Contraction Width in Graph & Tensor Networks

• Contraction width is a family of measures that quantifies structural irregularities during vertex contractions, as seen in twin-width and tensor-network analyses.
• It underpins methods that compute bounded contraction sequences, enabling linear-time algorithms for important graph classes and facilitating FO model checking.
• Variants like total twin-width and component twin-width reveal tight links with clique-width and rank-width, shaping both theoretical insights and practical algorithms.

Searching arXiv for the cited works to ground the synthesis. {"query":"(Hliněný et al., 2022) contraction width twin-width planar graphs (Jakes-Schauer et al., 2019) carving-width contraction trees tensor networks (Bernstein et al., 2017) distance-preserving graph contractions (Kráľ et al., 2023) graphs on surfaces twin-width (Baril et al., 5 Sep 2025) component twin-width (Gajarský et al., 2022) twin-width and types", "max_results": 10} Contraction width denotes a family of context-dependent measurements attached to contraction processes. In graph structure theory, its most prominent incarnation is twin-width: a graph is viewed as a trigraph, vertices are merged until one vertex remains, and the width of a contraction sequence is the maximum red degree appearing along the sequence (Bonnet et al., 2020). In tensor-network algorithms, contraction width is expressed through contraction trees and the induced space and time bottlenecks, most notably $B_{\mathrm{s}}(G)$ and $B_{\mathrm{t}}(G)$, with carving-width capturing the logarithm of the spatial bottleneck (Jakes-Schauer et al., 2019). In distance-preserving graph compression, the term is not fixed as a single formal invariant, but the framework explicitly suggests width-like quantities based on how much edge or vertex mass can be contracted under a tolerance function $\varphi(x)=x/\alpha-\beta$ (Bernstein et al., 2017). This suggests that contraction width is best regarded as a family of contraction-based invariants rather than a single universally standardized parameter.

1. Twin-width as contraction width

In the twin-width setting, the basic object is a trigraph: a graph in which some edges are marked red. Starting from an ordinary graph $G$ with no red edges, one repeatedly contracts pairs of vertices. If $x_1,x_2$ are contracted into $x_0$, then the full neighborhood of $x_0$ is

$$N_{G'}(x_0)=\bigl(N_G(x_1)\cup N_G(x_2)\bigr)\setminus\{x_1,x_2\},$$

while the red neighborhood is

$$N_{G'}^r(x_0)=\bigl(N_G^r(x_1)\cup N_G^r(x_2)\cup (N_G(x_1)\Delta N_G(x_2))\bigr)\setminus\{x_1,x_2\}.$$

Red edges therefore record neighborhood disagreements created by contraction (Hliněný et al., 2022).

A contraction sequence is a sequence

$G=G_0,G_1,\dots,G_{n-1},$

where each $B_{\mathrm{t}}(G)$0 is obtained from $B_{\mathrm{t}}(G)$1 by contracting a pair of vertices and the final trigraph has one vertex. The width of such a sequence is

$B_{\mathrm{t}}(G)$2

that is, the maximum red degree ever attained. Twin-width is the minimum possible width: $B_{\mathrm{t}}(G)$3 Equivalently, $B_{\mathrm{t}}(G)$4 has twin-width at most $B_{\mathrm{t}}(G)$5 if and only if it admits a contraction sequence in which the red degree of every vertex, at every step, is at most $B_{\mathrm{t}}(G)$6 (Hliněný et al., 2022).

An equivalent formulation uses partitions and impurity graphs. For a partition $B_{\mathrm{t}}(G)$7 of $B_{\mathrm{t}}(G)$8, the quotient trigraph $B_{\mathrm{t}}(G)$9 has one vertex per part, with an impure pair corresponding to a red edge. The width of a contraction sequence is then the maximum degree of the impurity graphs $\varphi(x)=x/\alpha-\beta$0 over all times $\varphi(x)=x/\alpha-\beta$1; the twin-width of $\varphi(x)=x/\alpha-\beta$2 is the minimum such width (Gajarský et al., 2022). In this sense, “contraction width” in the twin-width literature is exactly the maximal local irregularity generated during iterative merging.

2. Explicit contraction-width bounds in graph classes

A major line of work studies explicit bounded-width contraction sequences for concrete graph classes. For planar graphs, the twin-width is at most $\varphi(x)=x/\alpha-\beta$3, and a witnessing contraction sequence can be found in linear time (Hliněný et al., 2022). The same paper also proves explicit upper bounds of $\varphi(x)=x/\alpha-\beta$4 for simple bipartite planar graphs, $\varphi(x)=x/\alpha-\beta$5 for simple 1-planar graphs, $\varphi(x)=x/\alpha-\beta$6 for the square of any bipartite planar graph, and $\varphi(x)=x/\alpha-\beta$7 for map graphs, all via a recursive decomposition based on skeletal trigraphs, BFS trees, wrapped faces, and level-respecting contractions (Hliněný et al., 2022).

