The Hyperbolic Volume of 3-manifolds Y. Liu, S. Wang and P. Derbez October 29, 2013 Volume of representations Let G be a real semi-simple Lie group and let K denote its maximal compact subgroup. The homogeneous space X = G/K , with base point x0 = {K }, is contractible and we fix a G-invariant Riemannian metric on it. Let ωX be the G-invariant volume form. Let M be a closed, connected, oriented n-manifold such that dimM = dimX = n. e →X For each ρ : π1 M → G, there is a developing map Dρ : M ∗ which is equivariant. Then one can identify Dρ (ωX ) with a n-form on M and Z ∗ volG (M, ρ) = Dρ (ωX ) M Constructing developing maps Fix a triangulation T = {∆1 , ..., ∆s } of M and a Dirichlet domain e Denote by {x1 , ..., xN } the vertices of T e lying in intΩ and Ω in M. let {y1 , ..., yN } be N arbitrary points in X . We first set Dρ (xi ) = yi for i = 1, ..., N. e and next We then extend Dρ equivariantly to all the vertices of T to all the edges, faces etc... straightening their image w.r.t. the homogeneous metric on X . e 1 , ..., ∆ e s a lift of ∆1 , ..., ∆s . The volume of ρ can be Denote by ∆ computed as follows Z volG (M, ρ) = M Dρ∗ (ωX ) = s X e i )) εi volX (Dρ (∆ i=1 e i is orientation where εi = ±1 depends on whether Dρ |∆ preserving or not. Volume and Chern-Simons functional The Chern Simons invariants are based upon the observation if ω is a connection on a principal G-bundle P over M then f (Ω ∧ Ω) is an exact form on H 4 (P), where f = Tr : g ⊗ g → C is an invariant polynomial. A primitive is given explicitly in their work by 1 cs(ω) = f (dω ∧ ω) + f (ω ∧ [ω, ω]) 3 This is the Chern-Simons class of ω. If the bundle P → M is trivial, the Chern-Simons functional is then defined by Z csM (ω, δ) = δ ∗ cs(ω) ∈ C M Suppose from now on G = PSL(2; C). For each representation ρ : π1 M → G admitting a lift into SL(2; C), we have the (trivial) principal bundle M ×ρ G and the associated bundle M ×ρ H3 e × G and M e × H3 by the diagonal action of π1 M. obtained from M Denote by A the flat connection over M corresponding to ρ. This connection is defined by A = q ∗ (ωM.C. ) where ωM.C. = dLg −1 denotes the Maurer Cartan form of G which is the unique left invariant g-valued 1-form such that ωM.C. (X ) = X for any X ∈ g. GO p / H3 O q q e ×G M − P = M ×ρ G p p O δ M s e × H3 /M − / E = M ×ρ H3 6 1 0 0 0 0 1 The matrices X = ,Y = ,Z = form a 0 −1 1 0 0 0 basis of the Lie algebra g = sl(2; C) with commutators relations [X , Y ] = −2Y , [X , Z ] = 2Z , [Y , Z ] = −X Denote by ϕX , ϕY , ϕZ the dual basis of sl∗ (2; C). The Maurer Cartan form of G is ωM.C. = ϕX ⊗ X + ϕY ⊗ Y + ϕZ ⊗ Z one can consider the 3-form split by a formula of Yoshida cs(ωM.C. ) = 1 1 ϕX ∧ ϕY ∧ ϕZ = 2 p∗ ωH3 + icsL.C. (H3 ) + dγ π2 π inducing a closed 3-form on P cs(A) = 1 ∗ q (ϕX ∧ ϕY ∧ ϕZ ) π2 The Chern-Simons functionnal is then defined by Z csM (A, δ) = δ ∗ cs(A) ∈ C M Plugging this formula in the above diagram we get csM (A, δ) = cs(Mρ ; δ) − i vol (M, ρ) π2 G Note that this formula was proved by Kirk and Klassen when M is closed hyperbolic and ρ is the discrete and faithful representation. The Chern-Simons invariant csM (A, δ) is well defined modulo Z when the section changes. Then define cs∗M (A) to be the class of csM (A, δ) in C/Z which can be seen as cs∗M (A) = e2iπcsM (A,δ) Abelian representation Any abelian representation of a closed oriented 3-manifold M into G has a zero volume. Suppose the image of ρ : π1 M → PSL(2; C) is torsion free. Then there is a path of representations ρt : π1 M → PSL(2; C) such that ρ1 = ρ and ρ0 is the trivial representation. Consider the associated path of flat connections At . This path defines a connection A on the