-
Connection between f-electron correlations and magnetic excitations in UTe2
Authors:
Thomas Halloran,
Peter Czajka,
Gicela Saucedo Salas,
Corey Frank,
Chang-Jong Kang,
J. A. Rodriguez-Rivera,
Jakob Lass,
Daniel G. Mazzone,
Marc Janoschek,
Gabi Kotliar,
Nicholas P. Butch
Abstract:
The detailed anisotropy of the low-temperature, low-energy magnetic excitations of the candidate spin-triplet superconductor UTe$_2$ is revealed using inelastic neutron scattering. The magnetic excitations emerge from the Brillouin zone boundary at the high symmetry $Y$ and $T$ points and disperse along the crystallographic $\hat{b}$-axis. In applied magnetic fields to at least $μ_0 H=11$~T along…
▽ More
The detailed anisotropy of the low-temperature, low-energy magnetic excitations of the candidate spin-triplet superconductor UTe$_2$ is revealed using inelastic neutron scattering. The magnetic excitations emerge from the Brillouin zone boundary at the high symmetry $Y$ and $T$ points and disperse along the crystallographic $\hat{b}$-axis. In applied magnetic fields to at least $μ_0 H=11$~T along the $\hat{c}-$axis, the magnetism is found to be field-independent in the $(hk0)$ plane. The scattering intensity is consistent with that expected from U$^{3+}$/U$^{4+}$ $f$-electron spins with preferential orientation along the crystallographic $\hat{a}$-axis, and a fluctuating magnetic moment of 2.3(7) $μ_B$. These characteristics indicate that the excitations are due to intraband spin excitons arising from $f$-electron hybridization.
△ Less
Submitted 6 September, 2024; v1 submitted 26 August, 2024;
originally announced August 2024.
-
Mamba State-Space Models Are Lyapunov-Stable Learners
Authors:
John T. Halloran,
Manbir Gulati,
Paul F. Roysdon
Abstract:
Mamba state-space models (SSMs) were recently shown to outperform state-of-the-art (SOTA) Transformer large language models (LLMs) across various tasks. Despite subsequent widespread adaptation, little work has focused on Mamba LLMs' amenability for fine-tuning frameworks ubiquitously used for Transformer-based LLMs, e.g., mixed-precision fine-tuning (MPFT) and parameter-efficient fine-tuning (PEF…
▽ More
Mamba state-space models (SSMs) were recently shown to outperform state-of-the-art (SOTA) Transformer large language models (LLMs) across various tasks. Despite subsequent widespread adaptation, little work has focused on Mamba LLMs' amenability for fine-tuning frameworks ubiquitously used for Transformer-based LLMs, e.g., mixed-precision fine-tuning (MPFT) and parameter-efficient fine-tuning (PEFT). For the former, it currently remains an open question whether Mamba's recurrent dynamics are robust to small input changes, such as those encountered during MPFT. Using dynamical systems theory (in particular, Lyapunov exponents), we answer this question in the affirmative. We empirically validate this result through several experiments, showing that Mamba SSMs are significantly more stable to changes introduced by mixed-precision than comparable Transformers, even when both MPFT and PEFT are combined. For PEFT, we show how targeting specific memory buffers in Mamba's customized CUDA kernels for low-rank adaptation regularizes SSM parameters, thus providing both parameter efficient learning and computational savings. Finally, with both MPFT and PEFT enabled, we explore the impact of instruction tuning Mamba SSMs for in-context learning (ICL) on natural language tasks. While pretrained Mamba and Mamba-2 models only achieve 38% and 82% (respectively) of the ICL improvements of comparable Transformer-based LLMs, we show that instruction tuning allows Mamba models to narrow this gap to 81% and Mamba-2 models to skyrocket over this gap to 132%.
△ Less
Submitted 15 October, 2024; v1 submitted 31 May, 2024;
originally announced June 2024.
