Laboratory* represents those who will not accept new students.
To understand natural, biological, and artificial systems, we perform mathematical modeling and analysis. Moreover, closely collaborating with experimentalists of various fields, we try to solve problems related to our lives.
Modeling and Theory Construction
By constructing simple models that describe complex dynamical phenomena, we try to understand, predict, and control such phenomena. Moreover, through the generalizing and abstraction of problems, we try to construct general theories. Examples of our subjects include biological rhythms, locomotion, hydrodynamic phenomena, power grids, transportation networks, traffic networks, pattern formation in biological and chemical systems, social systems, neural networks.
Collaboration with experimentalists
To solve problems closely related to our lives, we collaborate with researchers of various disciplines such as engineering and biology. Our roles are to provide theoretical ideas, to analyze and interpret experimental data, and to propose new experiments.
nonlinear phenomena, oscillations, synchronization, fluctuation, complex networks, control, optimization, biological rhythms, circadian rhythms, locomotion, biological physics
Web ScienceSee Details
To understand the world, e.g. biological and social systems, we are developing a method for modeling and forecasting time series, which were observed from complex systems. We are also applying the method to large dataset from the brain, the internet, and other social systems by collaborating with experts in the respective field. We are currently working in the following projects.
Event time series analysis
We focus on event time series, which means the timestamps of an event, appearing in various systems including tweets on Twitter, log data of product purchases, earthquake, and the action potentials in the brain. We are developing a method for discovering hidden rules underlying an event time series.
Modeling biological and social systems
We are developing a mathematical model that describes the observed data and analyzing the model to understand the way the brain processes information and the information diffusion on the internet. This research has implications for brain-inspired artificial intelligence and a current internet problem of the spread of fake news, misinformation, and flaming.
statistical mechanics, Nonlinear dynamicsSee Details
Many systems in our world, natural and artificial systems, are dynamical and are in nonequilibrium states accompanied by energy and mass flow. Their dynamics are typically governed by nonlinear equations. In our laboratory, through mathematical modeling and construction of phenomenological theory of specific systems, we explore fundamental laws that govern the nonlinear and nonequilibrium systems. We also aim at creating new technologies based on these laws. We have been mainly working on the following subjects:
(1) Efficiency bound of heat engines working at maximum power
While maximum efficiency of heat engines is given by Carnot efficiency, it is practically useless because power vanishes at the quasistatic limit. For this problem, we have proposed a finite-time Carnot cycle model and are approaching a fundamental aspect of nonequilibrium thermodynamics and statistical mechanics by investigating efficiency bound of heat engines working at maximum power.
(2) Physics of autonomous heat engines
Low-temperature-differential Stirling engines operate autonomously under quite a small temperature difference between body temperature and room temperature. We have elucidated a rotational mechanism of the engine by developing a nonlinear dynamics model and have been constructing a nonequilibrium thermodynamics theory of the engine. We are also interested in proposing a new type of energy devices.
(3) Energetics of synchronization in coupled oscillators
It is known that flagella of living organisms that play vital role in biological functions can be regarded as self-sustained oscillators and that they synchronize to work via hydrodynamic coupling. We have constructed energetics of synchronization in coupled oscillators and are approaching complex life phenomena by highlighting role of synchronization in biological functions.
(4) Shortcuts to adiabaticity (STA)
STA is a recently proposed method for achieving acceleration of adiabatic quantum dynamics accompanied by sufficiently slow variation of external parameters, which has been receiving attention for its potential applicability. STA has also been applied to adiabatic classical dynamics, and we are investigating an underlying mathematical mechanism of the acceleration.
Developing a Computer That Learns Like Humans
Together with the rapid progress and spread of the Internet and sensor technology,
vast amounts of data are collected in various fields of engineering, industries, and natural sciences such as speeches, images, texts, movies, social media,
E-commerce, power networks, medicine, and biology.
To create new value from such big data, machine learning plays a central role.
Machine learning is aimed at developing a computer that learns like humans. Our group studies various aspects of machine learning and statistical data analysis such as fundamental theory, practical algorithms, and application to real-world data analysis.
