Introduction
The manifolds constructed in [6 ] have the rational cohomology ring of S 2 × S 2 S^2 \times S^2 S 2 × S 2 . We call them Akhmedov’s fake quadrics; no complex structure is assumed. They are symplectic sums of ruled surfaces along cyclic multisections. We give handlebody decompositions of their complements, compute the integral handle chain complexes, identify the H 2 H_2 H 2 -torsion geometrically, and specify marked surface gluing maps. For the symmetric ( 3 , 3 ) (3,3) ( 3 , 3 ) sum with twist parameter m = 1 m = 1 m = 1 , we also give a simultaneous surface diagram of the relative product attachments and determine their surface framings.
The earlier construction in [5 ] used knot surgery and twisted fiber sums to obtain the integral cohomology ring of S 2 × S 2 S^2 \times S^2 S 2 × S 2 . The cyclic construction in [6 ] instead uses multisection complements. It determines their groups and boundary incompressibility, classifies the degree pairs with Euler characteristic four and signature zero, and gives a rational-homology gluing criterion. The resulting infinite families are distinguished by the orders of their finite first homology groups. For related non-isotopic symplectic surfaces, see Smith [12 ] and Hays and Park [13 ] .
In joint work with Sai-Kee Yeung [8 ] , we computed integral Mayer-Vietoris presentations and first-homology formulas for the three product families. Here we realize the matrices of [8 Theorem 4.5] by marked surface diffeomorphisms. Their based mapping classes are part of the gluing data: the matrices alone do not determine the smooth sums.
§ 2 gives the complement handles and homology; § 3 and § 4 give the markings and gluings. The simultaneous surface-framed relative attachments are in Theorem B.1 . In the follow-up paper with Selman Akbulut and Eylem Zeliha Yildiz [9 ] , we will study full handlebody diagrams of the closed sums, including the complete attaching links and their integer framing coefficients.
Removing a regular sphere fiber as well gives the Stein fillings of [7 ] . For fixed g ≥ 2 g \ge2 g ≥ 2 and p ≥ 4 p \ge4 p ≥ 4 , these include infinitely many pairwise nonhomeomorphic aspherical fillings of one contact three-manifold, with the same integral homology and zero intersection pairings. Their groups are distinguished by the first homology of finite covers. For related boundary gluings and handlebody constructions, see [1 Chapter 3 and Section 4.3] and [2 , 3 , 4 ] .
Throughout, Σ g \Sigma_g Σ g is a closed oriented surface of genus g ≥ 2 g \ge2 g ≥ 2 , and p ≥ 2 p \ge2 p ≥ 2 is the multisection degree. Identify the oriented equator of S 2 S^2 S 2 with S 1 = { z ∈ C : ∣ z ∣ = 1 } S^1 = \{z \in\mathbb{C}: |z| = 1\} S 1 = { z ∈ C : ∣ z ∣ = 1 } . Write S = Σ g × { p t } S = \Sigma_g \times\{pt\} S = Σ g × { pt } and choose a smooth map u : Σ g → S 1 u:\Sigma_g\to S^1 u : Σ g → S 1 for which u ∗ : π 1 ( Σ g ) → Z u_*:\pi_1(\Sigma_g)\to\mathbb{Z} u ∗ : π 1 ( Σ g ) → Z is surjective. Set
F g , p = { ( x , z ) ∈ Σ g × S 1 : z p = u ( x ) } ⊂ Σ g × S 2 , M g , p = ( Σ g × S 2 ) ∖ int ν F g , p . F_{g,p}=\{(x,z)\in\Sigma_g\times S^1:z^p=u(x)\}\subset\Sigma_g\times S^2,\qquad M_{g,p}=(\Sigma_g\times S^2)\setminus\operatorname{int}\nu F_{g,p}. F g , p = {( x , z ) ∈ Σ g × S 1 : z p = u ( x )} ⊂ Σ g × S 2 , M g , p = ( Σ g × S 2 ) ∖ int ν F g , p .
Here ν F g , p \nu F_{g,p} ν F g , p is a closed tubular neighborhood. The first projection F g , p → Σ g F_{g,p}\to\Sigma_g F g , p → Σ g is a connected, unbranched p p p -fold cover; connectedness follows from the surjectivity of u ∗ u_* u ∗ modulo p p p . The product symplectic form restricts to the pullback of the base area form, since the sphere coordinate lies in the equator. The tubular disks are chosen fiberwise and equivariantly. The multisection has class p [ S ] p[S] p [ S ] , square zero, and genus G = 1 + p ( g − 1 ) G=1+p(g-1) G = 1 + p ( g − 1 ) . The complement is a bundle with fiber P p P_p P p , a sphere with p p p disks removed. Figure 1 gives the g = 2 g=2 g = 2 , p = 3 p=3 p = 3 model as a sum of two connected degree-three braided tori in T 2 × S 2 T^2\times S^2 T 2 × S 2 .
Figure 1. The genus-four braided multisection in Σ 2 × S 2 \Sigma_2 \times S^2 Σ 2 × S 2 . Each braided torus meets the summing sphere { p t } × S 2 \{\mathrm{pt}\} \times S^2 { pt } × S 2 in three points. The punctured tori are joined by three annuli, giving χ ( F ) = − 3 − 3 = − 6 \chi(F)=-3-3=-6 χ ( F ) = − 3 − 3 = − 6 and g ( F ) = 4 g(F)=4 g ( F ) = 4 . At the seam, j L j_L j L is joined to j R j_R j R , with matching colors. The large silhouettes show the surface pieces and their annular joins; the insets record 1 → 2 → 3 → 1 1\to2\to3\to1 1 → 2 → 3 → 1 around a base loop. Strand colors follow moving points, while numbers label fixed fiber positions.
Here u u u represents the primitive class ( 1 , 0 , 1 , 0 ) (1,0,1,0) ( 1 , 0 , 1 , 0 ) . An orientation-preserving change of base marking gives the fixed convention used in § 2 . Take the two defining maps constant near the summing points, with the same value. Choose product tubular neighborhoods S 2 × D 2 S^2\times D^2 S 2 × D 2 of the summing spheres, with identical three-point configurations and deleted disks. On their boundaries the gluing is
Ψ ( v , e i θ ) = ( v , e − i θ ) , v ∈ S 2 . \Psi(v,e^{i\theta})=(v,e^{-i\theta}),\qquad v\in S^2. Ψ ( v , e i θ ) = ( v , e − i θ ) , v ∈ S 2 .
Thus j L j_L j L joins j R j_R j R for j = 1 , 2 , 3 j=1,2,3 j = 1 , 2 , 3 through a product annulus. The cyclic braid occurs around a base loop in each torus, with no additional braid at the sphere-sum seam.
The product cases of [6 ] are listed below; G G G is the common gluing genus.
Table 1.
We study the ( 3 , 3 ) (3,3) ( 3 , 3 ) and ( 2 , 4 ) (2,4) ( 2 , 4 ) families in product ruled surfaces. Theorem 2.3 gives
H 2 ( M g , p ; Z ) ≅ Z 1 + 2 ( g − 1 ) ( p − 1 ) ⊕ Z / p . H_2(M_{g,p};\mathbb{Z})\cong\mathbb{Z}^{1+2(g-1)(p-1)}\oplus\mathbb{Z}/p. H 2 ( M g , p ; Z ) ≅ Z 1 + 2 ( g − 1 ) ( p − 1 ) ⊕ Z / p .
and Corollary 2.4 identifies the torsion generator. The closed first-homology formulas in the following theorem recover those of [8 Theorem 4.5] .
Theorem 1.1. The complements M 2 , 3 M_{2,3} M 2 , 3 , M 3 , 2 M_{3,2} M 3 , 2 , and M 2 , 4 M_{2,4} M 2 , 4 admit the handlebody decompositions described in § 2 , with respective handle counts ( h 0 , h 1 , h 2 , h 3 ) (h_0,h_1,h_2,h_3) ( h 0 , h 1 , h 2 , h 3 ) given by
( 2 , 11 , 14 , 3 ) , ( 2 , 14 , 14 , 2 ) , ( 2 , 12 , 18 , 4 ) . (2,11,14,3),\qquad(2,14,14,2),\qquad(2,12,18,4). ( 2 , 11 , 14 , 3 ) , ( 2 , 14 , 14 , 2 ) , ( 2 , 12 , 18 , 4 ) . For each integer m ≥ 1 m\geq1 m ≥ 1 , let Z m 3 , 3 Z_m^{3,3} Z m 3 , 3 and Z m 2 , 4 Z_m^{2,4} Z m 2 , 4 be the symplectic sums formed using the marked surface diffeomorphisms f m f_m f m of § 3 and § 4 . In the chosen boundary trivializations, the gluing is
ϕ m ( w , e i θ ) = ( f m ( w ) , e − i θ ) , w ∈ Σ G , \phi_m(w,e^{i\theta})=(f_m(w),e^{-i\theta}),\qquad w\in\Sigma_G, ϕ m ( w , e i θ ) = ( f m ( w ) , e − i θ ) , w ∈ Σ G , with no surface-dependent normal-circle twist. Then
H 1 ( Z m 3 , 3 ; Z ) ≅ ( Z / 3 ) 2 ⊕ Z / ( 3 m ) , H_1(Z_m^{3,3};\mathbb{Z})\cong(\mathbb{Z}/3)^2\oplus\mathbb{Z}/(3m), H 1 ( Z m 3 , 3 ; Z ) ≅ ( Z /3 ) 2 ⊕ Z / ( 3 m ) , H 1 ( Z m 2 , 4 ; Z ) ≅ ( Z / 2 ) 2 ⊕ Z / ( 4 m ) . H_1(Z_m^{2,4};\mathbb{Z})\cong(\mathbb{Z}/2)^2\oplus\mathbb{Z}/(4m). H 1 ( Z m 2 , 4 ; Z ) ≅ ( Z /2 ) 2 ⊕ Z / ( 4 m ) . Both symplectic sums have Euler characteristic four, signature zero, and the rational cohomology ring of S 2 × S 2 S^2\times S^2 S 2 × S 2 . They are aspherical and symplectically minimal.
Acknowledgments. The author is very grateful to B. Doug Park for several valuable discussions in 2007–2010 concerning one of the fake quadrics studied here, and for his insights and encouragement. The author also thanks Selman Akbulut, Eylem Zeliha Yildiz, Sai-Kee Yeung, Çağrı Karakurt, and Sümeyra Sakallı for helpful discussions. An LLM-based tool assisted with some algebraic computations involving the homology presentations and gluing matrices, preparation of the computational scripts and interval-arithmetic verification described in Appendix § 9 , and figure preparation. All mathematical ideas and results are due to the author, who takes full responsibility for the content.
Computational materials. The scripts, input data, surface-routing and cap-tree data, and interval certificates used in this paper are available in the accompanying GitHub repository. The file COMPUTATIONAL_README.txt gives reproduction commands, dependencies, software versions, and the scope of each check. The homology computations are also proved in the text; the accompanying files provide the finite data used in the computer-assisted diagram verifications.
Handlebody decompositions of the complements
Products of paths are traversed from left to right. Write R θ R_\theta R θ for rotation about the polar axis through angle θ \theta θ , and put r p = R 2 π / p r_p=R_{2\pi/p} r p = R 2 π / p . Choose based generators a i , b i a_i,b_i a i , b i of π 1 ( Σ g ) \pi_1(\Sigma_g) π 1 ( Σ g ) with relation ∏ i = 1 g [ a i , b i ] = 1 \prod_{i=1}^{g}[a_i,b_i]=1 ∏ i = 1 g [ a i , b i ] = 1 , where [ a , b ] = a b a − 1 b − 1 [a,b]=aba^{-1}b^{-1} [ a , b ] = ab a − 1 b − 1 . We use the same letters for the corresponding symplectic homology basis, with a i ⋅ b j = δ i j a_i\cdot b_j=\delta_{ij} a i ⋅ b j = δ ij . Use a marking with u ∗ ( a 1 ) = − 1 u_*(a_1)=-1 u ∗ ( a 1 ) = − 1 , u ∗ ( b 1 ) = 0 u_*(b_1)=0 u ∗ ( b 1 ) = 0 , and u ∗ ( a i ) = u ∗ ( b i ) = 0 u_*(a_i)=u_*(b_i)=0 u ∗ ( a i ) = u ∗ ( b i ) = 0 for 2 ≤ i ≤ g 2\le i\le g 2 ≤ i ≤ g , and write χ p = − u ∗ m o d p : π 1 ( Σ g ) → Z / p \chi_p=-u_*\bmod p:\pi_1(\Sigma_g)\to\mathbb{Z}/p χ p = − u ∗ mod p : π 1 ( Σ g ) → Z / p . Such a marking exists since u ∗ u_* u ∗ is surjective. Our face-pairing convention uses r p r_p r p for a 1 a_1 a 1 and the identity for the other generators. Positive geometric transport is the inverse of this quotient action. Conjugation by the displayed a 1 a_1 a 1 acts by r p r_p r p on the fiber group. Positive passage through the corresponding 1-handle has the inverse angular shift. For a fiber loop x x x based at the fixed north pole, the transport square has boundary x a 1 ( r p − 1 x ) a 1 − 1 xa_1(r_p^{-1}x)a_1^{-1} x a 1 ( r p − 1 x ) a 1 − 1 , giving a 1 x a 1 − 1 = r p ( x ) a_1xa_1^{-1}=r_p(x) a 1 x a 1 − 1 = r p ( x ) . This accounts for the inverse in the passage convention. Let ρ : Σ ~ g → Σ g \rho:\widetilde{\Sigma}_g\to\Sigma_g ρ : Σ g → Σ g be the universal covering; γ ∈ π 1 ( Σ g ) \gamma\in\pi_1(\Sigma_g) γ ∈ π 1 ( Σ g ) acts on it by deck transformations. With this convention, the bundle is
M g , p ≅ ( Σ ~ g × P p ) / π 1 ( Σ g ) , γ ( x , v ) = ( γ x , r p χ p ( γ ) v ) . M_{g,p}\cong(\widetilde{\Sigma}_g\times P_p)/\pi_1(\Sigma_g),\qquad\gamma(x,v)=(\gamma x,r_p^{\chi_p(\gamma)}v). M g , p ≅ ( Σ g × P p ) / π 1 ( Σ g ) , γ ( x , v ) = ( γ x , r p χ p ( γ ) v ) .
Equivalently, glue the faces of a base polygon times P p P_p P p by these transition maps.
For the quotient identification, choose a real lift U U U of the argument of u u u on Σ ~ g \widetilde{\Sigma}_g Σ g . It satisfies U ( γ x ) = U ( x ) + 2 π u ∗ ( γ ) U(\gamma x)=U(x)+2\pi u_*(\gamma) U ( γ x ) = U ( x ) + 2 π u ∗ ( γ ) . Rotation of the fiber by U ( x ) / p U(x)/p U ( x ) / p carries the reference deleted disks to the disks centered at the roots of z p = u ( x ) z^p=u(x) z p = u ( x ) . The map ( x , v ) ↦ ( ρ ( x ) , R U ( x ) / p v ) (x,v)\mapsto(\rho(x),R_{U(x)/p}v) ( x , v ) ↦ ( ρ ( x ) , R U ( x ) / p v ) is invariant under ( x , v ) ↦ ( γ x , r p − u ∗ ( γ ) v ) (x,v)\mapsto(\gamma x,r_p^{-u_*(\gamma)}v) ( x , v ) ↦ ( γ x , r p − u ∗ ( γ ) v ) , hence descends to the stated diffeomorphism of complement bundles, carrying their pole sections and boundary collars with it.
The 1-handles are represented by paired attaching 3-balls. Product coordinates and their prescribed identifications determine the 2-handle framings. We do not convert these drawings to dotted-circle notation with integer coefficients relative to Seifert framings.
The fiber and product handle decompositions
We use I = [ − 1 , 1 ] I=[-1,1] I = [ − 1 , 1 ] ; subscripts distinguish the coordinate factors. Retain a small disk about each pole of the sphere, denoted N N N and S S S , and join them by p p p disjoint bands c 0 , … , c p − 1 c_0,\ldots,c_{p-1} c 0 , … , c p − 1 between the equatorial deleted disks. These form two surface 0-handles and p p p surface 1-handles. Rotation fixes the two disks setwise and sends c j c_j c j to c j + 1 c_{j+1} c j + 1 cyclically. The rotation preserves the orientation of each band.
Use c j c_j c j also for the oriented core path of the j j j th band, from its center in N N N to its center in S S S . Read fiber indices modulo p p p . Define
x j = c j − 1 c j − 1 ( 1 ≤ j ≤ p ) , x 1 ⋯ x p = 1 , r p ( x j ) = x j + 1 . x_j=c_{j-1}c_j^{-1}\quad(1\le j\le p),\qquad x_1\cdots x_p=1,\qquad r_p(x_j)=x_{j+1}. x j = c j − 1 c j − 1 ( 1 ≤ j ≤ p ) , x 1 ⋯ x p = 1 , r p ( x j ) = x j + 1 .
These are peripheral loops with the indicated base paths at N N N . After choosing c 0 c_0 c 0 as a maximal tree edge, put y j = c j c 0 − 1 y_j=c_jc_0^{-1} y j = c j c 0 − 1 , 1 ≤ j < p 1\le j<p 1 ≤ j < p . The free fiber group has the basis y 1 , … , y p − 1 y_1,\ldots,y_{p-1} y 1 , … , y p − 1 , and
r p ( y j ) = y j + 1 y 1 − 1 ( j < p − 1 ) , r p ( y p − 1 ) = y 1 − 1 . r_p(y_j)=y_{j+1}y_1^{-1}\quad(j<p-1),\qquad r_p(y_{p-1})=y_1^{-1}. r p ( y j ) = y j + 1 y 1 − 1 ( j < p − 1 ) , r p ( y p − 1 ) = y 1 − 1 .
The two-disk decomposition retains the cyclic symmetry of the bands, although r p r_p r p does not permute the based free generators.
For a base band k ∈ { a i , b i } k\in\{a_i,b_i\} k ∈ { a i , b i } and pole P ∈ { N , S } P\in\{N,S\} P ∈ { N , S } , k P k^P k P denotes the 1-handle given by its product with the pole disk. The fiber band c j c_j c j over the base 0-handle gives the connecting 1-handle C j C_j C j ; its core interval is I j I_j I j . Write k − , k + k^-,k^+ k − , k + for the entry and exit feet of the base band. Give Σ g \Sigma_g Σ g one surface 0-handle, 2 g 2g 2 g surface 1-handles, and one surface 2-handle. Since r p r_p r p preserves the fiber disk-and-band decomposition, products of base and fiber handles give four-dimensional handles after rounding corners. The boundary of each product core attaches through products of lower total index, so the handles may be ordered by their four-dimensional indices. Over a base 1-handle the two disk products attach first, followed by the p p p band products. Over the closing base 2-handle, the two disk products attach before the p p p band products. The a 1 a_1 a 1 face pairing is r p r_p r p , so positive passage through the corresponding 1-handles has the inverse shift; the other base handles use the identity.
Figure 2. Planar disk-and-band drawings of the annulus, the pair of pants, and the four-holed sphere. The turquoise disks are 0-handles; the colored strips are 1-handles. The arrow on each strip is its core direction from N N N to S S S . The outer boundary is one of the p p p boundary components, not an additional component. Cyclic rotation is a diffeomorphism of the sphere model; its planar drawing can move an inner boundary to the outer one.
Table 1 (PDF p. 5).
The unreduced handle counts are
( h 0 , h 1 , h 2 , h 3 ) = ( 2 , 4 g + p , 2 + 2 g p , p ) . (h_0,h_1,h_2,h_3)=(2,\,4g+p,\,2+2gp,\,p). ( h 0 , h 1 , h 2 , h 3 ) = ( 2 , 4 g + p , 2 + 2 g p , p ) .
