JHEP05(2001)047 May 24, 2001 May 24, 2001 Revised: Accepted: April 2, 2001, Received: D-, Differential and Algebraic Geometry. We analyze the role of RR fluxes in orientifold backgrounds from the [email protected] [email protected] [email protected] HYPER VERSION CIT/USC Center for ,Los Angeles Univ. CA, of USA SouthernE-mail: California School of Natural Sciences, InstituteEinstein for Drive, Advanced Princeton Study NJE-mail: 08540, USA California Institute of Technology, PasadenaCIT/USC CA Center 91125, for Theoretical USA Physics Univ. and of Southern California,E-mail: Los Angeles CA, USA Abstract: point of view ofing K-theory, these and fluxes demonstrate incertain some K-theory physical fractional rather implications shifts than ofnaturally in cohomology. describ- explained RR In in charge particular, K-theory. quantizationwhich we are We due considered show also to distinct that show in discretethe that the K-theory RR some cohomology description, fluxes orientifold classification, while backgrounds, become are others become equivalent in unphysical. Keywords: Shigeki Sugimoto Eric Gimon Oren Bergman Orientifolds, RR torsion, and K-theory JHEP05(2001)047 )27 0 , 6 S ( KH 1 )and 0 , 6 S ( KR using a spectral sequence 25 K A.3 Extending theA.4 AHSS to Illustrative K-theories example: with freely acting involutions 26 2.1 Basics2.2 Cohomology torsion2.3 variants realization of torsion3.1 Real K-theory3.2 RR3.3 fields in orientifold backgrounds Comparing with cohomology4.1 Charge quantization4.2 Equivalent orientifolds4.3 Unphysical orientifolds5.1 K-theory5.2 Anomalies5.3 4 Components of type-I vacuua 6 9 A.1 Defining theA.2 associated 10 graded Computing complex 3 Gr 8 12 15 17 21 24 17 19 1. Introduction Orientifold planes are intriguing objectsbatively, in to many be ways. void They of appear, dynamics,the at and other least in pertur- hand this respect orientifoldwith they planes the are are similar same to also dimensionalities, . carry similar RR On to charge, D-branes, and in have a that non-vanishing they tension. appear 3. Orientifolds in K-theory4. Applications5. The orientifold 6-plane6. Conclusions and open questionsA. AHSS in complex and 8 real K-theory 12 17 23 24 Contents 1. Introduction2. Review of orientifolds 3 1 JHEP05(2001)047 ,as- p 2charge / 1 ± ,therelevant 2 +1 / p 4 are fractionally C ), respectively. The +1 p ≤ 5 5 ) shifts the charge by 5 5 − p 8 − − − − p p p p p − P 8 2 2 R P − +2 − +2 charge ( variant, and yet another to is not an allowed value for p R ( − 5 − p 6 − − -plane in the reduced space of p 6 f H O4 p 2 H ± . The generator of this has = 6, both of these cohomologies are Z ) ) , and have traditionally been classi- p +1) p n n )and ) plane. So in the first case it seems to p − n n 6 − )= + gauge group 8 )-brane, and can be expressed in terms of p G p p 2 − P p 8 -plane is charged under R − SO(2 USp(2 SO(2 USp(2 D P p ( 3 R ( 5), depending on the values of the discrete fluxes H p -plane charge )-form p − p 8 ≤ p . One could imagine that the latter is due to the term − H variant + + − + − p p p p p f (for O4 f f O O O O O Orientifold plane variants and their charges. p -branes). This appears to be at odds with Dirac’s quanti- ) p p 2) / − 6 2) 1 0) , , 2 there are actually additional variants, which are associated with / 2 2 0) 1 plane, but not for the O Table 1: / , / , 2chargeof -flux [2]. H, G − / p p< (1 ( (0 (1 (0 p is the “sphere” that surrounds the O ,andtheRR(6 − 2 p H plane is consistent with an anomalous phase in the D2-brane path integral − G in the D2-brane action, but there is no obvious candidate for the former -brane charge, so the O 8 ± p 1 P C R ∧ 4. In addition, the presence of RR torsion in There are actually a variety of orientifold planes for each dimensionality In some cases the violation can be attributed to an in the path integral One immediately observes that orientifold planes with , so the fluxes are torsion classes. This means that there are four variants of the B ≤ 2 of the O4 charged (relative to D zation condition, suitablyor generalized equivalently to with higher thehomology. rank idea that anti-symmetric In RR tensor charges particular, fields, and since fields the take values O in integral co- a discrete spacetime fields [3], but this is not generally the case. For example, the Z the orientifold projection. Except for the case fied by the (integral) cohomologies 1/2 for the O explain the +1 cohomology group is given by space contribution. For other orientifoldanomaly planes contribution of one [3]. does not even have an analog of the R However the latter twosome differ cases from the their tension of D-brane an(and