Introduction
Tree-level scattering amplitudes in quantum field theories are rational functions on kinematic spaces encoding the probabilities of certain outcomes of particle interactions. In recent years, remarkable insights have been gained by finding combinatorial and geometric structures in these amplitudes, in particular for Tr φ 3 \operatorname{Tr}\varphi^3 Tr φ 3 theory and N = 4 \mathcal{N}=4 N = 4 supersymmetric Yang–Mills theory; see [13 ] for a survey. As pointed out in [16 ] the geometric and combinatorial features of N = 8 \mathcal{N}=8 N = 8 supergravity [FV12] amplitudes, however, are still more elusive. For example, these amplitudes will display double poles in soft limits, complicating the analysis. In the first part of this work we present a geometric interpretation of these amplitudes. This gives rise to the pentad varieties which we will study in the latter part from a purely algebro-geometric perspective. Motivated by the notion of little group scalings for scattering amplitudes, we will reduce the study of the pentad variety to that of the affine slope variety as first defined in [14 ] .
Our first central object will be the reduced maximally helicity violating (MHV) tree-level gravity amplitude in this theory. Remarkably, this amplitude has a compact determinantal formula originally due to Hodges [9 ] . Let n ≥ 5 n \ge5 n ≥ 5 , and let λ i , λ ~ i ∈ C 2 \lambda_i, \widetilde{\lambda}_i \in\mathbb{C}^2 λ i , λ i ∈ C 2 be the columns of two rank-two spinor matrices satisfying momentum conservation
∑ i = 1 n λ i λ ~ i T = 0. \sum_{i=1}^{n} \lambda_i \widetilde{\lambda}_i^{\mathsf{T}}=0. i = 1 ∑ n λ i λ i T = 0.
Write p i j = ⟨ i j ⟩ : = det ( λ i , λ j ) p_{ij}=\langle ij\rangle:=\det(\lambda_i,\lambda_j) p ij = ⟨ ij ⟩ := det ( λ i , λ j ) and q i j = [ i j ] : = det ( λ ~ i , λ ~ j ) q_{ij}=[ij]:=\det(\widetilde{\lambda}_i,\widetilde{\lambda}_j) q ij = [ ij ] := det ( λ i , λ j ) , for the Plücker coordinates on Gr ( 2 , n ) \operatorname{Gr}(2,n) Gr ( 2 , n ) . Assume that p i j ≠ 0 p_{ij}\ne0 p ij = 0 for all i < j i<j i < j . The Hodges matrix is the symmetric n × n n\times n n × n matrix Φ \Phi Φ with entries
Φ i j = { q i j p i j , i ≠ j , − ∑ k ≠ i q i k p i k ⟨ k x ⟩ ⟨ k y ⟩ ⟨ i x ⟩ ⟨ i y ⟩ , i = j , \Phi_{ij}=
\begin{cases}
\dfrac{q_{ij}}{p_{ij}}, & i\ne j,\\[6pt]
-\sum_{k\ne i}\dfrac{q_{ik}}{p_{ik}}\dfrac{\langle kx\rangle\langle ky\rangle}{\langle ix\rangle\langle iy\rangle}, & i=j,
\end{cases} Φ ij = ⎩ ⎨ ⎧ p ij q ij , − ∑ k = i p ik q ik ⟨ i x ⟩ ⟨ i y ⟩ ⟨ k x ⟩ ⟨ k y ⟩ , i = j , i = j ,
where x , y ∈ C 2 x,y\in\mathbb{C}^2 x , y ∈ C 2 are reference spinors chosen so that the denominators are nonzero. Hodges’ formula gives the reduced MHV gravity amplitude then as
M ~ n : = ± det Φ i j k r s t p r s p r t p s t p i j p i k p j k , \widetilde{M}_n:=\pm\frac{\det\Phi^{rst}_{ijk}}{p_{rs}p_{rt}p_{st}p_{ij}p_{ik}p_{jk}}, M n := ± p r s p r t p s t p ij p ik p j k det Φ ij k r s t ,
where i < j < k i<j<k i < j < k and r < s < t r<s<t r < s < t , and Φ i j k r s t \Phi^{rst}_{ijk} Φ ij k r s t is obtained by deleting rows { i , j , k } \{i,j,k\} { i , j , k } and columns labeled by { r , s , t } \{r,s,t\} { r , s , t } (the correct sign in front depends on the chosen indices of rows and columns); see also [16 ] . By the results in [7 ] , pairs of spinor brackets ( p i j , q i j ) (p_{ij},q_{ij}) ( p ij , q ij ) satisfying momentum conservation and the Plücker relations, form the spinor-helicity variety SH ( 2 , n , 0 ) \operatorname{SH}(2,n,0) SH ( 2 , n , 0 ) , which is in turn isomorphic to the two-step flag variety F n : = Fl ( 2 , n − 2 ; n ) \mathcal{F}_n:=\operatorname{Fl}(2,n-2;n) F n := Fl ( 2 , n − 2 ; n ) . Hence M ~ n \widetilde{M}_n M n is a rational expression on F n \mathcal{F}_n F n . It has been a long-standing open issue to find the geometric origin of Hodges’ formula. Our first main result is to give such a geometric interpretation by an intrinsic vector bundle construction on F n \mathcal{F}_n F n .
where a priori N n N_n N n is a rational function on S H ( 2 , n , 0 ) \mathrm{SH}(2,n,0) SH ( 2 , n , 0 ) . Upon defining on the coordinate ring Q n Q_n Q n of S H ( 2 , n , 0 ) \mathrm{SH}(2,n,0) SH ( 2 , n , 0 ) a physically-motivated Z n + 2 \mathbb{Z}^{n+2} Z n + 2 -valued multidegree, in [10 ] the authors prove that N n N_n N n is a polynomial class in Q n Q_n Q n of a specific multidegree d ( n ) ∈ Z n + 2 d(n) \in\mathbb{Z}^{n+2} d ( n ) ∈ Z n + 2 . They conjectured that N n N_n N n is the unique (up to scaling) polynomial of multidegree d ( n ) d(n) d ( n ) being in the ideal of Q n Q_n Q n given by the intersections of all monomial ideals ( p i j , q i j ) ⊂ Q n (p_{ij},q_{ij}) \subset Q_n ( p ij , q ij ) ⊂ Q n . We recently disproved this conjecture for every n ≥ 7 n \ge7 n ≥ 7 [12 ] .
We will show that the local expressions for Hodges matrix Φ \Phi Φ glue to a rational map from F n \mathcal{F}_n F n to the projectivized space of symmetric n × n n \times n n × n matrices P ( Sym C ( n ) ) \mathbb{P}(\operatorname{Sym}_{\mathbb{C}}(n)) P ( Sym C ( n )) :
H : F n ⇢ P ( Sym C ( n ) ) , ( L , W ) ⟼ [ Φ ( L , W ) ] . H:\mathcal{F}_n \dashrightarrow\mathbb{P}(\operatorname{Sym}_{\mathbb{C}}(n)),\qquad(L,W)\longmapsto[\Phi(L,W)]. H : F n ⇢ P ( Sym C ( n )) , ( L , W ) ⟼ [ Φ ( L , W )] .
which we will call the Hodges map . Since it is also well known that Hodges matrix is generically of rank n − 3 n-3 n − 3 , see [9 ] , Hodges map H H H induces another rational map
H ~ : F n ⇢ Σ ⊂ P ( Sym C ( n ) ) × Gr ( 3 , n ) \widetilde{H}:\mathcal{F}_n\dashrightarrow\Sigma\subset\mathbb{P}(\operatorname{Sym}_{\mathbb{C}}(n))\times\operatorname{Gr}(3,n) H : F n ⇢ Σ ⊂ P ( Sym C ( n )) × Gr ( 3 , n )
by sending ( L , W ) ↦ ( [ Φ ( L , W ) ] , K ( L , W ) ) (L,W)\mapsto([\Phi(L,W)],\mathcal{K}_{(L,W)}) ( L , W ) ↦ ([ Φ ( L , W )] , K ( L , W ) ) , where K ( L , W ) \mathcal{K}_{(L,W)} K ( L , W ) is the kernel of Φ ( L , W ) \Phi(L,W) Φ ( L , W ) , and Σ \Sigma Σ is the incidence variety defined as
Σ : = { ( [ A ] , K ) ∈ P ( Sym C ( n ) ) × Gr ( 3 , n ) ∣ K ⊆ ker A } . \Sigma:=\left\{([A],K)\in\mathbb{P}(\operatorname{Sym}_{\mathbb{C}}(n))\times\operatorname{Gr}(3,n)\mid K\subseteq\ker A\right\}. Σ := { ([ A ] , K ) ∈ P ( Sym C ( n )) × Gr ( 3 , n ) ∣ K ⊆ ker A } .
This incidence Σ \Sigma Σ is isomorphic to the blow-up of the variety of symmetric n n n matrices of rank at most n − 3 n-3 n − 3 in its singular locus, see Corollary 3.18, whose exceptional locus we denote by E E E . Then, our main result gives a geometric interpretation of Hodges formula.
Theorem 1.2 (The MHV gravity divisor). Let n ≥ 5 n \ge5 n ≥ 5 and F n ∘ \mathcal{F}_n^\circ F n ∘ be the Hodges locus in F n \mathcal{F}_n F n where all Plücker coordinates p i j p_{ij} p ij do not vanish.
The (open) gravity divisor is the effective Cartier divisor D H ∘ : = H ~ ∗ E D_H^\circ:=\widetilde{H}^{*}E D H ∘ := H ∗ E on the Hodges locus F n ∘ \mathcal{F}_n^\circ F n ∘ .
For choices of { i < j < k } \{i<j<k\} { i < j < k } and { r < s < t } \{r<s<t\} { r < s < t } , locally the open gravity divisor D H ∘ D_H^\circ D H ∘ coincides with the vanishing locus in F n ∘ \mathcal{F}_n^\circ F n ∘ of Hodges formula
Δ H : = ± det Φ i j k r s t p i j p i k p j k p r s p r t p s t . \Delta_H:=\pm\frac{\det\Phi_{ijk}^{rst}}{p_{ij}p_{ik}p_{jk}\,p_{rs}p_{rt}p_{st}}. Δ H := ± p ij p ik p j k p r s p r t p s t det Φ ij k r s t .
The construction of the Hodges matrix with off-diagonal entries q i j / p i j q_{ij}/p_{ij} q ij / p ij and of the Hodges map leads us to a second geometric object, which is an instance of what we will call a Hadamard ratio . Analogously to Hadamard products X ∗ Y X*Y X ∗ Y of two projective varieties, Hadamard ratios X : Y X:Y X : Y are defined by taking coordinatewise ratios of X X X and Y Y Y . In the context of MHV gravity amplitudes, we focus on the Hadamard ratio of two Grassmannians of spinors, which is the Zariski closure of the image of
Ψ n : Gr ( 2 , n ) × Gr ( 2 , n ) ⇢ P N , ( q i j , p i j ) i < j ⟼ [ q i j p i j ] i < j . \Psi_n:\operatorname{Gr}(2,n)\times\operatorname{Gr}(2,n)\dashrightarrow\mathbb{P}^{N},\qquad(q_{ij},p_{ij})_{i<j}\longmapsto\left[\frac{q_{ij}}{p_{ij}}\right]_{i<j}. Ψ n : Gr ( 2 , n ) × Gr ( 2 , n ) ⇢ P N , ( q ij , p ij ) i < j ⟼ [ p ij q ij ] i < j .
where N = ( n 2 ) − 1 N=\binom{n}{2}-1 N = ( 2 n ) − 1 , and the ratios q i j / p i j q_{ij}/p_{ij} q ij / p ij are precisely the off-diagonal entries of Hodges matrix Φ \Phi Φ . This notion generalizes in particular the idea of reciprocal spaces, which have attracted attention in recent years; see e.g. [11 ] . For n = 5 n=5 n = 5 , with coordinates on P 9 \mathbb{P}^{9} P 9 denoted by z i j z_{ij} z ij , Gr ( 2 , 5 ) : Gr ( 2 , 5 ) \operatorname{Gr}(2,5):\operatorname{Gr}(2,5) Gr ( 2 , 5 ) : Gr ( 2 , 5 ) is a 3 3 3 -dimensional hypersurface cut out by the classical pentad
\operatorname{pen}_{12345}={}&z_{12}z_{13}z_{24}z_{35}z_{45}-z_{12}z_{13}z_{25}z_{34}z_{45}-z_{12}z_{14}z_{23}z_{35}z_{45}+z_{12}z_{14}z_{25}z_{34}z_{35}\\
&+z_{12}z_{15}z_{23}z_{34}z_{45}-z_{12}z_{15}z_{24}z_{34}z_{35}+z_{13}z_{14}z_{23}z_{25}z_{45}-z_{13}z_{14}z_{24}z_{25}z_{35}\\
&-z_{13}z_{15}z_{23}z_{24}z_{45}+z_{13}z_{15}z_{24}z_{25}z_{34}+z_{14}z_{15}z_{23}z_{24}z_{35}-z_{14}z_{15}z_{23}z_{25}z_{34},
appearing in algebraic statistics [5 ] . This coincidence suggests the Hadamard ratio Gr ( 2 , n ) : Gr ( 2 , n ) \operatorname{Gr}(2,n) : \operatorname{Gr}(2,n) Gr ( 2 , n ) : Gr ( 2 , n ) should be viewed as a higher pentad variety; hence, we refer to it as the pentad variety Pen n : = Gr ( 2 , n ) : Gr ( 2 , n ) ⊂ P N \operatorname{Pen}_n := \operatorname{Gr}(2,n) : \operatorname{Gr}(2,n) \subset\mathbb{P}^N Pen n := Gr ( 2 , n ) : Gr ( 2 , n ) ⊂ P N . Moreover, we will see in Proposition 4.10 that, for generic data and up to a normalization factor, the pullback of any 2 × 2 2 \times2 2 × 2 minor of the symmetric 5 × 5 5 \times5 5 × 5 matrix Z 5 Z_5 Z 5 with off-diagonal entries q i j / p i j q_{ij}/p_{ij} q ij / p ij and diagonal entries filled in such that rank Z 5 ≤ 2 \operatorname{rank} Z_5 \le2 rank Z 5 ≤ 2 returns the MHV gravity amplitude M ‾ 5 \overline{M}_5 M 5 . This connects the study of pentad varieties with that of these amplitudes.
The remainder of our work is to study the pentad variety from an algebro-geometric point of view. We investigate its dimension, defining equations, and its connection to further combinatorial algebraic varieties. Let J n p J_n^p J n p be the ideal in the polynomial ring C [ z i j ∣ 1 ≤ i < j ≤ n ] \mathbb{C}[z_{ij} \mid1 \le i < j \le n] C [ z ij ∣ 1 ≤ i < j ≤ n ] generated by all pentad constraints pen σ \operatorname{pen}_{\sigma} pen σ on five-element subsets σ \sigma σ of [ n ] : = { 1 , … , n } [n] := \{1,\ldots,n\} [ n ] := { 1 , … , n } . Moreover, let I n p I_n^p I n p be defined as the saturation of J n p J_n^p J n p with respect to the product ∏ i < j z i j \prod_{i<j} z_{ij} ∏ i < j z ij of the N + 1 N+1 N + 1 homogeneous coordinates on P N \mathbb{P}^N P N . Then, our first main result about Pen n \operatorname{Pen}_n Pen n can be summarized as follows.
Theorem 1.3. The pentad variety Pen n ⊂ P N \operatorname{Pen}_n \subset\mathbb{P}^N Pen n ⊂ P N is irreducible of dimension 3 n − 7 3n - 7 3 n − 7 with prime vanishing ideal I n p \sqrt{I_n^p} I n p . In particular,
V ( I n p ) = V ( I n p ) = Gr ( 2 , n ) : Gr ( 2 , n ) = Pen n . \operatorname{V}\left(\sqrt{I_n^p}\right) = \operatorname{V}(I_n^p) = \operatorname{Gr}(2,n) : \operatorname{Gr}(2,n) = \operatorname{Pen}_n. V ( I n p ) = V ( I n p ) = Gr ( 2 , n ) : Gr ( 2 , n ) = Pen n . Moreover, for n ≤ 10 n \le10 n ≤ 10 , the ideal I n p I_n^p I n p is radical, hence I n p I_n^p I n p is the prime vanishing ideal.
Our third major object arises from the notion of little-group scaling invariant variables, which are natural for the scattering amplitudes. In our setup this little group scaling takes the following form: Write the columns of generic representatives of the two Grassmannian Gr ( 2 , n ) \operatorname{Gr}(2,n) Gr ( 2 , n ) as ν i ( 1 , x i ) T \nu_i(1,x_i)^T ν i ( 1 , x i ) T and μ i ( 1 , y i ) T \mu_i(1,y_i)^T μ i ( 1 , y i ) T with ν i , μ i ∈ C × \nu_i,\mu_i \in\mathbb{C}^{\times} ν i , μ i ∈ C × , respectively. Their Hadamard ratios are
z i j = q i j p i j = μ i μ j ν i ν j y j − y i x j − x i = t i t j y j − y i x j − x i , where t i : = μ i ν i . z_{ij} = \frac{q_{ij}}{p_{ij}} = \frac{\mu_i\mu_j}{\nu_i\nu_j}\frac{y_j-y_i}{x_j-x_i} = t_i t_j \frac{y_j-y_i}{x_j-x_i}, \qquad\text{where } t_i := \frac{\mu_i}{\nu_i}. z ij = p ij q ij = ν i ν j μ i μ j x j − x i y j − y i = t i t j x j − x i y j − y i , where t i := ν i μ i .
The little group scalings t i t j t_i t_j t i t j parametrize the dense torus of the toric variety X Δ ⊂ P N X_{\Delta} \subset\mathbb{P}^N X Δ ⊂ P N associated to the second hypersimplex Δ : = Conv { e i + e j ∣ 1 ≤ i < j ≤ n } \Delta:= \operatorname{Conv}\{e_i + e_j \mid1 \le i < j \le n\} Δ := Conv { e i + e j ∣ 1 ≤ i < j ≤ n } . The factors ( y j − y i ) / ( x j − x i ) (y_j-y_i)/(x_j-x_i) ( y j − y i ) / ( x j − x i ) are slopes determined by points ( x i , y i ) ∈ C 2 (x_i,y_i) \in\mathbb{C}^2 ( x i , y i ) ∈ C 2 . This connects the pentad variety to the affine slope variety S n S_n S n , as first defined in [14 ] , given as the Zariski closure in C N + 1 \mathbb{C}^{N+1} C N + 1 with N = ( n 2 ) − 1 N = \binom{n}{2} - 1 N = ( 2 n ) − 1 of the slope vectors of n n n points in the complex plane with x i ≠ x j x_i \ne x_j x i = x j for i ≠ j i \ne j i = j . The following result makes this connection explicit.
Corollary 1.4. The pentad variety Pen n \operatorname{Pen}_n Pen n is given by the Hadamard product of the projectivization of the affine slope variety P ( S n ) \mathbb{P}(S_n) P ( S n ) with X Δ X_{\Delta} X Δ . Put in symbols,
Pen n = P ( S n ) ∗ X Δ , \operatorname{Pen}_n = \mathbb{P}(S_n) * X_{\Delta}, Pen n = P ( S n ) ∗ X Δ , as subvarieties of P N \mathbb{P}^N P N .
Remark 1.5. The affine slope variety can also be defined by polynomials associated to graphs, see [14 ] , thereby reducing the study of the pentad variety to a problem in combinatorial commutative algebra. This allows us in § 5.2 to use these polynomials to study the defining equations of pentad varieties.
Lastly, we show that the slope variety has an interpretation in terms of phirotopes , the complex phase analogues of chirotopes introduced in [3 ] . A realizable rank-two phirotope records the phases of the 2 × 2 2 \times2 2 × 2 determinants of a complex vector configuration Z 0 , … , Z n ∈ C 2 Z_0,\ldots,Z_n \in\mathbb{C}^2 Z 0 , … , Z n ∈ C 2 . For a uniform phirotope these phases are nonzero on every pair of distinct labels i ≠ j i \ne j i = j . For uniform realizable phirotopes, these phases can be interpreted, up to normalization, as slope vectors of affine points, hence connecting the affine slope variety to phirotopes. The following result makes this precise.
Proposition 1.6. For a uniform rank 2 phirotope φ \varphi φ with index set E = { 0 , 1 , … , n } E = \{0,1,\ldots,n\} E = { 0 , 1 , … , n } , denote its normalized phases by φ ^ ( i , j ) : = φ ( i , j ) / ( φ ( 0 , i ) φ ( 0 , j ) ) \widehat{\varphi}(i,j) := \varphi(i,j)/(\varphi(0,i)\varphi(0,j)) φ ( i , j ) := φ ( i , j ) / ( φ ( 0 , i ) φ ( 0 , j )) for all $1 \le i < j \le n$.
Let φ ^ 2 : = ( φ ^ ( 1 , 2 ) 2 , … , φ ^ ( n − 1 , n ) 2 ) ∈ C ( n 2 ) \widehat{\varphi}^{2} := (\widehat{\varphi}(1,2)^2,\ldots,\widehat{\varphi}(n-1,n)^2) \in\mathbb{C}^{\binom{n}{2}} φ 2 := ( φ ( 1 , 2 ) 2 , … , φ ( n − 1 , n ) 2 ) ∈ C ( 2 n ) be the vector of their squares. Define R n ⊂ C ( n 2 ) R_n \subset\mathbb{C}^{\binom{n}{2}} R n ⊂ C ( 2 n ) as the set of all such points obtained from uniform realizable rank 2 phirotopes φ \varphi φ on E E E , that is
R n : = { φ ^ 2 ∣ φ is a uniform realizable rank 2 phirotope on E } ⊂ C ( n 2 ) . R_n := \left\{\widehat{\varphi}^{2} \mid\varphi\text{ is a uniform realizable rank 2 phirotope on }E\right\} \subset\mathbb{C}^{\binom{n}{2}}. R n := { φ 2 ∣ φ is a uniform realizable rank 2 phirotope on E } ⊂ C ( 2 n ) . Then, the complex Zariski closure of R n R_n R n coincides with the affine slope variety S n S_n S n .
Structure of the paper. Section 2 2 2 recalls the spinor-helicity formalism, and the identification of the spinor-helicity variety with the two-step flag variety F n F_n F n . Moreover, we introduce the notion of Hadamard ratios. In Section 3 3 3 , we put the Hodges matrix in the context of vector bundles. We start by constructing the Hodges bundle, and a generalized Hodges matrix, define the Hodges map, and identify the associated vanishing determinant with the MHV gravity divisor, leading to Theorem 1.2 . Section 4 4 4 studies the Hadamard ratio map and the resulting pentad varieties from an algebro-geometric perspective. The proof of the main result will use the connection of pentad varieties with the affine slope varieties (see Theorem 1.3 and Corollary 1.4 ). We end this section by showcasing another surprising incarnation of the affine slope variety as the locus of phases of realizable (uniform) phirotopes.
Acknowledgments. We are grateful for several discussions with Bernd Sturmfels, who also originally pointed out the observation that the pentad constraint gives the Hadamard ratio Gr ( 2 , 5 ) : Gr ( 2 , 5 ) \operatorname{Gr}(2,5) : \operatorname{Gr}(2,5) Gr ( 2 , 5 ) : Gr ( 2 , 5 ) , as well as the potential connection to the phirotopes. Moreover, we are very grateful to Ben Hollering, whose help and insights into the computational aspects of this work have been invaluable. We thank Hannah Friedman for pointing out the connection formalized in Proposition 4.26. Lastly, we thank Dario Antolini for a helpful discussion on Hadamard products. J. K. is grateful for the support received from the Institute for Advanced Study and an anonymous donor. H.Y. wishes to express sincere gratitude to the Max Planck Institute for Mathematics in the Sciences (MPI MIS) for its hospitality during the preparation of this work. E.V. and H.Y. are members of the GNSAGA group of INdAM. E.V. and H.Y. acknowledge that the preparation of this article was partially funded by the Italian Ministry of University and Research, through the project Applied Algebraic Geometry of Tensors , PRIN 2022 Protocol no. 2022NBN7TL.
AI Disclosure. AI was used as a search engine and for proofreading purposes of drafts of this work. Moreover, it was used for the implication (iii)⇒ \Rightarrow ⇒ (i) in Corollary 4.31, and for the pairing in (iii) of Lemma 3.19 . The text of the article is entirely written by the authors, and they take full responsibility for its content, including all mathematical results.
Preliminaries
In this section we introduce the spinor formalism, which is the language most frequently used in the study of MHV gravity scattering amplitudes.
We start by fixing two copies of the Grassmannian Gr ( 2 , n ) \operatorname{Gr}(2,n) Gr ( 2 , n ) with Plücker coordinates denoted by ⟨ i j ⟩ \langle ij\rangle ⟨ ij ⟩ and $[ij]w h e r e $ 1 ≤ i < j ≤ n where \$1 \le i < j \le n w h er e $1 ≤ i < j ≤ n , respectively. So in matrix coordinates, these are $2 \times 2$ minors of the corresponding 2 × n 2 \times n 2 × n matrices representing points in the Grassmannian. Let R n : = C [ [ i j ] , ⟨ i j ⟩ ∣ 1 ≤ i < j ≤ n ] R_n := \mathbb{C}[[ij],\langle ij\rangle\mid1 \le i < j \le n] R n := C [[ ij ] , ⟨ ij ⟩ ∣ 1 ≤ i < j ≤ n ] be the polynomial ring generated by these Plücker coordinates. The variables [ i j ] [ij] [ ij ] and ⟨ i j ⟩ \langle ij\rangle ⟨ ij ⟩ are called spinor variables . Let p i ∈ C 4 p_i \in\mathbb{C}^4 p i ∈ C 4 be the 4-complex momentum of a massless particle i i i . We say a particle is massless if p i ⋅ p i = p i 2 = 0 p_i \cdot p_i = p_i^2 = 0 p i ⋅ p i = p i 2 = 0 , where the product is the Lorentzian inner product on C 4 \mathbb{C}^4 C 4 . One can define an isomorphism between C 4 \mathbb{C}^4 C 4 and 2 × 2 2 \times2 2 × 2 complex matrices p i ↦ P i p_i \mapsto P_i p i ↦ P i such that the vanishing of the inner product p i 2 = 0 p_i^2 = 0 p i 2 = 0 is equivalent to the condition det ( P i ) = 0 \det(P_i) = 0 det ( P i ) = 0 . Since P i P_i P i is a rank-one matrix, one can write P i = λ i λ ~ i P_i = \lambda_i\widetilde{\lambda}_i P i = λ i λ i , for column and row vectors λ i , λ ~ i ∈ C 2 \lambda_i,\widetilde{\lambda}_i \in\mathbb{C}^2 λ i , λ i ∈ C 2 . The spinor variable [ i j ] [ij] [ ij ] is the determinant of the 2 × 2 2 \times2 2 × 2 matrix whose rows are λ ~ i , λ ~ j \widetilde{\lambda}_i,\widetilde{\lambda}_j λ i , λ j ; the spinor variable ⟨ i j ⟩ \langle ij\rangle ⟨ ij ⟩ is the determinant of the 2 × 2 2 \times2 2 × 2 matrix whose columns are λ i , λ j \lambda_i,\lambda_j λ i , λ j .
