Tohoku University Repository (TOUR) / 東北大学機関リポジトリ

Tohoku University博士（工学）要約のみthesi

CHIKHALIKAR, AKASH,KEDAR

論文内容要約 
 In this dissertation we present a new strategy to personalize object-landmark-room relationships using 
semantic maps for planning for object search tasks.  
Object search in an indoor space constitutes the task of planning, navigation and perception to look for 
movable objects (e.g., cup, remote) found in daily life. The ability to search for objects is a fundamental 
ability for service robots. It is a precursor to execute complex tasks like `Prepare Dinner Table' or `Tidy My 
Room'. The tasks of navigation and perception are more generalized and existing methods perform 
well-enough to be deployed for our scenario. The higher level task of planning is the main focus of this thesis, 
and is dependent on three aspects. The first is the discretization of the search space, i.e., the indoor 
environment. 
To this end, in Chapter 2 of this thesis, we present a novel framework that enables the robot to prepare a 
semantic map of the environment. A semantic map is an effective tool for discretization of search space. Our 
semantic map consists of three layers; metric layer, landmark layer and room layer. The RTAB-Map 
framework is used to track the camera poses and obtain the metric layer. To obtain the landmark layer, we 
use the YOLOv8 network for detection of landmarks. These detections need to be tracked as the robot 
navigation proceeds. The subsequent data-association problem is solved using the Hungarian algorithm for 
obtaining the best association. After assignment, we use Kalman filter to localize the landmarks. To obtain 
the room layer, we first geometrically segment the metric layer based on Voronoi region. The room labels for 
each segment are obtained based on the statistical data in the Places365 dataset. 
  アカシュ ケダル チカリカル 
氏 名   Akash Kedar CHIKHALIKAR 
研究科，専攻の名称 東北大学大学院工学研究科（博士課程）ロボティクス専攻 
学 位 論 文 題 目          Task-Driven Planning and Object Search by a Mobile Robot in Indoor 
Environments Using Semantic Maps and Personalized Ontology
(セマンティックマップと個人適用オントロジーを用いた屋内環境移動ロボットによるタスク駆動
型計画と物体探索) 
論 文 審 査 委 員          主査 東北大学教授 平田 泰久 東北大学教授 田所 諭 
  東北大学教授 吉田 和哉 東北大学助教 ホセ サラザル 
 
The second aspect is a probabilistic prior about the locations within the environment that conveys where the 
object is likely to be found. In the past, a common strategy is to discretize the search space based on the 
semantic maps and then populate the map with a prior according to the object. This prior is based on 
common-sense derived from statistical data. The statistical data cannot effectively capture 
environment-specific conditions such as the relative locations of landmarks within the environment or 
underlying temporal patterns of an individual, more generally referred to as behaviors. The ability to adapt 
and personalize the priors based on environment-specific conditions is a key aspect missing in previous 
approaches. 
 
In Chapter 3, we present a novel framework with the aim to personalize common-sense knowledge based on 
spatial and temporal environment-specific conditions. The common-sense knowledge is modelled as an 
object-landmark-room ontology. We consider two types of ontology between objects and landmarks which are 
relevant for object search: `object-isNear-landmark' and `object-isOn-landmark'. An initial generalized 
estimate of the ontology is obtained by counting the co-occurrence frequency of objects and landmarks in 
images of public datasets. This estimate is conditioned on the room category of the images as well. However, 
the environment-specific conditions will vary from statistical data. The spatial topology, i.e. the relative 
position of landmarks, of the environment is used to tweak this estimate. Along with this, the individuals 
present in the environment may have preferences for the likely locations of these objects. Thus, a 
personalization of the ontologies is necessary. We use episode-based learning approach for personalization of 
object-factored ontologies. This learning is coupled with a confidence-based termination criteria for accurate 
personalization. 
 
The third aspect is a strategy for search guidance that leverages the probabilistic priors. Usually, an iterative 
process is followed wherein the next best location to search for the object is determined and if the object is 
not found, the process is repeated. Heuristics that consider trade-off between distances and probabilities 
have been proposed for the same. The heuristics proposed either consist of a greedy horizon or a long-horizon 
that is only useful in worst-case scenarios. The previous approaches do not adapt the search horizon based 
on the probabilistic prior. 
 
In Chapter 4, we propose a novel inferencing strategy that uses an adaptive horizon to efficiently locate 
objects. Simply put, if the prior is highly skewed towards a particular landmark the robot visits that 
landmark. If the priors are uniformly distributed between the landmarks, the robot plans an optimal route 
purely based on distance. The robot still uses a heuristic to calculate the utility of any location, but the 
number of locations to consider for route planning adapts to the prior distribution. The optimal route is 
determined by transposing object search as a travelling salesperson (TSP) problem and using existing 
TSP-solvers to find a solution. In this chapter, we validate our strategy against multiple state-of-the-art 
(SOTA) approaches to quantify the improvements due to our approach. We also conduct experiments to 
prove that ontology personalization can act as a catalyst to improve the performance of SOTA methods. 
 
Finally, in Chapter 5, we propose a novel framework to extend the object search into the Open-vocabulary 
domain. Specifically, we provide the robot the capability to search for any object when instructed in natural 
language. Since it is not possible to store priors for all objects that may be present in an environment, our 
framework is capable of generating task-driven priors. These are obtained by a weighted recombination of 
personalized object-factored ontologies for a fixed set of dictionary objects. The initial step in this 
recombination is categorizing the target object based on its degree of membership to functional primitives 
(e.g. work, entertainment, edibles etc.). This categorization is obtained by actively querying a Large 
Language Model (LLM) at the start of the task. Then the degree of similarity between the target object and 
dictionary object is obtained. Along with this, the degree of membership of each room to the functional 
primitives is also obtained. This enables us to obtain a task-driven prior for an open-set of objects. 
 