values orientifold tension) plane in is is two negative, fractional. important and in ways. some cases In its charge sociated with the possibilitytensor of fields turning [1, on 2]. discretethree-form fluxes These of fluxes certain come anti-symmetric in two varieties corresponding to the NSNS orientifold plane for each (table 1). For lead to a further violation of Dirac’s condition. for the dual magnetic object, i.e. a D(6 p coming from the fermionto determinant the [3]. phase in However order one to needs explain another the contribution 0 charge of the a unit D JHEP05(2001)047 -flux. For p − 8 .Thusweare G 2 G -brane is characterized not only by its p alters the quantization of p 3 − 6 G and the half-integral part of 0 G , but also by a gauge bundle, which in turn determines +1 p C ) is integral rather than discrete, but there is an analogous 0 G The rest of the paper is organized as follows. In section 2 we review what is In this paper we investigate some of the consequences of describing RR fluxes All this indicates that integral cohomology is not the appropriate mathematical = 6 the former ( able to compute thethe discrete shift RR in flux. thethe quantization absolute We condition charges are ofcohomological however. not the one able In RR to we comparing charge establishwhile also the due others the discover to K-theory are quantization that obstructed. condition classificationsome some for of with discrete This the the lower fluxes has orientifold interesting are planes. physical actually implications trivial, concerning known about orientifold planes, andcohomology. their In classification section in 3 terms wedescribing describe of RR fields discrete the in necessary fluxes orientifold backgrounds, K-theory in also and machinery compute explain required the how relevant for to groups. comparesion We the of K-theory the results Atiyah-Hirzebruch spectral within sequence cohomology the (AHSS), using appendix. which a is In real furtherK-theory section ver- elaborated and 4 we cohomology discuss in threeRR orientifolds. implications charge quantization, The of the first the second difference concernstifold is between variants the the which identification question are in of distinct K-theory intain proper of cohomology, orientifold certain and variants as orien- the unphysical. third Section isthe 5 the orientifold is 6-plane, dismissal devoted which of to in cer- a many separate ways discussion was of the original motivation for this work. p correlation between the parity of 2. Review of orientifolds 2.1 Basics Orientifolds are defined bycludes gauging a a reversal of discrete the symmetry world-sheet in orientation string (denoted Ω). theory which The loci in- of points its charges under lowerfields rank RR take values fields. in It K-theorycreated has as by D-brane recently well sources, [7, been since 8]. suggestedas the the that This field source at the is which infinity intuitively RR creates should obvious it.RR belong for fluxes, But to RR the are the idea fields same classified is class in that all K-theory. RR fields, includingin source-free K-theory rather thanexplanation integral of the cohomology. observed fractional In charges;in particular, K-theory such correlates this a different RR way, provides that fields a the discrete partial flux of framework for RRproperly fields. RR charges, We are take[4, 5, values familiar 6]. in with This K-theorycharge the follows under rather from idea the than RR the that field in fact D-branes, that integral a or cohomology D more JHEP05(2001)047 2 + ). P p n R (2.1) (2.3) (2.4) (2.2) =+1. )cases, R 3mod4 and O n F , + − L 6) is to note p , F . =2 ) ≤ 1) p J p − projection ( ≡ (hence the sign in 2 2 for ( Z +  L Ω) p 2 F n ) and U Sp(2 I × n 1) 0 n . I 3mod4 1mod4 − i =( , , for O , 2 5 ) n 1 2 − =2 =0 L p × F n p p 0 I -planes (as compared with those 1) i p ), and in the second case U Sp(2 − L − +3mod4 n +1mod4 F  for the SO(2 p p Ω( -plane. A non-trivial holonomy con- 1) 2 p + ,and+2 I = = − p or − ( 1 0 0 p -brane invariant, and in fact preserves the p p 4 n  p 2 · =0 or × n Ω and O 2 π for O 0 0 B p I 2 p p − 5 − 2 9 C B C = p − P I p + R obius strip amplitude with the cylinder amplitude, / 2 Z p amplitude, and therefore exchanges the O − − 2 9 = −→ −→ − −→ − P R b 0 0 field [1], R p p B 1 on fermions, but ( × M, M C C B T − ,p -branes (and their images), there are two possible actions on ) in type-II is given by the obius strip amplitudes, and therefore also the vacuum 1 which surrounds the O λ p p 1 = R p D − 2 − is odd under the projection, which maps opposite points in the 8 n 1inthe M Ω) P − B 2 R : I → = p λ -plane (O ⊂ O p πib 2 -branes) is given by 2 Since is even in IIA and odd in IIB. The presence of ( P e p 1 R p If we include It isn’t obvious at this stage which variant should be identified with trivial holonomy, and which -planes. These are denoted O 1 p the open string Chan-Paton factors [9], where In the first case the resulting gauge group is SO(2 The corresponding M¨ where In particular, the projection leavessame the 16 D . can be understood from theFor requirement example, that the generator ( square to 1 on fermions. The action of the orientifold on the antisymmetric tensor fields is given by the notation). 