Assembly Addition Chains

• Assembly Addition Chains are a generalization of classical addition chains that define minimal-length sequences for constructing complex combinatorial or algebraic objects via iterative gluing operations.
• The framework establishes universal bounds, including a log-based lower limit and refined upper bounds for binary and 2-piece decompositions across diverse structures such as graphs, strings, and polyominoes.
• Algorithmic approaches like binary decomposition, m-ary window methods, and tree constructions enable efficient assembly with practical applications in exponentiation, graph assembly, and optimized matrix powering.

Assembly Addition Chains (AACs) generalize classical addition chains to combinatorial and algebraic structures beyond the positive integers, providing a unified framework for efficient assembly of complex structures from primitive building blocks. They capture minimal-length sequences that realize a specified target object via iterative “gluing” (or combining) operations, and admit precise theoretical bounds and efficient construction methods across a variety of settings including strings, graphs, polyominoes, and classical integer exponentiation (Cronin et al., 19 Dec 2025).

1. Formalism and Fundamental Definitions

Let $S$ be a set of combinatorial or algebraic objects and let %%%%1%%%% denote a distinguished set of building blocks (indecoposable objects). An Assembly Multi-Magma is the pair $(S, \circ)$, where $\circ: 2^S \times 2^S \to 2^S$ defines a set-valued binary operation (“gluing”), to account for potentially non-unique assembly pathways (e.g., multiple gluings of graphs along different subgraphs).

An Assembly Addition Chain (AAC) for $O \in S$ is a finite sequence

$(O_1, O_2, \dots, O_r = O)$

with $O_1 \in \mathrm{BB} \circ \mathrm{BB}$, and for each $i = 2,\dots, r$, there exist $X, Y \in \{O_1,\dots,O_{i-1}\} \cup \mathrm{BB}$ with $O_i \in \{X\} \circ \{Y\}$ (Cronin et al., 19 Dec 2025).

The assembly index $a(O)$ is the minimal chain length for $O$, i.e., $a(O) = \min\{ L(C) : C \in AAC(O, \mathrm{BB}) \}$, and an optimal AAC achieves this minimum.

In the classical case $S = \mathbb{Z}^+$, $\mathrm{BB} = \{1\}$, and $\circ$ is $+$, which reduces to traditional addition chaining for integer exponentiation (Cronin et al., 19 Dec 2025, Elias et al., 2016).

2. Theoretical Bounds

Let $s(O)$ denote the size of object $O$, that is, the number of building blocks required in a canonical assembly (e.g., the length of a string, the number of edges in a graph).

The universal lower bound is

$a(O) \geq \lceil \log_2 s(O) \rceil$

since each step at most doubles the assembly size (Cronin et al., 19 Dec 2025). Sending sizes to $\mathbb{Z}^+$ recovers the classical addition-chain lower bound $\ell(s(O)) \geq \lceil \log_2 s(O) \rceil$.

A trivial upper bound is $a(O)\leq s(O)-1$ (linear stepwise assembly). Substantial refinements are possible for specific structures:

• Binary-decomposable objects (objects partitionable into powers-of-two subpieces):

$$a(O) \leq (H-1) + \sum_{i=1}^{n_1} \min\left\{ \sum_{j=1}^H \left\lfloor 2^{n_j-i} \right\rfloor,\, |S(2^i)| \right\}$$

where $s(O) = 2^{n_1} + ... + 2^{n_H}$ and $|S(2^i)|$ is the number of objects in $S$ of size $2^i$ (Cronin et al., 19 Dec 2025).

• 2-piece decomposition: If $O$ may be assembled from $2k$ or $2k+1$ blocks via pairwise gluing:

$$a(O) \leq \min\left\{ \left\lfloor \frac{s(O)}{2} \right\rfloor,\, |S(2)| \right\} + \left\lceil \frac{s(O)}{2} \right\rceil - 1$$

For classical integer chains ($S = \mathbb{Z}^+$), these bounds specialize to the Schönhage bound $\ell(n) \leq \log_2 n + H(n) - 1$ (Cronin et al., 19 Dec 2025, Elias et al., 2016).

3. Algorithmic Constructions

AACs generalize the construction algorithms of classical (integer) addition chains:

• Binary decomposition: Split $O$ into subobjects whose sizes correspond to the binary expansion of $s(O)$. Each subobject is assembled recursively via doubling, then the subobjects are glued in decreasing order. This replicates the m-ary exponentiation method and achieves near-optimal lengths for many combinatorial targets (Cronin et al., 19 Dec 2025, Elias et al., 2016).
• 2-piece decomposition: Partition the building blocks into as many pairs as possible, glue each pair, and iteratively glue the results. Analogue to optimal binary tree construction in the sum-chain setting (Cronin et al., 19 Dec 2025).
• Generalized m-ary and window methods: For classical $S = \mathbb{Z}^+$ and general $g \geq 2$, a $g$-addition chain for $d$ is a sequence where each term is the sum of at most $g$ previous entries. The window method groups digits in base $g^w$ and precomputes optimal subchains for window values, leading to length bounds of form $\ell_g(d) \leq \lfloor \log_g d \rfloor + \mu_g(d)$, where $\mu_g(d)$ is the base-$g$ Hamming weight (Elias et al., 2016).
• Tree methods: For small instances, the tree method grows all possible sums of $g$ or fewer prior terms at each step, finding optimal length, though it is exponential in practice (Elias et al., 2016).

