The resolution hinges on a subtle but crucial distinction between two kinds of correlation functions computed in lattice QCD. Mesonic propagators, which track how quark-antiquark pairs propagate in Euclidean time, show a dramatic rearrangement above the chiral crossover: they fall into multiplets organized by newly emergent chiral spin and SU(4) symmetries. These symmetries indicate that the propagating degrees of freedom are chirally symmetric, essentially massless quarks tied together into color singlets by the chromoelectric confining string. When these full QCD mesonic correlators are compared with correlators computed on the same lattice with free, non-interacting quarks — the so-called free quark gas that represents an idealized quark-gluon plasma — the two are qualitatively completely different, especially for the scalar and pseudoscalar channels.

The conserved charge densities tell a strikingly different story. Quantities such as the conserved quark number density cannot propagate in time, but they do propagate in spatial directions, and it is their spatial propagators that determine the fluctuation cumulatives measured on the lattice and accessible to heavy-ion experiments. When Glozman compared the full QCD spatial correlators of the conserved charge densities at a temperature of 380 MeV with the corresponding free quark loop correlators, the two essentially coincided. Unlike the mesonic correlators, the conserved charges simply do not see the confining force. The effect sets in surprisingly early: already at about 193 MeV, soon above the chiral crossover, the difference between the full QCD and free quark gas charge density correlators is insignificant, which explains why the cumulants approach their free quark gas values around 220 MeV.

This dichotomy is the core of the new interpretation. The mesonic propagators, sensitive to the confining infrared physics, prove that the medium above the crossover is still in a confining regime — there are no deconfined gluons, and only color singlets propagate in time. Meanwhile, the conserved charge fluctuations are blind to that same confining force and therefore mimic free quarks. Fluctuation cumulants, in other words, are simply not a sensitive diagnostic of deconfinement, and their free-quark-like behavior should not be taken as proof that the quark-gluon plasma has formed.

Far from being an exotic novelty, this situation has a well-understood analogue at zero temperature: quark-hadron duality in electron-positron annihilation. In the process e+e- going to hadrons, the vacuum is fully confining and only color-singlet hadrons are registered in detectors, yet the inclusive cross-section at invariant masses above about 2 GeV is accurately reproduced by a free quark loop plus small perturbative corrections. Nobody concludes from that agreement that quarks are deconfined in the vacuum; it merely means that this particular observable is insensitive to the confining infrared contributions. Glozman argues that the analogy is direct: mesonic propagators above the crossover are sensitive to confinement, while fluctuations of conserved charges are not, in exactly the same way that inclusive annihilation rates are dual to free quarks while exclusive observables are not.

The analysis also situates the result within a broader three-regime picture of the QCD phase diagram at small baryon chemical potential. Below the chiral crossover temperature of roughly 155 MeV lies the hadron gas, in which pressure, energy density and entropy density scale with the number of colors Nc to the zeroth power. Between the crossover and roughly three times that temperature lies the stringy fluid, whose thermodynamic quantities scale as Nc — the signature of a medium built from color singlets rather than deconfined gluons. Only at higher temperatures does the true quark-gluon plasma emerge, with its Stefan-Boltzmann-like Nc-squared scaling and T to the fourth power behavior, which lattice studies show sets in only above roughly 400 to 500 MeV. The smooth transition out of the stringy fluid is most likely centered near 300 MeV, the deconfinement temperature of the pure glue theory, a value also supported by center vortex percolation calculations and the Hagedorn temperature of the string description.

To make sense of what the stringy fluid actually is microscopically, the paper draws on large-Nc model calculations that are manifestly confining yet chirally symmetric. In such models, chiral symmetry restoration above the crossover happens because thermal excitations of quarks and antiquarks Pauli-block the quark levels that are required for the existence of the chiral condensate. The same Pauli blocking causes mesons to swell to several times their vacuum size — becoming infinitely large in the chiral limit. The resulting medium is a densely packed, highly collective system of overlapping color-singlet meson-like objects with a very small mean free path for their constituents, and with no deconfined gluons anywhere in it.

Within this picture, the free-quark-like fluctuations of conserved charges acquire a natural explanation. Because the color-singlet clusters overlap so strongly, the Pauli principle forces quarks to be exchanged between neighboring clusters. These interchanges make the quarks quasi-free for the purposes of conserved charge transport across space, even though no colored quark ever propagates freely through time. Only the color singlets, in which massless quarks and antiquarks remain connected by the confining chromoelectric string, propagate in time. The free-quark-loop behavior of the spatial charge density propagator is, in this view, the dual manifestation of quark interchange dynamics in the dense stringy fluid.

The interpretation resolves several otherwise conflicting lattice observations into a single coherent framework. The apparent contradiction between free-quark-like cumulants and the confining evidence from mesonic correlators disappears once the different infrared sensitivities of the two observables are recognized. It is likewise consistent with the pion spectral function, which shows clean pion and pion-prime peaks above the crossover, and with the bottomonium spectrum, which remains essentially unchanged from its vacuum form below about 300 MeV even as the states broaden with temperature.

The proposed picture of the stringy fluid remains a hypothesis to be tested and potentially falsified in lattice simulations, and the paper outlines concrete checks. The predicted several-fold swelling of pion-like excitations above the crossover can be probed with standard lattice techniques. More decisively, simulations of pressure, energy density and entropy density at Nc equal to 5 and 7, performed exactly as they were for three colors, can settle whether the regime above the crossover scales as Nc — as expected for a fluid of overlapping color singlets — or as Nc-squared, which would signal the presence of deconfined gluons. Either outcome would sharpen our map of the phases of hot strongly interacting matter.