The product of a base band and a fiber band gives a 2-handle with core square [ − 1 , 1 ] s × [ − 1 , 1 ] u [-1,1]_s \times[-1,1]_u [ − 1 , 1 ] s × [ − 1 , 1 ] u and normal square [ − 1 , 1 ] t × [ − 1 , 1 ] v [-1,1]_t \times[-1,1]_v [ − 1 , 1 ] t × [ − 1 , 1 ] v . Rounding corners gives its core and normal disks. The s s s -faces attach to the fiber bands over the endpoints of the base band. The u u u -faces attach to the base-band products with N N N and S S S , which have already been added. The face pairing at a 1 a_1 a 1 identifies c j c_j c j with c j + 1 c_{j+1} c j + 1 ; positive passage through the corresponding 1-handle therefore joins the j + 1 j+1 j + 1 entry to the j j j exit; along any other base generator it lands on c j c_j c j . These four face maps, including their collar coordinates, are the attaching map on S 1 × D 2 S^1 \times D^2 S 1 × D 2 . The product framing ( ∂ t , ∂ v ) (\partial_t,\partial_v) ( ∂ t , ∂ v ) is transported by those same maps. Along positive a 1 a_1 a 1 passage the pole-disk map is z ↦ e − σ P 2 π i s / p z z \mapsto e^{-\sigma_P 2\pi i s/p}z z ↦ e − σ P 2 π i s / p z , 0 ≤ s ≤ 1 0 \le s \le1 0 ≤ s ≤ 1 , where σ N = 1 \sigma_N=1 σ N = 1 and σ S = − 1 \sigma_S=-1 σ S = − 1 ; its vertical derivative is the same rotation. Other base 1-handles use the identity. Across C j C_j C j the base angle is fixed and the fiber-band normal matches + ∂ ϕ N +\partial_{\phi_N} + ∂ ϕ N with − ∂ ϕ S -\partial_{\phi_S} − ∂ ϕ S , where ϕ N , ϕ S \phi_N,\phi_S ϕ N , ϕ S are angular coordinates in the positively oriented pole disks. These maps specify the product framing. The flat quotient model selects the closing trivialization over the base 2-handle. Along each commutator, the forward and backward rotation paths cancel, with zero total lifted angular rotation. Thus both the endpoint transport and the winding in this chosen trivialization are fixed.
Attaching maps, coordinate charts, and projections
Use coordinates ( s , t ) (s,t) ( s , t ) on a base band and ( u , v ) (u,v) ( u , v ) on a fiber band. The normal coordinates of the square core are ( t , v ) (t,v) ( t , v ) . Over each pole the 0-handle is D b a s e 2 × D p o l e 2 D^2_{\mathrm{base}} \times D^2_{\mathrm{pole}} D base 2 × D pole 2 . Its boundary contains the product faces ∂ D b a s e 2 × D p o l e 2 \partial D^2_{\mathrm{base}} \times D^2_{\mathrm{pole}} ∂ D base 2 × D pole 2 and D b a s e 2 × ∂ D p o l e 2 D^2_{\mathrm{base}} \times\partial D^2_{\mathrm{pole}} D base 2 × ∂ D pole 2 . Put the base feet on the first face in the order inherited from the oriented surface polygon and the connecting 1-handle feet on the second face in the cyclic order of the sphere model. Each square is rounded on these product faces. Parallel attaching strips retain their product collars under the face pairings. These collars determine the attaching curves and their crossings; the relator words alone do not.
Figure 3. The two 0-handles and eleven 1-handles of M 2 , 3 M_{2,3} M 2 , 3 . The twelve product 2-handles and two closing 2-handles attach to this handlebody; their attaching circles are drawn below. M 2 , 3 M_{2,3} M 2 , 3 : the 0/1-handlebody (two 0-handles and eleven 1-handles). Blue and red thick strips denote 1-handles. No 2-handles are shown. To obtain the complement add twelve transported-square 2-handles, two pole closing-disk 2-handles, and three product 3-handles.
Figure 4. The ten 1-handles in dotted-circle notation after canceling C 0 C_0 C 0 with one 0-handle. Only the 0- and 1-handles are shown. To express the full attaching link in this notation, the eight strands through C 0 C_0 C 0 must be carried through the boundary connected-sum collar. That conversion has not been carried out. M 2 , 3 M_{2,3} M 2 , 3 : the 0/1-handlebody after canceling the tree edge c 0 c_0 c 0 . One implicit 0-handle and ten dotted, pairwise unlinked 1-handles. y 1 = c 1 c 0 − 1 y_1=c_1c_0^{-1} y 1 = c 1 c 0 − 1 and y 2 = c 2 c 0 − 1 y_2=c_2c_0^{-1} y 2 = c 2 c 0 − 1 . Only the 0/1-handles are shown.
In attaching words, c j c_j c j denotes passage through C j C_j C j . The closing 2-handles are products of the base surface 2-handle with the pole disks. With the group convention that a 1 a_1 a 1 conjugates by r p r_p r p , the oriented attaching circle follows the path
R k , j = k N c j ( k S ) − 1 c j + δ ( k ) − 1 , δ ( a 1 ) = 1 , δ ( k ) = 0 ( k ≠ a 1 ) . R_{k,j}=k^N c_j(k^S)^{-1}c_{j+\delta(k)}^{-1}, \qquad\delta(a_1)=1,\quad\delta(k)=0\ (k\ne a_1). R k , j = k N c j ( k S ) − 1 c j + δ ( k ) − 1 , δ ( a 1 ) = 1 , δ ( k ) = 0 ( k = a 1 ) .
The label R k , j R_{k,j} R k , j denotes both the product 2-handle and its oriented attaching circle; δ ( k ) \delta(k) δ ( k ) records whether k = a 1 k=a_1 k = a 1 . Reversing the base-band convention reverses the shift. The indices for the two degree pairs are:
Table 1 (PDF p. 6).
The letters in each row abbreviate the ordered base generators ( a 1 , b 1 , … , a g , b g ) (a_1,b_1,\ldots,a_g,b_g) ( a 1 , b 1 , … , a g , b g ) . In particular, for M 2 , 3 M_{2,3} M 2 , 3 , ( a , b , c , d ) = ( a 1 , b 1 , a 2 , b 2 ) (a,b,c,d)=(a_1,b_1,a_2,b_2) ( a , b , c , d ) = ( a 1 , b 1 , a 2 , b 2 ) . Its rotating attaching curves follow
a N c 0 ( a S ) − 1 c 1 − 1 , a N c 1 ( a S ) − 1 c 2 − 1 , a^N c_0(a^S)^{-1}c_1^{-1}, \qquad a^N c_1(a^S)^{-1}c_2^{-1}, a N c 0 ( a S ) − 1 c 1 − 1 , a N c 1 ( a S ) − 1 c 2 − 1 ,
a N c 2 ( a S ) − 1 c 0 − 1 . a^N c_2(a^S)^{-1}c_0^{-1}. a N c 2 ( a S ) − 1 c 0 − 1 .
For R a , 0 R_{a,0} R a , 0 , enter a N a^N a N on strand 1 and leave on strand 0, cross C 0 C_0 C 0 to S S S , pass backward through a S a^S a S from strand 0 to strand 1, and return through C 1 C_1 C 1 . The corner charts below join these passages and transport their normal disks. The nine nonrotating product 2-handles use matching connecting-handle indices. For M 3 , 2 M_{3,2} M 3 , 2 the two rotating product 2-handles exchange c 0 , c 1 c_0,c_1 c 0 , c 1 ; for M 2 , 4 M_{2,4} M 2 , 4 the four rotating product 2-handles advance cyclically.
Denote the two additional 2-handles by H N , H S H_N,H_S H N , H S and the p p p 3-handles by T j T_j T j . The 2-handles have base-disk cores at the pole centers. Their attaching curves are the two base-surface attaching curves ∏ i [ a i N , b i N ] \prod_i[a_i^N,b_i^N] ∏ i [ a i N , b i N ] and ∏ i [ a i S , b i S ] \prod_i[a_i^S,b_i^S] ∏ i [ a i S , b i S ] , embedded by the surface polygon with their pole-disk normal frames. The p p p 3-handle attachments are also specified by product faces: for a base closing handle D ξ 2 × D 0 D_\xi^2 \times D^0 D ξ 2 × D 0 and a fiber band D u 1 × D v 1 D_u^1 \times D_v^1 D u 1 × D v 1 , their cores are D ξ 2 × D u 1 D_\xi^2 \times D_u^1 D ξ 2 × D u 1 , and the attaching sphere is
( ∂ D ξ 2 × D u 1 ) ∪ ( D ξ 2 × ∂ D u 1 ) . (\partial D_\xi^2 \times D_u^1)\cup(D_\xi^2\times\partial D_u^1). ( ∂ D ξ 2 × D u 1 ) ∪ ( D ξ 2 × ∂ D u 1 ) .
The first face meets the previously attached product 2-handles and the second meets the closing 2-handles H N , H S H_N,H_S H N , H S . The normal interval is D v 1 D_v^1 D v 1 , with the transported collar coordinates. Figure 10 shows these attaching spheres.
Figure 10. The two closing 2-handle attaching curves and the three 3-handle attaching spheres. Each sphere consists of two closing-core disk faces joined by an annular face in the corresponding fiber-band collar. The normal interval is the transverse coordinate of that band. These are local product-face schematics, not a simultaneous planar projection of the entire post-attachment boundary.
Table 1 (PDF p. 7).
The alternating sum is ∑ ( − 1 ) i h i = ( 2 − 2 g ) ( 2 − p ) \sum(-1)^i h_i=(2-2g)(2-p) ∑ ( − 1 ) i h i = ( 2 − 2 g ) ( 2 − p ) .
We draw the attaching curves in one 0- and 1-handlebody. Order the eight base feet around each pole by
a − , b + , a + , b − , c − , d + , c + , d − . a^-,b^+,a^+,b^-,c^-,d^+,c^+,d^-. a − , b + , a + , b − , c − , d + , c + , d − .
Write their distinct base-disk angles as θ k ± \theta_{k\pm} θ k ± . Use complex coordinates ( z b a s e , z p o l e ) (z_{\mathrm{base}},z_{\mathrm{pole}}) ( z base , z pole ) on each rounded S 3 ⊂ C 2 S^3\subset\mathbb{C}^2 S 3 ⊂ C 2 , with arguments θ , ϕ \theta,\phi θ , ϕ . In positively oriented pole coordinates put ϕ N , j = 2 π j / 3 + α \phi_{N,j}=2\pi j/3+\alpha ϕ N , j = 2 π j /3 + α and ϕ S , j = − ( 2 π j / 3 + α ) \phi_{S,j}=-(2\pi j/3+\alpha) ϕ S , j = − ( 2 π j /3 + α ) , where α = 0.23 \alpha=0.23 α = 0.23 . The positive constants ϵ , ζ , η \epsilon,\zeta,\eta ϵ , ζ , η are respectively a polar cutoff, an angular half-width for the feet, and a normal-disk scale. Choose disjoint base-foot neighborhoods { ∣ θ − θ k ± ∣ ≤ ζ , ∣ z p o l e ∣ ≤ sin ϵ } \{|\theta-\theta_{k\pm}|\le\zeta,\ |z_{\mathrm{pole}}|\le\sin\epsilon\} { ∣ θ − θ k ± ∣ ≤ ζ , ∣ z pole ∣ ≤ sin ϵ } and connecting-handle foot neighborhoods { ∣ ϕ − ϕ P , j ∣ ≤ ζ , ∣ z b a s e ∣ ≤ sin ϵ } \{|\phi-\phi_{P,j}|\le\zeta,\ |z_{\mathrm{base}}|\le\sin\epsilon\} { ∣ ϕ − ϕ P , j ∣ ≤ ζ , ∣ z base ∣ ≤ sin ϵ } . These are product 3-balls after rounding. The two types are disjoint because ϵ < π / 4 \epsilon<\pi/4 ϵ < π /4 ; within either type take ζ \zeta ζ smaller than half the foot-angle separation and choose 0 < η < ζ 0<\eta<\zeta 0 < η < ζ . For the three pieces, ζ = 0.05 \zeta=0.05 ζ = 0.05 gives foot neighborhoods containing all the attaching strands. Extend corner endpoints slightly into their rounded collars.
For ϵ ≤ τ ≤ π / 2 − ϵ \epsilon\le\tau\le\pi/2-\epsilon ϵ ≤ τ ≤ π /2 − ϵ , use the corner arcs
Γ P , k ± , j ( τ ) = ( cos τ e i θ k ± , sin τ e i ϕ P , j ) . \Gamma_{P,k\pm,j}(\tau)=(\cos\tau\,e^{i\theta_{k\pm}},\sin\tau\,e^{i\phi_{P,j}}). Γ P , k ± , j ( τ ) = ( cos τ e i θ k ± , sin τ e i ϕ P , j ) .
Their local normal-disk charts are obtained by varying the two arguments. When ( t , v ) (t,v) ( t , v ) denotes the oriented surface-band coordinates, use θ k ± + λ ± η t \theta_{k\pm}+\lambda_{\pm}\eta t θ k ± + λ ± η t and ϕ P , j + σ P η v \phi_{P,j}+\sigma_P\eta v ϕ P , j + σ P η v , where λ − = 1 \lambda_-=1 λ − = 1 , λ + = − 1 \lambda_+=-1 λ + = − 1 , σ N = 1 \sigma_N=1 σ N = 1 , and σ S = − 1 \sigma_S=-1 σ S = − 1 . These signs record the orientation reversal between the two ends of a surface band. An unsigned local argument chart is equivalent only when its endpoint identification carries the same reversal. In either notation the normal frame also rotates along the distinguished base 1-handle. Choose η \eta η smaller than half the angular foot separation. Two different pairs of arguments cannot meet in the interior, so all twenty-four ribbons at each pole are disjoint. The coordinate atlas uses ε = 0.18 \varepsilon= 0.18 ε = 0.18 , η = 0.022 \eta= 0.022 η = 0.022 ; normal disks may be shrunk for the projections below.
Figure 5. The degree-two and degree-four transport maps used in the mixed case. Here ε > 0 \varepsilon> 0 ε > 0 is the strand radius in the pole disk, j j j labels the fiber strand, and 0 ≤ s ≤ 1 0 \le s \le1 0 ≤ s ≤ 1 . The northern coordinate is z j ( s ) = ε exp ( i ( 2 π ( j + s ) / p + 0.23 ) ) z_j(s) = \varepsilon\exp(i(2\pi(j+s)/p+0.23)) z j ( s ) = ε exp ( i ( 2 π ( j + s ) / p + 0.23 )) . Here s s s increases opposite the positive displayed base generator. These drawings fix the crossings within one pole handle; they are not the entire attaching link of either complement.
Figure 6 projects the two distinct 0-handle boundaries. The transparent foot outlines are graphical silhouettes, not certified enclosures; strands may project across feet to which they do not attach. Pair k − k^{-} k − with k + k^{+} k + within each pole and C j C_j C j across poles, retaining all eleven 1-handles. The twelve product curves are colored; black intervals belong to the closing curves. Figure 7 and Figure 8 specify the strand identifications. In the base ribbon, a side at θ k − − ζ \theta_{k-}-\zeta θ k − − ζ matches θ k + + ζ \theta_{k+}+\zeta θ k + + ζ , and the other side has the opposite signs. This is the side reversal that gives the closing cycle 0, 3, 2, 1, 4, 7, 6, 5. Colors in Figure 7 and Figure 8 distinguish local strands; they do not identify components in the full projections.
Figure 6. M 2 , 3 M_{2,3} M 2 , 3 . Northern and southern projections of the attaching curves. Pair k − k^{-} k − with k + k^{+} k + within each pole and C j C_j C j across poles. Colored arcs belong to the product 2-handles and black arcs to the closing 2-handles. The transparent outlines mark the 1-handle feet. Figures 7 and 8 give the strand identifications and product framings.
Figure 7. Strand identifications through the base 1-handles over N N N and S S S . Here J j J_j J j labels the position of strand j j j in a pole-disk cross-section. The cyclic strands run from J j + 1 J_{j+1} J j + 1 to J j J_j J j along positive a a a ; the other base 1-handles use identity transport. Dashed push-offs indicate the fiber-band normal direction. Figure 8 gives the identifications through the connecting 1-handles and the framing of the closing 2-handle. Colors label local strands, not globally colored components in Figure 6 .
Figure 8. Strand identifications through the connecting 1-handles and normal transport for the closing 2-handle. The normal displacement is enlarged in the drawing; the push-off uses a small displacement with the same rotation. Product framings are specified by these normal coordinates; see the framing convention at the start of Section 2. Colors label local strands, not the global component colors of Figure 6.
For the projection, regard q ∈ S 3 ⊂ R 4 q \in S^3 \subset\mathbb{R}^4 q ∈ S 3 ⊂ R 4 as a row vector. Choose a unit pole ℓ \ell ℓ off the curves, a 4 × 3 4 \times3 4 × 3 matrix E E E whose columns are an orthonormal basis of ℓ ⊥ \ell^\perp ℓ ⊥ , and a 3 × 3 3 \times3 3 × 3 orthogonal matrix O O O . The projected spatial coordinates are
X ( q ) = q E 1 − q ⋅ ℓ O . X(q)=\frac{q\mathbf{E}}{1-q\cdot\boldsymbol{\ell}}\,\mathbf{O}. X ( q ) = 1 − q ⋅ ℓ q E O .
The first two coordinates give the planar projection; the third determines the overpasses. All curves in a drawing use this same map. For these projections use ϵ = 0.032 \epsilon=0.032 ϵ = 0.032 , ζ = 0.012 \zeta=0.012 ζ = 0.012 , and η p r o j = 0.005 < ζ \eta_{\mathrm{proj}}=0.005<\zeta η proj = 0.005 < ζ . These are smaller product neighborhoods of the same framed cores. To compare with the preceding atlas, vary the parameters through
( ϵ r , ζ r , η r ) = ( 1 − r ) ( 0.18 , 0.05 , 0.022 ) + r ( 0.032 , 0.012 , 0.005 ) , 0 ≤ r ≤ 1. (\epsilon_r,\zeta_r,\eta_r)=(1-r)(0.18,0.05,0.022)+r(0.032,0.012,0.005),\qquad0\le r\le1. ( ϵ r , ζ r , η r ) = ( 1 − r ) ( 0.18 , 0.05 , 0.022 ) + r ( 0.032 , 0.012 , 0.005 ) , 0 ≤ r ≤ 1.
The foot angles remain fixed. Throughout this deformation, 0 < ϵ r < π / 4 0<\epsilon_r<\pi/4 0 < ϵ r < π /4 and 0 < η r < ζ r 0<\eta_r<\zeta_r 0 < η r < ζ r , and the foot neighborhoods remain disjoint. Changing the cutoff extends each corner arc along its fixed pair of arguments; changing the foot width extends the closing intervals into the same collars. Reparameterize the adjacent handle collars accordingly and carry the normal frames with them. The product characteristic maps of the subsequent attaching spheres are transported by the same collar isotopy. Hence this change preserves the framed attaching maps and all higher attachments. The camera data are recorded in figures/usual_projection_data.json; the interval certificates are in certified_projection_crossings.json. The stored camera entries are rounded decimals; orthonormalizing them changes each entry by less than 10 − 12 10^{-12} 1 0 − 12 and preserves all the crossings and overpasses below. The attaching maps and framings are still those of the product charts.
Proposition 2.1. The analytic arcs in the six pole projections have only transverse double points, with the following numbers of crossings:
Table 3.
Their overpasses agree with the recorded figure data. No individual arc has a self-intersection and no three arcs have a common projected point.
Appendix § 9 proves Proposition 2.1 by interval subdivision over the full parameter domains, including endpoints. Numerical searches only choose candidate neighborhoods.
For R k , j R_{k,j} R k , j , let q = j + δ ( k ) q=j+\delta(k) q = j + δ ( k ) , with indices modulo three. The attaching curve follows
Γ N , k − , q − 1 , k N : q ⟶ j , Γ N , k + , j , C j : N ⟶ S , \Gamma^{-1}_{N,k-,q},\quad k^N:q\longrightarrow j,\quad\Gamma_{N,k+,j},\quad C_j:N\longrightarrow S, Γ N , k − , q − 1 , k N : q ⟶ j , Γ N , k + , j , C j : N ⟶ S ,
Γ S , k + , j − 1 , ( k S ) − 1 : j ⟶ q , Γ S , k − , q , C q : S ⟶ N . \Gamma^{-1}_{S,k+,j},\quad(k^S)^{-1}:j\longrightarrow q,\quad\Gamma_{S,k-,q},\quad C_q:S\longrightarrow N. Γ S , k + , j − 1 , ( k S ) − 1 : j ⟶ q , Γ S , k − , q , C q : S ⟶ N .
