arXiv subjects

Swastik Bhattacharya

Publications and source records attributed to Swastik Bhattacharya.

At least 19 recordsLinked to original sources

arXiv2026

Complexity Growth in Black Holes: A Comparison of the Volume and Action Proposals

In this article, we investigate the late-time growth of holographic complexity, defined via the complexity-volume (CV) and complexity-action (CA) prescriptions, for BTZ, Schwarzschild, Reissner-Nordström, and Kerr black holes. Extending previous analyses beyond asymptotically AdS spacetimes, we include asymptotically flat geometries and employ the CV and CA prescriptions as comparative geometric diagnostics of black hole interior dynamics. In all cases considered, the complexity growth rate is governed by horizon thermodynamic data and scales with $T_H S_H$. While the CV prescription exhibits geometry-dependent proportionality constants, the CA prescription yields a universal thermodynamic scaling across all black holes studied, including non-AdS cases. We further analyze variations in the complexity growth rate, $δ\dot{\mathcal{C}}$, under physical processes such as the Penrose process, superradiance, and particle accretion. We find that $δ\dot{\mathcal{C}}$ exhibits non-trivial behavior: it increases under the Penrose process and superradiance, while under particle accretion it can increase, remain unchanged, or decrease depending on the angular momentum of the infalling particle. In quasi-equilibrium regimes, the variation in complexity closely tracks the behavior of the horizon area and interior volume growth, whereas out-of-equilibrium processes render it sensitive to angular momentum transfer and may lead to negative values within an equilibrium approximation. This behavior highlights the limitations of equilibrium-based treatments and motivates a fully dynamical analysis incorporating horizon stresses and transient hair.

gr-qc↗

arXiv2026

Temperature Fluctuations and quantum corrections near Black Hole Horizon

Spatially varying near-horizon fluctuations of temperature of a Schwarzschild Black Hole is considered within the Euclidean Gravity approach. We present evidence that suggests that such fluctuations in temperature are closely related with the near-horizon supertranslations of a Black Hole. This allows one to express the temperature-dependent corrections to the near-horizon part of the Euclidean gravity action or Free energy in terms of polynomial functionals of a near-horizon supertranslation for all orders. The leading order term turns out to be proportional to the near-horizon supertranslation charge. We also show that this same term results from a microscopic state counting of the near-horizon field with the constraint of near-horizon supertranslation symmetry imposed. The constructed near-horizon partition function provides a physically appealing intuitive description of the low-energy part of horizon physics as a sum over all possible near-horizon supertranslations. This suggests a dual description of near-horizon physics in terms of alternate variables. The implications of these results are briefly discussed.

gr-qc↗

arXiv2025

COBRA: Catastrophic Bit-flip Reliability Analysis of State-Space Models

State-space models (SSMs), exemplified by the Mamba architecture, have recently emerged as state-of-the-art sequence-modeling frameworks, offering linear-time scalability together with strong performance in long-context settings. Owing to their unique combination of efficiency, scalability, and expressive capacity, SSMs have become compelling alternatives to transformer-based models, which suffer from the quadratic computational and memory costs of attention mechanisms. As SSMs are increasingly deployed in real-world applications, it is critical to assess their susceptibility to both software- and hardware-level threats to ensure secure and reliable operation. Among such threats, hardware-induced bit-flip attacks (BFAs) pose a particularly severe risk by corrupting model parameters through memory faults, thereby undermining model accuracy and functional integrity. To investigate this vulnerability, we introduce RAMBO, the first BFA framework specifically designed to target Mamba-based architectures. Through experiments on the Mamba-1.4b model with LAMBADA benchmark, a cloze-style word-prediction task, we demonstrate that flipping merely a single critical bit can catastrophically reduce accuracy from 74.64% to 0% and increase perplexity from 18.94 to 3.75 x 10^6. These results demonstrate the pronounced fragility of SSMs to adversarial perturbations.

cs.CR↗

arXiv2025

SafeCiM: Investigating Resilience of Hybrid Floating-Point Compute-in-Memory Deep Learning Accelerators