For graphs embeddable in a surface of Euler genus $\varphi(x)=x/\alpha-\beta$8, the twin-width is at most

$\varphi(x)=x/\alpha-\beta$9

and the proof yields a quadratic-time algorithm to construct a corresponding contraction sequence (Kráľ et al., 2023). The same work gives a lower bound of

$G$0

so the dependence on $G$1 is asymptotically $G$2 (Kráľ et al., 2023).

For posets, the paper on width-bounded posets defines a natural twin-width directly on red posets and proves that a poset of width $G$3 has natural twin-width at most $G$4, hence symmetric twin-width at most $G$5; the corresponding contraction sequence can be constructed in time $G$6 (Balabán et al., 2021). This is asymptotically tight up to constants, since there exists a poset of width $G$7 whose natural twin-width is at least $G$8 (Balabán et al., 2021). In the special case of width $G$9, the worst-case value is exactly $x_1,x_2$0, and a width-$x_1,x_2$1 contraction sequence can be computed in linear time (Balabán et al., 2021).

Bounded VC-dimension yields sub-linear contraction width. For graphs with VC-dimension at most $x_1,x_2$2, the twin-width is

$x_1,x_2$3

obtained by partitioning the vertex set into blocks whose members have similar neighborhoods toward later blocks (Biedl et al., 19 Jun 2026). Interval graphs admit a sharper upper bound $x_1,x_2$4, while there also exist $x_1,x_2$5-vertex interval graphs with twin-width at least

$x_1,x_2$6

(Biedl et al., 19 Jun 2026).

3. Variants of contraction width and relations to other width parameters

The contraction-sequence viewpoint supports several refinements of twin-width. Component twin-width measures the maximum size of a red-connected component; total twin-width measures the number of red edges; total vertex twin-width measures the number of vertices incident to a red edge (Baril et al., 5 Sep 2025). These variants admit unexpectedly tight comparisons with clique-width parameters: $x_1,x_2$7 and

$x_1,x_2$8

while total twin-width satisfies

$x_1,x_2$9

(Baril et al., 5 Sep 2025). A separate synthesis through contraction sequences shows that component twin-width is functionally equivalent to rank-width, and total twin-width is functionally equivalent to linear rank-width (Bonnet et al., 2021).

These relations clarify that the width measured along a contraction sequence can be tuned to recover older decomposition paradigms. In sparse regimes, this becomes particularly sharp. If a graph class excludes $x_0$0 and has twin-width at most $x_0$1, then its tree-width is bounded by a polynomial in $x_0$2, more precisely by $x_0$3 (Bergougnoux et al., 2023). Conversely, there is a $x_0$4-free graph class of twin-width $x_0$5 and unbounded tree-width, showing that the twin-width-$x_0$6 threshold is structurally special (Bergougnoux et al., 2023).

The “lens of contraction sequences” perspective also introduces oriented twin-width, where newly created red edges are oriented away from the contracted vertex and only the red out-degree is bounded. Although this looks weaker, bounded oriented twin-width and bounded twin-width coincide as graph classes (Bonnet et al., 2021). The same paper defines spanning twin-width and shows that, for monotone classes, bounded spanning twin-width is equivalent to being proper minor-closed (Bonnet et al., 2021).

4. Logical and algorithmic significance of bounded contraction width

Bounded contraction width is algorithmically meaningful only when a witnessing sequence is available. For graphs of bounded twin-width, given a $x_0$7-contraction sequence, FO model checking is fixed-parameter tractable in the formula size and runs in time

$x_0$8

where $x_0$9 is computable and $x_0$0 is the domain of the input binary structure (Bonnet et al., 2020). The same framework is preserved under FO interpretations and transductions, including operations such as squaring and complementation (Bonnet et al., 2020).