2.2 Cohomology torsion variants Another way to see that therethat are (at the least) orientifold two variants background (atto least admits the for a holonomy of discrete the degree of freedom corresponding tributes planes. amplitudes, have the oppositeO sign, so we see that there are actually two types of which are fixed underorientifold this symmetry are called orientifold planes. In particular, an respectively. Comparing theone M¨ finds that the RRof charge the and D tension of the O with non-trivial holonomy. This will be clarified in K-theory. JHEP05(2001)047 . b and ] U(1) (2.7) (2.6) (2.5) , but plane [1]. If − + .These − − 2 Z f O3 , corresponds + is related to an torsion, both of . 3 f + O3 2 ] U(1)), where G , Z p − = 8 plane as an O3 and , rather than P ) − Z R , , ) theory corresponding to H 4 5 ( ] 2 U(1)) are in turn classified , , n 2 2 1 f O3 , 2 3 Z )= p , , H − ) theory is self-dual. Z 8 , )theoryofO3 n 5 the relevant holonomy is is interchanged with p )= P =0 =1 n -planes, and is therefore similarly − e Z + R ≤ 8 − p p , ( p p P 2 p f − O R 8 H ( 0 P . Unlike the previous case, the charges of R H ) for ) for ± ( =0 or 5 p e 3 Z Z ∈ p , , f O ] p p H − π 0 5 . S-duality in type-IIB string theory exchanges − − 2 2 8 8 ∈ -field, G C -plane with p ] Z P P p p − 6 R R − . This exchanges the orthogonal gauge group with 5 ( ( dB G p p P − )= R − − e Z 6 6 Z f , are invariant, and O3 O3 is denoted ]=[ 5 5 (see table 1). H H c = + ). The theory is perturbatively the same as for O3 P H ( [ c n ≤ bundle. The choice of torsion coincides with the holonomy R f ( O3 p ∈ 3 Z , the holonomy is an element of ] p p H − − 8 and 6 P torsion on − G [ R = 6 is special, since [ H (see also [17]). p − = 4 supersymmetric . The SO(2 ,soO3 3 f O6 is the twisted N G + 1) theory [10]. We can therefore interpret the e Z n is invariant under this duality, but the U Sp(2 . However, we shall see in section 5 that there is another variant, which has the The case There are analogous strong coupling arguments for type-IIA O2 planes [11] and RR fields can also have a discrete holonomy, which leads to additional variants + − and O3 with an additional fractional (1/2) D3-brane. The fourth variant, which take values in The choice of a non-trivial is the twisted U(1) bundle. The elements of and is identified with a torsion the BPS dyon spectrum is different [2]. This U Sp(2 O4 planes [12]–[14] usingpattern persists M-theory. for all However, it is clear from T-duality that this O6 we include D3-branes, string S-duality becomesvolume Olive-Montonen duality in the world- these variants cannot bebackground. computed in Instead, perturbation one theory, mustexample, since appeal the they to O3-plane involve arguments has a four involving RR variants strong corresponding coupling. to For topologically by the characteristic class classes correspond to themassive allowed type-IIA (integral) values [15,distinct of 16]. O6-planes, the but It cosmological rather does constantsit. different in not One IIA might appear, backgrounds therefore at in conclude that this which the an stage, O6-plane O6-plane to only can has label two variants, O6 where of the orientifold plane. For the O a symplectic one, U Sp(2 to turning on H SO(2 denoted double cover of properties (namely charge and tension) of the JHEP05(2001)047 . p 1, Z , and (2.8) (2.9) ,and =0 inverts (2.10) (2.11) } H )= p 2 6 Z Z , ,x 6 +2)-brane 1 p − P p In both cases R ( 2 0 H ). We denote the ,wherethe ,...,x Z 2 1 , )inthecaseof Z , and therefore forms 0 x 8 / c } { P m 1 -plane. The NS5-brane field, or a D( + R − l p . ( p 2 R H 2 1 2 1 H ≡ field (figure 1). = p + 2)-brane interchanges the O l,m p ,...,x − 1 R − p π 6 5 x 2 )and 2 1 G { C e Z p 1 2 , = − 7 5 p = P P − π R 6 R =0 or π 2 Z 6 ( H 2 G . 