English

Task-Driven Planning and Object Search by a Mobile Robot in Indoor Environments Using Semantic Maps and Personalized Ontology

干川, 大和

Japanese

先進複合材料の材料非線形性を伴う破壊現象に関する数値解析的研究

清野, 智大

核融合炉における電子サイクロトロン波を用いた高効率プラズマ電流駆動と電流分布制御性に関する研究

MUTAI, KEVIN,KIPRUTO

 
論文内容要約 
  Imaging at microwave and millimeter wave frequencies continues to draw interest for various 
applications. This is because at these frequencies, the energy carried by the electromagnetic wave is 
sufficient to penetrate thin barriers to allow noninvasive detection and imaging and yet is too low to cause 
ionization and is therefore perceived to be relatively safe compared to other technologies such as X-ray. 
Additionally, as the skin depth is small, the scattered field from both conductors and non-conductors can be 
measured to enable the detection and subsequent imaging of both kinds of objects. Several types of imaging 
systems have been developed to exploit these physical phenomena however a common characteristic with 
existing approaches is their requirement for an array of sensors which leads to increased fabrication costs 
and non-portability. To address these shortcomings, an imaging system using leaky-wave antennas (LWAs) 
is envisioned where the array of sensors can be reduced to just two antennas in which the frequency 
scanning capability of the LWAs is exploited to instantaneously scan along one spatial dimension. 
    Chapter 1 is the introduction providing the background to this study. More specifically, the challenges 
facing the implementation of the proposed imaging system are discussed and the research objectives 
identified following which the outline of this dissertation is clarified. 
    In Chapter 2, analytical equations used to study the LWAs in the rest of the dissertation are introduced. 
The main motivation behind the derivation of these equations is to dramatically reduce the design and 
analysis time of the LWA structures. The derived closed form equations are also independent of the physical 
structure of the LWA and as such can be used to study any type of one-dimensional LWA structure. The 
accuracy of the equations is verified by comparison with full-wave simulations. 
           むたいけヴぃんきぷると 
氏 名           MUTAI KEVIN KIPRUTO 
研究科，専攻の名称 東北大学大学院工学研究科（博士課程）通信工学専攻 
学 位 論 文 題 目          Study on Leaky-Wave Antennas for Imaging Applications 
          (イメージング用漏れ波アンテナに関する研究) 
論 文 審 査 委 員          主査 東北大学教授 陳 強 東北大学教授 末松 憲治 
             東北大学教授 松浦 祐司 東北大学准教授 今野 佳祐 
 
    Low-profile, lightweight, inexpensive and relatively simple to fabricate leaky-wave antennas (LWAs) 
that are capable of near field focusing (NFF) are required to realize the proposed imaging scheme. To 
address these requirements, metallized plastic LWAs are introduced in Chapter 3. Based on the equations 
studied in Chapter 2, a simple design method generally applicable to rectangular waveguide based NFF 
LWAs is introduced. The proposed design method is then used to develop an acrylic NFF LWA with a 
tapered broad wall height and a 3D printed LWA with a tapered dielectric constant distribution physically 
achieved by using a perforation profile along the length of the waveguide. The fabricated acrylic and 3D 
printed prototypes are metallized by copper plating and by adhesive copper tape respectively resulting in 
light and portable structures. The accuracy in the positions of the focusing spots of these structures is 
verified theoretically and experimentally. 
    A major limitation of purely periodic or purely uniform NFF LWAs is their relatively limited scanning 
range and design flexibility. To tackle the problem of the limited scanning range, a structure composed of two 
waveguides sharing a single feed capable of supporting travelling waves with a positive and negative phase 
constant is proposed in Chapter 4. The forward and backward travelling waves supported by the combined 
antenna structure allow the illumination of the forward and backward quadrants simultaneously leading to 
a dramatically improved scanning range. 
    To increase the design flexibility of NFF LWA structures, an approach to decouple the phase tuning and 
feed (the wave guiding structure supporting the travelling wave) mechanisms by moving the phase tapering 
to a gradient index (GRIN) lens superstrate is proposed. The GRIN lens responsible for NFF is then fed by a 
mechanically untapered LWA resulting in a structure with an extra degree of freedom in the design process 
as the lens and LWA can be separately optimized. 
    Imaging measurements to confirm the practical applicability of the proposed structures to the target 
application scenario are also included in Chapter 4. In the first measurement, the NFF LWAs from Chapter 
3 are used to generate the image of a cylindrical object (with an angle independent scattering pattern in two 
dimensiona) positioned in front of a human body phantom to simulate the real application environment. 
Due to the inclined nature of the radiated field from these antennas illuminating only the forward direction 
and using only the magnitude information to generate the image, details on the shape or position of planar 
scattering objects are lost. To resolve this problem, a holographic reconstruction approach is introduced in 
Chapter 4 where the geometric shape of a planar scattering object could be recovered using the leaky-wave 
fed lens antennas.  
    As mechanically tapered LWAs are not capable of dynamically controlling the radiation direction owing 
to their fixed physical parameters, an electronically tunable guided-wave fed linear array antenna is 
introduced in Chapter 5 to realize dynamic control of the radiation direction at a single frequency. By virtue 
of using the phase constant of the travelling wave at each antenna element to accomplish the stated 
objective, the structure can also be considered as an LWA. The radiating element varies between two states 
(ON or '1' and OFF or '0')  determined by a PIN diode soldered across the slot-dipole element making up the 
array.  By using a substrate integrated waveguide (SIW) as the platform for the proposed concept, the 
resulting antenna structure is of low profile and has an easily integrable control circuit. Additionally, as 
scanning along the longitudinal direction is no longer dependent on frequency sweep, the stopband 
encountered close to broadside by traditional LWAs can be either completely avoided or significantly 
mitigated. 
    Chapter 6 is the conclusion. 
 