-
Continuum of magnetic excitations in the Kitaev honeycomb iridate D$_3$LiIr$_2$O$_6$
Authors:
Thomas Halloran,
Yishu Wang,
K. W. Plumb,
M. B. Stone,
Barry Winn,
M. K. Graves-Brook,
J. A. Rodriguez-Rivera,
Yiming Qui,
Prashant Chauhan,
Johannes Knolle,
Roderich Moessner,
N. P. Armitage,
Tomohiro Takayama,
Hidenori Takagi,
Collin Broholm
Abstract:
Inelastic neutron scattering (INS) measurements of powder D$_3(^{7}$Li)($^{193}$Ir)$_2$O$_6$ reveal low energy magnetic excitations with a scattering cross section that is broad in $|Q|$ and consistent with a Kitaev spin-liquid (KSL) state. The magnetic nature of the excitation spectrum is demonstrated by longitudinally polarized neutron studies. The total magnetic moment of 1.7(2)$μ_B$/Ir inferre…
▽ More
Inelastic neutron scattering (INS) measurements of powder D$_3(^{7}$Li)($^{193}$Ir)$_2$O$_6$ reveal low energy magnetic excitations with a scattering cross section that is broad in $|Q|$ and consistent with a Kitaev spin-liquid (KSL) state. The magnetic nature of the excitation spectrum is demonstrated by longitudinally polarized neutron studies. The total magnetic moment of 1.7(2)$μ_B$/Ir inferred from the total magnetic scattering cross section is consistent with the effective moment inferred from magnetic susceptibility data and expectations for the $J_{\rm eff}=1/2$ single ion state. The rise in the dynamic correlation function ${\cal S}(Q,ω)$ for $\hbarω<5~$meV can be described by a nearest-neighbor Kitaev model with interaction strength $K\approx-13(5)$~meV. Exchange disorder associated with the mixed D-Li site could play an important role in stabilizing the low $T$ quantum fluctuating state.
△ Less
Submitted 12 February, 2024;
originally announced February 2024.
-
Non-collinear 2k antiferromagnetism in the Zintl semiconductor Eu$_5$In$_2$Sb$_6$
Authors:
Vincent C. Morano,
Jonathan Gaudet,
Nicodemos Varnava,
Tanya Berry,
Thomas Halloran,
Chris J. Lygouras,
Xiaoping Wang,
Christina M. Hoffman,
Guangyong Xu,
Jeffrey W. Lynn,
Tyrel M. McQueen,
David Vanderbilt,
Collin L. Broholm
Abstract:
Eu$_5$In$_2$Sb$_6$ is an orthorhombic non-symmorphic small band gap semiconductor with three distinct Eu$^{2+}$ sites and two low-temperature magnetic phase transitions. The material displays one of the greatest (negative) magnetoresistances of known stoichiometric antiferromagnets and belongs to a family of Zintl materials that may host an axion insulator. Using single crystal neutron diffraction…
▽ More
Eu$_5$In$_2$Sb$_6$ is an orthorhombic non-symmorphic small band gap semiconductor with three distinct Eu$^{2+}$ sites and two low-temperature magnetic phase transitions. The material displays one of the greatest (negative) magnetoresistances of known stoichiometric antiferromagnets and belongs to a family of Zintl materials that may host an axion insulator. Using single crystal neutron diffraction, we show that the $T_{\mathrm{N1}}=14\mathrm{~K}$ second-order phase transition is associated with long-range antiferromagnetic order within the chemical unit cell $\left( k_1 = (000) \right)$. Upon cooling below $T_{\mathrm{N1}}$, the relative sublattice magnetizations of this structure vary until a second-order phase transition at $T_{\mathrm{N2}}=7\mathrm{~K}$ that doubles the unit cell along the $\hat{c}$ axis $\left( k_2 = \left(00\frac{1}{2}\right) \right)$. We show the anisotropic susceptibility and our magnetic neutron diffraction data are consistent with magnetic structures described by the $Γ_3$ irreducible representation with the staggered magnetization of the $k_1$ and $k_2$ components polarized along the $\hat{b}$ and $\hat{a}$ axis, respectively. As the $k_2$ component develops, the amplitude of the $k_1$ component is reduced, which indicates a 2k non-collinear magnetic structure. Density functional theory is used to calculate the energies of these magnetic structures and to show the $k_1$ phase is a metal so $T_{\mathrm{N1}}$ is a rare example of a unit-cell-preserving second-order phase transition from a paramagnetic semiconductor to an antiferromagnetic metal. DFT indicates the transition at $T_{\mathrm{N2}}$ to a doubled unit cell reduces the carrier density of the metal, which is consistent with resistivity data.
△ Less
Submitted 1 November, 2023;
originally announced November 2023.
-
Fragile superconductivity in a Dirac metal
Authors:
Chris J. Lygouras,
Junyi Zhang,
Jonah Gautreau,
Mathew Pula,
Sudarshan Sharma,
Shiyuan Gao,
Tanya Berry,
Thomas Halloran,
Peter Orban,
Gael Grissonnanche,
Juan R. Chamorro,
Kagetora Mikuri,
Dilip K. Bhoi,
Maxime A. Siegler,
Kenneth K. Livi,
Yoshiya Uwatoko,
Satoru Nakatsuji,
B. J. Ramshaw,
Yi Li,
Graeme M. Luke,
Collin L. Broholm,
Tyrel M. McQueen
Abstract:
Studying superconductivity in Dirac semimetals is an important step in understanding quantum matter with topologically non-trivial order parameters. We report on the properties of the superconducting phase in single crystals of the Dirac material LaCuSb2 prepared by the self-flux method. We find that chemical and hydrostatic pressure drastically suppress the superconducting transition. Furthermore…
▽ More
Studying superconductivity in Dirac semimetals is an important step in understanding quantum matter with topologically non-trivial order parameters. We report on the properties of the superconducting phase in single crystals of the Dirac material LaCuSb2 prepared by the self-flux method. We find that chemical and hydrostatic pressure drastically suppress the superconducting transition. Furthermore, due to large Fermi surface anisotropy, magnetization and muon spin relaxation measurements reveal Type-II superconductivity for applied magnetic fields along the $a$-axis, and Type-I superconductivity for fields along the $c$-axis. Specific heat confirms the bulk nature of the transition, and its deviation from single-gap $s$-wave BCS theory suggests multigap superconductivity. Our tight-binding model points to an anisotropic gap function arising from the spin-orbital texture near the Dirac nodes, providing an explanation for the appearance of an anomaly in specific heat well below $T_c$. Given the existence of superconductivity in a material harboring Dirac fermions, LaCuSb2 proves an interesting material candidate in the search for topological superconductivity.