Machine learning to retrieve important information from big data is recently becoming a common technology in wide areas. On the other hand, there are many situations where we have to take a decision on some problem without sufficient data. For example in the advertisement systems, one cannot know the click-through rate of an ad unless displaying it whereas it is a loss to display an ad with low click-through rate many times. Another example is development of new drugs or materials. In this problem it takes some days to compute property of one candidate and it is necessary to choose a promising candidate appropriately by some mathematical model to fully utilize limited resources.
Our laboratory researches this problem of decision making with dynamically gathered information from the viewpoint of, such as, bandit problems and Bayesian optimization. In this problem we face two different aspects, that is, the statistical aspect to realize good estimation from gathered data, and the algorithmic aspect to determine the next action for improvement of estimations and rewards. How to combining these aspects is the challenging and interesting point of our problem.
Understanding What's Happening on Earth from Multimodal Spatio-Temporal Data
We are studying geoinformatics to understand the current state of the Earth from multimodal spatio-temporal data such as satellite images, mobile data, and disaster data. In particular, we are working on intelligent data analysis based on image processing and machine learning to understand what is happening on the Earth in a timely manner from a large and diverse set of images obtained from visible, near-infrared, thermal infrared, and microwave based remote sensing.
(1) Beyond human vision
Data obtained by spectroscopic imaging and synthetic aperture radar allow us to see the world invisible to humans, but there are imperfections due to sensor characteristics and atmospheric conditions. Based on mathematical optimization, machine learning, and signal processing, we aim to further advance our sensing technology beyond human vision by recovering the original signal from incomplete multidimensional observation data.
(2) Understanding Earth today
In an emergency such as a disaster, it is important to get a full picture as quickly as possible. We are developing data fusion technology that extracts change information from spatio-temporal data obtained from a variety of sensors on different platforms, from satellites to the ground, and technology that can rapidly estimate unexpected complex changes on the ground surface by combining machine learning and numerical simulation.
(3) Towards a sustainable society
We are conducting research to support solutions to international social issues, such as disaster monitoring and forest monitoring. Our goal is to contribute to the realization of a sustainable society globally by working closely with the world-leading space agencies and disaster prevention organizations to explore technologies that are truly useful in solving real-world problems.
Machine learning is a technology for a machine to automatically learn a rule for Intelligently information processing by using a large amount of data. For example, machine learning has played an important role in a real life such as a face recognition in a smart phone, a recommendation system in on-line shopping site, and automatic car driving. In our laboratory, we are studying the following theme.
(1) Mathematical modeling
In statistical machine learning, we have to design a mathematical model for data and problems to slove. We are studying a statistical model with a latent variable that represents a hidden property in data.
(2) Learning algorithm
Learning is formulated as a parameter estimation of mathematical models. We are studying a fast algorithm to estimate model parameters from a large amount of data. In particular, we focus on a learning algorithm based on the Bayes estimation.
(3) Experimental design for statistical machine learning by statistical machine learning
Regardless of statistical machine learning, it takes longer times to perform experiments in many scientific fields. We are studying a support system for researchers' experiments with long times. Our system can perform trial and error in experiments instead of a researcher.
(4) Social application
We are applying our work and recent technology of machine learning to social problems. For example, we are developing a supporting system for a medical image analysis collaborating with the University of Tokyo hospital.
Quantum Calculation, Nonliner DynamicsSee Details
Theoretical neuroscience, Statistical physics
Human Interface, Network, Physical InformaticsSee Details
We investigate technical and scientific issues related to haptics. Our goal is to develop a human activity support system while clarifying the physical mechanisms of the human tactile organ and relationships between haptic stimulation to humans and human responses in physical actions and mental statuses.
This relates to research on signal and power transmission with electromagnetic waves traveling along a thin sheet. The technology enables a wireless and battery-free information environment that provides safe wireless power transmission to items touching the sheet and high-speed signal transmission with low interference from ordinary wireless signals. This technology also contributes to wearable computing and sensor embedding in various elastic materials.