At each connecting 1-handle C j C_j C j , strand ( k , + ) (k,+) ( k , + ) belongs to R k , j R_{k,j} R k , j and strand ( k , − ) (k,-) ( k , − ) belongs to R k , j − δ ( k ) R_{k,j-\delta(k)} R k , j − δ ( k ) . The strand’s base angle is constant from N N N to S S S . These eight strands through each connecting 1-handle match the same corner arcs at both ends.
The remaining attaching curves are the boundaries of the two base disk-and-four-band ribbons at pole coordinate zero. Their cycle visits foot indices 0, 3, 2, 1, 4, 7, 6, 5, giving [ a , b ] [ c , d ] [a,b][c,d] [ a , b ] [ c , d ] . Both sides of every band are drawn. The pole-coordinate normal disk fixes the framing. After the product 2-handles have been attached, the closing 2-handles are D c l o s i n g 2 × D N 2 D^2_{\mathrm{closing}}\times D^2_N D closing 2 × D N 2 and D c l o s i n g 2 × D S 2 D^2_{\mathrm{closing}}\times D^2_S D closing 2 × D S 2 , with the pole disks as normal factors. The parallel curves show the framing. Before the product 2-handles are attached, the normal framing extending over each punctured pole section agrees with the Seifert longitude of its boundary in the 0- and 1-handlebody. Here the closing curve is null-homologous and the 0- and 1-handlebody has H 2 = 0 H_2=0 H 2 = 0 . Capping with the closing core gives the pole section, whose normal Euler number is zero, since the commutator rotation paths cancel. Each closing attachment therefore has coefficient zero relative to this longitude, carried forward in a small neighborhood disjoint from the other attachments. To obtain an integer label in a dotted-circle presentation, this longitude must be carried through the chosen conversion of the paired feet. We may shrink the attaching neighborhoods while transporting the product normal coordinates and the face maps of the subsequent attaching spheres. Along a base band, the closing attachment follows the strips along the two boundary edges of the band, whereas the product 2-handle attachments use strips in its interior. On the base 0-handle these strips are separated by the corresponding foot-angle intervals. Taking disjoint product collars therefore separates the closing attaching ribbons from the attaching ribbons of the product 2-handles. The normal disk identifications for the higher attachments are taken in the boundary after attaching the product 2-handles. They need not extend as large normal disks in the original 0- and 1-handlebody. Write D P , j = D c l o s i n g 2 × { q P , j } D_{P,j}=D^2_{\mathrm{closing}}\times\{q_{P,j}\} D P , j = D closing 2 × { q P , j } for its disk face at pole P P P , where q P , j q_{P,j} q P , j is the fiber-band foot position in D 2 D^2 D 2 . The two disk faces of the j j j th attaching sphere are D N , j D_{N,j} D N , j and D S , j D_{S,j} D S , j . Their collars are those of the product construction.
Theorem 2.2. The attaching maps above give the handle decomposition of M 2 , 3 M_{2,3} M 2 , 3 in Theorem 1.1 . The same construction gives the decompositions of M 3 , 2 M_{3,2} M 3 , 2 and M 2 , 4 M_{2,4} M 2 , 4 drawn in the appendix.
Proof. The corner ribbons are disjoint by their two argument coordinates. The three strands within each 1-handle k P k^P k P remain at distinct disk angles, and the eight strands in each connecting 1-handle have distinct base angles. The specified endpoint matchings therefore give twelve disjoint framed product 2-handle attaching curves. They are exactly the products of the four base bands with the three fiber bands. The pole-center closing ribbons are separate from those curves and attach the two products of the closing base disk with the pole disks. Finally the three spheres are ∂ ( D c l o s i n g 2 × I j ) \partial(D^2_{\mathrm{closing}}\times I_j) ∂ ( D closing 2 × I j ) , with disjoint I j I_j I j collars and product normal intervals. These are the remaining three product attachments. Rounding the common faces recovers the quotient bundle. The drawings show these attachments in separate boundary charts. The scripts check_attachment_atlases.py and check_product_frames.py check the finite attachment recipes and symbolic frame identities. These consistency checks supplement the product-chart argument; they do not independently establish the geometric embeddings or frame transport. For the other two pieces replace the base foot list by ( a i − , b i + , a i + , b i − ) i = 1 g (a_i^-, b_i^+, a_i^+, b_i^-)_{i=1}^{g} ( a i − , b i + , a i + , b i − ) i = 1 g and use φ N , j = 2 π j / p + α \varphi_{N,j}=2\pi j/p+\alpha φ N , j = 2 π j / p + α , φ S , j = − φ N , j \varphi_{S,j}=-\varphi_{N,j} φ S , j = − φ N , j . There are 4 g p 4gp 4 g p corner charts at each pole, 2 g p 2gp 2 g p product 2-handles, 4 g 4g 4 g strands through each connecting 1-handle, two closing 2-handles, and p p p attaching spheres for the 3-handles. The same argument proves disjointness and compatibility of the framings. The closing foot cycle is the concatenation of ( 4 i , 4 i + 3 , 4 i + 2 , 4 i + 1 ) (4i,4i+3,4i+2,4i+1) ( 4 i , 4 i + 3 , 4 i + 2 , 4 i + 1 ) for 0 ≤ i < g 0\le i<g 0 ≤ i < g , and reads ∏ i = 1 g [ a i , b i ] \prod_{i=1}^{g}[a_i,b_i] ∏ i = 1 g [ a i , b i ] . The appendix draws these specializations.
Integral homology
We compute the handle chain complex from the attaching curves and the product faces of the attaching spheres.
Theorem 2.3. For g ≥ 2 g \ge2 g ≥ 2 and p ≥ 2 p \ge2 p ≥ 2 , the displayed decomposition of M g , p M_{g,p} M g , p has
H 0 ( M g , p ; Z ) = Z , H_0(M_{g,p};\mathbb{Z}) = \mathbb{Z}, H 0 ( M g , p ; Z ) = Z , H 1 ( M g , p ; Z ) = Z 2 g ⊕ Z / p , H_1(M_{g,p};\mathbb{Z}) = \mathbb{Z}^{2g} \oplus\mathbb{Z}/p, H 1 ( M g , p ; Z ) = Z 2 g ⊕ Z / p , H 2 ( M g , p ; Z ) = Z 1 + 2 ( g − 1 ) ( p − 1 ) ⊕ Z / p , H_2(M_{g,p};\mathbb{Z}) = \mathbb{Z}^{1+2(g-1)(p-1)} \oplus\mathbb{Z}/p, H 2 ( M g , p ; Z ) = Z 1 + 2 ( g − 1 ) ( p − 1 ) ⊕ Z / p , H 3 ( M g , p ; Z ) = H 4 ( M g , p ; Z ) = 0. H_3(M_{g,p};\mathbb{Z}) = H_4(M_{g,p};\mathbb{Z}) = 0. H 3 ( M g , p ; Z ) = H 4 ( M g , p ; Z ) = 0.
Proof. Orient each connecting core c j c_j c j from N N N to S S S , and orient R k , j R_{k,j} R k , j by the displayed attaching path. Use the handle labels for the corresponding integral chain generators; in particular, c j c_j c j is the generator associated to C j C_j C j . The 3-handle T j T_j T j has core D c l o s i n g 2 × I j D^2_{\mathrm{closing}} \times I_j D closing 2 × I j . The chain groups, in degrees zero through three, have ranks 2 2 2 , 4 g + p 4g+p 4 g + p , 2 + 2 g p 2+2gp 2 + 2 g p , and p p p . Their differentials are
Figure 9. The three rotating product 2-handles, shown as a sublink of the preceding drawings. Black arcs belong to the closing 2-handles; dashed arcs indicate the fiber-band normal direction, enlarged for visibility. Normal transport is as in Figure 8 , with opposite pole-coordinate signs.
∂ 1 ( c j ) = S − N , ∂ 1 ( k P ) = 0 , ∂ 2 ( R k , j ) = k N − k S + c j − c j + δ ( k ) , ∂ 2 ( H P ) = 0 , ∂ 3 ( T j ) = H S − H N + R b 1 , j − 1 − R b 1 , j . (1) \begin{aligned}
\partial_1(c_j)&=S-N, & \partial_1(k^P)&=0,\\
\partial_2(R_{k,j})&=k^N-k^S+c_j-c_{j+\delta(k)}, & \partial_2(H_P)&=0,\\
\partial_3(T_j)&=H_S-H_N+R_{b_1,j-1}-R_{b_1,j}. \tag*{(1)}
\end{aligned} ∂ 1 ( c j ) ∂ 2 ( R k , j ) ∂ 3 ( T j ) = S − N , = k N − k S + c j − c j + δ ( k ) , = H S − H N + R b 1 , j − 1 − R b 1 , j . ∂ 1 ( k P ) ∂ 2 ( H P ) = 0 , = 0 , (1)
Here all fiber indices are modulo p p p . The first two lines are the signed handle passages of the attaching curves. For the last line, the cap faces contribute H S − H N H_S-H_N H S − H N . Trace the annular face along [ a 1 , b 1 ] ⋯ [ a g , b g ] [a_1,b_1]\cdots[a_g,b_g] [ a 1 , b 1 ] ⋯ [ a g , b g ] . Starting at fiber index j j j , positive a 1 a_1 a 1 transport takes it to j − 1 j-1 j − 1 . The first four faces contribute
R a 1 , j − 1 + R b 1 , j − 1 − R a 1 , j − 1 − R b 1 , j . R_{a_1,j-1}+R_{b_1,j-1}-R_{a_1,j-1}-R_{b_1,j}. R a 1 , j − 1 + R b 1 , j − 1 − R a 1 , j − 1 − R b 1 , j .
The a 1 a_1 a 1 faces cancel. All later commutators have trivial fiber transport and their faces cancel in pairs. This proves the last line from the attaching sphere itself. The formulas satisfy ∂ 1 ∂ 2 = ∂ 2 ∂ 3 = 0 \partial_1\partial_2=\partial_2\partial_3=0 ∂ 1 ∂ 2 = ∂ 2 ∂ 3 = 0 .
A basis of ker ∂ 1 \ker\partial_1 ker ∂ 1 consists of the 4 g 4g 4 g base-handle generators k N , k S k^N,k^S k N , k S and the p − 1 p-1 p − 1 differences c j − c 0 c_j-c_0 c j − c 0 , 1 ≤ j < p 1\le j<p 1 ≤ j < p . For k ≠ a 1 k\ne a_1 k = a 1 the product 2-handle relations identify k N k^N k N and k S k^S k S . Put Δ a = a 1 N − a 1 S \Delta_a=a_1^N-a_1^S Δ a = a 1 N − a 1 S . The rotating 2-handle relations give c j + 1 − c j = Δ a c_{j+1}-c_j=\Delta_a c j + 1 − c j = Δ a , and their cyclic sum gives p Δ a = 0 p\Delta_a=0 p Δ a = 0 . Eliminating c j − c 0 = j Δ a c_j-c_0=j\Delta_a c j − c 0 = j Δ a and a 1 S = a 1 N − Δ a a_1^S=a_1^N-\Delta_a a 1 S = a 1 N − Δ a leaves the 2 g 2g 2 g northern base edges free and Δ a \Delta_a Δ a of exact order p p p .
For degree two, a free integral basis of ker ∂ 2 \ker\partial_2 ker ∂ 2 is
H N , H S , z k , j = R k , j − R k , 0 ( k ≠ a 1 , 1 ≤ j < p ) . H_N,\ H_S,\quad z_{k,j}=R_{k,j}-R_{k,0}\quad(k\ne a_1,\ 1\le j<p). H N , H S , z k , j = R k , j − R k , 0 ( k = a 1 , 1 ≤ j < p ) .
Indeed the connecting-handle coefficients force all coefficients of the rotating a 1 a_1 a 1 2-handles to be equal; the a 1 N a_1^N a 1 N coefficient is then p p p times that integer and forces it to vanish. For every other k k k , the sum of its product 2-handle coefficients must vanish, giving exactly the differences z k , j z_{k,j} z k , j . Put h = H S − H N h=H_S-H_N h = H S − H N and z j = z b 1 , j z_j=z_{b_1,j} z j = z b 1 , j , with z 0 = 0 z_0=0 z 0 = 0 . The 3-handle relations are
h + z j − 1 − z j = 0 ( j m o d p ) . h+z_{j-1}-z_j=0\qquad(j\mathbin{\bmod}p). h + z j − 1 − z j = 0 ( j mod p ) .
The relations give z j = j h z_j=jh z j = j h . Eliminating z 1 , … , z p − 1 z_1,\ldots,z_{p-1} z 1 , … , z p − 1 leaves p h = 0 ph=0 p h = 0 . The assignment h ↦ 1 h\mapsto1 h ↦ 1 , z j ↦ j z_j\mapsto j z j ↦ j to Z / p \mathbb{Z}/p Z / p respects every relation, so h h h has exact order p p p . The generator H N H_N H N and all z k , j z_{k,j} z k , j for k ∉ { a 1 , b 1 } k\notin\{a_1,b_1\} k ∈ / { a 1 , b 1 } remain free. This gives the asserted H 2 H_2 H 2 . Finally a vanishing combination of the 3-handle columns forces all its coefficients equal, by the cyclic b 1 b_1 b 1 differences. Its cap coefficient is p p p times that common integer, so it is zero. Thus ∂ 3 \partial_3 ∂ 3 is injective and H 3 = 0 H_3=0 H 3 = 0 . There are no 4-handles.
The complement exact sequence gives an independent check of the H 2 H_2 H 2 torsion; see Appendix § 8 .
Corollary 2.4. Let S N S_N S N and S S S_S S S be the two pole sections of M g , p → Σ g M_{g,p} \to\Sigma_g M g , p → Σ g , oriented by the base. Then [ S S ] − [ S N ] [S_S]-[S_N] [ S S ] − [ S N ] generates the torsion subgroup of H 2 ( M g , p ; Z ) H_2(M_{g,p};\mathbb{Z}) H 2 ( M g , p ; Z ) and has exact order p p p .
Proof. In the product handle complex the two sections have cellular 2-cycles H N H_N H N and H S H_S H S . Thus their difference is the class h h h in the preceding proof. Summing the p p p attaching-sphere relations cancels the R b 1 , j R_{b_1,j} R b 1 , j terms and gives p h = 0 ph=0 p h = 0 . Conversely, eliminating z 1 , … , z p − 1 z_1,\ldots,z_{p-1} z 1 , … , z p − 1 leaves exactly this relation on h h h , so its order is p p p . This also identifies the torsion class geometrically rather than only through a Smith normal form.
Table 2.
The Smith-form checks in check_diagram_homology.py, recorded in verified_diagram_homology.json, verify the chain algebra; Theorem 2.2 supplies its geometric identification.
Boundary markings and surface gluings
The covering surface and boundary inclusion
The boundary is connected even though the fiber has p p p boundary circles: r p r_p r p permutes them transitively. It is a circle bundle over F g , p F_{g,p} F g , p , and its map to the base factors through the covering q p : F g , p → Σ g q_p:F_{g,p}\to\Sigma_g q p : F g , p → Σ g . The chosen normal framing gives
∂ M g , p ≅ F g , p × S 1 ≅ Σ G × S 1 . \partial M_{g,p}\cong F_{g,p}\times S^1\cong\Sigma_G\times S^1. ∂ M g , p ≅ F g , p × S 1 ≅ Σ G × S 1 .
Cut the base along a nonseparating curve dual to the character χ p \chi_p χ p , take p p p copies of the cut surface, and cyclically identify the cut boundaries. Choose the basepoint lift in sheet zero, with the base paths from this cut-and-glue model. This gives the marked surface-group generators:
E = a 1 p , B = b 1 , C i , r = a 1 r a i + 1 a 1 − r , D i , r = a 1 r b i + 1 a 1 − r . E=a_1^p,\qquad B=b_1,\qquad C_{i,r}=a_1^r a_{i+1}a_1^{-r},\qquad D_{i,r}=a_1^r b_{i+1}a_1^{-r}. E = a 1 p , B = b 1 , C i , r = a 1 r a i + 1 a 1 − r , D i , r = a 1 r b i + 1 a 1 − r .
Here 1 ≤ i < g 1\leq i<g 1 ≤ i < g , 0 ≤ r < p 0\leq r<p 0 ≤ r < p . With the ordering by sheet r r r , they satisfy
[ E , B ] ∏ r = 0 p − 1 ∏ i = 1 g − 1 [ C i , r , D i , r ] = 1. [E,B]\prod_{r=0}^{p-1}\prod_{i=1}^{g-1}[C_{i,r},D_{i,r}]=1. [ E , B ] r = 0 ∏ p − 1 i = 1 ∏ g − 1 [ C i , r , D i , r ] = 1.
To check this directly, write B r = a 1 r b 1 a 1 − r B_r=a_1^r b_1a_1^{-r} B r = a 1 r b 1 a 1 − r and K r = ∏ i [ C i , r , D i , r ] K_r=\prod_i[C_{i,r},D_{i,r}] K r = ∏ i [ C i , r , D i , r ] . Lifting the base relation gives B r + 1 = K r − 1 B r B_{r+1}=K_r^{-1}B_r B r + 1 = K r − 1 B r for r < p − 1 r<p-1 r < p − 1 ; the last lift gives [ E , B 0 ] K 0 ⋯ K p − 1 = 1 [E,B_0]K_0\cdots K_{p-1}=1 [ E , B 0 ] K 0 ⋯ K p − 1 = 1 .
For the surface relator we order the pairs by sheet r r r . For homology, put ( e 0 , f 0 ) = ( [ E ] , [ B ] ) (e_0,f_0)=([E],[B]) ( e 0 , f 0 ) = ([ E ] , [ B ]) and ( e i , r , f i , r ) = ( [ C i , r ] , [ D i , r ] ) (e_{i,r},f_{i,r})=([C_{i,r}],[D_{i,r}]) ( e i , r , f i , r ) = ([ C i , r ] , [ D i , r ]) ; after the initial pair, use lexicographic order ( i , r ) (i,r) ( i , r ) . The pairs are represented by the cover’s initial torus and its lifted one-holed tori. Reordering the based generators requires Hurwitz exchanges; the resulting conjugations are recorded in the mixed-family marking of § 4.2 . Define Q p , g = ( q p ) ∗ : H 1 ( F g , p ; Z ) → H 1 ( Σ g ; Z ) Q_{p,g}=(q_p)_*:H_1(F_{g,p};\mathbb{Z})\to H_1(\Sigma_g;\mathbb{Z}) Q p , g = ( q p ) ∗ : H 1 ( F g , p ; Z ) → H 1 ( Σ g ; Z ) . Its values on this basis are
Q p , g ( e 0 ) = p a 1 , Q p , g ( f 0 ) = b 1 , Q p , g ( e i , r ) = a i + 1 , Q p , g ( f i , r ) = b i + 1 . Q_{p,g}(e_0)=pa_1,\qquad Q_{p,g}(f_0)=b_1,\qquad Q_{p,g}(e_{i,r})=a_{i+1},\qquad Q_{p,g}(f_{i,r})=b_{i+1}. Q p , g ( e 0 ) = p a 1 , Q p , g ( f 0 ) = b 1 , Q p , g ( e i , r ) = a i + 1 , Q p , g ( f i , r ) = b i + 1 .
Figure 11. A marked connected-sum model of the genus-four boundary surface. Each T j T_j T j is a one-holed torus attached to the planar core, and carries its indicated red-blue symplectic marking. The drawing records the surface marking used to define the gluing maps. The torus outlines depict subsurfaces.