A more refined local-type methodology shows that bounded contraction width supports much stronger data-structural consequences. For a fixed FO formula $x_0$1 and fixed width bound $x_0$2, one can, from a graph $x_0$3 together with a contraction sequence of width at most $x_0$4, build in $x_0$5 time a data structure answering tuple queries in $x_0$6 time; with $x_0$7-time preprocessing, the query time can be reduced to $x_0$8 (Gajarský et al., 2022). After $x_0$9-time preprocessing, one can also enumerate all satisfying tuples with $$N_{G'}(x_0)=\bigl(N_G(x_1)\cup N_G(x_2)\bigr)\setminus\{x_1,x_2\},$$0 delay (Gajarský et al., 2022). The same local-type machinery yields the VC-density bound that, for a graph $$N_{G'}(x_0)=\bigl(N_G(x_1)\cup N_G(x_2)\bigr)\setminus\{x_1,x_2\},$$1 of twin-width $$N_{G'}(x_0)=\bigl(N_G(x_1)\cup N_G(x_2)\bigr)\setminus\{x_1,x_2\},$$2, a subset $$N_{G'}(x_0)=\bigl(N_G(x_1)\cup N_G(x_2)\bigr)\setminus\{x_1,x_2\},$$3, and an FO formula $$N_{G'}(x_0)=\bigl(N_G(x_1)\cup N_G(x_2)\bigr)\setminus\{x_1,x_2\},$$4, the number of subsets of $$N_{G'}(x_0)=\bigl(N_G(x_1)\cup N_G(x_2)\bigr)\setminus\{x_1,x_2\},$$5 definable by $$N_{G'}(x_0)=\bigl(N_G(x_1)\cup N_G(x_2)\bigr)\setminus\{x_1,x_2\},$$6 using $$N_{G'}(x_0)=\bigl(N_G(x_1)\cup N_G(x_2)\bigr)\setminus\{x_1,x_2\},$$7-tuples of parameters is $$N_{G'}(x_0)=\bigl(N_G(x_1)\cup N_G(x_2)\bigr)\setminus\{x_1,x_2\},$$8 (Gajarský et al., 2022).

On the algorithmic side of computing contraction sequences, the feedback-edge-number parameterization gives concrete positive results. If $$N_{G'}(x_0)=\bigl(N_G(x_1)\cup N_G(x_2)\bigr)\setminus\{x_1,x_2\},$$9, then deciding whether $$N_{G'}^r(x_0)=\bigl(N_G^r(x_1)\cup N_G^r(x_2)\cup (N_G(x_1)\Delta N_G(x_2))\bigr)\setminus\{x_1,x_2\}.$$0 has twin-width at most $$N_{G'}^r(x_0)=\bigl(N_G^r(x_1)\cup N_G^r(x_2)\cup (N_G(x_1)\Delta N_G(x_2))\bigr)\setminus\{x_1,x_2\}.$$1 admits a linear bikernel, and a corresponding 2-contraction sequence can be computed in time

$$N_{G'}^r(x_0)=\bigl(N_G^r(x_1)\cup N_G^r(x_2)\cup (N_G(x_1)\Delta N_G(x_2))\bigr)\setminus\{x_1,x_2\}.$$2

(Balabán et al., 2023). More generally, there is an algorithm running in time $$N_{G'}^r(x_0)=\bigl(N_G^r(x_1)\cup N_G^r(x_2)\cup (N_G(x_1)\Delta N_G(x_2))\bigr)\setminus\{x_1,x_2\}.$$3 that outputs a contraction sequence of width at most $$N_{G'}^r(x_0)=\bigl(N_G^r(x_1)\cup N_G^r(x_2)\cup (N_G(x_1)\Delta N_G(x_2))\bigr)\setminus\{x_1,x_2\}.$$4 (Balabán et al., 2023). The same paper also proves that every graph with feedback edge number $$N_{G'}^r(x_0)=\bigl(N_G^r(x_1)\cup N_G^r(x_2)\cup (N_G(x_1)\Delta N_G(x_2))\bigr)\setminus\{x_1,x_2\}.$$5 has twin-width at most $$N_{G'}^r(x_0)=\bigl(N_G^r(x_1)\cup N_G^r(x_2)\cup (N_G(x_1)\Delta N_G(x_2))\bigr)\setminus\{x_1,x_2\}.$$6 (Balabán et al., 2023).

5. Tensor-network contraction width

In tensor-network algorithms, contraction width is formalized through contraction trees rather than red-degree sequences. A tensor network is modeled as an undirected weighted graph $$N_{G'}^r(x_0)=\bigl(N_G^r(x_1)\cup N_G^r(x_2)\cup (N_G(x_1)\Delta N_G(x_2))\bigr)\setminus\{x_1,x_2\}.$$7, and a contraction ordering is represented by a rooted or free contraction tree whose leaves correspond to tensors and whose arcs and internal nodes are labeled by cuts of the original graph (Jakes-Schauer et al., 2019).

