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

Tohoku University博士（情報科学）thesi

田中, 涼太

English

A Holistic Framework for Visual Document Understanding

Dansu, Margaret Adesewa

Participation in Online Peer Support Groups: A Qualitative Approach to Understanding the Experience of Marginalized Individuals

Xia, Tian

Enhancing Training Stability in Generative Adversarial Networks via Gradient Normalization

Tohoku University博士（工学）要約のみthesi

SION, MARIANNE

Most refrigeration devices are based on gas-compression, such as CFC in the last century, and nowadays HFC. The use 
of compressed gases has led to serious issues such as degradation of the ozone layer, or high global warming potential. Several 
generations of these gases have succeeded each other and have been banned each time due to their toxicity or for being harmful 
to the environment. There is a need to find alternatives to these refrigeration systems, that are less toxic, environmentally 
friendly and that have high performances. Apart from the search for a new refrigerant gas generation, different alternative 
systems could be used such as Peltier or Stirling systems. However, those systems have shown some drawbacks such as high 
heat dissipation or low efficiency for Peltier devices or low power to weight ratio for Stirling devices.  
Another alternative is to use caloric materials that can be used for liquid-solid regenerative cooling systems or for 
solid-solid refrigeration in single stage devices. Caloric materials are known to experience external loading induced 
temperature variations. They are divided into four categories: magnetocaloric, electrocaloric, barocaloric and elastocaloric 
materials. These materials exhibit reversible temperature variations when submitted to magnetic fields, electric fields, pressure 
and uniaxial stress, respectively. These materials were vastly studied, in particular the magnetocaloric materials.  
Elastocaloric materials can be divided in two categories, the Shape Memory Alloys, mainly the Ni-Ti alloys and the 
polymers, like Natural Rubber. Shape Memory alloys have been vastly studied as these materials have high thermal diffusivity 
and a large adiabatic temperature difference. On the contrary, polymers with low thermal diffusivity and small adiabatic 
temperature changes like Natural Rubber have not been studied much to design elastocaloric cooling devices. Nevertheless, 
Natural Rubber is an interesting material for solid-state refrigeration as this is a cheap material with low environmental impact 
and that is easy to load and unload using relatively low force compared to Shape Memory Alloys that requires high forces.  
However, using Natural Rubber, a material which can be considered as an insulating material, in an elastocaloric cooling 
device represents a challenge. The goal of this work is to get a better understanding of how Natural Rubber could become a 
good candidate for solid-state refrigeration in a single stage cooling device. To achieve this goal, it is important to find out how 
such an insulating material can be used for designing an elastocaloric cooling device. In order to get good performances of this 
   シオン・マリアンヌ 
氏名          SION Marianne 
研究科，専攻の名称 東北大学大学院工学研究科（博士課程）機械機能創成専攻 
学位論文題目  
Elastocaloric Heat Pump Proof of Concept for Near Room Temperature Refrigeration Using Natural Rubber 
（室温近傍における天然ゴムを用いた弾性熱量効果の冷却機能概念実証） 
論文審査委員  主査 東北大学教授 小宮 敦樹 東北大学教授 琵琶 哲志
東北大学教授 小川 和洋 東北大学客員教授  Gaël Sebald 
論文内容要約
cooling device, it is necessary to understand which parameters, such as stretching time, contact time, thickness, are the most 
important to obtain the best temperature span and cooling power possible.  
The key scientific questions to be answered in the work are how to get a better understanding of the heat transfer 
mechanisms and determine if the scaling up of the cooling device could reach a large enough temperature span and cooling 
power to be a viable alternative to compressed gases refrigerants.  
To answer these questions, the PhD work focused on the development and performance evaluation of an elastocaloric 
cooling device using Natural Rubber (NR). The purpose of this PhD is to obtain a better understanding of the heat transfer 
mechanisms by developing and testing the performances of a 4-step and a 2.5-step single stage cooling device to determine the 
key parameters to get an efficient heat transfer between the NR and heat exchangers.  
The first chapter of this work presents the context and a brief state of the art on the different caloric effects and some 
devices using those effects. The focus is on the elastocaloric material and devices, as it is the purpose of the thesis, and it is 
important to get the necessary information for the following chapters. The elastocaloric effect is a phenomenon happening 
when an elastocaloric material, such as Natural Rubber and Shape Memory Alloy, is put under a uniaxial stress. When the 
material is loaded, its temperature increases, as the polymers chains inside the material get oriented. Then, when the 
elastocaloric material is unloaded, and goes back to its original length, an endothermic reaction happens due to the 
disorientation of the polymer chains. In the case of Natural Rubber, the sample can be loaded (i.e., can be stretched) at high 
elongations, even up to an elongation of 6.5. The elongation is considered as the ratio between the loaded length and original 
length of the samples, meaning that for an elongation of 6.5, the sample is stretched at 650 % of its original length. Adiabatic 
temperature changes can be observed when the material is put under loading/unloading cycles. When a heat exchanger is put 
into contact with an elastocaloric material while the material is loaded/unloaded, the heat generated or absorbed can be 
transferred. This can be used to design an elastocaloric cooling device for solid-state refrigeration. To finish this chapter, a few 
thermodynamics cycles are explained to understand how they can be applied to caloric materials.  