Study on Leaky-Wave Antennas for Imaging Applications

黄, 緒平

Reversible Audio Information Hiding Based on Integer Discrete Cosine Transform for Tampering Detection

QI, YUE

 
 
論文内容要約 
Rechargeable batteries play an important role in mobile devices and electric vehicles. In the current battery 
market, Li-ion batteries (LIBs) have been widely used owing to the high energy density and excellent cyclability. 
However, commercial LIBs have almost reached the theoretical limits in energy density and the scarce element 
sources cause significant concern on the supple risk. As alternative battery technologies, multivalent cation 
batteries (e.g. Mg2+, Ca2+, Zn2+) are expected to be the candidates for post Li-ion batteries. These multivalent 
elements have large source abundance, which is promising for reducing the production costs. Besides, multivalent 
cations usually show low tendency to dendritic growth in electrodeposition, which is a necessary feature for the 
safe use of high-capacity metal anodes. Especially, the rechargeable magnesium batteries (RMBs) have been a 
promising candidate for next-generation batteries combining high energy density and high safety. However, 
developing suitable cathode materials for RMBs encounters many difficulties. As one of the major challenges, the 
strong coulomb interaction between Mg2+ and the host structures usually cause significantly sluggish solid-state 
diffusion and irreversible phase transformation during intercalation reactions. 
For the rechargeable magnesium batteries (RMBs), many efforts have been made to develop 
intercalation-type cathode materials. In the early stage, reversible Mg2+ insertion and extraction has been revealed 
in transition metal sulfides, such as Chevrel phase Mo6S8. In the Mo6S8 phase, an octahedral Mo cluster is 
contained in a cubic S cluster as a unit block stacking in the crystal, resulting in a robust host structure and 
moderate coulomb interaction for the Mg2+ intercalation. The Mo6S8 cathode usually exhibits a reversible capacity 
of ~80 mAh g-1 and an operating voltage of ~1.2 V vs. Mg at room temperature. To improve the energy density of 
RMBs, transition metal oxides have been expected to be used as cathode materials owing to the higher theoretical 
capacity and operating potential. There are several promising candidates that have been reported in recent years. 
For example, spinel oxides, such as MgCo2O4, ZnMnO3, have been revealed for reversible Mg2+ intercalation at an 
elevated temperature around 150 ℃. The elevated temperature markedly accelerate the diffusion kinetics while 
limiting the use of the RMBs and leading to a rapid structural degradation in some cases. Therefore, it requires 
effective strategies to improve the reaction kinetics of oxide cathodes at room temperature.  
 As a successful attempt to facilitate the topotactic Mg2+ intercalation, we have proposed Li-Mg 
dual-cation/dual-salt systems from 2013, where the Li+ ions are combined with the Mg2+ ions (e.g. Mg2+) as 
carrier ions. Since the Mg2+/Mg has a higher redox potential than Li+/Li, the formation of Li dendrites can be 
circumvented on the metal anode. For the cathode reaction, we have discovered a concomitant intercalation 
behavior of preceding Li+ ions and following Mg2+ ions in the Chevrel phase Mo6S8 cathode. The concomitant 
intercalation of the dual ions significantly reduces the activation energy in the solid-state diffusion, which 
consequently decreases the overpotential and increases the reversible capacity. Similar phenomena were also 
reported in oxide cathode materials, such as anatase TiO2. The concerted effects observed in the dual-cation 
systems show intriguing features for constructing rechargeable batteries with multivalent ions. However, whether 
concerted effects can be applied to high-potential, high-energy-density oxides, the conditions that trigger these 
           さい えつ 
氏 名          斉 悦 
研究科，専攻の名称 東北大学大学院工学研究科（博士課程）金属フロンティア専攻 
学 位 論 文 題 目          Study on the Concomitant Intercalation of Alkali and Alkaline Earth 
Ions in Hollandite-Type Manganese Dioxide（ホランダイト型二酸化マンガンにおけるアル
カリ金属イオンとアルカリ土類金属イオンの共挿入挙動に関する研究） 
論 文 審 査 委 員          主査 東北大学教授 市坪 哲  東北大学教授 高村 仁  
             東北大学教授 加藤 秀実 東北大学教授 熊谷 悠  
 
