OctoInformBeta

OpenAI preprint, September 2026 · §5. A hierarchy of branching lattice points

Raw block lattices, triangular maps and suffix branching: Raw products, exact covolumes and scalar branching over fixed suffixes

Lattice chains with a terminal shear.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

A hierarchy of lattice blocks lets later coordinates enter earlier blocks through linear shifts. Its coordinates are ordered block by block, with an additional integer lattice after the last block. Branching is a condition on finite point sets: it concerns the possible values of one coordinate when all later coordinates have been fixed.

A block BB consists of a dimension bB∈ℕb_B\in\mathbb N, a prime pBp_B, a scale hB>0h_B>0, a finite usable set UB⊂GBU_B\subset G_B, and a nonzero direction vB∈GBv_B\in G_B, where GB=𝔽pBbBG_B=\mathbb F_{p_B}^{b_B} is its block group.✦ Its real vector space is VB=ℝbBV_B=\mathbb R^{b_B}.✦

For a natural modulus pp, a coordinate set II, and an integer vector k∈ℤIk\in\mathbb Z^I, the residue vector rp(k)r_p(k) over the residue ring modulo pp and the scaled real vector gp(k)∈ℝIg_p(k)\in\mathbb R^I are defined by (1). rp(k)j=kj(mod⁡p),gp(k)j=kjp(j∈I).r_p(k)_j=k_j\pmod p, \qquad g_p(k)_j=\frac{k_j}{p}\quad(j\in I).(1) The Construction A integer submodule Λp(C)⊂ℝI\Lambda_p(C)\subset\mathbb R^I associated with a linear code CC over the residue ring modulo a natural modulus pp on a coordinate set II is the set in (2). Λp(C)={gp(k):k∈ℤI,rp(k)∈C}.\Lambda_p(C)=\{g_p(k):k\in\mathbb Z^I,\ r_p(k)\in C\}.(2) For a block BB with direction vBv_B, the block lattice LB⊂VBL_B\subset V_B is the Construction A lattice of the line generated by vBv_B, as in (3). LB=ΛpB(𝔽pBvB).L_B=\Lambda_{p_B}(\mathbb F_{p_B}v_B).(3)✦

For columns w0,…,wd−1∈GBw_0,\ldots,w_{d-1}\in G_B, the column shear SB[w]:ℝd→VBS_B[w]:\mathbb R^d\longrightarrow V_B is the linear map in (4). Here w̃k\widetilde w_k is the coordinatewise representative of wkw_k with entries in {0,…,pB−1}\{0,\ldots,p_B-1\}. SB[w](z)=hBpB∑k=0d−1zkw̃k.S_B[w](z)=\frac{h_B}{p_B}\sum_{k=0}^{d-1}z_k\widetilde w_k.(4)✦

Products and Triangular Maps

Coordinate concatenation identifies ℝm×ℝn\mathbb R^m\times\mathbb R^n with ℝm+n\mathbb R^{m+n}, in that order.✦ For integer submodules L⊂ℝmL\subset\mathbb R^m and N⊂ℝnN\subset\mathbb R^n, the direct product L×N⊂ℝm+nL\times N\subset\mathbb R^{m+n} is the integer submodule consisting of the concatenations of vectors of LL and NN.✦ The linear image of an integer submodule L⊂ℝnL\subset\mathbb R^n under an invertible linear map ee of ℝn\mathbb R^n is the integer submodule eL⊂ℝneL\subset\mathbb R^n given by eL={ex:x∈L}eL=\{e x:x\in L\}.✦ The integer lattice ℤn⊂ℝn\mathbb Z^n\subset\mathbb R^n is the integer span of the standard coordinate vectors.✦

Given linear automorphisms ee of ℝm\mathbb R^m and ff of ℝn\mathbb R^n, and a linear map S:ℝn→ℝmS:\mathbb R^n\longrightarrow\mathbb R^m, the block triangular map 𝒯(e,f,S)\mathcal T(e,f,S) is the linear map from ℝm×ℝn\mathbb R^m\times\mathbb R^n to itself given by (5). 𝒯(e,f,S)(u,v)=(eu+Sv,fv).\mathcal T(e,f,S)(u,v)=(e u+S v,f v).(5)✦

A branching block chain cc is a finite nonempty sequence of blocks, with a family ww of bBb_B columns in GAG_A whenever a block BB follows a block AA.✦ Write c⋆wBc\mathbin{\star_w}B for appending BB to a chain ending in AA. The chain dimension d(c)d(c), raw product lattice R(c)⊂ℝd(c)R(c)\subset\mathbb R^{d(c)}, and real determinant parameter D(c)D(c) are specified by (6); δB\delta_B denotes the block determinant parameter. d(B)=bB,d(c⋆wB)=d(c)+bB,R(B)=LB,R(c⋆wB)=R(c)×LB,D(B)=δB,D(c⋆wB)=D(c)δB.\begin{aligned} d(B)&=b_B,& d(c\mathbin{\star_w}B)&=d(c)+b_B,\\ R(B)&=L_B,& R(c\mathbin{\star_w}B)&=R(c)\times L_B,\\ D(B)&=\delta_B,& D(c\mathbin{\star_w}B)&=D(c)\delta_B. \end{aligned}(6) The chain vector space of cc is ℝd(c)\mathbb R^{d(c)}.