Choose the deleted disks to be geodesic disks of one fixed small radius, and let z 0 z_0 z 0 be the center of the reference deleted disk, whose boundary circle has based peripheral loop x 1 x_1 x 1 . For η ∈ S 1 \eta\in S^1 η ∈ S 1 , write R η = R arg η R_\eta=R_{\arg\eta} R η = R a r g η . On the pullback to the root cover q p : F g , p → Σ g q_p:F_{g,p}\to\Sigma_g q p : F g , p → Σ g , the map
F g , p × P p ⟶ q p ∗ M g , p , ( ( x , z ) , v ) ⟼ ( ( x , z ) , R z / z 0 v ) F_{g,p}\times P_p\longrightarrow q_p^*M_{g,p},\qquad((x,z),v)\longmapsto((x,z),R_{z/z_0}v) F g , p × P p ⟶ q p ∗ M g , p , (( x , z ) , v ) ⟼ (( x , z ) , R z / z 0 v )
is a global fiberwise diffeomorphism. It fixes the pole section. For the selected cuff, its boundary restriction identifies F g , p × S 1 F_{g,p}\times S^1 F g , p × S 1 with ∂ M g , p \partial M_{g,p} ∂ M g , p , using the equatorial framing of [6 , 8 ] , with the oriented equatorial tangent and transverse vector as the normal frame.
Fix a reference path γ : [ 0 , 1 ] → P p \gamma:[0,1]\to P_p γ : [ 0 , 1 ] → P p from the north pole to one point of the selected cuff. For a based root-cover loop ℓ ( s ) = ( x ( s ) , z ( s ) ) \ell(s)=(x(s),z(s)) ℓ ( s ) = ( x ( s ) , z ( s )) , transport the entire path by R z ( s ) / z 0 R_{z(s)/z_0} R z ( s ) / z 0 . The map
H ( s , t ) = ( x ( s ) , R z ( s ) / z 0 γ ( t ) ) , 0 ≤ s , t ≤ 1 , H(s,t)=\bigl(x(s),R_{z(s)/z_0}\gamma(t)\bigr),\qquad0\leq s,t\leq1, H ( s , t ) = ( x ( s ) , R z ( s ) / z 0 γ ( t ) ) , 0 ≤ s , t ≤ 1 ,
is a homotopy from the pole-section lift to the fixed-angle boundary lift, with the two vertical sides equal to the chosen reference path. In the pullback product trivialization, this path is constant in the P p P_p P p factor.
Thus the boundary lift of a 1 p a_1^p a 1 p has no additional meridional factor, even when z ( s ) z(s) z ( s ) winds around S 1 S^1 S 1 . The same argument applies to every covering generator, so their based images are exactly the displayed words in the kernel of χ p \chi_p χ p . Under the splitting of Theorem 2.3 , the boundary inclusion induces
j ∗ : H 1 ( F g , p ; Z ) ⊕ Z ⟶ H 1 ( Σ g ; Z ) ⊕ Z / p , j_* : H_1(F_{g,p};\mathbb{Z}) \oplus\mathbb{Z} \longrightarrow H_1(\Sigma_g;\mathbb{Z}) \oplus\mathbb{Z}/p, j ∗ : H 1 ( F g , p ; Z ) ⊕ Z ⟶ H 1 ( Σ g ; Z ) ⊕ Z / p ,
j ∗ ( v , k ) = ( Q p , g v , k m o d p ) . (2) j_*(v,k) = (Q_{p,g}v,k \bmod p). \tag*{(2)} j ∗ ( v , k ) = ( Q p , g v , k mod p ) . (2)
The cyclic target summand is generated by the positive meridian t = x 1 t=x_1 t = x 1 . Here positive means the boundary orientation of the oriented normal disk to the multisection, opposite to the cuff orientation induced from P p P_p P p . With the chosen connecting-handle orientations, [ t ] = [ c 0 − c 1 ] = − Δ a [t]=[c_0-c_1]=-\Delta_a [ t ] = [ c 0 − c 1 ] = − Δ a .
The same trivialization proves boundary incompressibility. The selected boundary component of the pullback has inclusion
π 1 ( F g , p ) × Z ⟶ π 1 ( F g , p ) × π 1 ( P p ) , \pi_1(F_{g,p}) \times\mathbb{Z} \longrightarrow\pi_1(F_{g,p}) \times\pi_1(P_p), π 1 ( F g , p ) × Z ⟶ π 1 ( F g , p ) × π 1 ( P p ) ,
equal to the identity in the first factor and the peripheral-circle inclusion in the second. The latter is injective for p ≥ 2 p \ge2 p ≥ 2 . The pullback has p p p boundary components, one for each reference cuff. Each maps with degree one to the original boundary: a boundary point determines its unique nearby root center, and the choice of reference cuff then fixes the root-cover coordinate. Thus the selected component maps diffeomorphically to the original boundary. The covering homomorphisms are injective. Thus the boundary inclusion is π 1 \pi_1 π 1 -injective.
Surface diffeomorphisms and symplectic compatibility
Identify each cover with a surface assembled from a planar core and ordered marked one-holed tori T 0 , … , T G − 1 T_0,\ldots,T_{G-1} T 0 , … , T G − 1 . This identification uses the covering basis above. Reindex the G G G homology pairs as ( e j , f j ) (e_j,f_j) ( e j , f j ) , with the initial pair indexed by zero and
j = 1 + ( i − 1 ) p + r j=1+(i-1)p+r j = 1 + ( i − 1 ) p + r
for the pair previously indexed by ( i , r ) (i,r) ( i , r ) . Choose based torus generators A j , B j A_j,B_j A j , B j with [ A j ] = e j [A_j]=e_j [ A j ] = e j and [ B j ] = f j [B_j]=f_j [ B j ] = f j , including the Hurwitz base-path corrections required by this order. They satisfy
∏ j = 0 G − 1 [ A j , B j ] = 1. \prod_{j=0}^{G-1}[A_j,B_j]=1. j = 0 ∏ G − 1 [ A j , B j ] = 1.
The indices of adjacent tori below refer to this ordered based marking. The mixed-family correction is given explicitly in § 4.2 . Cap the outer boundary of the planar core, keeping the basepoint in that cap. We use Dehn twists and exchanges of adjacent one-holed tori.
For an integer c c c , on T j T_j T j let U j , c U_{j,c} U j , c and L j , c L_{j,c} L j , c be the powers of the twists, identity on the boundary collar, with actions
U c = ( 1 c 0 1 ) , L c = ( 1 0 c 1 ) U_c=\begin{pmatrix}1&c\\0&1\end{pmatrix},\qquad L_c=\begin{pmatrix}1&0\\c&1\end{pmatrix} U c = ( 1 0 c 1 ) , L c = ( 1 c 0 1 )
on the column basis ( e j , f j ) (e_j,f_j) ( e j , f j ) . The twist signs are fixed by these matrices. On standard based torus generators A , B A,B A , B one can choose representatives
U c : ( A , B ) ↦ ( A , B A c ) , L c : ( A , B ) ↦ ( A B c , B ) . U_c:(A,B)\mapsto(A,BA^c),\qquad L_c:(A,B)\mapsto(AB^c,B). U c : ( A , B ) ↦ ( A , B A c ) , L c : ( A , B ) ↦ ( A B c , B ) .
Both preserve [ A , B ] [A,B] [ A , B ] exactly. All matrix products act on column vectors, so the rightmost twist is performed first.
Define E j \mathcal{E}_j E j to exchange the adjacent marked tori T j T_j T j and T j + 1 T_{j+1} T j + 1 . Move their attaching circles along complementary semicircles in the planar core, preserving the chosen collar coordinates. Choose the signed exchange and transported base paths by the Hurwitz convention below. Transport the attached tori by their marked parametrizations. The resulting diffeomorphism is the identity outside the genus-two subsurface containing these tori and their joining region and sends each marked pair to the corresponding pair in its new position. For based generators A j A_j A j , B j B_j B j , put h j = [ A j , B j ] h_j=[A_j,B_j] h j = [ A j , B j ] and choose the exchange sign by the following base path convention:
E j ( A j ) = A j + 1 , E j ( B j ) = B j + 1 , E j ( A j + 1 ) = h j + 1 − 1 A j h j + 1 , E j ( B j + 1 ) = h j + 1 − 1 B j h j + 1 . \begin{aligned}
\mathcal{E}_j(A_j)&=A_{j+1}, & \mathcal{E}_j(B_j)&=B_{j+1},\\
\mathcal{E}_j(A_{j+1})&=h_{j+1}^{-1}A_jh_{j+1}, & \mathcal{E}_j(B_{j+1})&=h_{j+1}^{-1}B_jh_{j+1}.
\end{aligned} E j ( A j ) E j ( A j + 1 ) = A j + 1 , = h j + 1 − 1 A j h j + 1 , E j ( B j ) E j ( B j + 1 ) = B j + 1 , = h j + 1 − 1 B j h j + 1 .
All other generators are fixed. The Hurwitz move is h j h j + 1 ↦ h j + 1 ( h j + 1 − 1 h j h j + 1 ) h_jh_{j+1}\mapsto h_{j+1}(h_{j+1}^{-1}h_jh_{j+1}) h j h j + 1 ↦ h j + 1 ( h j + 1 − 1 h j h j + 1 ) , which preserves the ordered surface relator exactly. The transported base path of the second torus follows the inverse boundary loop of the first. This path accounts for the conjugations in the formula. Together with the twist formulas, these substitutions determine the induced homomorphisms on the boundary fundamental groups.
Figure 12. Unbased trajectories of an exchange in the planar core. The Hurwitz formulas specify the exchange sign and the transported base paths.
Let B j \mathcal{B}_j B j be the 2 × 2 2\times2 2 × 2 homology matrix assigned to the j j j th torus. Its entries are specified for each family in § 4 . Write π ( j ) \pi(j) π ( j ) for the target position of this torus; a permutation is listed as ( π ( 0 ) , … , π ( G − 1 ) ) (\pi(0),\ldots,\pi(G-1)) ( π ( 0 ) , … , π ( G − 1 )) . Apply the specified maps on the marked tori to obtain D m D_m D m , then perform the specified exchanges to obtain P P P . The subscript m ≥ 1 m\geq1 m ≥ 1 is the integer twist parameter. Our surface gluing is f m = P ∘ D m : F L → F R f_m=P\circ D_m:F_L\to F_R f m = P ∘ D m : F L → F R , where F L , F R F_L,F_R F L , F R are the two marked covering surfaces. Its action on first homology is the 2 G × 2 G 2G\times2G 2 G × 2 G matrix S m = ( f m ) ∗ S_m=(f_m)_* S m = ( f m ) ∗ , with target/right rows and source/left columns in the chosen ordered bases. For v ∈ Σ G v\in\Sigma_G v ∈ Σ G and normal-circle angle θ \theta θ , the boundary gluing is
ϕ m ( v , e i θ ) = ( f m ( v ) , e − i θ ) . \phi_m(v,e^{i\theta})=(f_m(v),e^{-i\theta}). ϕ m ( v , e i θ ) = ( f m ( v ) , e − i θ ) .
It preserves the surface orientation and reverses the normal circle. The normal-circle reversal is independent of v v v .
Rescale the product symplectic forms to give the multisections equal area. Choose the marked identifications in local area coordinates so that the induced forms agree near one fixed disk containing the cap, foot patches, and tree; for the symmetric case use identical copies. Write ω L , ω R \omega_L,\omega_R ω L , ω R for the induced forms in the chosen markings. The tree joins the chosen foot patches and has a disk neighbourhood; it does not contain the essential cycles on the one-holed tori. For Appendix § 7 it is the lifted tree a 0 ∪ a 1 a_0\cup a_1 a 0 ∪ a 1 ; see the fixed-disk placement in Appendix § 7 . The complement of this disk retains the genus. Choose the marked tori and the paths of their exchanges in that complement. The forms have equal total area and agree near the disk. Choose all twists and exchanges to be the identity on this disk. Thus β = f m ∗ ω R − ω L \beta=f_m^*\omega_R-\omega_L β = f m ∗ ω R − ω L vanishes near the disk and has integral zero. On the connected complement of a smaller disk still containing the foot patches and tree, integration identifies compactly supported degree-two cohomology with R \mathbb{R} R . Thus β \beta β , which has integral zero, has a compactly supported primitive there. Extend this primitive by zero to obtain α M \alpha_M α M with d α M = β d\alpha_M=\beta d α M = β . The positive forms ω t = ω L + t β \omega_t=\omega_L+t\beta ω t = ω L + tβ , 0 ≤ t ≤ 1 0\leq t\leq1 0 ≤ t ≤ 1 , and the equation ι X t ω t = − α M \iota_{X_t}\omega_t=-\alpha_M ι X t ω t = − α M give a flow h t h_t h t fixed near these marked sets, with h 1 ∗ f m ∗ ω R = ω L h_1^*f_m^*\omega_R=\omega_L h 1 ∗ f m ∗ ω R = ω L . Replacing f m f_m f m by f m ∘ h 1 f_m\circ h_1 f m ∘ h 1 therefore preserves its based mapping class and foot charts while making the surface gluing area preserving. Extend the isotopy by
( w , z ) ⟼ ( h t ( w ) , z ) . (w,z)\longmapsto(h_t(w),z). ( w , z ) ⟼ ( h t ( w ) , z ) .
The normal coordinate is fixed, so this extension preserves its trivialization and introduces no surface-dependent normal-circle twist. The vertical oriented planes project isomorphically to the normal quotient of the multisection. Choose the transverse vector so that the equatorial tangent and transverse vector form a positive frame. Project this frame to the symplectic normal bundle and normalize by positive rescaling, without an additional circle-valued change of frame. Choose symplectic tubular charts in this framing. Normal-circle reversal then corresponds to a complex anti-linear identification of the trivial normal bundles, as required in the symplectic sum construction [11 ] . The inverse of the adjusted surface map is h t − 1 ∘ f m − 1 h_t^{-1} \circ f_m^{-1} h t − 1 ∘ f m − 1 . Transporting the entire attaching graph together with its framing through h t − 1 h_t^{-1} h t − 1 gives an isotopy fixed near the feet. Its product lift, with the normal-circle coordinate fixed, extends over a boundary collar by isotopy extension. These are the same framed attachments, relative to the feet, as those specified using f m − 1 f_m^{-1} f m − 1 in Appendix B. Composing f m f_m f m with a Torelli mapping class preserves S m S_m S m but may change the smooth sum.
The integral gluing calculation
Let M L = M g , p M_L=M_{g,p} M L = M g , p and M R = M h , q M_R=M_{h,q} M R = M h , q with 1 + p ( g − 1 ) = 1 + q ( h − 1 ) = G 1+p(g-1)=1+q(h-1)=G 1 + p ( g − 1 ) = 1 + q ( h − 1 ) = G , and let ϕ ( v , e i θ ) = ( f ( v ) , e − i θ ) \phi(v,e^{i\theta})=(f(v),e^{-i\theta}) ϕ ( v , e i θ ) = ( f ( v ) , e − i θ ) for an orientation-preserving surface diffeomorphism f f f . Using the equatorial framings, pole-section splittings, and inclusion (2) , we recall the product-framed Mayer–Vietoris formula to fix the boundary-map conventions.
Proposition 3.1 ([6 ] , Proposition 3.7). Put S = f ∗ S=f_* S = f ∗ and
A f = ( Q p , g − Q q , h S ) A_f=\begin{pmatrix} Q_{p,g} \\ -Q_{q,h}S \end{pmatrix} A f = ( Q p , g − Q q , h S ) For Z = M L ∪ ϕ M R Z=M_L\cup_{\phi}M_R Z = M L ∪ ϕ M R there is an isomorphism
H 1 ( Z ; Z ) ≅ coker A f ⊕ Z / gcd ( p , q ) . H_1(Z;\mathbb{Z})\cong\operatorname{coker} A_f\oplus\mathbb{Z}/\gcd(p,q). H 1 ( Z ; Z ) ≅ coker A f ⊕ Z / g cd( p , q ) . Here Z / 1 = 0 \mathbb{Z}/1=0 Z /1 = 0 . The matrix A f A_f A f has 2 ( g + h ) 2(g+h) 2 ( g + h ) rows and 2 G 2G 2 G columns, so it is square precisely when G = g + h G=g+h G = g + h . If it is also nonsingular, then
∣ H 1 ( Z ; Z ) ∣ = ∣ det A f ∣ gcd ( p , q ) . \lvert H_1(Z;\mathbb{Z})\rvert=\lvert\det A_f\rvert\gcd(p,q). ∣ H 1 ( Z ; Z )∣ = ∣ det A f ∣ g cd( p , q ) .
Proof. The pieces and their common boundary are connected. The map on H 0 H_0 H 0 in Mayer–Vietoris sends 1 1 1 to ( 1 , − 1 ) (1,-1) ( 1 , − 1 ) and is injective. Consequently H 1 ( Z ; Z ) H_1(Z;\mathbb{Z}) H 1 ( Z ; Z ) is the cokernel of the boundary map on H 1 H_1 H 1 . In the specified splittings this map is
( v , k ) ⟼ ( Q p , g v , − Q q , h S v ; k m o d p , k m o d q ) . (v,k)\longmapsto(Q_{p,g}v,-Q_{q,h}Sv;\ k\bmod p,k\bmod q). ( v , k ) ⟼ ( Q p , g v , − Q q , h S v ; k mod p , k mod q ) .
Indeed the minus sign in Mayer–Vietoris cancels the reversal of the normal circle. The surface and meridian variables lie in separate summands. The surface cokernel is coker A f \operatorname{coker} A_f coker A f , while
( ( Z / p ) ⊕ ( Z / q ) ) / ⟨ ( 1 , 1 ) ⟩ ≅ Z / gcd ( p , q ) . ((\mathbb{Z}/p)\oplus(\mathbb{Z}/q))/\langle(1,1)\rangle\cong\mathbb{Z}/\gcd(p,q). (( Z / p ) ⊕ ( Z / q )) / ⟨( 1 , 1 )⟩ ≅ Z / g cd( p , q ) .
This proves the direct-sum formula and, in the square nonsingular case, the order formula.
A gluing of the form ( v , e i θ ) ↦ ( f ( v ) , λ ( v ) e − i θ ) (v,e^{i\theta})\mapsto(f(v),\lambda(v)e^{-i\theta}) ( v , e i θ ) ↦ ( f ( v ) , λ ( v ) e − i θ ) , with λ : Σ G → S 1 \lambda:\Sigma_G\to S^1 λ : Σ G → S 1 , can add a right-meridian term to the surface columns through λ ∗ \lambda_* λ ∗ . The direct-sum calculation above therefore uses the stated product normal-circle gluing; it is not a consequence of the surface matrix alone.
The two gluing families
The symmetric (3, 3) case
Both pieces are M 2 , 3 M_{2,3} M 2 , 3 , with base generators ( a , b , c , d ) (a,b,c,d) ( a , b , c , d ) and handle counts ( 2 , 11 , 14 , 3 ) (2,11,14,3) ( 2 , 11 , 14 , 3 ) . Only the a a a face pairing rotates the pair-of-pants fiber.
Put x = x 1 x=x_1 x = x 1 , y = x 2 y=x_2 y = x 2 for the first two peripheral loops. In these coordinates a useful presentation is
π 1 ( M 2 , 3 ) = ⟨ a , b , c , d , x , y ∣ [ a , b ] [ c , d ] = 1 , \pi_1(M_{2,3})=\langle a,b,c,d,x,y\mid[a,b][c,d]=1, π 1 ( M 2 , 3 ) = ⟨ a , b , c , d , x , y ∣ [ a , b ] [ c , d ] = 1 ,
a x a − 1 = y , a y a − 1 = ( x y ) − 1 , axa^{-1}=y,\qquad aya^{-1}=(xy)^{-1}, a x a − 1 = y , a y a − 1 = ( x y ) − 1 ,
[ s , x ] = [ s , y ] = 1 ( s = b , c , d ) ⟩ . [s,x]=[s,y]=1\quad(s=b,c,d)\rangle. [ s , x ] = [ s , y ] = 1 ( s = b , c , d )⟩ .
The third peripheral loop is z = ( x y ) − 1 z=(xy)^{-1} z = ( x y ) − 1 . This presentation follows from the handle attachments. In the disk-and-band basis the same monodromy is y 1 ↦ y 2 y 1 − 1 y_1 \mapsto y_2y_1^{-1} y 1 ↦ y 2 y 1 − 1 , y 2 ↦ y 1 − 1 y_2 \mapsto y_1^{-1} y 2 ↦ y 1 − 1 .