 
effects, and the mechanisms behind them are still not fully understood. 
First, to investigate whether concerted effects could be applied to high-potential, high-energy-density 
oxides, hollandite-type α-MnO2 (I4/m) was employed as the cathode material, and the intercalation of Mg²⁺ 
alongside the alkali cations Li⁺, Na⁺, and K⁺ was examined. In the constant current discharge experiments, the 
alkali cation additives markedly improved the discharge capacity. The composition analysis results indicated that 
Mg²⁺ remained the primary carrier ion. Further GITT measurements showed that Mg²⁺ intercalation was preferred 
by the α-MnO2 cathode, while the concomitant intercalation of alkali cations reduced the overpotential during 
discharge. Owing to the concerted effects in the dual-cation electrolytes, the cycle capability of the α-MnO2 
cathode was improved in Mg batteries, although the insufficient delithiation during charge still limited 
performance and needed to be addressed in future studies. 
Based on the results with the α-MnO2 cathode, the effectiveness of alkali cation additives in rechargeable 
magnesium batteries required the following conditions. First, Mg²⁺ needed to be inserted into the cathode 
materials prior to the alkali cations from an energetic perspective, which was essential for a “rocking-chair” 
battery structure utilizing a high-capacity Mg metal anode. Second, Mg²⁺ and alkali cations needed to coexist in a 
single phase for topotactic intercalation without phase separation. These findings contributed to enhancing the 
performance of rechargeable batteries that utilized multivalent carrier ions. 
 Second, the mechanism behind the concerted effect was explored by observing α-MnO2 to identify the 
positions of inserted Mg²⁺ ions and assess structural degradation during cycling. The intercalated Mg²⁺ ions 
randomly occupied both center and off-center sites within the diffusion channels, leading to isotropic expansion of 
the tetrahedral unit cell, which contradicted the shear distortion predicted by computational approaches. Upon 
cycling, most Mg²⁺ ions were reversibly extracted from the host structure, leaving trace amounts at the center sites. 
Structural degradation of the host structure was primarily attributed to Mn dissolution rather than a conversion 
reaction driven by Mg²⁺ intercalation. 
Additionally, the impact of the concomitant intercalation of Na⁺ and Mg²⁺ dual cations was examined. The 
presence of Na⁺ promoted the alignment of Mg²⁺ at center sites, reducing random occupation at off-center sites, 
which contributed to improved diffusion kinetics and cyclability. The addition of alkali ions may have weakened 
the electrostatic attraction between Mg²⁺ and the surrounding ions or lattice atoms. This reduction in electrostatic 
interaction lowered the energy barrier for ion insertion, making it more favorable for Mg²⁺ to preferentially insert 
into the central position. Therefore, it is plausible that the presence of alkali ions promoted the insertion of Mg²⁺ 
into the central site, enhancing overall stability and ion diffusion dynamics.  
Third, the potential for concerted effects with alkaline earth metal ions other than Mg²⁺, such as Ca²⁺, was 
examined to further explore the applicability of this phenomenon in multivalent ion batteries. Similar to Mg²⁺ ions, 
the slow diffusion of divalent calcium ions (Ca²⁺) within the intercalated host remains a major challenge in the 
development of calcium-ion batteries (CIBs), significantly limiting their rate capability and overall 
electrochemical performance. To determine whether alkali metal ions can facilitate the diffusion of Ca²⁺, Na⁺ was 
introduced to create a dual-cation system. Electrochemical measurements revealed that the presence of Na⁺ led to 
a notable decrease in overpotential and an increase in discharge capacity, suggesting that concerted effects are 
applicable to alkali-Ca cation systems and can potentially enhance the practicality of CIBs. 
To gain deeper insight into the role of Na⁺ in Ca²⁺ intercalation, the occupation sites of Ca²⁺ before and after 
the addition of Na⁺ were investigated using structural characterization techniques. In the absence of Na⁺, Ca²⁺ 
primarily resided at the surface of the cathode material without significant diffusion into the bulk structure, likely 
due to strong electrostatic interactions and the high energy barrier for insertion. However, in the presence of Na⁺, 
Ca²⁺ was observed to preferentially insert into the central channels of the material, forming a more stable and 
diffusion-favorable configuration. This behavior contrasts with that of the single-ion system, where Ca²⁺ remained 
localized at the surface and could not diffuse further into the host lattice. These findings indicate that the addition 
 
 
of alkali ions effectively reduces the energy barrier for Ca²⁺ migration, thereby promoting its diffusion and 
improving overall electrochemical performance. The observed enhancement in Ca²⁺ diffusion through the 
introduction of Na⁺ underscores the broader applicability of concerted effects in multivalent ion batteries. By 
leveraging alkali metal co-intercalation, strategies can be developed to mitigate the intrinsic limitations associated 
with divalent carrier ions. 
  Based on the findings of this study, the introduction of alkali ions (Li⁺, Na⁺, and K⁺) into magnesium 
batteries to construct dual-cation electrolytes was confirmed to significantly reduce discharge overpotential and 
increase the discharge capacity of a hollandite-type α-MnO2 cathode. Additionally, the impact of the concomitant 
intercalation of Na⁺ and Mg²⁺ dual cations was elucidated. The presence of Na⁺ promoted the alignment of Mg²⁺ at 
center sites, reducing random occupation at off-center sites, which contributed to improved diffusion kinetics and 
enhanced cyclability. Furthermore, the potential for concerted effects with alkaline earth metal ions beyond 
Mg²⁺, such as Ca²⁺, was examined to explore the broader applicability of this phenomenon. The introduction of 
Na⁺ facilitated the intercalation of Ca²⁺ into the central channels of the host structure, overcoming the 
surface-trapping limitation observed in the single-ion system. This enhancement in Ca²⁺ diffusion highlights the 
versatility of the concerted effect in improving the electrochemical performance of multivalent ion batteries. 
Although the underlying mechanisms governing the concerted effects and the long-term stability of dual-cation 
systems are not yet fully understood, these findings provide critical insights into the design of high-energy-density 
multivalent batteries. By leveraging alkali metal co-intercalation, new strategies can be developed to mitigate the 
inherent challenges associated with multivalent ion transport, paving the way for next-generation rechargeable 
battery technologies. 
 


Study on the Concomitant Intercalation of Alkali and Alkaline Earth Ions in Hollandite-Type Manganese Dioxide