1 p 1 1 for the case of the 1 − , π H − 3 1 p, l 2 } S =∆ − P C 5 6 Z S R p π Z B − 2 ∼ 1 c O0. When both RR torsions are turned on we get planes, and the D( 2 P 0 P R l, for the case of the background, which takes values in ,...,x + R S Z 1 } 0 5) situated along the directions p Z 9 -plane. The brane is “fractional” in the sense that it is x = G p { ∆ ≤ 0 ,x c 8 Now remove two small disks at the intersection of the sphere c p O1 and c f O0. The charges of these variants were not known. Using K- 3 and O ,x 7 − p ,x 1 − -plane ( p c f O1 and p to denote the sphere in the reduced space l,m +2)-brane. S = 0 and 1 there is another possible holonomy given by p ,...,x 1 p planes. x p { are trivial (as differential forms) at infinity, but the above result shows that f O For p A similar construction can be made forWe the additional use RR torsion ( 2 3 − 6 of the coordinates. In particular more variants, theory, we will be able tovariants determine are their removed charges. when We will oneone also interprets could see RR that also fields some consider of in these a K-theory. For the O2-plane, This is related to the torsion groups for the NS5-brane, and corresponding variants along the brane intersects the orientifold plane along interchanges the O there is relative torsion between the two pieces of the O a “domain wall” onits the own O image underorientifold the plane. orientifold It projection, is therefore and a can source therefore of 1/2 not unit move of off flux the in the reduced space, i.e. for the D( either an NS5-brane along and with the orientifold plane, and apply Stoke’s law to get for the difference in the holonomy between the two sides. The field strengths The situation is similarsection to 5. the O6-plane case above,2.3 and Brane will realization also of be torsion addressedPure torsion in fields can be realizedConsider physically an by simple O brane constructions [18, 19, 2]. using an intersection with a D6-brane and D7-brane, respectively. G l JHEP05(2001)047 . . ± is 2 at p +1 P 2 p f O R w +2 C p (2.12) ∧ R F is integral p R 0 p 5-p ~ ~ O O G RP is created by a 6-p,1 0 S (p+2) G D torsion. p − 6 b (p+2) G ) D b , + 2)-brane satisfies [20, 17] p Z 5-p p O torsion, ( RP , and “opens up” into a flat field also has 1/2 unit of flux on mod p H O ) is special, in that B b ) a →∞ 7 6 1 2 x could only change by an integral amount due = at + + p π p 2 2 F 2 p ,sincethe O P ~ 2 O RP +1 R P p R 3,1 × C S Z p ∧ is not a homology cycle in the bulk, so the wrapped part plane would shift by 1/2, due to the coupling R is non-orientable, so the question remains open. 2 B − (p+2) 2 P p D R P Brane realization of ( a R Deforming the intersecting brane to a wrapped brane. R NS 5 2 Figure 1: p Figure 2: RP - O = 6 the D-brane picture (figure 1 p p + 2)-brane wraps O p We can try to use this picture to understand the relative charges of the For two coincident branes the flux is 1, but they areFor now free to move off the = 0 (figure 2). The 6 rather than torsion. the additional coupling D8-brane, which fills the entiremove transverse off space. the orientifold The plane, D8-branes and can this therefore is never consistent with the fact that On the other hand the charge of O This would therefore explainresult the similar observed to relative (2.12) charges was of obtained these in two [21] variants. for A orientable for which x of the brane shrinks toassume zero that size, the giving world-volume the gauge original field intersection on configuration. the If D( we non-trivial. However planes. Consider a “deformation” ofthe the D( intersecting D-brane configuration, in which then the charge of the O orientifold plane (in oppositetrivial, directions). and the The orientifold torsion plane part remains unchanged. of the field is therefore JHEP05(2001)047 ) ) X X ( ( (3.1) (3.4) (3.3) 1 ), and − KR X K .These ( ± KH are related by KH One way to think KH 4 )and projection on closed X 2 and ( ) in IIA, and by Z ± X ( KR , KR ) (3.2) . K ) , ) depends only on the difference X ) p,q ( , and similarly for X X ( R p,q ( +1 p,q ), respectively. ,q ,q +4 × R n X +1 +8 KR ( − X p p 1. In K-theoretic terms, one is given ( K − ). In addition KH KR KR 8 KR X . Physically this corresponds to the inclusion ( q F )= )and )= )= − )= p X X X X ( ( X = 3 mod 4 in (2.1)), which exchanges the branes ( ( 1 ( n and p coordinates in − , and an antilinear involution on the bundles that − KR p,q p,q p,q E p X K KR )= KR KR KR X ( (e.g. for p,q L . The first implies that ), respectively [7]. This complements the fact that RR charges, F D9 system in IIB, and non-BPS D9-branes in IIA. ), or more precisely by the kernel of the natural maps from these groups to KR X 1) ( X KH . This defines the real and quaternionic K-theory groups 1 ( − on the space τ − τ cpct K K ), depending on whether the involution on the bundles squares to 1 or X ( is non-compact, then D-branes are actually classified