Abbreviate C r = C 1 , r C_r=C_{1,r} C r = C 1 , r and D r = D 1 , r D_r=D_{1,r} D r = D 1 , r . The marked genus-four covering surface has pairs
( E , B ) (E,B) ( E , B ) , ( C 0 , D 0 ) (C_0,D_0) ( C 0 , D 0 ) , ( C 1 , D 1 ) (C_1,D_1) ( C 1 , D 1 ) , ( C 2 , D 2 ) (C_2,D_2) ( C 2 , D 2 ) ,
and surface relation [ E , B ] [ C 0 , D 0 ] [ C 1 , D 1 ] [ C 2 , D 2 ] = 1 [E,B][C_0,D_0][C_1,D_1][C_2,D_2]=1 [ E , B ] [ C 0 , D 0 ] [ C 1 , D 1 ] [ C 2 , D 2 ] = 1 . Thus ( A 0 , B 0 ) = ( E , B ) (A_0,B_0)=(E,B) ( A 0 , B 0 ) = ( E , B ) and ( A r + 1 , B r + 1 ) = ( C r , D r ) (A_{r+1},B_{r+1})=(C_r,D_r) ( A r + 1 , B r + 1 ) = ( C r , D r ) for 0 ≤ r < 3 0\le r<3 0 ≤ r < 3 . With the based paths from the covering polygon, its boundary inclusion is specified by the following words:
Table 1 (PDF p. 19).
The meridian maps to x x x . All these words commute with x x x because their monodromy is trivial. Thus the inclusion respects the boundary relation [ π 1 ( Σ 4 ) , x ] = 1 [\pi_1(\Sigma_4),x]=1 [ π 1 ( Σ 4 ) , x ] = 1 .
In the homology order ( e 0 , f 0 , e 1 , f 1 , e 2 , f 2 , e 3 , f 3 ) (e_0,f_0,e_1,f_1,e_2,f_2,e_3,f_3) ( e 0 , f 0 , e 1 , f 1 , e 2 , f 2 , e 3 , f 3 ) , the projection matrix is
Q 3 , 2 = ( 3 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 1 0 1 0 0 0 0 1 0 1 0 1 ) . Q_{3,2}=\begin{pmatrix}3&0&0&0&0&0&0&0\\0&1&0&0&0&0&0&0\\0&0&1&0&1&0&1&0\\0&0&0&1&0&1&0&1\end{pmatrix}. Q 3 , 2 = 3 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 1 .
Use the same construction and marking for the second piece. To obtain a presentation of the fundamental group of the symplectic sum, combine the two piece presentations, identify the eight surface boundary generators by f m f_m f m , defined below, and identify the meridians inversely.
The chosen genus-four gluing. Use the following torus permutation and blocks from [8 Theorem 4.5] :
Table 2 (PDF p. 19).
For the permutation perform exchanges with indices
0 , 2 , 1 , 2 0,\ 2,\ 1,\ 2 0 , 2 , 1 , 2 in this order.
The ordered torus labels change as
( 0 , 1 , 2 , 3 ) ⟶ ( 1 , 0 , 2 , 3 ) ⟶ ( 1 , 0 , 3 , 2 ) (0,1,2,3)\longrightarrow(1,0,2,3)\longrightarrow(1,0,3,2) ( 0 , 1 , 2 , 3 ) ⟶ ( 1 , 0 , 2 , 3 ) ⟶ ( 1 , 0 , 3 , 2 )
⟶ ( 1 , 3 , 0 , 2 ) ⟶ ( 1 , 3 , 2 , 0 ) \longrightarrow(1,3,0,2)\longrightarrow(1,3,2,0) ⟶ ( 1 , 3 , 0 , 2 ) ⟶ ( 1 , 3 , 2 , 0 ) .
The source-to-target permutation is π = ( 3 , 0 , 2 , 1 ) \pi=(3,0,2,1) π = ( 3 , 0 , 2 , 1 ) . The based generator actions of the twists and exchanges in Section 3 therefore specify f m f_m f m on the genus-four surface group, not only on homology. For m = 1 m=1 m = 1 , all eight images under f 1 − 1 f_1^{-1} f 1 − 1 are displayed in equation (3) of Appendix § 7 .
The 8 × 8 8\times8 8 × 8 matrix S m S_m S m has block B j B_j B j in row-pair π ( j ) \pi(j) π ( j ) and column-pair j j j , and zeros elsewhere. The surface Mayer-Vietoris matrix, with left base rows followed by right base rows and the boundary basis as columns, is
A m = ( Q 3 , 2 − Q 3 , 2 S m ) = ( 3 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 1 0 1 0 0 0 0 1 0 1 0 1 0 0 − 3 − 3 0 0 0 0 0 0 0 − 1 0 0 0 0 − 1 − 1 0 0 − 1 1 − 1 1 − m 0 − 1 0 0 0 − 1 − 1 − m ) . A_m=\begin{pmatrix} Q_{3,2} \\ -Q_{3,2}S_m \end{pmatrix}=\begin{pmatrix} 3 & 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 1 & 0 & 1 & 0 & 1 & 0 \\ 0 & 0 & 0 & 1 & 0 & 1 & 0 & 1 \\ 0 & 0 & -3 & -3 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & -1 & 0 & 0 & 0 & 0 \\ -1 & -1 & 0 & 0 & -1 & 1 & -1 & 1-m \\ 0 & -1 & 0 & 0 & 0 & -1 & -1 & -m \end{pmatrix}. A m = ( Q 3 , 2 − Q 3 , 2 S m ) = 3 0 0 0 0 0 − 1 0 0 1 0 0 0 0 − 1 − 1 0 0 1 0 − 3 0 0 0 0 0 0 1 − 3 − 1 0 0 0 0 1 0 0 0 − 1 0 0 0 0 1 0 0 1 − 1 0 0 1 0 0 0 − 1 − 1 0 0 0 1 0 0 1 − m − m .
Its determinant is − 9 m -9m − 9 m . For the specified boundary gluing, the meridian relations give an additional direct summand Z / 3 \mathbb{Z}/3 Z /3 .
Proposition 4.1. For every m ≥ 1 m \ge1 m ≥ 1 , H 1 ( Z m 3 , 3 ; Z ) ≅ ( Z / 3 ) 2 ⊕ Z / ( 3 m ) H_1(Z_m^{3,3};\mathbb{Z}) \cong(\mathbb{Z}/3)^2 \oplus\mathbb{Z}/(3m) H 1 ( Z m 3 , 3 ; Z ) ≅ ( Z /3 ) 2 ⊕ Z / ( 3 m ) .
Proof. Let a , b , c , d a,b,c,d a , b , c , d denote the left base homology generators and A , B , C , D A,B,C,D A , B , C , D the right generators. The columns of A m A_m A m give the relations
3 a − C = 0 , b − C − D = 0 , c − 3 A = 0 , d − 3 A − B = 0 , c − C = 0 , d + C − D = 0 , c − C − D = 0 , d + ( 1 − m ) C − m D = 0. \begin{aligned}
3a-C&=0, & b-C-D&=0,\\
c-3A&=0, & d-3A-B&=0,\\
c-C&=0, & d+C-D&=0,\\
c-C-D&=0, & d+(1-m)C-mD&=0.
\end{aligned} 3 a − C c − 3 A c − C c − C − D = 0 , = 0 , = 0 , = 0 , b − C − D d − 3 A − B d + C − D d + ( 1 − m ) C − m D = 0 , = 0 , = 0 , = 0.
Subtracting the fifth relation from the seventh gives D = 0 D=0 D = 0 . Consequently
c = C = 3 A = 3 a , d = − C , b = C , B = − 6 A . c=C=3A=3a,\qquad d=-C,\qquad b=C,\qquad B=-6A. c = C = 3 A = 3 a , d = − C , b = C , B = − 6 A .
The last relation gives m C = 0 mC=0 m C = 0 . Eliminating generators by these coefficient-one relations leaves
3 ( a − A ) = 0 , 3 m A = 0. 3(a-A)=0,\qquad3mA=0. 3 ( a − A ) = 0 , 3 m A = 0.
Let t , s t,s t , s be the left and right meridian classes. They satisfy 3 t = 3 s = 0 3t=3s=0 3 t = 3 s = 0 and t + s = 0 t+s=0 t + s = 0 ; hence one independent meridian class of order three remains. Therefore
H 1 ( Z m 3 , 3 ; Z ) ≅ ( Z / 3 ) 2 ⊕ Z / ( 3 m ) . H_1(Z_m^{3,3};\mathbb{Z})\cong(\mathbb{Z}/3)^2\oplus\mathbb{Z}/(3m). H 1 ( Z m 3 , 3 ; Z ) ≅ ( Z /3 ) 2 ⊕ Z / ( 3 m ) .
This agrees with [8 ] for the surface gluing chosen here. For m = 1 , 2 m=1,2 m = 1 , 2 , we obtain
H 1 ( Z 1 3 , 3 ; Z ) = ( Z / 3 ) 3 , H 1 ( Z 2 3 , 3 ; Z ) = ( Z / 3 ) 2 ⊕ Z / 6. H_1(Z_1^{3,3};\mathbb{Z})=(\mathbb{Z}/3)^3,\qquad H_1(Z_2^{3,3};\mathbb{Z})=(\mathbb{Z}/3)^2\oplus\mathbb{Z}/6. H 1 ( Z 1 3 , 3 ; Z ) = ( Z /3 ) 3 , H 1 ( Z 2 3 , 3 ; Z ) = ( Z /3 ) 2 ⊕ Z /6.
:::
Use the left complement’s absolute handles, insert a collar Σ 4 × S 1 × [ 0 , 1 ] \Sigma_4\times S^1\times[0,1] Σ 4 × S 1 × [ 0 , 1 ] with transition ϕ m \phi_m ϕ m , and attach the right complement upside down relative to its entire boundary. Its index-k k k handles become relative index-( 4 − k ) (4-k) ( 4 − k ) handles.
The unreduced closed counts are ( 2 , 14 , 28 , 14 , 2 ) (2,14,28,14,2) ( 2 , 14 , 28 , 14 , 2 ) , so χ = 4 \chi=4 χ = 4 . For m = 1 , 2 m=1,2 m = 1 , 2 , the first-homology orders 27 and 54 distinguish the diffeomorphism types. The signature and rational cohomology ring are computed in § 5 .
For m = 1 m=1 m = 1 , the right complement also admits a relative product presentation with three 1-handles, fourteen 2-handles, eleven 3-handles, and two 4-handles; its relative attachment charts are in Appendix § 7 . This is an alternative to the upside-down absolute presentation used for the counts above.
The mixed (2, 4) case
The pieces M 3 , 2 M_{3,2} M 3 , 2 and M 2 , 4 M_{2,4} M 2 , 4 are respectively annulus and four-holed-sphere bundles over Σ 3 \Sigma_3 Σ 3 and Σ 2 \Sigma_2 Σ 2 ; their common boundary surface has genus five.
Let μ \mu μ be the positively oriented peripheral loop with its chosen base path in the annulus; it also generates its fundamental group. With base generators ( a , b , c , d , e , f ) (a,b,c,d,e,f) ( a , b , c , d , e , f ) , the left group is
π 1 ( M 3 , 2 ) = ⟨ a , b , c , d , e , f , μ ∣ [ a , b ] [ c , d ] [ e , f ] = 1 , \pi_1(M_{3,2})=\langle a,b,c,d,e,f,\mu\mid[a,b][c,d][e,f]=1, π 1 ( M 3 , 2 ) = ⟨ a , b , c , d , e , f , μ ∣ [ a , b ] [ c , d ] [ e , f ] = 1 ,
a μ a − 1 = μ − 1 , [ s , μ ] = 1 ( s = b , c , d , e , f ) ⟩ . a\mu a^{-1}=\mu^{-1},\quad[s,\mu]=1\ (s=b,c,d,e,f)\rangle. a μ a − 1 = μ − 1 , [ s , μ ] = 1 ( s = b , c , d , e , f )⟩ .
For the right base ( A , B , C , D ) (A,B,C,D) ( A , B , C , D ) use x 1 , x 2 , x 3 , x 4 x_1,x_2,x_3,x_4 x 1 , x 2 , x 3 , x 4 with x 1 x 2 x 3 x 4 = 1 x_1x_2x_3x_4=1 x 1 x 2 x 3 x 4 = 1 . Its group is
π 1 ( M 2 , 4 ) = ⟨ A , B , C , D , x 1 , … , x 4 ∣ [ A , B ] [ C , D ] = 1 , x 1 x 2 x 3 x 4 = 1 , \pi_1(M_{2,4})=\langle A,B,C,D,x_1,\ldots,x_4\mid[A,B][C,D]=1,\quad x_1x_2x_3x_4=1, π 1 ( M 2 , 4 ) = ⟨ A , B , C , D , x 1 , … , x 4 ∣ [ A , B ] [ C , D ] = 1 , x 1 x 2 x 3 x 4 = 1 ,
A x j A − 1 = x j + 1 , [ s , x j ] = 1 ( s = B , C , D ) ⟩ , Ax_jA^{-1}=x_{j+1},\quad[s,x_j]=1\ (s=B,C,D)\rangle, A x j A − 1 = x j + 1 , [ s , x j ] = 1 ( s = B , C , D )⟩ ,
where x 5 = x 1 x_5=x_1 x 5 = x 1 . Eliminating x 4 x_4 x 4 gives seven generators. The two pieces have unreduced counts
( 2 , 14 , 14 , 2 ) , ( 2 , 12 , 18 , 4 ) . (2,14,14,2),\qquad(2,12,18,4). ( 2 , 14 , 14 , 2 ) , ( 2 , 12 , 18 , 4 ) .
The two genus-five boundary markings. For the left covering surface use the pair order
0 = ( E , B 0 ) , 1 = ( C 0 , D 0 ) , 2 = ( C 1 , D 1 ) , 0=(E,B_0),\quad1=(C_0,D_0),\quad2=(C_1,D_1), 0 = ( E , B 0 ) , 1 = ( C 0 , D 0 ) , 2 = ( C 1 , D 1 ) ,
3 = ( E 0 , F 0 ) , 4 = ( E 1 , F 1 ) . 3=(E_0,F_0),\quad4=(E_1,F_1). 3 = ( E 0 , F 0 ) , 4 = ( E 1 , F 1 ) .
Here ( E , B 0 ) (E,B_0) ( E , B 0 ) is the initial covering pair; ( C r , D r ) (C_r,D_r) ( C r , D r ) and ( E r , F r ) (E_r,F_r) ( E r , F r ) are the lifts of the second and third base pairs, with the base-path correction given below. Put W = [ a c a − 1 , a d a − 1 ] W=[aca^{-1},ada^{-1}] W = [ a c a − 1 , a d a − 1 ] . The images in the left complement are:
Table 1 (PDF p. 21).
The original sheet-first closing product is
[ a 2 , b ] [ c , d ] [ e , f ] W [ a e a − 1 , a f a − 1 ] . [a^2,b][c,d][e,f]W[aea^{-1},afa^{-1}]. [ a 2 , b ] [ c , d ] [ e , f ] W [ a e a − 1 , a f a − 1 ] .
In the displayed order its middle factors are instead [ c , d ] W ( W − 1 [ e , f ] W ) [c,d]W(W^{-1}[e,f]W) [ c , d ] W ( W − 1 [ e , f ] W ) , which equal [ c , d ] [ e , f ] W [c,d][e,f]W [ c , d ] [ e , f ] W . Hence the surface relation holds in this order. The base path of pair three winds along W − 1 W^{-1} W − 1 before reaching its original torus; omitting this winding changes the based boundary map even though Q 2 , 3 Q_{2,3} Q 2 , 3 is unchanged.
The right marking has initial pair ( A 4 , B ) (A^4,B) ( A 4 , B ) , followed by ( A r C A − r , A r D A − r ) (A^rCA^{-r},A^rDA^{-r}) ( A r C A − r , A r D A − r ) for r = 0 , 1 , 2 , 3 r=0,1,2,3 r = 0 , 1 , 2 , 3 . The left meridian is the based, positively oriented annulus peripheral loop μ = x 1 \mu=x_1 μ = x 1 , with [ μ ] = − Δ a [\mu]=-\Delta_a [ μ ] = − Δ a , and the right meridian is x 1 x_1 x 1 . The covering polygon specifies the surface marking and its base paths.
In the homology order ( e 0 , f 0 , … , e 4 , f 4 ) (e_0,f_0,\ldots,e_4,f_4) ( e 0 , f 0 , … , e 4 , f 4 ) the projection matrices are
Q 2 , 3 = ( 2 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 1 0 1 ) , Q_{2,3}=
\begin{pmatrix}
2&0&0&0&0&0&0&0&0&0\\
0&1&0&0&0&0&0&0&0&0\\
0&0&1&0&1&0&0&0&0&0\\
0&0&0&1&0&1&0&0&0&0\\
0&0&0&0&0&0&1&0&1&0\\
0&0&0&0&0&0&0&1&0&1
\end{pmatrix}, Q 2 , 3 = 2 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 1 ,
Q 4 , 2 = ( 4 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 1 0 1 0 1 0 0 0 0 1 0 1 0 1 0 1 ) . Q_{4,2}=
\begin{pmatrix}
4&0&0&0&0&0&0&0&0&0\\
0&1&0&0&0&0&0&0&0&0\\
0&0&1&0&1&0&1&0&1&0\\
0&0&0&1&0&1&0&1&0&1
\end{pmatrix}. Q 4 , 2 = 4 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 1 .
The chosen genus-five gluing. Use the following source blocks and destinations from [8 ] :
Perform adjacent exchanges with indices
3 , 2 , 1 , 0 , 1 , 2 3,\ 2,\ 1,\ 0,\ 1,\ 2 3 , 2 , 1 , 0 , 1 , 2
in this order.
The final order is ( 4 , 1 , 2 , 0 , 3 ) (4,1,2,0,3) ( 4 , 1 , 2 , 0 , 3 ) . Thus π = ( 3 , 1 , 2 , 4 , 0 ) \pi=(3,1,2,4,0) π = ( 3 , 1 , 2 , 4 , 0 ) gives the source-to-target permutation.
In this genus-five case, S m S_m S m is the 10 × 10 10 \times10 10 × 10 matrix with block B j B_j B j in row-pair π ( j ) \pi(j) π ( j ) and column-pair j j j . The surface presentation matrix is
A m = ( Q 2 , 3 − Q 4 , 2 S m ) = ( 2 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 − 4 4 0 0 0 0 0 0 0 0 0 − 1 − 1 0 m − 2 1 − 1 0 − 1 0 0 0 − 1 − 1 1 − m − 1 1 − 1 1 − 1 0 0 ) \begin{aligned}
A_m=\begin{pmatrix} Q_{2,3}\\ -Q_{4,2}S_m \end{pmatrix} \\
&=\begin{pmatrix}
2&0&0&0&0&0&0&0&0&0\\
0&1&0&0&0&0&0&0&0&0\\
0&0&1&0&1&0&0&0&0&0\\
0&0&0&1&0&1&0&0&0&0\\
0&0&0&0&0&0&1&0&1&0\\
0&0&0&0&0&0&0&1&0&1\\
0&0&0&0&0&0&0&0&-4&4\\
0&0&0&0&0&0&0&0&0&-1\\
-1&0&m-2&1&-1&0&-1&0&0&0\\
-1&-1&1-m&-1&1&-1&1&-1&0&0
\end{pmatrix}
\end{aligned} A m = ( Q 2 , 3 − Q 4 , 2 S m ) = 2 0 0 0 0 0 0 0 − 1 − 1 0 1 0 0 0 0 0 0 0 − 1 0 0 1 0 0 0 0 0 m − 2 1 − m 0 0 0 1 0 0 0 0 1 − 1 0 0 1 0 0 0 0 0 − 1 1 0 0 0 1 0 0 0 0 0 − 1 0 0 0 0 1 0 0 0 − 1 1 0 0 0 0 0 1 0 0 0 − 1 0 0 0 0 1 0 − 4 0 0 0 0 0 0 0 0 1 4 − 1 0 0
Its determinant is 8 m 8m 8 m . For this gluing, the meridian relations give an additional direct summand Z / 2 \mathbb{Z}/2 Z /2 . Both the sign of the normal circle and the pair convention are the same as in the symmetric case.