Spinor-Helicity Variety
From now on, we let p i j = ⟨ i j ⟩ p_{ij} = \langle ij\rangle p ij = ⟨ ij ⟩ and q i j = [ i j ] q_{ij} = [ij] q ij = [ ij ] for $1 \le i < j \le n$ and denote the polynomial ring in these variables also by R n = C [ p i j , q i j ∣ 1 ≤ i < j ≤ n ] R_n = \mathbb{C}[p_{ij},q_{ij} \mid1 \le i < j \le n] R n = C [ p ij , q ij ∣ 1 ≤ i < j ≤ n ] . We define the ideal L n ⊂ R n L_n \subset R_n L n ⊂ R n generated by the Plücker relations
p i j p k ℓ − p i k p j ℓ + p i ℓ p j k , p_{ij}p_{k\ell} - p_{ik}p_{j\ell} + p_{i\ell}p_{jk}, p ij p k ℓ − p ik p j ℓ + p i ℓ p j k ,
q i j q k ℓ − q i k q j ℓ + q i ℓ q j k , q_{ij}q_{k\ell} - q_{ik}q_{j\ell} + q_{i\ell}q_{jk}, q ij q k ℓ − q ik q j ℓ + q i ℓ q j k ,
for all $1 \le i < j < k < \ell \le n$, together with the momentum-conservation equations
M a b : = ∑ r = 1 n p a r q r b = 0 , 1 ≤ a , b ≤ n . M_{ab} := \sum_{r=1}^{n} p_{ar}q_{rb} = 0,\qquad1 \le a,b \le n. M ab := r = 1 ∑ n p a r q r b = 0 , 1 ≤ a , b ≤ n .
The ring Q n = R n / L n Q_n = R_n/L_n Q n = R n / L n is the coordinate ring of the spinor-helicity variety SH ( 2 , n , 0 ) \operatorname{SH}(2,n,0) SH ( 2 , n , 0 ) of El Maazouz-Pfister-Sturmfels [7 ] .
Two-step Flag Varieties
Lastly, we want to explicitly connect this construction with a flag variety. Let V = C n V = \mathbb{C}^n V = C n with its standard basis e 1 , … , e n e_1,\ldots,e_n e 1 , … , e n and its standard nondegenerate symmetric bilinear form. We look at the product of two Grassmannians Gr ( 2 , n ) × Gr ( 2 , n ) \operatorname{Gr}(2,n) \times\operatorname{Gr}(2,n) Gr ( 2 , n ) × Gr ( 2 , n ) . A point of the first Grassmannian is a 2-dimensional plane L ⊂ V L \subset V L ⊂ V , with Plücker coordinates p i j p_{ij} p ij . A point of the second factor Grassmannian is another 2-dimensional plane M ⊂ V M \subset V M ⊂ V , with Plücker coordinates q i j q_{ij} q ij .
Choose two 2 × n 2 \times n 2 × n matrices X X X and Y Y Y whose row spans are L L L and M M M . Then the momentum-conservation equations are the vanishing matrix multiplication X Y T = 0 XY^{T} = 0 X Y T = 0 . Equivalently, every vector in L L L is orthogonal to every vector in M M M . Letting W = M ⊥ W = M^{\perp} W = M ⊥ , the condition L ⊥ M L \perp M L ⊥ M becomes L ⊂ W L \subset W L ⊂ W . Hence a spinor-helicity point is the same datum as a two-step flag L ⊂ W ⊂ V L \subset W \subset V L ⊂ W ⊂ V , dim L = 2 \dim L = 2 dim L = 2 , dim W = n − 2 \dim W = n - 2 dim W = n − 2 . One can show that this identification yields an isomorphism between the spinor-helicity and the two-step flag variety F n = Fl ( 2 , n − 2 ; n ) \mathcal{F}_n = \operatorname{Fl}(2,n-2;n) F n = Fl ( 2 , n − 2 ; n ) of partial flags L ⊂ W ⊂ V L \subset W \subset V L ⊂ W ⊂ V [7 ] .
Hadamard Ratios
In this section we introduce Hadamard ratios, which are a natural extensions of Hadamard products. Let X X X and Y Y Y be two irreducible subvarieties of P n \mathbb{P}^n P n with coordinates labeled by x i x_i x i and y i y_i y i respectively. Then, the Hadamard product X ∗ Y X * Y X ∗ Y is defined as the Zariski closure of the image of the map
X × Y ⇢ P n , ( x i , y i ) ↦ x i y i . X \times Y \dashrightarrow\mathbb{P}^n,\quad(x_i,y_i) \mapsto x_i y_i. X × Y ⇢ P n , ( x i , y i ) ↦ x i y i .
Hadamard products are well-studied objects, see for example [1 ] . In this work, we take a slightly different route.
Definition 2.1. Let X X X and Y Y Y be two irreducible subvarieties of P n \mathbb{P}^n P n , with coordinates labeled by x i x_i x i and y i y_i y i , respectively. Then, we define the Hadamard ratio map by
Ψ : X × Y ⇢ P n , ( x i , y i ) ↦ x i / y i , \Psi: X \times Y \dashrightarrow\mathbb{P}^n,\quad(x_i,y_i) \mapsto x_i/y_i, Ψ : X × Y ⇢ P n , ( x i , y i ) ↦ x i / y i , and refer to the Zariski closure of the image im Ψ ‾ \overline{\operatorname{im}\Psi} im Ψ as the Hadamard ratio X : Y X:Y X : Y between X X X and Y Y Y .
Example 2.2. Suppose Y = L ⊂ P n Y = L \subset\mathbb{P}^n Y = L ⊂ P n is a d d d -dimensional linear subspace. Then, for the variety p = [ 1 : 1 : ⋯ : 1 ] ∈ P n p = [1:1:\cdots:1] \in\mathbb{P}^n p = [ 1 : 1 : ⋯ : 1 ] ∈ P n , the Hadamard ratio p : L p:L p : L is precisely the reciprocal linear space L − 1 L^{-1} L − 1 , which has attracted much attention in the literature; see e.g. [11 ] . ◇
Denote by T n ⊂ P n \mathbb{T}^n \subset\mathbb{P}^n T n ⊂ P n the algebraic torus , which is the set of all points in P n \mathbb{P}^n P n where none of the coordinates vanishes. Then, we get natural extensions of a result in [1 Section 5] .
Lemma 2.3. Let X X X and Y Y Y be two irreducible complex subvarieties of P n \mathbb{P}^n P n and suppose that X ∘ : = X ∩ T n ≠ ∅ X^\circ:= X \cap\mathbb{T}^n \ne\varnothing X ∘ := X ∩ T n = ∅ and Y ∘ : = Y ∩ T n ≠ ∅ Y^\circ:= Y \cap\mathbb{T}^n \ne\varnothing Y ∘ := Y ∩ T n = ∅ . Then tropicalization of the Hadamard ratio X : Y X:Y X : Y is the Minkowski difference of the tropicalizations of X X X and Y Y Y . Put in symbols
trop ( ( X : Y ) ) = trop X − trop Y . \operatorname{trop}((X:Y)) = \operatorname{trop} X - \operatorname{trop} Y. trop (( X : Y )) = trop X − trop Y . Proof. The Hadamard ratio map Ψ \Psi Ψ is a monomial map, hence by [15 Corollary 3.2.13] the tropicalization of the image of Ψ \Psi Ψ is given by image of trop ( X ∘ × Y ∘ ) \operatorname{trop}(X^\circ\times Y^\circ) trop ( X ∘ × Y ∘ ) under the tropicalization of Ψ \Psi Ψ . The latter is given by coordinate-wise subtraction, hence the statement. □ \square □
Hence, [1 Proposition 5.4] directly generalizes, which we record here in a slightly weaker form.
Lemma 2.4. Let X X X and Y Y Y be two irreducible subvarieties in P n \mathbb{P}^n P n . Denote by H ⊂ ( C ∗ ) n + 1 / C ∗ H \subset(\mathbb{C}^*)^{n+1}/\mathbb{C}^* H ⊂ ( C ∗ ) n + 1 / C ∗ the largest subtorus acting on both X X X and Y Y Y . Then
dim ( X : Y ) ≤ dim ( X ) + dim ( Y ) − dim ( H ) . \dim(X:Y) \le\dim(X) + \dim(Y) - \dim(H). dim ( X : Y ) ≤ dim ( X ) + dim ( Y ) − dim ( H ) .
Hodges Map and the Gravity Divisor
The goal of this section is to prove our main result, Theorem 1.2 , which gives the geometric interpretation of Hodges formula for the reduced MHV gravity scattering amplitude M ~ n \widetilde{M}_n M n . We will do this by introducing an intrinsic vector bundle construction on the two-step flag variety F n \mathcal{F}_n F n .
Hodges Matrix and Gravity Scattering Amplitudes
We start by recalling the precise definitions of M ~ n \widetilde{M}_n M n and the Hodges matrix. For an arbitrary spinor z ∈ C 2 z \in\mathbb{C}^2 z ∈ C 2 (not necessarily one of the λ i \lambda_i λ i ’s from § 2 ) and an index i i i , we keep the notation ⟨ j z ⟩ \langle jz\rangle ⟨ j z ⟩ to avoid confusion with the Plücker coordinates p i j p_{ij} p ij ’s. For i ∈ [ n ] i \in[n] i ∈ [ n ] let λ i ∈ C 2 \lambda_i \in\mathbb{C}^2 λ i ∈ C 2 and λ ~ i ∈ C 2 \widetilde{\lambda}_i \in\mathbb{C}^2 λ i ∈ C 2 be spinors satisfying momentum conservation : for every fixed index i i i and every spinor z ∈ C 2 z \in\mathbb{C}^2 z ∈ C 2 ,
∑ j = 1 n q i j ⟨ j z ⟩ = 0. \sum_{j=1}^{n} q_{ij}\langle jz\rangle= 0. j = 1 ∑ n q ij ⟨ j z ⟩ = 0.
Then we can define Hodges matrix.
Definition 3.1. Choose two auxiliary spinors x , y ∈ C 2 x,y \in\mathbb{C}^2 x , y ∈ C 2 such that ⟨ i x ⟩ ⟨ i y ⟩ ≠ 0 \langle ix\rangle\langle iy\rangle\ne0 ⟨ i x ⟩ ⟨ i y ⟩ = 0 for all i i i . The Hodges matrix is the symmetric n × n n \times n n × n matrix Φ = Φ ( x , y ) \Phi= \Phi(x,y) Φ = Φ ( x , y ) with entries
Φ i j : = q i j p i j , for i ≠ j , \Phi_{ij} := \frac{q_{ij}}{p_{ij}}, \qquad\text{for } i \ne j, Φ ij := p ij q ij , for i = j , and
Φ i i : = − ∑ j ≠ i q i j p i j ⟨ j x ⟩ ⟨ j y ⟩ ⟨ i x ⟩ ⟨ i y ⟩ . \Phi_{ii} := -\sum_{j\ne i} \frac{q_{ij}}{p_{ij}}\frac{\langle jx\rangle\langle jy\rangle}{\langle ix\rangle\langle iy\rangle}. Φ ii := − j = i ∑ p ij q ij ⟨ i x ⟩ ⟨ i y ⟩ ⟨ j x ⟩ ⟨ j y ⟩ .
Finally, we can introduce the scattering amplitude of our main interest.
Definition 3.2. The reduced n n n -point MHV gravity amplitude M ~ n \widetilde{M}_n M n in N = 8 \mathcal{N}=8 N = 8 supergravity (SUGRA) is the rational section on the spinor helicity variety F n \mathcal{F}_n F n given by
M ~ n : = det Φ i j k r s t p i j p i k p j k p r s p r t p s t , \widetilde{M}_n := \frac{\det\Phi^{rst}_{ijk}}{p_{ij}p_{ik}p_{jk}p_{rs}p_{rt}p_{st}}, M n := p ij p ik p j k p r s p r t p s t det Φ ij k r s t , where Φ i j k r s t \Phi^{rst}_{ijk} Φ ij k r s t is the ( n − 3 ) (n-3) ( n − 3 ) square matrix obtained from the Hodges matrix Φ \Phi Φ by deleting the rows i , j , k i,j,k i , j , k and columns r , s , t r,s,t r , s , t .
Example 3.3. The 5-point scattering amplitude M ~ 5 \widetilde{M}_5 M 5 is obtained from the 5 × 5 5 \times5 5 × 5 Hodges matrix:
Φ 5 = ( Φ 11 q 12 p 12 q 13 p 13 q 14 p 14 q 15 p 15 q 12 p 12 Φ 22 q 23 p 23 q 24 p 24 q 25 p 25 q 13 p 13 q 23 p 23 Φ 33 q 34 p 34 q 35 p 35 q 14 p 14 q 24 p 24 q 34 p 34 Φ 44 q 45 p 45 q 15 p 15 q 25 p 25 q 35 p 35 q 45 p 45 Φ 55 ) . \Phi_5 =
\begin{pmatrix}
\Phi_{11} & \frac{q_{12}}{p_{12}} & \frac{q_{13}}{p_{13}} & \frac{q_{14}}{p_{14}} & \frac{q_{15}}{p_{15}} \\
\frac{q_{12}}{p_{12}} & \Phi_{22} & \frac{q_{23}}{p_{23}} & \frac{q_{24}}{p_{24}} & \frac{q_{25}}{p_{25}} \\
\frac{q_{13}}{p_{13}} & \frac{q_{23}}{p_{23}} & \Phi_{33} & \frac{q_{34}}{p_{34}} & \frac{q_{35}}{p_{35}} \\
\frac{q_{14}}{p_{14}} & \frac{q_{24}}{p_{24}} & \frac{q_{34}}{p_{34}} & \Phi_{44} & \frac{q_{45}}{p_{45}} \\
\frac{q_{15}}{p_{15}} & \frac{q_{25}}{p_{25}} & \frac{q_{35}}{p_{35}} & \frac{q_{45}}{p_{45}} & \Phi_{55}
\end{pmatrix}. Φ 5 = Φ 11 p 12 q 12 p 13 q 13 p 14 q 14 p 15 q 15 p 12 q 12 Φ 22 p 23 q 23 p 24 q 24 p 25 q 25 p 13 q 13 p 23 q 23 Φ 33 p 34 q 34 p 35 q 35 p 14 q 14 p 24 q 24 p 34 q 34 Φ 44 p 45 q 45 p 15 q 15 p 25 q 25 p 35 q 35 p 45 q 45 Φ 55 . We obtain
M ~ 5 = − p 14 p 25 q 24 q 15 + p 24 p 15 q 14 q 25 p 12 p 13 p 23 p 14 p 24 p 34 p 15 p 25 p 35 p 45 . \widetilde{M}_5 = \frac{-p_{14}p_{25}q_{24}q_{15}+p_{24}p_{15}q_{14}q_{25}}{p_{12}p_{13}p_{23}p_{14}p_{24}p_{34}p_{15}p_{25}p_{35}p_{45}}. M 5 = p 12 p 13 p 23 p 14 p 24 p 34 p 15 p 25 p 35 p 45 − p 14 p 25 q 24 q 15 + p 24 p 15 q 14 q 25 .
Hodges bundle on the flag variety and quadrics
In this subsection, we introduce the construction intrinsic to the partial flag variety F n \mathcal{F}_n F n , formulated in the language of vector bundles, which allows us to give a geometric interpretation of the reduced MHV scattering amplitude M ~ n \widetilde{\mathcal{M}}_n M n .
Let O F n \mathcal{O}_{\mathcal{F}_n} O F n be the structure sheaf on F n \mathcal{F}_n F n , and let S ⊂ V ⊗ O F n \mathcal{S} \subset V \otimes\mathcal{O}_{\mathcal{F}_n} S ⊂ V ⊗ O F n , with V ≅ C n V \cong\mathbb{C}^n V ≅ C n be the rank-two tautological subbundle, whose fiber over ( L , W ) ∈ F n (L,W) \in\mathcal{F}_n ( L , W ) ∈ F n is L L L .
Definition 3.4. For i ∈ [ n ] i \in[n] i ∈ [ n ] , define s i ∈ H 0 ( F n , S ∨ ) s_i \in H^0(\mathcal{F}_n,\mathcal{S}^{\vee}) s i ∈ H 0 ( F n , S ∨ ) as the restriction of the i i i -th coordinate functional on C n \mathbb{C}^n C n to the 2 2 2 -plane L L L .
Remark 3.5. Choosing a basis of L L L , we may write it as span of the rows of a 2 × n 2 \times n 2 × n matrix
Λ = ( u 1 u 2 ⋯ u n v 1 v 2 ⋯ v n ) \Lambda= \begin{pmatrix} u_1 & u_2 & \cdots& u_n \\ v_1 & v_2 & \cdots& v_n \end{pmatrix} Λ = ( u 1 v 1 u 2 v 2 ⋯ ⋯ u n v n )
In these coordinates, s i = u i e 1 ∨ + v i e 2 ∨ ∈ L ∨ s_i = u_i e_1^{\vee} + v_i e_2^{\vee} \in L^{\vee} s i = u i e 1 ∨ + v i e 2 ∨ ∈ L ∨ , where e 1 e_1 e 1 and e 2 e_2 e 2 are elements from the chosen basis of L L L . Thus s i s_i s i can then be identified with the column spinor λ i = ( u i v i ) \lambda_i = \begin{pmatrix} u_i \\ v_i \end{pmatrix} λ i = ( u i v i ) .
Note that the angle bracket ⟨ i j ⟩ = p i j = s i ∧ s j ∈ H 0 ( F n , det S ∨ ) \langle ij\rangle= p_{ij} = s_i \wedge s_j \in H^0(\mathcal{F}_n,\det\mathcal{S}^{\vee}) ⟨ ij ⟩ = p ij = s i ∧ s j ∈ H 0 ( F n , det S ∨ ) . This motivates the following definition.
Definition 3.6. Define the line bundle O p ( 1 ) : = det S ∨ \mathcal{O}_p(1) := \det\mathcal{S}^{\vee} O p ( 1 ) := det S ∨ , whose sections are homogeneous polynomials of degree 1 1 1 in the Plücker coordinates p i j p_{ij} p ij , analogously for O q ( 1 ) \mathcal{O}_q(1) O q ( 1 ) . Then, define the line bundle L : = O q ( 1 ) ⊗ O p ( − 1 ) \mathcal{L} := \mathcal{O}_q(1) \otimes\mathcal{O}_p(-1) L := O q ( 1 ) ⊗ O p ( − 1 ) , such that z i j = q i j p i j z_{ij} = \frac{q_{ij}}{p_{ij}} z ij = p ij q ij are rational sections of L \mathcal{L} L . On the open set F n ∘ \mathcal{F}_n^{\circ} F n ∘ of F n \mathcal{F}_n F n where p i j ≠ 0 p_{ij} \ne0 p ij = 0 , they are regular sections of L \mathcal{L} L .
In the following, we shall restrict on an open locus of F n \mathcal{F}_n F n , the Hodges locus . This is the set F n ∘ : = { ( L , W ) ∈ F n ∣ p i j ≠ 0 for all i < j } \mathcal{F}_n^{\circ} := \{(L,W) \in\mathcal{F}_n \mid p_{ij} \ne0 \text{ for all } i < j\} F n ∘ := {( L , W ) ∈ F n ∣ p ij = 0 for all i < j } . On the Hodges locus, all off-diagonal entries z i j z_{ij} z ij of the Hodges matrix are regular sections of the line bundle L \mathcal{L} L .
Definition 3.7 (Quadratic evaluation map). There is a natural vector bundle morphism
ev 2 : Sym 2 S ⟶ O F n ⊕ n \operatorname{ev}_2 : \operatorname{Sym}^2 \mathcal{S} \longrightarrow\mathcal{O}_{\mathcal{F}_n}^{\oplus n} ev 2 : Sym 2 S ⟶ O F n ⊕ n defined fiberwise by
Q ⟼ ( Q ( s 1 ) , … , Q ( s n ) ) . Q \longmapsto\left(Q(s_1),\ldots,Q(s_n)\right). Q ⟼ ( Q ( s 1 ) , … , Q ( s n ) ) . Here Q ∈ Sym 2 L Q \in\operatorname{Sym}^2 L Q ∈ Sym 2 L is a quadric on L ∨ L^{\vee} L ∨ ; recall that s i ∈ L ∨ s_i \in L^{\vee} s i ∈ L ∨ . Denote its image by
K : = im ( ev 2 ) ⊂ O F n ⊕ n , \mathcal{K} := \operatorname{im}(\operatorname{ev}_2) \subset\mathcal{O}_{\mathcal{F}_n}^{\oplus n}, K := im ( ev 2 ) ⊂ O F n ⊕ n , where this image has constant rank.
Lemma 3.8 (Rank three on the Hodges locus). Let n ≥ 3 n \ge3 n ≥ 3 . On the Hodges locus F n ∘ \mathcal{F}_n^{\circ} F n ∘ , the evaluation map ev 2 : Sym 2 S → O F n ⊕ n \operatorname{ev}_2 : \operatorname{Sym}^2 \mathcal{S} \to\mathcal{O}_{\mathcal{F}_n}^{\oplus n} ev 2 : Sym 2 S → O F n ⊕ n is injective . Hence K ≅ Sym 2 S \mathcal{K} \cong\operatorname{Sym}^2 \mathcal{S} K ≅ Sym 2 S is a rank-three subbundle of O F n ∘ ⊕ n \mathcal{O}_{\mathcal{F}_n^{\circ}}^{\oplus n} O F n ∘ ⊕ n .
Proof. The assertion can be proven on each fiber. Fix ( L , W ) ∈ F n ∘ (L,W) \in\mathcal{F}_n^{\circ} ( L , W ) ∈ F n ∘ and let Q ∈ Sym 2 L Q \in\operatorname{Sym}^2 L Q ∈ Sym 2 L be a quadric satisfying Q ( s i ) = 0 Q(s_i) = 0 Q ( s i ) = 0 for all 1 ≤ i ≤ n 1 \le i \le n 1 ≤ i ≤ n . Since p i j = s i ∧ s j ≠ 0 p_{ij} = s_i \wedge s_j \ne0 p ij = s i ∧ s j = 0 for all i < j i < j i < j , the points [ s 1 ] , … , [ s n ] ∈ P ( L ∨ ) ≅ P 1 [s_1],\ldots,[s_n] \in\mathbb{P}(L^{\vee}) \cong\mathbb{P}^1 [ s 1 ] , … , [ s n ] ∈ P ( L ∨ ) ≅ P 1 must be pairwise distinct. Therefore, since n ≥ 3 n \ge3 n ≥ 3 and Q Q Q is a binary quadric, the vanishing at all the [ s i ] [s_i] [ s i ] implies that Q Q Q vanishes identically. Hence ev 2 \operatorname{ev}_2 ev 2 is injective on every fiber of F n ∘ \mathcal{F}_n^{\circ} F n ∘ . Clearly, Sym 2 S \operatorname{Sym}^2 \mathcal{S} Sym 2 S has rank three, and so does its image.
The usual Hodges diagonal is written using two reference spinors x , y x,y x , y . These two spinors give a quadric f x y f_{xy} f x y in Sym 2 S \operatorname{Sym}^{2}\mathcal{S} Sym 2 S defined by f x y ( s ) = x ∧ s ⊗ y ∧ s ∈ ( det S ∨ ) ⊗ 2 f_{xy}(s)=x\wedge s\otimes y\wedge s\in(\det\mathcal{S}^{\vee})^{\otimes2} f x y ( s ) = x ∧ s ⊗ y ∧ s ∈ ( det S ∨ ) ⊗ 2 , where s s s is a section of S ∨ \mathcal{S}^{\vee} S ∨ . Intrinsically, this means choosing a ( det S ∨ ) ⊗ 2 (\det\mathcal{S}^{\vee})^{\otimes2} ( det S ∨ ) ⊗ 2 -valued quadric on S ∨ \mathcal{S}^{\vee} S ∨ .
Definition 3.9 (Hodges bundle). We call the rank 3 bundle H = Sym 2 S ⊗ ( det S ∨ ) ⊗ 2 \mathcal{H}=\operatorname{Sym}^{2}\mathcal{S}\otimes(\det\mathcal{S}^{\vee})^{\otimes2} H = Sym 2 S ⊗ ( det S ∨ ) ⊗ 2 on F n \mathcal{F}_{n} F n Hodges bundle. A section f f f of H \mathcal{H} H is then a quadric on S ∨ \mathcal{S}^{\vee} S ∨ with values in the line bundle ( det S ∨ ) ⊗ 2 (\det\mathcal{S}^{\vee})^{\otimes2} ( det S ∨ ) ⊗ 2 . For such an f f f , define U f = { ( L , W ) ∣ f ( s i ) ≠ 0 for all i = 1 , … , n } ⊂ F n U_{f}=\{(L,W)\mid f(s_{i})\ne0\text{ for all }i=1,\ldots,n\}\subset\mathcal{F}_{n} U f = {( L , W ) ∣ f ( s i ) = 0 for all i = 1 , … , n } ⊂ F n .
Lemma 3.10 (The opens U f U_{f} U f cover the Hodges locus). Hodges bundles H \mathcal{H} H is globally generated. Moreover, if we denote by
C n ∘ : = { ( L , W ) ∈ F n ∣ 0 ≠ s i ∈ det L ∨ for all i } , \mathcal{C}_{n}^{\circ}:=\{(L,W)\in\mathcal{F}_{n}\mid0\ne s_{i}\in\det L^{\vee}\text{ for all }i\}, C n ∘ := {( L , W ) ∈ F n ∣ 0 = s i ∈ det L ∨ for all i } , then the open sets U f U_{f} U f , as f f f varies over the global sections H 0 ( F n , H ) H^{0}(\mathcal{F}_{n},\mathcal{H}) H 0 ( F n , H ) of H \mathcal{H} H , cover C n ∘ \mathcal{C}_{n}^{\circ} C n ∘ . In particular, they cover F n ∘ \mathcal{F}_{n}^{\circ} F n ∘ .
Proof. First observe that, because S \mathcal{S} S has rank two, there is a canonical isomorphism
S ≃ S ∨ ⊗ det S . \mathcal{S}\simeq\mathcal{S}^{\vee}\otimes\det\mathcal{S}. S ≃ S ∨ ⊗ det S .
Indeed, fiberwise, for a 2-dimensional vector space L L L , the map is
L ⟶ L ∨ ⊗ ( L ∧ L ) , v ⟼ ( w ↦ v ∧ w ) , L\longrightarrow L^{\vee}\otimes(L\wedge L),\qquad v\longmapsto(w\mapsto v\wedge w), L ⟶ L ∨ ⊗ ( L ∧ L ) , v ⟼ ( w ↦ v ∧ w ) ,
which is an isomorphism. Taking symmetric squares on both sides gives Sym 2 S ≃ Sym 2 S ∨ ⊗ ( det S ) ⊗ 2 \operatorname{Sym}^{2}\mathcal{S}\simeq\operatorname{Sym}^{2}\mathcal{S}^{\vee}\otimes(\det\mathcal{S})^{\otimes2} Sym 2 S ≃ Sym 2 S ∨ ⊗ ( det S ) ⊗ 2 . Tensoring by ( det S ∨ ) ⊗ 2 (\det\mathcal{S}^{\vee})^{\otimes2} ( det S ∨ ) ⊗ 2 , we obtain H = Sym 2 S ⊗ ( det S ∨ ) ⊗ 2 ≃ Sym 2 S ∨ \mathcal{H}=\operatorname{Sym}^{2}\mathcal{S}\otimes(\det\mathcal{S}^{\vee})^{\otimes2}\simeq\operatorname{Sym}^{2}\mathcal{S}^{\vee} H = Sym 2 S ⊗ ( det S ∨ ) ⊗ 2 ≃ Sym 2 S ∨ . Since S ⊂ V ⊗ O F n \mathcal{S}\subset V\otimes\mathcal{O}_{\mathcal{F}_{n}} S ⊂ V ⊗ O F n , dualizing gives a surjection V ∨ ⊗ O F n ↠ S ∨ V^{\vee}\otimes\mathcal{O}_{\mathcal{F}_{n}}\twoheadrightarrow\mathcal{S}^{\vee} V ∨ ⊗ O F n ↠ S ∨ and so S ∨ \mathcal{S}^{\vee} S ∨ is globally generated. Thus Sym 2 S ∨ ≃ H \operatorname{Sym}^{2}\mathcal{S}^{\vee}\simeq\mathcal{H} Sym 2 S ∨ ≃ H is globally generated too. This shows the first claim.