Deep Neural Networks (DNNs) continue to grow in complexity with Large Language Models (LLMs) incorporating vast numbers of parameters. Handling these parameters efficiently in traditional accelerators is limited by data-transmission bottlenecks, motivating Compute-in-Memory (CiM) architectures that integrate computation within or near memory to reduce data movement. Recent work has explored CiM designs using Floating-Point (FP) and Integer (INT) operations. FP computations typically deliver higher output quality due to their wider dynamic range and precision, benefiting precision-sensitive Generative AI applications. These include models such as LLMs, thus driving advancements in FP-CiM accelerators. However, the vulnerability of FP-CiM to hardware faults remains underexplored, posing a major reliability concern in mission-critical settings. To address this gap, we systematically analyze hardware fault effects in FP-CiM by introducing bit-flip faults at key computational stages, including digital multipliers, CiM memory cells, and digital adder trees. Experiments with Convolutional Neural Networks (CNNs) such as AlexNet and state-of-the-art LLMs including LLaMA-3.2-1B and Qwen-0.3B-Base reveal how faults at each stage affect inference accuracy. Notably, a single adder fault can reduce LLM accuracy to 0%. Based on these insights, we propose a fault-resilient design, SafeCiM, that mitigates fault impact far better than a naive FP-CiM with a pre-alignment stage. For example, with 4096 MAC units, SafeCiM reduces accuracy degradation by up to 49x for a single adder fault compared to the baseline FP-CiM architecture.

cs.AR↗

arXiv2025

GenBFA: An Evolutionary Optimization Approach to Bit-Flip Attacks on LLMs

Large Language Models (LLMs) have revolutionized natural language processing (NLP), excelling in tasks like text generation and summarization. However, their increasing adoption in mission-critical applications raises concerns about hardware-based threats, particularly bit-flip attacks (BFAs). BFAs, enabled by fault injection methods such as Rowhammer, target model parameters in memory, compromising both integrity and performance. Identifying critical parameters for BFAs in the vast parameter space of LLMs poses significant challenges. While prior research suggests transformer-based architectures are inherently more robust to BFAs compared to traditional deep neural networks, we challenge this assumption. For the first time, we demonstrate that as few as three bit-flips can cause catastrophic performance degradation in an LLM with billions of parameters. Current BFA techniques are inadequate for exploiting this vulnerability due to the difficulty of efficiently identifying critical parameters within the immense parameter space. To address this, we propose AttentionBreaker, a novel framework tailored for LLMs that enables efficient traversal of the parameter space to identify critical parameters. Additionally, we introduce GenBFA, an evolutionary optimization strategy designed to refine the search further, isolating the most critical bits for an efficient and effective attack. Empirical results reveal the profound vulnerability of LLMs to AttentionBreaker. For example, merely three bit-flips (4.129 x 10^-9% of total parameters) in the LLaMA3-8B-Instruct 8-bit quantized (W8) model result in a complete performance collapse: accuracy on MMLU tasks drops from 67.3% to 0%, and Wikitext perplexity skyrockets from 12.6 to 4.72 x 10^5. These findings underscore the effectiveness of AttentionBreaker in uncovering and exploiting critical vulnerabilities within LLM architectures.

cs.CR↗

arXiv2024

Bayesian evaluation of hadron-quark phase transition models through neutron star observables in light of nuclear and astrophysics data

We investigate the role of hybrid and nucleonic equations of state (EOSs) within neutron star (NS) interiors using Bayesian inference to evaluate their alignment with recent observational data from NICER and LIGO-Virgo (LV) collaborations. We find that smooth hybrid EOSs are slightly favoured in explaining NS mass-radius relations, particularly for pulsars such as PSR J0030+0451 and PSR J0740+6620. However, this preference is not definitive, as gravitational wave (GW) data does not significantly differentiate between our hybrid and nucleonic models. Our analysis also reveals tensions between older NICER data and recent measurements for PSR J0437-4715, highlighting the need for more flexible EOS models. Through two sampling approaches - one fixing the hadronic EOS set and the other without fixing the same, we demonstrate that the hybrid EOS model can incorporate stiffer EOSs, resulting in a better agreement with NICER data but leading to higher tidal deformability, which is less consistent with GW observations. In some recent publications a parameter $d_c$, related to the trace anomaly and its derivative, is used to indicate the presence of deconfined quark matter. We find that our hadronic model, which does not include phase transition to deconfined matter, under the influence of imposed constraints, is able to predict values below 0.2 for $d_c$ at around five times saturation density. The hybrid model goes below this threshold at lower densities under the same conditions.

