Daily Summary for 2026-10-30

Introduction to the show: ident: Condensed Matter. The day in condensed matter: materials, phases and the measurements under them.

Kai: It's the thirtieth of October, twenty twenty-six, and this is the day's research.

Mira: 4 new papers came out today.

Kai: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We'll take the day in one pass, then pull out the papers we're staying with.

Mira: Start us off with Symmetry-Based Parameter Efficient Low-Rank Fine-Tuning for High-Throughput Calculation of Lattice Thermal Conductivity.

The day's research: Kai: We are focusing on tools for simulating complex physical systems with materials and memory structures today.

Mira: The key finding is the methodology for high-throughput calculation of lattice thermal conductivity.

Lev: That shows a way to dramatically speed up predicting material properties for thermoelectric devices.

Kai: This approach uses symmetry to fine-tune a machine learning potential much faster than tuning every compound individually.

Mira: This efficiency builds upon using symmetry-based few-shot fine-tuning on compounds sharing common crystal symmetry.

Lev: That fine-tuned potential allows for accurate lattice thermal conductivity calculations across much larger datasets.

Kai: It shows improvements of two to three times over the non-fine-tuned version.

Mira: Furthermore, it reduces computation time by five to seven times for a hundred compounds.

Lev: Another piece investigates memory dynamics with sequential retrieval in an n-body asymmetric dense associative memory model.

Kai: That looks at how sequential retrieval works where the state settles into a stable limit-cycle attractor.

Mira: This analysis uses generating functional techniques to show that overlap dynamics follow an Amari-Maginu-type signal-to-noise analysis.

Lev: The work also notes that for interaction order three, normalized storage capacity decreases as the interaction order increases compared to symmetric models.

Kai: Finally, there is work examining collective flows in active tissue mechanics with uniaxially compressed 3D spheroids.

Mira: This study uses a minimal three-dimensional active vertex model to see how tissue mechanics and activity shape responses to uniaxial compression.

Lev: The findings link the activity-dependent reversal of speed response and compression-induced coherent rotation directly to contact dynamics.

Kai: We have two papers: Sequential Retrieval in Dense Associative Memory and Symmetry-Based Parameter Efficient Low-Rank Fine-Tuning for High-Throughput Calculation of Lattice Thermal Conductivity.

Mira: The first paper studies how patterns are retrieved one by one from a dense memory model using shifting indices.

Lev: That method uses symmetry to quickly fine-tune a machine learning potential for calculating thermal conductivity across many materials.

Kai: And the third paper examines how cell activity and compression shape the movement and rotation of cells in tissue spheroids.

Mira: We examine how cell activity and compression shape the movement and rotation of cells in tissue spheroids.

Lucky paper: 2610.29634: Kai: Okay, so we're digging into the Symmetry-Based Parameter Efficient Low-Rank Fine-Tuning for High-Throughput Calculation of Lattice Thermal Conductivity paper now.

Mira: This paper is about accurate and efficient prediction of lattice thermal conductivity for materials used in thermoelectric and photovoltaic applications.

Lev: The core idea is using symmetry to fine-tune the MatterSim 1M universal machine-learned interatomic potential, which then lets you calculate lattice thermal conductivity across large datasets much faster.

Kai: So the key claim here is that this symmetry-based few-shot fine-tuning approach yields LTC prediction accuracy comparable to tuning for each compound individually, and even two to three times higher than the non-fine-tuned version.

Mira: That speedup comes from using a small set of compounds sharing a common crystal lattice symmetry for the initial fine-tuning.

Lev: The paper details that they used three types of compounds: half-Heusler, materials with space group two hundred twenty-five and spinels with space group two hundred twenty-seven.

Kai: And for each dataset, they created a training subset of about twenty compounds, which is then used to form the configuration set by joining the configurations related to each compound in that subset.

Mira: That process of fine-tuning on that subset is what they refer to as SBFS-FT.