Next, we show that the U f U_{f} U f cover C n ∘ \mathcal{C}_{n}^{\circ} C n ∘ . Fix a point ( L , W ) ∈ C n ∘ (L,W)\in\mathcal{C}_{n}^{\circ} ( L , W ) ∈ C n ∘ . By definition of C n ∘ \mathcal{C}_{n}^{\circ} C n ∘ , the sections s 1 , … , s n ∈ L ∨ s_{1},\ldots,s_{n}\in L^{\vee} s 1 , … , s n ∈ L ∨ are all nonzero, and therefore determine finitely many points [ s 1 ] , … , [ s n ] ∈ P ( L ∨ ) ≃ P 1 [s_{1}],\ldots,[s_{n}]\in\mathbb{P}(L^{\vee})\simeq\mathbb{P}^{1} [ s 1 ] , … , [ s n ] ∈ P ( L ∨ ) ≃ P 1 . Recall that Sym 2 L = Sym 2 ( ( L ∨ ) ∨ ) \operatorname{Sym}^{2}L=\operatorname{Sym}^{2}((L^{\vee})^{\vee}) Sym 2 L = Sym 2 (( L ∨ ) ∨ ) is naturally the vector space of homogeneous quadratic forms on L ∨ L^{\vee} L ∨ . For every i i i , the condition that a quadratic form Q ∈ Sym 2 L Q\in\operatorname{Sym}^{2}L Q ∈ Sym 2 L vanishes at s i s_{i} s i , i.e. Q ( s i ) = 0 Q(s_{i})=0 Q ( s i ) = 0 , defines a two-plane in Sym 2 L \operatorname{Sym}^{2}L Sym 2 L . Hence, there is a Zariski dense open set of elements Q ∈ Sym 2 L Q\in\operatorname{Sym}^{2}L Q ∈ Sym 2 L , such that Q ( s i ) ≠ 0 Q(s_{i})\ne0 Q ( s i ) = 0 for every i = 1 , … , n i=1,\ldots,n i = 1 , … , n . Choose also a nonzero element ω ∈ ( det L ∨ ) ⊗ 2 \omega\in(\det L^{\vee})^{\otimes2} ω ∈ ( det L ∨ ) ⊗ 2 . Then
f ( L , W ) : = Q ⊗ ω ∈ Sym 2 L ⊗ ( det L ∨ ) ⊗ 2 = H ( L , W ) f_{(L,W)}:=Q\otimes\omega\in\operatorname{Sym}^{2}L\otimes(\det L^{\vee})^{\otimes2}=\mathcal{H}_{(L,W)} f ( L , W ) := Q ⊗ ω ∈ Sym 2 L ⊗ ( det L ∨ ) ⊗ 2 = H ( L , W )
satisfies f ( L , W ) ( s i ) = Q ( s i ) ω ≠ 0 f_{(L,W)}(s_{i})=Q(s_{i})\omega\ne0 f ( L , W ) ( s i ) = Q ( s i ) ω = 0 , for every i i i . Here f ( L , W ) ( s i ) ∈ ( det L ∨ ) ⊗ 2 f_{(L,W)}(s_{i})\in(\det L^{\vee})^{\otimes2} f ( L , W ) ( s i ) ∈ ( det L ∨ ) ⊗ 2 is the value of the quadratic form f ( L , W ) f_{(L,W)} f ( L , W ) at the vector s i ∈ L ∨ s_{i}\in L^{\vee} s i ∈ L ∨ . Since H \mathcal{H} H is globally generated, the evaluation map H 0 ( F n , H ) ⟶ H ( L , W ) H^{0}(\mathcal{F}_{n},\mathcal{H})\longrightarrow\mathcal{H}_{(L,W)} H 0 ( F n , H ) ⟶ H ( L , W ) is surjective. Hence there exists a global section f ∈ H 0 ( F n , H ) f\in H^{0}(\mathcal{F}_{n},\mathcal{H}) f ∈ H 0 ( F n , H ) such that f ( L , W ) = f ( L , W ) f(L,W)=f_{(L,W)} f ( L , W ) = f ( L , W ) . Consequently, we have f ( L , W ) ( s i ) ≠ 0 f(L,W)(s_{i})\ne0 f ( L , W ) ( s i ) = 0 for every i i i , and therefore ( L , W ) ∈ U f (L,W)\in U_{f} ( L , W ) ∈ U f . So the U f U_{f} U f 's cover C n ∘ \mathcal{C}_{n}^{\circ} C n ∘ . Finally, it is clear that C n ∘ ⊃ F n ∘ \mathcal{C}_{n}^{\circ}\supset\mathcal{F}_{n}^{\circ} C n ∘ ⊃ F n ∘ , because p i j ≠ 0 p_{ij}\ne0 p ij = 0 implies that s i , s j ≠ 0 s_{i},s_{j}\ne0 s i , s j = 0 . □
Remark 3.11. Note that using the sections s k ∈ H 0 ( F n , S ∨ ) s_{k}\in H^{0}(\mathcal{F}_{n},\mathcal{S}^{\vee}) s k ∈ H 0 ( F n , S ∨ ) , the momentum-conservation equations can be written as
∑ k = 1 n q i k s k = 0 for every i = 1 , … , n . \sum_{k=1}^{n}q_{ik}s_{k}=0\qquad\text{for every }i=1,\ldots,n. k = 1 ∑ n q ik s k = 0 for every i = 1 , … , n .
In coordinates this is the identity ∑ k = 1 n q i k λ k = 0 \sum_{k=1}^{n}q_{ik}\lambda_{k}=0 ∑ k = 1 n q ik λ k = 0 . Indeed, as we mentioned in § 2 , if λ k \lambda_{k} λ k denotes the k k k -th angle spinor column and λ ~ k \tilde{\lambda}_{k} λ ~ k the square spinor, then momentum conservation is ∑ k = 1 n λ k λ ~ k = 0 \sum_{k=1}^{n}\lambda_{k}\tilde{\lambda}_{k}=0 ∑ k = 1 n λ k λ ~ k = 0 . Contracting with λ ~ i \tilde{\lambda}_{i} λ ~ i on the square-spinor side gives ∑ k = 1 n q i k λ k = 0 \sum_{k=1}^{n}q_{ik}\lambda_{k}=0 ∑ k = 1 n q ik λ k = 0 .
Let f f f be a section of H \mathcal{H} H . We now define a generalized Hodges matrix.
Definition 3.12. On F n ∘ ∩ U f \mathcal{F}_n^\circ\cap U_f F n ∘ ∩ U f , define the n × n n \times n n × n symmetric matrix Φ ( f ) \Phi^{(f)} Φ ( f ) , called generalized Hodges matrix , with entries in the line bundle L \mathcal{L} L as follows:
Φ i j ( f ) : = z i j = q i j p i j , and Φ i i ( f ) : = − ∑ k ≠ i z i k f ( s k ) f ( s i ) . \Phi_{ij}^{(f)} := z_{ij} = \frac{q_{ij}}{p_{ij}}, \qquad\text{and} \qquad\Phi_{ii}^{(f)} := -\frac{\sum_{k\ne i}z_{ik}f(s_k)}{f(s_i)}. Φ ij ( f ) := z ij = p ij q ij , and Φ ii ( f ) := − f ( s i ) ∑ k = i z ik f ( s k ) .
We will need the following technical lemma in order to prove the independence of the choice of quadric f ∈ H 0 ( F n , H ) f\in H^0(\mathcal{F}_n,\mathcal{H}) f ∈ H 0 ( F n , H ) in this definition.
Lemma 3.15. Fix i i i and let h h h be any local section of Sym 2 S ⊗ E \operatorname{Sym}^2 S\otimes\mathcal{E} Sym 2 S ⊗ E , for some line bundle E \mathcal{E} E on F n \mathcal{F}_n F n . Suppose that, at a given point of F n ∘ \mathcal{F}_n^\circ F n ∘ , h ( s i ) = 0 h(s_i)=0 h ( s i ) = 0 . Then, at that point, we also have ∑ k ≠ i z i k h ( s k ) = 0 \sum_{k\ne i}z_{ik}h(s_k)=0 ∑ k = i z ik h ( s k ) = 0 .
Proof. The statement is local, so we may trivialize all the line bundles involved and choose a basis of the rank-two vector space L L L . On such a local chart, h h h is a quadric on the two-dimensional vector space L ∨ L^\vee L ∨ . The condition h ( s i ) = 0 h(s_i)=0 h ( s i ) = 0 says that the quadratic polynomial h h h vanishes at the point [ s i ] ∈ P ( L ∨ ) [s_i]\in\mathbb{P}(L^\vee) [ s i ] ∈ P ( L ∨ ) . A quadric vanishing at [ s i ] [s_i] [ s i ] is divisible by the linear form s ⟼ s ∧ s i s\longmapsto s\wedge s_i s ⟼ s ∧ s i . Hence there exists a linear form ℓ \ell ℓ such that h ( s ) = ( s ∧ s i ) ℓ ( s ) h(s)=(s\wedge s_i)\ell(s) h ( s ) = ( s ∧ s i ) ℓ ( s ) . Evaluating at s k s_k s k , we obtain h ( s k ) = ( s k ∧ s i ) ℓ ( s k ) h(s_k)=(s_k\wedge s_i)\ell(s_k) h ( s k ) = ( s k ∧ s i ) ℓ ( s k ) . Now, s k ∧ s i = − s i ∧ s k = − p i k s_k\wedge s_i=-s_i\wedge s_k=-p_{ik} s k ∧ s i = − s i ∧ s k = − p ik . So h ( s k ) = − p i k ℓ ( s k ) h(s_k)=-p_{ik}\ell(s_k) h ( s k ) = − p ik ℓ ( s k ) . Therefore
∑ k ≠ i z i k h ( s k ) = ∑ k ≠ i q i k p i k ( − p i k ℓ ( s k ) ) = − ∑ k ≠ i q i k ℓ ( s k ) = − ∑ k = 1 n q i k ℓ ( s k ) . \sum_{k\ne i}z_{ik}h(s_k)=\sum_{k\ne i}\frac{q_{ik}}{p_{ik}}(-p_{ik}\ell(s_k))=-\sum_{k\ne i}q_{ik}\ell(s_k)=-\sum_{k=1}^{n}q_{ik}\ell(s_k). k = i ∑ z ik h ( s k ) = k = i ∑ p ik q ik ( − p ik ℓ ( s k )) = − k = i ∑ q ik ℓ ( s k ) = − k = 1 ∑ n q ik ℓ ( s k ) .
where, in the last equality, we added the term for k = i k=i k = i which does not contribute. By momentum conservation, ∑ k = 1 n q i k s k = 0 \sum_{k=1}^{n}q_{ik}s_k=0 ∑ k = 1 n q ik s k = 0 . Pairing this identity with the linear form ℓ \ell ℓ , we obtain that ∑ k = 1 n q i k ℓ ( s k ) = 0 \sum_{k=1}^{n}q_{ik}\ell(s_k)=0 ∑ k = 1 n q ik ℓ ( s k ) = 0 . Thus ∑ k ≠ i z i k h ( s k ) = 0 \sum_{k\ne i}z_{ik}h(s_k)=0 ∑ k = i z ik h ( s k ) = 0 .
Proposition 3.16 (Independence of the quadric). Let f f f and g g g be two local sections of H \mathcal{H} H . On F n ∘ ∩ U f ∩ U g \mathcal{F}_n^\circ\cap U_f\cap U_g F n ∘ ∩ U f ∩ U g , one has Φ ( f ) = Φ ( g ) \Phi^{(f)}=\Phi^{(g)} Φ ( f ) = Φ ( g ) . Consequently, the local matrices Φ ( f ) \Phi^{(f)} Φ ( f ) glue to a canonical matrix
Φ ∈ H 0 ( F n ∘ , Sym 2 ( O F n ∘ ⊕ n ) ⊗ L ) . \Phi\in H^0\left(\mathcal{F}_n^\circ,\operatorname{Sym}^2\left(\mathcal{O}_{\mathcal{F}_n^\circ}^{\oplus n}\right)\otimes\mathcal{L}\right). Φ ∈ H 0 ( F n ∘ , Sym 2 ( O F n ∘ ⊕ n ) ⊗ L ) . Equivalently, Φ \Phi Φ gives a rational projective map, the Hodges map:
H : F n ⇢ P ( Sym C ( n ) ) , ( L , W ) ⟼ [ Φ ( L , W ) ] . H:\mathcal{F}_n\dashrightarrow\mathbb{P}(\operatorname{Sym}_{\mathbb{C}}(n)),\qquad(L,W)\longmapsto[\Phi(L,W)]. H : F n ⇢ P ( Sym C ( n )) , ( L , W ) ⟼ [ Φ ( L , W )] .
Proof. The off-diagonal entries are the same by definition. We only need to compare the diagonal entries. Fix i i i . After trivializing H \mathcal{H} H locally, regard f ( s i ) f(s_i) f ( s i ) and g ( s i ) g(s_i) g ( s i ) as nonzero scalars. Define h = g ( s i ) f − f ( s i ) g h=g(s_i)f-f(s_i)g h = g ( s i ) f − f ( s i ) g . Then h ( s i ) = 0 h(s_i)=0 h ( s i ) = 0 . By Lemma 3.15 ,
∑ k ≠ i z i k h ( s k ) = 0. \sum_{k\ne i}z_{ik}h(s_k)=0. k = i ∑ z ik h ( s k ) = 0.
Expanding this identity gives
g ( s i ) ∑ k ≠ i z i k f ( s k ) − f ( s i ) ∑ k ≠ i z i k g ( s k ) = 0. g(s_i)\sum_{k\ne i}z_{ik}f(s_k)-f(s_i)\sum_{k\ne i}z_{ik}g(s_k)=0. g ( s i ) k = i ∑ z ik f ( s k ) − f ( s i ) k = i ∑ z ik g ( s k ) = 0.
Dividing by f ( s i ) g ( s i ) f(s_i)g(s_i) f ( s i ) g ( s i ) yields
∑ k ≠ i z i k f ( s k ) f ( s i ) = ∑ k ≠ i z i k g ( s k ) g ( s i ) . \frac{\sum_{k\ne i}z_{ik}f(s_k)}{f(s_i)}=\frac{\sum_{k\ne i}z_{ik}g(s_k)}{g(s_i)}. f ( s i ) ∑ k = i z ik f ( s k ) = g ( s i ) ∑ k = i z ik g ( s k ) .
Hence Φ i i ( f ) = Φ i i ( g ) \Phi_{ii}^{(f)}=\Phi_{ii}^{(g)} Φ ii ( f ) = Φ ii ( g ) . This proves the independence of f f f . Since the open sets U f U_f U f cover F n ∘ F_n^\circ F n ∘ , the local matrices glue. Finally, the entries of Φ \Phi Φ are sections of the same line bundle L \mathcal{L} L . Hence, after trivializing L \mathcal{L} L , they give an ordinary symmetric matrix; changing the trivialization scales all entries by the same nonzero function. Therefore the projective point [ Φ ] ∈ P ( Sym n ) [\Phi]\in\mathbb{P}(\operatorname{Sym}_n) [ Φ ] ∈ P ( Sym n ) is well-defined. This describes the Hodges rational map H H H .
The Kernel of the Generalized Hodges Matrix
It was already shown in Hodges’ original work [9 ] , that the Hodges matrix from Definition 3.1 has rank n − 3 n-3 n − 3 . Motivated by this observation here we will study the kernel of the generalized Hodges matrix from Definition 3.12 in the global language of line bundles. This will allow us to derive the global version of Hodges formula in Definition 3.2 . The main result can be stated as follows.
Proposition 3.17 (The Hodges kernel). On the Hodges locus F n ∘ F_n^\circ F n ∘ , the rank-three bundle K \mathcal{K} K is contained in the kernel of Φ \Phi Φ . Equivalently, for every local quadric Q ∈ Sym 2 S Q\in\operatorname{Sym}^2 S Q ∈ Sym 2 S , the vector v Q = ( Q ( s 1 ) , … , Q ( s n ) ) v_Q=(Q(s_1),\ldots,Q(s_n)) v Q = ( Q ( s 1 ) , … , Q ( s n )) satisfies $\Phi v_Q=0.
Proof. By Proposition 3.16 it suffices to work on an open set U f U_f U f , where we can use the local formula Φ ( f ) \Phi^{(f)} Φ ( f ) for Φ \Phi Φ . Fix i ∈ [ n ] i\in[n] i ∈ [ n ] . We have to show that Φ i i Q ( s i ) + ∑ k ≠ i z i k Q ( s k ) = 0 \Phi_{ii}Q(s_i)+\sum_{k\ne i}z_{ik}Q(s_k)=0 Φ ii Q ( s i ) + ∑ k = i z ik Q ( s k ) = 0 . By definition of the generalized Hodges matrix as in Definition 3.12 , we locally have
Φ i i Q ( s i ) + ∑ k ≠ i z i k Q ( s k ) = − Q ( s i ) f ( s i ) ∑ k ≠ i z i k f ( s k ) + ∑ k ≠ i z i k Q ( s k ) = 1 f ( s i ) ∑ k ≠ i z i k ( − Q ( s i ) f ( s k ) + f ( s i ) Q ( s k ) ) . \begin{aligned}
\Phi_{ii}Q(s_i)+\sum_{k\ne i}z_{ik}Q(s_k)=-\frac{Q(s_i)}{f(s_i)}\sum_{k\ne i}z_{ik}f(s_k)+\sum_{k\ne i}z_{ik}Q(s_k) \\
&=\frac{1}{f(s_i)}\sum_{k\ne i}z_{ik}\bigl(-Q(s_i)f(s_k)+f(s_i)Q(s_k)\bigr).
\end{aligned} Φ ii Q ( s i ) + k = i ∑ z ik Q ( s k ) = − f ( s i ) Q ( s i ) k = i ∑ z ik f ( s k ) + k = i ∑ z ik Q ( s k ) = f ( s i ) 1 k = i ∑ z ik ( − Q ( s i ) f ( s k ) + f ( s i ) Q ( s k ) ) .
Consider the quadric h = f ( s i ) Q − Q ( s i ) f ∈ Sym 2 S ⊗ ( det S ∨ ) ⊗ 2 h=f(s_i)Q-Q(s_i)f\in\operatorname{Sym}^2 S\otimes(\det S^\vee)^{\otimes2} h = f ( s i ) Q − Q ( s i ) f ∈ Sym 2 S ⊗ ( det S ∨ ) ⊗ 2 . Then h ( s i ) = 0 h(s_i)=0 h ( s i ) = 0 . Hence ∑ k ≠ i z i k h ( s k ) = 0 \sum_{k\ne i}z_{ik}h(s_k)=0 ∑ k = i z ik h ( s k ) = 0 by Lemma 3.15 . Therefore Φ i i Q ( s i ) + ∑ k ≠ i z i k Q ( s k ) = 0 \Phi_{ii}Q(s_i)+\sum_{k\ne i}z_{ik}Q(s_k)=0 Φ ii Q ( s i ) + ∑ k = i z ik Q ( s k ) = 0 . Since this holds for every row i i i , we derive Φ v Q = 0 \Phi v_Q=0 Φ v Q = 0 .
Corollary 3.18. The Hodges rational map induces a rational map
H ~ : F n ⇢ { ( [ A ] , K ) ∈ P ( Sym n ) × Gr ( 3 , n ) : K ⊆ ker A } , \widetilde{H}:F_n\dashrightarrow\{([A],K)\in\mathbb{P}(\operatorname{Sym}_n)\times\operatorname{Gr}(3,n):K\subseteq\ker A\}, H : F n ⇢ {([ A ] , K ) ∈ P ( Sym n ) × Gr ( 3 , n ) : K ⊆ ker A } , on the Hodges locus F n ∘ F_n^\circ F n ∘ defined by
( L , W ) ⟼ ( [ Φ ( L , W ) ] , K ( L , W ) ) , (L,W)\longmapsto\bigl([\Phi(L,W)],\mathcal{K}_{(L,W)}\bigr), ( L , W ) ⟼ ( [ Φ ( L , W )] , K ( L , W ) ) , where K ( L , W ) \mathcal{K}_{(L,W)} K ( L , W ) is the fiber of K \mathcal{K} K at ( L , W ) (L,W) ( L , W ) . Moreover, the incidence space on the right is isomorphic to the blowup of the variety of symmetric matrices of rank at most ( n − 3 ) (n-3) ( n − 3 ) along the subvariety of those whose rank is at most ( n − 4 ) (n-4) ( n − 4 ) .
The Induced Determinant and the MHV gravity divisor
In this paragraph we finally give the geometric interpretation of the MHV gravity amplitude. We start, however, with a technical result.
Lemma 3.19. Let A A A be an n × n n \times n n × n symmetric complex matrix and suppose that we are given a 3-dimensional subspace K ⊆ ker ( A ) K \subseteq\ker(A) K ⊆ ker ( A ) . Let Q = C n / K Q = \mathbb{C}^{n}/K Q = C n / K .
(i) Then A ‾ \overline{A} A descends to a symmetric linear map
A ‾ : Q ⟶ Q ∨ . \overline{A}: Q \longrightarrow Q^{\vee}. A : Q ⟶ Q ∨ . (ii) Choosing the standard volume element on C n \mathbb{C}^{n} C n , the induced determinant det A ‾ ∈ ( det Q ∨ ) ⊗ 2 \det\overline{A} \in(\det Q^{\vee})^{\otimes2} det A ∈ ( det Q ∨ ) ⊗ 2 can be identified with an element Δ ‾ ∈ ( det K ) ⊗ 2 \overline{\Delta} \in(\det K)^{\otimes2} Δ ∈ ( det K ) ⊗ 2 .
(iii) Denote by A R C A^{C}_{R} A R C the ( n − 3 ) × ( n − 3 ) (n-3) \times(n-3) ( n − 3 ) × ( n − 3 ) matrix obtained from A A A by deleting the rows indexed by R R R and the columns indexed by C C C . Then,
det A R C = ± ⟨ κ R ⊗ κ C , Δ ‾ ⟩ ∈ C , \det A^{C}_{R} = \pm\langle\kappa_{R} \otimes\kappa_{C}, \overline{\Delta} \rangle\in\mathbb{C}, det A R C = ± ⟨ κ R ⊗ κ C , Δ ⟩ ∈ C , where κ R \kappa_{R} κ R and κ C \kappa_{C} κ C in ( det K ) ∨ (\det K)^{\vee} ( det K ) ∨ are the Plücker coordinate functionals of K K K labeled by R R R and C C C , respectively, and ⟨ − , − ⟩ \langle-, - \rangle ⟨ − , − ⟩ is the natural pairing ( det K ∨ ) ⊗ 2 × ( det K ) ⊗ 2 ⟶ C (\det K^{\vee})^{\otimes2} \times(\det K)^{\otimes2} \longrightarrow\mathbb{C} ( det K ∨ ) ⊗ 2 × ( det K ) ⊗ 2 ⟶ C . In particular, for any R R R and C C C the minor det A R C \det A^{C}_{R} det A R C is obtained by evaluating the induced determinant det ( A ‾ ) \det(\overline{A}) det ( A ) on the functionals κ R \kappa_{R} κ R and κ C \kappa_{C} κ C .
Proof. We start by showing (i). Because K ⊆ ker ( A ) K \subseteq\ker(A) K ⊆ ker ( A ) , the bilinear form associated to A A A vanishes whenever one argument lies in K K K . Let Q = C n / K Q = \mathbb{C}^{n}/K Q = C n / K . Hence it induces a well-defined bilinear form on the quotient Q Q Q , equivalently a symmetric map A ‾ : Q ⟶ Q ∨ \overline{A}: Q \longrightarrow Q^{\vee} A : Q ⟶ Q ∨ . In the rest of the proof, we implicitly apply the canonical identification between ( det Q ) ∨ (\det Q)^{\vee} ( det Q ) ∨ and det Q ∨ \det Q^{\vee} det Q ∨ .
We show statement (ii). Choose the standard volume form ω = e 1 ∧ ⋯ ∧ e n \omega= e_{1} \wedge\cdots\wedge e_{n} ω = e 1 ∧ ⋯ ∧ e n on C n \mathbb{C}^{n} C n , then det Q ∨ ≅ det K \det Q^{\vee} \cong\det K det Q ∨ ≅ det K , as can be seen by the short exact sequence
0 ⟶ K ⟶ C n ⟶ Q ⟶ 0. (1) 0 \longrightarrow K \longrightarrow\mathbb{C}^{n} \longrightarrow Q \longrightarrow0. \tag*{(1)} 0 ⟶ K ⟶ C n ⟶ Q ⟶ 0. (1)
Now recall that the determinant of a map Q ⟶ Q ∨ Q \longrightarrow Q^{\vee} Q ⟶ Q ∨ is an element of det Q ∨ ⊗ det Q ∨ ≅ ( det K ) ⊗ 2 \det Q^{\vee} \otimes\det Q^{\vee} \cong(\det K)^{\otimes2} det Q ∨ ⊗ det Q ∨ ≅ ( det K ) ⊗ 2 . Then, Δ ‾ ∈ ( det K ) ⊗ 2 \overline{\Delta} \in(\det K)^{\otimes2} Δ ∈ ( det K ) ⊗ 2 is the image of det A ‾ \det\overline{A} det A under this isomorphism.
It remains to show (iii). Let U = ( u 1 , u 2 , u 3 ) U = (u_{1}, u_{2}, u_{3}) U = ( u 1 , u 2 , u 3 ) be the n × 3 n \times3 n × 3 matrix spanning K K K , and let Z = ( z 1 , … , z n − 3 ) Z = (z_{1}, \ldots, z_{n-3}) Z = ( z 1 , … , z n − 3 ) be a n × ( n − 3 ) n \times(n-3) n × ( n − 3 ) matrix such that the n × n n \times n n × n matrix B = [ U Z ] B = [UZ] B = [ U Z ] has non-vanishing determinant. Let z = z ‾ 1 ∧ ⋯ ∧ z ‾ n − 3 ∈ det Q z = \overline{z}_{1} \wedge\cdots\wedge\overline{z}_{n-3} \in\det Q z = z 1 ∧ ⋯ ∧ z n − 3 ∈ det Q , and take its dual z ∨ ∈ det Q ∨ z^{\vee} \in\det Q^{\vee} z ∨ ∈ det Q ∨ with ⟨ z ∨ , z ⟩ = 1 \langle z^{\vee}, z \rangle= 1 ⟨ z ∨ , z ⟩ = 1 . Then, A ‾ \overline{A} A has a matrix representative Z T A Z Z^{T}AZ Z T A Z (with respect to the basis given by the z i z_{i} z i ). We obtain
det ( A ‾ ) = det ( Z T A Z ) ( z ∨ ) ⊗ 2 ∈ ( det Q ∨ ) ⊗ 2 . (2) \det(\overline{A}) = \det(Z^{T}AZ)(z^{\vee})^{\otimes2} \in(\det Q^{\vee})^{\otimes2}. \tag*{(2)} det ( A ) = det ( Z T A Z ) ( z ∨ ) ⊗ 2 ∈ ( det Q ∨ ) ⊗ 2 . (2)
The determinant isomorphism from the short exact sequence (1)
μ : det K ⊗ det Q ≅ det C n \mu: \det K \otimes\det Q \cong\det\mathbb{C}^{n} μ : det K ⊗ det Q ≅ det C n
satisfies μ ( u 1 ∧ u 2 ∧ u 3 ⊗ z ) = det ( B ) ω \mu(u_{1} \wedge u_{2} \wedge u_{3} \otimes z) = \det(B)\omega μ ( u 1 ∧ u 2 ∧ u 3 ⊗ z ) = det ( B ) ω . Thus, the identification induced by ω \omega ω gives a canonical isomorphism G ω : det K ⟶ det Q ∨ G_{\omega}: \det K \longrightarrow\det Q^{\vee} G ω : det K ⟶ det Q ∨ defined by G ω ( k ) ( q ) ω = μ ( k ⊗ q ) G_{\omega}(k)(q)\omega= \mu(k \otimes q) G ω ( k ) ( q ) ω = μ ( k ⊗ q ) . One can check that its inverse is the isomorphism
F ω : det Q ∨ ⟶ det K , λ ⟼ ( id det K ⊗ λ ) ( μ − 1 ( ω ) ) . F_{\omega}: \det Q^{\vee} \longrightarrow\det K,\qquad\lambda\longmapsto(\operatorname{id}_{\det K} \otimes\lambda)(\mu^{-1}(\omega)). F ω : det Q ∨ ⟶ det K , λ ⟼ ( id d e t K ⊗ λ ) ( μ − 1 ( ω )) .