Subject of Research: Why conserved charge fluctuations in hot QCD appear quasifree in the confining regime above the chiral crossover temperature

Article Title: Why fluctuations of conserved charges in the confining regime above \(T_{ch}\) behave as if the quarks were free?

Article References: Glozman, L. Y. (2026). Why fluctuations of conserved charges in the confining regime above $$T_{ch}$$ behave as if the quarks were free?. The European Physical Journal C, 86(10), Article 1223. https://doi.org/10.1140/epjc/s10052-026-16479-1

Image Credits: AI Generated

DOI: 10.1140/epjc/s10052-026-16479-1

Keywords: lattice QCD, confinement, chiral crossover, conserved charge fluctuations, stringy fluid, quark-hadron duality, quark-gluon plasma, chiral symmetry restoration, mesonic correlators, quark-gluon plasma threshold, Pauli blocking, heavy-ion physics

The Bc system occupies a unique position in heavy-quark physics. Because the quark and antiquark carry different flavors, the meson cannot annihilate directly into gluons, so every excited state below the lowest strong-decay threshold decays exclusively through radiative or weak transitions. This produces a sub-threshold spectrum that is both wider and populated by narrower states than in charmonium or bottomonium. The absence of charge conjugation symmetry has a further consequence: spin-singlet and spin-triplet states with the same orbital angular momentum can mix through the antisymmetric spin–orbit interaction, an effect entirely absent in equal-flavor quarkonia and directly sensitive to the interplay between the one-gluon-exchange and confinement components of the interquark potential. The characteristic dynamical scales of the Bc also lie between those of charmonium and bottomonium, placing it in an intermediate kinematic regime that must be calibrated on its own terms rather than borrowed from either neighbor.

Experimentally, the Bc family remains sparsely mapped. The complete Particle Data Group listing contains only the ground state Bc(1S) at 6274.47 ± 0.32 MeV, first observed by the CDF collaboration and confirmed by LHCb, CMS, and ATLAS, and the first radial excitation Bc(2S) at 6871.2 ± 1.0 MeV. LHCb has observed a wide structure in the Bc+γ mass spectrum consistent with the lowest excited P-wave states, while ATLAS has reported the first observation of the vector Bc*+ state. The individual 1P fine-structure components remain unresolved, however, and the low production rate — which requires simultaneous charm-anticharm and bottom-antibottom pair creation — means that the bulk of the predicted spectrum, including all D-wave states and radial excitations beyond 2S, still awaits discovery.

The theoretical framework is a non-relativistic potential model in which the heavy quark–antiquark interaction combines a short-distance Coulombic term from one-gluon exchange with a long-distance confining term. The authors solved the radial Schrödinger equation numerically using the Runge–Kutta method and added spin-dependent corrections — spin–spin, spin–orbit, and tensor interactions — perturbatively at order 1/m_Q². The logarithmic deformation modifies the intermediate-distance behavior through an effective string tension that decreases smoothly with separation while preserving asymptotic linearity, whereas the screened potential progressively suppresses the confining force at large distances altogether. Parameters for each potential were sampled with the affine-invariant ensemble sampler in the emcee package, using uniform priors within physically motivated bounds and a likelihood built from the two PDG masses, their mass difference, and the lattice QCD hyperfine splitting of 55 ± 3 MeV.

The Bayesian treatment matters because the Bc sector offers only three effectively independent experimental constraints while the models carry four to six free parameters, making deterministic chi-squared fits formally under-determined and susceptible to multiple acceptable solutions whose excited-state predictions diverge. Roughly 5000 posterior samples per potential, checked against convergence diagnostics including integrated autocorrelation times, effective sample sizes, and Gelman–Rubin statistics, yielded median predictions with asymmetric 1σ credible intervals. The ground state emerged with sub-MeV precision in all three models — 6274.48, 6274.47, and 6274.47 MeV — and the 1S hyperfine splittings of 55.71, 55.65, and 55.27 MeV agree with the lattice determinations of 55 ± 3 and 54 ± 3 MeV, validating the spin–contact parameterization.

A tension does appear with the newest data. ATLAS recently measured the 1S hyperfine splitting as 64.5 ± 1.4 MeV statistically, with a systematic uncertainty of −1.4 to +1.0 MeV, notably higher than the posterior medians, which were based on the lattice input available when the analysis was performed. To test the sensitivity of the inference to that constraint, the authors broadened the lattice uncertainty from 3 to 10 MeV through posterior importance reweighting using 300,000 uniformly thinned samples. The posterior median of the smearing parameter ρ shifted by roughly +0.21, +0.24, and +0.18 GeV for the three potentials while remaining within the original 68 percent credible intervals, indicating that the width of the ρ posterior is not an artifact of the assumed lattice precision and that the hyperfine splitting constrains a correlated combination of parameters rather than fixing any single one.

The picture changes sharply at higher excitation. The mass gap between the Cornell and logarithmically modified potentials accumulates at roughly 14 MeV per unit of radial quantum number, reaching about 70 MeV by the 6S state, with the modified potential clustering excited levels more tightly because of its shallower effective confining slope. The screened potential diverges even further: its 6S prediction lies about 260 MeV below the Cornell result, with an uncertainty of +150 to −241 MeV. Uncertainties widen monotonically with excitation — the Cornell 6³S₁ mass carries an interval of +39.46 to −80.35 MeV — reflecting the absence of experimental constraints on higher levels. The D-wave sector shows the largest inter-potential separation, about 90 MeV at 6D, because higher-l states are more spatially extended and therefore more sensitive to long-range potential differences, although that separation remains comparable to the credible interval, limiting the discriminating power to at most the 1σ level.