Human InterfaceSee Details
We are interested in the mechanisms of how the neural circuits develop and function to generate specific behavior, by using the nervous system of the fruit fly Drosophila as a model. In this organism, the relative simplicity and highly sophisticated genetic techniques allow one to identify and manipulate specific neurons. We focus on the larval peristalsis (waves of muscular contraction that propagate along the body) and try to understand how the motor outputs are generated by the neural circuits. For this, we use a variety of genetic and biophysical techniques. For example, we use calcium imaging to record the activity of specific population of neurons. By using a recently developed technique, called optogenetics, we manipulate the activity of specific neurons with light at high resolution. By recording and manipulating the spatio-temporal pattern of neural activity, we aim to understand the operational principle of the neural circuits.
Our main research goal is to elucidate the neuronal and network mechanisms underlying reward-related cognitive functions, particularly the contribution of the dopaminergic circuits of the basal ganglia, the prefrontal cortex and the hippocampus. Studying the physiological mechanisms of cognitive functions requires understandings not only the responses of single neurons to external stimuli but also circuit computations at the level of networks of neurons during cognitive processing. To understand the 'syntax' underlying neuronal communications, methods for monitoring and quantifying cooperative neuronal activities during cognition are required. To this end, we have been performing large-scale high-density recordings of local circuits with multi-channel silicon probes, enabling the observation of simultaneous neuronal firing activities in up to 100 neurons, as well as local field potentials in behaving animals. In addition, we are developing a new technique that combines large-scale recording and targeted simultaneous optogenetic stimulations of specific cells, such as dopaminergic neurons, to clarify the role of the different types of neurons in network processing, in freely behaving mice. Taking advantage of these methods, we try to decipher circuit computations within local and between the inter-regional networks during reward-related cognitions such as decision making and working memory.
The world is full of moving objects. Among them, animals are a special group that have survived and evolved on the earth. The most conspicuous property of animals is that they possess networks of neurons (“neural circuits”) in the body. Although these networks are just a cluster of electrically-charged cells, neural circuits have an amazing ability of conducting multiple tasks including detecting the surrounding physical world by sensory networks, changing neural circuit states by internal network dynamics, and generating adequate motor outputs by muscle contraction. How neural circuits generate coordinated and coherent activity patterns is a fundamental question in neuroscience. Recent technical advances in genetics, optical control, and image analyses enable dissecting how neural circuits work in cellular level. In our lab, in collaboration with the Nose Lab, we study physical basis in locomotion control by applying state-of-the-art techniques to fruit fly larvae. Because neuroscience is interdisciplinary, we focus on multiple axes in multiple scales:
1) Spatial scale axis: DNA, proteins (nanometer); synapses, cells, circuits (micrometer), whole animal (millimeter), and animal populations2) Time scale axis: neuronal activity (millisecond), circuit dynamics (second), motor control (minute), and evolution/speciation 3) Concrete-Abstract axis: from identification of genes and interneurons to construction of neural circuit models Based on these multiple perspectives, we are currently searching for key interneurons in the motor circuits and examining the mechanical dynamics of their motor outputs. The goal of our study is elucidating the physical basis for motor circuits by constant efforts and openminded thinking.
Fused brain measurement scienceSee Details
Neural Circuit ModelSee Details
The circulation of energy and substances in a planetary atmosphere controls the development of the planetary environment, thereby governing the possibility of a biosphere. The planetary atmospheres observed so far show surprising diversity, whose origin is still unclear. Our laboratory focuses on planetary atmosphere physics and resultant climate formation. The ultimate goal is general understanding of common physical processes behind the apparent diversity. The following observational and theoretical approaches are ongoing.
(1) Exploration of planetary atmospheres
Exploration of Venus atmosphere by a Japanese Venus explorer AKATSUKI is ongoing. We use AKATSUKI's data to unveil the mysteries of Venusian meteorology such as the high-speed westward circulation "super-rotation" and thick sulfuric acid clouds. Development of a Mars exploration program including the studies of water cycle and dust transport is also ongoing.
In a radio occultation experiment, a spacecraft transmits radio waves toward a tracking station on the earth and sequentially goes behind the planet's atmosphere; during such occultation events the planetary atmosphere cause frequency and amplitude fluctuation, from which information on the atmosphere is obtained. We apply this technique to planets and the solar corona. (3) Numerical modeling
Common physical processes behind the apparent diversity of atmospheric phenomena on the planets are investigated with numerical modeling and theories.