OBAYASHI, Takeshi

FUJII, Toyonobu

KITAMURA, Akane

KUMANO, Gaku

IKEDA, Minoru

 43
Poster Session
*PG-7
PlanDyO: Meta-Epigenomic Monitoring of Marine Plankton 
Dynamics in Onagawa and Mutsu Bay
Takeshi OBAYASHI1,2, Toyonobu FUJII1,3, Akane KITAMURA1, Gaku KUMANO1,4 and Minoru IKEDA1,3
1Tohoku University & JAMSTEC Advanced Institute for Marine Ecosystem Change (WPI-AIMEC), Miyagi, 980-8578, Japan 
2Graduate School of Information Sciences, Tohoku University, Miyagi, 980-8579, Japan 
3Onagawa Field Center, Graduate School of Agricultural Science, Tohoku University, Miyagi, 986-2248, Japan 
4Asamushi Research Center for Marine Biology, Graduate School of Life Sciences, Tohoku University, Aomori, 039-3501, Japan
Marine	ecosystems,	characterized	by	complex	physicochemical	properties	and	diverse	biomes,	present	significant	challenges	for	
ecological	research.	Coastal	areas,	in	particular,	exhibit	heightened	complexity	as	they	are	influenced	by	both	local	and	open	ocean	
environments,	as	well	as	human	activities	such	as	fisheries,	aquaculture,	and	coastal	development.	The	Sanriku	Coast,	 located	in	
northeast	Japan,	is	a	dynamic	region	where	the	Kuroshio	and	Oyashio	currents	converge,	creating	a	highly	productive	yet	sensitive	
marine	ecosystem	that	is	particularly	vulnerable	to	global	environmental	changes,	including	climate	change	and	ocean	acidification.
In	Onagawa	Bay,	the	southernmost	part	of	the	Sanriku	Coast,	we	are	conducting	monthly	sampling	to	perform	physicochemical	
measurements	and	genomic	analyses.	These	efforts	aim	to	provide	a	comprehensive	understanding	of	ecosystem	dynamics	by	
quantifying	the	abundance,	condition,	and	function	of	all	individual	planktonic	species	in	the	bay.	This	approach	allows	us	to	monitor	
seasonal	and	interannual	variations	in	biodiversity,	nutrient	cycling,	and	ecosystem	health.
To	address	 these	challenges,	we	propose	a	novel	marine	metagenomic	platform	leveraging	Nanopore	sequencing	 technology	
to	analyze	the	epigenomics	of	marine	plankton.	This	platform	focuses	on	two	key	objectives:	 (A)	generating	reference	genomes	
by	reconstructing	genome	sequences	from	metagenomic	data	(MAGs),	particularly	for	eukaryotic	species	with	 larger	genomes,	
and	(B)	characterizing	genes	and	genomes	through	epigenetic	profiles,	abundance	profiles,	and	Single	Nucleotide	Polymorphism	
(SNP)	profiles.	By	integrating	metagenomic	and	epigenomic	data,	we	aim	to	uncover	the	molecular	mechanisms	driving	ecosystem	
responses	to	environmental	changes.	We	present	the	current	status	of	this	platform’s	development,	which	is	expected	to	be	publicly	
available	in	2025,	providing	valuable	resources	for	marine	ecological	research	and	conservation	efforts.


【Poster Presentation】PlanDyO: Meta-Epigenomic Monitoring of Marine Plankton Dynamics in Onagawa and Mutsu Bay

LI, ZIHE

氏  名（本籍地） 李
り
 子
し
荷
か
 （中国） 
 
学 位 の 種 類 博 士 （経済学） 
 
学 位 記 番 号 経博 第 220 号 
 
学位授与年月日 令和 7 年 3 月 25 日 
 
学位授与の要件 学位規則第 4 条第 1 項該当 
 
研 究 科 、 専 攻 東北大学大学院経済学研究科（博士課程後期 3 年の課程） 
 経済経営学専攻 
 
学 位 論 文 題 目 Currency Internationalization and Investor Behavior in International 
Financial Markets 
(国際金融市場における通貨の国際化と投資家行動） 
 
博士論文審査委員 （主査） 
 教 授  永 易  淳 准教授  菅 原  歩 
  講 師  B A I  Y U 
 
論 文 内 容 の 要 旨 
 
The doctoral thesis contains three chapters on currency internationalization and investor behavior 
in international financial markets. All three chapters together constitute a comprehensive analysis of 
the process complexity and dynamics of currency internationalization. These chapters complement 
each other from micro to macro and provide a comprehensive framework of understanding the 
multidimensional characteristics of currency internationalization. 
 
Chapter 1: A Bayesian Causal Impact Analysis of the Chinese Yuan’s Internationalization 
During the 2008 global financial crisis, US dollar-dominated global monetary system has been 
suffering significant shocks. Investors and researchers gradually realize that the US dollar-dominated 
monetary system may not be the most stable system, and eager to find one or several new currencies 
to take the international currency role together with US dollar, or directly replace it. Meanwhile, as 
Chinese economy grows rapidly and China’s international status has been rising, people increasingly 
concern about Chinese Yuan (RMB), and expect it to become a new international currency in the 
future. A landmark event is that in 2016, International Monetary Fund has included RMB into Special 
Drawing Rights (SDR) basket. Therefore, RMB internationalization topic is getting more and more 
popular in recent decades. 
Chapter 1 mainly analyzes the internationalization process of RMB, and quantifies the impact of 
COVID-19 pandemic. In literature review part, we examine previous works related to the current 
progress of RMB internationalization and the impact of COVID-19 pandemic on financial markets, 
including both qualitative and quantitative works. We conclude that there are very few works 
discussing the impact of COVID-19 pandemic on currency internationalization, and therefore, the 
topic is worth researching. In this paper, although exploratory factor analysis is a traditional method, 
we explain it in a relatively new way to extract common factors including information of the most 
developed key currencies. We infer possible latent factors behind internationalization of key 
currencies. Together with rolling correlations between original and estimated exchange rate series, 
we evaluate and visualize the recent internationalization status of RMB and its trend considering the 
correlation among key currencies with different currency combinations. Rolling correlation proxy 
variable measures the level of RMB internationalization, because higher correlation implies a higher 
internationalization degree of RMB. Data used for these empirical analysis is monthly-frequency 
nominal exchange rate data from January 1999 to December 2021. The reason of using exchange rate 
data is that we consider exchange rate can convey money functions including unit of account, medium 
of exchange and store of value better and more directly comparing with other financial variables. 
The above proxy variable is also used to analyze the possible causal relationship between RMB 
internationalization process and COVID-19 pandemic event. A Bayesian-based causal impact 
analysis with Bayesian structural time series is conducted to analyze this possible causality. Using 
naive methods and moving average method, we simulate counterfactual series to represent what might 
happen if there were no COVID-19 outbreak. We examine the differences between reality and 
counterfactual series to evaluate the causal impact. 
We finally conclude that while RMB has not been internationalized, it is internationalizing with a 
fast speed. Besides, COVID-19 pandemic has a short-term negative impact on RMB 
internationalization, only lasting for a couple of months. However, this negative impact does not 
result in institutional changes such as alternation in settlement currencies. 
 