△ Less
Submitted 4 July, 2023;
originally announced July 2023.
-
Formation of a simple cubic antiferromagnet through charge ordering in a double Dirac material
Authors:
T. Berry,
V. C. Morano,
T. Halloran,
X. Zhang,
T. J. Slade,
A. Sapkota,
S. L. Budko,
W. Xie,
D. H. Ryan,
Z. Xu,
Y. Zhao,
J. W. Lynn,
T. Fennell,
P. C. Canfield,
C. L. Broholm,
T. M. McQueen
Abstract:
The appearance of spontaneous charge order in chemical systems is often associated with the emergence of novel, and useful, properties. Here we show through single crystal diffraction that the Eu ions in the mixed valent metal EuPd$_3$S$_4$ undergo long-range charge ordering at $T_{\mathrm{CO}} = 340 \mathrm{~K}$ resulting in simple cubic lattices of Eu$^{2+}$ ($J = 7/2$) and Eu$^{3+}$ ($J = 0$) i…
▽ More
The appearance of spontaneous charge order in chemical systems is often associated with the emergence of novel, and useful, properties. Here we show through single crystal diffraction that the Eu ions in the mixed valent metal EuPd$_3$S$_4$ undergo long-range charge ordering at $T_{\mathrm{CO}} = 340 \mathrm{~K}$ resulting in simple cubic lattices of Eu$^{2+}$ ($J = 7/2$) and Eu$^{3+}$ ($J = 0$) ions. As only one of the two sublattices has a non-magnetic ground state, the charge order results in the emergence of remarkably simple G-type antiferromagnetic order at $T_{\mathrm{N}} = 2.85(6) \mathrm{~K}$, observed in magnetization, specific heat, and neutron diffraction. Application of a $0.3 \mathrm{~T}$ field is sufficient to induce a spin flop transition to a magnetically polarized, but still charge ordered, state. Density functional theory calculations show that this charge order also modifies the electronic degeneracies present in the material: without charge order, EuPd$_3$S$_4$ is an example of a double Dirac material containing 8-fold degenerate electronic states, greater than the maximum degeneracy of six possible in molecular systems. The symmetry reduction from charge order transmutes 8-fold double Dirac states into 4-fold Dirac states, a degeneracy that can be preserved even in the presence of the magnetic order. Our results show not only how charge order can be used to produce interesting magnetic lattices, but also how it can be used to engineer controlled degeneracies in electronic states.
△ Less
Submitted 1 November, 2023; v1 submitted 3 March, 2023;
originally announced March 2023.
-
Spin-orbital order and excitons in magnetoresistive HoBi
Authors:
J. Gaudet,
H. -Y. Yang,
E. M. Smith,
T. Halloran,
J. P. Clancy,
J. A. Rodriguez-Rivera,
Guangyong Xu,
Y. Zhao,
W. C. Chen,
G. Sala,
A. A. Aczel,
B. D. Gaulin,
F. Tafti,
C. Broholm
Abstract:
The magnetism of the rock-salt $fcc$ rare-earth monopnictide HoBi, a candidate topological material with extreme magnetoresistance, is investigated. From the Ho$^{3+}$ non-Kramers $J$=8 spin-orbital multiplet, the cubic crystal electric field yields six nearly degenerate low-energy levels. These constitute an anisotropic magnetic moment with a Jahn-Teller-like coupling to the lattice. In the cubic…
▽ More
The magnetism of the rock-salt $fcc$ rare-earth monopnictide HoBi, a candidate topological material with extreme magnetoresistance, is investigated. From the Ho$^{3+}$ non-Kramers $J$=8 spin-orbital multiplet, the cubic crystal electric field yields six nearly degenerate low-energy levels. These constitute an anisotropic magnetic moment with a Jahn-Teller-like coupling to the lattice. In the cubic phase for $T>T_N~=~5.72(1)~K$, the paramagnetic neutron scattering is centered at $\mathbf{k}=(\frac{1}{2}\frac{1}{2}\frac{1}{2})$ and was fit to dominant antiferromagnetic interactions between Ho spins separated by $\{100\}$ and ferromagnetic interactions between spins displaced by $\{\frac{1}{2}\frac{1}{2}0\}$. For $T<T_N$, a type-II AFM long-range order with $\mathbf{k}=(\frac{1}{2}\frac{1}{2}\frac{1}{2})$ develops along with a tetragonal lattice distortion. While neutron diffraction from a multi-domain sample cannot unambiguously determine the spin orientation within a domain, the bulk magnetization, structural distortion, and our measurements of the magnetic excitations all show the easy axis coincides with the tetragonal axis. The weakly dispersive excitons for $T<T_N$ can be accounted for by a spin Hamiltonian that includes the crystal electric field and exchange interactions within the Random Phase Approximation.