Exploration of the Solar System, Exploration of ExoplanetsSee Details
火星や金星は生命にとって，どれほど苛酷な環境なのだろうか？ 地球の大気環境は変化しないのだろうか？ 火星のようになったりしないだろうか？
2013年9月にイプシロンロケットで打ち上げられた惑星分光観測衛星「ひさき]（SPRINT-A）は，極端紫外光の目を持つ “宇宙望遠鏡” である． 極端紫外光は，紫外線の中でも波長の短い光で，この光で見た地球は，ふだん私たちが見ている青くて丸い地球とはずいぶん違って見える． 例えば，北極と南極を付け根にして，地球半径の5～6倍の空間に広がる蝶のような姿（双極子磁場）に満たされたプラズマ（電離した気体）が写る．
Space physicsSee Details
世界に比類がなくわが国独自の月惑星探査を遂行するための探査機器の開発およびプロジェクトの遂行を行っている． これまでJAXA（宇宙科学研究所）において「かぐや」，「はやぶさ」などの月惑星探査を成功させてきたが， それらに続くものとして月着陸探査ミッション（SELENE-2）やC型小天体サンプルリターンミッション（はやぶさ2）などが進められている． これらのミッションの科学的側面から探査戦略の追及，それを実現するたの搭載器機の性能をつきつめて実現化し， 世界トップクラスと賞されるだけでなく将来にわたって活用し続けられるデータの取得を目指す．
これからの月惑星探査の主流は内部構造探査である．これを遂行するための搭載インフラや機器（地震計や熱流量計）の開発が重要である． 我々は長年にわたり地震計などを搭載可能なペネトレータとよばれる高速貫入型の観測装置の開発に携わり，技術的に高いレベルにまで完成させた． 我々が開発したこの装置を月惑星に送り込んで内部構造に関するデータを取得し，月惑星の起源と進化に重要な制約条件を得ることが私の究極的な目標である．
Planetary Georogy, Numerical. Fluid Mechanics, Planetary ExplorationSee Details
惑星探査技術の進歩により，太陽系の天体に探査機を送り込んで調査することが可能となった． 火星探査車は砂だらけの火星表面を走り回り，小惑星探査機は，弾丸を使って岩石を破砕し岩石サンプルを収集した． 人類は，こうした太陽系の直接探査を通じて地球外の天体に関する情報を猛烈な勢いで獲得している． 太陽系科学は，革命的な発展を遂げていると言って良い．
私たちの研究室では，太陽系探査に直接関連した，以下の2つの方向性の研究を推進している． 1つ目は，探査データの解析である． 特に天体の表層環境に関する研究に重点を置いており，主に固体天体表層地形の解析を通じて，地球表層環境の持つ普遍性と特異性を明らかにするという， 比較惑星学（特に惑星地質学）分野の研究を行っている． 「人類が地球に誕生した事に必然性が存在するか」というアストロバイオロジーの大問題に，惑星探査データの解析から迫ろうとしているとも言える．
Planetary Science, Earth Initial Evolution, Super High Speed Phys.,
Super High Speed Collision Phys.See Details
私は，惑星の起源と進化を理解するため，室内実験と惑星探査の両面から研究を行っている． 室内実験では，惑星初期進化で支配的な役割を果たした小天体衝突の機構解明に力を注いでいる． 地球の基本形が作られた地球集積期やその直後の時代の表層環境の解明が目的である． 特に，大気圏や水圏の質量と組成を決定する衝突蒸発現象機構の解明のため，高速度衝突実験と高速分光計測を用いた研究を行っている． こちらは，自分だけの自由な発想で行えるタイプの研究である．
惑星探査は，他の惑星や衛星を調査して，地球との違いを明らかにすることが目的である． 2014年の打ち上げを目指す「はやぶさ2」計画に参画し，可視分光カメラ開発のサイエンス担当者を務めている． 米国がOSIRISRex計画を打ち出したので競争が大変だが，日本の計画を米国がまねた珍しいケースであり，競争の甲斐もある． どちらも水や有機物を豊富に含んだC型小惑星から試料を持ち帰る計画である． 可視分光カメラは小惑星上の物質分布や地形を調べ，どこから試料を採るか決めるための重要な情報を得る． こちらの研究は，大型プロジェクトの動向に左右されるリスクもあるが，宇宙を実感できるメリットがある．
Extreme Matter Module
[Nucl. Fusion Res. Edu. Program]See Details
Nuclear fusion is a power source of active stars and future commercial power plants. Fusion reaction occurs in plasmas (hot ionized gases) under appropriate temperature, density, and confinement conditions. To develop smaller, more economical thermo-nuclear fusion reactors, we are making researches to achieve a magnetic configuration to sustain higher plasma pressure for a given magnetic field strength (i,e, high-beta plasma confinement). Since a spherical tokamak (ST) with a much tighter ring shape than a conventional tokamak provides both a high beta limit and good confinement, it is one of the promising candidates for an economical fusion reactor core in the future. We are developing a novel formation method of a high-beta ST by using a plasma merging scheme which utilizes magnetic reconnection as an effective conversion process from magnetic to plasma thermal energy. ST merging devices such as UTST and TS-4 are employed to conduct the experimental studies.