For a contraction tree $$N_{G'}^r(x_0)=\bigl(N_G^r(x_1)\cup N_G^r(x_2)\cup (N_G(x_1)\Delta N_G(x_2))\bigr)\setminus\{x_1,x_2\}.$$8, the paper defines the spatial and temporal bottlenecks. If an arc $$N_{G'}^r(x_0)=\bigl(N_G^r(x_1)\cup N_G^r(x_2)\cup (N_G(x_1)\Delta N_G(x_2))\bigr)\setminus\{x_1,x_2\}.$$9 induces a bipartition $G=G_0,G_1,\dots,G_{n-1},$0, its weight is

$G=G_0,G_1,\dots,G_{n-1},$1

which is the size of the corresponding intermediate tensor. The space bottleneck is

$G=G_0,G_1,\dots,G_{n-1},$2

If an internal node $G=G_0,G_1,\dots,G_{n-1},$3 induces a tripartition $G=G_0,G_1,\dots,G_{n-1},$4, then

$G=G_0,G_1,\dots,G_{n-1},$5

and the time bottleneck is

$G=G_0,G_1,\dots,G_{n-1},$6

The total time is

$G=G_0,G_1,\dots,G_{n-1},$7

(Jakes-Schauer et al., 2019).

In this setting, contraction width is best understood as these bottleneck quantities, especially $G=G_0,G_1,\dots,G_{n-1},$8 and $G=G_0,G_1,\dots,G_{n-1},$9. The paper proves

$B_{\mathrm{t}}(G)$00

and identifies carving-width as the logarithm of the spatial bottleneck: $B_{\mathrm{t}}(G)$01 (Jakes-Schauer et al., 2019). For planar tensor networks, the Ratcatcher algorithm of Seymour and Thomas computes carving-width, and the resulting contraction trees provide a practical heuristic for controlling both memory and time in exact contraction (Jakes-Schauer et al., 2019).

6. Distance-preserving contractions and adjacent contraction-centered notions

A different notion arises in metric compression. Given an edge-weighted graph $B_{\mathrm{t}}(G)$02 with positive lengths and a tolerance function $B_{\mathrm{t}}(G)$03, a set of edges $B_{\mathrm{t}}(G)$04 is a strong $B_{\mathrm{t}}(G)$05-distance-preserving contraction if

$B_{\mathrm{t}}(G)$06

where $B_{\mathrm{t}}(G)$07 sets contracted edges to length $B_{\mathrm{t}}(G)$08 (Bernstein et al., 2017). For affine tolerances

$B_{\mathrm{t}}(G)$09

the framework studies how much a graph can be compressed subject to multiplicative and additive distortion (Bernstein et al., 2017).

The paper explicitly remarks that it does not use the term “contraction width,” but it naturally suggests width-like parameters such as edge-contraction width

$B_{\mathrm{t}}(G)$10

vertex-contraction width

$B_{\mathrm{t}}(G)$11

and tolerance width for a target compressed size (Bernstein et al., 2017). In this metric setting, contraction width quantifies compressibility under distortion rather than red-degree irregularity. The paper gives linear-time greedy algorithms for paths and cycles with unit lengths, dynamic programming for strong contractions on trees in $B_{\mathrm{t}}(G)$12 time and for weak contractions in $B_{\mathrm{t}}(G)$13 time, as well as NP-hardness and inapproximability results on more general graph classes (Bernstein et al., 2017).

Adjacent literature uses contraction as an algorithmic mechanism even when no standalone “contraction width” parameter is defined. A contraction-recursive treewidth algorithm interleaves an exact treewidth routine with recursive calls on edge contractions $B_{\mathrm{t}}(G)$14, and may return a minimal contraction $B_{\mathrm{t}}(G)$15 of $B_{\mathrm{t}}(G)$16 such that $B_{\mathrm{t}}(G)$17 as a certificate (Tamaki, 2023). By contrast, clique-width behaves badly under unrestricted edge contraction: graphs of clique-width at most $B_{\mathrm{t}}(G)$18 can yield, after edge contractions, graphs of unbounded clique-width; however, contractions of edges incident with a degree-2 vertex preserve bounded clique-width up to the bound $B_{\mathrm{t}}(G)$19 when the original graph has clique-width $B_{\mathrm{t}}(G)$20 (Courcelle, 2013). These neighboring results reinforce a general lesson: contraction is a pervasive operation across graph theory and algorithms, but the width quantity it induces depends sharply on what structural information is preserved and on which contractions are allowed.