This linear map is such that
z ∨ ⟼ ( id det K ⊗ z ∨ ) ( μ − 1 ( ω ) ) = 1 det B ⟨ id det K , u 1 ∧ u 2 ∧ u 3 ⟩ ⊗ ⟨ z ∨ , z ⟩ = u 1 ∧ u 2 ∧ u 3 det B . (3) z^\vee\longmapsto(\operatorname{id}_{\det K}\otimes z^\vee)(\mu^{-1}(\omega))=\frac{1}{\det B}\langle\operatorname{id}_{\det K},u_1\wedge u_2\wedge u_3\rangle\otimes\langle z^\vee,z\rangle
=\frac{u_1\wedge u_2\wedge u_3}{\det B}. \tag*{(3)} z ∨ ⟼ ( id d e t K ⊗ z ∨ ) ( μ − 1 ( ω )) = det B 1 ⟨ id d e t K , u 1 ∧ u 2 ∧ u 3 ⟩ ⊗ ⟨ z ∨ , z ⟩ = det B u 1 ∧ u 2 ∧ u 3 . (3)
since μ − 1 ( ω ) = ( u 1 ∧ u 2 ∧ u 3 ⊗ z ) / det B \mu^{-1}(\omega)=(u_1\wedge u_2\wedge u_3\otimes z)/\det B μ − 1 ( ω ) = ( u 1 ∧ u 2 ∧ u 3 ⊗ z ) / det B and ⟨ z ∨ , z ⟩ = 1 \langle z^\vee,z\rangle=1 ⟨ z ∨ , z ⟩ = 1 . Combining {(3.2)} and (3) yields
Δ ‾ : = F ω ⊗ 2 ( det ( A ‾ ) ) = det Z T A Z ( det B ) 2 ( u 1 ∧ u 2 ∧ u 3 ) ⊗ 2 . \overline{\Delta}:=F_\omega^{\otimes2}\left(\det(\overline{A})\right)=\frac{\det Z^T A Z}{(\det B)^2}(u_1\wedge u_2\wedge u_3)^{\otimes2}. Δ := F ω ⊗ 2 ( det ( A ) ) = ( det B ) 2 det Z T A Z ( u 1 ∧ u 2 ∧ u 3 ) ⊗ 2 .
Moreover, as A U = 0 AU=0 A U = 0 we have
B T A B = ( 0 3 × 3 0 3 × ( n − 3 ) 0 ( n − 3 ) × 3 Z T A Z ) . B^T A B=
\begin{pmatrix}
0_{3\times3} & 0_{3\times(n-3)}\\
0_{(n-3)\times3} & Z^T A Z
\end{pmatrix}. B T A B = ( 0 3 × 3 0 ( n − 3 ) × 3 0 3 × ( n − 3 ) Z T A Z ) .
Therefore, A = B − T ( 0 3 × 3 0 3 × ( n − 3 ) 0 ( n − 3 ) × 3 Z T A Z ) B − 1 A=B^{-T}\begin{pmatrix}0_{3\times3}&0_{3\times(n-3)}\\0_{(n-3)\times3}&Z^T A Z\end{pmatrix}B^{-1} A = B − T ( 0 3 × 3 0 ( n − 3 ) × 3 0 3 × ( n − 3 ) Z T A Z ) B − 1 . Recall that for a matrix A A A we denote by A R C A^C_R A R C the matrix obtained from A A A by deleting the rows and columns indexed by R R R and C C C , respectively. Then, by applying Cauchy–Binet formula twice to the matrix equation above, we get
det A R C = ∑ I , J ⊂ { 1 , … , n } ∣ I ∣ = ∣ J ∣ = 3 det ( B − T ) R I det ( B T A B ) I J det ( B − 1 ) J C = det ( B − T ) R S c det ( Z T A Z ) det ( B − 1 ) S c C . \begin{aligned}
\det A^C_R=\sum_{\substack{I,J\subset\{1,\ldots,n\}\\|I|=|J|=3}}\det(B^{-T})^I_R\det(B^T A B)^J_I\det(B^{-1})^C_J \\
&=\det(B^{-T})^{S^c}_R\det(Z^T A Z)\det(B^{-1})^{C}_{S^c}.
\end{aligned} det A R C = I , J ⊂ { 1 , … , n } ∣ I ∣ = ∣ J ∣ = 3 ∑ det ( B − T ) R I det ( B T A B ) I J det ( B − 1 ) J C = det ( B − T ) R S c det ( Z T A Z ) det ( B − 1 ) S c C .
where S = { 4 , … , n } S=\{4,\ldots,n\} S = { 4 , … , n } , and R c , C c R^c,C^c R c , C c denote the complements of R , C R,C R , C in [ n ] [n] [ n ] . The last equality follows because the first three rows and columns of B T A B B^T A B B T A B vanish, so only I = J = S c I=J=S^c I = J = S c can contribute. Then, using Jacobi’s identity on complementary minors,
det ( B − T ) R S c = ± det ( B T ) S R c det B = ± det B S R c det B , \det(B^{-T})^{S^c}_R=\pm\frac{\det(B^T)^{R^c}_S}{\det B}=\pm\frac{\det B^{R^c}_S}{\det B}, det ( B − T ) R S c = ± det B det ( B T ) S R c = ± det B det B S R c ,
det ( B − 1 ) S c C = ± det B C c S det B . \det(B^{-1})^{C}_{S^c}=\pm\frac{\det B^S_{C^c}}{\det B}. det ( B − 1 ) S c C = ± det B det B C c S .
Note that B R c S = [ U Z ] R c S = U R c B^{S}_{R^c}=[UZ]^S_{R^c}=U_{R^c} B R c S = [ U Z ] R c S = U R c and B C c S = [ U Z ] C c S = U C c B^S_{C^c}=[UZ]^S_{C^c}=U_{C^c} B C c S = [ U Z ] C c S = U C c ; it follows that
det A R C = ± det U R c det U C c det ( B ) 2 det ( Z T A Z ) . \det A^C_R=\pm\frac{\det U_{R^c}\det U_{C^c}}{\det(B)^2}\det(Z^T A Z). det A R C = ± det ( B ) 2 det U R c det U C c det ( Z T A Z ) .
To establish (iii), we now compute the pairing ⟨ κ R ⊗ κ C , Δ ‾ ⟩ \langle\kappa_R\otimes\kappa_C,\overline{\Delta}\rangle ⟨ κ R ⊗ κ C , Δ ⟩ . For R = { r 1 < r 2 < r 3 } R=\{r_1<r_2<r_3\} R = { r 1 < r 2 < r 3 } let
κ R = ( e r 1 ∨ ∧ e r 2 ∨ ∧ e r 3 ∨ ) ∣ det K ∈ det K ∨ . \kappa_R=(e^\vee_{r_1}\wedge e^\vee_{r_2}\wedge e^\vee_{r_3})|_{\det K}\in\det K^\vee. κ R = ( e r 1 ∨ ∧ e r 2 ∨ ∧ e r 3 ∨ ) ∣ d e t K ∈ det K ∨ .
and analogously for κ C \kappa_C κ C . Then, κ R ( u 1 ∧ u 2 ∧ u 3 ) = det U R c \kappa_R(u_1\wedge u_2\wedge u_3)=\det U_{R^c} κ R ( u 1 ∧ u 2 ∧ u 3 ) = det U R c and κ C ( u 1 ∧ u 2 ∧ u 3 ) = det U C c \kappa_C(u_1\wedge u_2\wedge u_3)=\det U_{C^c} κ C ( u 1 ∧ u 2 ∧ u 3 ) = det U C c . Therefore,
⟨ κ R ⊗ κ C , Δ ‾ ⟩ = det ( Z T A Z ) ( det B ) 2 κ R ( u 1 ∧ u 2 ∧ u 3 ) κ C ( u 1 ∧ u 2 ∧ u 3 ) = det ( Z T A Z ) ( det B ) 2 det U R c det U C c = ± det A R C . \begin{aligned}
\langle\kappa_R\otimes\kappa_C,\overline{\Delta}\rangle
=\frac{\det(Z^T A Z)}{(\det B)^2}\kappa_R(u_1\wedge u_2\wedge u_3)\kappa_C(u_1\wedge u_2\wedge u_3) \\
&=\frac{\det(Z^T A Z)}{(\det B)^2}\det U_{R^c}\det U_{C^c}=\pm\det A^C_R.
\end{aligned} ⟨ κ R ⊗ κ C , Δ ⟩ = ( det B ) 2 det ( Z T A Z ) κ R ( u 1 ∧ u 2 ∧ u 3 ) κ C ( u 1 ∧ u 2 ∧ u 3 ) = ( det B ) 2 det ( Z T A Z ) det U R c det U C c = ± det A R C .
Now, on the Hodges locus F n ∘ \mathcal{F}^{\circ}_{n} F n ∘ , let Q : = O F n ∘ ⊕ n / K \mathcal{Q} := \mathcal{O}^{\oplus n}_{\mathcal{F}^{\circ}_{n}}/\mathcal{K} Q := O F n ∘ ⊕ n / K . This is a vector bundle of rank n − 3 n-3 n − 3 . Since Φ \Phi Φ is symmetric and K \mathcal{K} K is in the kernel of Φ \Phi Φ by Proposition 3.17 , it descends to a symmetric L \mathcal{L} L -valued form on Q \mathcal{Q} Q , i.e. Φ ‾ : Q ⟶ Q ∨ ⊗ L \overline{\Phi} : \mathcal{Q} \longrightarrow\mathcal{Q}^{\vee} \otimes\mathcal{L} Φ : Q ⟶ Q ∨ ⊗ L by Lemma 3.19 . Equivalently, it is a section of Sym 2 Q ∨ ⊗ L \operatorname{Sym}^{2}\mathcal{Q}^{\vee} \otimes\mathcal{L} Sym 2 Q ∨ ⊗ L . Recall that the determinant of a map Q → Q ∨ ⊗ L \mathcal{Q} \to\mathcal{Q}^{\vee} \otimes\mathcal{L} Q → Q ∨ ⊗ L is a (global) section of
det ( Q ∨ ⊗ L ) ⊗ ( det Q ) ∨ = det Q ∨ ⊗ L ⊗ ( n − 3 ) ⊗ det Q ∨ = ( det Q ∨ ) ⊗ 2 ⊗ L ⊗ ( n − 3 ) . \det(\mathcal{Q}^{\vee} \otimes\mathcal{L}) \otimes(\det\mathcal{Q})^{\vee} = \det\mathcal{Q}^{\vee} \otimes\mathcal{L}^{\otimes(n-3)} \otimes\det\mathcal{Q}^{\vee} = (\det\mathcal{Q}^{\vee})^{\otimes2} \otimes\mathcal{L}^{\otimes(n-3)}. det ( Q ∨ ⊗ L ) ⊗ ( det Q ) ∨ = det Q ∨ ⊗ L ⊗ ( n − 3 ) ⊗ det Q ∨ = ( det Q ∨ ) ⊗ 2 ⊗ L ⊗ ( n − 3 ) .
Thus, Δ H : = det ( Φ ‾ ) ∈ H 0 ( F n ∘ , ( det Q ∨ ) ⊗ 2 ⊗ L ⊗ ( n − 3 ) ) \Delta_{H} := \det(\overline{\Phi}) \in H^{0}\left(\mathcal{F}^{\circ}_{n},(\det\mathcal{Q}^{\vee})^{\otimes2} \otimes\mathcal{L}^{\otimes(n-3)}\right) Δ H := det ( Φ ) ∈ H 0 ( F n ∘ , ( det Q ∨ ) ⊗ 2 ⊗ L ⊗ ( n − 3 ) ) . This motivates the following definition.
Definition 3.20. Denote the vanishing locus of the regular section Δ H \Delta_{H} Δ H on Hodges locus F n ∘ \mathcal{F}^{\circ}_{n} F n ∘ by D H ∘ : = V ( Δ H ) ⊂ F n ∘ D^{\circ}_{H} := V(\Delta_{H}) \subset\mathcal{F}^{\circ}_{n} D H ∘ := V ( Δ H ) ⊂ F n ∘ , which we call the open gravity divisor . Its closure D H ⊂ F n D_{H} \subset\mathcal{F}_{n} D H ⊂ F n in the flag variety F n \mathcal{F}_{n} F n is called the gravity divisor .
Remark 3.21. It is known by [9 ] that Φ \Phi Φ has generic rank n − 3 n-3 n − 3 on F n ∘ \mathcal{F}^{\circ}_{n} F n ∘ .
We can now record the following version of our main Theorem 1.2 , using the language of vector bundles introduced above.
Theorem 3.22 (The MHV gravity divisor). Let n ≥ 5 n \ge5 n ≥ 5 and F n ∘ \mathcal{F}^{\circ}_{n} F n ∘ be the Hodges locus in F n \mathcal{F}_{n} F n where all p i j ≠ 0 p_{ij} \ne0 p ij = 0 .
(i) The open gravity divisor D H ∘ = V ( Δ H ) ⊂ F n ∘ D^{\circ}_{H} = V(\Delta_{H}) \subset\mathcal{F}^{\circ}_{n} D H ∘ = V ( Δ H ) ⊂ F n ∘ is an effective Cartier divisor. Set-theoretically it is the locus of points ( L , W ) ∈ F n ∘ (L,W) \in\mathcal{F}^{\circ}_{n} ( L , W ) ∈ F n ∘ where the induced bi-linear form Φ ‾ : C n / K ( L , W ) ⟶ ( C n / K ( L , W ) ) ∨ ⊗ L ( L , W ) \overline{\Phi} : \mathbb{C}^{n}/\mathcal{K}_{(L,W)} \longrightarrow(\mathbb{C}^{n}/\mathcal{K}_{(L,W)})^{\vee} \otimes\mathcal{L}_{(L,W)} Φ : C n / K ( L , W ) ⟶ ( C n / K ( L , W ) ) ∨ ⊗ L ( L , W ) fails to be invertible, with K ( L , W ) \mathcal{K}_{(L,W)} K ( L , W ) being the fiber of K \mathcal{K} K over ( L , W ) (L,W) ( L , W ) . In symbols we have
D H ∘ = { ( L , W ) ∈ F n ∘ ∣ det ( Φ ‾ ( L , W ) ) = 0 } ⊂ F n ∘ . D^{\circ}_{H} = \left\{(L,W) \in\mathcal{F}^{\circ}_{n} \mid\det(\overline{\Phi}(L,W)) = 0\right\} \subset\mathcal{F}^{\circ}_{n}. D H ∘ = { ( L , W ) ∈ F n ∘ ∣ det ( Φ ( L , W )) = 0 } ⊂ F n ∘ . (ii) For R = { i < j < k } R = \{i < j < k\} R = { i < j < k } and C = { r < s < t } C = \{r < s < t\} C = { r < s < t } , we have
Δ H = ± det Φ R C p i j p i k p j k p r s p r t p s t ∈ H 0 ( F n ∘ , ( det Q ∨ ) ⊗ 2 ⊗ L ⊗ ( n − 3 ) ) , \Delta_{H} = \pm\frac{\det\Phi^{C}_{R}}{p_{ij}p_{ik}p_{jk}\,p_{rs}p_{rt}p_{st}} \in H^{0}\left(\mathcal{F}^{\circ}_{n},(\det\mathcal{Q}^{\vee})^{\otimes2} \otimes\mathcal{L}^{\otimes(n-3)}\right), Δ H = ± p ij p ik p j k p r s p r t p s t det Φ R C ∈ H 0 ( F n ∘ , ( det Q ∨ ) ⊗ 2 ⊗ L ⊗ ( n − 3 ) ) , so it coincides with the MHV amplitude M ~ n \widetilde{M}_{n} M n for n n n particles in N = 8 \mathcal{N}=8 N = 8 supergravity.
(iii) The pullback under the Hodges rational map H ~ \widetilde{H} H of the exceptional divisor E E E of the blowup of the variety of symmetric n n n matrices of rank at most $n-3 in its singular locus gives the open gravity divisor D H ∘ D^{\circ}_{H} D H ∘ . Put in symbols we have
H ~ ∗ E ≅ D H ∘ . \widetilde{H}^{*}E \cong D^{\circ}_{H}. H ∗ E ≅ D H ∘ .
Proof. We start with (i). This follows from the discussion after Lemma 3.19 . Firstly, Δ H \Delta_{H} Δ H is by definition a regular global section of the line bundle ( det Q ∨ ) ⊗ 2 ⊗ L ⊗ ( n − 3 ) (\det\mathcal{Q}^{\vee})^{\otimes2} \otimes\mathcal{L}^{\otimes(n-3)} ( det Q ∨ ) ⊗ 2 ⊗ L ⊗ ( n − 3 ) on the smooth variety F n ∘ \mathcal{F}^{\circ}_{n} F n ∘ hence its vanishing locus is Cartier and effective. Since Δ H = det Φ ‾ \Delta_{H} = \det\overline{\Phi} Δ H = det Φ it is precisely the locus of points ( L , W ) ∈ F n ∘ (L,W) \in\mathcal{F}^{\circ}_{n} ( L , W ) ∈ F n ∘ where Φ ‾ ( L , W ) \overline{\Phi}(L,W) Φ ( L , W ) is not invertible.
Next we prove statement (ii). Choose a basis of L L L , then write s a = ( u a , v a ) ∈ det L ∨ s_{a} = (u_{a},v_{a}) \in\det L^{\vee} s a = ( u a , v a ) ∈ det L ∨ , and choose a basis of Sym 2 L \operatorname{Sym}^{2}L Sym 2 L as ( x 2 , x y , y 2 ) (x^{2},xy,y^{2}) ( x 2 , x y , y 2 ) . Then, by Proposition 3.17 K \mathcal{K} K is generated by evaluating Sym 2 L \operatorname{Sym}^{2}L Sym 2 L on the s a s_{a} s a . Hence, the a a a -th row of the n × 3 n \times3 n × 3 matrix U U U spanning (a subspace of) the kernel K ( L , W ) \mathcal{K}_{(L,W)} K ( L , W ) is given by ( u a 2 , u a v a , v a 2 ) (u_{a}^{2},u_{a}v_{a},v_{a}^{2}) ( u a 2 , u a v a , v a 2 ) . Therefore,
κ C = det U C c = ( u r v s − u s v r ) ( u r v t − u t v r ) ( u s v t − u t v s ) = p r s p r t p s t . \kappa_{C} = \det U_{C^{c}} = (u_{r}v_{s}-u_{s}v_{r})(u_{r}v_{t}-u_{t}v_{r})(u_{s}v_{t}-u_{t}v_{s}) = p_{rs}p_{rt}p_{st}. κ C = det U C c = ( u r v s − u s v r ) ( u r v t − u t v r ) ( u s v t − u t v s ) = p r s p r t p s t .
Similarly for κ R \kappa_R κ R . Then, the claim follows from Lemma 3.19 . Moreover, by Proposition 3.16 Φ \Phi Φ is independent of the choice of quadrics, used to define the matrix locally, hence we can assume that Φ \Phi Φ is locally the Hodges matrix. It remains to prove (iii): this follows directly from Corollary 3.18 together with (i).
Pentad and Affine Slope Varieties
Motivated by the construction of Hodges matrix and the Hodges rational map in the previous section, we study the Hadamard ratio of two Grassmannians Gr ( 2 , n ) \operatorname{Gr}(2,n) Gr ( 2 , n ) , which we will call the pentad variety Pen n \operatorname{Pen}_n Pen n . The goal of this section is to study the dimension, defining equations, its tropicalization. The results are summarized in Theorem 1.3 . As mentioned in the introduction we will also consider the notion of little group scaling in our setup. This will lead us to the affine slope variety Definition 4.14 and Corollary 1.4 makes this connection precise. The affine slope variety will also be a key ingredient in the proof of our main result about the pentad variety itself.
Pentad Variety
We start by introducing the pentad variety more formally. Let n ≥ 3 n \ge3 n ≥ 3 . The Grassmannian Gr ( 2 , n ) ↪ P ( ⋀ 2 C n ) \operatorname{Gr}(2,n) \hookrightarrow\mathbb{P}(\bigwedge^{2}\mathbb{C}^{n}) Gr ( 2 , n ) ↪ P ( ⋀ 2 C n ) is the subvariety of skew-symmetric n × n n \times n n × n matrices P ( ⋀ 2 C n ) = P N \mathbb{P}(\bigwedge^{2}\mathbb{C}^{n}) = \mathbb{P}^{N} P ( ⋀ 2 C n ) = P N with N = ( n 2 ) − 1 N = \binom{n}{2} - 1 N = ( 2 n ) − 1 given by matrices Q = ( q i j ) Q = (q_{ij}) Q = ( q ij ) such that rank Q = 2 \operatorname{rank} Q = 2 rank Q = 2 . Take a second copy of Gr ( 2 , n ) \operatorname{Gr}(2,n) Gr ( 2 , n ) and denote the entries of its rank 2 skew symmetric matrices by p i j p_{ij} p ij . In this section we denote the polynomial ring generated by the z i j z_{ij} z ij by Z n : = C [ z i j ∣ 1 ≤ i < j ≤ n ] Z_n := \mathbb{C}[z_{ij} \mid1 \le i < j \le n] Z n := C [ z ij ∣ 1 ≤ i < j ≤ n ] .
Definition 4.1. Denote the Hadamard ratio map for the two Grassmannians by
Ψ n : Gr ( 2 , n ) × Gr ( 2 , n ) ⇢ P N , ( q i j , p i j ) ↦ z i j = q i j p i j . (4) \Psi_n : \operatorname{Gr}(2,n) \times\operatorname{Gr}(2,n) \dashrightarrow\mathbb{P}^{N}, \qquad(q_{ij},p_{ij}) \mapsto z_{ij} = \frac{q_{ij}}{p_{ij}}. \tag*{(4)} Ψ n : Gr ( 2 , n ) × Gr ( 2 , n ) ⇢ P N , ( q ij , p ij ) ↦ z ij = p ij q ij . (4)
Then, the pentad variety Pen n ⊂ P N \operatorname{Pen}_n \subset\mathbb{P}^{N} Pen n ⊂ P N is the Hadamard ratio Pen n : = Gr ( 2 , n ) : Gr ( 2 , n ) \operatorname{Pen}_n := \operatorname{Gr}(2,n) : \operatorname{Gr}(2,n) Pen n := Gr ( 2 , n ) : Gr ( 2 , n ) .
Remark 4.2. The rational Hadamard ratio map can be resolved to a morphism using the De Concini-Procesi sequential blowup construction [4 ] .
Definition 4.3. Let Pen n ⊂ P N \operatorname{Pen}_n \subset\mathbb{P}^{N} Pen n ⊂ P N be the closure of the image of the n n n -th Hadamard ratio map Ψ n \Psi_n Ψ n . This is called pentad variety .
Remark 4.4. The Hadamard ratio map is dominant for n ≤ 4 n \le4 n ≤ 4 . The first interesting case is n = 5 n = 5 n = 5 . Here we find that Pen 5 \operatorname{Pen}_5 Pen 5 is a quintic hypersurface in P 9 \mathbb{P}^{9} P 9 . Namely, this hypersurface is defined by the polynomial
pen 12345 = z 12 z 13 z 24 z 35 z 45 − z 12 z 13 z 25 z 34 z 45 − z 12 z 14 z 23 z 35 z 45 + z 12 z 14 z 25 z 34 z 35 + z 12 z 15 z 23 z 34 z 45 − z 12 z 15 z 24 z 34 z 35 + z 13 z 14 z 23 z 25 z 45 − z 13 z 14 z 24 z 25 z 35 − z 13 z 15 z 23 z 24 z 45 + z 13 z 15 z 24 z 25 z 34 + z 14 z 15 z 23 z 24 z 35 − z 14 z 15 z 23 z 25 z 34 . (5) \begin{aligned}
\operatorname{pen}_{12345} ={}& z_{12}z_{13}z_{24}z_{35}z_{45} - z_{12}z_{13}z_{25}z_{34}z_{45} - z_{12}z_{14}z_{23}z_{35}z_{45} + z_{12}z_{14}z_{25}z_{34}z_{35} \\
&+ z_{12}z_{15}z_{23}z_{34}z_{45} - z_{12}z_{15}z_{24}z_{34}z_{35} + z_{13}z_{14}z_{23}z_{25}z_{45} - z_{13}z_{14}z_{24}z_{25}z_{35} \\
&- z_{13}z_{15}z_{23}z_{24}z_{45} + z_{13}z_{15}z_{24}z_{25}z_{34} + z_{14}z_{15}z_{23}z_{24}z_{35} - z_{14}z_{15}z_{23}z_{25}z_{34}. \tag*{(5)}
\end{aligned} pen 12345 = z 12 z 13 z 24 z 35 z 45 − z 12 z 13 z 25 z 34 z 45 − z 12 z 14 z 23 z 35 z 45 + z 12 z 14 z 25 z 34 z 35 + z 12 z 15 z 23 z 34 z 45 − z 12 z 15 z 24 z 34 z 35 + z 13 z 14 z 23 z 25 z 45 − z 13 z 14 z 24 z 25 z 35 − z 13 z 15 z 23 z 24 z 45 + z 13 z 15 z 24 z 25 z 34 + z 14 z 15 z 23 z 24 z 35 − z 14 z 15 z 23 z 25 z 34 . (5)
This is known as the pentad in algebraic statistics [5 ] , which gives the name to the variety we are interested in. Its 12 monomials correspond to the 12 oriented five-cycles of a pentagon.
Suppose now that n ≥ 6 n \ge6 n ≥ 6 . For any 5-tuple I I I in [ n ] [n] [ n ] , there is an analogous pentad, but now in the unknowns z i j z_{ij} z ij where i , j ∈ I i,j \in I i , j ∈ I . This is obtained by relabeling (5) . We denote this pentad by pen I \operatorname{pen}_I pen I . Let J n p J_n^p J n p denote the ideal in the polynomial ring Z n Z_n Z n which is generated by all ( n 5 ) \binom{n}{5} ( 5 n ) pentads, i.e.
J n p : = ( pen I ∣ I ⊂ [ n ] s.t. ∣ I ∣ = 5 ) ⊂ Z n . (6) J_n^p := \left(\operatorname{pen}_I \mid I \subset[n]\ \text{s.t.}\ |I|=5\right) \subset Z_n. \tag*{(6)} J n p := ( pen I ∣ I ⊂ [ n ] s.t. ∣ I ∣ = 5 ) ⊂ Z n . (6)
By Remark 4.4 the pentads vanish on the image of the Hadamard ratio map (4) , so we have Pen n ⊆ V ( J n p ) \operatorname{Pen}_n \subseteq V(J_n^p) Pen n ⊆ V ( J n p ) . However, the ideal J n p J_n^p J n p is not prime. Since the base locus of the Hadamard ratio map Ψ n \Psi_n Ψ n includes the Plücker hyperplane section of Gr ( 2 , n ) \operatorname{Gr}(2,n) Gr ( 2 , n ) , we consider the saturation
I n p : = ( J n p : [ ∏ 1 ≤ i < j ≤ n z i j ] ∞ ) ⊂ Z n . (7) I_n^p := \left(J_n^p : \left[\prod_{1 \le i<j \le n} z_{ij}\right]^\infty\right) \subset Z_n. \tag*{(7)} I n p := ( J n p : [ 1 ≤ i < j ≤ n ∏ z ij ] ∞ ) ⊂ Z n . (7)
This leads us to the following conjecture.
Conjecture 4.5. The ideal I n p I_n^p I n p is prime of codimension ( n − 3 2 ) \binom{n-3}{2} ( 2 n − 3 ) . It defines the irreducible variety
Pen n = V ( I n p ) . \operatorname{Pen}_n = V(I_n^p). Pen n = V ( I n p ) .
In words, the Hadamard ratio of the two Grassmannians is cut out by the pentad constraints up to saturation.
Remark 4.6. This conjecture can be proven for the case n = 6 , 7 n=6,7 n = 6 , 7 with a direct computation in Macaulay2. An interesting combinatorial question is to understand the combinatorics of the generators of I n p I_n^p I n p that are not in J n p J_n^p J n p . We collect the generators of I n p I_n^p I n p for n = 5 , 6 , 7 , 8 n=5,6,7,8 n = 5 , 6 , 7 , 8 in Table 1 below. (For n = 8 n=8 n = 8 we could only verify that there are no more generators of degree 8.)