Proposition 4.2. For every m ≥ 1 m \ge1 m ≥ 1 , H 1 ( Z m 2 , 4 ; Z ) ≅ ( Z / 2 ) 2 ⊕ Z / ( 4 m ) H_1(Z_m^{2,4};\mathbb{Z}) \cong(\mathbb{Z}/2)^2 \oplus\mathbb{Z}/(4m) H 1 ( Z m 2 , 4 ; Z ) ≅ ( Z /2 ) 2 ⊕ Z / ( 4 m ) .
Proof. Let a , b , c , d , e , f a,b,c,d,e,f a , b , c , d , e , f be the left base homology generators and A , B , C , D A,B,C,D A , B , C , D the right ones. The ten columns of the preceding matrix give
2 a − C − D = 0 , b − D = 0 , c + ( m − 2 ) C + ( 1 − m ) D = 0 , d + C − D = 0 , c − C + D = 0 , d − D = 0 , e − C + D = 0 , f − D = 0 , e − 4 A = 0 , f + 4 A − B = 0. \begin{aligned}
2a-C-D&=0, & b-D&=0,\\
c+(m-2)C+(1-m)D&=0, & d+C-D&=0,\\
c-C+D&=0, & d-D&=0,\\
e-C+D&=0, & f-D&=0,\\
e-4A&=0, & f+4A-B&=0.
\end{aligned} 2 a − C − D c + ( m − 2 ) C + ( 1 − m ) D c − C + D e − C + D e − 4 A = 0 , = 0 , = 0 , = 0 , = 0 , b − D d + C − D d − D f − D f + 4 A − B = 0 , = 0 , = 0 , = 0 , = 0. The fourth minus the sixth relation gives C = 0 C=0 C = 0 . Then
b = d = f = D , c = e = − D , D = − 4 A , B = 0. b=d=f=D,\qquad c=e=-D,\qquad D=-4A,\qquad B=0. b = d = f = D , c = e = − D , D = − 4 A , B = 0. The third relation gives m D = 0 mD=0 m D = 0 and the first gives 2 a = D 2a=D 2 a = D . Put w = a + 2 A w=a+2A w = a + 2 A ; the remaining surface relations are precisely
2 w = 0 , 4 m A = 0. 2w=0,\qquad4mA=0. 2 w = 0 , 4 m A = 0. For the left and right meridian classes t , s t,s t , s , the relations 2 t = 0 2t=0 2 t = 0 , 4 s = 0 4s=0 4 s = 0 , t + s = 0 t+s=0 t + s = 0 leave one independent order-two class. Hence
H 1 ( Z m 2 , 4 ; Z ) ≅ ( Z / 2 ) 2 ⊕ Z / ( 4 m ) , H_1(Z_m^{2,4};\mathbb{Z})\cong(\mathbb{Z}/2)^2\oplus\mathbb{Z}/(4m), H 1 ( Z m 2 , 4 ; Z ) ≅ ( Z /2 ) 2 ⊕ Z / ( 4 m ) ,
and specifically
H 1 ( Z 1 2 , 4 ; Z ) = ( Z / 2 ) 2 ⊕ Z / 4 , H 1 ( Z 2 2 , 4 ; Z ) = ( Z / 2 ) 2 ⊕ Z / 8. H_1(Z_1^{2,4};\mathbb{Z})=(\mathbb{Z}/2)^2\oplus\mathbb{Z}/4,\qquad H_1(Z_2^{2,4};\mathbb{Z})=(\mathbb{Z}/2)^2\oplus\mathbb{Z}/8. H 1 ( Z 1 2 , 4 ; Z ) = ( Z /2 ) 2 ⊕ Z /4 , H 1 ( Z 2 2 , 4 ; Z ) = ( Z /2 ) 2 ⊕ Z /8.
Use the same collar construction with Σ 5 \Sigma_5 Σ 5 in place of Σ 4 \Sigma_4 Σ 4 , M 3 , 2 M_{3,2} M 3 , 2 on the left, and M 2 , 4 M_{2,4} M 2 , 4 on the right. Its unreduced closed handle counts are
( h 0 , h 1 , h 2 , h 3 , h 4 ) = ( 2 , 18 , 32 , 14 , 2 ) , 2 − 18 + 32 − 14 + 2 = 4. (h_0,h_1,h_2,h_3,h_4)=(2,18,32,14,2),\qquad2-18+32-14+2=4. ( h 0 , h 1 , h 2 , h 3 , h 4 ) = ( 2 , 18 , 32 , 14 , 2 ) , 2 − 18 + 32 − 14 + 2 = 4.
For m = 1 , 2 m=1,2 m = 1 , 2 , the first-homology orders are 16 and 32.
Topology of the symplectic sums
Write Z m Z_m Z m for either family Z m 3 , 3 Z_m^{3,3} Z m 3 , 3 or Z m 2 , 4 Z_m^{2,4} Z m 2 , 4 .
Proposition 5.1. Let M L = M g , p M_L=M_{g,p} M L = M g , p and M R = M h , q M_R=M_{h,q} M R = M h , q , where g , h , p , q ≥ 2 g,h,p,q\ge2 g , h , p , q ≥ 2 and their boundary surfaces have the same genus. Gluing these pieces by any orientation-reversing boundary diffeomorphism gives an aspherical closed manifold with infinite fundamental group. In particular, this holds for both families Z m 3 , 3 Z_m^{3,3} Z m 3 , 3 and Z m 2 , 4 Z_m^{2,4} Z m 2 , 4 , for every m ≥ 1 m\ge1 m ≥ 1 .
Proof. The cyclic cover of each complement associated to ker χ p \ker\chi_p ker χ p is Σ G × P p \Sigma_G\times P_p Σ G × P p , where G = 1 + p ( g − 1 ) ≥ 2 G=1+p(g-1)\ge2 G = 1 + p ( g − 1 ) ≥ 2 . Both factors are aspherical. Each boundary component of this cover is Σ G × S 1 \Sigma_G\times S^1 Σ G × S 1 , and its fundamental group injects because a peripheral circle of P p P_p P p has infinite order. Boundary injectivity descends to M g , p M_{g,p} M g , p , recovering the incompressibility result proved in [6 ] . The chosen boundary diffeomorphism gives an amalgamated product with injective edge maps,
π 1 ( Z m ) = π 1 ( M L ) ∗ π 1 ( Σ G ) × Z π 1 ( M R ) . \pi_1(Z_m)=\pi_1(M_L)*_{\pi_1(\Sigma_G)\times\mathbb{Z}}\pi_1(M_R). π 1 ( Z m ) = π 1 ( M L ) ∗ π 1 ( Σ G ) × Z π 1 ( M R ) .
The universal cover is a tree of universal covers of the pieces and boundary collars, over the Bass-Serre tree of the amalgam. Its vertex and edge spaces are contractible. The lifted boundary collars make the edge inclusions cofibrations. The union over a finite subtree is contractible by induction: adjoining a terminal vertex gives a pushout of contractible spaces along a contractible cofibration, hence a homotopy pushout with the homotopy type of a point.
Choose a finite triangulation of the closed manifold compatible with the splitting collar. Its lifted triangulation is locally finite, so the image of any sphere map is contained in a finite subcomplex and meets only finitely many lifted pieces and collars. Adjoin the finitely many paths joining their vertices in the tree. The sphere map then lies in the contractible union over that finite subtree and is null-homotopic. Thus the universal cover is contractible by Whitehead’s theorem, and Z m Z_m Z m is aspherical. The amalgam normal-form theorem embeds the infinite boundary group in π 1 ( Z m ) \pi_1(Z_m) π 1 ( Z m ) .
Corollary 5.2. Every symplectic sum in Theorem 1.1 is symplectically minimal.
Proof. In an aspherical manifold every map from S 2 S^2 S 2 is null-homotopic, so every embedded sphere represents zero in integral homology and has self-intersection zero. In particular there is no embedded sphere of square − 1 -1 − 1 .
Since χ ( ∂ M L ) = 0 \chi(\partial M_L)=0 χ ( ∂ M L ) = 0 , additivity gives χ ( Z m ) = χ ( M L ) + χ ( M R ) = 4 \chi(Z_m)=\chi(M_L)+\chi(M_R)=4 χ ( Z m ) = χ ( M L ) + χ ( M R ) = 4 . The removed tubular neighborhood is Σ G × D 2 \Sigma_G\times D^2 Σ G × D 2 and has signature zero. Novikov additivity applied to the ambient product ruled surface gives σ ( M g , p ) = σ ( Σ g × S 2 ) = 0 \sigma(M_{g,p})=\sigma(\Sigma_g\times S^2)=0 σ ( M g , p ) = σ ( Σ g × S 2 ) = 0 . A second application across the gluing boundary gives σ ( Z m ) = 0 \sigma(Z_m)=0 σ ( Z m ) = 0 .
Corollary 5.3. For either family and every m ≥ 1 m\ge1 m ≥ 1 , put T = H 1 ( Z m ; Z ) T=H_1(Z_m;\mathbb{Z}) T = H 1 ( Z m ; Z ) . Then
H 0 ( Z m ; Z ) = H 4 ( Z m ; Z ) = Z , H 1 ( Z m ; Z ) = T , H_0(Z_m;\mathbb{Z})=H_4(Z_m;\mathbb{Z})=\mathbb{Z},\qquad H_1(Z_m;\mathbb{Z})=T, H 0 ( Z m ; Z ) = H 4 ( Z m ; Z ) = Z , H 1 ( Z m ; Z ) = T , H 2 ( Z m ; Z ) ≅ Z 2 ⊕ T , H 3 ( Z m ; Z ) = 0. H_2(Z_m;\mathbb{Z})\cong\mathbb{Z}^2\oplus T,\qquad H_3(Z_m;\mathbb{Z})=0. H 2 ( Z m ; Z ) ≅ Z 2 ⊕ T , H 3 ( Z m ; Z ) = 0.
The rational cohomology ring is that of S 2 × S 2 S^2\times S^2 S 2 × S 2 .
Proof. Since T = H 1 ( Z m ; Z ) T = H_{1}(Z_{m};\mathbb{Z}) T = H 1 ( Z m ; Z ) is finite, integral Poincare duality and the universal coefficient theorem give
H 3 ( Z m ; Z ) ≅ H 1 ( Z m ; Z ) = Hom ( T , Z ) = 0. H_{3}(Z_{m};\mathbb{Z}) \cong H^{1}(Z_{m};\mathbb{Z}) = \operatorname{Hom}(T,\mathbb{Z}) = 0. H 3 ( Z m ; Z ) ≅ H 1 ( Z m ; Z ) = Hom ( T , Z ) = 0.
The Euler characteristic is four, so the free rank of H 2 H_{2} H 2 is two. The universal coefficient theorem gives Tor H 2 ( Z m ; Z ) = Ext ( T , Z ) ≅ T \operatorname{Tor} H^{2}(Z_{m};\mathbb{Z}) = \operatorname{Ext}(T,\mathbb{Z}) \cong T Tor H 2 ( Z m ; Z ) = Ext ( T , Z ) ≅ T , noncanonically. Poincare duality therefore gives H 2 ( Z m ; Z ) ≅ Z 2 ⊕ T H_{2}(Z_{m};\mathbb{Z}) \cong\mathbb{Z}^{2} \oplus T H 2 ( Z m ; Z ) ≅ Z 2 ⊕ T . Connectedness and orientation give H 0 = H 4 = Z H_{0} = H_{4} = \mathbb{Z} H 0 = H 4 = Z .
The free intersection pairing is unimodular of rank two and signature zero, hence has determinant − 1 -1 − 1 . It is hyperbolic over Q \mathbb{Q} Q : for a matrix ( a b b c ) \begin{pmatrix} a & b \\ b & c \end{pmatrix} ( a b b c ) , the equality b 2 − a c = 1 b^{2} - ac = 1 b 2 − a c = 1 gives a nonzero rational isotropic vector u u u . Choose v v v with u ⋅ v = 1 u \cdot v = 1 u ⋅ v = 1 , and replace v v v by v − 1 2 ( v ⋅ v ) u v - \frac{1}{2}(v \cdot v)u v − 2 1 ( v ⋅ v ) u . The resulting basis has matrix ( 0 1 1 0 ) \begin{pmatrix} 0 & 1 \\ 1 & 0 \end{pmatrix} ( 0 1 1 0 ) . Together with H 1 ( Z m ; Q ) = H 3 ( Z m ; Q ) = 0 H^{1}(Z_{m};\mathbb{Q}) = H^{3}(Z_{m};\mathbb{Q}) = 0 H 1 ( Z m ; Q ) = H 3 ( Z m ; Q ) = 0 , this identifies the graded rational cohomology ring with that of S 2 × S 2 S^{2} \times S^{2} S 2 × S 2 . □
We do not determine whether these sums admit complex structures.
Appendix B describes the relative product attachments for the symmetric sum. The smooth sums obtained from other Torelli lifts have not been compared.
Handlebodies of M 3 , 2 M_{3,2} M 3 , 2 and M 2 , 4 M_{2,4} M 2 , 4
The complement M 3 , 2 M_{3,2} M 3 , 2
Use the genus-three base with six bands and two fiber bands interchanged by the monodromy. In the figures, ( a 1 , b 1 , a 2 , b 2 , a 3 , b 3 ) = ( a , b , c , d , e , f ) (a_{1}, b_{1}, a_{2}, b_{2}, a_{3}, b_{3}) = (a, b, c, d, e, f) ( a 1 , b 1 , a 2 , b 2 , a 3 , b 3 ) = ( a , b , c , d , e , f ) . There are twenty-four corner charts at each pole, twelve product 2-handles, two closing 2-handles, and two 3-handles. The chart conventions and normal frames are those of Section 2.
The complement M 2 , 4 M_{2,4} M 2 , 4
Use the genus-two base with four bands and four fiber bands permuted cyclically by the monodromy. All figures use ( a , b , c , d ) (a, b, c, d) ( a , b , c , d ) for the ordered base generators ( a 1 , b 1 , a 2 , b 2 ) (a_{1}, b_{1}, a_{2}, b_{2}) ( a 1 , b 1 , a 2 , b 2 ) ; these correspond to ( A , B , C , D ) (A, B, C, D) ( A , B , C , D ) in Section 4. There are thirty-two corner charts at each pole, sixteen product 2-handles, two closing 2-handles, and four 3-handles.
Relative attachments for the symmetric sum
We describe the right-hand M 2 , 3 M_{2,3} M 2 , 3 relative to its entire boundary and pull its attachments back by ϕ 1 − 1 \phi_{1}^{-1} ϕ 1 − 1 to form Z 1 3 , 3 Z_{1}^{3,3} Z 1 3 , 3 . Unlike this relative product construction, the dual decomposition recovers the original 2-handle belt circles only after adding the three 1-handles dual to the original 3-handles; those circles need not survive intact in the final boundary.
The relative fiber. Use longitude θ \theta θ and latitude λ \lambda λ on the unit sphere. Choose a longitude origin α 0 \alpha_{0} α 0 (α 0 = π / 3 \alpha_{0} = \pi/3 α 0 = π /3 in the figure) and put α j = α 0 + 2 π j / 3 \alpha_{j} = \alpha_{0} + 2\pi j/3 α j = α 0 + 2 π j /3 , with α 3 = α 0 + 2 π \alpha_{3} = \alpha_{0} + 2\pi α 3 = α 0 + 2 π . Take geodesic disks of radius 0 < ε < π / 6 0 < \varepsilon< \pi/6 0 < ε < π /6 about the equatorial points z j = ( cos α j , sin α j , 0 ) z_{j} = (\cos\alpha_{j}, \sin\alpha_{j}, 0) z j = ( cos α j , sin α j , 0 ) of the sphere. Its framed boundary circles, parametrized by the meridional angle ψ ∈ R / 2 π Z \psi\in\mathbb{R}/2\pi\mathbb{Z} ψ ∈ R /2 π Z , are
w j ( ψ ) = cos ε z j + sin ε ( cos ψ τ j + sin ψ n ) , w_{j}(\psi) = \cos\varepsilon\, z_{j} + \sin\varepsilon\,(\cos\psi\,\tau_{j} + \sin\psi\,n), w j ( ψ ) = cos ε z j + sin ε ( cos ψ τ j + sin ψ n ) ,
where τ j = ( − sin α j , cos α j , 0 ) \tau_{j} = (-\sin\alpha_{j}, \cos\alpha_{j}, 0) τ j = ( − sin α j , cos α j , 0 ) and n = ( 0 , 0 , 1 ) n = (0,0,1) n = ( 0 , 0 , 1 ) . Put δ = ε / 2 \delta= \varepsilon/2 δ = ε /2 and Δ ( λ ) = arccos ( cos ε / cos λ ) \Delta(\lambda) = \arccos(\cos\varepsilon/\cos\lambda) Δ ( λ ) = arccos ( cos ε / cos λ ) . The invariant band between holes j j j and j + 1 j+1 j + 1 is
c j ( u , v ) : λ = δ u , θ = α j + α j + 1 2 + v ( π 3 − Δ ( δ u ) ) , − 1 ≤ u , v ≤ 1. c_{j}(u,v):\quad\lambda= \delta u,\qquad\theta= \frac{\alpha_{j} + \alpha_{j+1}}{2} + v\left(\frac{\pi}{3} - \Delta(\delta u)\right),\qquad-1 \leq u,v \leq1. c j ( u , v ) : λ = δ u , θ = 2 α j + α j + 1 + v ( 3 π − Δ ( δ u ) ) , − 1 ≤ u , v ≤ 1.
Increasing u u u runs from south to north. The absolute band core, oriented from north to south in Section 2, is therefore traversed in decreasing u u u , with tangent − ∂ u -\partial_{u} − ∂ u . This constant reversal introduces no winding of the product normal frame. The sides v = − 1 v=-1 v = − 1 and v = 1 v=1 v = 1 lie on cuffs j j j and j + 1 j+1 j + 1 , respectively: their equations are cos λ cos ( θ − α j ) = cos ε \cos\lambda\cos(\theta- \alpha_{j}) = \cos\varepsilon cos λ cos ( θ − α j ) = cos ε and cos λ cos ( θ − α j + 1 ) = cos ε \cos\lambda\cos(\theta- \alpha_{j+1}) = \cos\varepsilon cos λ cos ( θ − α j + 1 ) = cos ε . The cocore arc of this surface 1-handle is
Figure 13. M 3 , 2 M_{3,2} M 3 , 2 . Northern and southern projections of the product and closing 2-handle attaching curves. Pair k − k^- k − with k + k^+ k + at each pole and C j C_j C j across poles. Black arcs belong to the closing 2-handles. Figure 15 specifies the strand identifications and product framings.
Figure 15. z k , j N ( s ) = ρ e ( i ( 2 π ( j + δ k ( 1 − s ) ) / p + α ) ) z^{N}_{k,j}(s)=\rho e^{\left(i\left(2\pi\left(j+\delta_k(1-s)\right)/p+\alpha\right)\right)} z k , j N ( s ) = ρ e ( i ( 2 π ( j + δ k ( 1 − s ) ) / p + α ) ) , z S = z N ‾ z^{S}=\overline{z^{N}} z S = z N ; product frames follow these derivatives; crossings use imaginary depth. Strand identifications through the 1-handles, with the normal coordinates inducing their framings.
Figure 14. M 3 , 2 M_{3,2} M 3 , 2 . The two closing 2-handle attaching curves and the product faces of the subsequent 3-handle attaching spheres, with their normal coordinates. These are local product schematics; their colors distinguish the displayed faces.
η j ( v ) = c j ( 0 , v ) \eta_j(v)=c_j(0,v) η j ( v ) = c j ( 0 , v ) , with endpoints w j ( 0 ) w_j(0) w j ( 0 ) and w j + 1 ( π ) w_{j+1}(\pi) w j + 1 ( π ) . The three cocore arcs cut the fiber into north and south disks. In the relative decomposition, these cocore arcs serve as the cores of three relative surface 1-handles E j E_j E j , followed by two relative surface 2-handles. Their attaching curves Θ N \Theta_N Θ N , Θ S \Theta_S Θ S use the two edges of each relative 1-handle and the complementary upper and lower cuff arcs, as drawn below.