nucl-th↗

arXiv2024

Sb2Se3 and SbBiSe3 Surface Capping and Biaxial Strain Co-Engineering for Tuning the Surface Electronic Properties of Bi2Se3 Nanosheet- A Density Functional Theory based Investigation

In this work, for the first time, a density functional theory (DFT) based comprehensive theoretical study is performed on the surface electronic properties of Bi2Se3 nanosheet in the presence of a surface capping layer as well as mechanical strain. The study systematically introduces a biaxial compressive and tensile strain up to 5% in natural, Sb2Se3 surface capped, and SbBiSe3 surface capped Bi2Se3, and the subsequent effects on the electronic properties are assessed from the surface energy band (E-k) structure, the density of states (DOS), band edge energy and bandgap variations, surface conducting state localization, and Fermi surface spin-textures. The key findings of this work are systematically analyzed from conducting surface state hybridization through bulk in the presence of surface capping layers and applied biaxial strain. The result demonstrates that the interplay of surface capping and strain can simultaneously tune the surface electronic structure, spin-momentum locking results from change in electronic localization and interactions. In essence, this work presents an extensive theoretical and design-level insight into the surface capping and biaxial strain co-engineering in Bi2Se3, which can potentially facilitate different topological transport for modern optoelectronics, spintronics, valleytronics, bulk photovoltaics applications of engineered nanostructured topological materials in the future.

cond-mat.mtrl-sci↗

arXiv2024

Analysis of Neutron Star $f-$mode Oscillations in General Relativity with Spectral Representation of Nuclear Equations of State

We study quasinormal $f-$mode oscillations in neutron star(NS) interiors within the linearized General Relativistic formalism. We utilize approximately 9000 nuclear Equations of State (EOS) using spectral representation techniques, incorporating constraints on nuclear saturation properties, chiral Effective Field Theory ($χ$EFT) for pure neutron matter, and perturbative Quantum Chromodynamics (pQCD) for densities pertinent to NS cores. The median values of f-mode frequency, $ν_f$ (damping time, $τ_f$) for NS with masses ranging from 1.4 - 2.0 $M_\odot$ lie between 1.80 - 2.20 kHz (0.13 - 0.22 s) for our entire EOS set. Our study reveals a weak correlation between $f-$mode frequencies and individual nuclear saturation properties, prompting the necessity for more intricate methodologies to unveil multi-parameter relationships. We observe a robust linear relationship between the radii and $f-$mode frequencies for different NS masses. Leveraging this correlation alongside NICER observations of PSR J0740+6620 and PSR J0030+0451, we establish constraints that exhibit partial and minimal overlap for observational data from Riley et al. and Miller et al. respectively with our nucleonic EOS dataset. Moreover, NICER data aligns closely with radius and frequency values for a few hadron-quark hybrid EOS models. This indicates the need to consider additional exotic particles such as deconfined quarks at suprasaturation densities. We conclude that future observations of the radius or $f-$mode frequency for more than one NS mass, particularly at the extremes of viable NS mass scale, would either rule out nucleon-only EOS or provide definitive evidence in its favour.

astro-ph.HE↗

arXiv2023

Surface Gravity of Dynamical Horizons: A Causal Perspective

We consider marginally trapped surfaces in a spherically symmetric spacetime evolving due to the presence of a perfect fluid in D-dimensions and look at the various definitions of the surface gravity for these marginally trapped surfaces. We show that using Einstein equations it is possible to simplify and obtain general formulae for the surface gravity in terms of invariant quantities defined at these marginally trapped surfaces like area radius, cosmological constant and principal values of the energy-momentum tensor \r{ho}, p. We then correlate these expressions of surface gravity to the cases of dynamical horizons and timelike tubes and find which proposals of surface gravity are causally sensitive as these surfaces undergo causal transitions from spacelike to timelike and vice versa.