product M × [0, 1] that is no longer flat but whose curvature satisfies the equation Ω ∧ Ω = 0 (this latter point follows from the fact that ΩAt = 0 for any t). Since dcs(A) = Tr(Ω ∧ Ω) = 0 it follows from the Stokes formula that cs∗M (A0 ) = cs∗M (A1 ) = 1. Therefore vol(M, ρ) = 0. From local to global representations Suppose M = V ∪T W is a closed oriented 3-manifold which is a union two manifolds with incompressible toral boundary such that ∂V = ∂W = T and fix a regular neighborhood T × [−1, 1] with T × [−1, 0] ⊂ V and T × [0, 1] ⊂ W . We suppose each component T of T is endowed with a meridian-parallel system (mT , lT ) and we denote by xT = mT ∩ lT . Let ρ : π1 V → PSL(2; C) and ϕ : π1 W → PSL(2; C) denote two representations which lift into SL(2; C). They both induce representation still denoted by ρ and ϕ of π1 (T × [−1, 1], xT ) into PSL(2; C) and we will consider the case of non-parabolic representations. T'x[-1,1] V l_T m_T Tx[-1,1] W After conjugation in PSL(2; C), we may assume that there exist αT , βT ∈ C such that 2iπα 2iπβ T T e 0 e 0 ρ(mT ) ∼ ρ(lT ) ∼ 0 e−2iπαT 0 e−2iπβT and αT0 , βT0 ∈ C 2iπα0 T e ϕ(mT ) ∼ 0 0 0 e−2iπαT ϕ(lT ) ∼ 0 e2iπβT 0 0 0 e−2iπβT Denote by A and B two flat connections over V and W corresponding to ρ and ϕ. Then A|T × [−1, 1] and B|T × [−1, 1] can be turned into normal form after gauge-transformation. Namely, there exist gT ∈ G and gT0 ∈ G such that iαT dx + iβT dy 0 ∗ gT A|T × [−1, 1] = 0 −iαT dx − iβT dy where gT : T × [−1, 1] → SL(2; C) and hT∗ A|T 0 iαT dx + iβT0 dy × [−1, 1] = 0 0 0 −iαT dx − iβT0 dy where hT : T × [−1, 1] → SL(2; C) By obstruction theory one can extend a a gT , hT : T × [−1, 1] → SL(2; C) T ∈T T ∈T to gauge transformations g : V → SL(2; C), h : W → SL(2; C). Eventually if we manage to find representations ρ and ϕ with the same eigenvalues when restricted to each torus T then the connections g ∗ A and g ∗ B match on T × [−1, 1] and their union define a flat and smooth connection C over M and an associated representation ψ of π1 M into PSL(2; C). Additivity principle Denote by [M] the orientation class of M and by [V , ∂V ] and [W , ∂W ] the induced orientations so that [M] = [(V , ∂V )] + [(W , ∂W )] (the orientations induced on ∂V and ∂W are opposite on T ). By linearity of the integration cs∗M (C) = cs∗V (CV )cs∗W (CW ) Along each component T of T we perform a Dehn filling identifying each mT with the meridian of a solid torus and denote b = V ∪ V − and W c = W ∪ V + the resulting closed manifolds. by V Suppose both CV and CW smoothly extend to flat connections cV , resp. C d b , resp. W c , and denoted by C over V W. These extensions are possible provided [mT ] lies in ker ρ and ker ϕ. Again by the linearity we have cV )cs∗ (C d cs∗Vb (C W) = c W + cV |V − )cs∗ (CW )cs∗ + (C d = cs∗V (CV )cs∗V − (C W |V ) W V Since the extensions from ∂V , resp. ∂W , to V − , resp. V + based on the normal form on [−1, 1] × T , are the same on the T direction but opposite on the [−1, 1] direction + cV |V − )cs∗ + (C d cs∗V − (C W |V ) = 1 V Eventually we get b , ρb) ± volG (W c , ϕ) volG (M, ψ) = volG (V b b , resp. of ϕ, to π1 W c. where ρb, resp. ϕ, b is the extension of ρ to π1 V Maximal volumes Given (G, K , X ) and M the Chern Simons rigidity theorem claims that the forms Dρ∗ (ωX ) represent at most finitely many classes in H 3 (M). Hence it makes sense to define VG (M) as the maximum of volG (M, ρ) when ρ : π1 M → G runs over the representations. The invariant VG is a lower volume, namely for any map f : M → N then VG M ≥ |degf |VG N According to the terminology of Reznikov, a non-negative manifolds invariant V is a volume if for any finite covering map f : M → N