by K-theory with compact support, )and )and X X X ( ,sowedenote ( KH q If 1 4 K − 1 [22]. Higher groups can then be defined by (assuming compact support) cpct − where the involution acts on of the operator ( about this, which will beinto two useful regions in using generalizing aand to domain orientifolds, non-BPS wall is D8-branes made to in of dividethe IIB. D8-branes the two and Differences space regions anti-D8-branes in are in theby then IIA, the values in D8-brane of one-to-one system, the correspondence and are RR with therefore fluxes the classified between RR by charges carried K 3.1 Real K-theory The orientifold backgrounds of interest are defined by a and antibranes. The corresponding groups are denoted an involution commutes with strings that includes world-sheetPaton parity factors Ω, and that an squares action to on either the open 1 string or Chan- p In ordinary (non-orientifold) type-IIAin and -IIB backgrounds RR fluxes take values 3. Orientifolds in K-theory satisfy the periodicity conditions Another variant of realalso K-theory exchanges the can two be bundles defined by considering an involution that and similarly for in IIB. The groupswe are start exchanged with relative a to D9- the D-brane classification since there and or D-branes, are classified by the ones without compact support. − JHEP05(2001)047 , 8 p S − 8 ). In (3.5) ,and ), and P X − ( X R 2 p ( Z ± K .InIIAwe KR p − ) for O 8 0 S p, − 9 ). Orientation reversal S ( X ) rather than ( , respectively. This is , 2 10 ± X Z ) − ( ), depending on whether p K 0 ± 1. Following the general X , ( 9 -plane we divide the trans- K KH − 2 → p S KR − ( . The former gives ) -plane, and the latter to the 0 1 ± , , − 2 1 and R R KR ± KH × and )= X ( 2 KR 2 , 2 , )or 0 0 Z 0 , R Ω in IIA maps a D8-brane wrapping R 9 K 9 S 9 . Note that, as expected, the groups which ( × I + ± / p 9 X ( R KR KR ) for O 0 p, )= − 9 X (3.2), and fails for ( S 1 ( ± , 6 1 5 − R . p KR ± ). The latter can be understood as the failure of ordinary Bott periodicity [23]–[25]. For example, D-branes in an O8 background (type IA) 1 , planes the groups are also KR 1 9 by wrapping D8-branes on the transverse sphere KH R − + p ). p × )= − X 9 0 ( X p, ( ± R KH 2 − Z 9 and O8 S exchanges the two spinor chiralities, and therefore gives K KH ( and 2 − 10 9 R )= − − -plane. p The appropriate groups for the other orientifold planesUsing (3.2)–(3.5) can we be can re-express similarly these deter- groups as For example, the O0 projection The usual 8-fold periodicity of orientifolds (and real K-theory) implies that for p X This relation is the analog in real K-theory of Hopkins’ relation in complex equivariant K-theory, ( + 5 ± O0 the action onarguments the of Chan-Paton [9], bundle we squares associate to the 1 former or to the O0 somewhat counterintuitive, in thatand a not D8-brane an gets anti-D8-brane. mappedsphere, to However which an recall in image that thismeans D8-brane the that case D8-branes the is wrap orientation of just theThe the transverse two projection “wrapped” points D8-brane maps on isother the reversed either side, on D8-brane side the so to other it of side. a looks the like D8-brane O8-plane. a of “wrapped” This themined. anti-D8-brane. same The results orientation are on summarized the in table 2. again consider both D8-branesD8-branes. and anti-D8-branes, The and orientifold in projection IIB (2.1) we acts use freely the on non-BPS the sphere, giving and the corresponding K-groups areacts obtained on by the keeping track D8-branes of and how their the bundles. projection KH classify RR fields are shiftedKR by one relative to the groups which classify RR charges, and similarly for verse space with a given orientationIt to therefore a exchanges D8-brane wrappedversa. D8-branes wrapping The with it group wrapped with is anti-D8-branes, therefore the either and opposite vice- orientation. 3.2 RR fields in orientifoldTo backgrounds classify RR fields in the background of an orientifold higher versions are defined preciselyreal as K-theory in (3.1). as follows, These groups are related to ordinary the O8 K when one tries to construct the Dirac index map: on the the real case it’s exactlyBott the periodicity opposite, works since for complex conjugation also exchanges the two chiralities. the latter JHEP05(2001)047 6 (3.7) (3.8) (3.9) (3.6) relative − 2 2 ). In the real Z f O5 Z 6thereisan Q . To compute Z Z , ⊕ , 2 ⊕ − X, Z Z ⊕ Z Z ⊕ ( 3) Z Z Z =2 Z Z ≥ KR . even )= )= flux of p )= )= 0 0 )= )= )= 0 0 , , H , , 3 m 0 0 = 0 8 )= 2 , , 6 4 )= , 0 9 1 5 0 , G S S ' , S S ± . 