The flavor-asymmetry-driven mixing produces distinctive predictions. P-wave singlet–triplet mixing angles start at 4.82 and 6.62 degrees for the Cornell and modified potentials, with uncertainties of roughly ±9 to 10 degrees at 1P, and rise to about 19 to 20 degrees by 6P with much smaller uncertainties. D-wave mixing angles are large and negative throughout, near −52.5 degrees at 1D, with uncertainties of only about 1.5 degrees. The analysis also finds a nonstandard D-wave mass ordering — 1D₂ below 1³D₃ below 1³D₁ below 1D₂′ — persisting through 4D for the first two potentials, driven by a sign reversal in the symmetric spin–orbit contribution as scalar confinement overtakes one-gluon exchange. The LHCb P-wave structure, described at current resolution by a two-peak fit at 6704.8 and 6752.4 MeV, cannot yet be assigned to individual 1P states, but the predicted 1P masses populate the observed region, providing a qualitative consistency check.

Wave-function and velocity diagnostics trace where the model itself reaches its limits. The squared radial wave function at the origin, which governs leptonic decays and production rates, falls by about 59 percent from 1S to 6S under the Cornell potential, while root-mean-square radii grow from 1.71 to 8.47 GeV⁻¹. The squared charm-quark velocity climbs steadily with excitation, exceeding the diagnostic benchmark of 0.5 from approximately the 3S, 3P, and 3D states onward — or from the 4S, 4P, and 4D states under the screened potential — while the bottom quark remains non-relativistic throughout, with its squared velocity staying below 0.09. This means the higher-excitation predictions should be read as indicative rather than precision benchmarks, since higher-order relativistic corrections grow relatively more important even as the leading spin-dependent splittings shrink in absolute size.

Regge trajectories, which correlate squared masses with radial and orbital quantum numbers, add a complementary view. The radial trajectories of S-wave states show pronounced non-linearity under all three potentials, diminishing for P- and D-wave families as the centrifugal barrier pushes the wave functions outward toward the linearly confining region. The independently fitted curvature coefficients are not fixed to a universal value but shift systematically with the potential form, and the orbital slopes fall with increasing radial excitation. Model comparison using the widely applicable information criterion and Pareto-smoothed leave-one-out cross-validation found no statistically significant preference among the three confinement forms, so the authors retained the uniform-prior results as the baseline. They also caution that the credible intervals for states above the lowest open-flavor threshold are conditional on single-channel potentials and exclude coupled-channel effects, which existing calculations suggest can shift masses by tens to hundreds of MeV in a strongly state- and model-dependent way. Resolving the predicted fine structure and mixing pattern will require higher-resolution measurements at LHCb and other facilities, with access to the higher excitations depending on increased luminosity and dedicated reconstruction strategies.

Subject of Research: Bayesian MCMC analysis of the Bc meson spectrum using three confining potential models

Article Title: \({B_c}\) meson spectroscopy from Bayesian MCMC: probing confinement and state mixing

Article References: Mony, A. C., & Dhir, R. (2026). $${B_c}$$ meson spectroscopy from Bayesian MCMC: probing confinement and state mixing. The European Physical Journal C, 86(10), Article 1221. https://doi.org/10.1140/epjc/s10052-026-16315-6

Image Credits: AI Generated

DOI: 10.1140/epjc/s10052-026-16315-6

Keywords: Bc meson, spectroscopy, Bayesian inference, Markov chain Monte Carlo, Cornell potential, confinement, quarkonium, Regge trajectories, spin-orbit mixing, lattice QCD, LHCb, potential models

The CKM matrix sits at the heart of the Standard Model’s description of quark flavour mixing. Its element |Vcb| quantifies the charged-current transition strength between the charm and bottom quarks, and it plays a double role in particle physics: it is used to verify the unitarity of the CKM matrix, and it serves as a benchmark against which possible new physics beyond the Standard Model can be tested. A precise determination of |Vcb| has therefore long been one of the central goals of B-physics phenomenology. What makes the problem genuinely difficult is a long-standing discrepancy between exclusive extractions, which rely on specific decays such as B → D(*)ℓν, and inclusive extractions, which sum over many final states. That tension has motivated extensive theoretical and experimental work for years, and it remains unresolved.

The experimental side of the analysis draws on a decade of increasingly sophisticated measurements. The Belle collaboration measured the semileptonic decay B → Dℓν using fully reconstructed B meson tag events, obtained the isospin-averaged branching fraction, and extracted |Vcb| with two different form-factor parameterization frameworks, presenting binned differential decay rates in bins of the recoil variable w. Belle II then performed an analogous measurement adopting the BCL parameterization. For the vector-vector channel B⁰ → D*⁻ℓ⁺ν, independent fits using the CLN and BGL parameterizations have been carried out alongside full binned differential distributions, response matrices, and statistical and systematic correlation matrices, all designed to alleviate the inclusive–exclusive |Vcb| puzzle. Belle II has also provided separate partial decay rates for electron and muon final states as functions of the recoil parameter ω and three decay angular observables.

These experimental fits require form factors evaluated near zero recoil as external theoretical inputs, and lattice QCD has undergone steady methodological and precision improvements over the past decade. The first unquenched lattice gauge simulation of the B → Dℓν form factors came from the Fermilab-MILC collaborations. HPQCD then produced full-q² unquenched calculations of the scalar and vector form factors f₊(q²) and f₀(q²). For the exclusive mode B → Dℓν, unquenched lattice-QCD form factors at nonzero recoil were computed by Fermilab-MILC with 2+1 flavours, and HPQCD later delivered complete vector, axial-vector, and tensor form factors covering the entire q² range. JLQCD introduced Möbius domain-wall fermion discretization for the B → Dℓν transition, offering an independent cross-check against the widely used clover heavy-quark schemes. Collectively, these efforts trace a trajectory from narrow zero-recoil single-channel studies toward full-q², multi-flavour, multi-operator form factor sets with multiple independent discretization frameworks, drastically reducing lattice systematic uncertainties in the |Vcb| extraction.