△ Less
Submitted 12 January, 2023;
originally announced January 2023.
-
Chemical tuning of a honeycomb magnet through a critical point
Authors:
Austin M. Ferrenti,
Maxime A. Siegler,
Shreenanda Ghosh,
Xin Zhang,
Nina Kintop,
Hector K. Vivanco,
Chris Lygouras,
Thomas Halloran,
Sebastian Klemenz,
Collin Broholm,
Natalia Drichko,
Tyrel M. McQueen
Abstract:
BaCo2(AsO4)2 (BCAO) has seen extensive study since its initial identification as a proximate Kitaev quantum spin liquid candidate. Thought to be described by the highly anisotropic XXZ-J_1-J_3 model, the ease with which magnetic order is suppressed in the system indicates proximity to a spin liquid phase. Upon chemical tuning via partial arsenic substitution with vanadium, we show an initial suppr…
▽ More
BaCo2(AsO4)2 (BCAO) has seen extensive study since its initial identification as a proximate Kitaev quantum spin liquid candidate. Thought to be described by the highly anisotropic XXZ-J_1-J_3 model, the ease with which magnetic order is suppressed in the system indicates proximity to a spin liquid phase. Upon chemical tuning via partial arsenic substitution with vanadium, we show an initial suppression of long-range incommensurate order in the BCAO system to T = 3.0 K, followed by increased spin freezing at higher substitution levels. Between these two regions, at around 10% substitution, the system is shown to pass through a critical point where the competing J_1/J_3 exchange interactions become more balanced, producing a more complex magnetic ground state, likely stabilized by quantum fluctuations. This state shows how slight compositional change in magnetically-frustrated systems may be leveraged to tune ground state degeneracies and potentially realize a quantum spin liquid state.
△ Less
Submitted 12 April, 2023; v1 submitted 25 October, 2022;
originally announced October 2022.
-
Geometrical frustration versus Kitaev interactions in BaCo$_2$(AsO$_4$)$_2$
Authors:
Thomas Halloran,
Félix Desrochers,
Emily Z. Zhang,
Tong Chen,
Li Ern Chern,
Zhijun Xu,
Barry Winn,
M. K. Graves-Brook,
M. B. Stone,
Alexander I. Kolesnikov,
Yiming Qui,
Ruidan Zhong,
Robert Cava,
Yong Baek Kim,
Collin Broholm
Abstract:
Recently, Co-based honeycomb magnets have been proposed as promising candidate materials to host the Kitaev spin liquid state. One of the front-runners is BaCo$_2$(AsO$_4$)$_2$ (BCAO), where it was suggested that the exchange processes between Co$^{2+}$ ions via the surrounding edge-sharing oxygen octahedra could give rise to bond-dependent Kitaev interactions. In this work, we present and analyze…
▽ More
Recently, Co-based honeycomb magnets have been proposed as promising candidate materials to host the Kitaev spin liquid state. One of the front-runners is BaCo$_2$(AsO$_4$)$_2$ (BCAO), where it was suggested that the exchange processes between Co$^{2+}$ ions via the surrounding edge-sharing oxygen octahedra could give rise to bond-dependent Kitaev interactions. In this work, we present and analyze comprehensive inelastic neutron scattering studies of BCAO with fields in the honeycomb plane. Combining the constraints from the magnon excitations in the high-field polarized state and the inelastic spin structure factor measured in zero magnetic field, we examine two leading theoretical models: the Kitaev-type \JKG model and the \XXZ model. We show that the existing experimental data can be consistently accounted for by the \XXZ model but not by the \JKG model, and we discuss the implications of these results for the realization of a spin liquid phase in BCAO and more generally for the realization of the Kitaev model in cobaltates.
△ Less
Submitted 30 May, 2022;
originally announced May 2022.
