done by standard BV procedures, leading to a and the OPE reduces to a conventional multiplication topological string theory with some similarities both with law. For Eq. (162), this is the Moyal product Eq. (21), the type A and type B sigma models. Its field content is i and by linearity this extends to the product of two gen- conveniently described by promoting pi and x to twisted eral functions. superfields, with an expansion in world-sheet differential We have again found that background Bij leads to forms dσi with components of all degrees 0, 1, 2. The noncommutativity. A precise connection to the previous original fields are the 0-form part of xi and the 1-form discussion of toroidal compactification can be made by part of pi; the other components are the additional ghosts applying T-duality to turn the D2-branes of the present and auxiliary fields of the BV framework. discussion into D0-branes, and then taking the zero vol- A basic observable in this theory is the three-point ume limit of the torus. This T-duality leads to the rela- function on the disc, ij 2 1 tion θ = 2πls (Bij )− . However, the present argument also works for IRd. θ f(x(0))g(x(1)) [h(x)χ...χ] ( ) θ,Σ=disc = f⋆gh(165) A rather similar argument shows that world-volumes h ∞ i ZX of D-branes in the Wess-Zumino-Witten model are de- for f,g C(X), h ΩdimX (X). As the notation indi- scribed by field theory on “fuzzy spheres” as described in ∈ ∈ Sec. VI.B (Alekseev, Recknagel, and Schomerus, 1999). cates, this will be identified with the star product of a An interesting difference is that here dB = 0 and the cor- deformation quantization associated to θ. This can be responding algebra is not associative, except6 in suitable seen by developing the perturbation series in powers of limits. θ; each term in the expansion can be expressed as an explicit sum over Feynman diagrams, producing Kontse- vich’s construction of deformation quantization. 2. Deformation quantization from topological open string An important advantage of this approach is that many theory properties of the formalism have simple arguments from string theory. In particular, the associativity of the ⋆- The idea that by considering only vertex operators of product defined by Eq. (165) follows in a sense from dimension zero, a vertex operator algebra such as Eq. associativity of the OPE. This is best expressed in terms 45 of a more general set of Ward identities obeyed by the relation between the mass of a state and the world-sheet string amplitudes in the theory, which allow making con- dimension of the corresponding vertex operator, which tact with and generalizing the discussion in Sec. VI.G. (as is visible in Eq. (162)) is controlled by Gij . Thus the As an example of a new result derived from Eq. (163), decoupling limit, sometimes referred to as the Seiberg- we give an expression for the θ-deformed action of in- Witten limit in this context, takes ls 0 while holding finitesimal diffeomorphisms δxi = vi(x) on functions and G and θ fixed. → covariant derivatives (Baulieu et al., 2001), which can be The nontrivial relation between the original metric gij used to make Eq. (24) generally covariant. One defines and Gij will show up at many points in the subsequent these in terms of the conventional action on functions discussion. One refers to gij as the “closed string metric” i ij δf = v ∂if, with θ transforming as a conventional ten- and G as the “open string metric” as they each govern ij k ij ki j kj i sor, δθ = v ∂kθ θ ∂kv + θ ∂kv . In the TFT, this the kinetic terms and the energies of the lightest states becomes − in their respective sectors. Once this realization is made, the subsequent steps (1) δf = U f(x)= f(x(0)) p vi(x) [δ(x( ) x)] in the derivation of the D-brane world-volume action v h i ∞ − iθ,Σ I go through without major changes from the conven- tional case, leading to noncommutative MSYM and even where the integral is taken over an arc surrounding the the noncommutative NBI action, defined by taking Eq. point 0 and ending on the boundary of the disc Σ. This (155), replacing products with star products, and setting leads to the metric to Gij and B = 0. The remaining step is to determine the prefactor and U f(x)= vi∂ f(x)+ θkl∂ vi∂2 f(x)+ .... v i k li thus the gauge coupling. This follows directly if we accept that the Seiberg-Witten map of Sec. II.D.4 between con- ventional and noncommutative gauge theories maps the 3. The decoupling limit conventional action Eq. (155) into the noncommutative ˆ Many of the deeper aspects of the connection to string action Eq. (155), as we can just specialize to F = F = 0. theory require a more careful treatment of the decoupling This leads to the relation limit leading to noncommutative field theory, as was first det(G) det(g +2πl2B) = s . (166) made by Seiberg and Witten (1999). G l (2πl )p g l (2πl )p String theories have many more perturbative degrees ps s s p s s s of freedom than field theories, so to get field theories one which determines an “open string coupling constant” Gs looks for controlled limits in which almost all of these de- and the corresponding noncommutative gauge coupling. grees of freedom go away, usually by sending their masses The decoupling limit now will take ls 0 in Eqs. to infinity. In the context of D-branes, the generic such (159), (160) and (166), scaling the original→ string cou- 2 (3 p+r)/4 limit takes ls 0 and thus the string tension T = 1/ls pling as g l − to end up with noncommutative → s s to infinity, and simultaneously takes the transverse dis- Yang-Mills∼ theory with finite parameters, tance L between branes to zero holding LT fixed. This 1 2 2 kl is the energy scale of the lightest open strings stretched θ = B− ; Gij = (2πls ) BikBjlg . (167) between branes, so in this limit we keep these degrees of freedom while sending excited string energies to infinity 4. Gauge invariance and the SW map as 1/ls. Finally, one rescales the coupling constant to keep it fixed in the limit. This leads to a field theory of the lightest open strings, which for flat D-branes will be An important point which can be seen by carrying MSYM. Similar limits in other brane theories can lead to out this discussion more explicitly is the precise point more exotic results, as we will mention below. at which conventional gauge invariance is replaced by However there are other massless states in the string noncommutative gauge invariance. As we mentioned in theory as well, the closed strings which lead to the gravi- Sec. II.D.4, there is a very general world-sheet argument tational sector, and we need to argue that these decouple. for conventional gauge invariance, and indeed this argu- A naive but often correct argument for this is that their ment is not incorrect; rather, one obtains noncommuta- couplings are gravitational and are suppressed by a fac- tive gauge invariance by choosing different conventions, and there is a formal equivalence between conventional tor GN E which wlil also go to zero in this limit. This requires detailed consideration in examples, however. It and noncommutative gauge theories. (Seiberg and Wit- will turn out to be true in the case at hand at least up ten, 1999; Andreev and Dorn, 2000b; Seiberg, 2000). to 3 + 1 dimensions; we will discuss a potential subtlety The origins of gauge invariance can be seen in open below. bosonic string theory. In this theory, the vertex operator The key new point in the present context is that the for a gauge boson is masses of open strings in this limit are determined by the i ip x i ei(p) : ∂tx e · : A = Ai(x)dx (168) metric Gij defined in Eq. (159), which has nontrivial B ←→ dependence. This follows from the standard string theory as was already implicit in Eq. (152). 46 The abelian gauge invariance, form of Eq. (155), 1 δAj = ∂j ε, (169) 2 ˆ SNCNBI = p Tr det(G +2πls(F + Φ)). Gsls(2πls) Z q then follows by varying Aj in Eq. (152), by taking the (173) integral of the total derivative dz∂t(ε) as zero. Extra Using Eq. (42), one can show that for slowly varying 2 terms can appear in Eq. (169) if there are divergences background fields lsF << 1, the descriptions with differ- R when the operator under consideration (say Vi) coincides ent values of θ are equivalent. 