7 ( ( 3 S S S ( ( 2 1 1 ( ( ( 1 1 S Q . S .For ( − − ± ± ± Z − − ( ± ± mod 8 p KR − + × )( − ) ⊕ 8 } 0 pt. K-group KH KH KH KR KR KR KR KH KH X G Z ( ( , n 0 K , − for O5 for O5 + + + + + + + + + + 0 )= 0 1 2 3 4 5 6 7 8 , e Z flux of the O5-plane, and the second 2 Op O O O O O O O O O Z KR , , 3 3 Z 0 ⊕ P G , ⊕ ) 2 R , which also agrees with cohomology. The ( 10 Z Z Z 0 1 pt. , ( 2 G H Z 2 +1 )= )= , 2 ⊕ n 0 0 -plane itself, i.e. Z , , Z Z − p ) 4 4 { Z Z S S ⊕ m f Z O5 variants. Let us once again stress that these ⊕ ( ( , Z Z 1 1 ⊕ Z Z 3 ⊕ Z Z Z − − )= P KR Z Z R Orientifold K-theory groups for RR fields. )= )= )= )= pt. ( 0 0 )= )= )= KR 0 0 ( , , 3 KH , , 0 0 0 )= 6 4 )= , )= 2 8 , , m 0 0 5 1 9 0 H , S S , S S − 3 ( ( , whereas RR fields take values in 7 S S S ( ( m, 1 1 ( 1 ( ( 1 1 S S S ( − − − ± ± ( − − ( ± ± ± KR n Table 2: − KR KR KH KH KH KH KR KR K-group KR KR KR − − − − − − − − − − ) part of K-theory always agrees with cohomology. This corresponds 2 3 4 5 6 7 8 0 1 Z O O O O O O O Op O O . In K-theory we find torsion, which gives the − 1 For example, the relevant cohomologies for orientifold 5-planes are This is a consequence of the Chern isomorphism, e.g. G 6 to the RR flux of the orientifold 3.3 Comparing with cohomology The above results agreeThe with free cohomology ( in some cases, and differ in other cases. The first term corresponds to the integral to O5 How should we interpret the absencewe of need the torsion to subgroup understand in the better first how case? to Clearly relate (or “lift”) cohomology to K-theory. cohomologies do not incorporate the fractional shift in the additional free part corresponding to are classified by the groups we used the following decomposition isomorphism due to Atiyah [22] is difference between K-theory and cohomology can only be in the torsion subgroup. and our knowlege of the real K-groups of a point, case the cohomologies willcorresponding be RR twisted field or (2.2). normal, according to the action of the orientifold on the JHEP05(2001)047 3 1)- g Sq − is not (3.10) (3.11) (3.12) p 3 . d 5 -exact are d r ,where d H + . 3 0 ), as . g Sq X → ( = ) itself, requires solving ) K odd even 3 X X d , ( ( is the Steenrod square [27]. p p ) )= ). This is defined in terms of 3 K +1 The basic idea of the AHSS p X X X ( Sq ( ( 0 7 /K +1 K ) p ) for K ) which are trivial on the ( maps between ordinary and twisted X Z ( /K X 3 p ) ( d X, ,where K ( X K p ( H p → ⊆···⊆ 0 for H 11 + K ) ) p 3  X X ⊕ -plane. We shall assume that the first term is ( ( Sq p + 1 )= ), and in particular p − = K )= n X X 3 ( ( K X d p → ( +1 ). Each successive step is given by the cohomology of ) K p ⊆ K Z ) . As we have seen before, this provides the difference X 2 ( /K Gr X ) Z X, is just the usual co-boundary operator acting on co-chains ( +1 ( p n ∗ 1 X ( d K K p H ) using a sequence of successive approximations, starting with )= K ), e → Z X , -plane and the O ( X 0 p ( that acts on classes in the previous step. Thus classes which are not − − K p r 8 K ) (or any of its variants), denoted Gr d P ) is the subgroup of classes in X R ( X ( . The successive ratios appearing in (3.10) are precisely the objects one ( 3 p K X H K ∈ Actually, what one really computes in the AHSS is an “associated graded com- There is in fact a systematic algorithm which does exactly this, known as the The AHSS can be extended to real K-theory as well. In the appendix we de- Introductory reviews of the AHSS for complex K-theory can be found in the appendix, as well (which is untwisted) is trivial in all cases. ] 7 -closed are removed in the new approximation, and classes which are 5 r H trivial in both cases, andd justify this assumption in hindsight. We shall also see that trivial. For example, The only other differential which may be non-trivial in ten dimensions is cohomologies, and must therefore itself be twisted. The precise form of is to compute known, but it can presumably be expressed in the form between the O is a twisted version[ of the Steenrod square. Here the second term is torsion, since (which are the zero’th step), so thedifferentials, first if step they is ordinary are de-Rahm non-trivial, cohomology.cohomology Higher will classes, refine and the trivializing approximationin others. the by complex removing The case some first is given non-trivial by higher differential d as in the appendices of [27] and [28]. plex” of Atiyah-Hirzebruch spectral sequence (AHSS) [26]. scribe how toappearing do in this table 2. incase, In the this with case case the the modification of AHSStion that follows a can the a cohomologies be freely similar appearing twisted pattern actingfields in to or (2.2). involution, the the normal, complex first Correspondingly, in