Keywords: RMB internationalization; Factor analysis; COVID-19 pandemic; Bayesian Causal impact 
 
Chapter 2: Searching for Safe Haven Currencies: A Risk Premium View 
Chapter 2 attempts to empirically understand a benchmark to identify safe haven currencies from a 
view of currency risk premium (CRP) for investors’ reference. Considering the increasing risks in 
international financial market and frequently occurred adverse geopolitical events, the paper aims at 
providing scholarly responses and explanations and helping investors make proper investment 
decisions. Potential contributions have two strands. First, we fill in the gap that no papers have 
examined the relationship between CRP and political uncertainties, especially adverse geopolitical 
events. Through different types of regressions, we prove that CRP can be a reliable benchmark for 
determining safe haven currencies for investors, especially when the environment is risky. An 
additional contribution is that based on our empirical results when defining safe haven currencies, we 
examine a nominal interest rate differential as a threshold to evaluate safe haven performances, using 
various empirical analyses to confirm the robustness of our results. 
In this paper, we use two relatively new indices, geopolitical risk index (GPR) and economic policy 
uncertainty index (EPU), to measure political uncertainties, where GPR is the main uncertainty 
variable and EPU is used for robustness check. With the use of these two indices, it fills the gap that 
previously there are no benchmarks quantifying political uncertainties. We consider both global and 
country specific uncertainties in the empirical analysis. Besides, CRP is also clearly defined in our 
work as a compensation that measures the risk aversion degree of investors in the forex market, 
meaning when investors consider their holdings risky, they will require more compensation. All 
empirical results of this work is based on the relationship between CRP and political uncertainty, 
because we believe that CRP can directly reflect investors’ attitudes toward financial markets and 
investment risks when the environment is risky. Proxy 1, based on real interest rate differential, and 
Proxy 2, based on Fama’s premium equation, are used to quantify CRP. Proxy 1 is used as the main 
proxy and Proxy 2 is used as robustness check substitute. All data are monthly data, ranging from 
May 2003 to December 2023. 
Benchmark CRP properly reflects the risk aversion degree of investors and the safe haven 
performances of currencies. The relationship between CRP and political uncertainty can help investor 
to identify which currencies will probably appreciate under risky environments. Using seemingly 
unrelated regression and threshold regression, selecting safe haven currencies with CRP is proved to 
be reliable. We confirm that US dollar and Japanese Yen are currently the most reliable safe haven 
currencies in international financial markets, while Pound Sterling is not safe haven, and there is a 
great possibility that Chinese RMB is not, either. We are not able to confirm whether Canadian dollar 
and Australian dollar are safe havens because their coefficients show instability of significance. 
Through examining this safe haven identification benchmark, we also find that nominal interest rate 
difference, an important sign when examining currency flows between two countries, can influence 
investor expectations and safe haven performances of currencies. 
There are three main conclusions. First, the relationship between CRP and global political 
uncertainties is positive, significant and robust. Results of the relationship between CRP and country 
specific political uncertainties are similar. Second, we confirm that US dollar and Japanese Yen are 
safe haven currencies, while Pound Sterling is not. Chinese RMB has a great possibility of not being a 
safe haven, and the status of Canadian dollar and Australian dollar cannot be ensured. Also, we 
provide a more direct criterion to identify safe haven is to see whether the currency appreciates when 
the degree of political uncertainty is high. In the end, we also conclude that both CRP and nominal 
interest rate differentials can be measurements of safe haven performances. 
 
Keywords: Exchange rate; currency risk premium; risk aversion; geopolitical risks; safe haven 
currency. 
 