1 with its neighbors Vi 1 and Vi+1. In Yang-Mills theory, In the special case of Φ = θ− , these relations sim- − − this leads to contact terms Vi 1ε εVi+1 which become plify even further to Eq. (167), even at finite ls. As the nonlinear terms in the gauge− transformation− law. mentioned in Sec. II.B.1, this is what one gets if one The previous discussion makes it very plausible that defines the derivatives as inner derivations, so this is such an argument will carry through in the noncommu- what emerges naturally from matrix model arguments tative case with the star product appearing as in Eq. (Seiberg, 2000). (162). However, making this argument precise requires More generally, it has been shown that the techniques choosing a cutoff on the world-sheet, and different choices of Kontsevich’s deformation quantization can be used to at this step can lead to different results. If one point- write a Seiberg-Witten map for arbitrary (non-constant) splits and uses the propagator Eq. (158), one obtains θ (Jurco, Schupp, and Wess, 2000). the star product, but one can also find prescriptions in which the term proportional to θij goes away in the co- incidence limit, leading to conventional gauge invariance D. Stringy explanations of the solitonic solutions instead. In particular, this will be the result if one de- fines a propagator and cutoff at B = 0, treating the B So far we have only discussed the simplest arrangement term in Eq. (150) using world-sheet perturbation theory. of Dirichlet branes, N parallel branes of the same dimen- Physics cannot depend on this choice and in general sion. There are a bewildering variety of more complicated a change of renormalization prescription on the string possibilities, with branes of lower dimension sitting inside world-sheet corresponds to a field redefinition in space- of those of higher dimension, intersecting branes, curved time. By the preceding arguments, this field redefinition branes and so forth. must be the Seiberg-Witten map of Sec. II.D.4. The two Not all of the possibilities are actually distinct, how- descriptions are therefore equivalent, at least in some for- ever. For example, a configuration of lower dimensional mal sense, but they are each adapted to different regimes, D-branes sitting inside a higher dimensional D-brane, is with conventional gauge theory simpler for small B and topologically equivalent to and can often be realized as noncommutative gauge theory simpler for large B. a limit of a smooth configuration of gauge fields on the Once we understand that the resulting gauge invari- higher dimensional brane. Another example is that an ance depends on the choice of world-sheet regularization, intersecting brane configuration, say of branes A and B, we can consider choices that lead to different star prod- can often be described as a single object, a nontrivial ucts. A simple choice to consider is one which leads to embedding of brane A in the higher dimensional space the same star product Eq. (21), but defined using a pa- and thus by some configuration of scalar fields on brane rameter θ which does not satisfy Eq. (160), which would A. Thus, the solitons and instantons in D-brane world- be obtained by treating part of the B term in Eq. (150) volume gauge theory themselves have interpretations as as a perturbation. Denote this part as Φ; it will enter D-branes. We refer to (Polchinski, 1996b, 1998) for an Eq. (155) as did B, so we would obtain a description in introduction and overview of this subject. terms of a noncommutative action depending on Fˆ + Φ. A great deal of this structure survives the limit tak- This description can be obtained from string theory ing string theory to noncommutative field theory, lead- by a simple generalization of the discussion above. The ing to stringy pictures for the gauge theory solutions of result is again Eqs. (159) and (160) but now with an an- Sec. III. What is quite striking is that the agreement is tisymmetric term in the metric on both sides (analogous not just qualitative but even quantitative, with brane to the B term in Eq. (150)). Combining this with Eq. tensions and other properties agreeing, apparently for (166) to determine the gauge coupling, we obtain reasons other than supersymmetry (Dasgupta, Mukhi, and Rajesh, 2000; Gross and Nekrasov, 2000b; Harvey, 1 1 Kraus, Larsen, and Martinec, 2000b). Establishing this G+2πl2Φ = g+2πl2B (170) s S s S detailed agreement of course requires a good deal of string 1 1 1 theory input and we refer to Nekrasov (2000) and Harvey G+2πl2Φ + 2πl2 θ = g+2πl2B (171) s A s s A (2001a) for detailed reviews of this class of results, but 2 1/2 provide an introduction here. det(G+2πlsΦ) Gs = gs det(g+2πl2B) (172) The simplest examples are the D(p k)-branes embed- s − ded in Dp-branes. There is a general result to the effect which determines the parameters in the noncommutative that a collection of N Dp-branes carrying gauge fields 47 with Chern character the D3-branes. The background B-field acts as a mag- netic field, which pulls these charges apart. Tilting of ch(F ) = Tr eF = N + ch(1) + ch(2) + . . . i1i2 i1i2i3i4 the D1-brane makes its tension work in the opposite di- carries the same charges as (and is topologically equiv- rection and stabilizes the configuration. This qualitative alent to) a collection with no gauge fields, but with ad- picture is confirmed by the profile of the explicit analytic ditional D-branes whose number and orientation is given solution (Gross and Nekrasov, 2000b). by the quantized values of the Chern characters. For ex- Finally, let us mention the work of Myers (1999), which (1) shows that Dp-branes in background Ramond-Ramond ample, ch counts D(p 2)-branes which are localized ij fields form bound states which can be thought of as in the i and j dimension,− and so forth. D(p + 2)-branes whose world-volume is a product space Carrying this over to the noncommutative theory in with a fuzzy S2. This noncommutativity plays a num- the obvious way, the fluxon of III.B is an embedded D(p ber of interesting roles in M theory (Myers, 2001) and 2)-brane, and the instanton of III.C is an embedded D(p− its slightly different origin poses the challenge to find a 4)-brane. According to this identification and the results− broader picture which would incorporate it into our pre- of Sec. III, the D(p 2) should be unstable to decay, while vious discussion. the D(p 4) should− be stable, and this is borne out by computation− in string theory; the destabilizing mode in Eq. (87) is a “tachyonic open string.” The agreement is much more detailed and striking than E. Quantum effects and closed strings this. Another qualitative point is that m coincident D(p 2)-branes must themselves carry U(m) gauge symmetry;− The string theory derivation of noncommutative gauge this is manifestly true of the m-fluxon solution Eq. (83) theory allows one to compute its coupling to a back- with U = Sm. A similar discussion can be made for ground metric, at least for small variations of the flat higher codimension branes and even for processes such background, by adding closed string graviton (and other) as the annihilation of a D-brane with an anti-D-brane, vertex operators in the world-sheet computations dis- which are quite difficult to study in the original string cussed above. Results of this type are given by many theory. A very general relation between the topology authors; we mention (Das and Trivedi, 2001; Liu and of gauge field configurations in noncommutative gauge Michelson, 2001b; Okawa and Ooguri, 2001c,a) as rep- theory, and the classification of D-branes in string theory, resentative. Analysis of these couplings to Ramond- has been found in (Harvey and Moore, 2000), based on Ramond closed string fields (Liu and Michelson, 2001a; the K-theory of operator algebras. Okawa and Ooguri, 2001b; Mukhi and Suryanarayana, Even better, paying careful attention to conventions, 2001) was an important input into the result Eq. (45). one finds that Eq. (84) exactly reproduces the tension As a particularly simple example, the coupling to the of the D(p 2)-brane in string theory. It is somewhat graviton defines a stress-energy tensor in noncommuta- surprising that− the noncommutative field theory limit tive gauge theory, which turns out to be precisely Eq. would preserve any quantitative properties of the solu- (40). tions. The agreement of the tension is related to a deep The general type of UV/IR relation we discussed at conjecture of Sen (1998), as explained by Harvey et al. length in Sec. IV is very common in string theory. Per- (2000b). Witten (2000) has shown how the noncommu- haps its simplest form is visible in the computation of the tative field theory and these arguments can be embedded annulus world-sheet with one boundary on each of a pair in the framework of string field theory, and much recent of D-branes. This amplitude admits two pictures and two progress has been made in this direction (Gross and Tay- corresponding field theoretic interpretations: it can be lor, 2001; Rastelli, Sen, and Zwiebach, 2001; Shatashvili, thought of as describing emission of a closed string by one 2001). D-brane and its absorption by the other, a purely clas- The monopole in Sec. III provides another example, sical interpretation, and it can equally well be thought which is now related to intersecting brane configurations. of as the sum of one-loop diagrams over all modes of One can start from the B = 0 description of a monopole the open string stretched between the pair of branes, a in the 3 + 1 U(2) MSYM gauge theory of 2 D3-branes, purely quantum interpretation. World-sheet duality im- which is a D1-brane “suspended between” the branes, plies that the two descriptions must be equal. i.e. extending in a transverse dimension perpendicular As discussed by Douglas, Kabat, Pouliot, and Shenker to the D3-branes, and with one end on each brane. One (1997), this leads very generally to the