[Nucl. Fusion Res. Edu. Program]See Details
In the plasma, which is a collection of charged particles, a variety of collective phenomena mediated by electromagnetic fields occur. In our laboratory, research on magnetohydrodynamic (MHD) instability, heating and current drive by high frequency waves, and energy and particle transport processes due to plasma turbulence, as well as development of high temperature plasma diagnostics based on various physical processes are conducted using the TST-2 spherical tokamak (photo) at the University of Tokyo. We foster world-class scientists through domestic and international collaborations on world-leading spherical tokamak experiments including MAST-U (UK) and NSTX-U (US). Furthermore, we provide scientific leadership on the JT-60SA Project, the flagship tokamak being constructed in Japan in collaboration with EU, and on the ITER Project, the international 'burning plasma' experiment being constructed in France.
[Nucl. Fusion Res. Edu. Program]See Details
Plasma consists of charged particles, which generate electric and magnetic fields, and they interact via electromagnetic forces. In high temperature plasmas, however, low collisionality and low diffusivity prevent an equilibrium. From these two features, high temperature plasmas often become a nonlinear and non equilibrium system. One of the approaches for these issues is fluctuation measurements. In our laboratory, we are developing advanced diagnostics and analysis algorithms for fluctuation measurements, as well as conventional plasma physics. Our main target plasma is the TST-2 spherical tokamak plasma in our laboratory. In addition, we study also LHD plasma in Gifu, QUEST plasma in Fukuoka, LATE plasma in Kyoto, MAST plasma in UK as collaboration research.
[Nucl. Fusion Res. Edu. Program]See Details
Creating an ideal energy source
Massively parallel computing reveals the physics of fusion plasmas
Fusion energy has the potential to generate base-load electricity in a clean and safe way. Although a tokamak based fusion reactor is now expected to be able to create net energy, there are still many topics which need further research to realize a commercial reactor, such as improvement of plasma confinement and establishment of steady-state operation scheme. In our laboratory, with Professor Takase and Associate Professor Ejiri, we study basic plasma physics and develop technologies which may lead to faster realization of fusion. On the TST-2 spherical tokamak at the University of Tokyo, we perform experiments with a tokamak plasma driven by RF waves which can be operated in steady-state. We also collaborate with world-leading fusion research groups at institutes such as NIFS and MIT. Simulation of fusion plasmas often requires simultaneous treatment of a wide range of spatial and temporal scales, and description of the phase-space dynamics, which, in turn, requires large computational resources. Recent development of massively parallel computing technology has allowed us to describe fusion plasmas with many interacting physics, with accuracy sufficient to make quantitative predictions. On TST-2, microwave interferometer and hard X-ray radiation measurement is developed and installed to validate the predictions of the simulations. By comparing the experimental measurements with the numerical simulation performed at the supercomputer on NIFS, we are slowly gaining better understanding of wave propagation and damping prcoesses, and interaction between current drive and magneto-hydrodynamic equilibrium.
