Holonomy-based Diagnostic of Strain Compatibility in Birefringence Imaging of Stress-induced Ferroelectric SrTiO$_3$

Abstract: We introduce a holonomy-based geometric diagnostic for birefringence-derived director fields and apply it to stress-induced ferroelectric SrTiO$_3$. Treating the director as a line field in $\mathbb{R}P^2$, we define a holonomy angle $ω$ from residual rotations accumulated along closed loops in real space and compare it with a conventional local-gradient metric. Whereas the gradient quantifies local orientational variation, $ω$ probes the global compatibility of rotations along closed paths. The resulting $ω$ map cannot be reproduced by simple coarse-graining of local gradients, indicating sensitivity to loop-level orientational incompatibility. Analysis of alignment of holonomy rotation axes reveals a cooling-induced reorganization of the electromechanical response, consistent with strain- or stress-related inhomogeneity above the ferroelectric transition and additional ordering below it. These results demonstrate holonomy as a loop-based geometric diagnostic of strain compatibility in orientational fields derived from birefringence.

• The paper introduces a holonomy-based diagnostic method to assess strain compatibility in stress-induced ferroelectric SrTiO3.
• It contrasts traditional local gradient measures with a closed-loop holonomy approach that captures non-integrable, path-dependent orientational features.
• The results reveal distinct spatial and temperature-dependent patterns that link strain incompatibility to domain evolution across phase transitions.

Holonomy-based Geometric Diagnostics for Strain Compatibility in Birefringence Imagery of Stress-induced Ferroelectric SrTiO$_3$

Introduction

Birefringence imaging enables the real-space reconstruction of strain and polarization textures in ferroic materials, capturing domains, walls, and emergent inhomogeneities relevant to device applications and fundamental phase-transition mechanisms. Conventional analysis predominantly relies on local gradient (e.g., “grad”) metrics to characterize orientational variation in the director (optical fast axis), which lacks sensitivity to global, path-dependent features such as strain incompatibility and non-integrability. This study introduces and implements a holonomy-based diagnostic for birefringence-derived line fields in ferroelectric SrTiO$_3$, conceptualizing the director as a field on $\mathbb{R}P^2$ and extracting a holonomy angle, $\omega$, as the residual rotation over closed spatial loops. The methodology is contrasted with local gradient measures and applied across the cubic-to-tetragonal and ferroelectric transitions, with spatial and temperature dependence resolved at high fidelity.

Holonomy Formulation and Implementation

The director field is reconstructed from polarization-resolved birefringence imaging, taking the Stokes vector at each pixel and defining a normalized director on $\mathbb{R}P^2$ (director $\sim$ $-$director). For each closed loop of size $L \times L$, the minimal SU(2) rotations between adjacent pixels are composed into a loop quaternion, $Q_L$, whose axis-angle form yields the holonomy angle $\omega_L$ as

$_3$0

where $_3$1 and $_3$2. This definition isolates strictly nonintegrable (path-dependent) orientational features; fields with finite gradients but globally compatible rotations exhibit vanishing holonomy by construction.

Application to 575 nm birefringence images of SrTiO$_3$3, stressed along [001], employs $_3$4 as the nominal loop size, matching typical domain wall and flexoelectric texture scales. The holonomy maps $_3$5 are aggregated in 5-K temperature intervals and compared to corresponding local gradient maps and ferroelectric transition temperature ($_3$6) fields derived from previous work.

Figure 1: Real-space maps (302×140 pixels) of (a) the ferroelectric transition temperature $_3$7, (b) the holonomy angle $_3$8, and (c) the nearest-neighbor $_3$9 edge-angle variation, for the (40.0K, 45.0K] interval.

Comparison with Local Measures and Statistical Characterization

The holonomy-based metric exhibits both spatial and statistical distinctiveness when compared to local gradient-based analysis. High-holonomy values are confined within clusters embedded in regions of elevated local gradient, but the two do not strictly overlap (intersection-over-union for top 10% pixels: 0.391), and correlation is moderate but far from deterministic for large values (Pearson $\mathbb{R}P^2$0 for upper 10%). Holonomy reveals loop-level geometric incompatibility hidden within domains that appear as simple rapid orientational change by conventional (grad) measures.