Chapter 3: The Role of Global Investor Sentiments in Currency 
Chapter 3 has two research objectives. The first is to quantify currency internationalization 
process to update currency status. The second is to discuss the relationship between currency 
internationalization and global investor sentiments. From these perspectives, we update 
internationalization status of selected currencies including US dollar, Japanese Yen, Chinese Yuan, 
Canadian dollar and Australian dollar. We provide both theoretical and empirical evidence to prove 
that policy makers and governments should take investor sentiments into consideration when making 
currency internationalization related policies. 
Around 2008, a huge global financial crisis occurred and negative sentiments towards financial 
markets spread widely, leading to more attention of the relationship between changes in financial 
markets and investor sentiments. Since sentiments are very subjective and abstract, there have been 
many attempts to quantify them, such as Chicago Board Options Exchange’s Volatility Index (VIX), 
Twitter happiness index and COVID-19 sentiment index. In this paper, we use risk appetite index 
(RAI), a newly constructed index by State Street Global Exchange (SSGE), to measure global investor 
sentiments. According to SSGE, if the value of RAI is closer to 1, it means the sentiment tends to be 
risk-loving, while tends to risk-aversion if the values is closer to -1. A zero RAI indicates 
risk-neutral. This RAI can capture time-varying sentiment flows across various financial markets such 
as currencies, bonds, stocks, assets and even commodities under the influence of 22 risk-taking 
dimension components. We consider RAI can be a better measurement because it covers a wider 
range of aspects of financial markets and world economy. 
One of the main contributions is that we develop a currency internationalization index (CII) to 
quantify and visualize the progress of internationalization of selected currencies with principal 
component analysis (PCA). Although PCA is widely used as a technique to construct such indices, 
the main difference between our index and other indices is that we separate internal factors and 
external factors. We use internal factors, defined as factors come from currencies themselves and 
convey money functions, to construct the CII, while external factors, defined as those factors not 
strictly following money functions, but can indirectly influence currency performances or the forex 
markets, are used as independent variables when examining the relationship between CII and RAI. In 
this chapter, variables including bank credit share, currency share as a global payment method, 
foreign exchange transaction weight, debt security volume weight in international financial market, 
SDR holdings weight and foreign exchange reserve proportion are regarded as internal factors. These 
CIIs can also help us to confirm and rank the internationalization status of selected currencies. 
Another contribution is that we differentiate short-term and possible long-term relationship 
between CII and RAI with relatively advanced techniques, including de-trended cross correlation 
analysis (DCCA) with time lags and autoregressive distributed lag model, both linear and nonlinear 
versions. The main assumption is that the risk preference of investors can influence the process of 
currency internationalization. With DCCA method, we confirm that the short-term relationship 
between CII and RAI are time-lagged, meaning it takes a period of time for the impact of sentiments 
to reflect to currency internationalization process. Results from DCCA indicate that for most 
selected currencies including US dollar, Australian dollar and Japanese Yen, risk-loving attitudes of 
investors can better promote currency internationalization, but there are time lags ranging from 3 to 
5 years for the impact to reflect on currency internationalization. Conversely, results of RMB and 
Canadian dollar show that risk-averse attitude is probably better to promote their 
internationalization. Since we consider results from DCCA method are not completely persuadable, 
we construct ARDL models to (1) provide more evidence of short-term relationship, and (2) examine 
whether there is a long-term relationship. 
We use external factors including GDP growth, foreign direct investment outflows, exchange rate 
volatility and geopolitical risk as control variables in both linear and nonlinear ARDL models. We 
mainly use F-statistic to judge whether there are long-term relationships between investor 
sentiments and currency internationalization. Our results show that F-statistic values are smaller 
than the 5% level of critical values, and therefore, prove that there are no long-term relationships 
between investor sentiments and currency internationalization. 
In summary, we update internationalization status of selected currencies including US dollar, 
Japanese Yen, Chinese RMB, Canadian dollar and Australian dollar. We have confirmed that US 
Dollar and Japanese Yen are still most powerful international currencies, while Canadian Dollar and 
Australian Dollar are not completely internationalized and becoming weaker. Chinese Yuan 
internationalization becomes faster, and tends to surpass Canadian Dollar and Australian Dollar in 
the near future. Using de-trended cross correlation analysis (DCCA) with time lags and 
autoregressive distributed lag model (ARDL), we conclude that there is only short-term relationship 
between currency internationalization and investor sentiments. A risk-loving attitude can promote 
internationalization of US Dollar, Australian Dollar and Japanese Yen, while a risk-averse attitude is 
better for Chinese Yuan and Canadian Dollar internationalization. Our research shows that 
considering investor sentiments is necessary when analyzing currency internationalization and making 
related policies. 
Keywords: currency internationalization; investor sentiments; de-trended cross correlation; short- 
and long-term relationship 
 
論 文 審 査 結 果 の 要 旨 
 
本論文はアメリカドルや日本円など主な為替レートの国際化について分析した実証研究である。
その研究内容は次の三つのテーマに大別することができる。１）高頻度データを用いて各為替レー
トの国際化への進展と基軸通貨を特定する研究、２）Safe Haven 為替レートの特徴から基軸通貨
を特定する研究、そして３）投資家感情(Investor Sentiment)と為替レートの相関から基軸通貨を
探求する研究である。研究１）では、近年目覚ましく発展した因果推論法により、新型コロナショ
ックに対する為替レートの国際化への影響を検証している。基軸通貨はショックに対して高い頑強
性を示すことが研究の背景にある。論文２）は、対象国の経済状況が悪化してもその国の信用度に
より通貨安にならない傾向があることから、非線形時系列モデルを用いて、高リスクに直面しても
通貨高の状態を維持できる通貨を探している。そして研究３）では通貨の国際化レベルを測定する 
インデックスを作成し、このインデックスと投資家感情の関係から基軸通貨を特定している。 
近年高経済成長を経験した中国の通貨も研究の対象としたこと、最新の統計手法を用いるなど、
研究者だけにとどまらず、政府や中央銀行の政策担当者や投資家にも興味深い研究テーマである。
データ分析の結果として、アメリカドルや日本円は各章で定義した基軸通貨の条件を満たす傾向が
あると論じている。他方で、IMF の SDR に組み込まれた中国元については、国際化が進展した証
拠は統計的に見 つけることができるが、基軸通貨と言えるほど国際化されていないと論じている。
また先行研究の焦点でなかったカナダドルやオーストラリアドルも分析しており、これらの通貨に
関しては国際化の鈍化が示されている。最後に、投資家感情は短期的に通貨の国際化に影響してい
るが、その効果は長期的ではないこと論じている。 
仮論文審査で指摘された意見も本論文に反映されており、博士論文としても丁寧に書かれている
ことが分かる。そのため学術論文として査読付き雑誌に掲載されてもよい程の内容に到達している
と判断したため、本論文は博士（経済学）の学位を授与するに十分な知識と能力を有していると判
定した。 


Currency Internationalization and Investor Behavior in International Financial Markets