-
Magnetic Excitations and Interactions in the Kitaev Hyperhoneycomb Iridate $β$-Li$_2$IrO$_3$
Authors:
Thomas Halloran,
Yishu Wang,
Mengqun Li,
Ioannis Rousochatzakis,
Prashant Chauhan,
M. B. Stone,
Tomohiro Takayama,
Hidenori Takagi,
N. P. Armitage,
Natalia B. Perkins,
Collin Broholm
Abstract:
We present a thorough experimental study of the three-dimensional hyperhoneycomb Kitaev magnet $β$-Li$_2$IrO$_3$, using a combination of inelastic neutron scattering (INS), time-domain THz spectroscopy, and heat capacity measurements. The main results include a massive low-temperature reorganization of the INS spectral weight that evolves into a broad peak centered around 12 meV, and a distinctive…
▽ More
We present a thorough experimental study of the three-dimensional hyperhoneycomb Kitaev magnet $β$-Li$_2$IrO$_3$, using a combination of inelastic neutron scattering (INS), time-domain THz spectroscopy, and heat capacity measurements. The main results include a massive low-temperature reorganization of the INS spectral weight that evolves into a broad peak centered around 12 meV, and a distinctive peak in the THz data at 2.8(1) meV. A detailed comparison to powder-averaged spin-wave theory calculations reveals that the positions of these two features are controlled by the anisotropic $Γ$ coupling and the Heisenberg exchange $J$, respectively. The refined microscopic spin model places $β$-Li$_2$IrO$_3$ in close proximity to the Kitaev spin liquid phase.
△ Less
Submitted 25 July, 2022; v1 submitted 26 April, 2022;
originally announced April 2022.
-
A magnetic continuum observed by terahertz spectroscopy in a quantum spin liquid candidate BaCo$_2$(AsO$_4$)$_2$
Authors:
Xinshu Zhang,
Yuanyuan Xu,
T. Halloran,
Ruidan Zhong,
R. J. Cava,
C. Broholm,
N. Drichko,
N. P. Armitage
Abstract:
Quantum spin liquids (QSLs) are topologically ordered exotic states of matter that host fractionalized excitations. Kitaev proposed a particular route towards a QSL via strongly bond-dependent interactions on the hexagonal lattice. A number of candidate Kitaev QSL materials have been pursued, but all have appreciable non-Kitaev interactions, which put these systems far from the QSL regime. Using t…
▽ More
Quantum spin liquids (QSLs) are topologically ordered exotic states of matter that host fractionalized excitations. Kitaev proposed a particular route towards a QSL via strongly bond-dependent interactions on the hexagonal lattice. A number of candidate Kitaev QSL materials have been pursued, but all have appreciable non-Kitaev interactions, which put these systems far from the QSL regime. Using time-domain terahertz spectroscopy (TDTS) we observed a broad magnetic continuum over a wide range of temperature and field in the honeycomb cobalt-based magnet, BaCo$_2$(AsO$_4$)$_2$, which has been proposed to be more ideal versions of a Kitaev QSL. Applying a small in-plane magnetic field of $\sim$ 0.5 T suppresses the magnetic order and at at even higher fields gives rise to a spin-polarized state. With 4T magnetic field oriented principally out-of-plane, a broad magnetic continuum was observed that could be consistent with a field induced QSL. Our results indicate BaCo$_2$(AsO$_4$)$_2$ to be a promising QSL candidate.
△ Less
Submitted 19 May, 2022; v1 submitted 25 June, 2021;
originally announced June 2021.
-
Quasi-one-dimensional exchange interactions and short-range magnetic correlations in CuTeO4
Authors:
Zubia Hasan,
Eli Zoghlin,
Michal J. Winiarski,
Kathryn E. Arpino,
Thomas Halloran,
Thao T. Tran,
Tyrel M. McQueen
Abstract:
CuTeO4 has been proposed as a crystallographically distinct, yet electronic structure analog, of the superconducting cuprates. Here, we present detailed characterization of the of the physical properties of CuTeO4 to address this proposal. Fitting of magnetic susceptibility data indicates unexpected quasi-one-dimensional, antiferromagnetic correlations at high temperature, with a nearest neighbour…
▽ More
CuTeO4 has been proposed as a crystallographically distinct, yet electronic structure analog, of the superconducting cuprates. Here, we present detailed characterization of the of the physical properties of CuTeO4 to address this proposal. Fitting of magnetic susceptibility data indicates unexpected quasi-one-dimensional, antiferromagnetic correlations at high temperature, with a nearest neighbour Heisenberg exchange of J1=165(4) K. Low temperature heat capacity measurements reveal a sizable T-linear contribution of $γ= 9.58(8) mJ mol^{-1} K^{-2}$, qualitatively consistent with expectations for a S=1/2, uniform, Heisenberg spin chain. Below T ~ 40 K, the susceptibility shows an upturn inconsistent with quasi-one-dimensional behaviour. While heat capacity measurements show no signs of magnetic order down to low temperature, the upturn in the magnetic susceptibility coincides with the emergence of a diffuse peak (centered at |Q| ~ 0.7 Angstrom) in the neutron diffraction data, indicative of persistent, short-range, antiferromagnetic order with a correlation length of $ξ$ = 10.1(9) Angstrom at T = 10 K. The onset of nonlinearity and hysteresis in the isothermal magnetization curves suggest the presence of a small ferromagnetic component. This persistent, short-range order is understood in the context of structural modeling of the x-ray and neutron diffraction data that show the presence of a significant density of stacking faults. No evidence for substantive dopability is observed and CuTeO4 appears qualitatively, to have a larger band gap than predicted by density functional theory. We ascribe this finding to the inductive withdrawal effect from high-valence Te and suggest that superconductivity in the copper tellurates is more likely to be found in compounds with a decreased reductive withdrawal effect from Te.