such approxima- the accordance as differential with the the ones orientifold action on the RR a differential where integral cohomology To get the actual subgroups tree of computes by the AHSS. In particular, in the first order approximation extension problems of the form a filtration of JHEP05(2001)047 ) 5 P R (3.14) (3.13) (3.15) ( 2 H ) will differ X ( . K 0 ) → 5 ) P 5 R P ( R 4 ( 2 ). If all the extensions are 3 H Z X , ⊕ ( /K = 4 ) ). In other words, K-theory is ) 2 5 Z 5 +1 p X P P H ( R R )= ( /K K ( 5 2 ) 2 , for which . P K 5 X H 2 ( R P p )). However it will be extremely useful ( Z ⊕ → R e K 12 ) X K )=Gr ⊕ ) ) 5 ( 5 5 = 2 ⊕ P X K P P ( Z ) )= R X R R ( 5 K ( ( 0 X ⊕ P 2 1 ( H Z R K K +1 ( p 1 = → K K )= 5 ) 2 5 P Z P )= R ( kk k = R X ( 4 ( K 3 p H K Gr K ) (as opposed to Gr → X 0 ( ) become correlated in K-theory; twice the generator of the former cor- K 5 P R ( 4 H In practical terms, one often needs additional information to determine whether Consider for example the space On the other hand it is known (see for example the appendix of [28]) that responds to the generator of the latter. an extension iscomputing trivial or not, which precludes the AHSS as a useful method for and 4. Applications Having elaborated the K-theory machineryorientifolds, needed we to would classify now RRfact like fields that to surrounding that demonstrate RR three fields important are implications valued in of4.1 K-theory the Charge rather quantization than cohomology. As was previously stated,proper there RR are charge two quantization. separate issues The related first to has the to question do of with the absolute fractional If the above exactto sequence be splits, i.e. trivial, if and it is exact both ways, the extension is said The relevant extension is given by so this extension is non-trivial. The RR 2-torsion and 4-torsion appearing in trivial we deduce recursively that in our case, sinceable we to determine already all know theK-theory the extensions data precise (3.12) with unambiguously, groups cohomology. and (table thereby compare 2). the We are therefore isomorphic to the ringAHSS. On of the cohomologies other that hand, survive if some the of higher the differentials extensions are of non-trivial the from its graded complex.different from In that particular, of theof the different additive cohomology degree classes. structure become in Physically correlated. this K-theory means will that be RR fields JHEP05(2001)047 . + p (4.2) (4.4) (4.3) (4.5) and O − p , but both are 0 G and 0 2 ) (4.1) G → X . ( 1 . Z ) ). The first issue encompasses K Z Z Z , respectively, we can determine + = /K , p + 0 3 0 ∈ P K )= )= H . R e Z Z ( Z , , 3 → 2 2 Z and O6 = P P H Z ,x , 0 for O ⊕ 2 0 − R R ) − ( ( 3 ⊕ 13 Z K x ) is obviously trivial in this case, the AHSS p d 0 2 K 5 −→ d H H = = ) ch( e 0 1 Z b = =0 = → A 4, for example), and the second with the relative , ) for O6 K K 3 1 2 2 0 ZZ , (and p could be in the map ≤ P 3 K 1 3 3 R S p /K /K = = kkk d ( ( d 0 1 K 2 0 ) K K x H H π ( KR 2 G → 0 )and + 1) in these cases [17]. 0 , 3 -planes with n S p ( KH = 0 K The first case where this procedure gives a non-trivial result is the O5-plane. The O6-plane background supports two RR fluxes, Note first that for the O8 and O7-planes K-theory is in complete agreement Using does not have ahow unique the solution, two so RR one fluxes are cannot correlated. determine We in willThe this return cohomologies way and to whether K-theory the and for O6-planeonly the in possible O5-planes section were non-triviality given 5. of in (3.8) and (3.9). The However the extension given by the filter groups uniquely, the second, and should in principle be resolved by an analog of the relation [7] integral. Since the differential In particular the O8-planebrane). (O7-plane) This cannot is support consistentgauge group with a SO(2 other fractional arguments D8-brane which (D7- rule out the possibility of a terminates at the first approximation, i.e. charges (of the O charge induced by RR torsion (1/2 for O for real K-theory.centrate Unfortunately on the the generalization issueprevious is section, of not is the known, that relative sovia in non-trivial K-theory charge. we extensions. RR will fields The con- of main different idea, degree become aswith correlated alluded cohomology. to There in isnon-trivial the only extension. a single There RR are flux, only and two therefore variants no of possibility these for a planes, O JHEP05(2001)047 is 3 d (4.9) (4.7) (4.6) (4.10) . The AHSS + . In the latter plane, and has − ) (appropriately , i.e. − p p − − 8 . 