The theoretical backbone of the new analysis is the BGL dispersive parameterization, a model-independent z-expansion scheme in which the form factors are written as series in a conformal variable z(q²) that maps the physical kinematic region into the unit disk. Blaschke factors remove poles associated with Bc resonances below the pair-production threshold, and outer functions φᵢ, built from perturbative QCD dispersion relations, control the series convergence. The authors impose strong unitarity bounds derived from four heavy-to-heavy transition channels — B → D, B → D*, B* → D and B* → D* — covering twenty form factors in total, with heavy-quark effective theory power corrections retained through O(αs), O(εb), O(εc) and O(εc²) terms. These bounds act as hard constraints in the fit, effectively restricting the space of BGL expansion coefficients and suppressing uncontrolled systematic uncertainties from z-series extrapolation. Sensitivity tests varying the HQET inputs over their quoted ranges shift the extracted |Vcb| by at most about 0.16σ of the nominal uncertainty.

A key innovation is the treatment of the light-cone sum-rule (LCSR) inputs at large hadronic recoil. The team used correlated SCET sum-rule predictions including next-to-leading-order QCD corrections and four sources of next-to-leading-power contributions, evaluated at six representative kinematic points q² ∈ {−3, −2, −1, 0, 1, 2} GeV². Rather than treating these predictions as independent numbers, the authors generated 1000 complete input-parameter sets by sampling the underlying LCSR parameters simultaneously according to their specified distributions, so that each replica is a full correlated prediction vector. This preserves parameter variations that alter the q² dependence, including those associated with the parameter λB. The empirical mean, standard deviations and correlation matrix of the replica ensemble then define the LCSR input to a correlated Gaussian χ². Because the predictions at different q² values are extremely strongly correlated, the raw correlation matrix is nearly singular; the authors regularize it mildly while retaining all correlation directions, and they verified that the regularization choice does not materially control the extracted |Vcb| or its uncertainty.

The numerical analysis proceeds through three nested fit scenarios with progressively enlarged constraint sets. The first relies purely on lattice-QCD form-factor simulations; the second adds the correlated SCET sum-rule predictions; the third supplements both with binned differential decay spectra and angular observables from Belle and Belle II, which is the only scenario in which |Vcb| can be determined. Comparing across the scenarios shows that incorporating the correlated large-recoil LCSR predictions dramatically shrinks the uncertainties of high-order BGL coefficients, while the inclusion of precise Belle II decay observables yields further moderate tightening of the parameter space. The complete global fit yields χ²min = 218.505 for 229 degrees of freedom, corresponding to a healthy p-value of 0.680, with the lattice-QCD, LCSR and experimental contributions to the χ² being 42.91, 14.44 and 161.16 respectively.

From the full global fit the team extracts |Vcb| = (39.18 ± 0.47) × 10⁻³. Individual channel fits give |Vcb| = (40.3 ± 0.8) × 10⁻³ from B → Dℓν and |Vcb| = (38.6 ± 0.6) × 10⁻³ from B → D*ℓν, and the authors document in detail how methodological choices — the exclusion of the Belle 2017 dataset, the retention of only the muonic mode of the Belle 2018 measurement to mitigate well-documented electron–muon tension, and the use of constrained frequentist χ² minimization rather than Bayesian MCMC or nested sampling — shift the central value relative to previous analyses. Robustness checks are thorough: a constrained profile-likelihood scan gives |Vcb| = (39.1801 +0.4611/−0.4577) × 10⁻³, consistent with the Hessian uncertainty, and increasing the BGL truncation from O(z²) to O(z³) shifts the central value by only 0.032 × 10⁻³ while the four strong unitarity bounds become essentially saturated, showing that the extra cubic coefficients are only weakly constrained by the present data.

Using the fitted form-factor coefficients from the combined lattice-QCD and LCSR fit, the team derives updated Standard Model predictions for the lepton-flavour-universality ratios R(D) = 𝓑𝓡(B → Dτντ)/𝓑𝓡(B → Dμνμ) = 0.3069 ± 0.0080 and R(D*) = 0.2548 ± 0.0043, alongside full differential decay distributions in the momentum transfer and angular variables for both the muon and tau channels. The muonic spectrum rises sharply near q² → 0 and decreases monotonically, while the tau channel has a nonvanishing kinematic threshold at finite q² owing to the tauon’s larger mass. Belle II measurements broadly follow the theoretical trend for the muonic channel, though mild local deviations appear in certain angular regions. At present precision, the authors emphasize, these discrepancies do not provide compelling evidence for physics beyond the Standard Model.

The persistent tension, however, remains vivid. Experimental world averages for both R(D) and R(D) lie systematically above the aggregate Standard Model theoretical band, and the team’s two-dimensional correlated 68% confidence contour in the (R(D), R(D)) plane shows little overlap with the experimental average. Their R(D*) central value aligns well with the global theory average, while their R(D) prediction sits slightly above pure lattice averages such as the FLAG 2024 result. The discrepancy may originate from underestimated lattice or LCSR systematic uncertainties, residual shape biases in the BGL z-expansion, missing higher-order or higher-power corrections, or unvalidated experimental reconstruction efficiencies. Improved lattice-QCD calculations, refined sum-rule predictions, and high-statistics Belle II measurements will be essential for determining whether the remaining gaps are theoretical artefacts or genuine signals of new physics in charged-current b → cℓν transitions — and this complete set of correlated Standard Model predictions now provides a robust, up-to-date reference for that decisive comparison.