△ Less
Submitted 29 May, 2024; v1 submitted 11 December, 2020;
originally announced December 2020.
-
Monopolar and dipolar relaxation in spin ice Ho$_2$Ti$_2$O$_7$
Authors:
Yishu Wang,
T. Reeder,
Y. Karaki,
J. Kindervater,
T. Halloran,
N. Maliszewskyj,
Yiming Qiu,
J. A. Rodriguez,
S. Gladchenko,
S. M. Koohpayeh,
S. Nakatsuji,
C. Broholm
Abstract:
When degenerate states are separated by large energy barriers, the approach to thermal equilibrium can be slow enough that physical properties are defined by the thermalization process rather than the equilibrium. The exploration of thermalization pushes experimental boundaries and provides refreshing insights into atomic scale correlations and processes that impact steady state dynamics and prosp…
▽ More
When degenerate states are separated by large energy barriers, the approach to thermal equilibrium can be slow enough that physical properties are defined by the thermalization process rather than the equilibrium. The exploration of thermalization pushes experimental boundaries and provides refreshing insights into atomic scale correlations and processes that impact steady state dynamics and prospects for realizing solid state quantum entanglement. We present a comprehensive study of magnetic relaxation in Ho$_2$Ti$_2$O$_7$ based on frequency-dependent susceptibility measurements and neutron diffraction studies of the real-time atomic-scale response to field quenches. Covering nearly ten decades in time scales, these experiments uncover two distinct relaxation processes that dominate in different temperature regimes. At low temperatures (0.6K<T<1K) magnetic relaxation is associated with monopole motion along the applied field direction through the spin-ice vacuum. The increase of the relaxation time upon cooling indicates reduced monopole conductivity driven by decreasing monopole concentration and mobility as in a semiconductor. At higher temperatures (1K<T<2K) magnetic relaxation is associated with the reorientation of monopolar bound states as the system approaches the single-spin tunneling regime. Spin fractionalization is thus directly exposed in the relaxation dynamics.
△ Less
Submitted 12 November, 2020;
originally announced November 2020.
-
Low Energy Magneto-optics of Tb$_{2}$Ti$_{2}$O$_{7}$ in [111] Magnetic Field
Authors:
Xinshu Zhang,
Yi Luo,
T. Halloran,
J. Gaudet,
Huiyuan Man,
S. M. Koohpayeh,
N. P. Armitage
Abstract:
The pyrochlore magnet Tb$_{2}$Ti$_{2}$O$_{7}$ shows a lack of magnetic order to low temperatures and is considered to be a quantum spin liquid candidate. We perform time-domain THz spectroscopy on high quality Tb$_{2}$Ti$_{2}$O$_{7}$ crystal and study the low energy excitations as a function of [111] magnetic field with high energy resolution. The low energy crystal field excitations change their…
▽ More
The pyrochlore magnet Tb$_{2}$Ti$_{2}$O$_{7}$ shows a lack of magnetic order to low temperatures and is considered to be a quantum spin liquid candidate. We perform time-domain THz spectroscopy on high quality Tb$_{2}$Ti$_{2}$O$_{7}$ crystal and study the low energy excitations as a function of [111] magnetic field with high energy resolution. The low energy crystal field excitations change their energies anomalously under magnetic field. Despite several sharp field dependent changes, we show that the material's spectrum can be described not by a phase transitions, but by field dependent hybridization between the low energy crystal field levels. We highlight the strong coupling between spin and lattice degrees of freedom in Tb$_{2}$Ti$_{2}$O$_{7}$ as evidenced by the magnetic field tunable crystal field environment. Calculations based on single ion physics with field induced symmetry reduction of the crystal field environment can reproduce our data.
△ Less
Submitted 25 September, 2020;
originally announced September 2020.