In Chapter 2, the 4-step single stage cooling device is presented. The system is made of 2 perpendicular actuators and two 
heat exchangers, the Hot Heat Exchanger (HHEX) and Cold Heat Exchanger (CHEX). The NR samples of 0.4 mm and 0.6 
mm were characterized using Infrared measurement to determine the adiabatic temperature changes occurring when the NR 
film is loaded and unloaded. The stress-elongation hysteresis curves showed a maximum stress of 3.5 MPa and 3.9 MPa, for 
the samples of 0.4 and 0.6 mm, respectively. Then, the samples were tested for 16 different experimental conditions (4 
stretching times and 4 contact times). In the last part of this chapter, the performances of the 4-step single stage cooling device 
are studied. For a stretching time (i.e., loading and unloading time) of 0.3 s, the contact time of 3 s was found to be the best 
contact time to enhance the performance of the system. A maximum specific cooling power of 430 W/kg and 540 W/kg were 
found for 0.6 and 0.4 mm, respectively. The temperature span was found to be around 1.9 K and the heat transfer effectiveness 
reached about 60 % for both 0.4 and 0.6 mm. The maximum coefficient of performance was found to be around 15. 
In Chapter 3, a comparison of nine different samples is done to determine which parameters influence the most the 
temperature span, cooling power and heat transfer effectiveness. The composition of the sample was tested using several NR 
films with the same thickness of 1 mm but with different compositions. To test the impact of the thickness, samples of the same 
composition but with three different thicknesses (0.5, 1 and 2 mm) were used to understand the role of the thickness in the heat 
exchange between the NR film and the heat exchangers. It was found that the composition of the samples only has a small 
impact on the performance of the 4-step single stage cooling device. Indeed, for the different compositions, the temperature 
span was found to be on average of 1.5 K and the cooling power about 0.3 W. The heat transfer effectiveness was found to 
vary depending on the thickness, from 20 to 70 %. The coefficient of performance was found to reach as high as 24, in the case 
of a 0.5 mm sample. The thickness of the sample seems to be a key parameter for the enhancement of performance.  
Chapter 4 presents a second single stage cooling device using Natural Rubber, with a different actuation cycle of 
2.5-steps instead of 4-steps. The 2.5-step cooling device is made of only one actuator and two heat exchangers, the HHEX and 
CHEX. The specificity of this device is to load the Natural Rubber while putting it in contact with the heat exchanger at the 
same time. This chapter first presents the 12 experimental conditions tested for three stretching (0.5, 1 and 2 s) and four contact 
times (1, 3, 5 and 7 s). The temperature span, cooling power and coefficient of performance is discussed to understand if the 
performances of the NR film could be enhanced using another actuation principle. It was found that the 2.5-step setup has 
lower performances than the 4-step setup. Indeed, for the 2.5-step setup, the specific cooling was found to be 175 W/kg, and 
the temperature span was found at 1.3 K. The maximum coefficient of performance was found to be about 2. A time contact of 
1 or 3 s was also found to be the contact time maximizing temperature span and cooling power, however, the coefficient of 
performance was maximized for a contact time of 5 or 7 s. 
Chapter 5 presents an analytical model that was developed to study the heat transfer between the NR and heat exchangers. 
The model development based on the heat transfer equations is presented. Then, a validation of the model was done by 
comparison with the experiment. The model was found to be a good fit to the experiment. The model was then used to study 
the impact of the convection, the heat exchanger losses as well as the thermal contact resistance influence. It was found that the 
impact of the convection was small, as increasing the heat transfer coefficient did not lead to a high decrease of cooling power 
or heat transfer effectiveness. Only a decrease of 0.04 W and 20 % was observed for a heat transfer coefficient nine times 
higher. The heat exchangers losses were found to be larger than expected, suggesting the need of improvement of the 
insulation of the experimental setup. Finally, a parametric study of the analytical model was done, using simplified heat transfer 
equation and studying the time constant of the heat exchange between the NR and the heat exchangers. The study of the time 
constant has shown a linear dependence on small thicknesses, however, for higher thicknesses, the dependence was found to be 
quadratic. A critical thickness was determined, and it was found that under this critical value, the interface is the most 
impacting factor while for values higher than this critical thickness, the diffusivity of the material becomes the most important 
factor. 
The final chapter is a general conclusion to remind the principal results of this PhD and gives prospects about how to 
further understand and enhance the performances of the elastocaloric cooling devices using Natural Rubber films. A prospect 
for the enhancement of the performances would be a multistage 4-step cooling device. The system could be made into a 
multistage using several parallel NR sheets and intermediate heat exchangers. The heat would then be transferred from the 
CHEX to the intermediate heat exchangers and finally to the HHEX, through the different Natural Rubber films. A case study, 
for an ideal heat exchange, shows that to get a temperature span of 15 K, 8 sheets of NR would be necessary as each NR film 
can provide a temperature span of about 2 K. However, to improve the cooling power, in order to get 100 W, the initial surface 
of the NR sheet should be 250 higher than the 1600 mm² needed to get 0.4 W. Working with such a large surface might be a 
challenge so using several devices working in parallel might be a better way to enhance the cooling power. 


