---
title: ℓ₁-Fiedler Value
url: https://www.emergentmind.com/topics/l_1-fiedler-value
type: topic
---

# ℓ₁-Fiedler Value

The $\ell_1$-Fiedler value $b(G)$ is a combinatorial graph invariant introduced as an $\ell_1$-norm analogue of graph algebraic connectivity. Formally, $b(G)$ is defined for a simple undirected graph $G=(V,E)$ as the minimum $\ell_1$-smoothness over all zero-sum, unit $\ell_1$-norm vectors, optimizing the sum of edge differences. This parameter provides a direct connection to the sparsest-cut problem and exposes new relationships between spectral graph theory and combinatorial expansion properties, with deep ties to Laplacian eigenvalues, isoperimetric numbers, and extremal graph constructions [2306.13032], [2601.05771].

## 1. Definition and Variational Formulation

Let $G=(V,E)$, $|V|=n$, and $x\in\mathbb{R}^n$ with $\|x\|_1=\sum_{v\in V}|x_v|=1$, $\sum_{v\in V}x_v=0$. The $\ell_1$-Fiedler value is:
\[
b(G) = \min_{x\in \mathbb{R}^n}\left\{
  \sum_{uv\in E} |x_u - x_v|\,:\,
  \sum_{v\in V} x_v=0,\;\|x\|_1=1
\right\}\,.
\]
Any optimal $x^*$ is called an $\ell_1$-Fiedler vector. This minimization can alternatively be formulated using nonnegative vectors $x^1,x^2$ with orthogonal supports and equal sums:
\[
b(G) = \min_{\substack{
  x^1,x^2\geq 0,\\
  (x^1)^Tx^2=0,\\
  \sum x^1 = \sum x^2 = 1/2
}}
\sum_{uv\in E} |x^1_u - x^2_u - x^1_v + x^2_v|\,.
\]
This variational principle establishes $b(G)$ as a combinatorial analogue to the second Laplacian eigenvalue $a(G)=\lambda_2(L_G)$, substituting the quadratic $\ell_2$ smoothing with $\ell_1$ total variation [2306.13032], [2601.05771].

## 2. Connectivity and Sparsest-Cut Equivalence

A central property (Theorem 2) is that $b(G)>0$ if and only if $G$ is connected. If $G$ is disconnected, a feasible $x$ supported on a component yields zero edge contributions, so $b(G)=0$. When $G$ is connected, any nonzero $x$ forces at least one edge to have nonzero difference [2306.13032].

Crucially, $b(G)$ has an explicit combinatorial characterization as the edge-density of the sparsest cut:
\[
b(G) = \frac{n}{2} \min_{\substack{S\subset V\\S\neq\emptyset,\,S\neq V}} \rho(S)
\qquad\text{where}\qquad
\rho(S) = \frac{|\partial S|}{|S|(n-|S|)}
\]
If a set $S^*$ achieves the minimum, the corresponding $\ell_1$-Fiedler vector is:
\[
x_v = 
  \begin{cases}
    1/(2\,|S^*|) & v\in S^* \\
    -1/(2(n-|S^*|)) & v\notin S^*
  \end{cases}
\]
This equivalence directly connects $b(G)$ to the classic sparsest-cut problem and leverages cut-based expansion tools for analysis [2306.13032], [2601.05771].

## 3. Fundamental Inequalities and Bounds

Multiple sharp inequalities relate $b(G)$ to spectral, degree, and isoperimetric invariants:
- $\frac{1}{2} a(G) \leq b(G) \leq \frac{1}{2} \lambda_1(L_G)$, where $a(G)$ is the algebraic connectivity and $\lambda_1(L_G)$ the largest Laplacian eigenvalue.
- $b(G) \leq \sqrt{m\,a(G)}$ for $m=|E|$.
- Vertex-degree bound: $b(G) \leq \frac{n}{2(n-1)}\,\delta$, with $\delta$ the minimum degree.
- Isoperimetric lower bound: $b(G) \geq i(G)$, $i(G)=\min_{S:0<|S|\leq n/2} |\partial S|/|S|$ (Cheeger constant).

Nordhaus-Gaddum inequalities (Theorem 3.7 [2601.05771]) constrain the sum of $b(G)$ and $b(G^c)$:
\[
\frac{1}{2} < b(G) + b(G^c) \leq \frac{n}{2}
\]
with equality characterized by complete graphs and limiting cases for stars and their complements [2601.05771].