-
GPU-Accelerated Primal Learning for Extremely Fast Large-Scale Classification
Authors:
John T. Halloran,
David M. Rocke
Abstract:
One of the most efficient methods to solve L2-regularized primal problems, such as logistic regression and linear support vector machine (SVM) classification, is the widely used trust region Newton algorithm, TRON. While TRON has recently been shown to enjoy substantial speedups on shared-memory multi-core systems, exploiting graphical processing units (GPUs) to speed up the method is significantl…
▽ More
One of the most efficient methods to solve L2-regularized primal problems, such as logistic regression and linear support vector machine (SVM) classification, is the widely used trust region Newton algorithm, TRON. While TRON has recently been shown to enjoy substantial speedups on shared-memory multi-core systems, exploiting graphical processing units (GPUs) to speed up the method is significantly more difficult, owing to the highly complex and heavily sequential nature of the algorithm. In this work, we show that using judicious GPU-optimization principles, TRON training time for different losses and feature representations may be drastically reduced. For sparse feature sets, we show that using GPUs to train logistic regression classifiers in LIBLINEAR is up to an order-of-magnitude faster than solely using multithreading. For dense feature sets--which impose far more stringent memory constraints--we show that GPUs substantially reduce the lengthy SVM learning times required for state-of-the-art proteomics analysis, leading to dramatic improvements over recently proposed speedups. Furthermore, we show how GPU speedups may be mixed with multithreading to enable such speedups when the dataset is too large for GPU memory requirements; on a massive dense proteomics dataset of nearly a quarter-billion data instances, these mixed-architecture speedups reduce SVM analysis time from over half a week to less than a single day while using limited GPU memory.
△ Less
Submitted 14 October, 2020; v1 submitted 7 August, 2020;
originally announced August 2020.
-
Learning Concave Conditional Likelihood Models for Improved Analysis of Tandem Mass Spectra
Authors:
John T. Halloran,
David M. Rocke
Abstract:
The most widely used technology to identify the proteins present in a complex biological sample is tandem mass spectrometry, which quickly produces a large collection of spectra representative of the peptides (i.e., protein subsequences) present in the original sample. In this work, we greatly expand the parameter learning capabilities of a dynamic Bayesian network (DBN) peptide-scoring algorithm,…
▽ More
The most widely used technology to identify the proteins present in a complex biological sample is tandem mass spectrometry, which quickly produces a large collection of spectra representative of the peptides (i.e., protein subsequences) present in the original sample. In this work, we greatly expand the parameter learning capabilities of a dynamic Bayesian network (DBN) peptide-scoring algorithm, Didea, by deriving emission distributions for which its conditional log-likelihood scoring function remains concave. We show that this class of emission distributions, called Convex Virtual Emissions (CVEs), naturally generalizes the log-sum-exp function while rendering both maximum likelihood estimation and conditional maximum likelihood estimation concave for a wide range of Bayesian networks. Utilizing CVEs in Didea allows efficient learning of a large number of parameters while ensuring global convergence, in stark contrast to Didea's previous parameter learning framework (which could only learn a single parameter using a costly grid search) and other trainable models (which only ensure convergence to local optima). The newly trained scoring function substantially outperforms the state-of-the-art in both scoring function accuracy and downstream Fisher kernel analysis. Furthermore, we significantly improve Didea's runtime performance through successive optimizations to its message passing schedule and derive explicit connections between Didea's new concave score and related MS/MS scoring functions.
△ Less
Submitted 4 September, 2019;
originally announced September 2019.
-
Gradients of Generative Models for Improved Discriminative Analysis of Tandem Mass Spectra
Authors:
John T. Halloran,
David M. Rocke
Abstract:
Tandem mass spectrometry (MS/MS) is a high-throughput technology used toidentify the proteins in a complex biological sample, such as a drop of blood. A collection of spectra is generated at the output of the process, each spectrum of which is representative of a peptide (protein subsequence) present in the original complex sample. In this work, we leverage the log-likelihood gradients of generati…
▽ More
Tandem mass spectrometry (MS/MS) is a high-throughput technology used toidentify the proteins in a complex biological sample, such as a drop of blood. A collection of spectra is generated at the output of the process, each spectrum of which is representative of a peptide (protein subsequence) present in the original complex sample. In this work, we leverage the log-likelihood gradients of generative models to improve the identification of such spectra. In particular, we show that the gradient of a recently proposed dynamic Bayesian network (DBN) may be naturally employed by a kernel-based discriminative classifier. The resulting Fisher kernel substantially improves upon recent attempts to combine generative and discriminative models for post-processing analysis, outperforming all other methods on the evaluated datasets. We extend the improved accuracy offered by the Fisher kernel framework to other search algorithms by introducing Theseus, a DBN representing a large number of widely used MS/MS scoring functions. Furthermore, with gradient ascent and max-product inference at hand, we use Theseus to learn model parameters without any supervision.
△ Less
Submitted 4 September, 2019;
originally announced September 2019.