The temperature evolution of holonomy is anomalous at both the cubic-tetragonal ($\mathbb{R}P^2$1) and ferroelectric ($\mathbb{R}P^2$2) transitions, with statistically significant peaks in the median of the top 10% of holonomy values per window. These responses are highly localized within the real-space map, with a majority of pixels near zero holonomy, underscoring the ability of holonomy to serve as a selective and sensitive diagnostic of loop-level director incompatibility.

Figure 2: Temperature dependences for $\mathbb{R}P^2$3 of (a) the holonomy angle $\mathbb{R}P^2$4 and (b) the axis-alignment order parameter $\mathbb{R}P^2$5 for all pixels and the upper 10% subset, revealing anomalies at $\mathbb{R}P^2$6 and $\mathbb{R}P^2$7.

Synthetic data tests reinforce the distinction between gradient and holonomy metrics. Integrable, smooth director fields present finite grad but zero holonomy. Domain walls and vortex-like defects (topological or geometric incompatibilities) yield locally intense, spatially confined holonomy peaks exactly where the closed-loop transport detects incompatibility.

Axis-alignment Structure and Cooling-induced Reorganization

The spatial organization of the holonomy rotation axes is probed using the maximum eigenvalue of the axis alignment tensor, yielding an order parameter $\mathbb{R}P^2$8 that quantifies axis alignment. The high-holonomy (top 10%) regions universally show reduced axis alignment compared to the full field, attesting to greater rotational disorder in these areas. Cooling through $\mathbb{R}P^2$9 and $\omega$0 produces a non-monotonic temperature dependence: axis disorder increases toward the transition, recovers partially in the ferroelectric regime, but fails to fully regain the high-$\omega$1 alignment, consistent with the absence of a single-domain state.

The spatial map of the change in alignment order, $\omega$2, between reference high-$\omega$3 and lower-$\omega$4 windows exposes marked reorganization during cooling. Above $\omega$5, stripe-like features reflect underlying stress and strain inhomogeneity. Below $\omega$6, the spatial structure shifts, reflecting the superposition of domain-related configurational reordering onto strain-driven inhomogeneity.

Figure 3: Maps of the change in the axis-alignment order parameter, $\omega$7, for (a) 14K-wnd and (b) 42K-wnd, illustrating qualitative reorganization of axis order across and below the ferroelectric transition.

Systematic metrics for inter-window pattern reorganization (normalized RMSE, sign-flip rate, $\omega$8, Pearson correlation) clearly outline three temperature regimes: substantial pattern evolution above $\omega$9, a stable intermediate region, and reorganization at the ferroelectric transition. The composite $\mathbb{R}P^2$0 peaks at $\mathbb{R}P^2$1 and $\mathbb{R}P^2$2, designating these as the primary loci for real-space electromechanical pattern reconfiguration.

Figure 4: Temperature dependence of the $\mathbb{R}P^2$3, quantifying the degree of real-space pattern reorganization of $\mathbb{R}P^2$4 between adjacent temperature windows.

Implications and Future Directions

This work establishes holonomy as a robust geometric diagnostic of strain compatibility from birefringence-imaged director fields. Holonomy is strictly more discriminating than gradient magnitudes: it isolates path-dependent, loop-level incompatibility that could be associated with real, physical features such as inhomogeneous strain, flexoelectric effects, dislocations, or domain-wall arrangements. The analysis reveals both extended strain-related features and localized, domain-related rearrangements in the real-space texture associated with the electromechanical response across ferroelastic and ferroelectric transitions.

Practically, the approach admits straightforward extension to other ferroic systems and any real-space imaging context yielding director-like fields (e.g., nematic liquid crystals, domain engineering in oxides, patterned mesostructures). Unlike purely topological invariants, the holonomy angle described here is sensitive to geometric, non-topological incompatibility and can be applied in topologically trivial fields. This positions holonomy analysis as a general purpose, path-sensitive complement to local measures in orientational field imaging. A direction for development is deriving closer analytic connections between the holonomy of the optical director and the (in)compatibility tensors of elasticity theory.

Conclusion

Holonomy-based diagnostics extract loop-level (non-integrability) features from birefringence images of stress-induced ferroelectric SrTiO$\mathbb{R}P^2$5, enabling the identification of localized strain and electromechanical incompatibility not resolvable by gradient-based methods. The holonomy measure tracks reorganization of orientational structure across ferroelastic and ferroelectric transitions, mapping directly to spatial domains of strain and domain evolution. This work provides not only a geometric tool for analyzing director fields but also a substrate for future studies into the interplay of topology, elasticity, and electromechanics in complex ferroic materials.