idea that for D- can show that this configuration not only reproduces the branes at “substringy” distances (i.e. with separation magnetic charge of the solution, but even obtain the full L as above satisfying L << ls), conventional gravity is Nahm formalism from D-brane world-volume considera- replaced by quantum effects in the world-volume gauge tions (Diaconescu, 1997). theory. In general this leads to different predictions from If one turns on B along the D3-branes, the essential Einstein gravity or supergravity, but in special circum- point is now that the D1-brane is no longer perpendicu- stances (e.g. with enough supersymmetry or in the large lar, it “tilts.” The endpoints of the D1-brane are mag- N limit) the substringy predictions can agree with grav- netically charged with respect to the U(1) gauge fields on ity. Conversely, since L/T is a world-volume energy scale, 48 taking L>>ls accesses the UV limit of gauge theory am- the D3-brane realization of noncommutative gauge the- plitudes, and one sees that these are replaced by the IR ory has a supergravity dual in the large N, strong ’t limit of the gravitational description. Hooft coupling limit. This was found in (Hashimoto and Although the decoupling limit takes L 0 and is thus Itzhaki, 1999; Maldacena and Russo, 1999) and takes the in the substringy regime, one can ask whether→ neverthe- form less this potential connection to gravity can shed light on ds2 = l2√λ the nature of UV/IR mixing. s × 2 2 2 U 2 2 2 dU 2 2 A strong sense in which this could be true would be U ( dt + dx1)+ 1+λ∆4U 4 (dx2 + dx3)+ U 2 + dΩ5 if the new IR divergences could be described by adding − h 2 1 2 2 i U = 22 x4 + . . . + x9 additional light degrees of freedom in the effective theory, λls which would be directly analogous to the closed strings. eφ = λ 1 This picture was explored in (Minwalla et al., 2000; Van 4πN √1+λ∆4U 4 2 λ∆2U 4 Raamsdonk and Seiberg, 2000). For example, if we in- B = ls 1+λ∆4U 4 dx2 dx3 (174) troduce a field χ describing the new mode, we can then − ∧ It is dual to the = 4 U(N) noncommutative gauge reproduce a singularity such as Eq. (107) by adding N 2 theory, with the noncommutativity [x2, x3] ∆ , and ’t 2 ∼ Hooft coupling λ = gY M N. The transverse geometry is a d 2 2√ 1/4 ik jl 2 1 five-sphere of radius R, R = ls λ. For U 1/(∆λ ) χ(θ θ Gkl∂i∂j ) − χ + χφ ≪ 5 the solution Eq. (174) approaches the AdS5 S su- Z pergravity background, dual to ordinary large N× =4 to the effective Lagrangian. This particular example super-Yang-Mills theory. However, for large U,N corre- looks quite natural in d = 4; one can also produce dif- sponding to the large energies in the gauge theory, the ferent power laws by postulating that χ propagates in a solution differs considerably: the dilaton flows, the B- different number of dimensions than the original gauge field approaches a constant in 23 directions, and the 23- theory, as is true of the closed strings in the analogy. directions collapse. The main observation is then that, comparing with Eq. The large U limit of anti de-Sitter space is a time-like (167), the kinetic term for χ contains precisely the closed boundary, and in the usual AdS/CFT correspondence the ij string metric g , which is compatible with the idea. In- boundary values of fields are related to couplings of local deed, there are cases in which this interpretation seems operators in the gauge theory. The drastic modifications to be valid (Rajaraman and Rozali, 2000), namely those to Eq. (174) in this region have been argued by many in which the divergence is produced by a finite number of authors to be associated with the lack of conventional closed string modes because of supersymmetric cancella- local gauge invariant observables. Das and Rey (2000) tions, e.g. as in = 2 SYM (Douglas and Li, 1996). have argued that the proportionality Eq. (16) between N In general, however, the closed string picture is more length and momenta, characteristic of the open Wilson complicated than this and one cannot identify a simple loop and exploited in Eq. (113), emerges naturally from set of massless modes which reproduce the new IR effects this picture. (Gomis, Kleban, Mehen, Rangamani, and Shenker, 2000; See (Danielsson, Guijosa, Kruczenski, and Sundborg, Andreev and Dorn, 2000a; Kiem and Lee, 2000; Bilal, 2000; Li and Wu, 2000; Berman et al., 2001; Russo and Chu, and Russo, 2000). Furthermore the effective field Sheikh-Jabbari, 2001) for further physics of this corre- theory required to reproduce higher loop effects does not spondence, and (Elitzur, Pioline, and Rabinovici, 2000) look natural (Van Raamsdonk and Seiberg, 2000). for a discussion of duality and Morita equivalence. One can also argue that if this had worked in a more complicated situation, it would signal the breakdown of the decoupling limit we used to derive the theory from G. Time-like θ and exotic theories string theory (Gomis et al., 2000). This is because ex- change of a finite number of closed string modes would So far we discussed the theories with spatial noncom- correspond to exchange of an infinite number of open mutativity, which arise from Dirichlet branes with a B string modes, including the massive open strings we field along the spatial directions. There are also lim- dropped in the limit. Explicit consideration of the an- its with timelike B field, leading to exotic noncommuta- nulus diagram however shows that these massive open tive string and membrane theories, the noncommutative strings do not contribute in the limit. open string (NCOS) theory (Gopakumar, Maldacena, Minwalla, and Strominger, 2000a; Seiberg, Susskind, and Toumbas, 2000b), the open membrane (OM) the- F. AdS duals of noncommutative theories ory (Bergshoeff, Berman, van der Schaar, and Sundell, 2000; Gopakumar, Minwalla, Seiberg, and Strominger, One of the beautiful outcomes of string theory is 2000b) and the open Dp-brane (ODp) theories (Gopaku- the description of strongly coupled large N gauge theo- mar et al., 2000b; Harmark, 2001). These appear to ries by the AdS/CFT correspondence (Maldacena, 1998; evade the arguments against timelike noncommutativity Aharony et al., 2000). In particular, one expects that in field theory. 49 Let us start with the D3-brane in IIb string theory in theories have recently been reviewed by Aharony (2000). a large spatial B field. This theory exhibits S-duality, To get a noncommutative version of the open mem- which maps electric field to magnetic field and vice versa. brane theory, we start with N coincident M5-branes with Since constant spatial B-field is gauge equivalent to a the background 3-form strength and the metric: constant magnetic field on the brane, S-duality must map 3 3 sinhβ this background into one with large electric field. This H012 = Mp tanhβ, H345 = √8Mp 3 − (eβ coshβ) 2 should, in turn, lead to space-time noncommutativity, with θi0 = 0. gµν = ηµν , µ, ν =0, 1, 2 6 1 Thus, combining accepted elements of M/string dual- gij = eβ coshβ δij , i,j =3, 4, 5 ity, one concludes that our previous arguments against 1 gMN = eβ coshβ δMN M,N =6, 7, 8, 9, 11 (175) time-like noncommutativity must have some loophole. This is correct and one can in fact take the large elec- where Mp is the gravitational mass scale, and β is the pa- tric field limit for any of the Dp-brane theories, not just rameter to tune. The critical limit is achieved by taking p = 3. However, the details are rather different from the β , while keeping spacelike case. An electric field in open string theory can- → ∞ 1 1 not be taken to be larger than a critical value Ec = 2πl2 1 3 s M = M (Burgess, 1987; Bachas, 1996). As one approaches this eff p eβcoshβ limit, since the ends of the open string carry opposite electric charges, its effective tension goes to zero, and – the mass scale for the membrane stretched spatially in any attempt to reach E > Ec will be screened by string the 1, 2 directions – finite. In this limit Mp/Meff formation in the vacuum. and so the open membranes propagating along the→ M5- ∞ So, one takes the limit E E while keeping the effec- branes will decouple from gravity. → c 2 2 E2 Finally, one can start with the NS fivebrane in either of tive open string tension ls o = ls E2 E2 finite. It turns c − the type II string theories. Dirichlet p-branes can end on out that while one manages to decouple closed string q the NS fivebrane, with all even p in IIa and all odd p in modes, the open string excitations remain in the spec- IIb, leading to open Dp-brane degrees of freedom. Each of trum. The resulting theory, NCOS theory, is apparently these is charged under a specific Ramond-Ramond gauge a true string theory with an infinite number of particle- field, and by taking a near critical electric background for like degrees of freedom. This is consistent both with the one of these fields, one can again reach a decoupling limit earlier arguments and with the general idea that an ac- in which only the corresponding open brane degrees of tion with timelike noncommutativity effectively contains freedom remain, discussed in (Gopakumar et al., 