gr-qc↗

arXiv2023

Microscopic Model building for Black Hole Membranes from Constraints of Symmetry

Einstein equations projected on black-hole horizons give rise to the equations of motion of a viscous fluid. This suggests a way to understand the microscopic degrees of freedom on the black-hole horizon by focusing on the physics of this fluid. In this talk, we shall approach this problem by building a crude microscopic model for the Horizon-fluid(HF) corresponding to asymptotically flat black-holes in 3+1 dimensions. The symmetry requirement for our model is that it should incorporate the S1 diffeo-symmetry on the black-hole horizon. The second constraint comes from the demand that the correct value of the Coefficient of the Bulk Viscosity of the HF can be deduced from the model. Both these requirements can be satisfied by an adoption of the eight vertex Baxter model on a S2 surface. We show that the adiabatic entropy quantisation proposed by Bekenstein also follows from this model. Finally, we argue the results obtained so far suggest that a perturbed black-hole can be described by a CFT perturbed by relevant operators and discuss the physical implications.

gr-qc↗

arXiv2022

Effective field theory description of horizon-fluid determines the scrambling time

Black hole horizons interact with external fields when matter-energy falls through them. Such non-stationary black hole horizons can be described using viscous fluid equations. This work attempts to describe this process using effective field theory methods. Such a description can provide important insights beyond classical black hole physics. In this work, we construct a low-energy effective field theory description for the horizon fluid of a 4-dimensional, asymptotically flat, Einstein black hole. The effective field theory of the dynamical horizon has two ingredients: degrees of freedom involved in the interaction with external fields and symmetry. The dual requirements of incorporating near-horizon symmetries (${\cal S}1$ diffeomorphism) and possessing length scales due to external perturbations is naturally satisfied if the theory on the non-stationary black hole horizon is a deformed Conformal Field Theory (CFT). For the homogeneous external perturbations, at the lowest order, this leads to $(2 + 1)-$dimensional massive scalar field where the mass is related to the extent of the deformation of the CFT. We determine the mass by obtaining the correlation function corresponding to the effective field and relating it to the bulk viscosity of the horizon fluid. Additionally, we show that the coefficient of bulk viscosity of the horizon fluid determines the time required for black holes to scramble. Furthermore, we argue that matter-field modes with energy less than $m_{\rm eff}$ falling into the horizon thermalize more slowly. Finally, we construct a microscopic toy model for the horizon fluid that reduces to the effective field theory with a single scalar degree of freedom. We then discuss the usefulness of the effective field model in understanding how information escapes from a black hole at late times.

gr-qc↗

arXiv2018

Is there an upper bound on the size of a black-hole?

According to the third law of Thermodynamics, it takes an infinite number of steps for any object, including black-holes, to reach zero temperature. For any physical system, the process of cooling to absolute zero corresponds to erasing information or generating pure states. In contrast with the ordinary matter, the black-hole temperature can be lowered only by adding matter-energy into it. However, it is impossible to remove the statistical fluctuations of the infalling matter-energy. The fluctuations lead to the fact the black-holes have a finite lower temperature and, hence, an upper bound on the horizon radius. We make an estimate of the upper bound for the horizon radius which is curiosly comparable to Hubble horizon. We compare this bound with known results and discuss its implications.

gr-qc↗

arXiv2017

Black-hole event horizons-Teleology and Predictivity

General Relativity predicts the existence of black-holes. Access to the complete space-time manifold is required to describe the black-hole. This feature necessitates that black-hole dynamics is specified by future or teleological boundary condition. Here we demonstrate that the statistical mechanical description of black-holes, the raison d'etre behind the existence of black-hole thermodynamics, requires teleological boundary condition. Within the fluid-gravity paradigm --- Einstein's equations when projected on space-time horizons resemble Navier-Stokes equation of a fluid --- we show that the specific heat and the coefficient of bulk viscosity of the horizon-fluid are negative only if the teleological boundary condition is taken into account. We argue that in a quantum theory of gravity, the future boundary condition plays a crucial role. We briefly discuss the possible implications of this at late stages of black-hole evaporation.