Elastocaloric Heat Pump Proof of Concept for Near Room Temperature Refrigeration Using Natural Rubber

渡部, 英治

論文内容要約 
 本論文では，カーボンニュートラルの実現に向けて，固相成膜法であるコールドスプレー法を用い，軟磁性粉
末の成形条件検討から，モーター用軟磁性材料の性能向上について検討をおこなった．本論文は以下の6つの章
から成り立っている． 
第1章では，コールドスプレー法によって軟磁性ナノ粒子の粒子径を維持したまま，高充填成形を実現する必
要性について述べた．世界的な気候変動，地球温暖化の抑制のため，二酸化炭素の排出量の削減やエネルギー利
用効率の向上が進められている．自動車産業においては，ハイブリッド車や電気自動車といった電動化車両の普
及率向上が進められており，電動化車両用部品の性能向上が求められている．電動化車両の主要部材の1つであ
るモーターには多くの軟磁性材料が使用されており，更なる性能向上のため優れた加工性を有しながら，高飽和
磁束密度と低保磁力の実現が課題となっている．そこで，付加造形手法としても応用が進められており，かつ材
料の融点以下の温度で高充填成形が可能なコールドスプレー法に注目して，シングルナノサイズの軟磁性ナノ粒
子の高充填成形を検討した．ナノ粒子径を維持したまま，バルク体相当の高充填成形を実現することによって，
粒子間に働く磁気的な相互作用を利用し，保磁力の起源である磁気異方性を平均化させることによって，保磁力
を低下させることが可能であることを述べた． 
 第2章では，軟磁性ナノ粒子のコールドスプレー成形可否検討用材料として，大気中でも比較的安全に取り扱
いができるナノ粒子モデル材料の選定を行い，初期検討を実施した．用いたモデル材は熱プラズマ法で作製した
球状粒子で，平均粒子径は60nmとした．成形にあたり粉塵爆発が起こらないよう，爆発下限界濃度測定試験を
行い，大気中でコールドスプレー成形をおこなっても安全な成形条件を明らかにした．安全な成形条件確認後，
モデル材ナノ粒子をそのままの状態でコールドスプレー成形を試みたが，通常ミクロンサイズ粒子を使用し，高
速ガス気流で基材と衝突させ成形するコールドスプレー法では，凝集力の高い金属ナノ粒子をそのままの状態で
はコールドスプレー成形することが困難であることが明らかとなった．そのため，モデル材ナノ粒子をミクロン
           わたなべ えいじ 
氏 名          渡 部 英 治 
研究科，専攻の名称 東北大学大学院工学研究科（博士課程）機械機能創成専攻 
学 位 論 文 題 目          コールドスプレー法によるモーター特性向上のための軟磁性ナノ粒子高充
填成形法の確立 
論 文 審 査 委 員          主査 東北大学教授 小川 和洋 東北大学教授 水谷 正義 
             東北大学教授 野村 直之 東北大学准教授 小川 智之 
             東北大学准教授 市川 裕士 
サイズまで造粒することによって，均一な搬送を実現し，各種基材上にナノ粒子の均一な成膜が可能であること
を見出した．また，成膜状態や磁気測定への影響を加味して，成形用基材としてアルミニウムが適していること
を明らかとした． 
 第3章では，軟磁性ナノ粒子の高充填化によって保磁力が低下するか検証をおこなうために，更なる高充填成
形を試み，成膜されたナノ粒子膜についての充填率評価，軟磁気特性評価を行った．得られた成形体の断面をSEM
で観察したところ，比較試料として作製した圧縮成形に比べ，コールドスプレー成形体の方が高充填化しており，
更にスプレーガス圧の増加に伴って充填率が向上していることが明らかとなった．さらに，三次元透過型電子顕
微鏡（3D-TEM）を用いた充填率評価でも断面SEMと同様の充填率が得られていた．高充填化要因として，粒
子の塑性変形により空隙が埋まったことが考えられるため，粒子の形状について詳細に分析したところ，原料粉
末や圧縮成形体では球状に近い形状であったのに対して，コールドスプレー成形体はロッド状や円盤状に変形し
ている粒子が確認された．そしてその変形方向はコールドスプレーの圧力負荷方向と垂直方向に変形している傾
向が確認されており，粒子が基材に衝突した際に変形したことが明らかとなった．さらに変形量と粒子径の関係
性について評価したところ、変形量は粒子径の小さな粒子ほど変形が進んでいることが明らかとなった．
3D-TEM上での最大のアスペクト比は2倍であった．また，得られたサンプルの磁気測定評価を実施したところ，
高充填化によって飽和磁束密度の向上や透磁率の向上が確認されたものの，直接交換相互作用による保磁力の低
下は確認できなかった．この要因として，粒子表面に非磁性の酸化膜が存在するため，コアの磁気的な相互作用
エネルギーが減少した影響と交換相互作用長内に結晶磁気異方性を平均化させるのに充分な粒子数が存在しない
可能性があるため，不活性雰囲気下で，表面酸化膜のないシングルナノサイズの軟磁性ナノ粒子の成形の必要性
が明らかとなった． 
 第4章では，性能実証用材料として低保磁力化が期待されるシングルナノサイズの合成が可能なホットソープ
法を用いて作製した，シングルナノサイズ鉄粒子を用いてコールドスプレー成形を試みた．シングルナノサイズ
鉄粒子は非常に酸化しやすく，大気中は取り扱いができないため，通常は大気中で使用するコールドスプレー装
置を不活性雰囲気管理が可能なグローブボックスの中に入れた装置を開発した．これによって，1ppm 以下の酸
素濃度下でコールドスプレー成形が可能となり，シングルナノサイズ鉄粒子を安全に取り扱えるようになった．
成形に使用した純鉄ナノ粒子は，平均粒子径が5.4nmの球状粒子であり，不活性雰囲気下での圧縮成形と分級に
よって20-200μmにサイズを整えて使用した．実際にコールドスプレー成形をおこない，粒子表面に残留した界
面活性剤の除去と純鉄ナノ粒子の結晶性向上のためにポストアニール処理行った．その結果，コールドスプレー
法成形体はバルク体相当の充填率を実現し，かつ原料粉末の保磁力よりも成形体の保磁力を 1/6まで低下させる
ことに成功した．さらに，この低保磁力化の要因解明のため，磁区観察を行ったところ，ポストアニール前のサ
ンプルは粒子同士が分断されており，直接交換相互作用が働いておらず，微小な磁区を形成していた．それに対
して，ポストアニール後は数十ミクロンサイズの大きな磁区を形成しており，磁壁幅から直接交換相互作用によ
る磁気異方性の平均化が確認された．これにより，交換結合作用長内に充分な数の軟磁性ナノ粒子が存在し，か
つ表面酸化膜の無い状態で結合することによって保磁力の起源である結晶磁気異方性を平均化することが可能と
なり，保磁力を低下させることが可能になったと結論付けた． 
 第5章では，コールドスプレー法においてナノ粒子が接合するメカニズムについて粒子形態から原子拡散，新
生界面生成，静電気力による接合が有力と考え，酸化膜のある粒子と水素処理によって表面酸化膜を除去した粒
子とをそれぞれコールドスプレー成形をおこない，断面TEMによる基板付近での粒子形態の詳細な観察と微小
引張試験を行うことで，接合界面の状態とその皮膜強度を検証した．その結果，断面TEMの結果から，酸化膜
のあるナノ粒子は基板付近では塑性変形によって原料の球状を維持していないほど大きな塑性変形を起こしてい
るのに対して，酸化膜を除去したナノ粒子は球状を維持していた．一方で，微小引張試験の結果からは，酸化膜
を除去したナノ粒子は基材としっかりと接合しており，界面でのはく離は起こらず，Al基材側で破断が起こった
のに対して，酸化膜のあるナノ粒子では Al 基材とのはく離が生じていた．また，酸化膜の有無に関わらず，ナ
ノ粒子膜内での引張試験ではバルク材を超える顕著に高い引張強度を有していることが明らかとなった．以上の
結果から，コールドスプレー法によるナノ粒子と Al 基材およびナノ粒子同士の接合メカニズムとして静電気力
によるものとは考えにくく，金属結合もしくは共有結合に相当する強固な結合が実現できており、接合には新生
界面による接合が重要であることが示唆された． 
 第6章は，結言として本論文をまとめた．軟磁性材料は優れた加工性と高い飽和磁束密度，低保磁力の実現が
課題であったが，本論文によって付加造形手法であるコールドスプレー法で，磁粉レベルでは優れた軟磁性特性
を示していた純鉄ナノ粒子の粒子径を維持したまま高充填成形が可能となり，バルク相当の高い飽和磁束密度を
実現することができ，磁粉保磁力を下回る低保磁力成形体の成形が可能となった．以上の結果から，軟磁性ナノ
粒子の高充填化成形における基盤技術の構築と基礎的な検討が完了した． 
 本論文では，車載用軟磁性材料の性能向上を目指し，シングルナノサイズの純鉄ナノ粒子を高充填成形する成
形手法や成形装置および成形体の分析技術の構築を行ったが，これらの結果は軟磁性材料だけでなく，ナノ粒子
特有の特性を活かしつつ，高充填成形が必要な触媒や誘電材料などの様々な機能性ナノ材料への応用が可能であ
ると考えられる．今後は他材料への応用も検討することによって，種々の機能性ナノ材料の実用化が期待できる． 
 


Japanese

コールドスプレー法によるモーター特性向上のための軟磁性ナノ粒子高充填成形法の確立

KANEKO, SAYUMI

High-Resolution PIV Measurement and State-Space-Model-Based Spatio-Temporal Superresolution of Underexpanded Supersonic Jets