化学反応を実現する際には，多くの場合に触媒が不可欠となる． 特に，固体触媒は，生成物からの分離が容易で，再利用にも有利なことから実用的な意味が高い． 不活性分子の有用物質への変換を目指して，ナノ金属酸化物結晶，金属錯体をベースにした固体触媒，メソポーラス金属酸化物の開発を行っている．
イオン液体は有機物であり，かつイオン対から構成される塩であることから物性のデザインが可能な溶媒として注目を受けている． 我々は，イオン液体分子を固体表面に固定化して（下図参照）固体触媒として有用であることを示している． イオン液体は二酸化炭素との親和性が高いことからこの性質を利用した反応を開発している．
電子線や光を入射することにより固体表面上の電子状態を励起することで，化学結合の切断や組み替えが起こる． これらのダイナミクスは光触媒作用とも直接かかわる重要なプロセスである．パルスの電子線，レーザーパルスを入射した後のイオン発生， 発光現象を時間分解測定するための装置開発，およびそれらを用いたダイナミクスの研究を行っている． また誘電体バリア放電による大気圧近傍のプラズマを利用した化学反応過程の研究を行っている．
計算化学的手法は現在非常に重要かつ有用なツールとなっている． 我々は，1）モンテカルロ法による固体表面上の吸着種の振る舞いと化学反応の記述，2）遺伝的アルゴリズムを取り入れたテンソルLEED法による複雑な固体表面構造の解析， 3）密度汎関数法による触媒の活性構造と反応過程の解明に取り組んでいる．
Strongly Correlated Electron System, Superconductivity,
Photoemission SpectroscopySee Details
Angle-resolved photoemission spectroscopy is a very powerful experimental technique that can directly observe a dispersion relation between momentum and energy of the electrons in solid-state materials, whereas by utilizing a femtosecond laser as pumping light and its high harmonic generation as probing light, we can observe ultrafast transient properties of the band structures in a non-equilibrium state. In our group, we are developing and improving a time-resolved photoemission apparatus that utilize high harmonic generations of an ultrashort-pulse laser in collaboration with a laser-developing group. We are aiming for understanding the mechanisms of electron relaxations from photo-excited states and mechanisms of photo-induced phase transitions by direct observations of transient electronic states with a pump-probe type time-resolved photoemission spectroscopy. Also, we are aiming for understanding the mechanisms of unconventional superconductivity by direct observations of the electronic structures and superconducting-gap structures of unconventional superconductors with a laser-based angle-resolved photoemission apparatus with a world-record performance that achieves a maximum energy resolution of 70 micro eV and lowest cooling temperature of 1 K.
Plasma consists of a huge number of charged particle. The various behaviors of plasma brought about by the interaction between charged particles and the electromagnetic field are complex nonlinear problems. In order to clarify and predict the nonlinear phenomena of plasmas, we promote large-scale computer simulations using supercomputers.
Our primary research subject is the interaction between fast ions and magnetohydrodynamic (MHD) waves in magnetically confined fusion plasmas. This is an especially important subject for ITER, which is under construction in France with the international collaboration.
In addition, we conduct simulation studies of energetic particles and MHD phenomena, and develop computational physics models and high-performance computing methods. We also actively promote both domestic and international collaborations for many experiment devices such as the Large Helical Device (National Institute for Fusion Science) and the TST-2 Spherical Tokamak (University of Tokyo).
Figure caption: Fluid velocity fluctuation profiles of MHD waves in the Large Helical Device (left) and ITER (middle), and snapshot of laboratory (right).
By means of first-principles calculation, we study non-trivial electronic properties of correlated/topological materials. We also aim at predicting intriguing phenomena originating from many-body correlations and designing novel functional materials/systems. The long-term goal of our research is to establish new guiding principles for materials design. We are also interested in the development of new methods for electronic structure calculation. Our recent research projects include
- Development of ab initio downfolding methods
- Superconductivity in iron-based superconductors, cuprates, fullerides, hydrides under high pressures
- Development of density functional theory for superconductors
- Interplay between the spin-orbit coupling and electron correlations in 5d electron systems
- Exotic electronic structure of skyrmion systems, Weyl semimetals, toporogical insulators
- Giant thermopower in transition-metal compounds
- Multi-pole physics in heavy fermion compounds