In the higher-handle figures, B 2 = D B 2 = D c l o s i n g 2 B_2=D^2_B=D^2_{\mathrm{closing}} B 2 = D B 2 = D closing 2 is the base closing disk, and D N 2 , D S 2 D^2_N,D^2_S D N 2 , D S 2 denote the relative caps with boundaries Θ N , Θ S \Theta_N,\Theta_S Θ N , Θ S . Their polar coordinates are ( r B , θ B ) (r_B,\theta_B) ( r B , θ B ) and ( r c a p , θ c a p ) (r_{\mathrm{cap}},\theta_{\mathrm{cap}}) ( r cap , θ cap ) . The intervals I s , I v I_s,I_v I s , I v follow the indicated core coordinates; e j e_j e j in these figures is the core of E j E_j E j . Multiplying by one base disk, four base bands, and a closing base disk gives the relative handle counts
( h 0 , h 1 , h 2 , h 3 , h 4 ) = ( 0 , 3 , 14 , 11 , 2 ) . (h_0,h_1,h_2,h_3,h_4)=(0,3,14,11,2). ( h 0 , h 1 , h 2 , h 3 , h 4 ) = ( 0 , 3 , 14 , 11 , 2 ) .
The fourteen 2-handles consist of two products of the base 0-handle with the relative fiber caps and twelve products of a base 1-handle with a relative surface 1-handle. The latter have core coordinates ( s , v ) (s,v) ( s , v ) and
Figure 16. M 2 , 4 M_{2,4} M 2 , 4 . Northern and southern projections of the product and closing 2-handle attaching curves. Pair k − k^- k − with k + k^+ k + at each pole and C j C_j C j across poles. Black arcs belong to the closing 2-handles. Figure 18 specifies the strand identifications and product framings.
Figure 17. M 2 , 4 M_{2,4} M 2 , 4 . The two closing 2-handle attaching curves and the product faces of the subsequent 3-handle attaching spheres, with their normal coordinates. These are local product schematics; their colors distinguish the displayed faces.
Figure 18. z k , j N ( s ) = ρ e i ( 2 π ( j + δ k ( 1 − s ) ) / p + α ) , z S = z N ‾ z^{N}_{k,j}(s)=\rho e^{i(2\pi(j+\delta_k(1-s))/p+\alpha)},\ z^{S}=\overline{z^{N}} z k , j N ( s ) = ρ e i ( 2 π ( j + δ k ( 1 − s )) / p + α ) , z S = z N ; product frames follow these derivatives; crossings use imaginary depth. M 2 , 4 M_{2,4} M 2 , 4 . Strand identifications through the 1-handles, with the normal coordinates inducing their framings.
normal coordinates ( t , u ) (t,u) ( t , u ) . The eleven 3-handles are the three products of the base 2-handle with relative surface 1-handles and eight base-band/cap products. Their attaching spheres are the rounded boundaries of the product cores; the remaining interval is the normal coordinate. The two closing-disk/cap products are the final 4-handles.
As a check on these relative counts, Poincaré–Lefschetz duality and Theorem 2.3 give
H 4 ( M 2 , 3 , ∂ M 2 , 3 ; Z ) = Z , H 3 ( M 2 , 3 , ∂ M 2 , 3 ; Z ) = Z 4 , H_{4}(M_{2,3},\partial M_{2,3};\mathbb{Z})=\mathbb{Z},\qquad H_{3}(M_{2,3},\partial M_{2,3};\mathbb{Z})=\mathbb{Z}^{4}, H 4 ( M 2 , 3 , ∂ M 2 , 3 ; Z ) = Z , H 3 ( M 2 , 3 , ∂ M 2 , 3 ; Z ) = Z 4 ,
H 2 ( M 2 , 3 , ∂ M 2 , 3 ; Z ) ≅ Z 5 ⊕ Z / 3 , H 1 ( M 2 , 3 , ∂ M 2 , 3 ; Z ) = Z / 3 , H_{2}(M_{2,3},\partial M_{2,3};\mathbb{Z})\cong\mathbb{Z}^{5}\oplus\mathbb{Z}/3,\qquad H_{1}(M_{2,3},\partial M_{2,3};\mathbb{Z})=\mathbb{Z}/3, H 2 ( M 2 , 3 , ∂ M 2 , 3 ; Z ) ≅ Z 5 ⊕ Z /3 , H 1 ( M 2 , 3 , ∂ M 2 , 3 ; Z ) = Z /3 ,
with H 0 ( M 2 , 3 , ∂ M 2 , 3 ; Z ) = 0 H_{0}(M_{2,3},\partial M_{2,3};\mathbb{Z})=0 H 0 ( M 2 , 3 , ∂ M 2 , 3 ; Z ) = 0 . Both alternating sums are two.
The surface marking and the six feet. Let v j v_j v j be the three surface lifts above the center of the chosen base disk. Place their small patches in one fixed disk of the planar core of the marked genus-four surface. Include a regular neighborhood of the lifted tree a 0 ∪ a 1 a_0 \cup a_1 a 0 ∪ a 1 in this disk, carrying the covering marking and its base paths by the same surface isotopy. The remaining, essential parts of the lifted graph lie outside the disk. Take the twists and exchanges to be supported outside this disk. The resulting surface diffeomorphism induces the gluing map on first homology specified in [8 ] . The six relative 1-handle feet are
( v j , ψ = 0 ) , ( v j + 1 , ψ = π ) , j m o d 3. (v_j,\psi=0), \qquad(v_{j+1},\psi=\pi), \qquad j\ \mathrm{mod}\ 3. ( v j , ψ = 0 ) , ( v j + 1 , ψ = π ) , j mod 3.
Write V r V_r V r for the vertex on sheet r ∈ Z / 3 r \in\mathbb{Z}/3 r ∈ Z /3 of the lifted base polygon. The feet are ( V − j , 0 ) (V_{-j},0) ( V − j , 0 ) and ( V − ( j + 1 ) , π ) (V_{-(j+1)},\pi) ( V − ( j + 1 ) , π ) . The inverse gluing fixes their surface patches and reverses ψ \psi ψ . Thus it fixes each foot center and reflects its latitude coordinate in the foot chart ( x , y , u ) (x,y,u) ( x , y , u ) , where x , y x,y x , y are local coordinates on the covering surface and u u u is the transverse fiber coordinate. Extend this reflection across each relative 1-handle by ( v , x , y , u ) ↦ ( v , x , y , − u ) (v,x,y,u) \mapsto(v,x,y,-u) ( v , x , y , u ) ↦ ( v , x , y , − u ) . Thus the core coordinate v v v and base-disk coordinates ( x , y ) (x,y) ( x , y ) are fixed. The cap ribbons in Figure 19 are then interchanged by the circle reversal; their base-disk normal frames are preserved.
Figure 19. The relative fiber in longitude and latitude coordinates, with left and right edges paired. The colored arcs are cocores of the absolute fiber bands and cores of the relative 1-handles. The purple curves attach the two relative surface 2-handles. Their products with the base 0-handle give four-dimensional 2-handles, with the base disk as normal factor. The longitude origin is chosen so that all three deleted disks lie inside the chart.
The twelve relative product 2-handles for m = 1 m=1 m = 1 . Cuff indices decrease along the positively oriented a a a 1-handle. Use Schreier sheet index r = − j m o d 3 r=-j\ \mathrm{mod}\ 3 r = − j mod 3 on the covering surface: the positive a a a band sends j j j to j − 1 j-1 j − 1 and r r r to r + 1 r+1 r + 1 . The cuff patches and Schreier vertices satisfy v j = V − j v_j=V_{-j} v j = V − j . For k ∈ { a , b , c , d } k \in\{a,b,c,d\} k ∈ { a , b , c , d } , let k r k_r k r be its lift starting at V r V_r V r . In particular, a r a_r a r runs from V r V_r V r to V r + 1 V_{r+1} V r + 1 . Choose a 0 , a 1 a_0,a_1 a 0 , a 1 as a tree in the fixed cap. The lifted graph, with its endpoints fixed, has the following based edge labels:
h i = [ A i , B i ] h_i=[A_i,B_i] h i = [ A i , B i ] .
Figure 20. The attaching spheres of the eight relative 3-handles obtained from base bands and fiber caps. Each core is D 1 × D 2 D^1 \times D^2 D 1 × D 2 , with the base-band transverse interval as normal. The faces are identified according to the relative product decomposition. Labels a 1 , b 1 , a 2 , b 2 a_1,b_1,a_2,b_2 a 1 , b 1 , a 2 , b 2 denote a , b , c , d a,b,c,d a , b , c , d .
The entries 1 for tree edges mean that their closed based labels are trivial; the embedded tree edges and endpoints are retained. Let Q k , j Q_{k,j} Q k , j denote the relative product 2-handle obtained from the base band k k k and relative fiber handle E j E_j E j . Here δ ( a ) = 1 \delta(a)=1 δ ( a ) = 1 and δ ( b ) = δ ( c ) = δ ( d ) = 0 \delta(b)=\delta(c)=\delta(d)=0 δ ( b ) = δ ( c ) = δ ( d ) = 0 . Its two sides on the boundary surface are the lifts of k k k joining the relative 1-handle feet on sheets − j -j − j and − ( j + 1 ) -(j+1) − ( j + 1 ) . In Figure 23 , these sides are labeled λ k , j 0 \lambda^0_{k,j} λ k , j 0 and λ k , j π \lambda^\pi_{k,j} λ k , j π , respectively, before transport by the inverse surface gluing. Its sides in the relative 1-handles are E j E_j E j and E j − δ ( k ) E_{j-\delta(k)} E j − δ ( k ) . With endpoints fixed, the attaching path is
Figure 23. The twelve relative product 2-handle charts, showing how the sides attach to the relative 1-handles and the boundary surface. The sides on the boundary surface are transported by the inverse gluing map specified below. These are local product charts. The framing is induced by the transported normal coordinates ( t , u ) (t,u) ( t , u ) .
k − j E j − δ ( k ) ( k − ( j + 1 ) ) − 1 E j − 1 . k_{-j} E_{j-\delta(k)} (k_{-(j+1)})^{-1} E_j^{-1}. k − j E j − δ ( k ) ( k − ( j + 1 ) ) − 1 E j − 1 .
These curves bound the product core squares, with the side curves and their endpoints fixed by the product construction.
With h i = [ A i , B i ] h_i=[A_i,B_i] h i = [ A i , B i ] , the based substitutions for T = f − 1 T=f^{-1} T = f − 1 are
T ( A 0 ) = h 0 A 1 h 0 − 1 , T ( B 0 ) = h 0 B 1 A 1 − 1 h 0 − 1 , T ( A 1 ) = h 0 h 2 A 3 B 3 − 1 h 2 − 1 h 0 − 1 , T ( B 1 ) = h 0 h 2 B 3 h 2 − 1 h 0 − 1 , T ( A 2 ) = h 0 A 2 h 0 − 1 , T ( B 2 ) = h 0 B 2 A 2 h 0 − 1 , T ( A 3 ) = A 0 , T ( B 3 ) = B 0 A 0 − 1 . (3) \begin{aligned}
T(A_0)&=h_0A_1h_0^{-1}, & T(B_0)&=h_0B_1A_1^{-1}h_0^{-1},\\
T(A_1)&=h_0h_2A_3B_3^{-1}h_2^{-1}h_0^{-1}, & T(B_1)&=h_0h_2B_3h_2^{-1}h_0^{-1},\\
T(A_2)&=h_0A_2h_0^{-1}, & T(B_2)&=h_0B_2A_2h_0^{-1},\\
T(A_3)&=A_0, & T(B_3)&=B_0A_0^{-1}. \tag*{(3)}
\end{aligned} T ( A 0 ) T ( A 1 ) T ( A 2 ) T ( A 3 ) = h 0 A 1 h 0 − 1 , = h 0 h 2 A 3 B 3 − 1 h 2 − 1 h 0 − 1 , = h 0 A 2 h 0 − 1 , = A 0 , T ( B 0 ) T ( B 1 ) T ( B 2 ) T ( B 3 ) = h 0 B 1 A 1 − 1 h 0 − 1 , = h 0 h 2 B 3 h 2 − 1 h 0 − 1 , = h 0 B 2 A 2 h 0 − 1 , = B 0 A 0 − 1 . (3)
Figure 21. Relative closing-band attaching spheres. Core D B 2 × I v D^{2}_{B} \times I_{v} D B 2 × I v ; caps are B 2 B_{2} B 2 at the two cuff feet. Side ∂ B 2 × I v \partial B_{2} \times I_{v} ∂ B 2 × I v ; normal u u u . Attach after all relative 2-handles. Relative 4-handle attaching faces. S 3 = ( ∂ D B 2 × D c a p 2 ) ∪ ( D B 2 × ∂ D c a p 2 ) S^{3} = (\partial D^{2}_{B} \times D^{2}_{\mathrm{cap}}) \cup(D^{2}_{B} \times\partial D^{2}_{\mathrm{cap}}) S 3 = ( ∂ D B 2 × D cap 2 ) ∪ ( D B 2 × ∂ D cap 2 ) , glued on T 2 T^{2} T 2 . Attach after the eleven 3-handles; round the product corners. Figure 21: Above: the attaching spheres for the three products of the base surface 2-handle with relative surface 1-handles. Below: the boundaries of the two 4-handle cores, each obtained by joining two solid tori along their common boundary. The labels a 1 , b 1 , a 2 , b 2 a_{1}, b_{1}, a_{2}, b_{2} a 1 , b 1 , a 2 , b 2 denote a , b , c , d a, b, c, d a , b , c , d .
The final pulled-back attaching path is
T ( k − j ) E j − δ ( k ) T ( k − ( j + 1 ) ) − 1 E j − 1 . T(k_{-j})\,E_{j-\delta(k)}\,T(k_{-(j+1)})^{-1}\,E_{j}^{-1}. T ( k − j ) E j − δ ( k ) T ( k − ( j + 1 ) ) − 1 E j − 1 .
Here the E E E labels are retained, with their foot parametrizations reflected as above; T T T acts on the marked surface paths. Apply T T T to the based edge labels in the table and reverse ψ \psi ψ along their cuff lifts. Apply the following maps, in order,
E 2 − 1 , E 1 − 1 , E 2 − 1 , E 0 − 1 , L 3 , − 1 , U 2 , 1 , U 1 , − 1 , U 0 , − 1 . \mathcal{E}_{2}^{-1},\ \mathcal{E}_{1}^{-1},\ \mathcal{E}_{2}^{-1},\ \mathcal{E}_{0}^{-1},\ \mathcal{L}_{3,-1},\ \mathcal{U}_{2,1},\ \mathcal{U}_{1,-1},\ \mathcal{U}_{0,-1}. E 2 − 1 , E 1 − 1 , E 2 − 1 , E 0 − 1 , L 3 , − 1 , U 2 , 1 , U 1 , − 1 , U 0 , − 1 .
Figure 22. The six marked feet of the three relative 1-handles for the closed symmetric sum, with labels v j = V − j v_j=V_{-j} v j = V − j . The connectors indicate the foot pairings; they are not 2-handle attaching arcs. The displayed foot reflection and fixed surface patches specify their attaching 3-ball parameterizations.
to one embedded lifted ribbon graph. Transport the framing with this embedded graph. The individual based words do not specify the framed attachments.
A simultaneous surface diagram. Let D D D be the fixed disk containing the tree and the three vertex patches, and put S = Σ 4 ∖ int D S = \Sigma_4 \setminus\operatorname{int} D S = Σ 4 ∖ int D . After contracting a 0 ∪ a 1 a_0 \cup a_1 a 0 ∪ a 1 within D D D , number the twenty ports in their positive boundary order on D D D as follows:
Here + + + and − - − denote the initial and terminal ends of an oriented edge. The order on ∂ S \partial S ∂ S is the reverse order. Deleting the b 1 , b 2 b_1,b_2 b 1 , b 2 ends leaves the four consecutive blocks ( A i + , B i − , A i − , B i + ) (A_i^+, B_i^-, A_i^-, B_i^+) ( A i + , B i − , A i − , B i + ) , 0 ≤ i ≤ 3 0 \le i \le3 0 ≤ i ≤ 3 . Thus the eight selected edges give the standard generators for the genus-four surface. Choose the boundary basepoint in the sector of the third lifted polygon face, with port paths along the boundary of the tree neighborhood avoiding that sector. Equivalently, extend the tree neighborhood by a small tongue into that face. The first two face relations are b 0 − 1 h 1 b 1 = 1 b_0^{-1}h_1b_1=1 b 0 − 1 h 1 b 1 = 1 and b 1 − 1 h 2 b 2 = 1 b_1^{-1}h_2b_2=1 b 1 − 1 h 2 b 2 = 1 . They give b 1 = h 1 − 1 B 0 b_1=h_1^{-1}B_0 b 1 = h 1 − 1 B 0 and b 2 = h 2 − 1 h 1 − 1 B 0 b_2=h_2^{-1}h_1^{-1}B_0 b 2 = h 2 − 1 h 1 − 1 B 0 in the free group π 1 ( S ) \pi_1(S) π 1 ( S ) with these port paths; the third records R = ∏ i = 0 3 [ A i , B i ] R=\prod_{i=0}^{3}[A_i,B_i] R = ∏ i = 0 3 [ A i , B i ] as the boundary word. This specifies the punctured-surface marking used in the edge table, including its boundary paths.
Cut S S S into a disk with thirty-two sides, alternating eight pairs of identified sides with sixteen boundary intervals. The identified sides have the cyclic labels
A 0 + , B 0 − , A 0 − , B 0 + , … , A 3 + , B 3 − , A 3 − , B 3 + . A_0^+, B_0^-, A_0^-, B_0^+, \ldots, A_3^+, B_3^-, A_3^-, B_3^+. A 0 + , B 0 − , A 0 − , B 0 + , … , A 3 + , B 3 − , A 3 − , B 3 + .
Oppositely signed sides with the same letter are identified in reverse boundary order. The sixteen intervening intervals form ∂ S \partial S ∂ S ; all twenty ports are placed on one of them. The original graph has the routing shown in Figure 25 . Figure 27 shows the ten non-tree edges after transport by T T T . The retained tree is reinserted in D D D in both cases.
Figure 25. The original non-tree edges in the paired polygon for S S S . Thick gray intervals are parts of ∂ S \partial S ∂ S ; paired labeled sides are read in reverse order. The cap is shown separately in Figure 26 .
Figure 26. The three vertex patches and retained tree in D D D . Equally numbered endpoints are identified with those in the polygon along the oppositely oriented boundaries.
Figure 27. The ten non-tree edges of the transported graph T ( Γ ) T(\Gamma) T ( Γ ) for the symmetric m = 1 m=1 m = 1 sum. Each legend entry k r k_r k r denotes T ( k r ) T(k_r) T ( k r ) . Pair equally labeled sides with reversed endpoint order, then reinsert the cap from Figure 26 . Two copies of the polygonal diagram, joined through E j E_j E j along arcs of constant angle, give the twelve relative product curves on H H H . The zero surface framings refer to H H H . They do not assign integer labels in a dotted-circle presentation.
Theorem B.1. For m = 1 m=1 m = 1 , the paired-polygon routing and the retained tree give a simultaneous surface diagram of the transported lifted graph, relative to D D D . Two copies of this diagram, joined through the three relative 1 1 1 -handles with the prescribed product charts, carry all twelve relative product attaching curves Q k , j Q_{k,j} Q k , j on an oriented surface H H H of genus ten. Their product framings are the surface framings of H H H .
Proof. The routing is recorded by ordered side endpoints and their chord pairings in surface_graph_routing.json. There are thirty-two chords before transport and 126 after transport. The integer check check_surface_routing.py verifies that every endpoint is used once, that identified sides have reversed endpoint orders, and that no two chord pairs alternate on the polygon boundary. Straight chords in any strictly convex realization therefore have disjoint interiors. It also reads each edge word from the side crossings, reduces it in the free group, and checks exact agreement with the edge table or its image under T T T , respectively. In particular this comparison is not made only after imposing R = 1 R=1 R = 1 . The relation T ( R ) = R T(R)=R T ( R ) = R holds freely as well. The cap check check_cap_tree.py uses rational coordinates to restore the two tree edges and the twenty port segments without unintended intersections. It checks the eight incident directions at every vertex against the original lifted rotation order.
