arXiv subjects

Jiajun Li

Publications and source records attributed to Jiajun Li.

101 records · Page 6Linked to original sources

arXiv2019

Revealing Hund's multiplets in Mott insulators under strong electric fields

We investigate the strong-field dynamics of a paramagnetic two-band Mott insulator using real-time dynamical mean-field theory. A dielectric breakdown occurs due to many-body Landau-Zener tunnelling, with a threshold field determined by the gap. For a large range of fields, however, we predict that the tunnelling currents are small enough to allow the observation of field-induced localization of electrons, which becomes most strikingly evident in atomic-like local spin multiplets determined by the Hund's coupling $J$. This field-induced localization might provide a way of measuring the value of $J$ in correlated materials. It should be observable in transition metal oxides using time-resolved photo-emission spectroscopy or optical measurements in the presence of strong THz field transients.

cond-mat.str-el↗

arXiv2018

Control of Competing Superconductivity and Charge Order by Non-equilibrium Currents

We study the competing charge-density-wave and superconducting order in the attractive Hubbard model under a voltage bias, using steady-state non-equilibrium dynamical mean-field theory. We show that the charge-density-wave is suppressed in a current-carrying non-equilibrium steady state. This effect is beyond a simple Joule-heating mechanism and a "supercooled" metallic state is stabilized at a non-equilibrium temperature lower than the equilibrium superconducting $T_c$. On the other hand, a current-carrying superconducting state is always in equilibrium. It is not subject to the same non-thermal suppression, and can therefore nucleate out of the supercooled metal, e.g. in a resistive switching experiment. The fact that an electric current can change the relative stability of different phases compared to thermal equilibrium, even when a system appears locally thermal due to electron-eletron scattering, provides a general perspective to control intertwined orders out of equilibrium.

cond-mat.supr-con↗

arXiv2018

Theory of photoinduced ultrafast switching to a spin-orbital ordered `hidden' phase

Photoinduced hidden phases are often observed in materials with intertwined orders. Understanding the formation of these non-thermal phases is challenging and requires a resolution of the cooperative interplay between different orders on the ultrashort timescale. In this work, we demonstrate that non-equilibrium photo-excitations can induce a state with spin-orbital orders entirely different from the equilibrium state in the three-quarter-filled two-band Hubbard model. We identify a general mechanism governing the transition to the hidden state, which relies on a non-thermal partial melting of the intertwined orders mediated by photoinduced charge excitations in the presence of strong spin-orbital exchange interactions. Our study theoretically confirms the crucial role played by orbital degrees of freedom in the light-induced dynamics of strongly correlated materials and it shows that the switching to hidden states can be controlled already on the fs timescale of the electron dynamics.

cond-mat.str-el↗

arXiv2018

Learning Multi-touch Conversion Attribution with Dual-attention Mechanisms for Online Advertising

In online advertising, the Internet users may be exposed to a sequence of different ad campaigns, i.e., display ads, search, or referrals from multiple channels, before led up to any final sales conversion and transaction. For both campaigners and publishers, it is fundamentally critical to estimate the contribution from ad campaign touch-points during the customer journey (conversion funnel) and assign the right credit to the right ad exposure accordingly. However, the existing research on the multi-touch attribution problem lacks a principled way of utilizing the users' pre-conversion actions (i.e., clicks), and quite often fails to model the sequential patterns among the touch points from a user's behavior data. To make it worse, the current industry practice is merely employing a set of arbitrary rules as the attribution model, e.g., the popular last-touch model assigns 100% credit to the final touch-point regardless of actual attributions. In this paper, we propose a Dual-attention Recurrent Neural Network (DARNN) for the multi-touch attribution problem. It learns the attribution values through an attention mechanism directly from the conversion estimation objective. To achieve this, we utilize sequence-to-sequence prediction for user clicks, and combine both post-view and post-click attribution patterns together for the final conversion estimation. To quantitatively benchmark attribution models, we also propose a novel yet practical attribution evaluation scheme through the proxy of budget allocation (under the estimated attributions) over ad channels. The experimental results on two real datasets demonstrate the significant performance gains of our attribution model against the state of the art.

cs.IR↗

arXiv2018

Nonequilibrium Excitations and Transport of Dirac Electrons in Electric-Field-Driven Graphene

We investigate nonequilibrium excitations and charge transport in charge-neutral graphene driven with DC electric field by using the nonequilibrium Green's function technique. Due to the vanishing Fermi surface, electrons are subject to non-trivial nonequilibrium excitations such as highly anisotropic momentum distribution of electron-hole pairs, an analog of the Schwinger effect. We show that the electron-hole excitations, initiated by the Landau-Zener tunneling with a superlinear IV relation $I \propto E^{3/2}$, reaches a steady-state dominated by the dissipation due to optical phonons, resulting in a marginally sublinear IV with $I \propto E$, in agreement with recent experiments. The linear IV starts to show the sign of current saturation as the graphene is doped away from the Dirac point, and recovers the semi-classical relation for the saturated velocity. We give a detailed discussion on the nonequilibrium charge creation and the relation between the electron-phonon scattering rate and the electric field in the steady-state limit. We explain how the apparent Ohmic IV is recovered near the Dirac point. We propose a mechanism where the peculiar nonequilibrium electron-hole creation can be utilized in a novel infra-red device.