Lev: They selected the LoTR method because it achieved a Tanimoto index of zero point eight four three for phonon density of states prediction, and it outperformed specific fine-tuning in computational efficiency across all three datasets by about one point five to two times less time.

Kai: What about the accuracy metric? They used the mean absolute relative difference between the MTP and MatterSim LTC values as a measure of how accurate they were.

Mira: The SBFS-FT reduced that mean absolute relative difference between MTP and MatterSim LTC values by about three times, getting accuracies comparable to those from specific fine-tuning.

Lev: That improved accuracy extended to compounds not even in the training subset, which suggests good generalization of the model.

Kai: So for a hundred compounds, they're projecting a speedup of approximately five to six times for each dataset using this SBFS-FT strategy.

Mira: It seems like an optimal balance between accuracy and computational cost for high-throughput LTC calculations is what they are showing here in the Symmetry-Based Parameter Efficient Low-Rank Fine-Tuning for High-Throughput Calculation of Lattice Thermal Conductivity paper.

Lucky paper: 2610.29931: Kai: Okay, let's look at paper number two of two today, which is Two-stage Slowdown of Vacancy-mediated Diffusion in a Lattice Gas Model with Quenched Disorder and Spins.

Mira: This model is defined on a squared lattice of size L with an energy given by E equals X J i n j s p i s p j under the constraint P i n i equals L two rho.

Lev: The interaction J i j is drawn randomly from a bimodal distribution with equal probability and is spatially quenched for bonding between nearest neighbors.

Kai: So, occupation of a lattice site is either by a particle or a vacancy, indicated by P n i n i equals one or P n i n i equals zero.

Mira: Particle hopping is allowed when one of its nearest neighbors is occupied by a vacancy.

Lev: We see a two-stage slowdown in both the mean squared displacement A r squared I (t) and the spatial overlaps Q(t).

Kai: Those plateaus are due to anti-correlation at the corresponding length scale.

Mira: This hierarchical dynamics is further corroborated by strong anti-correlations at the corresponding length scale where the plateau emerges.

Lev: The ensemble-averaged MSD defined as A r I (t) minus A r I (zero), where the brackets indicate thermal average and the bar indicates disorder average, shows this.

Kai: You also need to carefully magnify the relaxation curve of particles to see those two-stage slowdown details in Fig two a.

Mira: The motion of a typical vacancy first, we look at Delta r (t) and its two components Delta X (t) and Delta Y (t).

Lev: For short timescales of order one x ten-six the vacancies are trapped in relative-small regions bouncing back and forth.

Kai: And then for the timescale of unity, the vacancy can escape that first type of localization but gets trapped again before traveling between those two traps.

Mira: Those motions of Type II are responsible for the second slowdown we see in both MSD and Q(t).

Lev: The displacement correlation function C(k tau) is defined by Delta r I (zero) delta r I (k tau), where delta r I (k tau) equals r I

(k plus one) tau: minus r I (k tau).

Kai: Fig four a and b plot the normalized C(k tau) at Tau equals two x ten-six and Tau equals one.

Mira: The two time intervals are chosen to approximately match the timescale of the two slowdown regimes of MSD and Q(t) at T equals zero point two two.

Lev: The strong anti-correlations at T equals zero point two two for both Tau echo the picture in real space, meaning traps at different length scale.

Kai: This suggests vacancies' normal diffusion happens in the large time limit then.

Mira: The hierarchical relaxation exhibited by this model appears naturally consistent with the rugged potential energy landscape of Stillinger It is attempting to associate hopping motion at different length scale to intrabasin and intercrater diffusion on the PEL.

Lev: The emergence of glassy dynamics induced by frustration and crowdedness corroborates the idea of existence of deeper connection between spin glasses and other general glassy systems.

Kai: So, it’s a lattice gas model with spatially-quenched interactions and particle-quenched spins that shows this two-stage slowdown for diffusion and structural relaxation.