then V (M) = |degf |V (N) Example: The Euler characteristic of surfaces and more generally the Gromov simplicial volume. Three-dimensional examples of VG The universal covering + 2 ^ SL 2 (R) → PSL(2; R) = Iso H is endowed with the (pull-backed) hyperbolic metric. ^ ^ One gets (SL 2 (R), 1, SL2 (R)). g2 ), SO(2), SL g2 ), where Isome (SL g2 ) More generally (Isome (SL denotes the identity component of the whole isometry group. When G = PSL(2; C) we get the hyperbolic geometry (PSL2 (C), SO(3), H3 ) and the associated hyperbolic volume HV = VPSL(2;C) . The other 3-dimensional geometries give always a 0 maximal volume. Hyperbolic manifolds Theorem (Reznikov, Gromov) Let M be a closed hyperbolic manifold. Then HV (M) = volH3 (M) = µ3 kMk and the maximum is reached only by the discrete and faithful representation. One can check basically that for any closed oriented 3-manifolds M (not necessarily hyperbolic) HV (M) ≤ µ3 ∆3 (M) Reznikov gave a sharper bound proving that HV (M) ≤ µ3 kMk Structure of 3-manifolds Let M be a closed oriented 3-manifold. By the Kneser Milnor decomposition theorem, M splits into M = M1 ]....]Mr ](S1 × S2 )]....](S1 × S2 ) where the Mi ’s are irreducible 3-manifolds. By the works including those of Jaco, Shalen, Johannson, Thurston, Hamilton, Perelman, each irreducible closed oriented 3-manifold M is either geometric or can be cut out along a family of tori and Klein bottles TM such that M \ TM = S(M) ∪ H(M) where the components of S(M) are H2 × R and those of H(M) are hyperbolic. When both TM and H(M) are non-empty M is termed a mixed manifold. Gromov proved that k.k is additive w.r.t connected sum and toral decomposition. Moreover when M is irreducible kMk = X Q∈H(M) kHk = X Q∈H(M) volH Q µ3 In particular if HV (M) is positive then M is necessarily hyperbolic or a mixed 3-manifold. Questions for mixed 3-manifolds 1) Is HV (virtually) positive for mixed 3-manifolds ? 2) Is HV a volume? 3) What kind of information is detected by the hyperbolic volume? If the answer to the first question is yes then HV detects the hyperbolic pieces of M as the simplicial volume does but the latter only depends on the geometric pieces of M and not on the way they are glued together. One can therefore wonder whether HV measures the complexity of the sewing involution between the geometric pieces. l s FxS^1 m h Bi-reducible hyperbolic manifold Solid Torus Manifolds with HV = 0 and kMk > 0 Let M1 = F × S1 where F is a surface with positive genus and connected boundary. Let (s, h) denote the associated section-fibre basis of H1 (∂M1 ; Z). Let M2 be a one-cusped, complete, finite volume hyperbolic manifolds endowed with a basis (m, l) of H1 (∂M2 ) such that both M2 (l) and M2 (m) are connected sums of lens spaces. Such manifolds do exist by a construction of Hoffman and Matignon. Denote by ϕ : ∂M1 → ∂M2 the homeomorphism defined by ϕ(s) = m and ϕ(h) = l −1 . Then M = M1 ∪ϕ M2 satisfies kMk > 0 and HV (M) = 0. Let ρ : π1 Mϕ → PSL(2; C) be any representation and denote by A the resulting connection over Mϕ . Notice that either ρ(s) or ρ(h) is trivial. Indeed if ρ(h) 6= 1, its centralizer Z (ρ(h)) in PSL(2; C) must be abelian. Since h is central in π1 M1 , this means that ρ(π1 M1 ) is abelian. Since s is homologically zero in M1 , then ρ(s) = 1. Let ζ be either s or h so that ρ(ζ) = 1. After putting A in normal form with respect to T , denote by A1 and A2 the flat connections over M1 and M2 respectively. Since ρ(ζ) = 1 then both A1 and A2 b 1 and A b2 do extend over M1 (ζ) and M2 (ζ) to flat connections A such that b b 1 ) × cs∗ cs∗Mϕ (A) = cs∗M1 (ζ) (A M2 (ζ) (A2 ) Eventually taking the imaginary part we get vol(Mϕ , ρ) = vol(M1 (ζ), ρb1 ) ± vol(M2 (ζ), ρb2 ) where ρbi denotes the extension of ρ|π1 Mi to π1 Mi (ζ). Since both vol(M1 (ζ), ρb1 ) and vol(M2 (ζ), ρb2 ) do vanish by the construction, this proves that HV (M) = 0. Answer to Questions 1 and 2 Theorem (L-W-D) Any closed mixed 3-manifold has a virtually positive hyperbolic volume. Remark Since there are mixed 3-manifolds with zero hyperbolic volume then HV is only a lower volume and not a volume whenever M is not hyperbolic which provides a negative answer to Question 2. However HV can be rescaled to define a honest volume ( ) e HV (M) HV (M) = sup ≤ µ3 kMk e → M) deg(M and HV (M) > 0 iff kMk > 0. Proof of the theorem Some examples Let f : M → N be a map between oriented closed 3-manifolds. If ρ : π1 N → G is a representation then it induces a representation ρ ◦ f∗ : π1 M → G and volG (M, ρ ◦ f∗ ) = |deg(f )|volG (N, ρ) Let M = Q ∪T (F × S1 ) where Q has a hyperbolic interior and T = ∂Q = ∂F × S1 is connected. The torus T has a natural meridian-parallel basis (m, l) given by ∂F and the fiber S1 . There is a degree one ”pinching” map π : F × S1 → D 2 × S1 b inducing a degree one map M → Q(m) implying that b HV (M) ≥ HV (Q(m)) \tilde{Q} \tilde{N} T Q N Consider M = Q ∪T N where Q has a hyperbolic interior and T = ∂Q = ∂N is connected. Then i] Rank H1 (∂N; R) → H1 (N; R) = 1 We may therefore choose a meridian-parallel basis (m, l) of H1 (∂N; Z) such that i] (l) has infinite order whereas i] (m) is a torsion element in H1 (N; Z). Lemma There is a prime number p0 such that for any slope m ⊂ ∂Q and any prime number q ≥ p0 one can find e → Q inducing the q × q-characteristic - a finite covering p : Q covering over ∂Q, and e −1 (m)) → PSL(2; C) of positive - a representation ρ : π1 Q(p volume. Choose q big enough so that Q(q.m) is a hyperbolic orbifold. There is therefore a representationρ : π1 Q → PSL(2; C) with eiπ/q 0 positive volume such that ρ(m) ∼ and 0 e−iπ/q x 0 ρ(l) ∼ for some x ∈ C∗ . 0 x −1 By a result of Hempel, if q is big enough there exists a finite e → Q inducing the q × q-characteristic covering covering p : Q on the boundary. Meanwhile p induces an orbifold covering e −1 (m)) → Q(q.m). Accordingly the composition ρ ◦ p∗ is b : Q(p p a representation factoring through the fundamental group of the e −1 (m)) whose volume is (non-singular) hyperbolic manifold Q(p deg(p)volH3 Q(q.m) > 0 e → M be a finite covering such that each component of Let p : M e Each component Q e of p−1 (Q), p−1 (Q) is homeomorphic to Q. has a representation e m) e → PSL(2; C) ρ : π1 Q( q x 0 e of positive volume where ρ(l) ∼ . 0 x −q Since l ∈ H1 (N; Z)/TorH1 (N; Z) there is a homomorphism x 0 H1 (N; Z) → PSL(2; C) sending m trivially and l to . For 0 x −1 e of p−1 (N), consider the representation η each component N p] e → H1 (N; e Z) → π1 N H1 (N; Z)→PSL(2; C) q x 0 e e with η(m) = Id and η(l) = and such that 0 x −q e m), e η) = 0. volG (N( By the additivity principle we conclude e m)) e ≥ dvol 3 (Q( e HV (M) H where d is the number of components of p−1 (Q). Przytycki-Wise surfaces Let M be a closed oriented mixed 3-manifold without essential Klein bottles. By a PW -surface of M we mean a π1 -injective map j : S → M of a closed orientable surface in minimal general position w.r.t. the geometric decomposition such that: 1. for each maximal graph chunk N ⊂ M the components of j −1 (N) are virtually embedded in N 2. for each hyperbolic piece Q the components of j −1 (Q) are geometrically finite in Q Remark Any component of TM is a trivial example of a PW-surface. A partial PW-surface is a connected subsurface R ⊂ S such that j(∂R) is immersed in a single torus T0 of TM and for every component