2000b; an infinite number of time derivatives, thereby enhancing Harmark, 2001). the number of degrees of freedom. All of these theories are connected by a web of dualities The NCOS theory contradicts standard arguments analogous to those connecting the conventional decou- from world-sheet duality that open string theory must pled brane theories and the bulk theories which contain contain closed strings. How this works can be seen ex- them. For example, compactifying OM theory on a circle plicitly in the annulus diagram; extra phases present in leads to the NCOS 4 + 1 theory; conversely the strong the non-planar open string diagrams make the would-be coupling limit of this NCOS theory has a geometric de- closed string poles vanish. These effects apparently re- scription (OM theory) just as did the strong coupling solve the unitarity problems of IV.C as well. limit of the IIa string (M theory). Since this string theory decouples from gravity, it pro- vides a system in which the Hagedorn transition of string theory can be analyzed in a clean situation, free from VIII. CONCLUSIONS black hole thermal effects and other complications of gravitating systems (Gubser, Gukov, Klebanov, Ranga- Field theory can be generalized to space-time with non- mani, and Witten, 2000; Barbon and Rabinovici, 2001). commuting coordinates. Much of the formalism is very A similar limit can be taken starting with the M-theory parallel to that of conventional field theory and especially fivebrane (the “M5-brane”). Its world-volume theory is with the large N limit of conventional field theory. Al- also a gauge theory, but now involving a rank two anti- though not proven, it appears that quantum noncommu- symmetric tensor potential. The membrane is allowed to tative field theories under certain restrictions (say with end on an M5-brane and thus parallel M5-branes come spacelike noncommutativity and some supersymmetry) with light degrees of freedom (open membranes) directly are renormalizable and sensible. analogous to the open strings which end on a Dirichlet Their physics is similar enough to conventional field brane. In the limit that the branes coincide, the result- theory to make comparisons possible, and different ing light degrees of freedom are governed by a nontrivial enough to make them interesting. To repeat some of the fixed point theory in 5 + 1 dimensions, usually called the highlights, we found that noncommutative gauge sym- “(2, 0) theory” after its supersymmetry algebra. A simi- metry includes space-time symmetries, that nonsingular lar limit in IIa theory leads to “little string theory,” these soliton solutions exist in higher dimensional scalar field 50 theory and in noncommutative Maxwell theory, that UV from the fact that it emerges from limits of M theory divergences can be transmuted into new IR effects, and and string theory, and it seems clear at this point that that noncommutative gauge theories can have more du- the subject will have lasting importance in this context. alities than their conventional counterparts. So far its most fruitful applications have been to duality, Much of our knowledge of conventional field theory still and to the understanding of solitons and branes in string awaits a noncommutative counterpart. Throughout the theory. It is quite striking how much structure which review many questions were left open, such as the mean- had been considered “essentially stringy” is captured by ing of the IR divergences found in Sec. IV.C, the poten- these much simpler theories. tial nonperturbative role of the solitons and instantons Noncommutativity enters into open string theory es- of Sec. III, the meaning of the high energy behavior dis- sentially because open strings interact by joining at their cussed in Sec. IV.D, and the high temperature behavior. ends, and the choice of one or the other of the two ends One central problem is to properly understand the corresponds formally to acting on the corresponding field renormalization group. Even if one can directly adapt by multiplication on the left or on the right; these are dif- existing RG technology, it seems very likely that theories ferent. This is such a fundamental level that it has long with such a different underlying concept of space and been thought that noncommutativity should be central time will admit other and perhaps more suitable formu- to the subject. So far, the developments we discussed lations of the RG. This might lead to insights into nonlo- look like a very promising start towards realizing this cality of the sort hoped for in the introduction. Questions idea. Progress is also being made on the direct approach, about the existence of quantized noncommutative theo- through string field theory based on noncommutative ge- ries could then be settled by using the RG starting with a ometry, and we believe that many of the ideas we dis- good regulated nonperturbative definition of the theory, cussed will reappear in this context. perhaps that of (Ambjorn et al., 1999) or perhaps along Whether noncommutativity is a central concept in the other lines as discussed in Sec. VI.A. full string or M theory is less clear. Perhaps the best The techniques of exactly solvable field theory, which reason to think this is that it appears so naturally from are so fruitful in two dimensions, await possible noncom- the M(atrix) theory definition, which can define all of M mutative generalization. These might be particularly rel- theory in certain backgrounds. On the other hand, this evant for the quantum Hall application. also points to the weakness of our present understanding: It is not impossible that noncommutative field the- these are very special backgrounds. We do not now have ory has some direct relevance for particle physics phe- formulations of M theory in general backgrounds; this in- nomenology, or possible relevance in the early universe. cludes the backgrounds of primary physical interest with Possible signatures of noncommutativity in QED and four observable dimensions. A related point is that in the standard model are discussed in (Mocioiu, Pospelov, string theory, one thinks of the background as defined and Roiban, 2000; Arfaei and Yavartanoo, 2000; Hewett, within the gravitational or closed string sector, and the Petriello, and Rizzo, 2000; Baek, Ghosh, He, and Hwang, role of noncommutativity in this sector is less clear. 2001; Mathews, 2001; Mazumdar and Sheikh-Jabbari, An important question in noncommutative field the- 2000) and by Carroll, Harvey, Kostelecky, Lane, and ory is to what extent the definitions can be generalized Okamoto (2001), who work with a general extension of to spaces besides Minkowski space and the torus, which the standard model allowing for Lorentz violation (see are not flat. D-brane constructions in other backgrounds (Kostelecky, 2001) and references there) and argue that analogous to what we discussed for flat space seem to atomic clock-comparison experiments lead to a bound lead to theories with finitely many degrees of freedom, as 2 in the QED sector of θ < (10TeV)− . A noncom- in (Alekseev et al., 1999). It might be that noncommu- mutative “brane world”| scenario| is developed in (Pilo tative field theories can arise as large N limits of these and Riotto, 2001), and cosmological applications are dis- models, but at present this is not clear. cussed in (Chu, Greene, and Shiu, 2000; Alexander and Even for group manifolds and homogeneous spaces, Magueijo, 2001). where mathematical definitions exist, the physics of these This motivation as well as the motivation mentioned in theories is not clear, and deserves more study. As we the introduction of modeling position-space uncertainty discussed in section VI, there are many more interest- in quantum gravity might be better served by Lorentz ing noncommutative algebras arising from geometric con- invariant theories, and in pursuing the second of these structions, which would be interesting test cases as well. motivations it has been suggested in (Doplicher, Freden- At the present state of knowledge, it is conceivable hagen, and Roberts, 1994; Doplicher, 2001) that such the- that, contrary to our intuition both from the study ories could be defined by treating the noncommutativity of gravity and perturbative string theory, special back- parameter θ as a dynamical variable. The space-time grounds such as flat space, anti-de Sitter space, orbifolds, stringy uncertainty principle of Yoneya (1987) leads to and perhaps others, which correspond in M theory to related considerations (Yoneya, 2001). simple gauge theories and noncommutative gauge the- While we hope that our discussion has demonstrated ories, play a preferred role in the theory, and that all that noncommutative field theory is a subject of intrinsic others will be derived from these. In this picture, the interest, at present its primary physical application stems gravitational or closed string degrees of freedom would 51 be derived from the gauge theory or open string theory, Akhmedov, E. T., P. DeBoer, and G. W. Semenoff, 2000, as has been argued to happen at substringy distances, in Running couplings and triviality of field theories on non- M(atrix) theory, and in the AdS/CFT correspondence. commutative spaces, eprint eprint hep-th/0010003. If physically realistic backgrounds could be derived this Akhmedov, E. T., P. DeBoer, and G. W. Semenoff, 2001, Non- commutative gross-neveu model at large n, eprint eprint way, then this might be a satisfactory outcome. It would hep-th/0103199. radically change our viewpoint on space-time and might Alekseev, A. Y., A. Recknagel, and V. Schomerus, 1999, Non- predict that many backgrounds which would be accept- commutative world-volume geometries: Branes on su(2) able solutions of gravity are in fact not allowed in M the- and fuzzy spheres, JHEP 09, 023, eprint eprint hep- ory. It is far too early to judge this point however and it th/9908040. seems to us that at present such hopes are founded more Alexander, S. and J. Magueijo, 2001, Non-commutative ge- on our lack of understanding of M theory in general back- ometry as a realization of varying speed of light cosmology, grounds than on anything else. Perhaps noncommutative eprint eprint hep-th/0104093. field theories in more general backgrounds, or in a more Alvarez-Gaume, L., J. L. F. Barbon, and R. Zwicky, 2001, Re- background-independent formulation, will serve as useful marks on time-space noncommutative field theories, eprint eprint hep-th/0103069. analogs to M/string theory for this question as well. Ambjorn, J., Y. M. Makeenko, J. Nishimura, and R. J. Szabo, In any case, our general conclusion has to be that 1999, Finite n matrix models of noncommutative gauge the- the study of noncommutative field theory, as well as the ory, JHEP 11, 029, eprint eprint hep-th/9911041. more mysterious theories which have emerged from the Ambjorn, J., Y. M. Makeenko, J. Nishimura, and R. J. Sz- study of superstring duality (a few of which we men- abo, 2000, Nonperturbative dynamics of noncommutative tioned in Sec. VII.G), has shown that field theory is a gauge theory, Phys. Lett. B480, 399, eprint eprint hep- much broader concept than had been dreamed of even a th/0002158. few years ago. It surely has many more surprises in store Andreev, O. and H. Dorn, 2000a, Diagrams of noncommuta- tive phi**3 theory from string theory, Nucl. Phys. B583, for us, and we hope this review will stimulate the reader 145, eprint eprint hep-th/0003113. to pick up and continue the story. Andreev, O. and H. Dorn, 2000b, On open string sigma-model and noncommutative gauge fields, Phys. Lett. B476, 402, eprint eprint hep-th/9912070. Arcioni, G., J. L. F. Barbon, J. Gomis, and M. A. Vazquez- Acknowledgements Mozo, 2000, On the stringy nature of winding modes in noncommutative thermal field theories, JHEP 06, 038, We would like to thank Alain Connes, Jeff Harvey, Al- eprint eprint hep-th/0004080. bert Schwarz and Nathan Seiberg for comments on the Arcioni, G. and M. A. Vazquez-Mozo, 2000, Thermal effects 01 manuscript, and the ITP, Santa Barbara for hospital- in perturbative noncommutative gauge theories, JHEP , 028, eprint eprint hep-th/9912140. ity. This work was supported in part by DOE grant DE- Ardalan, F. and N. Sadooghi, 2000, Anomaly and nonplanar FG05-90ER40559, and by RFFI grants 01-01-00549 and diagrams in noncommutative gauge theories, eprint eprint 00-15-96557. hep-th/0009233. Arfaei, H. and M. H. Yavartanoo, 2000, Phenomenological consequences of non-commutative qed, eprint eprint hep- th/0010244. References Armoni, A., 2001, Comments on perturbative dynamics of non-commutative yang- mills theory, Nucl. Phys. B593, Ablowitz, M. J. and P. A. Clarkson, 1991, Solitons, nonlin- 229, eprint eprint hep-th/0005208. ear evolution equations and inverse scattering (Cambridge Armoni, A., R. Minasian, and S. Theisen, 2001, On non- Univ. Press), (LMS lecture note series, 149). commutative n = 2 super yang-mills, eprint eprint hep- Abou-Zeid, M. and H. Dorn, 2001a, Comments on the energy- th/0102007. momentum tensor in non-commutative field theories, eprint Asakawa, T. and I. Kishimoto, 2000, Noncommutative gauge eprint hep-th/0104244. theories from deformation quantization, Nucl. Phys. B591, Abou-Zeid, M. and H. Dorn, 2001b, Dynamics of wilson ob- 611, eprint eprint hep-th/0002138. servables in non-commutative gauge theory, Phys. Lett. Atiyah, M., V. Drinfeld, N. Hitchin, and Y. Manin, 1978, B504, 165, eprint eprint hep-th/0009231. Construction of instantons, Phys. Lett. A65, 185. Abouelsaood, A., C. G. Callan, C. R. Nappi, and S. A. Yost, Bachas, C., 1996, D-brane dynamics, Phys. Lett. B374, 37, 1987, Open strings in background gauge fields, Nucl. Phys. eprint hep-th/9511043. B280, 599. Baek, S., D. K. Ghosh, X.-G. He, and W. Y. P. Hwang, Aganagic, M., R. Gopakumar, S. Minwalla, and A. Stro- 2001, Signatures of non-commutative qed at photon col- minger, 2001, Unstable solitons in noncommutative gauge liders, eprint eprint hep-ph/0103068. theory, JHEP 04, 001, eprint eprint hep-th/0009142. Bak, D., 2000, Exact multi-vortex solutions in noncommuta- Aharony, O., 2000, A brief review of ’little string theories’, tive abelian- higgs theory, Phys. Lett. B495, 251, eprint Class. Quant. Grav. 17, 929, eprint hep-th/9911147. hep-th/0008204. Aharony, O., S. S. Gubser, J. Maldacena, H. Ooguri, and Bak, D., K. Lee, and J.-H. Park, 2000, Noncommutative vor- Y. Oz, 2000, Large n field theories, string theory and grav- tex solitons, eprint eprint hep-th/0011099. ity, Phys. Rept. 323, 183, eprint hep-th/9905111. Banks, T., W. Fischler, S. H. Shenker, and L. Susskind, 1997, 52 M theory as a matrix model: A conjecture, Phys. Rev. linear sigma models, eprint eprint hep-th/0102022. D55, 5112, eprint hep-th/9610043. Carroll, S. M., J. A. Harvey, V. A. Kostelecky, C. D. Lane, Barbon, J. L. F. and E. Rabinovici, 2001, On the nature of and T. Okamoto, 2001, Noncommutative field theory and the hagedorn transition in ncos systems, eprint eprint hep- lorentz violation, eprint eprint hep-th/0105082. th/0104169. Cattaneo, A. S. and G. Felder, 2000, A path integral approach Bars, I., H. Kajiura, Y. Matsuo, and T. Takayanagi, 2001a, to the kontsevich quantization formula, Commun. Math. Tachyon condensation on noncommutative torus, Phys. Phys. 212, 591, eprint math.qa/9902090. Rev. D63, 086001, eprint eprint hep-th/0010101. Cattaneo, A. S. and G. Felder, 2001, Poisson sigma models Bars, I. and D. Minic, 2000, Non-commutative geometry on a and deformation quantization, Mod. Phys. Lett. A16, 179, discrete periodic lattice and gauge theory, Phys. Rev. D62, eprint hep-th/0102208. 105018, eprint eprint hep-th/9910091. Chen, G.-H. and Y.-S. Wu, 2001, On critical phenomena in a Bars, I., M. M. Sheikh-Jabbari, and M. A. Vasiliev, 2001b, noncommutative space, eprint eprint hep-th/0103020. Noncommutative o*(n) and usp*(2n) algebras and the Chepelev, I. and R. Roiban, 2000, Renormalization of quan- corresponding gauge field theories, eprint eprint hep- tum field theories on noncommutative r**d. i: Scalars, th/0103209. JHEP 05, 037, eprint eprint hep-th/9911098. Bassetto, A., L. Griguolo, G. Nardelli, and F. Vian, 2001, On Cheung, Y.-K. E. and M. Krogh, 1998, Noncommutative ge- the unitarity of quantum gauge theories on non- commuta- ometry from 0-branes in a background b- field, Nucl. Phys. tive spaces, eprint hep-th/0105257. B528, 185, eprint eprint hep-th/9803031. Baulieu, L., A. Losev, and N. Nekrasov, 2001, Target Chu, C.-S., B. R. Greene, and G. Shiu, 2000, Remarks on space symmetries in topological theories i, eprint hep- inflation and noncommutative geometry, eprint eprint hep- th/0106042. th/0011241. Bayen, F., M. Flato, C. Fronsdal, A. Lichnerowicz, and Coleman, S., 1985, Aspects of Symmetry (Cambridge Univ. D. Sternheimer, 1978, Deformation theory and quantiza- Press). tion. 1. deformations of symplectic structures, Ann. Phys. Connes, A., 1980, C* algebras and differential geometry, 111, 61. Compt. Rend. Acad. Sci. (Ser. I Math. ) A290, 599, eprint Bellisai, D., J. M. Isidro, and M. Matone, 2000, On the struc- eprint hep-th/0101093. ture of noncommutative n = 2 super yang-mills theory, Connes, A., 1982, Noncommutative differential geometry, JHEP 10, 026, eprint eprint hep-th/0009174. parts i, ii, preprints I.H.E.S. M/82/53, M/83/19. Bellissard, J., A. van Elst, and H. Schulz-Baldes, 1993, The Connes, A., 1994, Noncommutative Geometry (Academic non-commutative geometry of the quantum hall effect, Press, San Diego). eprint cond-mat/9301005. Connes, A., 1995, in Quantum symmetries, edited by Bergman, A., O. J. Ganor, and J. L. Karczmarek, 2001, A A. Connes, K. Gawedzki, and J. Zinn-Justin (Elsevier, Am- note on intersecting and fluctuating solitons in 4d noncom- sterdam), proceedings, NATO Advanced Study Institute, mutative field theory, eprint eprint hep-th/0101095. 