FARIDAFSHIN, MOHAMMADJAVAD

論文内容要約 
 The divertor is one of the highest thermally loaded components in a nuclear fusion reactor and one of its 
responsibilities is to remove heat and ash produced by nuclear fusion reaction. In one of the designs of the 
divertor target that is called the “mono-block”, tungsten (W) blocks are bonded to the cooling pipe made from 
copper-chromium-zirconium (CuCrZr), through an interlayer of oxygen-free copper (OFCu). The W blocks 
provide a plasma facing surface and are subject to high heat flux loads, and to keep their surface from 
overheating, an active cooling technique is applied. In the water-cooled divertor, the heat removal 
performance from the W blocks to the coolant water highly relies on the quality of the bonded interface 
between the blocks and the pipe. Due to the large difference between the thermal expansion coefficient of W 
and OFCu, a large stress would be concentrated on the W-OFCu bonded interface during fabrication of the 
bond that makes the bond susceptible to developing defects that can turn into thermal barriers, block the 
heat removal performance, and cause failure of the divertor. Therefore, it is necessary to perform pre-service 
non-destructive inspections on the divertor to evaluate the quality of the W-OFCu bonded interface before 
the operation of the nuclear fusion reactor. Among the several non-destructive inspection methods whose 
applicability has been analyzed for the inspection of the W-OFCu bonded interface of the divertor, ultrasonic 
testing (UT) in the pulse-echo immersion technique has been reported to have the best performance. 
Furthermore, it has been shown recently that performing UT at sufficiently high frequencies – a method 
that is called High-Frequency UT (HF-UT) – further improves the inspection performance and enables the 
detection of small defects that were invisible to lower frequencies. The aim of this study is to 1) analyze the 
HF-UT performance to inspect an internally grooved cooling pipe called the “screw pipe” to open the 
discussion on the ways to improve the inspectability of this pipe design, and, 2) discuss how the performance 
of HF-UT will be affected when OFCu grains undergo enlargement, especially, in the special cases that grain 
enlargements have high locality in a small area. The contents of this dissertation are organized as follows:  
氏 名          Mohammadjavad Faridafshin 
研究科，専攻の名称 東北大学大学院工学研究科（博士課程）量子エネルギー工学専攻 
学 位 論 文 題 目          Study on Non-Destructive Inspection of the Tungsten/Copper Bonded 
Interface in the Divertor of a Nuclear Fusion Reactor using 
High-Frequency Ultrasonic Testing (高周波超音波による核融合炉ダイバー
タのタングステンー銅接合界面の検査技術開発に関する研究) 
論 文 審 査 委 員          主査 東北大学教授 遊佐 訓孝 東北大学教授 橋爪 秀利 
             東北大学教授 笠田 竜太 東北大学教授  内一 哲哉 
             東北大学准教授 江原 真司          
 
Chapter 1 contains the background, preliminary results, and objectives of this study.  
Chapter 2 contains theories of the influence of grooves of screw pipe on ultrasonic waves. The details of 
experiments performed in this study are also explained in this chapter.  
Chapter 3 analyzes the inspection of screw pipe using HF-UT. Tungsten-OFCu bonded samples were 
prepared with grooves on the OFCu side, and a flat-bottom hole on the center of the face of the W block to 
simulate a defect at the interface. The grooves had rectangular and triangular geometries with various 
geometric parameters to compare their influence on HF-UT signals. The tests were performed from the 
grooved surface of the samples, and the signals were analyzed by the Snell’s law and Time-of-Flight (ToF) 
analysis. The analysis results showed that the original design of the screw pipe has a low capability to be 
inspected with HF-UT. The Snell’s law analysis showed that the slope of the grooves in the original screw 
pipe design prevented the ultrasonic waves to penetrate  
OFCu at the water-OFCu interface that made the areas of the W-OFCu interface under the slope of the 
grooves to be invisible to HF-UT. To enhance the HF-UT performance for the inspeciton of the screw pipe 
geometric modifications were proposed for the grooves. The signals obtained from samples with a modified 
grooved geometry showed that employing step-like grooves with a reduced ratio of height to the pipe wall 
thickness considerably enhanced the visibility of the drilled hole at the W-OFCu interface. The results were 
validated by destructive testing of the samples to evaluate their physical condition, and the cross-sectional 
images of the W-OFCu bonded interface confirmed that the inspection results were in agreement with the 
physical condition of the interface.  
Chapter 4 contains the numerical analysis of the influence of the geometric modifications that were 
proposed in Chapter 3 on the heat removal performance of screw pipe in the water-cooled divertor of a 
nuclear fusion reactor. Numerical simulations were performed in the finite element software COMSOL 
Multiphysics version 6.1, and screw pipes with rectangular and triangular geometries and varying 
geometrical parameters were modelled. The single-phase analysis results indicated that screw pipes with 
original triangular grooves had comparatively the same heat removal performance as the pipe with 
rectangular grooves with shallower height and shorter length. Also, it was shown that in pipes with 
rectangular grooves with a higher ratio between the grooves’ height to length, the water inside the grooves 
develops two vortices on top of each other with the temperature near the grooves’ top surface being close to 
the boiling temperature. It is suggested that in the case of boiling, the double vortices can help remove the 
bubbles and prevent them from entering the main channel of the pipe.  
Chapter 5 investigates the influence of OFCu and CuCrZr grain enlargement on HF-UT signals. Block 
OFCu and CuCrZr samples were prepared and heat treated in different conditions to have grains with 
various sizes. Some samples contained drilled holes with the same size on different locations on one of their 
sides to facilitate analyzing the effect of local grain enlargement on the defect detection performance of 
HF-UT. Experiments were performed from the flat surface of the samples using 10 MHz, 15 MHz, 25 MHz, 
and 35 MHz frequencies and the signals were analyzed quantitatively and qualitatively. To quantify the 
noise caused by ultrasonic scattering by grains, standard deviation of the ultrasonic C-scan signals was 
calculated and compared for samples with different average grain areas. The analysis results showed that 
grain enlargement is the reason of increased fluctuation in HF-UT signals as was previously suggested, and 
that the noise increases with the increase in the average grain area of the samples. The results also 
indicated that the detection performance of HF-UT is lower for samples that contain local grain enlargement, 
compared to the samples with the same average grain size but without local grain enlargement.  
Chapter 6 summarizes this study and presents several future tasks for further improvement. 
 


Study on Non-Destructive Inspection of the Tungsten/Copper Bonded Interface in the Divertor of a Nuclear Fusion Reactor using High-Frequency Ultrasonic Testing