For a chain cc ending in AA, the last block inclusion Jc:VA→ℝd(c)J_c:V_A\longrightarrow\mathbb R^{d(c)} places a vector in the last block and sets all preceding coordinates to zero.✦ For blocks AA and BB with column data ww, write EBE_B for the block linear automorphism on VBV_B and SA,B,w:VB→VAS_{A,B,w}:V_B\longrightarrow V_A for the column shear of (4) with block AA and columns ww. The chain triangular automorphism Tc:ℝd(c)→ℝd(c)T_c:\mathbb R^{d(c)}\longrightarrow\mathbb R^{d(c)} is defined recursively by (7). TB=EB,Tc⋆wB=𝒯(Tc,EB,Jc∘SA,B,w).T_B=E_B, \qquad T_{c\mathbin{\star_w}B} =\mathcal T(T_c,E_B,J_c\circ S_{A,B,w}).(7)✦

For a block BB, let ℓB\ell_B be the least nonnegative integer with 2ℓB≥pB2^{\ell_B}\ge p_B. For nonnegative integers bb and pp, the binary terminal columns are the vectors 2rej2^r e_j over the residue ring modulo pp, indexed by 0≤j<b0\leq j<b and 0≤r<ℓ0\leq r<\ell, where eje_j is the jjth standard coordinate vector over that ring and ℓ\ell is the least nonnegative integer with p≤2ℓp\leq2^\ell. For a block BB, with b=bBb=b_B and p=pBp=p_B, the terminal dimension tBt_B, these columns wtermw^{\mathrm{term}}, and the terminal binary shear QBQ_B are given in (8). The columns are indexed by pairs (j,r)(j,r) in lexicographic order. tB=bBℓB,wj,rterm=2rej(0≤j<bB,0≤r<ℓB),QB=SB[wterm].\begin{aligned} t_B&=b_B\ell_B,\\ w^{\mathrm{term}}_{j,r}&=2^r e_j &&(0\le j<b_B,\ 0\le r<\ell_B),\\ Q_B&=S_B[w^{\mathrm{term}}]. \end{aligned}(8) For a chain cc ending in a block BB, the extended dimension is d+(c)=d(c)+tBd^+(c)=d(c)+t_B, and the extended vector space is ℝd+(c)\mathbb R^{d^+(c)}.

The Claim

Definition 1 (Terminal extension and suffix branching). For a branching block chain cc ending in a block BB, the terminal product lattice R+(c)⊂ℝd+(c)R^+(c)\subset\mathbb R^{d^+(c)} is the direct product in (9). R+(c)=R(c)×ℤtB.R^+(c)=R(c)\times\mathbb Z^{t_B}.(9)✦ The extended triangular automorphism Tc+T_c^+ acts on a pair (x,z)∈ℝd(c)×ℝtB(x,z)\in\mathbb R^{d(c)}\times\mathbb R^{t_B} as in (10). Tc+(x,z)=(Tcx+JcQBz,z).T_c^+(x,z)=(T_c x+J_c Q_B z,z).(10)✦ The full vertical lattice L+(c)⊂ℝd+(c)L^+(c)\subset\mathbb R^{d^+(c)} is the linear image in (11). L+(c)=Tc+R+(c).L^+(c)=T_c^+R^+(c).(11)✦

A finite set P⊂ℝmP\subset\mathbb R^m is suffix binary if, for each coordinate jj and each fixed strict suffix yy, its jjth coordinate fiber lies in a translate of {0,1}\{0,1\}.✦ Thus there are real numbers aj,ya_{j,y} satisfying (12). {xj:x∈P,xk=yk for every k>j}⊂{aj,y,aj,y+1}(0≤j<m,y∈ℝm).\{x_j:x\in P,\ x_k=y_k\text{ for every }k>j\} \subset\{a_{j,y},a_{j,y}+1\} \quad(0\le j<m,\ y\in\mathbb R^m).(12)✦

Theorem 1 (Block and full covolumes). For every block BB, the covolume of its block lattice is given by (13). covol⁡(LB)=pB−1.\operatorname{covol}(L_B)=p_B^{-1}.(13)✦ For every branching block chain cc, the full vertical lattice has the covolume in (14). covol⁡(L+(c))=D(c).\operatorname{covol}(L^+(c))=D(c).(14)✦

What Is Checked

The lattice construction is stated over blocks with prime moduli, positive scales and nonzero directions, together with arbitrary branching chains of column data; the suffix condition is stated over finite subsets of real coordinate spaces. For the block triangular map with invertible diagonal maps ee and ff, the inverse is the linear map in (15). 𝒯(e,f,S)−1(u,v)=(e−1(u−Sf−1v),f−1v).\mathcal T(e,f,S)^{-1}(u,v) =\bigl(e^{-1}(u-S f^{-1}v),f^{-1}v\bigr).(15)