64th Session, Les Houches. Bergshoeff, E., D. S. Berman, J. P. van der Schaar, and Connes, A., 1996, Gravity coupled with matter and the foun- P. Sundell, 2000, Critical fields on the m5-brane and non- dation of non- commutative geometry, Commun. Math. commutative open strings, Phys. Lett. B492, 193, eprint Phys. 182, 155, eprint eprint hep-th/9603053. eprint hep-th/0006112. Connes, A., 2000a, Noncommutative geometry: Year 2000, Berman, D. S. et al., 2001, Holographic noncommutativity, eprint eprint math.qa/0011193. JHEP 05, 002, eprint eprint hep-th/0011282. Connes, A., 2000b, A short survey of noncommutative geome- Bigatti, D. and L. Susskind, 2000, Magnetic fields, branes try, J. Math. Phys. 41, 3832, eprint eprint hep-th/0003006. and noncommutative geometry, Phys. Rev. D62, 066004, Connes, A., M. R. Douglas, and A. Schwarz, 1998, Noncom- eprint eprint hep-th/9908056. mutative geometry and matrix theory: Compactification Bilal, A., C.-S. Chu, and R. Russo, 2000, String theory and on tori, JHEP 02, 003, eprint eprint hep-th/9711162. noncommutative field theories at one loop, Nucl. Phys. Connes, A. and J. Lott, 1991, Particle models and noncom- B582, 65, eprint eprint hep-th/0003180. mutative geometry (expanded version), Nucl. Phys. Proc. Bogomolny, E. B., 1976, Stability of classical solutions, Sov. Suppl. 18B, 29. J. Nucl. Phys. 24, 449. Connes, A. and M. Rieffel, 1987, Yang-mills for noncommuta- Bonora, L., M. Schnabl, M. M. Sheikh-Jabbari, and tive two-tori, Contemp. Math. Oper. Algebra. Math. Phys. A. Tomasiello, 2000a, Noncommutative so(n) and sp(n) 62, 237. gauge theories, Nucl. Phys. B589, 461, eprint eprint hep- Cornalba, L., 1999, D-brane physics and noncommutative th/0006091. yang-mills theory, eprint eprint hep-th/9909081. Bonora, L., M. Schnabl, and A. Tomasiello, 2000b, A note Cremmer, E., B. Julia, and J. Scherk, 1978, Supergravity on consistent anomalies in noncommutative ym theories, theory in 11 dimensions, Phys. Lett. B76, 409. Phys. Lett. B485, 311, eprint hep-th/0002210. Curtright, T., T. Uematsu, and C. Zachos, 2000, Generating Bordemann, M. and J. Hoppe, 1993, The dynamics of rela- all wigner functions, eprint eprint hep-th/0011137. tivistic membranes. 1. reduction to two-dimensional fluid Dai, J., R. G. Leigh, and J. Polchinski, 1989, New connections dynamics, Phys. Lett. B317, 315, eprint hep-th/9307036. between string theories, Mod. Phys. Lett. A4, 2073. Brace, D. and B. Morariu, 1999, A note on the bps spectrum Danielsson, U. H., A. Guijosa, M. Kruczenski, and B. Sund- of the matrix model, JHEP 02, 004, eprint eprint hep- borg, 2000, D3-brane holography, JHEP 05, 028, eprint th/9810185. eprint hep-th/0004187. Burgess, C. P., 1987, Open string instability in background Das, S. R. and S.-J. Rey, 2000, Open wilson lines in non- electric fields, Nucl. Phys. B294, 427. commutative gauge theory and tomography of holographic Campbell, B. A. and K. Kaminsky, 2001, Noncommutative dual supergravity, Nucl. Phys. B590, 453, eprint eprint 53 hep-th/0008042. space, Phys. Lett. B376, 53. Das, S. R. and S. P. Trivedi, 2001, Supergravity couplings to Fischler, W., J. Gomis, E. Gorbatov, A. Kashani-Poor, S. Pa- noncommutative branes, open wilson lines and generalized ban, and P. Pouliot, 2000a, Evidence for winding states star products, JHEP 02, 046, eprint eprint hep-th/0011131. in noncommutative quantum field theory, JHEP 05, 024, Dasgupta, K., S. Mukhi, and G. Rajesh, 2000, Noncommuta- eprint eprint hep-th/0002067. tive tachyons, JHEP 06, 022, eprint eprint hep-th/0005006. Fischler, W., E. Gorbatov, A. Kashani-Poor, R. McNees, Derrick, G. M., 1964, J. Math. Phys. 5, 1252. S. Paban, and P. Pouliot, 2000b, The interplay between DeWilde, M. and P. Lecomte, 1983, Lett. Math. Phys. 7, 487. theta and t, JHEP 06, 032, eprint eprint hep-th/0003216. DeWitt, B., 1962, in Gravitation, edited by L. Witten, pp. Fradkin, E., 1991, Field Theories of Condensed Matter Sys- 266–381. tems (Addison-Wesley). Dhar, A. and Y. Kitazawa, 2001a, High energy behavior of Fradkin, E. S. and A. A. Tseytlin, 1985, Nonlinear electrody- wilson lines, JHEP 02, 004, eprint eprint hep-th/0012170. namics from quantized strings, Phys. Lett. B163, 123. Dhar, A. and Y. Kitazawa, 2001b, Wilson loops in strongly Furuuchi, K., 2000, Topological charge of u(1) instantons on coupled noncommutative gauge theories, Phys. Rev. D63, noncommutative r**4, eprint eprint hep-th/0010006. 125005, eprint eprint hep-th/0010256. Garousi, M. R., 2000, Non-commutative world-volume in- Diaconescu, D.-E., 1997, D-branes, monopoles and nahm teractions on d-branes and dirac-born-infeld action, Nucl. equations, Nucl. Phys. B503, 220, eprint hep-th/9608163. Phys. B579, 209, eprint eprint hep-th/9909214. Dijkgraaf, R., 1998, in Quantum symmetries, edited by Gerhold, A., J. Grimstrup, H. Grosse, L. Popp, M. Schweda, A. Connes, K. Gawedzki, and J. Zinn-Justin (Elsevier, Am- and R. Wulkenhaar, 2000, The energy-momentum tensor sterdam), proceedings, NATO Advanced Study Institute, on noncommutative spaces: Some pedagogical comments, 64th Session, Les Houches, eprint hep-th/9703136. eprint hep-th/0012112. Dijkgraaf, R., E. Verlinde, and H. Verlinde, 1997, Ma- Girotti, H. O., M. Gomes, V. O. Rivelles, and A. J. da Silva, trix string theory, Nucl. Phys. B500, 43, eprint hep- 2000, A consistent noncommutative field theory: The wess- th/9703030. zumino model, Nucl. Phys. B587, 299, eprint eprint hep- Doplicher, S., 2001, Spacetime and fields, a quantum texture, th/0005272. eprint hep-th/0105251. Girotti, H. O., M. Gomes, V. O. Rivelles, and A. J. Doplicher, S., K. Fredenhagen, and J. E. Roberts, 1994, da Silva, 2001, The noncommutative supersymmetric non- Space-time quantization induced by classical gravity, Phys. linear sigma model, eprint hep-th/0102101. Lett. B331, 39. Girvin, S., 1999, proceedings, Ecole d’Ete Les Houches, July Dorey, N., T. J. Hollowood, and V. V. Khoze, 2000, A brief 1998, eprint cond-mat/9907002. history of the stringy instanton, eprint hep-th/0010015. Girvin, S., A. H. MacDonald, and P. M. Platzman, 1985, Douglas, M. R., 1996, Superstring dualities, dirichlet branes Phys. Rev. Lett. 54, 581. and the small scale structure of space, eprint hep- Girvin, S. and R. Prange, 1987, The Quantum Hall Effect. th/9610041. Gomis, J., K. Kamimura, and J. Llosa, 2001, Hamiltonian Douglas, M. R., 1998, D-branes and discrete torsion, eprint formalism for space-time non-commutative theories, Phys. eprint hep-th/9807235. Rev. D63, 045003, eprint hep-th/0006235. Douglas, M. R., 1999, Two lectures on d-geometry and non- Gomis, J., M. Kleban, T. Mehen, M. Rangamani, and commutative geometry, eprint eprint hep-th/9901146. S. Shenker, 2000, Noncommutative gauge dynamics from Douglas, M. R. and C. Hull, 1998, D-branes and the non- the string worldsheet, JHEP 08, 011, eprint eprint hep- commutative torus, JHEP 02, 008, eprint eprint hep- th/0003215. th/9711165. Gomis, J. and T. Mehen, 2000, Space-time noncommutative Douglas, M. R., D. Kabat, P. Pouliot, and S. H. Shenker, field theories and unitarity, Nucl. Phys. B591, 265, eprint 1997, D-branes and short distances in string theory, Nucl. eprint hep-th/0005129. Phys. B485, 85, eprint hep-th/9608024. Gonzalez-Arroyo, A. and C. P. Korthals-Altes, 1983, Reduced Douglas, M. R. and M. Li, 1996, D-brane realization of n=2 model for large n continuum field theories, Phys. Lett. super yang-mills theory in four dimensions, eprint hep- B131, 396. th/9604041. Gonzalez-Arroyo, A. and M. Okawa, 1983, The twisted Durhuus, B., T. Jonsson, and R. Nest, 2001, Noncommutative eguchi-kawai model: a reduced model for large n lattice scalar solitons: Existence and nonexistence, Phys. Lett. gauge theory, Phys. Rev. D27, 2397. B500, 320, eprint eprint hep-th/0011139. Gopakumar, R., M. Headrick, and M. Spradlin, 2001, On non- Eguchi, T. and R. Nakayama, 1983, Simplification of quench- commutative multi-solitons, eprint eprint hep-th/0103256. ing procedure for large n spin models, Phys. Lett. B122, Gopakumar, R., J. Maldacena, S. Minwalla, and A. Stro- 59. minger, 2000a, S-duality and noncommutative gauge the- Elitzur, S., B. Pioline, and E. Rabinovici, 2000, On the short- ory, JHEP 06, 036, eprint eprint hep-th/0005048. distance structure of irrational non- commutative gauge Gopakumar, R., S. Minwalla, N. Seiberg, and A. Strominger, theories, JHEP 10, 011, eprint eprint hep-th/0009009. 2000b, Om theory in diverse dimensions, JHEP 08, 008, Fairlie, D., 1964, Proc. Camb. Phil. Soc. 60, 581. eprint hep-th/0006062. Fedosov, B., 1996, Deformation Quantization and Index The- Gopakumar, R., S. Minwalla, and A. Strominger, 2000c, Non- ory (Akademie Verlag, Berlin). commutative solitons, JHEP 05, 020, eprint eprint hep- Fedosov, B. V., 1994, J. Diff. Geom. 40, 213. th/0003160. Ferrara, S. and M. A. Lledo, 2000, Some aspects of defor- Gracia-Bondia, J. M. and C. P. Martin, 2000, Chiral gauge mations of supersymmetric field theories, JHEP 05, 008, anomalies on noncommutative r**4, Phys. Lett. B479, 321, eprint eprint hep-th/0002084. eprint eprint hep-th/0002171. Filk, T., 1996, Divergencies in a field theory on quantum Gracia-Bondia, J. M., J. C. Varilly, and H. Figueroa, 2001, El- 54 ements of noncommutative geometry (Birkhaeuser, Boston, commutative gauge theory, JHEP 12, 023, eprint eprint USA). hep-th/0010251. Green, M. B., J. H. Schwarz, and E. Witten, 1987, Superstring Hayakawa, M., 1999, Perturbative analysis on infrared and Theory (Cambridge Univ. Press). ultraviolet aspects of noncommutative qed on r**4, eprint Griguolo, L. and M. Pietroni, 2001a, Hard non-commutative eprint hep-th/9912167. loops resummation, eprint eprint hep-th/0102070. Hellerman, S. and M. Van Raamsdonk, 2001, Quantum hall Griguolo, L. and M. Pietroni, 2001b, Wilsonian renormal- physics equals noncommutative field theory, eprint eprint ization group and the non-commutative ir/uv connection, hep-th/0103179. eprint eprint hep-th/0104217. Hewett, J. L., F. J. Petriello, and T. G. Rizzo, 2000, Signals Gross, D. J., A. Hashimoto, and N. Itzhaki, 2000, Observ- for non-commutative interactions at linear colliders, eprint ables of non-commutative gauge theories, eprint eprint hep- eprint hep-ph/0010354. th/0008075. Ho, P.