MIYAUCHI, KEN

論文内容要約 
Complementary-metal-oxide-semiconductor (CMOS) image sensors are used widely not only in the 
imaging field for conventional cameras but also in the sensing field for object detection purposes, such as 
automotive, factory automation (FA), and augmented reality and virtual reality (AR/VR), as the eyes for 
machines. The two most important requirements in such image sensors for sensing fields are high dynamic 
range (HDR) and global shutter (GS) to suppress the clipped white, crushed shadows, and rolling shutter 
distortion issues. Although several GS and HDR technologies have been proposed and implemented, 
effective technologies for achieving both GS and over 120 dB HDR, which is required in our daily life, in the 
practical pixel size of less than 5 μm, have not been established yet. 
Taking these backgrounds into consideration, this doctoral thesis gives the effectiveness of a novel 
pixel-level stacking in-pixel quantization technology through the implementation, fabrication, and 
evaluation of small size HDR-GS CMOS image sensors. This doctoral thesis consists of 7 Chapters. 
In Chapter 2, several prior works of HDR, GS, and pixel size shrinkage technologies are summarized. 
These advantages and drawbacks are summarized. HDR technologies are classified in terms of 
input-output characteristics and the method of generating multiple signals with different responsivity. GS 
technologies are classified by the in-pixel memory structure, which is needed for GS CMOS image sensors. 
In Chapter 3, a novel stacked HDR voltage mode GS (VMGS) technology with back-side illumination 
(BSI) type dual conversion gain (Dual-CG) and lateral overflow integration capacitor (LOFIC) pixels is 
demonstrated. In this technology, the reset and the signal levels of high conversion gain (HCG) and low 
conversion gain (LCG) signals are stored in the in-pixel capacitors to realize both HDR and GS. Moreover, 
smaller pixel size is realized by pixel-level stacking with hybrid bonding. Through the implementation of a 
4.0 μm pixel image sensor in a 45 nm/65 nm stacked process, 90 dB HDR and GS are demonstrated. 
In Chapter 4, a novel pixel-level stacking in-pixel quantization technology, as shown in Fig. 1, for further 
DR extension, low power consumption, and high frame rate, is proposed, and its effectiveness is 
           みやうち けん 
氏 名          宮 内 健 
研究科，専攻の名称 東北大学大学院工学研究科（博士課程）電子工学専攻 
学 位 論 文 題 目          A Study on In-Pixel Quantization and Dynamic Range Extension 
Technologies for 3D Stacked Global Shutter CMOS Image Sensors(3 次元積層型グローバルシ
ャッタCMOSイメージセンサの画素内量子化とダイナミックレンジ拡大技術に関する研究) 
論 文 審 査 委 員          主査 東北大学教授 黒田 理人 東北大学教授 金子 俊郎 
             東北大学教授 吉信 達夫 東北大学教授 日暮 栄治 
demonstrated by the chip fabrication and evaluation. The in-pixel quantization of this thesis is called triple 
quantization (3Q) digital pixel sensor (DPS), where the equivalent saturation signal is extended by the 
saturation-time detection (Time-to-saturation: TTS) for the overflow integration charges of LOFIC pixel, as 
shown in Fig. 2. In other words, TTS is a DR extension technology where the integration time of each pixel 
is automatically set depending on the input light intensity. With the LOFIC readout operation under low 
and middle light conditions, the low light performance is maintained as high as in the conventional CMOS 
image sensors. Moreover, an in-pixel illuminance-adaptive mode selection technology is proposed to reduce 
the number of readout signals and in-pixel memories for further reduction of pixel size and power 
consumption, and improvement of frame rate. Through the chip fabrication in 45 nm/65 nm stacked 
process and the evaluation, 127 dB HDR, as shown in Fig. 3, and GS function with an extremely low power 
consumption of 5.75 mW have been achieved in a 4.6 μm, 512 × 512 pixel DPS. 
In Chapter 5, a floating diffusion (FD) shared dual-bank static random access memory (SRAM) 3Q-DPS 
technology with a finer process node, as shown in Fig. 4, is demonstrated. In the stacked-DPS, it is possible 
to use an advanced process node for the analog/digital circuits in the bottom pixel. Through the chip 
fabrication in a 45 nm/40 nm stacked process and the evaluation, 124 dB HDR and GS function have been 
achieved in a 640 × 640 pixel DPS with a further shrunk pixel size of 3.96 μm. Moreover, two additional 
functions are demonstrated. One is the simultaneous acquisition of visible and depth images. It is effective 
for small form factor AR/VR devices. The other is the dark noise cancellation in a single frame by 
subtracting the reset and signal levels, which are stored in the dual-bank SRAM. 
In Chapter 6, modeling of the 3Q-DPS, especially for analysis of mode transition, is described. From the 
established theoretical model, it is demonstrated that the photodiode full well capacity (PDFWC), the 
LOFIC capacitance, and the input range of the analog-to-digital converter are the most important among 
many design parameters. It is verified by the evaluation results with several pixel layout patterns, which 
have different PDFWC and LOFIC capacitance. 
The conclusion is given in Charter 7. This doctoral thesis gives the effectiveness of a novel pixel-level 
stacking in-pixel quantization technology through the implementation, fabrication, and evaluation of small 
size HDR-GS CMOS image sensors, as shown in Fig. 5 and Fig. 6. Table 1 and Table 2 show the achieved 
performance of the 3Q-DPS in Chapter 4 and the dual-bank 3Q-DPS in Chapter 5, respectively. The 3D 
stacked in-pixel quantization technology is effective for further pixel size shrinkage and DR extension by 
the progress of the advanced semiconductor process. This in-pixel quantization technology will play a key 
role in future CMOS image sensors for sensing applications. 
Top pixel
Bottom 
pixel
・・・
Pixel parallel ADC
Pixel level HB
・・・
Mode 
control
CS
ADC
Incident light
Overflow
BSI LOFIC base pixel
 
RST
TG
DCG
AB
TTS PDADC FDADC
AZ
COMP_SET
VRAMP
WL_SEL
COMP_CHK
1023 768 511 0 512 767counter
WL_DATA
High illum.
Middle illum.
Low illum.
VREF
VTTS_th
Code latch
t0 t1 t2 t3 t4 t5 t6 t7 t8 t9
CL
Wake Exposure & Pixel ADC SRAMReadout Sleep
Trigger (V sync)
Sensor state
1 frame period
 
(a) Pixel-level stacking structure. (b) Pixel operation timing. 
AZ
WL_SEL
State 
latch
COMP
_CHK
COMP_SET
WL_DATAVOUT1
STATE
VOUT2 VOUT3
SET
10-bit 
SRAM
AB TG
DCG
RST
CS
SF
Vbn
PD
HI
LO
CL
BSI LOFIC base pixel Comparator Mode control Digital memories
Overflow
 
(c) Pixel schematic. 
Fig. 1 Triple-quantization (3Q)-DPS with in-pixel quantization technology. 
 