gr-qc↗

arXiv2017

Negative specific heat of black-holes from Fluid-Gravity Correspondence

Black-holes in asymptotically flat space-times have negative specific heat --- they get hotter as they loose energy. A clear statistical mechanical understanding of this has remained a challenge. In this work, we address this issue using fluid-gravity correspondence which aims to associate fluid degrees of freedom to the horizon. Using linear response theory and the teleological nature of event horizon, we show explicitly that the fluctuations of the horizon-fluid lead to negative specific heat for Schwarzschild black Hole. We also point out how the specific heat can be positive for Kerr-Newman or AdS black holes. Our approach constitutes an important advance as it allows us to apply canonical ensemble approach to study thermodynamics of asymptotically flat black-hole space-times.

gr-qc↗

arXiv2016

Hints of quantum gravity from the horizon fluid

For many years researchers have tried to glean hints about quantum gravity from black hole thermodynamics. However, black hole thermodynamics suffers from the problem of Universality --- at leading order, several approaches with different microscopic degrees of freedom lead to Bekenstein-Hawking entropy. We attempt to bypass this issue by using a minimal statistical mechanical model for the horizon fluid based on Damour-Navier-Stokes (DNS) equation. For stationary asymptotically flat black hole spacetimes in General Relativity, we show explicitly that at equilibrium the entropy of the horizon fluid is the Bekenstein-Hawking entropy. Further we show that, for the bulk viscosity of the fluctuations of the horizon fluid to be identical to Damour, a confinement scale exists for these fluctuations, implying quantization of the horizon area. The implications and possible mechanisms from the fluid point of view are discussed.

gr-qc↗

arXiv2016

Statistical modeling of the fluid dual to Boulware-Deser Black hole

In this work we study the statistical and thermodynamic properties of the horizon fluid corresponding to the Boulware-Deser (BD) black hole of Einstein-Gauss-Bonnet (EGB) gravity. Using mean field theory, we show explicitly that the BD fluid exhibits the coexistence of two phases, a BEC and a non-condensed phase corresponding to the Einstein term and the Gauss-Bonnet term in the gravity action, respectively. In the fluid description, the high-energy corrections associated to Gauss-Bonnet gravity are modeled as excitations of the fluid medium. We provide statistical modeling of the excited part of the fluid and explicitly show that it is characterized by a generalized dispersion relation which in $D=6$ dimensions corresponds to a non-relativistic fluid. We also shed light on the ambiguity found in the literature regarding the expression of the entropy of the horizon fluid. We provide a general prescription to obtain the entropy and show that it is indeed given by Wald entropy.

gr-qc↗

arXiv2016

Fluctuations in horizon-fluid lead to negative bulk viscosity

Einstein equations projected on to a black hole horizon gives rise to Navier-Stokes equations. Horizon-fluids typically possess unusual features like negative bulk viscosity and it is not clear whether a statistical mechanical description exists for such fluids. In this work, we provide an explicit derivation of the Bulk viscosity of the horizon-fluid based on the theory of fluctuations a la Kubo. The main advantage of our approach is that our analysis remains for the most part independent of the details of the underlying microscopic theory and hence the conclusions reached here are model independent. We show that the coefficient of bulk viscosity for the horizon-fluid matches exactly with the value found from the equations of motion for the horizon-fluid.

gr-qc↗

arXiv2015

How Emergent is Gravity?

General theory of relativity (or Lovelock extensions) is a dynamical theory; given an initial configuration on a space-like hypersurface, it makes a definite prediction of the final configuration. Recent developments suggest that gravity may be described in terms of macroscopic parameters. It finds a concrete manifestation in the fluid-gravity correspondence. Most of the efforts till date has been to relate equilibrium configurations in gravity with fluid variables. In order for the emergent paradigm to be truly successful, it has to provide a statistical mechanical derivation of how a given initial static configuration evolves into another. In this essay, we show that the energy transport equation governed by the fluctuations of the horizon-fluid is similar to Raychaudhuri equation and, hence gravity is truly emergent.

gr-qc↗