-M. and Y.-S. Wu, 1997, Noncommutative geom- Gross, D. J. and N. A. Nekrasov, 2000a, Dynamics of strings etry and d-branes, Phys. Lett. B398, 52, eprint hep- in noncommutative gauge theory, JHEP 10, 021, eprint th/9611233. eprint hep-th/0007204. Ho, P.-M. and Y.-S. Wu, 1998, Noncommutative gauge theo- Gross, D. J. and N. A. Nekrasov, 2000b, Monopoles and ries in matrix theory, Phys. Rev. D58, 066003, eprint eprint strings in noncommutative gauge theory, JHEP 07, 034, hep-th/9801147. eprint eprint hep-th/0005204. Hofman, C. and E. Verlinde, 1999, Gauge bundles and born- Gross, D. J. and N. A. Nekrasov, 2001, Solitons in noncom- infeld on the noncommutative torus, Nucl. Phys. B547, mutative gauge theory, JHEP 03, 044, eprint eprint hep- 157, eprint eprint hep-th/9810219. th/0010090. Hollowood, T. J., V. V. Khoze, and G. Travaglini, 2001, Exact Gross, D. J. and W. Taylor, 2001, Split string field theory. i, results in noncommutative n = 2 supersymmetric gauge eprint hep-th/0105059. theories, eprint eprint hep-th/0102045. Grosse, H., C. Klimcik, and P. Presnajder, 1996, Towards ’t Hooft, G., 1974, A planar diagram theory for strong inter- finite quantum field theory in noncommutative geometry, actions, Nucl. Phys. B72, 461. Int. J. Theor. Phys. 35, 231, eprint arXiv:hep-th/9505175. Hoppe, J., 1982, Mit ph.d. thesis. Gubser, S. S., S. Gukov, I. R. Klebanov, M. Ranga- Hoppe, J., 1989, Diffeomorphism groups, quantization and mani, and E. Witten, 2000, The hagedorn transition in su(infinity), Int. J. Mod. Phys. A4, 5235. non-commutative open string theory, eprint eprint hep- Ishibashi, N., S. Iso, H. Kawai, and Y. Kitazawa, 2000, Wilson th/0009140. loops in noncommutative yang-mills, Nucl. Phys. B573, Gubser, S. S. and S. L. Sondhi, 2000, Phase structure of 573, eprint eprint hep-th/9910004. non-commutative scalar field theories, eprint eprint hep- Ishibashi, N., H. Kawai, Y. Kitazawa, and A. Tsuchiya, 1997, th/0006119. A large-n reduced model as superstring, Nucl. Phys. B498, Gurarie, V. and A. Zee, 2000, Quantum hall transition in the 467, eprint hep-th/9612115. classical limit, eprint cond-mat/0008163. Jackson, M. G., 2001, The stability of noncommutative scalar Hadasz, L., U. Lindstrom, M. Rocek, and R. von Unge, solitons, eprint eprint hep-th/0103217. 2001, Noncommutative multisolitons: Moduli spaces, quan- Jatkar, D. P., G. Mandal, and S. R. Wadia, 2000, Nielsen- tization, finite theta effects and stability, eprint hep- olesen vortices in noncommutative abelian higgs model, th/0104017. JHEP 09, 018, eprint eprint hep-th/0007078. Hamanaka, M. and S. Terashima, 2001, On exact noncom- Jurco, B., L. Moller, S. Schraml, P. Schupp, and J. Wess, mutative bps solitons, JHEP 03, 034, eprint eprint hep- 2001, Construction of non-abelian gauge theories on non- th/0010221. commutative spaces, eprint eprint hep-th/0104153. Harmark, T., 2001, Open branes in space-time non- Jurco, B., P. Schupp, and J. Wess, 2000, Noncommutative commutative little string theory, Nucl. Phys. B593, 76, gauge theory for poisson manifolds, Nucl. Phys. B584, 784, eprint hep-th/0007147. eprint eprint hep-th/0005005. Harvey, J. A., 2001a, Komaba lectures on noncommutative Kallin, C. and B. I. Halperin, 1984, Phys. Rev. B 30, 5655. solitons and d-branes, eprint eprint hep-th/0102076. Kapusta, J., 1990, Finite Temperature Field Theory (Cam- Harvey, J. A., 2001b, Topology of the gauge group in non- bridge Univ. Press). commutative gauge theory, eprint hep-th/0105242. Khoze, V. V. and G. Travaglini, 2001, Wilsonian effective ac- Harvey, J. A., P. Kraus, and F. Larsen, 2000a, Exact non- tions and the ir/uv mixing in noncommutative gauge the- commutative solitons, JHEP 12, 024, eprint eprint hep- ories, JHEP 01, 026, eprint eprint hep-th/0011218. th/0010060. Kiem, Y. and S. Lee, 2000, Uv/ir mixing in noncommutative Harvey, J. A., P. Kraus, F. Larsen, and E. J. Martinec, 2000b, field theory via open string loops, Nucl. Phys. B586, 303, D-branes and strings as non-commutative solitons, JHEP eprint eprint hep-th/0003145. 07, 042, eprint eprint hep-th/0005031. Kinar, Y., G. Lifschytz, and J. Sonnenschein, 2001, Uv/ir Harvey, J. A. and G. Moore, 2000, Noncommutative tachyons connection, a matrix perspective, eprint hep-th/0105089. and k-theory, eprint eprint hep-th/0009030. Klauder, J. R. and E. C. G. Sudarshan, 1968, Fundamentals Hashimoto, A. and N. Itzhaki, 1999, Non-commutative yang- of Quantum Optics (Benjamin, New York). mills and the ads/cft correspondence, Phys. Lett. B465, Konechny, A. and A. Schwarz, 1999, Bps states on noncom- 142, eprint eprint hep-th/9907166. mutative tori and duality, Nucl. Phys. B550, 561, eprint Hashimoto, A. and N. Itzhaki, 2000, Traveling faster than the eprint hep-th/9811159. speed of light in non-commutative geometry, eprint eprint Konechny, A. and A. Schwarz, 2000a, Compactification of hep-th/0012093. m(atrix) theory on noncommutative toroidal orbifolds, Hashimoto, K., 2000, Fluxons and exact bps solitons in non- Nucl. Phys. B591, 667, eprint eprint hep-th/9912185. 55 Konechny, A. and A. Schwarz, 2000b, Introduction to eprint hep-ph/0012363. m(atrix) theory and noncommutative geometry, eprint Mehen, T. and M. B. Wise, 2000, Generalized *-products, wil- eprint hep-th/0012145. son lines and the solution of the seiberg-witten equations, Kontsevich, M., 1997, Deformation quantization of poisson JHEP 12, 008, eprint eprint hep-th/0010204. manifolds, i, eprint q-alg/9709040. Minwalla, S., M. Van Raamsdonk, and N. Seiberg, 2000, Non- Kostelecky, V. A., 2001, Topics in lorentz and cpt violation, commutative perturbative dynamics, JHEP 02, 020, eprint eprint hep-ph/0104227. eprint hep-th/9912072. Kuiper, N. H., 1965, Topology 3, 19. Mocioiu, I., M. Pospelov, and R. Roiban, 2000, Low-energy Landau, L. D., E. M. Lifshitz, and A. Pitaevskii, 1960, Field limits on the antisymmetric tensor field background on Theory of Continuous Media (Butterworth Heinemann, the brane and on the non-commutative scale, Phys. Lett. Oxford). B489, 390, eprint eprint hep-ph/0005191. Landsteiner, K., E. Lopez, and M. H. G. Tytgat, 2001, Insta- Montonen, C. and D. Olive, 1977, Magnetic monopoles as bility of non-commutative sym theories at finite tempera- gauge particles?, Phys. Lett. B72, 117. ture, eprint eprint hep-th/0104133. Moreno, E. F. and F. A. Schaposnik, 2000, The wess-zumino- Laughlin, R. B., 1985, Physica 126B, 254. witten term in non-commutative two- dimensional fermion Lee, B.-H., K. Lee, and H. S. Yang, 2001, The cp(n) model models, JHEP 03, 032, eprint eprint hep-th/0002236. on noncommutative plane, Phys. Lett. B498, 277, eprint Moreno, E. F. and F. A. Schaposnik, 2001, Wess-zumino- eprint hep-th/0007140. witten and fermion models in noncommutative space, Nucl. Li, M., 2000, Quantum corrections to noncommutative soli- Phys. B596, 439, eprint eprint hep-th/0008118. tons, eprint hep-th/0011170. Motl, L., 1997, Proposals on nonperturbative superstring in- Li, M. and Y.-S. Wu, 2000, Holography and noncommutative teractions, eprint hep-th/9701025. yang-mills, Phys. Rev. Lett. 84, 2084, eprint eprint hep- Mukhi, S. and N. V. Suryanarayana, 2001, Gauge-invariant th/9909085. couplings of noncommutative branes to ramond-ramond Lindstrom, U., M. Rocek, and R. von Unge, 2000, Non- backgrounds, eprint eprint hep-th/0104045. commutative soliton scattering, JHEP 12, 004, eprint Myers, R. C., 1999, Dielectric-branes, JHEP 12, 022, eprint eprint hep-th/0008108. eprint hep-th/9910053. Liu, H., 2000, *-trek ii: *n operations, open wilson lines and Myers, R. C., 2001, Nonabelian d-branes and noncommuta- the seiberg-witten map, eprint eprint hep-th/0011125. tive geometry, eprint hep-th/0106178. Liu, H. and J. Michelson, 2000, *-trek: The one loop n = 4 Nahm, W., 1980, A simple formalism for the bps monopole, noncommutative sym action, eprint eprint hep-th/0008205. Phys. Lett. B90, 413. Liu, H. and J. Michelson, 2001a, Ramond-ramond couplings Nekrasov, N., 2001, Lectures on d-branes and noncommuta- of noncommutative d-branes, eprint eprint hep-th/0104139. tive gauge theories. Liu, H. and J. Michelson, 2001b, Supergravity couplings of Nekrasov, N. and A. Schwarz, 1998, Instantons on noncom- noncommutative d-branes, eprint eprint hep-th/0101016. mutative r**4 and (2,0) superconformal six dimensional Lozano, G. S., E. F. Moreno, and