Table 1. Number of generators of I n p I_n^p I n p for small n n n by degree.
For convenience, we recall our main result for the pentad variety, Theorem 1.3, here again as follows.
Theorem 4.7. The pentad variety Pen n ⊂ P N \operatorname{Pen}_n \subset\mathbb{P}^N Pen n ⊂ P N with N = ( n 2 ) − 1 N=\binom{n}{2}-1 N = ( 2 n ) − 1 is irreducible of dimension 3 n − 7 3n-7 3 n − 7 , with prime vanishing ideal I n p \sqrt{I_n^p} I n p . So in particular,
V ( I n p ) = V ( I n p ) = Pen n . V\left(\sqrt{I_n^p}\right)=V(I_n^p)=\operatorname{Pen}_n. V ( I n p ) = V ( I n p ) = Pen n . Moreover, for n ≤ 10 n \le10 n ≤ 10 the ideal I n p I_n^p I n p is radical, hence I n p I_n^p I n p is the prime vanishing ideal. Equivalently, for n ≤ 10 n \le10 n ≤ 10 Conjecture 4.5 holds.
In the remainder of this section we prove the statements in this theorem. We start with the claim about the dimension, which is asserted by the following proposition. Denote by Gr ˚ ( 2 , n ) \mathring{\operatorname{Gr}}(2,n) Gr ˚ ( 2 , n ) the open Grassmannian defined by removing all Plücker hyperplane sections from Gr ( 2 , n ) \operatorname{Gr}(2,n) Gr ( 2 , n ) in its Plücker embedding. Moreover, let Pen ˚ n \mathring{\operatorname{Pen}}_{n} Pen ˚ n be the image of the restriction of Ψ n \Psi_n Ψ n to these open sets.
Proposition 4.8. The pentad variety Pen n ⊂ P N \operatorname{Pen}_n \subset\mathbb{P}^{N} Pen n ⊂ P N has dimension 3 n − 7 3n-7 3 n − 7 , equivalently it has codimension ( n − 3 2 ) \binom{n-3}{2} ( 2 n − 3 ) in P N \mathbb{P}^{N} P N with N = ( n 2 ) − 1 N=\binom{n}{2}-1 N = ( 2 n ) − 1 .
Proof. We start by giving an upper bound for the dimension. First note that H = ( C × ) n − 1 ≅ ( C × ) n / C × H=(\mathbb{C}^{\times})^{n-1}\cong(\mathbb{C}^{\times})^n/\mathbb{C}^{\times} H = ( C × ) n − 1 ≅ ( C × ) n / C × is a subtorus of ( C × ) N (\mathbb{C}^{\times})^N ( C × ) N acting on Gr ˚ ( 2 , n ) \mathring{\operatorname{Gr}}(2,n) Gr ˚ ( 2 , n ) . Hence,
dim Pen ˚ n = dim ( Gr ˚ ( 2 , n ) : Gr ˚ ( 2 , n ) ) (8) \dim\mathring{\operatorname{Pen}}_n=\dim\bigl(\mathring{\operatorname{Gr}}(2,n):\mathring{\operatorname{Gr}}(2,n)\bigr) \tag*{(8)} dim Pen ˚ n = dim ( Gr ˚ ( 2 , n ) : Gr ˚ ( 2 , n ) ) (8)
≤ 2 dim Gr ˚ ( 2 , n ) − dim ( C × ) n − 1 = 4 n − 8 − ( n − 1 ) = 3 n − 7 \leq2\dim\mathring{\operatorname{Gr}}(2,n)-\dim(\mathbb{C}^{\times})^{n-1}=4n-8-(n-1)=3n-7 ≤ 2 dim Gr ˚ ( 2 , n ) − dim ( C × ) n − 1 = 4 n − 8 − ( n − 1 ) = 3 n − 7
by Lemma 2.4 . Next, we show that this bound is tight. To that end we construct a 3 n − 7 3n-7 3 n − 7 dimensional set in Gr ˚ ( 2 , n ) × Gr ˚ ( 2 , n ) \mathring{\operatorname{Gr}}(2,n)\times\mathring{\operatorname{Gr}}(2,n) Gr ˚ ( 2 , n ) × Gr ˚ ( 2 , n ) such that the Jacobian Jac Ψ n \operatorname{Jac}_{\Psi_n} Jac Ψ n has full rank. Consider the following family of 2 × n 2\times n 2 × n matrices A A A and B B B
A = [ ν 1 ⋯ ν n ν 1 x 1 ⋯ ν n x n ] , B = [ μ 1 ⋯ μ n μ 1 y 1 ⋯ μ n y n ] . A=\begin{bmatrix}
\nu_1 & \cdots& \nu_n\\
\nu_1x_1 & \cdots& \nu_nx_n
\end{bmatrix},
\qquad
B=\begin{bmatrix}
\mu_1 & \cdots& \mu_n\\
\mu_1y_1 & \cdots& \mu_ny_n
\end{bmatrix}. A = [ ν 1 ν 1 x 1 ⋯ ⋯ ν n ν n x n ] , B = [ μ 1 μ 1 y 1 ⋯ ⋯ μ n μ n y n ] .
Then, by the PGL 2 \operatorname{PGL}_2 PGL 2 action on A A A and B B B , we can assume that x 1 = 0 = y 1 x_1=0=y_1 x 1 = 0 = y 1 , x 2 = 1 = y 2 x_2=1=y_2 x 2 = 1 = y 2 and x 3 = 2 = y 3 x_3=2=y_3 x 3 = 2 = y 3 . Fixing in addition ν i = 1 \nu_i=1 ν i = 1 for all i i i and μ 1 = 1 \mu_1=1 μ 1 = 1 , we get a ( 3 n − 7 ) (3n-7) ( 3 n − 7 ) -dimensional family V ⊂ Gr ( 2 , n ) × Gr ( 2 , n ) \mathcal{V}\subset\operatorname{Gr}(2,n)\times\operatorname{Gr}(2,n) V ⊂ Gr ( 2 , n ) × Gr ( 2 , n ) of pairs ( B , A ) (B,A) ( B , A ) . Denote the 2 × 2 2\times2 2 × 2 matrices with columns ( 1 , x i ) T (1,x_i)^{T} ( 1 , x i ) T and ( 1 , x j ) T (1,x_j)^{T} ( 1 , x j ) T by ( x i , x j ) (x_i,x_j) ( x i , x j ) , similarly for ( y i , y j ) (y_i,y_j) ( y i , y j ) . Then, for all 1 ≤ i < j ≤ n 1\leq i<j\leq n 1 ≤ i < j ≤ n we get
z i j = det ( λ i y i , λ j y j ) det ( x i , x j ) = λ i λ j y j − y i x j − x i , z_{ij}=\frac{\det(\lambda_i y_i,\lambda_j y_j)}{\det(x_i,x_j)}
=\lambda_i\lambda_j\frac{y_j-y_i}{x_j-x_i}, z ij = det ( x i , x j ) det ( λ i y i , λ j y j ) = λ i λ j x j − x i y j − y i ,
where λ i = μ i ∈ C × \lambda_i=\mu_i\in\mathbb{C}^{\times} λ i = μ i ∈ C × and λ 1 = 1 \lambda_1=1 λ 1 = 1 . Hence, the free parameters are λ 2 , … , λ n \lambda_2,\ldots,\lambda_n λ 2 , … , λ n , x 4 , … , x n x_4,\ldots,x_n x 4 , … , x n and y 4 , … , y n y_4,\ldots,y_n y 4 , … , y n , giving ( n − 1 ) + 2 ( n − 3 ) = 3 n − 7 (n-1)+2(n-3)=3n-7 ( n − 1 ) + 2 ( n − 3 ) = 3 n − 7 free parameters. We will use the following 3 n − 7 3n-7 3 n − 7 cross-ratios on P N \mathbb{P}^{N} P N defined for all j ≥ 4 j\geq4 j ≥ 4 :
u j = z 1 j z 23 z 12 z 13 , u 2 = z 23 z 13 , u 3 = z 23 z 12 , u_j=\frac{z_{1j}z_{23}}{z_{12}z_{13}},\qquad
u_2=\frac{z_{23}}{z_{13}},\qquad
u_3=\frac{z_{23}}{z_{12}}, u j = z 12 z 13 z 1 j z 23 , u 2 = z 13 z 23 , u 3 = z 12 z 23 ,
v j = z 2 j z 13 z 1 j z 23 , v_j=\frac{z_{2j}z_{13}}{z_{1j}z_{23}}, v j = z 1 j z 23 z 2 j z 13 ,
w j = z 3 j z 12 z 1 j z 23 . w_j=\frac{z_{3j}z_{12}}{z_{1j}z_{23}}. w j = z 1 j z 23 z 3 j z 12 .
These cross ratios give rise to a morphism ϕ : Pen ˚ n → C ( u , v , w ) 3 n − 7 \phi:\mathring{\operatorname{Pen}}_n\to\mathbb{C}^{3n-7}_{(u,v,w)} ϕ : Pen ˚ n → C ( u , v , w ) 3 n − 7 . Note with our choice of V \mathcal{V} V from above, these cross ratios simplify on Ψ n ( V ) \Psi_n(\mathcal{V}) Ψ n ( V ) to
u j = λ j y j x j , u 2 = λ 2 , u 3 = λ 3 , u_j=\lambda_j\frac{y_j}{x_j},\qquad u_2=\lambda_2,\qquad u_3=\lambda_3, u j = λ j x j y j , u 2 = λ 2 , u 3 = λ 3 ,
v j = x j ( y j − 1 ) y j ( x j − 1 ) , v_j=\frac{x_j(y_j-1)}{y_j(x_j-1)}, v j = y j ( x j − 1 ) x j ( y j − 1 ) ,
w j = x j ( y j − 2 ) y j ( x j − 2 ) . w_j=\frac{x_j(y_j-2)}{y_j(x_j-2)}. w j = y j ( x j − 2 ) x j ( y j − 2 ) .
Moreover, note that each ratio only depends on x j x_j x j , y j y_j y j , and λ j \lambda_j λ j for the same index j j j , so in particular
∂ u j ∂ x i = 0 , ∂ u j ∂ y i = 0 , ∂ u j ∂ λ i = 0 , ∂ v j ∂ x i = 0 , ∂ v j ∂ y i = 0 , ∂ v j ∂ λ i = 0 , ∂ w j ∂ x i = 0 , ∂ w j ∂ y i = 0 , ∂ w j ∂ λ i = 0 , (9) \begin{aligned}
\frac{\partial u_j}{\partial x_i}&=0,\quad\frac{\partial u_j}{\partial y_i}=0,\quad\frac{\partial u_j}{\partial\lambda_i}=0,\\
\frac{\partial v_j}{\partial x_i}&=0,\quad\frac{\partial v_j}{\partial y_i}=0,\quad\frac{\partial v_j}{\partial\lambda_i}=0,\\
\frac{\partial w_j}{\partial x_i}&=0,\quad\frac{\partial w_j}{\partial y_i}=0,\quad\frac{\partial w_j}{\partial\lambda_i}=0, \tag*{(9)}
\end{aligned} ∂ x i ∂ u j ∂ x i ∂ v j ∂ x i ∂ w j = 0 , ∂ y i ∂ u j = 0 , ∂ λ i ∂ u j = 0 , = 0 , ∂ y i ∂ v j = 0 , ∂ λ i ∂ v j = 0 , = 0 , ∂ y i ∂ w j = 0 , ∂ λ i ∂ w j = 0 , (9)
for all i ≠ j i \ne j i = j . Next consider the Jacobian J = Jac ϕ ∘ Ψ n J=\operatorname{Jac}_{\phi\circ\Psi_n} J = Jac ϕ ∘ Ψ n of the composition of the crossratio and the Hadamard ratio ϕ ∘ Ψ n : C ( λ , x , y ) 3 n − 7 ⇢ C ( u , v , w ) 3 n − 7 \phi\circ\Psi_n:\mathbb{C}^{3n-7}_{(\lambda,x,y)}\dashrightarrow\mathbb{C}^{3n-7}_{(u,v,w)} ϕ ∘ Ψ n : C ( λ , x , y ) 3 n − 7 ⇢ C ( u , v , w ) 3 n − 7 , and note rank J ≤ rank Jac Ψ n \operatorname{rank}J\leq\operatorname{rank}\operatorname{Jac}_{\Psi_n} rank J ≤ rank Jac Ψ n , hence it suffices to show that rank J = 3 n − 7 \operatorname{rank}J=3n-7 rank J = 3 n − 7 . We compute the rank of J J J with respect to an ordered basis ( λ 2 , λ 3 , λ 4 , x 4 , y 4 , … , λ n , x n , y n ) (\lambda_2,\lambda_3,\lambda_4,x_4,y_4,\ldots,\lambda_n,x_n,y_n) ( λ 2 , λ 3 , λ 4 , x 4 , y 4 , … , λ n , x n , y n ) of C ( λ , x , y ) 3 n − 7 \mathbb{C}^{3n-7}_{(\lambda,x,y)} C ( λ , x , y ) 3 n − 7 and ( u 2 , u 3 , u 4 , v 4 , w 4 , … , u n , v n , w n ) (u_2,u_3,u_4,v_4,w_4,\ldots,u_n,v_n,w_n) ( u 2 , u 3 , u 4 , v 4 , w 4 , … , u n , v n , w n ) of C ( u , v , w ) 3 n − 7 \mathbb{C}^{3n-7}_{(u,v,w)} C ( u , v , w ) 3 n − 7 . Then, J J J is block diagonal with
J = 1 ⊕ 1 ⊕ ⨁ i = 4 n J i , J=1\oplus1\oplus\bigoplus_{i=4}^{n}J_i, J = 1 ⊕ 1 ⊕ i = 4 ⨁ n J i ,
by (9) where
J i = [ ∂ u i ∂ λ i ∂ u i ∂ x i ∂ u i ∂ y i ∂ v i ∂ λ i ∂ v i ∂ x i ∂ v i ∂ y i ∂ w i ∂ λ i ∂ w i ∂ x i ∂ w i ∂ y i ] for all i ≥ 4. J_i=
\begin{bmatrix}
\frac{\partial u_i}{\partial\lambda_i} & \frac{\partial u_i}{\partial x_i} & \frac{\partial u_i}{\partial y_i}\\
\frac{\partial v_i}{\partial\lambda_i} & \frac{\partial v_i}{\partial x_i} & \frac{\partial v_i}{\partial y_i}\\
\frac{\partial w_i}{\partial\lambda_i} & \frac{\partial w_i}{\partial x_i} & \frac{\partial w_i}{\partial y_i}
\end{bmatrix}
\quad\text{for all }i\geq4. J i = ∂ λ i ∂ u i ∂ λ i ∂ v i ∂ λ i ∂ w i ∂ x i ∂ u i ∂ x i ∂ v i ∂ x i ∂ w i ∂ y i ∂ u i ∂ y i ∂ v i ∂ y i ∂ w i for all i ≥ 4.
Then, note that for all i i i
det J i = 2 y i − x i y i 2 ( x i − 1 ) 2 ( x i − 2 ) 2 , \det J_i=2\frac{y_i-x_i}{y_i^2(x_i-1)^2(x_i-2)^2}, det J i = 2 y i 2 ( x i − 1 ) 2 ( x i − 2 ) 2 y i − x i ,
hence
det J = ∏ i = 4 n det J i = 2 n − 3 ∏ i = 4 n y i − x i y i 2 ( x i − 1 ) 2 ( x i − 2 ) 2 . \det J=\prod_{i=4}^{n}\det J_i=2^{n-3}\prod_{i=4}^{n}\frac{y_i-x_i}{y_i^2(x_i-1)^2(x_i-2)^2}. det J = i = 4 ∏ n det J i = 2 n − 3 i = 4 ∏ n y i 2 ( x i − 1 ) 2 ( x i − 2 ) 2 y i − x i .
For a generic choice of ( λ , x , y ) ∈ ( C × ) n − 1 × C n − 3 × C n − 3 (\lambda,x,y)\in(\mathbb{C}^{\times})^{n-1}\times\mathbb{C}^{n-3}\times\mathbb{C}^{n-3} ( λ , x , y ) ∈ ( C × ) n − 1 × C n − 3 × C n − 3 , this determinant will not vanish, hence J J J has generically full rank 3 n − 7 3n-7 3 n − 7 . This shows that rank Jac Ψ n ≥ rank J = 3 n − 7 \operatorname{rank}\operatorname{Jac}_{\Psi_n}\geq\operatorname{rank}J=3n-7 rank Jac Ψ n ≥ rank J = 3 n − 7 . Therefore, dim im Ψ n ≥ dim im Ψ n ∣ V ≥ 3 n − 7 \dim\operatorname{im}\Psi_n\geq\dim\operatorname{im}\left.\Psi_n\right|_V\geq3n-7 dim im Ψ n ≥ dim im Ψ n ∣ V ≥ 3 n − 7 . Since dim Pen n ≤ 3 n − 7 \dim\operatorname{Pen}_n\leq3n-7 dim Pen n ≤ 3 n − 7 by (8) , the set Ψ ( V ) \Psi(V) Ψ ( V ) is dense in Pen n \operatorname{Pen}_n Pen n . Therefore, dim Pen n = 3 n − 7 \dim\operatorname{Pen}_n=3n-7 dim Pen n = 3 n − 7 as had to be shown.
Remark 4.9. Proposition 4.8 also shows that the generic fiber of Ψ n \Psi_n Ψ n has dimension n − 1 n-1 n − 1 . Note that there is an ( n − 1 ) (n-1) ( n − 1 ) -dimensional algebraic torus contained in each fiber as follows. On the Plücker coordinates on each of the two factors, there is an action of the ( n − 1 ) (n-1) ( n − 1 ) -dimensional torus ( C × ) n − 1 (\mathbb{C}^{\times})^{n-1} ( C × ) n − 1 , rescaling both p i j p_{ij} p ij and q i j q_{ij} q ij by λ i λ j \lambda_i\lambda_j λ i λ j . Their ratio is constant and the fiber over each point contains ( C × ) n − 1 (\mathbb{C}^{\times})^{n-1} ( C × ) n − 1 . Since the generic fiber is ( n − 1 ) (n-1) ( n − 1 ) -dimensional, an irreducible component thereof is a Zariski closure of ( C × ) n − 1 (\mathbb{C}^{\times})^{n-1} ( C × ) n − 1 .
The pentad variety and the MHV gravity amplitudes can be computationally connected via the following matrix-completion property of the pentad from (5) .
Proposition 4.10. For n = 5 n=5 n = 5 , there exists a Zariski dense subset of the affine cone over the pentad variety Pen 5 \operatorname{Pen}_5 Pen 5 , where each point lifts to a unique symmetric matrix Z 5 Z_5 Z 5 with the following propety: its diagonal entries z i i z_{ii} z ii are given by rational functions f i = g i / h i f_i=g_i/h_i f i = g i / h i , g i , h i ∈ C [ z i j ] g_i,h_i\in\mathbb{C}[z_{ij}] g i , h i ∈ C [ z ij ] , in the off-diagonal entries z i j z_{ij} z ij in such a way that rank Z 5 ≤ 2 \operatorname{rank}Z_5\leq2 rank Z 5 ≤ 2 .
Proof. Introduce the following rational functions in the off-diagonal entries:
f 1 = z ~ 12 z ~ 14 z ~ 35 − z ~ 12 z ~ 15 z ~ 34 − z ~ 13 z ~ 14 z ~ 25 + z ~ 13 z ~ 15 z ~ 24 z ~ 24 z ~ 35 − z ~ 25 z ~ 34 , f_1=\frac{\widetilde{z}_{12}\widetilde{z}_{14}\widetilde{z}_{35}-\widetilde{z}_{12}\widetilde{z}_{15}\widetilde{z}_{34}-\widetilde{z}_{13}\widetilde{z}_{14}\widetilde{z}_{25}+\widetilde{z}_{13}\widetilde{z}_{15}\widetilde{z}_{24}}{\widetilde{z}_{24}\widetilde{z}_{35}-\widetilde{z}_{25}\widetilde{z}_{34}}, f 1 = z 24 z 35 − z 25 z 34 z 12 z 14 z 35 − z 12 z 15 z 34 − z 13 z 14 z 25 + z 13 z 15 z 24 ,
f 2 = z ~ 12 z ~ 23 z ~ 45 − z ~ 12 z ~ 24 z ~ 35 + z ~ 13 z ~ 24 z ~ 25 − z ~ 14 z ~ 23 z ~ 25 z ~ 13 z ~ 45 − z ~ 14 z ~ 35 , f_2=\frac{\widetilde{z}_{12}\widetilde{z}_{23}\widetilde{z}_{45}-\widetilde{z}_{12}\widetilde{z}_{24}\widetilde{z}_{35}+\widetilde{z}_{13}\widetilde{z}_{24}\widetilde{z}_{25}-\widetilde{z}_{14}\widetilde{z}_{23}\widetilde{z}_{25}}{\widetilde{z}_{13}\widetilde{z}_{45}-\widetilde{z}_{14}\widetilde{z}_{35}}, f 2 = z 13 z 45 − z 14 z 35 z 12 z 23 z 45 − z 12 z 24 z 35 + z 13 z 24 z 25 − z 14 z 23 z 25 ,
f 3 = − z ~ 13 z ~ 24 z ~ 35 − z ~ 13 z ~ 25 z ~ 34 − z ~ 14 z ~ 23 z ~ 35 + z ~ 15 z ~ 23 z ~ 34 z ~ 14 z ~ 25 − z ~ 15 z ~ 24 , f_3=-\frac{\widetilde{z}_{13}\widetilde{z}_{24}\widetilde{z}_{35}-\widetilde{z}_{13}\widetilde{z}_{25}\widetilde{z}_{34}-\widetilde{z}_{14}\widetilde{z}_{23}\widetilde{z}_{35}+\widetilde{z}_{15}\widetilde{z}_{23}\widetilde{z}_{34}}{\widetilde{z}_{14}\widetilde{z}_{25}-\widetilde{z}_{15}\widetilde{z}_{24}}, f 3 = − z 14 z 25 − z 15 z 24 z 13 z 24 z 35 − z 13 z 25 z 34 − z 14 z 23 z 35 + z 15 z 23 z 34 ,
f 4 = z ~ 12 z ~ 34 z ~ 45 − z ~ 13 z ~ 24 z ~ 45 + z ~ 14 z ~ 24 z ~ 35 − z ~ 14 z ~ 25 z ~ 34 z ~ 12 z ~ 35 − z ~ 13 z ~ 25 , f_4=\frac{\widetilde{z}_{12}\widetilde{z}_{34}\widetilde{z}_{45}-\widetilde{z}_{13}\widetilde{z}_{24}\widetilde{z}_{45}+\widetilde{z}_{14}\widetilde{z}_{24}\widetilde{z}_{35}-\widetilde{z}_{14}\widetilde{z}_{25}\widetilde{z}_{34}}{\widetilde{z}_{12}\widetilde{z}_{35}-\widetilde{z}_{13}\widetilde{z}_{25}}, f 4 = z 12 z 35 − z 13 z 25 z 12 z 34 z 45 − z 13 z 24 z 45 + z 14 z 24 z 35 − z 14 z 25 z 34 ,
f 5 = z ~ 13 z ~ 25 z ~ 45 − z ~ 14 z ~ 25 z ~ 35 − z ~ 15 z ~ 23 z ~ 45 + z ~ 15 z ~ 24 z ~ 35 z ~ 13 z ~ 24 − z ~ 14 z ~ 23 . f_5=\frac{\widetilde{z}_{13}\widetilde{z}_{25}\widetilde{z}_{45}-\widetilde{z}_{14}\widetilde{z}_{25}\widetilde{z}_{35}-\widetilde{z}_{15}\widetilde{z}_{23}\widetilde{z}_{45}+\widetilde{z}_{15}\widetilde{z}_{24}\widetilde{z}_{35}}{\widetilde{z}_{13}\widetilde{z}_{24}-\widetilde{z}_{14}\widetilde{z}_{23}}. f 5 = z 13 z 24 − z 14 z 23 z 13 z 25 z 45 − z 14 z 25 z 35 − z 15 z 23 z 45 + z 15 z 24 z 35 .
Let
Z 5 = ( f 1 z ~ 12 z ~ 13 z ~ 14 z ~ 15 z ~ 12 f 2 z ~ 23 z ~ 24 z ~ 25 z ~ 13 z ~ 23 f 3 z ~ 34 z ~ 35 z ~ 14 z ~ 24 z ~ 34 f 4 z ~ 45 z ~ 15 z ~ 25 z ~ 35 z ~ 45 f 5 ) . Z_5=\begin{pmatrix}
f_1 & \widetilde{z}_{12} & \widetilde{z}_{13} & \widetilde{z}_{14} & \widetilde{z}_{15}\\
\widetilde{z}_{12} & f_2 & \widetilde{z}_{23} & \widetilde{z}_{24} & \widetilde{z}_{25}\\
\widetilde{z}_{13} & \widetilde{z}_{23} & f_3 & \widetilde{z}_{34} & \widetilde{z}_{35}\\
\widetilde{z}_{14} & \widetilde{z}_{24} & \widetilde{z}_{34} & f_4 & \widetilde{z}_{45}\\
\widetilde{z}_{15} & \widetilde{z}_{25} & \widetilde{z}_{35} & \widetilde{z}_{45} & f_5
\end{pmatrix}. Z 5 = f 1 z 12 z 13 z 14 z 15 z 12 f 2 z 23 z 24 z 25 z 13 z 23 f 3 z 34 z 35 z 14 z 24 z 34 f 4 z 45 z 15 z 25 z 35 z 45 f 5 .
Let U U U be the Zariski dense subset in the affine space of symmetric 5 × 5 5 \times5 5 × 5 matrices in the z ~ i j \widetilde{z}_{ij} z ij ’s (including the diagonal entries), where all the f i f_i f i ’s make regular functions. One can check in Macaulay2 that each 3 × 3 3 \times3 3 × 3 minor of Z 5 Z_5 Z 5 is either identically zero or it is a multiple of the pentad pen 12345 \operatorname{pen}_{12345} pen 12345 . Now, let project U U U to the affine space of the off-diagonal entries and obtain an open set V V V . Letting W = V ∩ Pen ^ 5 ≠ ∅ W=V\cap\widehat{\operatorname{Pen}}_5\ne\varnothing W = V ∩ Pen 5 = ∅ , the intersection of V V V with the affine cone over Pen 5 \operatorname{Pen}_5 Pen 5 , we see that any point in W W W is such that its lift Z 5 ∈ U Z_5\in U Z 5 ∈ U has rank Z 5 = 2 Z_5=2 Z 5 = 2 . By the construction of the f i f_i f i ’s, such a lift is unique on W W W . □
Remark 4.11. It is easy to see that the uniqueness statement fails on on the entire variety Pen 5 \operatorname{Pen}_5 Pen 5 .
We end this subsection by recording a technical result that will be useful later to prove our main theorem.
Lemma 4.12. Recall that Z n = C [ z ~ i j ∣ 1 ≤ i < j ≤ n ] Z_n=\mathbb{C}[\widetilde{z}_{ij}\mid1\le i<j\le n] Z n = C [ z ij ∣ 1 ≤ i < j ≤ n ] and let Z n ± Z_n^{\pm} Z n ± its Laurent polynomial ring. Let J n p J_n^p J n p and I n p I_n^p I n p be the ideals in Z n Z_n Z n defined in (6) and (7) . Then, I n p = ( J n p ⋅ Z n ± ) ∩ Z n I_n^p=(J_n^p\cdot Z_n^{\pm})\cap Z_n I n p = ( J n p ⋅ Z n ± ) ∩ Z n . The same holds true for the radicals, i.e. I n p = ( J n p ⋅ Z n ± ) ∩ Z n \sqrt{I_n^p}=(\sqrt{J_n^p}\cdot Z_n^{\pm})\cap Z_n I n p = ( J n p ⋅ Z n ± ) ∩ Z n .