Subject of Research: Exclusive determination of the CKM matrix element |Vcb| from semileptonic B meson decays

Article Title: Exclusive determination of (|V_{cb}|) from semileptonic decays (B\rightarrow D^{(*)}\ell \nu _{\ell })

Article References: Chen, X.-W., Cui, B.-Y., Zhang, J.-W., Zhao, X.-C., & Chen, Y.-H. (2026). Exclusive determination of $$|V_{cb}|$$ from semileptonic decays $$B\rightarrow D^{(*)}\ell \nu _{\ell }$$. The European Physical Journal C, 86(10), Article 1174. https://doi.org/10.1140/epjc/s10052-026-16486-2

Image Credits: AI Generated

DOI: 10.1140/epjc/s10052-026-16486-2

Keywords: CKM matrix, Vcb, semileptonic decays, B mesons, lattice QCD, BGL parameterization, Belle II, R(D), R(D*), lepton flavour universality, Standard Model, flavour physics

A team of researchers from the Jülich Supercomputing Centre, KTH Royal Institute of Technology, the University of Bonn and the University of Groningen, led by Carlos Falquez, Shiting Long, Nam Ho, Estela Suarez and Dirk Pleiter, now propose a surprising answer: use a processor simulator not just to design chips, but as a full performance engineering workbench. In a study published in Frontiers in High Performance Computing, they enhanced the widely used open-source simulator gem5 with a performance analysis framework that extends the available performance counters, tracks the temporal behaviour of running applications, and links those statistics automatically to specific code regions. The result is a set of performance profiles that can identify code regions with optimisation potential, monitoring quantities that are usually not directly accessible on real hardware.

The motivation is rooted in the sheer complexity of modern processors. Even a single processor core is a system of interacting subsystems, from the micro-architecture of the front-end pipeline, where various factors influence how efficiently functional units are utilised, to multi-level cache hierarchies and memory prefetchers whose combined behaviour of data transport is notoriously difficult to extrapolate from the behaviour of individual components. On real hardware, Performance Management Units (PMUs) can be programmed at run time to count events such as cache misses, but the set of trackable events is limited, cannot be changed once a processor design is final, and is highly vendor-specific. Even when advanced monitoring features such as Intel’s Processor Event Based Sampling, AMD’s Instruction-Based Sampling or Arm’s Statistical Profiling Extension are available, the context for a particular processor behaviour often cannot be recorded, making root cause analysis a matter of expert guesswork.

Simulators offer an alternative, but they come with a fundamental trade-off. Because cycle-accurate simulation runs many orders of magnitude slower than real hardware, in practice it is only affordable when restricting the scope to a single or a small number of processor cores. gem5, which supports several instruction set architectures including x86, Arm and RISC-V, has become a popular choice and is continuously updated to follow new architectural developments. It has already served as a co-design tool in the development of the A64FX processor and in the European Processor Initiative’s design space explorations. Yet gem5 has been notoriously difficult for computational scientists to use: it was built by and for processor architects, requires significant training, and lacks tools to correlate simulation statistics with application code. Even when an application is instrumented to restrict profiling to a region of interest, gem5’s counters are accumulated over the entire region and typically reported as a single aggregate value, hiding the multiple execution phases with very different performance characteristics that may exist within it.

The new framework tackles this with two key additions: time-binned event counters and automatic labelling of executed code regions. The execution timeline of a region of interest is partitioned into fixed-length intervals, each of duration 5000 cycles, and events such as the commit of an instruction are aggregated per bin, producing a sequence of event counts that provides much finer temporal resolution. To reduce stochastic noise in the plots, a median filter from the scipy Python package is applied, suppressing outliers while preserving sharp changes. For code region detection, the instructions of the executable are grouped according to function entry point labels parsed from the objdump tool of GNU Binutils. For each time bin, the five most frequently occurring events are determined, and a function label is associated with the bin when at least five of those instructions belong to that function. The resulting label sequence is converted into labelled time ranges that can be mapped directly onto the time-binned event counters, revealing which routines were executing during each phase.

The framework also implements algorithms to detect specific classes of performance bottlenecks. One concerns sensitivity to instruction data dependencies: in out-of-order processors, fetched instructions wait in an issue queue until their operands are available, and when an application is ported to an architecture with deeper execution pipelines, the probability of instructions waiting for data dependencies increases while pressure on the execution units decreases, causing unexpectedly low performance. The team added event counters that can differentiate between instruction classes and track both the average number of instructions waiting in the issue queue for dependencies to resolve and the average number of instructions stalled due to busy execution units. A second class concerns memory traffic and the average filling of the queue for pending requests at the memory controller, observables that are difficult or impossible to extract reliably from real hardware. A third concerns long-latency instructions, such as conditioned load instructions whose latency cannot be hidden by out-of-order execution; the simulator detects these by measuring the instruction squashing rate, that is, the rate at which speculatively executed instructions must be discarded because of misspeculation.

Three case studies demonstrate the approach across a broad range of computational science. The first is GROMACS, a very widely used molecular dynamics package, whose most expensive kernel computes non-bonded interactions. Using a recent version of GROMACS with a solvation free energy benchmark, the simulations modelled a single Arm Neoverse V2 core supporting SVE instructions, the same core used by NVIDIA in its Grace processor. The profiles showed that the main numerical kernel achieves a throughput of around 1.7 SIMD floating-point instructions per cycle, almost fully utilising the two available SVE pipelines. Crucially, the new counters revealed that for effective execution latencies of four cycles or more, the vectorised kernel becomes sensitive to data dependencies: the average number of stalled instructions drops below one, signalling a decrease in the rate of dependency resolution, while the number of instructions waiting in the issue queue for missing dependencies remains large, verifying that the front-end keeps sending instructions at sufficient rate.