These checks identify the base routing with the original marked ribbon graph. The transported routing has the same endpoints and relative arc homotopy classes as T T T of that graph, using the same specified boundary port paths. These are properly embedded arcs, with individually fixed, distinct labeled endpoints. Embedded arc systems with the same relative homotopy classes and fixed distinct endpoints are isotopic relative to the boundary: remove innermost bigons to make corresponding arcs coincide, and proceed in the surface cut along the arcs already matched. Applying this to the ten arcs, and reinserting the fixed tree, gives the asserted surface isotopy. The normal ribbons are transported by this isotopy.
Take the two pages Σ 4 × { 0 , π } \Sigma_4 \times\{0,\pi\} Σ 4 × { 0 , π } , remove the three vertex disks in each, and join the corresponding boundary circles through E 0 , E 1 , E 2 E_0,E_1,E_2 E 0 , E 1 , E 2 . In a relative 1 1 1 -handle, use a round 3 3 3 -ball of radius ρ \rho ρ and write
E j = D v 1 × D x , y 2 , H ∩ E j = { u = 0 , x 2 + y 2 = ρ 2 } , ρ > 0. E_j=D_v^1 \times D_{x,y}^2,\qquad H\cap E_j=\{u=0,\ x^2+y^2=\rho^2\},\qquad\rho>0. E j = D v 1 × D x , y 2 , H ∩ E j = { u = 0 , x 2 + y 2 = ρ 2 } , ρ > 0.
We specify the lane matching and the transverse fields at both feet. Write r = − j r=-j r = − j mod 3. The v = − 1 v=-1 v = − 1 foot of E j E_j E j is on the ψ = 0 \psi=0 ψ = 0 page at V r V_r V r , and its v = 1 v=1 v = 1 foot is on the ψ = π \psi=\pi ψ = π page at V r − 1 V_{r-1} V r − 1 . A superscript + + + denotes an outgoing halfedge of the oriented lifted edge, and − - − denotes an incoming halfedge. In the common oriented ( x , y ) (x,y) ( x , y ) coordinates, the cyclic order at V r V_r V r is
a r + , b r − , a r − 1 − , b r + , c r + , d r − , c r − , d r + . a_r^+,\, b_r^-,\, a_{r-1}^-,\, b_r^+,\, c_r^+,\, d_r^-,\, c_r^-,\, d_r^+. a r + , b r − , a r − 1 − , b r + , c r + , d r − , c r − , d r + .
At the other foot, replace r r r by r − 1 r-1 r − 1 . These are the vertex rotations of the retained ribbon graph. The map T T T fixes the foot patches, so transporting the graph does not change this order. Choose the transverse coordinate t t t on each oriented base band so that ( ∂ s , ∂ t ) (\partial_s,\partial_t) ( ∂ s , ∂ t ) gives the surface orientation. The signed angular field at an outgoing end is then positive and at an incoming end negative. A constant choice of sign for a product normal vector does not change its framing class. The following table gives the matching; all indices are modulo three, and ϵ ℓ \epsilon_\ell ϵ ℓ is the sign of ∂ t \partial_t ∂ t relative to the positive angular field at either foot. The labels stand for their transported arcs under T T T .
Table 1 (PDF p. 34).
The two occurrences of Q k , j Q_{k,j} Q k , j for k = b , c , d k=b,c,d k = b , c , d are its distinct passages through E j E_j E j in opposite directions. For k = a k=a k = a , the two passages lie in E j E_j E j and E j − 1 E_{j-1} E j − 1 , in agreement with the attaching word above.
Choose one oriented parametrization of the base 0-handle so that its eight halfedges meet the boundary at θ ℓ = 2 π ℓ / 8 \theta_\ell=2\pi\ell/8 θ ℓ = 2 π ℓ /8 in the listed order, and lift this same parametrization to all three vertex disks on both pages. Carry the base-band collars and their transverse coordinates by the same parametrization. This common choice respects the existing product transport, which fixes the base coordinates; it does not introduce an independent change of coordinates at either foot. The map T T T fixes these charts. Near a lane endpoint, use the collar coordinates
( q , t , u ) ⟼ ( R ( q , u ) cos ( θ ℓ + ϵ ℓ t / ρ ) , R ( q , u ) sin ( θ ℓ + ϵ ℓ t / ρ ) , u ) , R ( q , u ) = ρ 2 − u 2 + q . (q,t,u)\longmapsto\left(R(q,u)\cos\left(\theta_\ell+\epsilon_\ell t/\rho\right),R(q,u)\sin\left(\theta_\ell+\epsilon_\ell t/\rho\right),u\right),\qquad R(q,u)=\sqrt{\rho^2-u^2}+q. ( q , t , u ) ⟼ ( R ( q , u ) cos ( θ ℓ + ϵ ℓ t / ρ ) , R ( q , u ) sin ( θ ℓ + ϵ ℓ t / ρ ) , u ) , R ( q , u ) = ρ 2 − u 2 + q .
Here q = 0 q=0 q = 0 is the foot sphere and u = 0 u=0 u = 0 is the page; the coordinates are restricted to a sufficiently small collar. At the endpoint q = t = u = 0 q=t=u=0 q = t = u = 0 , the derivatives send
∂ t ⟼ ϵ ℓ ρ ∂ θ , ∂ u ⟼ ∂ u . \partial_t\longmapsto\frac{\epsilon_\ell}{\rho}\partial_\theta,\qquad\partial_u\longmapsto\partial_u. ∂ t ⟼ ρ ϵ ℓ ∂ θ , ∂ u ⟼ ∂ u .
The same formulas hold at the two feet, since the fiber transport changes the sheet index but fixes the base-band transverse coordinate. Extend the eight lanes through the handle by
v ⟼ ( v , ρ cos θ ℓ , ρ sin θ ℓ , 0 ) , − 1 ≤ v ≤ 1 , v\longmapsto(v,\rho\cos\theta_\ell,\rho\sin\theta_\ell,0),\qquad-1\le v\le1, v ⟼ ( v , ρ cos θ ℓ , ρ sin θ ℓ , 0 ) , − 1 ≤ v ≤ 1 ,
and extend their normal fields by ( ϵ ℓ ∂ θ , ∂ u ) (\epsilon_\ell\partial_\theta,\partial_u) ( ϵ ℓ ∂ θ , ∂ u ) . Their angles and signs are constant along v v v , so the lanes are disjoint and the extension has no annular winding. The two page orientations used to orient H H H do not reverse these fixed ( x , y ) (x,y) ( x , y ) foot coordinates. The curve Q k , j Q_{k,j} Q k , j uses T ( k − j ) T(k-j) T ( k − j ) on the first page and T ( k − ( j + 1 ) ) T(k-(j+1)) T ( k − ( j + 1 )) in reverse on the second, with its other sides in E j − δ ( k ) E_{j-\delta(k)} E j − δ ( k ) and E j E_j E j . Every lifted edge occurs once on each page, including the two tree edges. Thus all twenty-four page arcs are represented together. Orient the two pages oppositely. The annuli join them to an oriented connected surface with
χ ( H ) = 2 χ ( Σ 4 ) − 6 = − 18 , g ( H ) = 10. \chi(H)=2\chi(\Sigma_4)-6=-18,\qquad g(H)=10. χ ( H ) = 2 χ ( Σ 4 ) − 6 = − 18 , g ( H ) = 10.
On the pages, the product vector ∂ t \partial_t ∂ t is tangent to the surface and transverse to each lifted band core. The table and collar derivatives identify this vector at both endpoints of every lane with its specified constant angular field. Reversal of ψ \psi ψ acts by ( x , y , u ) ↦ ( x , y , − u ) (x,y,u) \mapsto(x,y,-u) ( x , y , u ) ↦ ( x , y , − u ) at both feet. Its derivative fixes the displayed first normal vector and negates the second at both ends, so it introduces no rotation in the first normal vector. The opposite orientations of the pages agree with the opposite traversals of the two band sides. After rounding each product corner, the matching vectors give a continuous nonzero transverse field on H H H around every Q k , j Q_{k,j} Q k , j . This is the first vector of the transported product framing. Its push-off stays on H H H , and hence the product framing has coefficient zero relative to the surface framing.
Figure 24 shows the construction of H H H from its two genus-four pages and the three annuli through the relative 1-handles.
Figure 24. The genus-ten surface H H H . Remove three vertex disks from each of the pages Σ 4 × { 0 } \Sigma_4 \times\{0\} Σ 4 × { 0 } and Σ 4 × { π } \Sigma_4 \times\{\pi\} Σ 4 × { π } and join their boundary circles by the annuli H ∩ E j H \cap E_j H ∩ E j . The annulus through E j E_j E j joins V − j V_{-j} V − j on the first page to V − ( j + 1 ) V_{-(j+1)} V − ( j + 1 ) on the second, with indices modulo three. The pages are oriented oppositely. Thus χ ( H ) = 2 ( 2 − 8 − 3 ) = − 18 \chi(H)=2(2-8-3)=-18 χ ( H ) = 2 ( 2 − 8 − 3 ) = − 18 and g ( H ) = 10 g(H)=10 g ( H ) = 10 . This is a schematic of the surface construction; the twelve attaching curves are not shown.
Pullback by ( T , ψ ↦ − ψ ) (T,\psi\mapsto-\psi) ( T , ψ ↦ − ψ ) gives the closed attaching maps; the two cap curves retain their base-disk normal framings. Theorem B.1 gives the simultaneous product graph and surface framings. Full handlebody diagrams of the closed sums, with their complete attaching links and integer framing coefficients, will be studied in the follow-up paper [9 ] . No further handle cancellations are made here.
A complement homology check
The torsion in H 2 H_2 H 2 also follows from the complement exact sequence. Put X = Σ g × S 2 X = \Sigma_g \times S^2 X = Σ g × S 2 and F = F g , p F = F_{g,p} F = F g , p . Excision and the Thom isomorphism give H k ( X , M g , p ; Z ) ≅ H k − 2 ( F ; Z ) H_k(X,M_{g,p};\mathbb{Z}) \cong H_{k-2}(F;\mathbb{Z}) H k ( X , M g , p ; Z ) ≅ H k − 2 ( F ; Z ) . The map H 4 ( X ) → H 2 ( F ) H_4(X) \to H_2(F) H 4 ( X ) → H 2 ( F ) is an isomorphism. The relevant part of the long exact sequence is
H 3 ( X ) ⟶ H 1 ( F ) ⟶ H 2 ( M g , p ) ⟶ H 2 ( X ) ⟶ H 0 ( F ) . H_3(X) \longrightarrow H_1(F) \longrightarrow H_2(M_{g,p}) \longrightarrow H_2(X) \longrightarrow H_0(F). H 3 ( X ) ⟶ H 1 ( F ) ⟶ H 2 ( M g , p ) ⟶ H 2 ( X ) ⟶ H 0 ( F ) .
The first map is Poincaré dual to q ∗ : H 1 ( Σ g ; Z ) → H 1 ( F ; Z ) q^*: H^1(\Sigma_g;\mathbb{Z}) \to H^1(F;\mathbb{Z}) q ∗ : H 1 ( Σ g ; Z ) → H 1 ( F ; Z ) . The matrix Q p , g Q_{p,g} Q p , g in Section 3 represents q ∗ q_* q ∗ and has nonzero Smith entries p , 1 , … , 1 p,1,\ldots,1 p , 1 , … , 1 . Transposition and the unimodular Poincaré duality identifications preserve these entries. Thus the first map has cokernel Z 2 G − 2 g ⊕ Z / p \mathbb{Z}^{2G-2g} \oplus\mathbb{Z}/p Z 2 G − 2 g ⊕ Z / p . The intersection map H 2 ( X ) → H 0 ( F ) H_2(X) \to H_0(F) H 2 ( X ) → H 0 ( F ) has image p Z p\mathbb{Z} p Z and kernel Z \mathbb{Z} Z , since [ F ] = p [ S ] [F] = p[S] [ F ] = p [ S ] . Hence
0 ⟶ Z 2 G − 2 g ⊕ Z / p ⟶ H 2 ( M g , p ; Z ) ⟶ Z ⟶ 0. 0 \longrightarrow\mathbb{Z}^{2G-2g} \oplus\mathbb{Z}/p \longrightarrow H_2(M_{g,p};\mathbb{Z}) \longrightarrow\mathbb{Z} \longrightarrow0. 0 ⟶ Z 2 G − 2 g ⊕ Z / p ⟶ H 2 ( M g , p ; Z ) ⟶ Z ⟶ 0.
Since the quotient is free, this sequence splits as abelian groups and recovers the stated H 2 H_2 H 2 . This establishes the abstract group; Corollary 2.4 identifies the difference of the two pole sections as a generator of its torsion subgroup.
Certification of the pole projections
We prove Proposition 2.1 for the analytic arcs outside the paired feet, not for the closed sums. The calculation uses 128-bit Arb arithmetic [10 ] through python-flint. The source package contains certify_projection_crossings.py, camera data in figures/usual_projection_data.json, seeds in verified_usual_projections.json, and certificates in certified_projection_crossings.json. The validator projection_parameters.py checks the drawing parameters; test_projection_parameters.py tests parameter drift. We use ϵ = 0.032 \epsilon= 0.032 ϵ = 0.032 , ζ = 0.012 \zeta= 0.012 ζ = 0.012 , and η = 0.005 \eta= 0.005 η = 0.005 , obtained by the framed collar deformation of § 2 . Only the center arcs are certified, not the enlarged normal markers. Seeds choose candidate neighborhoods; full-domain interval subdivision proves completeness. Execution instructions and software versions for all checks are in COMPUTATIONAL_README.txt.
The foot angles are θ m = 2 π ( m + 0.13 ) / ( 4 g ) \theta_m = 2\pi(m + 0.13)/(4g) θ m = 2 π ( m + 0.13 ) / ( 4 g ) , and the fiber angles are ϕ = σ P ( 2 π j / p + 0.23 ) \phi= \sigma_P(2\pi j/p + 0.23) ϕ = σ P ( 2 π j / p + 0.23 ) . A closing interval has base angle between θ m + ζ \theta_m + \zeta θ m + ζ and θ m + 1 − ζ \theta_{m+1} - \zeta θ m + 1 − ζ . Each arc has the form q ( t ) = A cos t + B sin t q(t) = A\cos t + B\sin t q ( t ) = A cos t + B sin t , with A , B A,B A , B orthonormal. For a corner arc, take A = ( cos θ , sin θ , 0 , 0 ) A = (\cos\theta,\sin\theta,0,0) A = ( cos θ , sin θ , 0 , 0 ) and B = ( 0 , 0 , cos ϕ , sin ϕ ) B = (0,0,\cos\phi,\sin\phi) B = ( 0 , 0 , cos ϕ , sin ϕ ) , with ϵ ≤ t ≤ π / 2 − ϵ \epsilon\le t \le\pi/2 - \epsilon ϵ ≤ t ≤ π /2 − ϵ . For a closing interval, recenter its base angle at θ c = ( θ m + θ m + 1 ) / 2 \theta_c = (\theta_m + \theta_{m+1})/2 θ c = ( θ m + θ m + 1 ) /2 and use A = ( cos θ c , sin θ c , 0 , 0 ) A = (\cos\theta_c,\sin\theta_c,0,0) A = ( cos θ c , sin θ c , 0 , 0 ) ,
B = ( − sin θ c , cos θ c , 0 , 0 ) B=(-\sin\theta_c,\cos\theta_c,0,0) B = ( − sin θ c , cos θ c , 0 , 0 ) , with ∣ t ∣ ≤ π / ( 4 g ) − ζ |t|\leq\pi/(4g)-\zeta ∣ t ∣ ≤ π / ( 4 g ) − ζ . With h = tan ( t / 2 ) h=\tan(t/2) h = tan ( t /2 ) and M = E O M=\mathbf{E}\mathbf{O} M = EO , write
X i ( h ) = n i ( h ) d ( h ) , n i ( h ) = ( A M ) i ( 1 − h 2 ) + 2 ( B M ) i h , d ( h ) = 1 + h 2 − ( A ⋅ ℓ ) ( 1 − h 2 ) − 2 ( B ⋅ ℓ ) h . X_i(h)=\frac{n_i(h)}{d(h)},\qquad n_i(h)=(AM)_i(1-h^2)+2(BM)_i h,\qquad d(h)=1+h^2-(A\cdot\ell)(1-h^2)-2(B\cdot\ell)h. X i ( h ) = d ( h ) n i ( h ) , n i ( h ) = ( A M ) i ( 1 − h 2 ) + 2 ( B M ) i h , d ( h ) = 1 + h 2 − ( A ⋅ ℓ ) ( 1 − h 2 ) − 2 ( B ⋅ ℓ ) h .
Thus a crossing of arcs a , b a,b a , b is a zero of the two polynomials
F i ( h , k ) = n i a ( h ) d b ( k ) − n i b ( k ) d a ( h ) , i = 1 , 2. F_i(h,k)=n_i^a(h)d^b(k)-n_i^b(k)d^a(h),\qquad i=1,2. F i ( h , k ) = n i a ( h ) d b ( k ) − n i b ( k ) d a ( h ) , i = 1 , 2.
The script encloses each stored camera entry in a ball of radius 10 − 12 10^{-12} 1 0 − 12 . It checks that the exact normalized pole and the Gram-Schmidt orthonormalizations of the stored matrices lie in these balls, projecting the columns of E \mathbf{E} E to ℓ ⊥ \ell^\perp ℓ ⊥ first. All subsequent tests hold uniformly over the resulting coefficient intervals. The denominators are positive on the entire arc domains. For each arc the coefficient matrix of ( n 1 , n 2 , d ) (n_1,n_2,d) ( n 1 , n 2 , d ) , in the basis ( 1 , h , h 2 ) (1,h,h^2) ( 1 , h , h 2 ) , has determinant bounded away from zero. Its planar map is therefore a projective image of a nonsingular conic and is injective on the parameter interval.
For every pair of distinct arcs, subdivide the full parameter rectangle. Discard a box if either interval evaluation of F i F_i F i excludes zero. All other terminal boxes must lie in one of the certified root neighborhoods recorded in certified_projection_crossings.json. The entries are grouped by chart name; each crossing records its two arc indices, root box, root enclosure, and depth gap. For a candidate box B B B , with center x 0 x_0 x 0 , choose a nonsingular approximate inverse C C C to the Jacobian and form
K = x 0 − C F ( x 0 ) + ( I − C D F ( B ) ) ( B − x 0 ) . K = x_0 - CF(x_0) + (I - CDF(B))(B - x_0). K = x 0 − C F ( x 0 ) + ( I − C D F ( B )) ( B − x 0 ) .
The interval tests give K ⊂ int B K \subset\operatorname{int} B K ⊂ int B and ∥ I − C D F ( B ) ∥ ∞ < 1 / 2 \lVert I - CDF(B)\rVert_{\infty} < 1/2 ∥ I − C D F ( B ) ∥ ∞ < 1/2 . The contraction mapping theorem then gives exactly one zero in B B B , with enclosure K K K ; the same bound makes its Jacobian nonsingular. Each root box lies strictly inside both arc domains. Distinct root neighborhoods for one pair are disjoint, and the projected enclosures of all crossing points are pairwise disjoint. Finally the interval for X 3 a − X 3 b X^a_3 - X^b_3 X 3 a − X 3 b at each root excludes zero and gives the recorded overpass.
The exhaustive subdivision terminates for all six charts. The total numbers of tested boxes are, in the chart order M 2 , 3 N M^N_{2,3} M 2 , 3 N , M 2 , 3 S M^S_{2,3} M 2 , 3 S , M 3 , 2 N M^N_{3,2} M 3 , 2 N , M 3 , 2 S M^S_{3,2} M 3 , 2 S , M 2 , 4 N M^N_{2,4} M 2 , 4 N , M 2 , 4 S M^S_{2,4} M 2 , 4 S , 67880, 48700, 123596, 97272, 94428, and 159638. Lower bounds for the absolute depth gaps are respectively 0.910, 0.180, 1.226, 0.120, 0.402, and 0.180. These checks establish the stated crossing counts and overpasses for the complement projections.
Anar Akhmedov
School of Mathematics, University of Minnesota
Visiting Scholar, Department of Mathematics, Harvard University
Email: akhmedov@umn.edu