T of TM the components of j −1 (T ) covers the same slope of T . Theorem (Haglund-Wise) Every PW-surface subgroup of π1 M is separable and every double coset of two PW-surface subgroups of π1 M with non-trivial intersection is separable. Passing to some finite covering we may assume that 1. M containg no Klein bottles, 2. each JSJ-torus is shared by two distinct geometric pieces, 3. each Seifert piece is homeomorphic to a product F × S1 , 4. each maximal graph-manifold chunk has the antenna property. The antenna property of Przytycki-Wise means that if N is a graph manifold with ∂N 6= ∅ then for any vertex v0 of ΓN there exists an edge-path (v0 , ..., vn ) which is a full subgraph of ΓN where vn is a boundary-vertex. S_1 T_0 Q_1 Q_2 R Antenna Q_0 S_2 Proposition (Virtual boundary) Let Q0 be a hyperbolic piece of M0 , T0 be a component of ∂Q0 and let ξ0 be a slope of T0 . There exists a partial PW -surface R which virtually bounds ξ0 outside Q0 . This means that j : S → M induces a proper immersion j| : (R, ∂R) → (M \ intQ0 , T0 ) such that ∂R covers ξ0 under j. We will denote by X (R) the carrier chunk of R which is the minimal chunk containing R. We are going to explain how to construct a PW-surface j : S → M and the desired partial PW-surface R ⊂ S. By Przytycki-Wise there exists a geometrically finite properly immersed incompressible surface R0 in Q0 whose ∂R0 6= ∅ and covers ξ0 . Let Q be the geometric piece adjacent to Q0 along T0 (Q 6= Q0 ). If Q is hyperbolic too, by the same argument as above we choose (R, ∂R) → (Q, T0 ) where ∂R covers ξ0 . We use the merging trick: choose a finite (possibly f0 → R0 and R e → R such that R f0 and R e disconnected) covering R have the same number of boundary component and all of these components cover ξ0 with the same (unsigned) degree. This is possible because for any hyperbolic surface S there exists an integer K > 0 such that if we assign a positive degree ni to each e→S component Ci of ∂S then one can find a finite covering S whose degree on each component over Ci is Kni . Then eventually S is obtained taking a component of [ f0 e R R f0 =∂ R e ∂R If Q is Seifert denote by Qmax the maximal graph chunk ⊃ Q. If ∂R0 is parallel to the fibre of Q then taking two copies of R0 and connecting each pair of boundary components by a properly immersed ∂-essential vertical annulus in (Q, ∂Q) we get S. Suppose ∂R0 is not parallel to the fibre of Q. We denote by A = Q0 ∪ Q1 ∪ ... ∪ Qn the antenna of Qmax with Q1 = Q. There exists a properly embedded horizontal surface E in Q1 ∪ ... ∪ Qn meeting T0 along parallel copies of ξ0 . This is based on the following result Lemma (Wang-Yu) Let S = F × S1 with boundary T1 , ..., Tl and assume there are families of disjoint identically oriented circles C1 ⊂ T1 , ..., Cl−1 ⊂ Tl−1 such that [C1 ] ∩ [fibre] = ... = [Cl−1 ] ∩ [fibre] 6= 0. Then there exists a family of disjoint identically oriented circles Cl ⊂ Tl such that ∪Ci is the boundary of a horizontal surface. Denote by E a surface of the antenna obtained by matching up horizontal surfaces in Q1 ∪ ... ∪ Qn such that the components of ∂E ∩ T0 are parallel to ξ0 . If a component of ∂E is not in ∂Qmax then one can make sure that it lies in some JSJ-torus inside Qmax parallel to the adjacent fibre. Then these boundaries can be closed up by an immersed vertical annuli. If a component ∂E in ∂Qmax then the adjacent pieces are hyperbolic and the surface can be closed up by merging some geometrically finite surfaces. Proposition e → M in which every elevation of There exists a finite covering M (S, R) is an embedding and intersects any elevation of T0 in at most one slope. The separability of PW-surface