The second case study concerns a kernel from the Lattice QCD solver DD-αAMG, representative of memory-bandwidth-limited stencil applications. The team investigated a recently extended version of the solver that supports solving linear equations with multiple right-hand sides, an optimisation that increases arithmetic intensity but had not delivered the expected performance improvements on real hardware. Using the A64FX processor on the Ookami system, they had identified that effective memory bandwidth varied across blocking sizes, but comprehensive information from cache and memory controllers is almost impossible to extract from real hardware. The gem5 simulations of a read-intensive inner-product kernel showed that larger blocking sizes enhance performance, mirroring real-hardware trends, but also revealed an unexpected performance drop at a blocking size of eight, where the compiler failed to vectorise the code and generated scalar load instructions instead, halving the rate of read requests.

The third case study targeted the breadth-first search algorithm implemented in the Ligra graph framework, whose irregular control flow and memory access patterns stress the branch predictor. The profiles showed poor execution unit utilisation, with an average commit rate below one instruction per cycle most of the time, and a high rate of squashed instructions. The team traced most of the branch mispredictions to the EdgeMapDense routine, where three specific lines each caused about a third of the mispredictions. By moving a conditioned load instruction outside the branch so it could be initiated earlier, a trade-off accepting some unnecessary data transfers, they benchmarked both implementations on an AWS Graviton 4 bare-metal node with Kronecker graphs of scale 21 to 24 and measured a speed-up of around five percent for all problem sizes, confirming that a simulator-detected bottleneck could be fixed on real hardware.

The authors are careful about the limitations. Single-core simulations run about five to six orders of magnitude slower than execution on real hardware, so simulating many-core architectures or full applications is likely to be prohibitively expensive, although distributed frameworks such as COSSIM may help. Optimising code for specific processors also depends on the availability of validated gem5 configurations, and no library of ready-to-use configurations exists yet; previous validation studies have reported average deviations of about ten percent for simpler Arm cores, with discrepancies of up to forty percent in some cases for complex workloads, and there is a risk of mismatches between a specific processor solution and its gem5 model that only show up under particular circumstances. Still, the benefit of the approach is that gem5 already provides a large and diverse set of statistics that can illuminate the often difficult-to-uncover interplay between observed application performance and the complex hardware on which it executes, with much freedom to add further event monitors and analysis mechanisms. The team plans to extend the list of performance issues for which automatic detection can be integrated into gem5, and to conduct further case studies exploring the benefits and disadvantages of the approach. The modified simulator has been made openly available, potentially also benefiting compiler engineers and system software developers far beyond HPC.

Subject of Research: High Performance Computing

Article Title: Processor simulation as a tool for performance engineering

Article References: Falquez, C., Long, S., Ho, N., Suarez, E., & Pleiter, D. (2025). Processor simulation as a tool for performance engineering. Frontiers in High Performance Computing, 3, Article 1669101. https://doi.org/10.3389/fhpcp.2025.1669101

Image Credits: AI Generated

DOI: 10.3389/fhpcp.2025.1669101

Keywords: processor simulation, gem5, performance engineering, high-performance computing, Arm SVE, RISC-V, GROMACS, Lattice QCD, breadth-first search, performance counters, Processor, simulation

The research, carried out by Wei Shen, Zhen-Yan Lu, Muhammad Waqas, Xun Chen, Zhi-Jun Ma and Guang-Xiong Peng, focuses on pure SU(3) gauge theory, the gluon-only sector of quantum chromodynamics with all quark flavors switched off. This simplified world is not a mere toy. It captures the essential nonperturbative dynamics of confinement and deconfinement, and it undergoes a first-order phase transition at a critical temperature T_c. Because quarks are absent, the theory is cleaner to simulate on the lattice and serves as a rigorous testing ground for the phenomenological frameworks that ultimately must describe the full quark-gluon plasma created in experiments at facilities such as the Relativistic Heavy Ion Collider and the Large Hadron Collider.

The team’s approach, which they call the temperature-dependent mass or TDM model, rests on a deceptively simple idea. All the complicated medium effects that make hot gluon matter deviate from an ideal gas are encoded in a single effective parameter: a temperature-dependent gluon mass m_g(T). Crucially, the authors stress that this mass is not a physical pole mass of a propagating particle and certainly not evidence of free massive gluons floating around below T_c. Below the transition, where color-singlet glueball-like excitations dominate, the effective mass is simply a convenient bookkeeping device that parametrizes the suppression of colored gluonic degrees of freedom inferred from lattice data. Above the transition, it plays a similar role for the deconfined but still strongly interacting medium.

What sets this work apart is the austerity of its input. The only lattice data used to constrain the entire framework is the normalized pressure P/T^4 of pure SU(3) gauge theory. At each temperature point, the researchers numerically solved for the value of m_g/T that makes the model pressure match the lattice pressure, yielding 48 pressure-extracted mass values. Below T_c these were fitted with a cubic polynomial in the reduced temperature, while above T_c the fit is expressed as a cubic polynomial in the running strong coupling, whose temperature dependence follows a renormalization-group-motivated form. No energy density, entropy density or trace anomaly data entered the fit. Everything else, including the entropy, energy density and trace anomaly, emerged from thermodynamically consistent relations that carefully treat both the explicit and implicit temperature dependence carried by the effective mass.