2 , P and O5 − + planes. The relevant H maps to the generator = 0 (and we assume 0 R − ( 3 − . We therefore see that p 0 H 1 , which are appropriately → p H − , 2 7 . p for the O5 2 H ) Z for O5 for O5 → for both ) ) (4.8) − p H 3 Z 7 p p − 2 − − 8 G 2 2 =0 =1 = 8 8 Z )= =2 P /K Z p p p p P P p p /K R = − − − R R ( 1 8 ⊕ 6 6 1 ( ( p for for P p p Z K − Z H for K H − − R 8 6 11 ( Z p → )= → H H H is trivial since )= − 0 0 2 2 6 , p , 1 is trivial since 3 4 )= 4 Z Z Z − 14 d H . The generator of S 5 6 K S Z could arise from the (twist preserving) maps ( −→ −→ −→ ) is trivial for both O5 ( d − , 1 K 5 5 1 3 ) ) ) flux, i.e. the charge of the orientifold plane, is − → d d − )= )= )= p p p P and p e − − − plane. Z Z Z → R Z 8 8 8 − , , , ( Z 8 6 7 8 + KR KH 3 P P P 2 ZZ , but to twice the generator for O5 = kk G (and P P P + R R R p K  H 3 ( ( ( = 3 R R R )= − p p p ( ( ( d p 7 kk k p H = = 0 1 2 − − − − − 6 3 1 planes, and K 1 3 8 8 H H H maps to non-trivial torsion in → − P H H H K K torsion. for O5 H 1 p R 0 p 1 ( = = K − p 2 there is an additional RR cohomology → K 6 0 2 − )=0,so 8 0 ≤ G K K e Z H p , , but non-trivial for O5 3 + = 0 for O P R ( H 0 H = 0). Possible non-trivialities of Actually, for The story repeats along similar lines for the lower O torsion produces a half-integral shift in the flux of 3 1 g so the half-integral part of the is trivial for O5 again terminates at cohomology, and we find Sq twisted or untwisted. The differential case the generator of no effect on the charge of the O5 This time we find that the extension given by G however in all cases the maps are found to be trivial. The relevant extension is cohomologies are trivial since However determined by the As with the other cohomologies, these lift unobstructed to the graded complex; of (the free part of) JHEP05(2001)047 − Z are The ⊕ + (4.14) (4.13) (4.11) . 8 Z − f O3 plane . 0 + case above; C and − and O0 + ) transformation − ,whichis with Z planes, as we will 0 0 , K + ω plane, the extensions p . = + charge (and tension) as 0 p − -torsion. This will modify ) (4.12) 2 → + 1. On the other hand, the e H Z K p 0 , same 5 − ω 3 P p )theoriesonO1 − R → 2 k /K ( n ,and , 0 p 3 , H 0 − Z ω 2 . In particular, for the O3 H  ω case is similar to the O6 3 , such that K = d .AswiththeO5 ∪ 0 H − Z p ∪ torsion does not shift their charge. ω −→ 11 01 → H − 15 p 3 )  p − Z )= 2 K − = , 0 2 = , torsion therefore lifts to a trivial class, and the 5 G + 1. We therefore identify 6 3 K 3 P S 0 T ( G therefore have the G R C 1 1. The O2 ( → . − ± planes with non-trivial 0 − , 2 p → H − Z − c O0 0 3 p : KR =0 , cannot support the end of a D-string. C 3 to − K . We will return to this problem in section 5. For O1 p d 0 Z and G → for − 0 2 Z c O1 3 through the differential variants are related by a shift ⊕ ≤ + Z p fluxes are equivalent in K-theory, the two variants O3 3 the extension is trivial, so G − and a D7-brane for O1 − =2,and planes correspond to O The story becomes even more interesting for the lower O p This is also consistent with the fact that the branes which create this torsion, i.e. a D6-brane + 8 p for NSNS and RR 3-torsions of the two variants are related by an SL(2 which also shifts the axion see next. 4.2 Equivalent orientifolds O equivalent as well.implies Physically the this existence can of be a constant understood 0-form as follows. Equation (4.12) is a surjective map from there is no uniquedetermine solution the to effect the extension, of so one must use a different method to and the two O3 the AHSS for are trivial, so thethe charge doesn’t two shift. However, the new feature here is that, since One can show that in all three cases variants we called the original torsion-free planes.how) It the would gauge be theories on interestingorientifold D1-branes to variants and differ investigate D0-branes from whether in the (and the ordinary presence SO(2 of these exotic and O0 normal or twisted) is trivial. There is also an additional extension to solve, in type-IIB string theory given by [1, 2] graded complex is simply Gr for O0 JHEP05(2001)047 − + + + is f and O2 + case, plane (4.17) (4.15) (4.16) 9) are + f O2 planes, + p + 3 with flat + O f O2 1 ,..., ω The =3 . and one 9 i 9 ( + x i x )= 9 x planes correspond to an ( 2 + can be written as a derivative , respectively (figure 3), which ), where n i + dξ . G x ( ,ξ , ∧ plane, respectively (figure 3). The ξ ) torsion class which is not trivial p 2 planes, but in the second case only + π 0 − − B planes. This is unlike the 3 G 2 + ω < f <π and O3 O4 + = 9 f 9 O2 ∪ )= + 3 i torsion can be written as . On the other hand, the M-theory shift x = 2 there is again an interesting physical H 16 p f O3 π will not correspond to just the cup product