All methods inherit the asymptotic rate $\ell_g(d)\sim \log_g d$ for integer $g$-addition chains, and extend with minor modification to AAC calculation for finite-sized combinatorial targets (Elias et al., 2016, Cronin et al., 19 Dec 2025).

4. Applications in Combinatorial and Algebraic Structures

AAC methodology supports the efficient assembly of both classical and complex structures:

• j-Strings: For $S$ the set of all strings over an alphabet of size $j$, with singleton strings as $\mathrm{BB}$, gluing is concatenation. For example, to assemble “aababb,” optimal AAC length matches binary decomposition bounds, with empirical optimal chain length 4 for string length 6 (Cronin et al., 19 Dec 2025).
• Colored Connected Graphs: $S$ is the family of colored, connected graphs with edge-colored building blocks. Gluing comprises uniting graphs along complementary subgraphs. All connected graphs are 2-piece decomposable, yielding tight upper indices: $$a(G) \leq \min\{\lfloor s/2 \rfloor, m(m-1)/2\} + \lceil s/2 \rceil - 1$$ for $m$ edge colors (Cronin et al., 19 Dec 2025).
• Colored Polyominoes: $S$ is the set of colored polyominoes, with gluing by identification of touching subpolyominoes. Skeleton graph construction ensures all are 2-piece decomposable; optimal bounds closely track numerically observed behaviors (Cronin et al., 19 Dec 2025).
• Classical integer exponentiation: $S = \mathbb{Z}^+$, $\mathrm{BB} = \{1\}$, and AACs recover all $g$-addition-chain results. AAC formalism naturally extends to multi-operand/fused multiply-add regimes relevant in hardware (Elias et al., 2016).
• Data science and algebraic computation: AAC-based algorithms reduce the number of multiplications (or complex convolution-like operations) required for tasks such as matrix powering (minimizing full matrix multiplications) and subset convolution in Bayesian clustering, with concrete sublinear savings—for example, using an extremal stamp chain of length $k$ for target range $n$ achieves $k-1$ multiplications vs. $n-1$ naively (Kohonen et al., 2013).

5. Extremal Constructions and Stamp Chain Methodology

In the context of integer AACs, extremal addition chains correspond to solutions that also provide an additive 2-basis for appropriate intervals (the “stamp chain” or “postage stamp” approach) (Kohonen et al., 2013):

• A stamp chain is a finite sequence that is both an addition chain and an additive 2-basis for $[1, n]$.
• An extremal basis of length $k$ achieves range $n(k)$ maximal for that size.
• There is a systematic +1-shift transformer: if $A_k$ is an admissible basis (generating all $1 \leq c \leq a_k$), then $B_{k+1} = \{1\} \cup (A_k + 1)$ is an admissible stamp chain with range $n(A_k) + 2$ (Kohonen et al., 2013).
• There is a bijection between extremal 2-bases of length $k$ and extremal stamp chains of length $k+1$; the maximal range $T(k+1) = n(k) + 2$ (Kohonen et al., 2013).
• Empirical enumeration up to $k=24$ yields maximum $n(24)=212$ with explicit constructions, and the induced chains of length 25 have range 214 (Kohonen et al., 2013).

In algorithmic applications, stamp chains enable evaluation of functions $y_i = u(x^i)$ for $i=1,\dots,n$ using at most $T_k-1 = O(\sqrt{n})$ multiplications, asymptotically improving upon the trivial $n-1$ method (Kohonen et al., 2013).

6. Generalizations: g-Addition Chains and Assembly Level Implementation

AACs subsume $g$-addition chains, where each step allows sums of up to $g$ prior entries. For integer targets, the minimal length function $\ell_g(d)$ satisfies

$$\lfloor \log_g d \rfloor \leq \ell_g(d) \leq \lfloor \log_g d \rfloor + \mu_g(d)$$

with $\lim_{d \to \infty} \ell_g(d)/\log_g d = 1$, and explicit constructions via m-ary windowed and tree methods (Elias et al., 2016). The practical significance is heightened for architectures supporting multi-operand multiplication or fused multiply-add instructions. Optimal $g$-chains minimize instruction count and critical-path latency for repeated multiplications or exponentiations (Elias et al., 2016).

AACs extend these principles to any assembly scenario where objects can be recursively built from primitive units via structured gluing, thus serving as a broadly applicable combinatorial tool for resource-optimal construction sequences in both theoretical and applied contexts (Cronin et al., 19 Dec 2025).