That thermodynamic consistency is not a technical nicety. Because the dispersion relation of each gluonic mode depends on temperature through m_g(T), naive differentiation misses an entire term, and response functions, which are derivatives of the thermodynamic potential, are exquisitely sensitive to exactly this kind of subtlety. The authors derive the entropy density and energy density with the implicit dependence included, and in the limit of a temperature-independent mass their expressions reduce to the standard ideal boson gas results. The payoff is that the model reproduces not only the lattice pressure but also the trace anomaly, defined as the energy density minus three times the pressure, which measures the violation of conformal, or scale-invariant, behavior. The pronounced peak of the trace anomaly near T_c, a hallmark of strong nonconformal dynamics, comes out naturally from the same pressure-constrained mass profile.

The heart of the new paper lies in its treatment of second-order response functions, quantities that involve second derivatives of the thermodynamic potential and are therefore far more delicate probes than bulk observables. The first of these is the specific heat at constant volume, which measures how the energy density responds to changes in temperature. In the normalized form C_V/T^3, the model predicts a striking nonmonotonic pattern: the quantity is strongly suppressed at low temperatures, rises rapidly as the deconfinement region is approached, reaches a sharp maximum near T_c, dips to a local minimum, and then climbs again toward a constant. This behavior directly reflects the rapid temperature variation of the energy density across the transition, amplified by the derivative structure of the specific heat formula.

Independent lattice estimates of the specific heat, which were not used as input, show a qualitatively similar enhancement near T_c, and the model reproduces their overall trend and magnitude above the transition, though noticeable differences appear right at the peak. The authors attribute this to the intrinsic sensitivity of derivative quantities to interpolation procedures and finite-volume effects near a first-order transition. They also verified that the sharp peak is not a numerical artifact: refining the temperature step near T_c only sharpens the structure further, leaving the behavior away from the critical region unchanged. At high temperatures, C_V/T^3 approaches the massless Stefan-Boltzmann reference value of 32 pi squared over 15, approximately 21.06, although the authors caution that because their fitted mass ratio tends to a small nonzero constant, this conformal value should be read as a reference point rather than an exact asymptote of the model.

The second response function, and arguably the more novel one, is the isentropic bulk modulus K_S, which quantifies the mechanical stiffness of the medium under compression at constant entropy. It is defined as minus the volume times the pressure derivative with respect to volume at fixed entropy, and it equals the enthalpy density times the square of the speed of sound. This makes the bulk modulus a bridge between thermodynamics and mechanics: it governs how efficiently the plasma converts energy density into pressure, and thereby how sound waves and density fluctuations propagate through it. The calculations show that K_S/T^4 is strongly suppressed in the confined phase, where both pressure and enthalpy are small and the medium offers little resistance to squeezing, but rises sharply across the deconfinement region, signaling a substantial stiffening of the equation of state. At high temperatures it settles toward the conformal reference value of 32 pi squared over 135, roughly 2.34, while the inverse quantity, the isentropic compressibility, correspondingly falls toward about 0.427 in the same normalized units.

Supporting this picture, the study also tracks the gluon number density and the energy per thermally active gluonic mode. Although gluon number is not conserved, the scaled density n_g/T^3 serves as a measure of how many gluonic degrees of freedom are thermally accessible. It is nearly zero below and near T_c, then rises rapidly as deconfinement liberates the effective degrees of freedom, before flattening onto a high-temperature plateau consistent with the ultrarelativistic scaling n_g proportional to T cubed. The energy per mode, E_g divided by n_g times T_c, shows a sharp peak just above the transition: near T_c, a small number of thermally excited modes each carry a comparatively large share of energy, a signature of the large effective mass suppressing the thermal population. As the temperature climbs, the population becomes denser and more evenly energetic, and the ratio eventually grows linearly with temperature, exactly as dimensional scaling demands.

Taken together, these results paint a coherent portrait of gluon matter evolving from a confined, nonconformal regime into an approximately scale-invariant plasma. The authors emphasize that the specific heat and the isentropic bulk modulus are complementary: the former probes thermal response through energy fluctuations, the latter mechanical response through sound propagation and stiffness. Because both involve additional temperature derivatives, they constrain the effective mass profile far more stringently than bulk quantities alone could, and their nontrivial structures cannot be inferred from the magnitudes of pressure or energy density. The TDM framework, minimal as it is, thus offers a predictive baseline for derivative-sensitive observables in pure gauge matter. The natural next steps, the authors suggest, are extensions toward transport properties and toward full QCD with dynamical quarks, bringing the framework closer to the conditions of the real quark-gluon plasma that heavy-ion experiments continue to probe with ever-greater precision.

Subject of Research: Thermodynamic response functions of pure SU(3) gluon matter near the deconfinement transition, derived from a lattice-data-constrained effective gluon mass model

Article Title: Lattice-data-driven specific heat and isentropic bulk modulus of SU(3) gluon matter at finite temperature

Article References: Shen, W., Lu, Z.-Y., Waqas, M., Chen, X., Ma, Z.-J., & Peng, G.-X. (2026). Lattice-data-driven specific heat and isentropic bulk modulus of SU(3) gluon matter at finite temperature. The European Physical Journal C, 86(9), Article 1079. https://doi.org/10.1140/epjc/s10052-026-16286-8

Image Credits: AI Generated

DOI: 10.1140/epjc/s10052-026-16286-8

Keywords: lattice QCD, SU(3) gauge theory, deconfinement, quark-gluon plasma, specific heat, bulk modulus, equation of state, trace anomaly, effective gluon mass, thermodynamics, speed of sound, conformal symmetry

GPDs are the mathematical objects that go beyond the familiar parton distribution functions measured in inclusive deep inelastic scattering. While ordinary PDFs tell you the probability of finding a quark carrying a given fraction of the proton’s momentum, GPDs also encode what happens when the proton recoils, absorbing a momentum transfer that shifts it sideways. That extra information, accessed through processes like deeply virtual Compton scattering, where an electron knocks a virtual photon out of the proton and a real photon emerges, is what allows physicists to reconstruct genuinely three-dimensional portraits of nucleon structure, including the elusive orbital angular momentum carried by quarks and gluons.