F. A. Schaposnik, 2001, theory, Commun. Math. Phys. 198, 689, eprint eprint hep- Nielsen-olesen vortices in noncommutative space, Phys. th/9802068. Lett. B504, 117, eprint eprint hep-th/0011205. Nekrasov, N. A., 2000, Trieste lectures on solitons in noncom- Lugo, A. R., 2000, Correlation functions in the non- mutative gauge theories, eprint eprint hep-th/0011095. commutative wess-zumino- witten model, eprint eprint hep- Nielsen, H. B. and P. Olesen, 1973, Nucl. Phys. B61, 45. th/0012268. Okawa, Y. and H. Ooguri, 2001a, Energy-momentum tensors Madore, J., 1992, The fuzzy sphere, Class. Quant. Grav. 9, in matrix theory and in noncommutative gauge theories, 69. eprint eprint hep-th/0103124. Madore, J., 1999, An Introduction to Noncommutative Ge- Okawa, Y. and H. Ooguri, 2001b, An exact solution to seiberg- ometry and its Physical Applications (Cambridge Univ. witten equation of noncommutative gauge theory, eprint Press), (LMS lecture note Series 257, 2nd ed.). eprint hep-th/0104036. Maldacena, J., 1998, The large n limit of superconformal field Okawa, Y. and H. Ooguri, 2001c, How noncommutative gauge theories and supergravity, Adv. Theor. Math. Phys. 2, 231, theories couple to gravity, Nucl. Phys. B599, 55, eprint eprint hep-th/9711200. eprint hep-th/0012218. Maldacena, J. M. and J. G. Russo, 1999, Large n limit of non- Olshanetsky, M. and A. Perelomov, 1976, Inv. Math. 31, 93. commutative gauge theories, JHEP 09, 025, eprint eprint Palais, R. S., 1965, Topology 3, 271. hep-th/9908134. Pernici, M., A. Santambrogio, and D. Zanon, 2001, The one- Martin, C. P., 2001, Chiral gauge anomalies on noncommu- loop effective action of noncommutative n = 4 super yang- tative minkowski space- time, Mod. Phys. Lett. A16, 311, mills is gauge invariant, Phys. Lett. B504, 131, eprint eprint eprint hep-th/0102066. eprint hep-th/0011140. Martinec, E. J. and G. Moore, 2001, Noncommutative solitons Petriello, F. J., 2001, The higgs mechanism in non- on orbifolds, eprint eprint hep-th/0101199. commutative gauge theories, Nucl. Phys. B601, 169, eprint Mathews, P., 2001, Compton scattering in noncommutative eprint hep-th/0101109. space-time at the nlc, Phys. Rev. D63, 075007, eprint Pilo, L. and A. Riotto, 2001, The non-commutative brane eprint hep-ph/0011332. world, JHEP 03, 015, eprint eprint hep-ph/0012174. Matusis, A., L. Susskind, and N. Toumbas, 2000, The ir/uv Polchinski, J., 1984, Renormalization and effective la- connection in the non-commutative gauge theories, JHEP grangians, Nucl. Phys. B231, 269. 12, 002, eprint eprint hep-th/0002075. Polchinski, J., 1995, Dirichlet-branes and ramond-ramond Mazumdar, A. and M. M. Sheikh-Jabbari, 2000, Noncom- charges, Phys. Rev. Lett. 75, 4724, eprint hep-th/9510017. mutativity in space and primordial magnetic field, eprint Polchinski, J., 1996a, String duality: A colloquium, Rev. 56 Mod. Phys. 68, 1245, eprint hep-th/9607050. in background electric field, space/time noncommutativity Polchinski, J., 1996b, Tasi lectures on d-branes, eprint hep- and a new noncritical string theory, JHEP 06, 021, eprint th/9611050. eprint hep-th/0005040. Polchinski, J., 1998, Superstring Theory (Cambridge Univ. Seiberg, N. and E. Witten, 1994, Electric - magnetic dual- Press). ity, monopole condensation, and confinement in n=2 super- Polyakov, A. M., 1987, Gauge Fields and Strings (Harwood, symmetric yang-mills theory, Nucl. Phys. B426, 19, eprint Chur, Switzerland). hep-th/9407087. Polychronakos, A. P., 1991, Integrable systems from gauged Seiberg, N. and E. Witten, 1999, String theory and non- matrix models, Phys. Lett. B266, 29. commutative geometry, JHEP 09, 032, eprint eprint hep- Polychronakos, A. P., 2000, Flux tube solutions in noncom- th/9908142. mutative gauge theories, Phys. Lett. B495, 407, eprint Sen, A., 1998, Tachyon condensation on the brane antibrane eprint hep-th/0007043. system, JHEP 9808, 012, eprint hep-th/9805170. Polychronakos, A. P., 2001, Quantum hall states as ma- Shatashvili, S. L., 2001, On field theory of open strings, trix chern-simons theory, JHEP 04, 011, eprint hep- tachyon condensation and closed strings, eprint hep- th/0103013. th/0105076. Rajaraman, A. and M. Rozali, 2000, Noncommutative gauge Sheikh-Jabbari, M. M., 1999, Open strings in a b-field back- theory, divergences and closed strings, JHEP 04, 033, ground as electric dipoles, Phys. Lett. B455, 129, eprint eprint eprint hep-th/0003227. eprint hep-th/9901080. Rastelli, L., A. Sen, and B. Zwiebach, 2001, Half strings, pro- Sinova, J., V. Meden, and S. Girvin, 2000, Phys. Rev. B62, jectors, and multiple d-branes in vacuum string field theory, 2008. eprint hep-th/0105058. Snyder, H. S., 1947, Quantized space-time, Phys. Rev. 71, 38. Rieffel, M., 1981, Pacific J. Math. 93, 415. Sternheimer, D., 1998, in Particles, fields, and gravitation Rieffel, M. A., 1988, Projective modules over higher- (Lodz, 1998) (Amer. Inst. Phys., Woodbury, NY), pp. 107– dimensional noncommutative tori, Canadian J. Math. 40, 145, aIP Conf. Proc., 453, eprint math.QA/9809056. 257. Susskind, L., 2001, The quantum hall fluid and non- Rieffel, M. A., 1998, Questions on quantization, Contemp. commutative chern simons theory, eprint eprint hep- Math. 228, 315, eprint quant-ph/9712009. th/0101029. Rieffel, M. A. and A. Schwarz, 1998, Morita equiva- Taylor, W., 1997, D-brane field theory on compact spaces, lence of multidimensional noncommutative tori, eprint Phys. Lett. B394, 283, eprint hep-th/9611042. math.QA/9803057. Taylor, W., 2001, M(atrix) theory: Matrix quantum mechan- Rozali, M. and M. Van Raamsdonk, 2000, Gauge invariant ics as a fundamental theory, eprint eprint hep-th/0101126. correlators in non-commutative gauge theory, eprint eprint Tseng, L.-S., 2000, Noncommutative solitons and intersecting hep-th/0012065. d-branes, eprint eprint hep-th/0101125. Ruiz, F. R., 2001, Gauge-fixing independence of ir divergences Vafa, C., 1986, Modular invariance and discrete torsion on in non- commutative u(1), perturbative tachyonic instabil- orbifolds, Nucl. Phys. B273, 592. ities and supersymmetry, Phys. Lett. B502, 274, eprint Van Raamsdonk, M. and N. Seiberg, 2000, Comments on non- hep-th/0012171. commutative perturbative dynamics, JHEP 03, 035, eprint Russo, J. G. and M. M. Sheikh-Jabbari, 2001, Strong cou- eprint hep-th/0002186. pling effects in non-commutative spaces from om theory Varilly, J. C., 1997, An introduction to noncommutative ge- and supergravity, Nucl. Phys. B600, 62, eprint eprint hep- ometry, eprint physics/9709045. th/0009141. Varilly, J. C. and J. M. Gracia-Bondia, 1999, On the ultra- Sarkar, S. and B. Sathiapalan, 2001, Comments on the renor- violet behaviour of quantum fields over noncommutative malizability of the broken phase in noncommutative scalar manifolds, Int. J. Mod. Phys. A14, 1305, eprint eprint hep- field theory, eprint eprint hep-th/0104106. th/9804001. Schafer, T. and E. V. Shuryak, 1998, Instantons in qcd, Rev. Verlinde, H., 2001, talk at Davidfest, ITP Santa Barbara, Feb Mod. Phys. 70, 323, eprint hep-ph/9610451. 2001. Schnabl, M., 2000, String field theory at large b-field and Wilczek, F., 1982, Magnetic flux, angular momentum, and noncommutative geometry, JHEP 11, 031, eprint eprint statistics, Phys. Rev. Lett. 48, 1144. hep-th/0010034. de Wit, B., J. Hoppe, and H. Nicolai, 1988, On the quantum Schomerus, V., 1999, D-branes and deformation quantization, mechanics of supermembranes, Nucl. Phys. B305, 545. JHEP 06, 030, eprint eprint hep-th/9903205. Witten, E., 1986, Noncommutative geometry and string field Schwarz, A., 1998, Morita equivalence and duality, Nucl. theory, Nucl. Phys. B268, 253. Phys. B534, 720, eprint eprint hep-th/9805034. Witten, E., 1988, Topological quantum field theory, Commun. Schwarz, A., 2001a, Noncommutative instantons: A new ap- Math. Phys. 117, 353. proach, eprint eprint hep-th/0102182. Witten, E., 1996, Small instantons in string theory, Nucl. Schwarz, J. H., 2001b, Comments on born-infeld theory, Phys. B460, 541, eprint hep-th/9511030. eprint hep-th/0103165. Witten, E., 2000, Noncommutative tachyons and string field Seiberg, N., 2000, A note on background independence in theory, eprint eprint hep-th/0006071. noncommutative gauge theories, matrix model and tachyon Yoneya, T., 1987, in Wandering in the Fields, edited by condensation, JHEP 09, 003, eprint eprint hep-th/0008013. K. Kawarabayashi and A. Ukawa (World Scientific), p. 419. Seiberg, N., L. Susskind, and N. Toumbas, 2000a, Space/time Yoneya, T., 2001, Space-time uncertainty and noncommuta- non-commutativity and causality, JHEP 06, 044, eprint tivity in string theory, Int. J. Mod. Phys. A16, 945, eprint eprint hep-th/0005015. hep-th/0010172. Seiberg, N., L. Susskind, and N. Toumbas, 2000b, Strings Zanon, D., 2001a, Noncommutative n = 1,2 super u(n) yang- 57 mills: Uv/ir mixing and effective action results at one loop, th/0009196. Phys. Lett. B502, 265, eprint eprint hep-th/0012009. Zhou, C.-G., 2000, Noncommutative scalar solitons at finite Zanon, D., 2001b, Noncommutative perturbation in su- theta, eprint eprint hep-th/0007255. perspace, Phys. Lett. B504, 101, eprint eprint hep-