cond-mat.mes-hall↗

arXiv2018

Nonequilibrium Mean-Field Theory of Resistive Phase Transitions

We investigate the quantum mechanical origin of resistive phase transitions in solids driven by a constant electric field in the vicinity of a metal-insulator transition. We perform a nonequilibrium mean-field analysis of a driven-dissipative anti-ferromagnet, which we solve analytically for the most part. We find that the insulator-to-metal transition (IMT) and the metal-to-insulator transition (MIT) proceed by two distinct electronic mechanisms: Landau-Zener processes, and the destabilization of metallic state by Joule heating, respectively. However, we show that both regimes can be unified in a common effective thermal description, where the effective temperature $T_{\rm eff}$ depends on the state of the system. This explains recent experimental measurements in which the hot-electron temperature at the IMT was found to match the equilibrium transition temperature. Our analytic approach enables us to formulate testable predictions on the non-analytic behavior of $I$-$V$ relation near the insulator-to-metal transition. Building on these successes, we propose an effective Ginzburg-Landau theory which paves the way to incorporating spatial fluctuations, and to bringing the theory closer to a realistic description of the resistive switchings in correlated materials.

cond-mat.str-el↗

arXiv2017

Personalizing a Dialogue System with Transfer Reinforcement Learning

It is difficult to train a personalized task-oriented dialogue system because the data collected from each individual is often insufficient. Personalized dialogue systems trained on a small dataset can overfit and make it difficult to adapt to different user needs. One way to solve this problem is to consider a collection of multiple users' data as a source domain and an individual user's data as a target domain, and to perform a transfer learning from the source to the target domain. By following this idea, we propose "PETAL"(PErsonalized Task-oriented diALogue), a transfer-learning framework based on POMDP to learn a personalized dialogue system. The system first learns common dialogue knowledge from the source domain and then adapts this knowledge to the target user. This framework can avoid the negative transfer problem by considering differences between source and target users. The policy in the personalized POMDP can learn to choose different actions appropriately for different users. Experimental results on a real-world coffee-shopping data and simulation data show that our personalized dialogue system can choose different optimal actions for different users, and thus effectively improve the dialogue quality under the personalized setting.

cs.AI↗

arXiv2017

Microscopic Theory of Resistive Switching in Ordered Insulators: Electronic vs. Thermal Mechanism

We investigate the dramatic switch of resistance in ordered correlated insulators, when driven out of equilibrium by a strong voltage bias. Microscopic calculations on a driven-dissipative lattice of interacting electrons explain the main experimental features of resistive switching (RS), such as the hysteretic $I$-$V$ curves and the formation of hot conductive filaments. The energy-resolved electron distribution at the RS reveals the underlying nonequilibrium electronic mechanism, namely Landau-Zener tunneling, and also justifies a thermal description where the hot-electron temperature, estimated from the first moment of the distribution, matches the equilibrium phase transition temperature. We discuss the tangled relationship between filament growth and negative differential resistance, and the influence of crystallographic structure and disorder in the RS.

cond-mat.str-el↗

arXiv2016

RGBD-based Parameter Extraction for Door Opening Tasks with Human Assists in Nuclear Rescue

The ability to open a door is essential for robots to perform home-serving and rescuing tasks. A substantial problem is to obtain the necessary parameters such as the width of the door and the length of the handle. Many researchers utilize computer vision techniques to extract the parameters automatically which lead to fine but not very stable results because of the complexity of the environment. We propose a method that utilizes an RGBD sensor and a GUI for users to 'point' at the target region with a mouse to acquire 3D information. Algorithms that can extract important parameters from the selected points are designed. To avoid large internal force induced by the misalignment of the robot orientation and the normal of the door plane, we design a module that can compute the normal of the plane by pointing at three non-collinear points and then drive the robot to the desired orientation. We carried out experiments on real robot. The result shows that the designed GUI and algorithms can help find the necessary parameters stably and get the robot prepared for further operations.

cs.RO↗

arXiv2015

Electric-field-driven resistive switching in dissipative Hubbard model

We study how strongly correlated electrons on a dissipative lattice evolve from equilibrium when driven by a constant electric field, focusing on the extent of the linear regime and hysteretic non-linear effects at higher fields. We access the non-equilibrium steady states, non-perturbatively in both the field and the electronic interactions, by means of a non-equilibrium dynamical mean-field theory in the Coulomb gauge. The linear response regime is limited by Joule heating effects and breaks down at fields orders of magnitude smaller than the quasi-particle energy scale. For large electronic interactions, strong but experimentally accessible electric fields can induce a resistive switching by driving the strongly correlated metal into a Mott insulator. Hysteretic $I$-$V$ curves suggest that the non-equilibrium current is carried through a spatially inhomogeneous metal-insulator mixed state.

cond-mat.str-el↗

arXiv2013

Energy dissipation in DC-field driven electron lattice coupled to fermion baths

Electron transport in electric-field-driven tight-binding lattice coupled to fermion baths is comprehensively studied. We reformulate the problem by using the scattering state method within the Coulomb gauge. Calculations show that the formulation justifies direct access to the steady-state bypassing the time-transient calculations, which then makes the steady-state methods developed for quantum dot theories applicable to lattice models. We show that the effective temperature of the hot-electron induced by a DC electric field behaves as $T_{\rm eff}=Cγ(Ω/Γ)$ with a numerical constant $C$, tight-binding parameter $γ$, the Bloch oscillation frequency $Ω$ and the damping parameter $Γ$. In the small damping limit $Γ/Ω\to 0$, the steady-state has a singular property with the electron becoming extremely hot in an analogy to the short-circuit effect. This leads to the conclusion that the dissipation mechanism cannot be considered as an implicit process, as treated in equilibrium theories. Finally, using the energy flux relation, we derive a steady-state current for interacting models where only on-site Green's functions are necessary.

cond-mat.mes-hall↗