The central difficulty is that experiments do not measure GPDs directly. What they measure are Compton form factors, convolutions of the GPDs with perturbatively calculable kernels that are sharply singular at particular points. Recovering the underlying GPDs from these convolutions is an inverse problem that is, in principle, intractable without additional input. The Virginia team’s strategy is to narrow the space of possible solutions by building the parametrization directly on the symmetries of quantum chromodynamics: crossing symmetry under the exchange of quark and antiquark momenta, polynomiality of the Mellin moments, and the area rules dictated by baryon number and momentum conservation.

The mathematical backbone of UVA2 is the reggeized spectator model, in which the proton is pictured as a struck parton bound to a remnant whose invariant mass is allowed to fluctuate. This yields compact analytic expressions for the two leading vector GPDs, H and E, for each quark flavor, including the valence u and d quarks, the antiquark sea, and the gluons. In a first for this framework, the authors provide fully analytic closed forms in all kinematic variables, eliminating the need for numerically expensive transverse momentum integrations and making the parametrization immediately usable in regression analyses, neural network training, and Monte Carlo event generators.

Constraining such a multidimensional function requires an enormous amount of input. In the forward limit, where the GPD H reduces to ordinary PDFs, the team fitted their parameters to reproduce the NNPDF21 parton distributions, checking the quality of the match with Kolmogorov-Smirnov tests. The t-dependence, describing how the distributions fall off with transverse momentum transfer, was then pinned down by fitting the flavor-separated Dirac and Pauli form factors extracted from lepton-nucleon scattering, together with the second Mellin moments recently computed in lattice QCD. The gluon sector was anchored to lattice determinations of the gluon form factors evaluated at a pion mass closer to the physical value than in previous versions.

All of these ingredients are defined at a common initial evolution scale of 0.58 square gigaelectronvolts, an improvement over earlier versions in which quarks and gluons started from inconsistent scales. From there, the distributions are evolved to higher energy scales using the leading-order off-forward DGLAP and ERBL evolution equations, which the authors solved with the Adams predictor-corrector method rather than the conventional Runge-Kutta approach. Because the right-hand side of the evolution equations involves convolution integrals that are notoriously expensive to evaluate, this choice of integrator delivers a substantial gain in computational speed, particularly in the low-momentum-fraction region that will dominate at future colliders.

The practical payoff comes in the form of predictions for the Compton form factors across a vast kinematic range, from the fixed-target regime of Jefferson Lab, where valence quarks dominate, all the way to the low Bjorken-x frontier of the upcoming Electron-Ion Collider, where sea quarks and gluons take over. The team’s calculations show that at EIC kinematics the Compton form factors are expected to be several orders of magnitude larger than at fixed-target facilities, and that the valence contribution is strongly suppressed in the collider setting. Crucially, because the flavor-singlet quark combination couples to gluons in the evolution equations, the gluon GPDs can in principle be extracted directly from DVCS data at the EIC, turning a machine designed to map gluon structure into a precision instrument for proton tomography.

The authors are careful about uncertainties. The error budget on their GPDs originates entirely from the fits to the t-dependence of the electromagnetic form factors, the lattice gluon form factors, and the second moments of the quark distributions, while the forward-limit parameters, obtained by matching to PDF parametrizations rather than raw data, carry no statistical errors and are instead validated through distribution-to-distribution comparisons. They flag this Hessian-based treatment as a first step, with maximum-likelihood and Markov-chain Monte Carlo analyses, along with neural network approaches, planned to properly capture correlations among the seven to eight parameters that define each flavor sector.

Everything is openly available. The UVA2 parametrization, including analytic forms at the initial scale and precomputed grids spanning momentum fractions from ten thousandths up to nearly one, momentum transfers up to four square gigaelectronvolts, and evolution scales up to a thousand square gigaelectronvolts, can be downloaded from a public GitHub repository in easily readable CSV formats. This openness matters, because GPD parametrizations are becoming shared infrastructure: they feed event generators, guide experimental analyses of exclusive processes, and provide the benchmark curves against which lattice QCD results and future DVCS measurements will be judged.

As the Electron-Ion Collider moves from blueprint to beamline, tools like UVA2 will determine how much physics can be squeezed out of the first data. By fusing elastic scattering, inclusive deep inelastic data, and lattice QCD into a single symmetry-respecting framework, the Virginia collaboration has delivered what amounts to a flexible, evolving atlas of the proton’s interior, one that promises to turn the coming decade of scattering experiments into a genuine three-dimensional census of the most abundant visible matter in the universe.

Subject of Research: Global parametrization of generalized parton distributions describing the three-dimensional quark and gluon structure of the proton

Article Title: Updated flexible global parametrization of generalized parton distributions from elastic and deep inelastic inclusive scattering data

Article References: Panjsheeri, Z., Adams, D. Q., Khawaja, A., Pandey, S., Tezgin, K., & Liuti, S. (2026). Updated flexible global parametrization of generalized parton distributions from elastic and deep inelastic inclusive scattering data. The European Physical Journal C, 86(9), Article 1078. https://doi.org/10.1140/epjc/s10052-026-15872-0

Image Credits: AI Generated

DOI: 10.1140/epjc/s10052-026-15872-0

Keywords: generalized parton distributions, proton structure, quantum chromodynamics, deeply virtual Compton scattering, lattice QCD, parton distribution functions, Electron-Ion Collider, QCD evolution, elastic form factors, Compton form factors, nucleon tomography, reggeized spectator model