-
Jensen: An Easily-Extensible C++ Toolkit for Production-Level Machine Learning and Convex Optimization
Authors:
Rishabh Iyer,
John T. Halloran,
Kai Wei
Abstract:
This paper introduces Jensen, an easily extensible and scalable toolkit for production-level machine learning and convex optimization. Jensen implements a framework of convex (or loss) functions, convex optimization algorithms (including Gradient Descent, L-BFGS, Stochastic Gradient Descent, Conjugate Gradient, etc.), and a family of machine learning classifiers and regressors (Logistic Regression…
▽ More
This paper introduces Jensen, an easily extensible and scalable toolkit for production-level machine learning and convex optimization. Jensen implements a framework of convex (or loss) functions, convex optimization algorithms (including Gradient Descent, L-BFGS, Stochastic Gradient Descent, Conjugate Gradient, etc.), and a family of machine learning classifiers and regressors (Logistic Regression, SVMs, Least Square Regression, etc.). This framework makes it possible to deploy and train models with a few lines of code, and also extend and build upon this by integrating new loss functions and optimization algorithms.
△ Less
Submitted 17 July, 2018;
originally announced July 2018.
-
A Geometrically Frustrated Trimer-Based Mott Insulator
Authors:
Loi T. Nguyen,
T. Halloran,
Weiwei Xie,
Tai Kong,
C. L. Broholm,
R. J. Cava
Abstract:
The crystal structure of Ba4NbRu3O12 is based on triangular planes of elongated Ru3O12 trimers oriented perpendicular to the plane. We report that it is semiconducting, that its Weiss temperature and effective magnetic moment are -155 K and 2.59 μB/f.u. respectively, and that magnetic susceptibility and specific heat data indicate that it exhibits magnetic ordering near 4 K. The presence of a high…
▽ More
The crystal structure of Ba4NbRu3O12 is based on triangular planes of elongated Ru3O12 trimers oriented perpendicular to the plane. We report that it is semiconducting, that its Weiss temperature and effective magnetic moment are -155 K and 2.59 μB/f.u. respectively, and that magnetic susceptibility and specific heat data indicate that it exhibits magnetic ordering near 4 K. The presence of a high density of low energy states is evidenced by a substantial Sommerfeld-like T-linear term (gamma = 31(2) mJ/mole-K^2) in the specific heat. Electronic structure calculations reveal that the electronic states at the Fermi Energy reside on the Ru3O12 trimers and that the calculated density of electronic states is high and continuous around the Fermi Energy - in other words density functional theory calculates the material to be a metal. Our results imply that Ba4NbRu3O12 is a geometrically frustrated trimer-based Mott insulator.
△ Less
Submitted 8 May, 2018;
originally announced May 2018.
-
Faster graphical model identification of tandem mass spectra using peptide word lattices
Authors:
Shengjie Wang,
John T. Halloran,
Jeff A. Bilmes,
William S. Noble
Abstract:
Liquid chromatography coupled with tandem mass spectrometry, also known as shotgun proteomics, is a widely-used high-throughput technology for identifying proteins in complex biological samples. Analysis of the tens of thousands of fragmentation spectra produced by a typical shotgun proteomics experiment begins by assigning to each observed spectrum the peptide hypothesized to be responsible for g…
▽ More
Liquid chromatography coupled with tandem mass spectrometry, also known as shotgun proteomics, is a widely-used high-throughput technology for identifying proteins in complex biological samples. Analysis of the tens of thousands of fragmentation spectra produced by a typical shotgun proteomics experiment begins by assigning to each observed spectrum the peptide hypothesized to be responsible for generating the spectrum, typically done by searching each spectrum against a database of peptides. We have recently described a machine learning method---Dynamic Bayesian Network for Rapid Identification of Peptides (DRIP)---that not only achieves state-of-the-art spectrum identification performance on a variety of datasets but also provides a trainable model capable of returning valuable auxiliary information regarding specific peptide-spectrum matches. In this work, we present two significant improvements to DRIP. First, we describe how to use word lattices, which are widely used in natural language processing, to significantly speed up DRIP's computations. To our knowledge, all existing shotgun proteomics search engines compute independent scores between a given observed spectrum and each possible candidate peptide from the database. The key idea of the word lattice is to represent the set of candidate peptides in a single data structure, thereby allowing sharing of redundant computations among the different candidates. We demonstrate that using lattices in conjunction with DRIP leads to speedups on the order of tens across yeast and worm data sets. Second, we introduce a variant of DRIP that uses a discriminative training framework, performing maximum mutual entropy estimation rather than maximum likelihood estimation. This modification improves DRIP's statistical power, enabling us to increase the number of identified spectrum at a 1% false discovery rate on yeast and worm data sets.
△ Less
Submitted 29 October, 2014;
originally announced October 2014.