Proof. Let m = ∏ 1 ≤ i < j ≤ n z ~ i j m=\prod_{1\le i<j\le n}\widetilde{z}_{ij} m = ∏ 1 ≤ i < j ≤ n z ij . Define the multiplicative set S = { m i , i ≥ 0 } S=\{m^i,i\ge0\} S = { m i , i ≥ 0 } . Then the localization of Z n Z_n Z n at S S S is
S − 1 Z n = Z n [ m − 1 ] = Z n ± , S^{-1}Z_n=Z_n[m^{-1}]=Z_n^{\pm}, S − 1 Z n = Z n [ m − 1 ] = Z n ± ,
which is the coordinate ring of the torus T = { z ~ i j ≠ 0 for all i < j } T=\{\widetilde{z}_{ij}\ne0\text{ for all }i<j\} T = { z ij = 0 for all i < j } . Let J n p ⊆ Z n J_n^p\subseteq Z_n J n p ⊆ Z n be the ideal generated by the pentads, and recall that
I n p = ( J n p : m ∞ ) = { f ∈ Z n ∣ ∃ k ≥ 0 such that m k f ∈ J n p } . I_n^p=(J_n^p:m^\infty)=\{f\in Z_n\mid\exists k\ge0\text{ such that }m^k f\in J_n^p\}. I n p = ( J n p : m ∞ ) = { f ∈ Z n ∣ ∃ k ≥ 0 such that m k f ∈ J n p } .
We claim that
I n p = ( J n p ⋅ Z n ± ) ∩ C [ z ~ i j ] . I_n^p=(J_n^p\cdot Z_n^{\pm})\cap\mathbb{C}[\widetilde{z}_{ij}]. I n p = ( J n p ⋅ Z n ± ) ∩ C [ z ij ] .
In other words, I n p I_n^p I n p is exactly the contraction to Z n Z_n Z n of the extension of J n p J_n^p J n p to the torus coordinate ring. Let f ∈ I n p f\in I_n^p f ∈ I n p . By definition of saturation, there exists k ≥ 0 k\ge0 k ≥ 0 such that m k f ∈ J n p m^k f\in J_n^p m k f ∈ J n p . Passing to the localization Z n [ m − 1 ] Z_n[m^{-1}] Z n [ m − 1 ] , we may divide by m k m^k m k , and obtain
f = m k f m k ∈ J n p ⋅ Z n [ m − 1 ] = J n p ⋅ Z n ± . f=\frac{m^k f}{m^k}\in J_n^p\cdot Z_n[m^{-1}]=J_n^p\cdot Z_n^{\pm}. f = m k m k f ∈ J n p ⋅ Z n [ m − 1 ] = J n p ⋅ Z n ± .
Since already f ∈ Z n f \in Z_n f ∈ Z n , it follows that f ∈ ( J n p ⋅ Z n ± ) ∩ C [ z i j ] f \in(J_n^p \cdot Z_n^{\pm}) \cap\mathbb{C}[z_{ij}] f ∈ ( J n p ⋅ Z n ± ) ∩ C [ z ij ] . Now we show the converse, so assume that f ∈ ( J n p ⋅ Z n ± ) ∩ C [ z i j ] f \in(J_n^p \cdot Z_n^{\pm}) \cap\mathbb{C}[z_{ij}] f ∈ ( J n p ⋅ Z n ± ) ∩ C [ z ij ] . So in the localized ring Z n [ m − 1 ] Z_n[m^{-1}] Z n [ m − 1 ] we can write f = ∑ ℓ = 1 r g ℓ h ℓ f = \sum_{\ell=1}^{r} g_{\ell}h_{\ell} f = ∑ ℓ = 1 r g ℓ h ℓ , g ℓ ∈ J n p g_{\ell} \in J_n^p g ℓ ∈ J n p , h ℓ ∈ Z n [ m − 1 ] h_{\ell} \in Z_n[m^{-1}] h ℓ ∈ Z n [ m − 1 ] . Each h ℓ h_{\ell} h ℓ has the form h ℓ = a ℓ m N ℓ h_{\ell} = \frac{a_{\ell}}{m^{N_{\ell}}} h ℓ = m N ℓ a ℓ where a ℓ ∈ Z n a_{\ell} \in Z_n a ℓ ∈ Z n . Choose N N N large enough so that N ≥ N ℓ N \geq N_{\ell} N ≥ N ℓ for all ℓ \ell ℓ . Then
m N f = ∑ ℓ = 1 r g ℓ a ℓ m N − N ℓ . m^N f = \sum_{\ell=1}^{r} g_{\ell}a_{\ell}m^{N-N_{\ell}}. m N f = ℓ = 1 ∑ r g ℓ a ℓ m N − N ℓ .
Since g ℓ ∈ J n p g_{\ell} \in J_n^p g ℓ ∈ J n p and J n p J_n^p J n p is an ideal of Z n Z_n Z n , every term on the right-hand side lies in J n p J_n^p J n p . Therefore m N f ∈ J n p m^Nf \in J_n^p m N f ∈ J n p . This implies that f ∈ J n p : m ∞ = I n p f \in J_n^p:m^\infty= I_n^p f ∈ J n p : m ∞ = I n p . The proof for the statement about the radical ideals is identical, after replacing f f f with a suitable power f ℓ f^\ell f ℓ .
Remark 4.13. Suppose that the localized ideal J n p ⋅ Z n ± J_n^p \cdot Z_n^{\pm} J n p ⋅ Z n ± is a prime ideal of Z n ± Z_n^{\pm} Z n ± . Then its contraction I n p = ( J n p ⋅ Z n ± ) ∩ Z n I_n^p=(J_n^p\cdot Z_n^{\pm})\cap Z_n I n p = ( J n p ⋅ Z n ± ) ∩ Z n is prime in Z n Z_n Z n . Indeed, if a b ∈ I n p ab \in I_n^p ab ∈ I n p for some a , b ∈ Z n a,b \in Z_n a , b ∈ Z n , then a b ∈ J n p ⋅ Z n ± ab \in J_n^p\cdot Z_n^{\pm} ab ∈ J n p ⋅ Z n ± . Since the latter ideal is prime, either a ∈ J n p ⋅ Z n ± a \in J_n^p\cdot Z_n^{\pm} a ∈ J n p ⋅ Z n ± or b ∈ J n p ⋅ Z n ± b \in J_n^p\cdot Z_n^{\pm} b ∈ J n p ⋅ Z n ± . Because a , b ∈ Z n a,b \in Z_n a , b ∈ Z n , this implies a ∈ I n p a \in I_n^p a ∈ I n p or b ∈ I n p b \in I_n^p b ∈ I n p . Hence I n p I_n^p I n p is prime.
Affine Slope Variety
The purpose of this section is to prove our main result, Theorem 1.3 , on the pentad variety. In order to do so, we define affine slope varieties as first introduced by Martin in [14 ] . It turns out that there is a nice connection between these affine slope varieties and the pentad via a Hadamard product with a toric variety, which plays the role of the little-group scaling, see Corollary 1.4 . Moreover, we also realize the affine slope variety as the Hadamard ratio of two copies of the moduli space of n n n marked points on P 1 \mathbb{P}^{1} P 1 , see Proposition 4.19 .
Definition 4.14 (Affine slope variety). Consider h ≥ 2 h \geq2 h ≥ 2 distinct points p 1 , … , p h p_1,\ldots,p_h p 1 , … , p h with coordinates ( x i , y i ) (x_i,y_i) ( x i , y i ) in the affine complex plane A 2 \mathbb{A}^{2} A 2 , such that none of the ( h 2 ) \binom{h}{2} ( 2 h ) lines L i j L_{ij} L ij connecting any pair of points p i p_i p i and p j p_j p j is vertical. Let m i j = y j − y i x j − x i m_{ij}=\frac{y_j-y_i}{x_j-x_i} m ij = x j − x i y j − y i be the indeterminates given by the slopes of the lines L i j L_{ij} L ij . The Zariski closure of all slope vectors ( m 12 , … , m h − 1 h ) ∈ A N (m_{12},\ldots,m_{h-1\,h}) \in\mathbb{A}^{N} ( m 12 , … , m h − 1 h ) ∈ A N with N = ( h 2 ) N=\binom{h}{2} N = ( 2 h ) is the algebraic variety S h ⊂ A N S_h \subset\mathbb{A}^{N} S h ⊂ A N called the affine slope variety .
There is an equivalent description of the affine slope variety, also due to [14 ] , in terms of graphs, which will be useful to us. Let v 1 , v 2 , … , v h v_1,v_2,\ldots,v_h v 1 , v 2 , … , v h be distinct vertices. The ( h − 1 ) (h-1) ( h − 1 ) -wheel W = W ( v 1 ; v 2 , … , v h ) W=W(v_1;v_2,\ldots,v_h) W = W ( v 1 ; v 2 , … , v h ) is defined as the graph with the edge set
E ( W ) : = { v 1 v 2 , v 1 v 3 , … , v 1 v h } ∪ { v 2 v 3 , v 3 v 4 … , v h v 2 } . E(W):=\{v_1v_2,v_1v_3,\ldots,v_1v_h\}\cup\{v_2v_3,v_3v_4\ldots,v_hv_2\}. E ( W ) := { v 1 v 2 , v 1 v 3 , … , v 1 v h } ∪ { v 2 v 3 , v 3 v 4 … , v h v 2 } .
Next consider the ( h − 1 ) × ( h − 1 ) (h-1)\times(h-1) ( h − 1 ) × ( h − 1 ) matrix
M v 1 , v 2 , … , v h : = [ m v 1 v 2 − m v 2 v 3 m v 2 v 3 − m v 1 v 3 0 ⋯ 0 0 m v 1 v 3 − m v 3 v 4 m v 3 v 4 − m v 1 v 4 ⋯ 0 ⋮ ⋮ ⋮ ⋱ ⋮ m v 2 v h − m v 1 v 2 0 0 ⋯ m v 1 v h − m v 2 v h ] . M_{v_1,v_2,\ldots,v_h}:=
\begin{bmatrix}
m_{v_1v_2}-m_{v_2v_3} & m_{v_2v_3}-m_{v_1v_3} & 0 & \cdots& 0\\
0 & m_{v_1v_3}-m_{v_3v_4} & m_{v_3v_4}-m_{v_1v_4} & \cdots& 0\\
\vdots& \vdots& \vdots& \ddots& \vdots\\
m_{v_2v_h}-m_{v_1v_2} & 0 & 0 & \cdots& m_{v_1v_h}-m_{v_2v_h}
\end{bmatrix}. M v 1 , v 2 , … , v h := m v 1 v 2 − m v 2 v 3 0 ⋮ m v 2 v h − m v 1 v 2 m v 2 v 3 − m v 1 v 3 m v 1 v 3 − m v 3 v 4 ⋮ 0 0 m v 3 v 4 − m v 1 v 4 ⋮ 0 ⋯ ⋯ ⋱ ⋯ 0 0 ⋮ m v 1 v h − m v 2 v h .
We adopt the convention that v h + i = v i + 1 v_{h+i}=v_{i+1} v h + i = v i + 1 for all 1 ≤ i ≤ h − 1 1\leq i\leq h-1 1 ≤ i ≤ h − 1 . We call the determinants
τ v 1 , v 2 , … , v h : = det M v 1 , v 2 , … , v h = ∏ i = 2 h ( m v 1 , v i − m v i , v i + 1 ) − ∏ i = 2 h ( m v 1 , v i + 1 − m v i , v i + 1 ) . (10) \tau_{v_1,v_2,\ldots,v_h}:=\det M_{v_1,v_2,\ldots,v_h}
=\prod_{i=2}^{h}(m_{v_1,v_i}-m_{v_i,v_{i+1}})
-\prod_{i=2}^{h}(m_{v_1,v_{i+1}}-m_{v_i,v_{i+1}}). \tag*{(10)} τ v 1 , v 2 , … , v h := det M v 1 , v 2 , … , v h = i = 2 ∏ h ( m v 1 , v i − m v i , v i + 1 ) − i = 2 ∏ h ( m v 1 , v i + 1 − m v i , v i + 1 ) . (10)
the tree-polynomial of the wheel W = W ( v 1 ; v 2 , … , v h ) W = W(v_1; v_2, \ldots, v_h) W = W ( v 1 ; v 2 , … , v h ) , and sometimes also write τ ( W ) \tau(W) τ ( W ) for τ v 1 , v 2 , … , v h \tau_{v_1,v_2,\ldots,v_h} τ v 1 , v 2 , … , v h .
Example 4.15. Fix a set of vertices labeled as { 1 , 2 , 3 , 4 } \{1,2,3,4\} { 1 , 2 , 3 , 4 } . Then, the 3-wheel W 3 = W ( 1 ; 2 , 3 , 4 ) W_3 = W(1;2,3,4) W 3 = W ( 1 ; 2 , 3 , 4 ) is the complete graph K 4 K_4 K 4 . The matrix from above reads
M W 3 = [ m 12 − m 23 m 23 − m 13 0 0 m 13 − m 34 m 34 − m 14 m 24 − m 12 0 m 14 − m 24 ] . M_{W_3} =
\begin{bmatrix}
m_{12}-m_{23} & m_{23}-m_{13} & 0 \\
0 & m_{13}-m_{34} & m_{34}-m_{14} \\
m_{24}-m_{12} & 0 & m_{14}-m_{24}
\end{bmatrix}. M W 3 = m 12 − m 23 0 m 24 − m 12 m 23 − m 13 m 13 − m 34 0 0 m 34 − m 14 m 14 − m 24 .
We obtain the tree polynomial of W 3 W_3 W 3 as
τ ( W 3 ) = det M W 3 = − m 12 m 13 m 24 + m 12 m 13 m 34 + m 12 m 14 m 23 − m 12 m 14 m 34 − m 12 m 23 m 34 + m 12 m 24 m 34 − m 13 m 14 m 23 + m 13 m 14 m 24 + m 13 m 23 m 24 − m 13 m 24 m 34 − m 14 m 23 m 24 + m 14 m 23 m 34 . \tau(W_3) = \det M_{W_3} = -m_{12}m_{13}m_{24} + m_{12}m_{13}m_{34} + m_{12}m_{14}m_{23} - m_{12}m_{14}m_{34}
\\
{} - m_{12}m_{23}m_{34} + m_{12}m_{24}m_{34} - m_{13}m_{14}m_{23} + m_{13}m_{14}m_{24}
\\
{} + m_{13}m_{23}m_{24} - m_{13}m_{24}m_{34} - m_{14}m_{23}m_{24} + m_{14}m_{23}m_{34}. τ ( W 3 ) = det M W 3 = − m 12 m 13 m 24 + m 12 m 13 m 34 + m 12 m 14 m 23 − m 12 m 14 m 34 − m 12 m 23 m 34 + m 12 m 24 m 34 − m 13 m 14 m 23 + m 13 m 14 m 24 + m 13 m 23 m 24 − m 13 m 24 m 34 − m 14 m 23 m 24 + m 14 m 23 m 34 .
Note that this polynomial is square-free and has 12 monomials much like the pentad constraint in (5) .
We fix some more notation. Denote by A h A_h A h the tensor product
A h : = C [ λ i ∣ 1 ≤ i ≤ h ] ⊗ C C [ m i j ∣ 1 ≤ i < j ≤ h ] . A_h := \mathbb{C}[\lambda_i \mid1 \leq i \leq h] \otimes_{\mathbb{C}} \mathbb{C}[m_{ij} \mid1 \leq i < j \leq h]. A h := C [ λ i ∣ 1 ≤ i ≤ h ] ⊗ C C [ m ij ∣ 1 ≤ i < j ≤ h ] .
Recall that Z n = C [ z i j ∣ 1 ≤ i < j ≤ n ] Z_n = \mathbb{C}[z_{ij} \mid1 \leq i < j \leq n] Z n = C [ z ij ∣ 1 ≤ i < j ≤ n ] . Denote by A h ± A_h^{\pm} A h ± the Laurent polynomial ring in the variables of A h A_h A h , analogously for Z n ± Z_n^{\pm} Z n ± . We then define the following two ideals in C [ m i j ∣ 1 ≤ i < j ≤ h ] \mathbb{C}[m_{ij} \mid1 \leq i < j \leq h] C [ m ij ∣ 1 ≤ i < j ≤ h ] . Let J h s J_h^s J h s be the ideal generated by the tree polynomials of any 3-wheel subgraph of the complete graph K h K_h K h on h h h vertices, i.e.,
J h s : = ⟨ τ ( W 3 ) ∣ W 3 wheel subgraph of K h ⟩ = ⟨ det M i j k l ∣ W 3 = W ( i ; j , k , l ) with { i , j , k , l } ⊂ [ h ] ⟩ ⊂ C [ m i j ∣ 1 ≤ i < j ≤ h ] . J_h^s := \langle\tau(W_3) \mid W_3\text{ wheel subgraph of }K_h\rangle
\\
= \langle\det M_{ijkl} \mid W_3 = W(i;j,k,l)\text{ with }\{i,j,k,l\} \subset[h]\rangle\subset\mathbb{C}[m_{ij} \mid1 \leq i < j \leq h]. J h s := ⟨ τ ( W 3 ) ∣ W 3 wheel subgraph of K h ⟩ = ⟨ det M ij k l ∣ W 3 = W ( i ; j , k , l ) with { i , j , k , l } ⊂ [ h ]⟩ ⊂ C [ m ij ∣ 1 ≤ i < j ≤ h ] .
Similarly, let I h s I_h^s I h s be the ideal in C [ m i j ∣ 1 ≤ i < j ≤ h ] \mathbb{C}[m_{ij} \mid1 \leq i < j \leq h] C [ m ij ∣ 1 ≤ i < j ≤ h ] generated by all tree polynomials of any wheel subgraph of K h K_h K h .
Example 4.16. For h = 5 h = 5 h = 5 the affine slope variety S 5 ⊂ C 10 S_5 \subset\mathbb{C}^{10} S 5 ⊂ C 10 is of dimension 7, the ideal J 5 s J_5^s J 5 s has 5 generators, which correspond to the 5 complete graphs K 4 K_4 K 4 in K 5 K_5 K 5 . The ideal I 5 s I_5^s I 5 s in turn has 15 extra generators corresponding to the 15 = ( h − 1 ) ! 2 ( h 4 ) 15 = \frac{(h-1)!}{2}\binom{h}{4} 15 = 2 ( h − 1 )! ( 4 h ) wheels W 4 W_4 W 4 of size 4 contained in K 5 K_5 K 5 . However, the five generators of J 5 s J_5^s J 5 s minimally generate I 5 s I_5^s I 5 s .
The following is the main result [14 Theorem 1.1] and a remark from the end of cit. loc.
Theorem 4.17 ([14 ] ). The affine slope variety S h ⊂ A ( h 2 ) S_h \subset\mathbb{A}^{\binom{h}{2}} S h ⊂ A ( 2 h ) is irreducible of affine dimension 2 h − 3 2h - 3 2 h − 3 . Its prime vanishing ideal in C [ m i j ∣ i < j ≤ h ] \mathbb{C}[m_{ij} \mid i < j \leq h] C [ m ij ∣ i < j ≤ h ] is I h s I_h^s I h s . Moreover, for h ≤ 9 h \leq9 h ≤ 9 its vanishing ideal is equal to the ideal generated by the tree polynomials of all 3-wheels W 3 W_3 W 3 contained in K h K_h K h i.e.
I h s = J h s ⊂ C [ m i j ∣ 1 ≤ i < j ≤ h ] . I_h^s = J_h^s \subset\mathbb{C}[m_{ij} \mid1 \leq i < j \leq h]. I h s = J h s ⊂ C [ m ij ∣ 1 ≤ i < j ≤ h ] .
Remark 4.18. Note that the affine slope variety S h ⊂ C ( h 2 ) S_h \subset\mathbb{C}^{\binom{h}{2}} S h ⊂ C ( 2 h ) is a cone. More precisely, for t ∈ C × t \in\mathbb{C}^{\times} t ∈ C × , and m ∈ S h m \in S_h m ∈ S h , we have t m ∈ S h tm \in S_h t m ∈ S h as t ∙ ( x i , y i ) ↦ ( x i , t y i ) t \bullet(x_i,y_i) \mapsto(x_i,ty_i) t ∙ ( x i , y i ) ↦ ( x i , t y i ) scales all m i j m_{ij} m ij by t t t . Therefore, let P ( S h ) ⊂ P ( h 2 ) − 1 \mathbb{P}(S_h) \subset\mathbb{P}^{\binom{h}{2}-1} P ( S h ) ⊂ P ( 2 h ) − 1 be the projectivized slope variety , which is the projectivization of the affine cone i.e. ( S h ∖ { 0 } ) / C × (S_h \setminus\{0\})/\mathbb{C}^{\times} ( S h ∖ { 0 }) / C × . Equivalently the projectivization is simply given by taking the vanishing locus of I h s I_h^s I h s in P ( h 2 ) − 1 = P ( C ( h 2 ) ) \mathbb{P}^{\binom{h}{2}-1} = \mathbb{P}(\mathbb{C}^{\binom{h}{2}}) P ( 2 h ) − 1 = P ( C ( 2 h ) ) as all generators are homogeneous. Conversely, taking the affine cone of the projectivization again recovers the affine slope variety. We will use these facts freely throughout the text.
First let us connect affine slope varieties and Hadamard ratios.
Proposition 4.19. Let M 0 , n M_{0,n} M 0 , n be the moduli space of n n n -marked points on P 1 \mathbb{P}^{1} P 1 . Then, the Hadamard ratio M 0 , n + 1 : M 0 , n + 1 M_{0,n+1}:M_{0,n+1} M 0 , n + 1 : M 0 , n + 1 is the projectivized slope variety P ( S n ) \mathbb{P}(S_n) P ( S n ) .
Proof. We construct a homogenized parametrization of M 0 , n + 1 M_{0,n+1} M 0 , n + 1 as follows. Consider the 2 × ( n + 1 ) 2 \times(n+1) 2 × ( n + 1 ) matrix
X = [ 0 1 ⋯ 1 1 x 1 ⋯ x n ] , X = \begin{bmatrix} 0 & 1 & \cdots& 1 \\ 1 & x_1 & \cdots& x_n \end{bmatrix}, X = [ 0 1 1 x 1 ⋯ ⋯ 1 x n ] , in the indeterminates x i x_i x i with x j ≠ x i x_j \ne x_i x j = x i . This corresponds to a choice of n n n marked affine points on U 0 ⊂ P 1 U_0 \subset\mathbb{P}^{1} U 0 ⊂ P 1 together with the point at infinity [ 0 : 1 ] ∈ P 1 [0:1] \in\mathbb{P}^{1} [ 0 : 1 ] ∈ P 1 . We fix the remaining affine coordinate freedom by setting x 1 = 0 x_1=0 x 1 = 0 and x n = 1 x_n=1 x n = 1 . Taking the $2 \times 2$ minors of X X X then embeds M 0 , n + 1 M_{0,n+1} M 0 , n + 1 into P M \mathbb{P}^{M} P M , where M = ( n + 1 2 ) − 1 M=\binom{n+1}{2}-1 M = ( 2 n + 1 ) − 1 . Label the columns of X X X from 0 0 0 through n n n . For each $1 \le i < j \le n$, denote by p i j = x j − x i p_{ij}=x_j-x_i p ij = x j − x i the homogeneous 2 × 2 2 \times2 2 × 2 minors of X X X formed by columns i i i and j j j , and note that p 0 j = − 1 p_{0j}=-1 p 0 j = − 1 for all j > 0 j>0 j > 0 . Then, after projecting to the P N \mathbb{P}^{N} P N with N = ( n 2 ) − 1 N=\binom{n}{2}-1 N = ( 2 n ) − 1 and coordinates p i j p_{ij} p ij with 1 ≤ i < j ≤ n 1 \le i < j \le n 1 ≤ i < j ≤ n , the Hadamard ratio M 0 , n + 1 : M 0 , n + 1 M_{0,n+1}:M_{0,n+1} M 0 , n + 1 : M 0 , n + 1 is the closure of the image of the map
Ψ : M 0 , n + 1 × M 0 , n + 1 ⟶ P N , ( q i j ) × ( p i j ) ⟼ y j − y i x j − x i = m i j , 1 ≤ i < j ≤ n . \Psi:M_{0,n+1}\times M_{0,n+1}\longrightarrow\mathbb{P}^{N},\qquad(q_{ij})\times(p_{ij})\longmapsto\frac{y_j-y_i}{x_j-x_i}=m_{ij},\quad1\le i<j\le n. Ψ : M 0 , n + 1 × M 0 , n + 1 ⟶ P N , ( q ij ) × ( p ij ) ⟼ x j − x i y j − y i = m ij , 1 ≤ i < j ≤ n . After taking the affine cone over im Ψ \operatorname{im}\Psi im Ψ , we obtain a Zariski dense subset of the affine slope variety S n ⊂ C N + 1 S_n\subset\mathbb{C}^{N+1} S n ⊂ C N + 1 . Therefore, M 0 , n + 1 : M 0 , n + 1 = P ( S n ) M_{0,n+1}:M_{0,n+1}=\mathbb{P}(S_n) M 0 , n + 1 : M 0 , n + 1 = P ( S n ) , as the Hadamard ratio is defined by taking the Zariski closure of im Ψ \operatorname{im}\Psi im Ψ . This shows the claim.
Recall that the tropicalization of the moduli space M 0 , n M_{0,n} M 0 , n is precisely the space of all metric trees with n n n labeled leaves. Then, the following result addresses the relation between the affine slope variety and the BHV-phylogenetic tree space [2 ] , as first speculated in [14 ] .
Corollary 4.20. The tropicalization of the projectivized slope variety P ( S n ) \mathbb{P}(S_n) P ( S n ) is given as the Minkowski difference of the tropicalization of two copies of M 0 , n + 1 M_{0,n+1} M 0 , n + 1 . In symbols:
trop ( P ( S n ) ) = trop ( M 0 , n + 1 ) − trop ( M 0 , n + 1 ) , \operatorname{trop}(\mathbb{P}(S_n))=\operatorname{trop}(M_{0,n+1})-\operatorname{trop}(M_{0,n+1}), trop ( P ( S n )) = trop ( M 0 , n + 1 ) − trop ( M 0 , n + 1 ) , where M 0 , n + 1 M_{0,n+1} M 0 , n + 1 is embedded as in Proposition 4.19 .
Proof. This follows from combining Lemma 2.3 with Proposition 4.19 .
The remainder of this section is devoted to prove Theorem 4.7 . We start by introducing a series of technical results. To that end consider the ring isomorphism
ϕ : Z n ± 1 ⟶ A n − 1 ± 1 z i n ⟼ λ i i < n , z i j ⟼ λ i λ j m i j i < j < n . (11) \begin{aligned}
\phi:\mathbb{Z}_{n}^{\pm1}&\longrightarrow A_{n-1}^{\pm1}\\
z_{in}&\longmapsto\lambda_i \qquad i<n,\\
z_{ij}&\longmapsto\lambda_i\lambda_jm_{ij}\qquad i<j<n. \tag*{(11)}
\end{aligned} ϕ : Z n ± 1 z in z ij ⟶ A n − 1 ± 1 ⟼ λ i i < n , ⟼ λ i λ j m ij i < j < n . (11)
Then, for the pentads pen i j k l n \operatorname{pen}_{ijkl n} pen ij k l n , we have
ϕ ( pen i j k l n ) = − ( λ i λ j λ k λ l ) 2 det M i j k l . (12) \phi(\operatorname{pen}_{ijkl n})=-(\lambda_i\lambda_j\lambda_k\lambda_l)^2\det M_{ijkl}. \tag*{(12)} ϕ ( pen ij k l n ) = − ( λ i λ j λ k λ l ) 2 det M ij k l . (12)
Lemma 4.21. Let n ≥ 5 n \ge5 n ≥ 5 . The image of p e n a b c d e \mathrm{pen}_{abcde} pen ab c d e with { a , b , c , d , e } ⊂ [ n − 1 ] \{a,b,c,d,e\} \subset[n-1] { a , b , c , d , e } ⊂ [ n − 1 ] under ϕ \phi ϕ is contained in the following ideal
⟨ ϕ ( p e n i j k l n ) ∣ { i , j , k , l } ⊂ { a , b , c , d , e } ⟩ \left\langle\phi(\mathrm{pen}_{ijkl n}) \mid\{i,j,k,l\} \subset\{a,b,c,d,e\} \right\rangle ⟨ ϕ ( pen ij k l n ) ∣ { i , j , k , l } ⊂ { a , b , c , d , e } ⟩ in C [ λ i ± 1 ∣ i ≤ n − 1 ] ⊗ C [ m i j ± 1 ∣ i < j ≤ n − 1 ] \mathbb{C}[\lambda_i^{\pm1} \mid i \le n-1] \otimes\mathbb{C}[m_{ij}^{\pm1} \mid i<j \le n-1] C [ λ i ± 1 ∣ i ≤ n − 1 ] ⊗ C [ m ij ± 1 ∣ i < j ≤ n − 1 ] .