groups in π1 M implies that for e of M such that any any j : S → M there exists a finite covering M e e e elevation j : S → M is an embedding. Suppose S is an embedded PW-surface in M crossing a component T of TM . The double coset separability of any two PW-surface groups in π1 M implies that there exists a finite e of M such that for any elevation T e of T and for any covering M e e e elevation S of S the components of S ∩ T cover the same component of S ∩ T . Suppose that S ∩ T contains two components c1 , c2 and fix a base points x1 ∈ c1 and x2 ∈ c2 . Let γS an arc in S and γT an arc in T from x1 to x2 and denote by γ the closed curve γS ∪ γT . By construction [γ] 6∈ π1 S.π1 T . There is therefore an epimorphism ϕ from π1 M into a finite group such that ϕ([γ]) does not belong to the image of π1 S.π1 T . e → M the elevations of In the corresponding regular covering M c1 and c2 never belong to the same component over T . b of S e b of M e such that any elevation S There a finite covering M e meets each JSJ-torus of M in at most one slope. e ∩T e be a JSJ-torus of M e and suppose S e contains two Let T components c1 , c2 . Since they cover the same component of S ∩ T then they are directly parallel with the direction induced from S. The curve γ = γS ∪ γT meets S once. Then the abelian covering corresponding to π1 M → Z → Z/2Z where the first arrow is the Poincaré Duality ∩[S] and the second is obtained after dividing by the image of π1 T satisfied the conclusion. For each boundary component Ti of Q0 , 1 ≤ i ≤ n, we fix a meridian parallel system ξi , li such that Q0 (ξ1 , ..., ξn ) is hyperbolic and we denote by ρ : π1 Q → PSL(2; C) the representation factoring through the discrete and faithful representation π1 Q(ξ1 , ..., ξn ) → PSL(2; C). Denote by xi±1 the eigenvalues of ρ(li ). For each i we fix a partial PW -surface Ri which virtually bounds e → M (regular and ξi and a finite covering p : M q × q-characteristic) such that any elevation of Ri is embedded and meets any elevation of Ti in at most one slope. Define for ei of X (Ri ) containing an elevation R e i of Ri a each elevation X homomorphism e i ] : π1 X ei → H1 (X ei ; Z) → Z ϕ ei = ∩[R e i ] = 1 so that By construction any elevation eli of li satisfies [eli ] ∩ [R e e ϕ ei (li ) = 1 ∈ Z and ϕ ei (ξi ) = 0 Denote by αi : Z → PSL(2; C) the morphism defined by q xi 0 1 7→ 0 xi−q We next define on each component outside of p−1 (Q0 ) ∪i p−1 (X (Ri )) the trivial representations. For each f0 of Q0 in M f0 )∗ the induced e denote by ρe = ρ ◦ (p|Q elevation Q ei of Ti the elements ρe(eli ) are representation. For any elevation T all conjugated in PSL(2; C) to αi ◦ ϕ ei . Eventually the representations ρe, αi ◦ ϕ ei , 1 match up to conjugation on the common boundary so that one can define a global flat connection and therefore a global representation e → PSL(2; C) such that ψ : π1 M X e ξ), e ψ) = e ρe) > volG Q(ξ1 , ..., ξn ) volG (M, volG (Q( where ξe is made of one elevation of each ξi on each component e of ∂ Q. Using the same construction and adapting the gauge theory of ^ PSL(2; C) to the group of Isome (SL 2 (R)) of the Seifert geometry we prove Theorem (L-W-D) Any closed mixed 3-manifold containing at least one Seifert piece has a virtually positive Seifert volume. Remark We proved the same result for non-trivial graph manifold and meanwhile we found some non-trivial graph manifolds with zero Seifert volume so that SV is neither a volume but a lower volume whenever M is not Seifert. These results lead to another Question Is SV virtually positive for hyperbolic and mixed 3-manifolds that contain only hyperbolic pieces in their geometric decomposition?