TRAN, QUANG THANH

氏  名（本籍地） TRAN
ト ラ ン
 QUANG
ク ア ン
 THANH
タ イ ン
 （ベトナム） 
 
学 位 の 種 類 博 士 （経済学） 
 
学 位 記 番 号 経博 第 221 号 
 
学位授与年月日 令和 7 年 3 月 25 日 
 
学位授与の要件 学位規則第 4 条第 1 項該当 
 
研 究 科 、 専 攻 東北大学大学院経済学研究科（博士課程後期 3 年の課程） 
 経済経営学専攻 
 
学 位 論 文 題 目 Three Essays on Fertility and Economic Growth 
（出生率と経済成長に関する3つの論考） 
 
博士論文審査委員 （主査） 
 教 授  北 川  章 臣 准教授  久 保 田  荘 
  名誉教授  秋 田  次 郎 
                                                         （東北大学） 
 
論 文 内 容 の 要 旨 
 
This thesis addresses the issue of underpopulation by developing models that explore the channels 
fertility can affect economic growth in the long run, with an emphasis on the possibility of a 
population decline and severely low fertility rates. The thesis contains five chapters and is organized 
as follows. 
Chapter 1 introduces the context and motivation for this research. 
In Chapter 2, we examine the long-term consequences of population decline in a quality-ladder 
endogenous growth framework. The key factor of fertility decision is the childrearing cost, which is 
modeled as a convex function of labor productivity. As technology grows, the costs of raising children 
(including childbearing, childcare, and educational investments) increase disproportionately. While 
the economy may experience rapid growth in the early stages of development, they are likely to face 
sharp contractions in both population and innovation as child-rearing costs outpace the incentives for 
having children. We show that consistent population decay is almost inevitable. Nevertheless, a 
pronatalist policy is shown to be effective at improving the long-run likelihood of achieving a state of 
high labor productivity and high living standards, especially if implemented early in the development 
process. 
In Chapter 3, we study how higher education incentives affect fertility decisions and influence 
long-term economic growth by examining an R&D-growth overlapping-generations model with 
endogenous education choice. In this model, higher education plays a dual role – it increases earnings 
for skilled labor and technological progress but also discourages childbearing. When the fertility of 
skilled workers is sufficiently low, too many higher education pursuers may lead to a long-run secular 
stagnation where technology and population remain constant or even persistent population decline. 
To avoid these scenarios, regulating access to higher education may be necessary, although it could 
impose welfare loss on some generations following the policy’s introduction. 
In Chapter 4, we pay attention to the dual caregiving duties – child-rearing and elderly care – and 
how they can affect the stability of an economy. In this model, economic agents are assumed to be 
siblings and make decisions on fertility, savings, and division of labor to maximize their collective 
welfare. In the long run, the dual care burden plays a significant role. If it is sufficiently small, the 
economy operates and grows sustainably. If it is sufficiently large, the economy can be locked in a 
“nursing hell” state where all resources can only be put into caring. In this case, the government can 
consider a minimum fertility requirement to solve the problem, although it is not a Pareto 
improvement policy. More importantly, if the burden of care exceeds some feasible threshold, the 
survival of the economy is at risk of a structural collapse where even the aforementioned policy 
becomes ineffective. 
Finally, Chapter 5 concludes the thesis. 
 
論 文 審 査 結 果 の 要 旨 
 
本論文は、出生率と経済成長率がともに内生的に決定される世代重複モデルを構築し、少子高齢
化や人口減少といった現象を理論面から解明することが、その目的となっている。論文は 5 章から成り、
次のように構成されている。 
第 1 章では、研究の背景と動機が紹介されている。 
第２章では、人口減少の長期的帰結が内生成長モデルの枠組みで検討されている。このモデルで
は、出生率決定の主要因は子育てコストであり、これは労働生産性の凸関数として定式化される。この
ため、技術が進歩すると、出産、育児、教育投資を含む子育て費用は急速に増加する。この仮定のも
とでは、経済は発展の初期段階では急速な成長を遂げる一方で、技術が進歩するにつれて子育て費
用が増加すると、人々の出生意欲は急速に低下し、人口とイノベーションの急激な縮小が生じる。この
縮小は不可避であるが、にもかかわらず、出生促進的な政策が発展の初期段階で実施された場合、
長期的に高い労働生産性と生活水準が実現する可能性の高いことが示されている。 
第３章では、高等教育が出生率と長期的な経済成長に与える影響が、教育選択を伴う内生成長モ
デルを用いて検討されている。このモデルでは、高等教育は熟練労働者の所得と技術進歩を増加さ
せる一方で出産を抑制するという二重の役割を果たしており、熟練労働者の出生率が十分に低い場
合、高等教育を受ける人が多すぎると、技術と人口が一定に保たれるか、あるいは、人口が減少し続け
る慢性的な停滞が実現することが示される。このような停滞は高等教育へのアクセスを規制すれば回
避できるが、それは政策導入後の一部の世代に厚生損失をもたらす。 
第４章では、育児と高齢者介護という二重介護に注目し、それらが経済の持続性にどのような影響を
与えるかが考察されている。この章のモデル経済が長期的に成長するか否かは、二重介護負担の程
度によって決まる。すなわち、負担が十分に小さければ、経済は持続的に成長するが、ある水準を超
えると、経済は「介護地獄」状態に陥り、すべての資源を介護に投入するしかなくなる。この「地獄」は政
府が人々に最低出生数を課することで回避できるものの、介護負担がさらにある水準を超えると、経済
の存続が不可能となり、前述の政策も無効になることが示される。 
最後に第 5 章で、本論文全体の結論が述べられている。 
本論文で検討された人口問題は近年人々の関心を集めるものであり、その重要性は極めて高い。
また、本論文の分析は正確であり、そこで得られた結論は示唆に富む。以上の理由により、本論文は
博士（経済学）を授与するに十分な内容を有すると判断した。 


Three Essays on Fertility and Economic Growth

Tohoku University博士（農学）博士学位論文 (Thesis(doctor))thesi

Not a member yet