VFD_RST
VFD_REF
RST
TG
0 tfliptOF
t
ΔVTTS
Nmin Nflip Nmax
tint
Time Code
F
D
 v
ol
ta
ge
High illum.
Low illum.
LOFIC
operation
ΔVTTS
VFD_REF
 
Fig. 2 3Q-DPS technology for dynamic range extension. 
0
100
200
300
400
500
600
700
800
900
1000
0
200
400
600
800
1000
1200
0.001 0.01 0.1 1 10 100 1000 10000 100000100 1 1 102 103 1 4 5 1 6 1 7
Light intensity [arb. unit]
O
ut
pu
t s
ig
na
l [
LS
B
]
0
1 8
1
2
3
4
5
6
7
8
9
1
M
od
e
ra
tio
 [%
]
TTS [%]FDADC [%]PDADC [%]
 
-10
0
10
20
30
40
50
60
70
80
90
100
0.1
1
10
100
1000
10000
100000
1000000
10000000
0.001 0.01 0.1 1 10 100 1000 100001000001 0 1 1 102 103 104 5 06 1 7
Light intensity [arb. unit]
10-1
100
1
1 2
1 3
1 4
1 5
1 6
1 7
R
ec
on
st
ru
ct
ed
 o
ut
pu
ts
ig
na
l [
LS
B
]
-10
0
108
10
20
30
40
50
60
70
80
90
10
S
N
R
 [
dB
]
 
(a) Non-linearized 127 dB photo-response curve. (b) Linearized 127 dB photo-response curve. 
Fig. 3 Measured photo-response curve of fabricated 3Q-DPS. 
Top pixel Bottom pixel
Hybrid bonding
DCG
RST
PCS
FD
Cs
VAAPIX
-
+
CC
VREF
Comp_RST
State
& bank 
control
SRAM
10bit
(Bank0)
Bit line
(10 bit)
SRAM
10bit
(Bank1)
AB1 TG1
AB2 TG2
AB3 TG3
AB4 TG4
Comp_CHK
State_Init
SF
MONO
PD1
MONO
PD2
NIR
PD1
NIR
PD2
CL
Overflow
drain
(OFD)
Overflow direction
40nm process
(65nm process in previous 3Q-DPS)
45nm process
Dual-bank SRAM
FD-shared pixel
 
AB4TG4TG1AB1
Clear
Y Y’
FDOFD OFD
Deep-PDN Deep-PDN
BDTI
PDN PDNBOFP
BSM
BOFP
TG : Transfer gate
AB : Anti-blooming gate
FD : Floating diffusion
OFD : Overflow drain
BOFP : Buried overflow path
PDN : N-layer of PD
Deep-PDN : Deep n-layer of PD
BDTI : Back-side deep trench isolation
BSM : Back-side metal
MONO PD1 NIR PD2
NIR Pass 
Filter
Large micro-lens
Small micro-lens
 
(a)  Pixel schematic. (b) Top pixel structure. 
Fig. 4 Dual-bank 3Q-DPS with 40 nm process. 
4.
0m
m
4.0mm
Pixel array
540H × 560V
(Total pixels)
512H × 512V
(Effective pixels)
 
Bottom pixel array
[ADC, SRAM]
512×512 (Effective)
540×560 (Total)
Clock gen
A
D
C
 
co
nt
ro
lle
r
Pixel SRAM bit-line driver
A
D
C
 r
o
w
 d
ri
ve
r
A
D
C
 p
ix
e
l b
ia
s
C
IS
 r
o
w
 d
ri
ve
r
C
IS
 p
ix
e
l b
ia
s
Read 
out
S
en
so
r 
I/
F
 &
 
co
nt
ro
lle
r
Ramp 
generator
Temp 
sensor
BGR
IREF
MIPI 
I2C/I3C
MIPI C/D-PHY
(1.5Gs/s)Pixel SRAM readout
CIS layer
ADC layer
C
IS
 r
o
w
 d
ri
ve
r 
C
IS
 p
ix
e
l b
ia
s
Digital
w/o HB
PLL
PHY
w/ HB
Hybrid bonding
(HB)
ADC
SRAM
Top pixel array
[PD, Pixel transistors]
512×512 (Effective)
540×560 (Total)
ADC
SRAM
ADC
SRAM
PD
FD
CS
PD
FD
CS
PD
FD
CS
Signal 
readout path
Pixel array
668H × 694V
(Total super pixels)
640H × 640V
(Effective super pixels)
4.4mm
4.
1m
m
 
Bottom pixel array
ADC, SRAM
668×694 (Total)
640×640 (Effective)
Clock gen
A
D
C
 
co
nt
ro
lle
r
SRAM bit-line driver
B
ot
to
m
 p
ix
el
 r
ow
 d
riv
er
B
ot
to
m
 p
ix
el
 b
ia
s
To
p
 p
ix
el
 r
ow
 d
riv
er
Top pixel bias
Data 
path
S
en
so
r 
I/
F
 &
 
co
nt
ro
lle
r
Ramp 
generator
Temp 
sensor
BGR
IREF
MIPI 
I2C/I3C
MIPI C-PHY
(2.5Gs/s)SRAM readout
CIS layer
ADC layer
To
p 
p
ix
e
l r
o
w
 d
riv
er
 
Digital
Analog w/o HB
PLL
PHY
Analog with HB
Hybrid 
bonding
(HB)
On-chip regulators
(charge pump, LDO)
Top pixel bias
Top pixel array
Sub-pixel
1,336×1,388 (Total)
1,280×1,280 (Effective)
Super-pixel
668×694 (Total)
640×640 (Effective)
ADC
SRAM
ADC
SRAM
ADC
SRAM
M
IR M
IRM
IR M
IRM
IR M
IR
 
Fig. 5 3Q-DPS chip in Chapter 4. Fig. 6 Dual-bank 3Q-DPS chip in Chapter 5. 
 
Table 1 Specifications of 3Q-DPS in Chapter 4. Table 2 Specifications of dual-channel 3Q-DPS in Chapter 5. 
Specification 
45nm/65nm stackProcess technology
4.6Pixel size [µm]
512H×512V# of pixels (H*V)
10b (1 pixel/ADC)In pixel memory bit #
Global shutterShutter operation
98 (Mono pixel)Max. QE (530nm) [%]
42 (Mono pixel)NIR QE (940nm) [%]
127Dynamic range [dB]
170/7Conversion Gain (HCG/LCG) [µV/e-]
3.8/51/9000Linear full well capacity [ke-]
4.2Noise floor [e-]
47Dark FPN (w/o FPN cancel) [e-]
300 at 1ms exposureFrame rate [fps]
5.75Power [mW] at 30fps  
Specification 
45nm/40nm stackProcess technology
3.96Pixel size [µm]
640H×640V# of pixels (H*V)
20b (1 pixel/ADC)In pixel memory bit #
Global shutterShutter operation
66 (Mono pixel)
1.6 (NIR pixel)
Max. QE (530nm) [%]
22 (Mono pixel)
19 (NIR pixel)
NIR QE (940nm) [%]
124Dynamic range [dB]
108/7.7Conversion Gain (HCG/LCG) [µV/e-]
2.6/35.7/3234 (Mono pixel)
6.3 (NIR pixel)
Linear full well capacity [ke-]
6.3 (w/o D-CDS)
6.4 (w/ D-CDS)
Noise floor [e-]
43.5 (w/o D-CDS)
5.4 (w/ D-CDS)
Dark FPN [e-]
280 at 1ms exposureFrame rate [fps]
6.2Power [mW] at 30fps   


A Study on In-Pixel Quantization and Dynamic Range Extension Technologies for 3D Stacked Global Shutter CMOS Image Sensors