Proof. It suffices to note we can write
ϕ ( p e n a b c d e ) = 1 2 ( λ a λ b λ c λ d λ e ) 2 ( A a b c d det M a b c d + A a b c e det M a b c e + A a b d e det M a b d e + A a c d e det M a c d e + A b c d e det M b c d e ) . \phi(\mathrm{pen}_{abcde}) = \frac{1}{2}(\lambda_a\lambda_b\lambda_c\lambda_d\lambda_e)^2\left(A_{abcd}\det M_{abcd}+A_{abce}\det M_{abce}
+A_{abde}\det M_{abde}+A_{acde}\det M_{acde}+A_{bcde}\det M_{bcde}\right). ϕ ( pen ab c d e ) = 2 1 ( λ a λ b λ c λ d λ e ) 2 ( A ab c d det M ab c d + A ab ce det M ab ce + A ab d e det M ab d e + A a c d e det M a c d e + A b c d e det M b c d e ) .
with
A a b c d = − 2 m a e m b e + m a b ( m a e + m b e − m c e − m d e ) , A a b c e = 2 m a d m b d + m a b ( − m a d − m b d + m c d + m d e ) , A a b d e = − 2 m a c m b c + m a b ( m a c + m b c − m c d − m c e ) , A a c d e = m a b ( m b c + m b d + m b e − m a b ) , A b c d e = m a b ( m a b − m a c − m a d − m a e ) . \begin{aligned}
A_{abcd} &= -2m_{ae}m_{be}+m_{ab}(m_{ae}+m_{be}-m_{ce}-m_{de}),\\
A_{abce} &= 2m_{ad}m_{bd}+m_{ab}(-m_{ad}-m_{bd}+m_{cd}+m_{de}),\\
A_{abde} &= -2m_{ac}m_{bc}+m_{ab}(m_{ac}+m_{bc}-m_{cd}-m_{ce}),\\
A_{acde} &= m_{ab}(m_{bc}+m_{bd}+m_{be}-m_{ab}),\\
A_{bcde} &= m_{ab}(m_{ab}-m_{ac}-m_{ad}-m_{ae}).
\end{aligned} A ab c d A ab ce A ab d e A a c d e A b c d e = − 2 m a e m b e + m ab ( m a e + m b e − m ce − m d e ) , = 2 m a d m b d + m ab ( − m a d − m b d + m c d + m d e ) , = − 2 m a c m b c + m ab ( m a c + m b c − m c d − m ce ) , = m ab ( m b c + m b d + m b e − m ab ) , = m ab ( m ab − m a c − m a d − m a e ) .
Then the claim follows as ( λ a λ b λ c λ d λ e ) 2 (\lambda_a\lambda_b\lambda_c\lambda_d\lambda_e)^2 ( λ a λ b λ c λ d λ e ) 2 is a unit in C [ λ i ± 1 ∣ i < n ] \mathbb{C}[\lambda_i^{\pm1}\mid i<n] C [ λ i ± 1 ∣ i < n ] .
This has an immediate consequence.
Lemma 4.22. Let J n p J_n^p J n p be the ideal generated by all pentads on [ n ] [n] [ n ] . Then, the image of the localization J n p ⋅ C [ z i j ± 1 ∣ i < j ≤ n ] J_n^p\cdot\mathbb{C}[z_{ij}^{\pm1}\mid i<j\le n] J n p ⋅ C [ z ij ± 1 ∣ i < j ≤ n ] under ϕ \phi ϕ , is the localization of the ideal J n − 1 s J_{n-1}^s J n − 1 s generated by all tree polynomials of 3-wheels inside K n − 1 K_{n-1} K n − 1 , i.e.
ϕ ( J n p ⋅ Z n ± ) = J n − 1 s ⋅ A n − 1 ± . \phi(J_n^p\cdot Z_n^\pm)=J_{n-1}^s\cdot A_{n-1}^\pm. ϕ ( J n p ⋅ Z n ± ) = J n − 1 s ⋅ A n − 1 ± . The same claim holds for the radicals of the ideals, i.e.
ϕ ( J n p ⋅ Z n ± ) = J n − 1 s ⋅ A n − 1 ± . \phi\left(\sqrt{J_n^p}\cdot Z_n^\pm\right)=\sqrt{J_{n-1}^s}\cdot A_{n-1}^\pm. ϕ ( J n p ⋅ Z n ± ) = J n − 1 s ⋅ A n − 1 ± . Proof. This follows from combining the description of the images of the pentad equations in (12) with Lemma 4.21. If f ∈ J n p f\in\sqrt{J_n^p} f ∈ J n p then f m ∈ J n p f^m\in J_n^p f m ∈ J n p for a suitable m ∈ Z ≥ 1 m\in\mathbb{Z}_{\ge1} m ∈ Z ≥ 1 , then the radical statement follows analogously.
Lemma 4.23. Let x i j = m 1 i − m i j x_{ij}=m_{1i}-m_{ij} x ij = m 1 i − m ij and identify m i j = m j i m_{ij}=m_{ji} m ij = m j i for i > j i>j i > j . Let W W W be the wheel W ( v 1 ; v 2 … , v h ) W(v_1;v_2\ldots,v_h) W ( v 1 ; v 2 … , v h ) . For each pair ( v a , v b ) ∉ E ( W ) (v_a,v_b)\notin E(W) ( v a , v b ) ∈ / E ( W ) , there exist two wheels U a b U_{ab} U ab and V a b V_{ab} V ab such that their vertex sets are proper subsets of { v 1 , v 2 , … , v h } \{v_1,v_2,\ldots,v_h\} { v 1 , v 2 , … , v h } , and
x v a v b τ ( W ) , x v b v a τ ( W ) ∈ ⟨ τ ( U v a v b ) , τ ( V v a v b ) ⟩ . (13) x_{v_av_b}\tau(W),x_{v_bv_a}\tau(W)\in\langle\tau(U_{v_av_b}),\tau(V_{v_av_b})\rangle. \tag*{(13)} x v a v b τ ( W ) , x v b v a τ ( W ) ∈ ⟨ τ ( U v a v b ) , τ ( V v a v b )⟩ . (13) In particular, if h ≥ 7 h\ge7 h ≥ 7 , then
τ ( W ) h ∈ ∑ ( v a , v b ) ∉ E ( W ) ⟨ τ ( U v a , v b ) , τ ( V v a , v b ) ⟩ . \tau(W)^h\in\sum_{(v_a,v_b)\notin E(W)}\langle\tau(U_{v_a,v_b}),\tau(V_{v_a,v_b})\rangle. τ ( W ) h ∈ ( v a , v b ) ∈ / E ( W ) ∑ ⟨ τ ( U v a , v b ) , τ ( V v a , v b )⟩ .
Proof. We start by splitting the h − 1 h-1 h − 1 wheel W ( v 1 , v 2 , … , v h ) W(v_1,v_2,\ldots,v_h) W ( v 1 , v 2 , … , v h ) as follows. Let U = W ( v 1 ; v 2 , … , v r ) U=W(v_1;v_2,\ldots,v_r) U = W ( v 1 ; v 2 , … , v r ) and V = W ( v 1 ; v r , … , v h , v 2 ) V=W(v_1;v_r,\ldots,v_h,v_2) V = W ( v 1 ; v r , … , v h , v 2 ) with 4 ≤ r ≤ h − 1 4\le r\le h-1 4 ≤ r ≤ h − 1 (see Figure 1 ).
Figure 1. Splitting of the h − 1 h-1 h − 1 wheel W ( v 1 ; v 2 , … , v h ) W(v_1;v_2,\ldots,v_h) W ( v 1 ; v 2 , … , v h ) .
First we show that
x v 2 v r τ ( W ) , x v r v 2 τ ( W ) ∈ ⟨ τ ( U ) , τ ( V ) ⟩ . x_{v_2v_r}\tau(W),x_{v_rv_2}\tau(W)\in\langle\tau(U),\tau(V)\rangle. x v 2 v r τ ( W ) , x v r v 2 τ ( W ) ∈ ⟨ τ ( U ) , τ ( V )⟩ .
It suffices to consider the case v i = i v_i=i v i = i . Then, by definition,
τ ( W ) = ∏ i = 2 h ( m 1 i − m i ( i + 1 ) ) − ∏ i = 2 h ( m 1 ( i + 1 ) − m i ( i + 1 ) ) = ∏ i = 2 h x i ( i + 1 ) − ∏ i = 2 h x ( i + 1 ) i = A B − A ′ B ′ , \tau(W)=\prod_{i=2}^{h}(m_{1i}-m_{i(i+1)})-\prod_{i=2}^{h}(m_{1(i+1)}-m_{i(i+1)})=\prod_{i=2}^{h}x_{i(i+1)}-\prod_{i=2}^{h}x_{(i+1)i}=AB-A'B', τ ( W ) = i = 2 ∏ h ( m 1 i − m i ( i + 1 ) ) − i = 2 ∏ h ( m 1 ( i + 1 ) − m i ( i + 1 ) ) = i = 2 ∏ h x i ( i + 1 ) − i = 2 ∏ h x ( i + 1 ) i = A B − A ′ B ′ ,
where A = ∏ i = 2 r − 1 x i ( i + 1 ) A=\prod_{i=2}^{r-1}x_{i(i+1)} A = ∏ i = 2 r − 1 x i ( i + 1 ) , B = ∏ i = r h x i ( i + 1 ) B=\prod_{i=r}^{h}x_{i(i+1)} B = ∏ i = r h x i ( i + 1 ) , A ′ = ∏ i = 2 r − 1 x ( i + 1 ) i A'=\prod_{i=2}^{r-1}x_{(i+1)i} A ′ = ∏ i = 2 r − 1 x ( i + 1 ) i , and B ′ = ∏ i = r h x ( i + 1 ) i B'=\prod_{i=r}^{h}x_{(i+1)i} B ′ = ∏ i = r h x ( i + 1 ) i . Moreover,
τ ( U ) = ∏ i = 2 r x i ( i + 1 ) − ∏ i = 2 r x ( i + 1 ) i = A x r 2 − A ′ x 2 r , \tau(U)=\prod_{i=2}^{r}x_{i(i+1)}-\prod_{i=2}^{r}x_{(i+1)i}=Ax_{r2}-A'x_{2r}, τ ( U ) = i = 2 ∏ r x i ( i + 1 ) − i = 2 ∏ r x ( i + 1 ) i = A x r 2 − A ′ x 2 r ,
τ ( V ) = ( ∏ i = r h x i ( i + 1 ) ) x 2 r − ( ∏ i = r h x ( i + 1 ) i ) x r 2 = B x 2 r − B ′ x r 2 . \tau(V)=\left(\prod_{i=r}^{h}x_{i(i+1)}\right)x_{2r}-\left(\prod_{i=r}^{h}x_{(i+1)i}\right)x_{r2}=Bx_{2r}-B'x_{r2}. τ ( V ) = ( i = r ∏ h x i ( i + 1 ) ) x 2 r − ( i = r ∏ h x ( i + 1 ) i ) x r 2 = B x 2 r − B ′ x r 2 .
It follows that
A τ ( V ) + B ′ τ ( U ) = A B x 2 r − A B ′ x r 2 + A B ′ x r 2 − A ′ B ′ x 2 r = ( A B − A ′ B ′ ) x 2 r = x 2 r τ ( W ) . \begin{aligned}
A\tau(V)+B'\tau(U)=ABx_{2r}-AB'x_{r2}+AB'x_{r2}-A'B'x_{2r} \\
&=(AB-A'B')x_{2r}=x_{2r}\tau(W).
\end{aligned} A τ ( V ) + B ′ τ ( U ) = A B x 2 r − A B ′ x r 2 + A B ′ x r 2 − A ′ B ′ x 2 r = ( A B − A ′ B ′ ) x 2 r = x 2 r τ ( W ) .
Similarly,
A ′ τ ( V ) + B τ ( U ) = A ′ B x 2 r − A ′ B ′ x r 2 + A B x r 2 − A ′ B x 2 r = ( A B − A ′ B ′ ) x r 2 = x r 2 τ ( W ) . \begin{aligned}
A'\tau(V)+B\tau(U)=A'Bx_{2r}-A'B'x_{r2}+ABx_{r2}-A'Bx_{2r} \\
&=(AB-A'B')x_{r2}=x_{r2}\tau(W).
\end{aligned} A ′ τ ( V ) + B τ ( U ) = A ′ B x 2 r − A ′ B ′ x r 2 + A B x r 2 − A ′ B x 2 r = ( A B − A ′ B ′ ) x r 2 = x r 2 τ ( W ) .
Moreover, if v a v_a v a and v b v_b v b are non-adjacent in W W W with a < b a<b a < b , then neither of them can be the center of the wheel since the center is adjacent to every other vertex. Then
W ( v 1 ; v 2 , … , v h ) = W ( v 1 ; v a , … , v b , … , v h , v 2 , … , v a − 1 ) = W ( v 1 ; v 2 ′ , … , v h ′ ) , W(v_1;v_2,\ldots,v_h)=W(v_1;v_a,\ldots,v_b,\ldots,v_h,v_2,\ldots,v_{a-1})=W(v_1;v'_2,\ldots,v'_h), W ( v 1 ; v 2 , … , v h ) = W ( v 1 ; v a , … , v b , … , v h , v 2 , … , v a − 1 ) = W ( v 1 ; v 2 ′ , … , v h ′ ) ,
where v i ′ = v i + a − 2 v'_i=v_{i+a-2} v i ′ = v i + a − 2 .
Now assume W = W ( 1 ; 2 , … , h ) W=W(1;2,\ldots,h) W = W ( 1 ; 2 , … , h ) is such that h ≥ 7 h\ge7 h ≥ 7 . The containment relations in (13) imply that
m 1 a τ ( W ) − m 1 b τ ( W ) = x a b τ ( W ) − x b a τ ( W ) ∈ ⟨ τ ( U a b ) , τ ( V a b ) ⟩ , m_{1a}\tau(W)-m_{1b}\tau(W)=x_{ab}\tau(W)-x_{ba}\tau(W)\in\langle\tau(U_{ab}),\tau(V_{ab})\rangle, m 1 a τ ( W ) − m 1 b τ ( W ) = x ab τ ( W ) − x ba τ ( W ) ∈ ⟨ τ ( U ab ) , τ ( V ab )⟩ ,
for each pair of non-adjacent vertices a a a and b b b of W W W . In particular,
( m 12 − m 1 i ) τ ( W ) ∈ ⟨ τ ( U 2 i ) , τ ( V 2 i ) ⟩ for all 4 ≤ i ≤ h − 1 , ( m 13 − m 1 j ) τ ( W ) ∈ ⟨ τ ( U 3 j ) , τ ( V 3 j ) ⟩ for all 5 ≤ j ≤ h , ( m 14 − m 1 k ) τ ( W ) ∈ ⟨ τ ( U 4 k ) , τ ( V 4 k ) ⟩ for all k = 6 , … , h , 2 , ( m 15 − m 1 l ) τ ( W ) ∈ ⟨ τ ( U 5 l ) , τ ( V 5 l ) ⟩ for all l = 7 , … , h , 2 , 3 , ⋮ ( m 1 h − m 1 s ) τ ( W ) ∈ ⟨ τ ( U s h ) , τ ( V s h ) ⟩ for all 3 ≤ s ≤ h − 2. \begin{aligned}
(m_{12}-m_{1i})\tau(W)&\in\langle\tau(U_{2i}),\tau(V_{2i})\rangle&&\text{for all }4\le i\le h-1,\\
(m_{13}-m_{1j})\tau(W)&\in\langle\tau(U_{3j}),\tau(V_{3j})\rangle&&\text{for all }5\le j\le h,\\
(m_{14}-m_{1k})\tau(W)&\in\langle\tau(U_{4k}),\tau(V_{4k})\rangle&&\text{for all }k=6,\ldots,h,2,\\
(m_{15}-m_{1l})\tau(W)&\in\langle\tau(U_{5l}),\tau(V_{5l})\rangle&&\text{for all }l=7,\ldots,h,2,3,\\
&\vdots\\
(m_{1h}-m_{1s})\tau(W)&\in\langle\tau(U_{sh}),\tau(V_{sh})\rangle&&\text{for all }3\le s\le h-2.
\end{aligned} ( m 12 − m 1 i ) τ ( W ) ( m 13 − m 1 j ) τ ( W ) ( m 14 − m 1 k ) τ ( W ) ( m 15 − m 1 l ) τ ( W ) ( m 1 h − m 1 s ) τ ( W ) ∈ ⟨ τ ( U 2 i ) , τ ( V 2 i )⟩ ∈ ⟨ τ ( U 3 j ) , τ ( V 3 j )⟩ ∈ ⟨ τ ( U 4 k ) , τ ( V 4 k )⟩ ∈ ⟨ τ ( U 5 l ) , τ ( V 5 l )⟩ ⋮ ∈ ⟨ τ ( U s h ) , τ ( V s h )⟩ for all 4 ≤ i ≤ h − 1 , for all 5 ≤ j ≤ h , for all k = 6 , … , h , 2 , for all l = 7 , … , h , 2 , 3 , for all 3 ≤ s ≤ h − 2.
It follows that m 1 i τ ( W ) = m 1 j τ ( W ) m_{1i}\tau(W)=m_{1j}\tau(W) m 1 i τ ( W ) = m 1 j τ ( W ) in C [ m i j ∣ 1 ≤ i < j ≤ h ] / H \mathbb{C}[m_{ij}\mid1\le i<j\le h]/H C [ m ij ∣ 1 ≤ i < j ≤ h ] / H for all 2 ≤ i < j ≤ h 2\le i<j\le h 2 ≤ i < j ≤ h , where
H = ∑ ( v a , v b ) ∉ E ( W ) ⟨ τ ( U a b ) , τ ( V a b ) ⟩ ⊆ C [ m i j ∣ 1 ≤ i < j ≤ h ] . H=\sum_{(v_a,v_b)\notin E(W)}\langle\tau(U_{ab}),\tau(V_{ab})\rangle\subseteq\mathbb{C}[m_{ij}\mid1\le i<j\le h]. H = ( v a , v b ) ∈ / E ( W ) ∑ ⟨ τ ( U ab ) , τ ( V ab )⟩ ⊆ C [ m ij ∣ 1 ≤ i < j ≤ h ] .
Hence
τ ( W ) h = τ ( W ) h − 1 ( ∏ i = 2 h ( m 1 i − m i ( i + 1 ) ) − ∏ i = 2 h ( m 1 ( i + 1 ) − m i ( i + 1 ) ) ) = ∏ i = 2 h ( m 1 i τ ( W ) − m i ( i + 1 ) τ ( W ) ) − ∏ i = 2 h ( m 1 ( i + 1 ) τ ( W ) − m i ( i + 1 ) τ ( W ) ) = ∏ i = 2 h ( m 12 τ ( W ) − m i ( i + 1 ) τ ( W ) ) − ∏ i = 2 h ( m 12 τ ( W ) − m i ( i + 1 ) τ ( W ) ) = 0 \begin{aligned}
\tau(W)^h
&=\tau(W)^{h-1}\left(\prod_{i=2}^{h}(m_{1i}-m_{i(i+1)})-\prod_{i=2}^{h}(m_{1(i+1)}-m_{i(i+1)})\right)\\
&=\prod_{i=2}^{h}(m_{1i}\tau(W)-m_{i(i+1)}\tau(W))-\prod_{i=2}^{h}(m_{1(i+1)}\tau(W)-m_{i(i+1)}\tau(W))\\
&=\prod_{i=2}^{h}(m_{12}\tau(W)-m_{i(i+1)}\tau(W))-\prod_{i=2}^{h}(m_{12}\tau(W)-m_{i(i+1)}\tau(W))=0
\end{aligned} τ ( W ) h = τ ( W ) h − 1 ( i = 2 ∏ h ( m 1 i − m i ( i + 1 ) ) − i = 2 ∏ h ( m 1 ( i + 1 ) − m i ( i + 1 ) ) ) = i = 2 ∏ h ( m 1 i τ ( W ) − m i ( i + 1 ) τ ( W )) − i = 2 ∏ h ( m 1 ( i + 1 ) τ ( W ) − m i ( i + 1 ) τ ( W )) = i = 2 ∏ h ( m 12 τ ( W ) − m i ( i + 1 ) τ ( W )) − i = 2 ∏ h ( m 12 τ ( W ) − m i ( i + 1 ) τ ( W )) = 0
in C [ m i j ∣ 1 ≤ i < j ≤ h ] / H \mathbb{C}[m_{ij}\mid1\le i<j\le h]/H C [ m ij ∣ 1 ≤ i < j ≤ h ] / H . That is, τ ( W ) h ∈ H \tau(W)^h\in H τ ( W ) h ∈ H . Up to relabeling and cyclic rotation of the indices, it follows that, for every wheel W = W ( v 1 ; v 2 , … , v h ) W=W(v_1;v_2,\ldots,v_h) W = W ( v 1 ; v 2 , … , v h ) with h ≥ 7 h\ge7 h ≥ 7 ,
τ ( W ) h ∈ ∑ ( v a , v b ) ∉ E ( W ) ⟨ τ ( U v a v b ) , τ ( V v a v b ) ⟩ . \tau(W)^h\in\sum_{(v_a,v_b)\notin E(W)}\langle\tau(U_{v_av_b}),\tau(V_{v_av_b})\rangle. τ ( W ) h ∈ ( v a , v b ) ∈ / E ( W ) ∑ ⟨ τ ( U v a v b ) , τ ( V v a v b )⟩ .
For a homogeneous ideal I I I in a standard graded polynomial ring, we write I ≤ d I_{\le d} I ≤ d for the ideal generated by all homogeneous polynomials in I I I whose degrees are less than or equal to d d d . Then Lemma 4.23 immediately implies the following.
Corollary 4.24. Let h ≥ 7 h\ge7 h ≥ 7 . Let τ ( W ) \tau(W) τ ( W ) be a tree polynomial of an i i i -wheel W W W , which is a subgraph of K h K_h K h , then τ ( W ) i + 1 ∈ ( I h s ) ≤ i − 1 \tau(W)^{i+1}\in(I_h^s)_{\le i-1} τ ( W ) i + 1 ∈ ( I h s ) ≤ i − 1 for all $6\le i\le h-1.
Proof. By Lemma 4.23 ,
τ ( W ) i + 1 ∈ ∑ ( v a , v b ) ∉ E ( W ) ⟨ τ ( U v a v b ) , τ ( V v a v b ) ⟩ \tau(W)^{i+1}\in\sum_{(v_a,v_b)\notin E(W)}\langle\tau(U_{v_av_b}),\tau(V_{v_av_b})\rangle τ ( W ) i + 1 ∈ ( v a , v b ) ∈ / E ( W ) ∑ ⟨ τ ( U v a v b ) , τ ( V v a v b )⟩
⊆ ⟨ tree polynomials of j -wheel subgraphs of K h ∣ j ≤ i − 1 ⟩ = ( I h s ) ≤ i − 1 . \subseteq\langle\text{tree polynomials of }j\text{-wheel subgraphs of }K_{h}\mid j\le i-1\rangle=(I_{h}^{s})_{\le i-1}. ⊆ ⟨ tree polynomials of j -wheel subgraphs of K h ∣ j ≤ i − 1 ⟩ = ( I h s ) ≤ i − 1 .
The following lemma was proved in Enkosky’s PhD thesis [6 Theorem 4.1.2] . We provide here an alternative and simpler proof.
Lemma 4.25. We have J h s = I h s \sqrt{J_h^s}=I_h^s J h s = I h s for any h ≥ 4 h\ge4 h ≥ 4 .
Proof. It is clear that J h s ⊆ I h s J_h^s\subseteq I_h^s J h s ⊆ I h s for any h ≥ 4 h\ge4 h ≥ 4 , and equality holds for h ≤ 9 h\le9 h ≤ 9 by Theorem 4.17 . Since I h s I_h^s I h s is a prime ideal, it follows that J h s ⊆ I h s = I h s \sqrt{J_h^s}\subseteq\sqrt{I_h^s}=I_h^s J h s ⊆ I h s = I h s for any h ≥ 4. I t r e m a i n s t o p r o v e t h e r e v e r s e i n c l u s i o n f o r h\ge4. It remains to prove the reverse inclusion for h ≥ 4. I t r e main s t o p r o v e t h er e v er se in c l u s i o n f or h\ge 10.
Fix h ≥ 10 h \ge10 h ≥ 10 . We prove by induction on i i i that ( I h s ) ≤ i ⊆ J h s (I_h^s)_{\le i} \subseteq\sqrt{J_h^s} ( I h s ) ≤ i ⊆ J h s for all 3 ≤ i ≤ h − 1 3 \le i \le h-1 3 ≤ i ≤ h − 1 . Since ( I h s ) ≤ h − 1 = I h s (I_h^s)_{\le h-1}=I_h^s ( I h s ) ≤ h − 1 = I h s , this will complete the proof. For 3 ≤ i ≤ 8 3 \le i \le8 3 ≤ i ≤ 8 , the claim follows by applying Theorem 4.17 again, which gives ( I h s ) ≤ i = J h s (I_h^s)_{\le i}=J_h^s ( I h s ) ≤ i = J h s . Now let 9 ≤ i ≤ h − 1 9 \le i \le h-1 9 ≤ i ≤ h − 1 and for the purpose of induction assume that ( I h s ) ≤ i − 1 ⊆ J h s (I_h^s)_{\le i-1} \subseteq\sqrt{J_h^s} ( I h s ) ≤ i − 1 ⊆ J h s . Let W W W be an i i i -wheel subgraph of K h K_h K h . By Corollary 4.24, we have τ ( W ) i + 1 ∈ ( I h s ) ≤ i − 1 ⊆ J h s \tau(W)^{i+1} \in(I_h^s)_{\le i-1} \subseteq\sqrt{J_h^s} τ ( W ) i + 1 ∈ ( I h s ) ≤ i − 1 ⊆ J h s . The inductive hypothesis therefore implies that τ ( W ) i + 1 ∈ J h s \tau(W)^{i+1} \in\sqrt{J_h^s} τ ( W ) i + 1 ∈ J h s , which in turn implies τ ( W ) ∈ J h s \tau(W) \in\sqrt{J_h^s} τ ( W ) ∈ J h s . Hence every degree-i i i generator of I h s I_h^s I h s belongs to J h s \sqrt{J_h^s} J h s , and consequently ( I h s ) ≤ i ⊆ J h s (I_h^s)_{\le i} \subseteq\sqrt{J_h^s} ( I h s ) ≤ i ⊆ J h s .
We can finally prove the main result in this section.