## 4. Explicit Formulas and Extremal Constructions for Trees

For trees $T$, $b(T)$ admits a closed formula:
\[
b(T) = \frac{1}{2}\left(\frac{1}{|V_u|} + \frac{1}{|V_v|}\right) = \frac{n}{2|V_u||V_v|}
\]
where $uv$ is a center-edge minimizing the partition size differential. Key special cases:
- Paths: $b(P_n)=2/n$ if $n$ is even; $b(P_n)=2n/(n^2-1)$ if $n$ is odd.
- Stars: $b(S_n)=\frac{1}{2} + 1/[2(n-1)]$.

Extremal tree results (Theorem 4.10 [2601.05771]) show the star graph $S_n$ globally maximizes $b(T)$ among $n$-vertex trees; the path $P_n$ minimizes it. Prescribed tree parameters (diameter $D$, maximum degree $\Delta$, number of pendant vertices $p$) yield construction schemes and explicit bounds for $b(T)$ [2601.05771].

## 5. Laplacian and Edge-Connectivity Connections

Given $S$ as the sparsest-cut subset and the induced $\ell_1$-Fiedler vector $x$, summations of Laplacian products yield
\[
\sum_{u\in S} (Lx)_u = b(G),\qquad
\sum_{u\in S^c} (Lx)_u = -b(G)
\]
If the edge-connectivity of $G$ is $\lambda(G)=k$, then
\[
b(G)\leq \frac{n\,k}{2(n-1)}
\]
Equality holds if and only if $G$ may be constructed by adding $k$ edges incident solely to an isolated vertex such that at each intermediate step, the singleton induces the unique sparsest cut [2601.05771].

## 6. Addition of Pendant Vertices

Successively attaching $k$ pendant vertices to $G$ yields
\[
b(G^k)\leq b(G)\prod_{i=0}^{k-1}\left(1-\frac{1}{(n+i)^2}\right)
\]
Each pendant attachment reduces $b$ multiplicatively by at most $1-1/m^2$ ($m$ the number of vertices at step $i$) [2601.05771].

## 7. Relationship with Isoperimetric Number and Cheeger-Type Bounds

The isoperimetric number $\mathrm{iso}(G)=\min_{S:\,|S|\leq\lfloor n/2\rfloor} |\partial S|/|S|$ upper bounds $b(G)$:
\[
b(G)\leq \mathrm{iso}(G)
\]
Combining with known bounds for $\mathrm{iso}(G)$, such as Mohar's $\mathrm{iso}(G)\leq\sqrt{a(G)\,(2d_{\max}-a(G))}$, one obtains
\[
b(G)\leq \sqrt{a(G)\left(2d_{\max}-a(G)\right)}
\]
For $r$-regular graphs, Cheeger’s inequality in the $\ell_1$ setting gives $b(G)\geq \frac{1}{2}r\,h(G)=\frac{1}{2}\mathrm{iso}(G)$. When the minimum sparsest cut is of size $\lfloor n/2\rfloor$ and $n$ is even, $b(G)=\mathrm{iso}(G)$; singleton minimizers yield $b(G)=n/(2(n-1))=\mathrm{iso}(G)$ [2601.05771].

## 8. Computational Complexity and Norm Variants

Computing $b(G)$ and associated $\ell_1$-Fiedler vectors is NP-hard, as it is equivalent to the sparsest cut problem. In contrast, the $\ell_\infty$-Fiedler value
\[
\gamma(G)=\min\{\max_{uv\in E}|x_u-x_v|:\;\sum_v x_v=0,\;\|x\|_\infty=1\}
\]
admits a polynomial-time solution through $n$ linear programs. For paths, $\gamma(P_n)=2/(n-1)$ [2306.13032].

---

The $\ell_1$-Fiedler value $b(G)$ thus bridges spectral and combinatorial connectivity, underpinning sparsest cut duality, tree extremal constructions, edge-connectivity, and isoperimetric bounds in both classical and parameterized graph families. Its direct link to NP-hard optimization and Cheeger-type inequalities positions $b(G)$ as a central object in modern combinatorial spectral theory [2306.13032], [2601.05771].

Source: https://www.emergentmind.com/topics/l_1-fiedler-value