廣瀬, 達也

論文内容要約 
Chapter 1: Introduction 
Peptides and proteins that bind to inorganic materials are expected to be used in various fields, such as 
creation of new inorganic materials by biomineralization and interface molecules in biosensors and 
bioimaging. Inorganic material-binding proteins exist in nature, but the available protein species are limited 
for industrial use. Therefore, it is necessary to establish a method that can rapidly and efficiently develop 
peptides and proteins binding to arbitrary inorganic materials. 
In this study, I aimed to obtain peptides and antibodies that bind to indium tin oxide (ITO) and zirconia 
(ZrO2) using phage display method, which can display various peptides or antibodies on the phage surface. 
Here, peptides are composed of 12 amino acids, and antibodies are variable domain of the heavy chain of 
heavy-chain antibodies (VHHs). Although phage display method to organic targets is widely used to prepare 
binding peptides and antibodies, no phage selection method suitable for obtaining peptides and antibodies 
binding to inorganic materials has been established. Additionally, as the sequence diversity of antibody 
complementarity-determining regions (CDRs) is as high as 1×1065, not all sequences can be explored 
exhaustively using phage display method. Therefore, in this study, I obtained binding peptides by phage 
display method, grafted the peptides into one of the CDRs of VHH, and obtained high binding affinity 
antibodies by optimizing the CDRs by random mutations using phage display method. Such a stepwise CDR 
optimization method does not require large phage library sizes because the length of amino acid sequences 
to be optimized will be reduced in each optimization step. Additionally, I aimed to establish a more accurate 
and efficient method for obtaining antibodies binding to inorganic materials by examining the conditions for 
the phage selection operation and by utilizing next generation sequencing (NGS) analysis. Furthermore, the 
binding mechanism of the ITO- and ZrO2 nanoparticle-binding peptides and antibodies was discussed by 
analysis of the obtained peptide and antibody sequences, analysis of the phage selection results, and 
evaluation of the binding properties of the obtained antibodies. 
           ひろ せ たつ や 
氏 名          廣 瀬 達 也 
研究科，専攻の名称 東北大学大学院工学研究科（博士課程）バイオ工学専攻 
学 位 論 文 題 目          無機ナノ粒子結合性タンパク質の取得法の構築と結合機構の解明 
 
論 文 審 査 委 員          主査 東北大学教授 梅津 光央 東北大学教授 魚住 信之 
             東北大学教授 珠玖 仁 教授 
 
 
Chapter 2: Generation of Indium Tin Oxide nanoparticle-binding peptides by phage display method 
Since ITO has been used as electrodes and substrates in a variety of devices, it is useful to fuse other 
materials via ITO-binding peptides. I examined various washing solutions and eluates in screening 
peptide-displayed phages to obtain ITO nanoparticle-binding peptides. When phages were washed with 
commonly used Tris buffer during phage selection, non-specific phage adsorption to ITO nanoparticles was 
observed. On the contrary, washing with phosphate buffer or acetonitrile greatly reduced the nonspecific 
adsorption. As eluates during phage selection, acidic buffer, high-phosphate buffer, and guanidine 
hydrochloride solution were examined, and different ITO nanoparticle-binding peptides were obtained from 
each eluate. The peptides obtained in high-phosphate buffer had electrostatically interacting lysine and 
coordinatively binding histidine rich sequences, and showed high binding affinity and phosphate-dependent 
binding to ITO nanoparticles. In contrast, peptides obtained in guanidine hydrochloride solution had fewer 
basic amino acids and had phosphate-independent binding. These results showed that it was possible to 
obtain ITO nanoparticle-binding peptides with different binding characteristics by changing the eluate in 
phage selection. 
 
Chapter 3: Generation of Zirconia nanoparticle-binding peptides and antibodies by phage display method 
  Since ZrO2 (Zirconia) is chemically and thermally stable, has high strength and is highly biocompatible, it 
is expected to be applied in combination with various biological molecules by ZrO2-binding peptides and 
antibodies. Firstly, ZrO2 nanoparticle-binding peptides were obtained using phage display method. As 
eluates during phage selection, high-phosphate buffer, high-salt buffer, and acidic solution were examined, 
and different ZrO2 nanoparticle-binding peptides were obtained from each eluate. The ZrO2 
nanoparticle-binding peptide ZrO2BPn obtained by high-salt buffer had a sequence rich in basic amino acids, 
while the peptide ZrO2BPa obtained by acidic solution had lower number of basic amino acid and low 
isoelectric point. From these results, similar to the results of obtaining ITO-binding peptides in Chapter 2, it 
was found that peptides with different binding characteristics could be obtained by varying the eluate. Then, 
based on the CAnIGET (construction of antibodies by integrating grafting and evolution technology) method 
developed in our laboratory, a peptide (ZrO2BPa or ZrO2BPn) was grafted into CDR1 of the template VHH, 
and CDR3 was optimized by screening a phage library randomized to αββαββαββα sequence (α: Arginine, 
histidine, β: Arginine, glycine, leucine, valine). ZrO2 nanoparticle-binding VHHs could be generated by 
grafting the peptide into CDR1 of template VHH, and their bindings were highly improved by optimizing 
the CDR3s. As a result, ZrO2A1 VHH with a KD of 251 nM was obtained from ZrO2BPa-derived library and 
ZrO2N3 VHH with a KD of 253 nM was obtained from ZrO2BPn-derived library. These antibodies showed 11- 
and 6-times higher affinity than the peptide-grafted antibody, respectively. These antibodies were evaluated 
for binding to ZrO2 nanoparticles in various buffers of varying phosphate, salt concentration, and pH. Since 
both antibodies bound in a manner dependent on the phosphate and salt concentrations, it was found that 
they bound mainly by electrostatic interactions. Both VHHs showed pH-dependent binding, with the 
highest binding observed at pH 6.0, likely due to the increased electrostatic interaction caused by the 
positive charge of histidine. Some degree of binding was maintained at pH 4.5, indicating that additional 
binding mechanisms other than electrostatic interactions such as hydrophobic interactions are involved. 
Although the peptides were likely to have different binding properties due to differences in peptide sequence, 
it was likely that during CDR3 optimization of the VHHs, CDR3s were changed to sequences rich in 
arginine and histidine, resulting in a similar binding mechanism in the final VHHs.  
 