Proof of Theorem 4.7 . The dimensional claim is Proposition 4.8 . Next, we show that I n p \sqrt{I_n^p} I n p is prime. By Theorem 4.17 , I n − 1 s I_{n-1}^s I n − 1 s is a prime ideal of C [ m i j ∣ 1 ≤ i < j ≤ n − 1 ] \mathbb{C}[m_{ij} \mid1 \le i < j \le n-1] C [ m ij ∣ 1 ≤ i < j ≤ n − 1 ] . By Lemma 4.25, it follows that J n − 1 s \sqrt{J_{n-1}^s} J n − 1 s is also a prime ideal of C [ m i j ∣ 1 ≤ i < j ≤ n − 1 ] \mathbb{C}[m_{ij} \mid1 \le i < j \le n-1] C [ m ij ∣ 1 ≤ i < j ≤ n − 1 ] . We can extend this ideal to A n − 1 A_{n-1} A n − 1 and obtain the prime ideal J n − 1 s ⋅ A n − 1 \sqrt{J_{n-1}^s} \cdot A_{n-1} J n − 1 s ⋅ A n − 1 . Now, we check that the monomial u : = ∏ 1 ≤ i < j ≤ n − 1 m i j ∏ k = 1 n − 1 λ k u := \prod_{1 \le i < j \le n-1} m_{ij} \prod_{k=1}^{n-1} \lambda_k u := ∏ 1 ≤ i < j ≤ n − 1 m ij ∏ k = 1 n − 1 λ k does not belong to J n − 1 s ⋅ A n − 1 \sqrt{J_{n-1}^s} \cdot A_{n-1} J n − 1 s ⋅ A n − 1 . Since J n − 1 s ⋅ A n − 1 \sqrt{J_{n-1}^s} \cdot A_{n-1} J n − 1 s ⋅ A n − 1 is prime, if u ∈ J n − 1 s ⋅ A n − 1 u \in\sqrt{J_{n-1}^s} \cdot A_{n-1} u ∈ J n − 1 s ⋅ A n − 1 , then at least one of the variables of A n − 1 A_{n-1} A n − 1 must belong to J n − 1 s ⋅ A n − 1 \sqrt{J_{n-1}^s} \cdot A_{n-1} J n − 1 s ⋅ A n − 1 . However, by definition J n − 1 s = I n − 1 s \sqrt{J_{n-1}^s}=I_{n-1}^s J n − 1 s = I n − 1 s is generated by homogeneous polynomials of degree at least 3. Consequently, its extension J n − 1 s ⋅ A n − 1 \sqrt{J_{n-1}^s}\cdot A_{n-1} J n − 1 s ⋅ A n − 1 is also generated by homogeneous polynomials of degree at least 3. In particular, it contains no nonzero homogeneous elements of degree 1. This yields a contradiction. Since u u u is not in the prime ideal J n − 1 s ⋅ A n − 1 \sqrt{J_{n-1}^s}\cdot A_{n-1} J n − 1 s ⋅ A n − 1 , localizing at u u u preserves the primality of J n − 1 s ⋅ A n − 1 \sqrt{J_{n-1}^s}\cdot A_{n-1} J n − 1 s ⋅ A n − 1 . Hence J n − 1 s ⋅ A n − 1 ± \sqrt{J_{n-1}^s}\cdot A_{n-1}^{\pm} J n − 1 s ⋅ A n − 1 ± is prime. Since ϕ \phi ϕ is an isomorphism we have that
J n p ⋅ Z n ± = ϕ − 1 ( J n − 1 s ⋅ A n − 1 ± ) \sqrt{J_n^p}\cdot Z_n^{\pm}=\phi^{-1}\left(\sqrt{J_{n-1}^s}\cdot A_{n-1}^{\pm}\right) J n p ⋅ Z n ± = ϕ − 1 ( J n − 1 s ⋅ A n − 1 ± )
is a prime ideal of Z n ± Z_n^{\pm} Z n ± . Moreover, Remark 4.13 shows that the primality of J n p ⋅ Z n ± \sqrt{J_n^p}\cdot Z_n^{\pm} J n p ⋅ Z n ± forces I n p \sqrt{I_n^p} I n p to be prime. Thus, we obtain the following chain of implications establishing primality:
I n − 1 s is prime ⟹ J n − 1 s is prime ⟹ J n p ⋅ Z n ± is prime ⟹ I n p is prime . I_{n-1}^s\text{ is prime}\Longrightarrow\sqrt{J_{n-1}^s}\text{ is prime}\Longrightarrow\sqrt{J_n^p}\cdot Z_n^{\pm}\text{ is prime}\Longrightarrow\sqrt{I_n^p}\text{ is prime}. I n − 1 s is prime ⟹ J n − 1 s is prime ⟹ J n p ⋅ Z n ± is prime ⟹ I n p is prime .
In particular, if n ≤ 10 n \le10 n ≤ 10 , then the same chain of implications shows that I n p I_n^p I n p is prime:
I n − 1 s is prime ⟹ J n − 1 s is prime ⟹ J n p ⋅ Z n ± is prime ⟹ I n p is prime . I_{n-1}^s\text{ is prime}\Longrightarrow J_{n-1}^s\text{ is prime}\Longrightarrow J_n^p\cdot Z_n^{\pm}\text{ is prime}\Longrightarrow I_n^p\text{ is prime}. I n − 1 s is prime ⟹ J n − 1 s is prime ⟹ J n p ⋅ Z n ± is prime ⟹ I n p is prime .
This concludes the proof.
We end this section by recording our main structural result relating the affine slope variety with the pentad variety, i.e. Corollary 1.4 . We restate this result here as follows.
Proposition 4.26. Let Δ \Delta Δ be the second hypersimplex in R n \mathbb{R}^n R n , i.e. the convex hull of all vectors e i + e j ∈ R n e_i+e_j \in\mathbb{R}^n e i + e j ∈ R n for i < j ≤ n i<j\le n i < j ≤ n , where e i e_i e i are the standard basis vectors of R n \mathbb{R}^n R n . Let X Δ ⊂ P N X_\Delta\subset\mathbb{P}^N X Δ ⊂ P N with N = ( n 2 ) − 1 N=\binom{n}{2}-1 N = ( 2 n ) − 1 denote the embedded toric variety associated to Δ \Delta Δ . Then, the Hadamard product P ( S n ) ∗ X Δ \mathbb{P}(S_n)*X_\Delta P ( S n ) ∗ X Δ equals to the pentad variety Pen n \operatorname{Pen}_n Pen n as subvarieties of P N \mathbb{P}^N P N .
Proof. First recall that X Δ ⊂ P N X_\Delta\subset\mathbb{P}^N X Δ ⊂ P N is defined as the Zariski closure of the monomial map
( C × ) n → P N , λ ↦ ( λ i λ j ) i < j . (\mathbb{C}^{\times})^n \to\mathbb{P}^N,\qquad\lambda\mapsto(\lambda_i\lambda_j)_{i<j}. ( C × ) n → P N , λ ↦ ( λ i λ j ) i < j .
Next, recall that the Hadamard product P ( S n ) ∗ X Δ \mathbb{P}(S_n) * X_{\Delta} P ( S n ) ∗ X Δ is the closure of the image of the following map
P ( S n ) × X Δ ⇢ P N , ( m i j , λ i λ j ) ↦ ( λ i λ j m i j ) = ( λ i λ j y j − y i x j − x i ) . \mathbb{P}(S_n) \times X_{\Delta} \dashrightarrow\mathbb{P}^{N},\qquad(m_{ij},\lambda_i\lambda_j) \mapsto(\lambda_i\lambda_jm_{ij}) = \left(\lambda_i\lambda_j\frac{y_j-y_i}{x_j-x_i}\right). P ( S n ) × X Δ ⇢ P N , ( m ij , λ i λ j ) ↦ ( λ i λ j m ij ) = ( λ i λ j x j − x i y j − y i ) .
Lastly, recall the parametrization in the proof of Proposition 4.8 via
( C × ) n × C n × C n ⇢ Pen n ⊂ P N , ( λ , x , y ) ↦ ( λ i λ j y j − y i x j − x i ) . (\mathbb{C}^{\times})^n \times\mathbb{C}^n \times\mathbb{C}^n \dashrightarrow\operatorname{Pen}_n \subset\mathbb{P}^{N},\qquad(\lambda,x,y) \mapsto\left(\lambda_i\lambda_j\frac{y_j-y_i}{x_j-x_i}\right). ( C × ) n × C n × C n ⇢ Pen n ⊂ P N , ( λ , x , y ) ↦ ( λ i λ j x j − x i y j − y i ) .
Hence, the closures of the images of the two parametrizations, which define the varieties Pen n \operatorname{Pen}_n Pen n and P ( S n ) ∗ X Δ \mathbb{P}(S_n) * X_{\Delta} P ( S n ) ∗ X Δ , agree. This shows the claim.
Affine Slopes and Phirotopes
In this section, we want to point out another surprising connection of the pentad and the affine slope variety with the combinatorial structure of phirotopes [3 ] , which are a generalization of chirotopes. We record here the more general definition.
Definition 4.27. Let E = { 1 , … , n } E = \{1,\ldots,n\} E = { 1 , … , n } be a finite index set. A function
φ : E d ⟶ S 1 ∪ { 0 } \varphi:E^d \longrightarrow S^1 \cup\{0\} φ : E d ⟶ S 1 ∪ { 0 } on all tuples of E E E is called a rank-d d d phirotope if:
(a) The function φ \varphi φ is alternating.
(b) For any two subsets τ = { τ 1 , … , τ d − 1 } \tau= \{\tau_1,\ldots,\tau_{d-1}\} τ = { τ 1 , … , τ d − 1 } and μ = { μ 1 , … , μ d + 1 } \mu= \{\mu_1,\ldots,\mu_{d+1}\} μ = { μ 1 , … , μ d + 1 } of E E E , define
ω k = ( − 1 ) k − 1 φ ( τ , μ k ) φ ( μ 1 , … , μ k ^ , … , μ d + 1 ) , \omega_k = (-1)^{k-1}\varphi(\tau,\mu_k)\varphi(\mu_1,\ldots,\widehat{\mu_k},\ldots,\mu_{d+1}), ω k = ( − 1 ) k − 1 φ ( τ , μ k ) φ ( μ 1 , … , μ k , … , μ d + 1 ) , for k = 1 , … , d + 1 k = 1,\ldots,d+1 k = 1 , … , d + 1 . Then there exist r 1 , … , r d + 1 ∈ R > 0 r_1,\ldots,r_{d+1} \in\mathbb{R}_{>0} r 1 , … , r d + 1 ∈ R > 0 such that
∑ k = 1 d + 1 r k ω k = 0. \sum_{k=1}^{d+1} r_k\omega_k = 0. k = 1 ∑ d + 1 r k ω k = 0. We call φ \varphi φ uniform if its image on subsets of E d E^d E d without repeated elements, is contained in S 1 S^1 S 1 . We say φ \varphi φ is realizable if there exists a configuration of points Z = ( Z 1 , … , Z n ) ∈ C d × n Z = (Z_1,\ldots,Z_n) \in\mathbb{C}^{d \times n} Z = ( Z 1 , … , Z n ) ∈ C d × n such that
φ ( i 1 , … , i d ) = ω ( det ( Z i 1 , … , Z i d ) ) , \varphi(i_1,\ldots,i_d) = \omega(\det(Z_{i_1},\ldots,Z_{i_d})), φ ( i 1 , … , i d ) = ω ( det ( Z i 1 , … , Z i d )) , where ω : C ⟶ S 1 ∪ { 0 } \omega:\mathbb{C} \longrightarrow S^1 \cup\{0\} ω : C ⟶ S 1 ∪ { 0 } is the phase map with
ω ( r e i x ) = { 0 r = 0 , e i x r ≠ 0. \omega(re^{ix}) =
\begin{cases}
0 & r = 0,\\
e^{ix} & r \ne0.
\end{cases} ω ( r e i x ) = { 0 e i x r = 0 , r = 0.
Example 4.28. Let us consider an explicit example of a rank-2 phirotope on 4 elements, with E = { 1 , 2 , 3 , 4 } E = \{1,2,3,4\} E = { 1 , 2 , 3 , 4 } . We define the map φ : E 2 ⟶ S 1 ∪ { 0 } \varphi:E^2 \longrightarrow S^1 \cup\{0\} φ : E 2 ⟶ S 1 ∪ { 0 } by setting
φ ( 1 , 2 ) = 1 , φ ( 1 , 3 ) = i , φ ( 1 , 4 ) = 2 + i 5 , \varphi(1,2) = 1,\qquad\varphi(1,3) = i,\qquad\varphi(1,4) = \frac{2+i}{\sqrt{5}}, φ ( 1 , 2 ) = 1 , φ ( 1 , 3 ) = i , φ ( 1 , 4 ) = 5 2 + i , φ ( 2 , 3 ) = − 1 + i 2 , φ ( 2 , 4 ) = 1 + i 2 , φ ( 3 , 4 ) = 1 , \varphi(2,3) = \frac{-1+i}{\sqrt{2}},\qquad\varphi(2,4) = \frac{1+i}{\sqrt{2}},\qquad\varphi(3,4) = 1, φ ( 2 , 3 ) = 2 − 1 + i , φ ( 2 , 4 ) = 2 1 + i , φ ( 3 , 4 ) = 1 ,
and set φ ( i , j ) = − φ ( j , i ) \varphi(i,j)=-\varphi(j,i) φ ( i , j ) = − φ ( j , i ) for i < j i<j i < j , as well as φ ( i , i ) = 0 \varphi(i,i)=0 φ ( i , i ) = 0 . Then, in particular φ \varphi φ is alternating, so it satisfies condition (a) from Definition 4.27 . Next, we verify condition (b). For the four indices 1 , 2 , 3 , 4 1,2,3,4 1 , 2 , 3 , 4 , set
ω 1 = φ ( 1 , 2 ) φ ( 3 , 4 ) = 1 , \omega_1=\varphi(1,2)\varphi(3,4)=1, ω 1 = φ ( 1 , 2 ) φ ( 3 , 4 ) = 1 ,
ω 2 = − φ ( 1 , 3 ) φ ( 2 , 4 ) = 1 − i 2 , \omega_2=-\varphi(1,3)\varphi(2,4)=\frac{1-i}{\sqrt{2}}, ω 2 = − φ ( 1 , 3 ) φ ( 2 , 4 ) = 2 1 − i ,
ω 3 = φ ( 1 , 4 ) φ ( 2 , 3 ) = − 3 + i 10 . \omega_3=\varphi(1,4)\varphi(2,3)=\frac{-3+i}{\sqrt{10}}. ω 3 = φ ( 1 , 4 ) φ ( 2 , 3 ) = 10 − 3 + i .
Therefore, we can obtain positive solutions with r 1 = 2 r_1=2 r 1 = 2 , r 2 = 2 r_2=\sqrt{2} r 2 = 2 , and r 3 = 10 r_3=\sqrt{10} r 3 = 10 , such that
0 = r 1 ω 1 + r 2 ω 2 + r 3 ω 3 = 2 + ( 1 − i ) + ( − 3 + i ) . 0=r_1\omega_1+r_2\omega_2+r_3\omega_3=2+(1-i)+(-3+i). 0 = r 1 ω 1 + r 2 ω 2 + r 3 ω 3 = 2 + ( 1 − i ) + ( − 3 + i ) .
Hence, φ \varphi φ indeed defines a phirotope. Moreover, by construction φ \varphi φ is uniform as im φ ⊂ S 1 \operatorname{im}\varphi\subset S^1 im φ ⊂ S 1 .
We now show that φ \varphi φ is realizable. Consider the following four affine complex points
z 1 = 0 , z 2 = 1 , z 3 = i , z 4 = 2 + i z_1=0,\qquad z_2=1,\qquad z_3=i,\qquad z_4=2+i z 1 = 0 , z 2 = 1 , z 3 = i , z 4 = 2 + i
in the complex plane C ≃ R 2 \mathbb{C}\simeq\mathbb{R}^2 C ≃ R 2 . Using homogeneous coordinates, these correspond to the vectors
Z 1 = ( 1 0 ) , Z 2 = ( 1 1 ) , Z 3 = ( 1 i ) , Z 4 = ( 1 2 + i ) ∈ C 2 . Z_1=\begin{pmatrix}1\\0\end{pmatrix},\qquad
Z_2=\begin{pmatrix}1\\1\end{pmatrix},\qquad
Z_3=\begin{pmatrix}1\\i\end{pmatrix},\qquad
Z_4=\begin{pmatrix}1\\2+i\end{pmatrix}\in\mathbb{C}^2. Z 1 = ( 1 0 ) , Z 2 = ( 1 1 ) , Z 3 = ( 1 i ) , Z 4 = ( 1 2 + i ) ∈ C 2 .
Then, indeed, the phases ω \omega ω of the determinants of Z i Z h Z_iZ_h Z i Z h are given as φ ( i , j ) \varphi(i,j) φ ( i , j ) , hence φ \varphi φ is a uniform realizable phirotope of rank 2 2 2 .
Remark 4.29. The original reason for trying connect phirotopes with the pentad variety, was the observation made by Bernd Sturmfels and Jürgen Richter-Gebert that the following realizability condition appearing in [3 Theorem 5.2]
∑ σ ∈ Sym ( 5 ) sgn ( σ ) φ ( σ ( 1 , 2 ) ) 2 φ ( σ ( 2 , 3 ) ) 2 φ ( σ ( 3 , 4 ) ) 2 φ ( σ ( 4 , 5 ) ) 2 φ ( σ ( 5 , 1 ) ) 2 = 0 , \sum_{\sigma\in\operatorname{Sym}(5)}\operatorname{sgn}(\sigma)\varphi(\sigma(1,2))^2\varphi(\sigma(2,3))^2\varphi(\sigma(3,4))^2\varphi(\sigma(4,5))^2\varphi(\sigma(5,1))^2=0, σ ∈ Sym ( 5 ) ∑ sgn ( σ ) φ ( σ ( 1 , 2 ) ) 2 φ ( σ ( 2 , 3 ) ) 2 φ ( σ ( 3 , 4 ) ) 2 φ ( σ ( 4 , 5 ) ) 2 φ ( σ ( 5 , 1 ) ) 2 = 0 ,
precisely recovers the pentad constraint from (5) after substituting z i j = ( φ ( i , j ) ) 2 z_{ij}=(\varphi(i,j))^2 z ij = ( φ ( i , j ) ) 2 .
Our main result in this paragraph makes the remark above rigorous. More specifically, it relates the realizable phirotopes of rank 2 2 2 on n + 1 n+1 n + 1 points to the affine slope variety S n \mathcal{S}_n S n .
Proposition 4.30. For a uniform rank 2 2 2 phirotope φ \varphi φ with index set E = { 0 , 1 , … , n } E=\{0,1,\ldots,n\} E = { 0 , 1 , … , n } , denote by φ ^ ( i , j ) : = φ ( i , j ) / ( φ ( 0 , i ) φ ( 0 , j ) ) \widehat{\varphi}(i,j):=\varphi(i,j)/(\varphi(0,i)\varphi(0,j)) φ ( i , j ) := φ ( i , j ) / ( φ ( 0 , i ) φ ( 0 , j )) with 1 ≤ i < j ≤ n 1\leq i<j\leq n 1 ≤ i < j ≤ n its normalized phases. Let φ ^ 2 : = ( φ ^ ( 1 , 2 ) 2 , … , φ ^ ( n − 1 , n ) 2 ) ∈ C ( n 2 ) \widehat{\varphi}^2:=(\widehat{\varphi}(1,2)^2,\ldots,\widehat{\varphi}(n-1,n)^2)\in\mathbb{C}^{\binom{n}{2}} φ 2 := ( φ ( 1 , 2 ) 2 , … , φ ( n − 1 , n ) 2 ) ∈ C ( 2 n ) be the vector of these coordinates. Define R n ⊂ C ( n 2 ) \mathcal{R}_n\subset\mathbb{C}^{\binom{n}{2}} R n ⊂ C ( 2 n ) as the set of all such points obtained from uniform realizable phirotopes φ \varphi φ on E E E , that is
R n : = { φ ^ 2 ∣ φ uniform realizable phirotope on E } ⊂ C ( n 2 ) . \mathcal{R}_n:=\{\widehat{\varphi}^2\mid\varphi\text{ uniform realizable phirotope on }E\}\subset\mathbb{C}^{\binom{n}{2}}. R n := { φ 2 ∣ φ uniform realizable phirotope on E } ⊂ C ( 2 n ) . Then, the complex Zariski closure of R n \mathcal{R}_n R n coincides with the affine slope variety S n \mathcal{S}_n S n .
Proof. First note that by [3 Lemma 2.3] any realizable phirotope of rank 2 on n + 1 n+1 n + 1 elements can be represented by a point configuration ( Z 0 , … , Z n ) (Z_0,\ldots,Z_n) ( Z 0 , … , Z n ) in P 1 \mathbb{P}^1 P 1 of φ \varphi φ . By loc. cit. we can also assume that Z 0 = ( 0 , 1 ) T Z_0=(0,1)^{\mathsf{T}} Z 0 = ( 0 , 1 ) T , and Z i = b i ( 1 , a i ) T Z_i=b_i(1,a_i)^{\mathsf{T}} Z i = b i ( 1 , a i ) T for i > 0 i>0 i > 0 , with b i ∈ C × b_i\in\mathbb{C}^{\times} b i ∈ C × and pairwise distinct a i ∈ C a_i\in\mathbb{C} a i ∈ C . Thus, for the phases we get φ ( 0 , i ) = − ω ( b i ) \varphi(0,i)=-\omega(b_i) φ ( 0 , i ) = − ω ( b i ) and φ ( i , j ) = ω ( b i ) ω ( b j ) ω ( a j − a i ) \varphi(i,j)=\omega(b_i)\omega(b_j)\omega(a_j-a_i) φ ( i , j ) = ω ( b i ) ω ( b j ) ω ( a j − a i ) . With a i = X i + − 1 Y i a_i=X_i+\sqrt{-1}Y_i a i = X i + − 1 Y i , where X i , Y i ∈ R X_i,Y_i\in\mathbb{R} X i , Y i ∈ R , the normalized squared phases satisfy
φ ^ ( i , j ) 2 = a j − a i a j − a i ‾ = ( X j − X i ) + − 1 ( Y j − Y i ) ( X j − X i ) − − 1 ( Y j − Y i ) . \widehat{\varphi}(i,j)^2=\frac{a_j-a_i}{\overline{a_j-a_i}}=\frac{(X_j-X_i)+\sqrt{-1}(Y_j-Y_i)}{(X_j-X_i)-\sqrt{-1}(Y_j-Y_i)}. φ ( i , j ) 2 = a j − a i a j − a i = ( X j − X i ) − − 1 ( Y j − Y i ) ( X j − X i ) + − 1 ( Y j − Y i ) .
where a j − a i ‾ \overline{a_j-a_i} a j − a i denotes the complex conjugate. This is well-defined, since the a i a_i a i are pairwise distinct. Then, after a complex change of parameters setting x i : = X i − − 1 Y i x_i:=X_i-\sqrt{-1}Y_i x i := X i − − 1 Y i and y i : = X i + − 1 Y i y_i:=X_i+\sqrt{-1}Y_i y i := X i + − 1 Y i , we get
φ ^ ( i , j ) 2 = y j − y i x j − x i . \widehat{\varphi}(i,j)^2=\frac{y_j-y_i}{x_j-x_i}. φ ( i , j ) 2 = x j − x i y j − y i .
Therefore, each point φ ^ 2 ∈ R n \widehat{\varphi}^{\,2}\in\mathcal{R}_n φ 2 ∈ R n corresponds to a point in the affine slope variety S n \mathcal{S}_n S n . Note that we have x i = y i ‾ x_i=\overline{y_i} x i = y i . It suffices to show that R n \mathcal{R}_n R n is Zariski dense in S n \mathcal{S}_n S n . To that end, let
{ x i = y i ‾ for all i } ⊂ C 2 n , and U : = { x i ≠ x j for all i ≠ j } ⊂ C 2 n . \{x_i=\overline{y_i}\text{ for all }i\}\subset\mathbb{C}^{2n},\qquad\text{and}\qquad U:=\{x_i\ne x_j\text{ for all }i\ne j\}\subset\mathbb{C}^{2n}. { x i = y i for all i } ⊂ C 2 n , and U := { x i = x j for all i = j } ⊂ C 2 n .
Since { x i = y i ‾ for all i } ≅ R 2 n \{x_i=\overline{y_i}\text{ for all }i\}\cong\mathbb{R}^{2n} { x i = y i for all i } ≅ R 2 n is complex Zariski dense in C 2 n \mathbb{C}^{2n} C 2 n , the intersection with U U U is Zariski dense in U U U . The morphism
U ⟶ C ( n 2 ) , ( x i , y i ) ⟼ ( y j − y i x j − x i ) i < j U\longrightarrow\mathbb{C}^{\binom{n}{2}},\qquad(x_i,y_i)\longmapsto\left(\frac{y_j-y_i}{x_j-x_i}\right)_{i<j} U ⟶ C ( 2 n ) , ( x i , y i ) ⟼ ( x j − x i y j − y i ) i < j
has image with Zariski closure S n \mathcal{S}_n S n , and its restriction to { x i = y i ‾ for all i } ∩ U \{x_i=\overline{y_i}\text{ for all }i\}\cap U { x i = y i for all i } ∩ U has image R n \mathcal{R}_n R n . Hence, R n \mathcal{R}_n R n is Zariski dense in S n \mathcal{S}_n S n . This shows the claim. □
In fact we can see that the affine slope variety completely characterizes the realizable uniform phirotopes.
Corollary 4.31. Let φ \varphi φ be a uniform rank 2 phirotope on E = { 0 , 1 , … , n } E=\{0,1,\ldots,n\} E = { 0 , 1 , … , n } . The following are equivalent:
(i) φ \varphi φ is realizable.
(ii) φ ^ 2 ∈ S n \widehat{\varphi}^{\,2}\in\mathcal{S}_n φ 2 ∈ S n .
(iii) τ i j k l ( φ ^ 2 ) = 0 \tau_{ijkl}(\widehat{\varphi}^{\,2})=0 τ ij k l ( φ 2 ) = 0 for every 1 ≤ i < j < k < l ≤ n 1\le i<j<k<l\le n 1 ≤ i < j < k < l ≤ n .
Proof. The implication (i) ⇒ \Rightarrow ⇒ (ii) follows from Proposition 4.30 . Every tree polynomial (10) vanishes on S n \mathcal{S}_n S n by Theorem 4.17 , so (ii) ⇒ \Rightarrow ⇒ (iii). Now assume (iii). Let φ ‾ i j : = φ ( i , j ) φ ( i , j ) ‾ \overline{\varphi}_{ij}:=\varphi(i,j)\overline{\varphi(i,j)} φ ij := φ ( i , j ) φ ( i , j ) be the non-normalized phases, and let φ ‾ : = ( φ ‾ i j ) 0 ≤ i < j ≤ n ∈ C N \overline{\varphi}:=(\overline{\varphi}_{ij})_{0\le i<j\le n}\in\mathbb{C}^{N} φ := ( φ ij ) 0 ≤ i < j ≤ n ∈ C N with N = ( n + 1 2 ) N=\binom{n+1}{2} N = ( 2 n + 1 ) . Note that φ ‾ i j ≠ 0 \overline{\varphi}_{ij}\ne0 φ ij = 0 for i ≠ j i\ne j i = j by uniformity of φ \varphi φ . Applying (11) , with its distinguished index n + 1 n+1 n + 1 replaced with 0 0 0 , gives
pen i j k l 0 ( φ ‾ ) = − ( φ ‾ 0 i φ ‾ 0 j φ ‾ 0 k φ ‾ 0 l ) 2 τ i j k l ( φ ^ 2 ) = 0. \operatorname{pen}_{ijkl0}(\overline{\varphi})=-(\overline{\varphi}_{0i}\overline{\varphi}_{0j}\overline{\varphi}_{0k}\overline{\varphi}_{0l})^2\tau_{ijkl}(\widehat{\varphi}^{\,2})=0. pen ij k l 0 ( φ ) = − ( φ 0 i φ 0 j φ 0 k φ 0 l ) 2 τ ij k l ( φ 2 ) = 0.
Consequently, every pentad whose index sets contains 0 vanishes at φ ‾ \overline{\varphi} φ . By Lemma 4.21 all remaining pentads also vanish at φ ‾ \overline{\varphi} φ . By the five-point realizability criterion of [3 Lemma 3.4 and Theorem 3.6] , the vanishing of all these pentads implies that φ \varphi φ is realizable. Thus (iii) ⇒ \Rightarrow ⇒ (i). This finishes the proof. □