Identification and Optimization of Accurate Spin Models for Open-Shell Carbon Ladders with Matrix Product States

Open-shell nanographenes offer a controlled setting to study correlated magnetism emerging from $π$-electron systems. We analyze oligo(indenoindene) molecules, non-bipartite carbon ladders whose tight-binding spectra feature a gapped, weakly dispersing manifold of quasi-zero modes, and show that their low-energy properties can be effectively mapped onto an interacting set of spin-1/2 degrees of freedom. Using Density Matrix Renormalization Group simulations of the full Fermi-Hubbard model, we obtain their excitation spectra, entanglement profiles, and spin-spin correlations. We then construct optimized delocalized fermionic modes that act as emergent spins and show that their interactions are well described by a frustrated $J_1$-$J_2$ Heisenberg chain. This effective description clarifies how spin degrees of freedom arise and interact in non-bipartite nanographene ladders, providing a compact and accurate representation of their correlated behavior.

💡 Research Summary

This paper investigates the magnetic properties of a class of non‑bipartite nanographene ladders known as oligo(indenoindene) (OInIn) molecules. These structures consist of alternating pentagon and hexagon rings, which give rise to a tight‑binding spectrum featuring a set of quasi‑zero‑energy modes whose number equals the number of pentagons, P. Because these modes are weakly dispersive and lie close to the Fermi level, at half‑filling each mode tends to be singly occupied even for moderate on‑site repulsion (U≈1.5 t). Consequently, the low‑energy sector can be mapped onto P effective spin‑½ degrees of freedom.

The authors first model the π‑electron system with the full Hubbard Hamiltonian
(H = -t\sum_{\langle i,j\rangle,\sigma}c^{\dagger}{i\sigma}c{j\sigma}+U\sum_i n_{i\uparrow}n_{i\downarrow})
using parameters t = 2.7 eV and U = 1.5 t, which are realistic for graphene‑based nanostructures. They then apply the Density Matrix Renormalization Group (DMRG) algorithm based on matrix product states (MPS) to obtain ground‑state and excited‑state energies, entanglement spectra, and spin–spin correlation functions for ladders with P = 2, 4, 6, 8 (up to 62 carbon sites). The DMRG results reveal that the energy gap between the lowest singlet and triplet alternates with P, consistent with a frustrated J₁‑J₂ Heisenberg chain in the regime J₂/|J₁| > ¼, where a Haldane‑dimer (valence‑bond‑solid) phase with AKLT‑type topological order is expected.

To construct an explicit mapping from the electronic degrees of freedom to the effective spins, the authors introduce delocalized fermionic modes
(c_{M(p),\sigma} = \sum_{i\in M(p)} \alpha^{(p)}i c{i\sigma})
where each set M(p) contains a subset of lattice sites associated with the p‑th effective spin. The coefficients α are optimized to maximize the “spin fidelity” (\langle 4(S_z^{M(p)})^2\rangle), i.e., the probability that the mode is singly occupied. Optimization is performed with the NLopt library, starting from a mode localized on the pentagon tip (site 7p). The resulting optimal modes have a robust spatial structure: roughly 40 % of the weight resides on the pentagon tip, while the remaining weight is distributed over six neighboring sites belonging to the adjacent hexagons. Including additional sites (the full ladder) raises the fidelity to 0.98, whereas a three‑site or eight‑site truncated mode already reaches 0.90–0.95. Importantly, the optimal mode shapes are essentially independent of the total system size, allowing the same set of coefficients obtained for a small ladder (e.g., P = 4) to be transferred to much larger ladders (P = 24) with negligible loss of fidelity.

Using these optimized modes, the authors define spin operators ( \mathbf{S}{M(p)} = \frac{1}{2} \sum{\sigma,\sigma’} c^\dagger_{M(p),\sigma},\boldsymbol{\tau}{\sigma\sigma’},c{M(p),\sigma’}) and evaluate spin–spin correlation functions in the electronic ground and excited states. They compare these correlations with those of the target frustrated Heisenberg chain (H_{\text{SC}} = J_1\sum_{p}\mathbf{S}p\cdot\mathbf{S}{p+1}+J_2\sum_{p}\mathbf{S}p\cdot\mathbf{S}{p+2}). By fitting the DMRG energy gaps, they extract effective couplings J₁ < 0 (ferromagnetic nearest‑neighbor) and J₂ > 0 (antiferromagnetic next‑nearest‑neighbor). The fitted values vary only slightly with P, and a single set of parameters (J₁* = −0.06 t, J₂* = 0.26 t) works well across all sizes, demonstrating the scalability of the effective model.

Entanglement entropy profiles of the electronic states mirror those of the spin chain, confirming that the mapping preserves the underlying quantum correlations. Moreover, the authors examine local spin polarization (\langle 4(S_z^i)^2\rangle) and find that pentagon tip sites exhibit slightly higher values (~0.66) than other carbon atoms (0.57–0.61), but the delocalized modes capture the majority of the magnetic moment, indicating that a simple site‑wise picture is insufficient.

The paper culminates with a systematic benchmark of four types of effective spin descriptions: (i) a naïve pentagon‑tip localized mode, (ii) a three‑site truncated mode, (iii) an eight‑site mode, and (iv) the full‑ladder mode. Correlation functions, magnetization, and spin‑flip fidelity are evaluated for representative eigenstates (ground singlet, first triplet, and a S = 2 Dicke‑like state). The eight‑site and full‑ladder modes reproduce the Heisenberg chain results with high accuracy, while the simpler modes capture the qualitative trends but miss quantitative details. The authors also introduce a “magnetization approximation ratio” (\bar r = \frac{1}{P}\sum_p \frac{M_p}{P/2}) to quantify overall performance; the full‑ladder mode reaches (\bar r \approx 0.98).

In summary, this work provides a comprehensive, multi‑scale methodology: (1) high‑precision DMRG simulations of the full Hubbard model for realistic nanographene ladders, (2) systematic construction and optimization of delocalized fermionic modes that act as emergent spin‑½ objects, and (3) validation that a frustrated J₁‑J₂ Heisenberg chain faithfully captures the low‑energy magnetic physics. The approach bridges the gap between ab‑initio electronic structure and compact spin models, offering a powerful framework for studying correlated magnetism in non‑bipartite carbon nanostructures and guiding the design of future spin‑based quantum devices.