Chapter 4: Affinity maturation of Zirconia nanoparticle-binding antibodies and establishment of phage 
screening conditions using next generation sequencing 
I aimed at further improving the affinity of ZrO2A1 VHH obtained in Chapter 3 by phage display utilizing 
NGS analysis. At first, phages library with single amino acid saturation mutations at all positions in the 
CDRs was screened, and beneficial mutations affecting the improvement of binding affinity were identified 
by identifying the enriched mutations by NGS analysis of phage library and screened phages. These 
beneficial mutations were commonly found in CDR1, which was predicted by the conformational prediction 
to be a region highly solvent exposed and likely to contribute to the binding. CDR3, on the other hand, was 
also predicted to be a region highly solvent exposed, but there were few useful mutations, and it was 
considered that there was little room for optimization because optimization was performed based on the 
αββαββαββα sequence in Chapter 3. Since most beneficial mutations were substitutions to basic amino acids, 
the addition of positively charged electrostatic interactions by arginine and lysine and coordination bonds by 
histidine were found to be useful in improving binding affinity. Most preferred mutations were lysine, but 
some preferred mutations were arginine or histidine, indicating that the preference for specific basic amino 
acids depends on the CDR location. Next, phage library displaying VHHs combining useful mutations was 
prepared and phage selection was performed. High-phosphate buffer, high-salt buffer, and acidic solution 
were used for elution in phage selection. Sequence analysis by NGS revealed that similar antibodies were 
concentrated by elution using high-phosphate buffer and high-salt buffer, but different antibodies containing 
more arginine-lysine were concentrated by elution using acidic solution. Subsequently, antibodies 
concentrated by each elution method were selected based on the results of NGS analysis, and the binding 
affinity to ZrO2 nanoparticles was evaluated. ZrO2A1AM1 VHH found by high-phosphate buffer had a 
sequence with one additional arginine in basic amino acids compared to the parent antibody, but showed a 
2.6 fold increase in binding affinity compared to the parent antibody. On the other hand, the binding affinity 
did not improve with the other obtained antibodies. Particularly, the antibody found by acid elution had a 
sequence containing more arginine and lysine, but the affinity did not improve. Therefore, instead of simply 
increasing the number of basic amino acids in the CDR to improve the affinity, it was found to be important 
to introduce an appropriate number of basic amino acids at appropriate positions. In addition, based on the 
results in the Chapter 2, 3 and 4, high binding affinity peptides and antibodies tended to be obtained more 
easily by elution using high-phosphate buffer and high-salt buffer than by acidic solution. Throughout this 
study, the Output/Input values for phage selection tended to be higher for elution using phosphate buffer 
and high-salt buffer than for acidic solution, suggesting that high-binding affinity phages cannot be 
dissociated by acid elution. On the other hand, since high binding affinity peptides and antibodies generally 
bind by electrostatic interactions, there is a possibility that peptides and antibodies with different binding 
mechanisms can be obtained by acid elution, as seen in the acquisition of ZrO2 nanoparticle-binding 
peptides. 
 
Chapter 5: Conclusion 
In this study, I obtained knowledge on an efficient preparation method and optimization method of 
inorganic material-binding peptides and antibodies through the preparation of peptides and antibodies 
against ITO and ZrO2 nanoparticles using the phage display method. This method can be applied to various 
other inorganic materials. In this study, I showed that peptides and antibodies with different binding 
mechanisms could be obtained by changing the washing and elution methods during phage selection, and I 
succeeded in identifying antibodies with phosphate/salt concentration- or pH-dependent binding properties. 
Applications that take advantage of environment-dependent changes in binding properties may be 
considered, for example, controlling the crystal growth of inorganic materials by binding to inorganic 
materials only under specific solution conditions, or exerting therapeutic effects by assembling inorganic 
materials only in cancer environments around pH 6.0. It is expected that the findings of this research will 
lead to rapid and reliable preparation method of peptides and antibodies with desirable binding properties 
and new utilization of inorganic materials in various fields. 


無機ナノ粒子結合性タンパク質の取得法の構築と結合機構の解明

XU, YIPU

 
 
          じょ いふ 
氏 名          XU YIPU 
研究科，専攻の名称 東北大学大学院工学研究科（博士課程）材料システム工学専攻 
学 位 論 文 題 目          Fabrication of Bimetallic Structure Consisting of Austenitic Stainless Steel and Ferritic 
Stainless Steel using Dual-wire Arc Additive Manufacturing  
                                          （デュアルワイヤーアーク積層造形法によるオーステナイト系ステン
レス鋼とフェライト系ステンレス鋼のバイメタル構造の作製） 
論 文 審 査 委 員          主査 東北大学教授 佐藤 裕  東北大学教授 吉見 享祐 
             東北大学教授 野村 直之 東北大学准教授 山中 謙太 
論文内容要約 
Chapter 1 Introduction 
Austenitic stainless steel (ASS) is the most versatile metal applied in diverse fields due to its excellent 
ductility, high corrosion resistance and suitable weldability. While in the context of carbon neutrality, the Ni-
free ferritic stainless steel (FSS) has regained the spotlight as a potential substitute of ASS for its lower cost. 
Moreover, owing to its good thermal conductivity and superior resistance to stress corrosion cracking (SCC), 
the bimetallic joining of FSS-ASS is widely used in industrial applications such as petrochemical power plants, 
energy conversion systems and nuclear reactors. However, the differences of chemical, physical, mechanical, 
and metallurgical properties make dissimilar metal welding (DMW) more difficult as compared to similar  metal 
welding. Accordingly, a series of issues still remain in DMW, including porosity, detrimental phase, grain 
coarsen in heat-affected zone, and structural distortion caused by different coefficient of thermal expansion. 
Functionally graded materials (FGMs) are advantageous in mitigating this challenge because their chemical 
composition changes gradually from one side to the other, resulting in continuous or quasi-continuous gradient 
changes in properties according to the design requirements. In the past decade, the emerging additive 
manufacturing (AM) technologies are acknowledged as a promising approach to fabricate FGMs in current 
industrial applications, featured as flexible geometry, cost optimization as well as higher productivity. Among 
them, dual-wire arc additive manufacturing (D-WAAM) offers the capability of fabricating FGMs owing to its 
adjustable wire feeding speeds. D-WAAM process utilizes an electric arc as the heat source to melt metal wires 
and deposit components layer by layer, which provides a combination of high deposition efficiency, low device 
and material cost, and unlimited building size. 
During WAAM process, each location of the fabricated part experiences the complicated thermal history 
including repeated heating and cooling, directional thermal dissipation, and inconstant cooling rate. This factor, 
 
 
combined with the compositional variation, results in a complex microstructural evolution in FGMs. Such 
complexity profoundly affects the microstructure distribution, ultimately reflecting in the mechanical properties. 
This is the first study focusing on the control of the microstructure and mechanical properties of the ASS-FSS 
bimetallic structure by compositionally graded design using D-WAAM. 
 
Chapter 2 Interfacial characteristics in austenitic stainless steel-ferritic stainless steel bimetallic 
component fabricated by dual-wire arc additive manufacturing 
In this chapter, a bimetallic structure composed of austenitic stainless steel (ASS) and ferritic stainless steel 
(FSS) is fabricated using the dual-wire arc additive manufacturing (D-WAAM) process. The interfacial 
characteristics of the additively manufactured bimetallic structure are extensively analyzed through detailed 
microstructural observations and mechanical performance assessments. The interface exhibited distinct 
morphological features between the ASS and FSS regions, characterized by the formation of lathy α’-martensite, 
which significantly enhanced hardness and 
tensile strength in the interfacial region 
(Fig.1).  The microstructural evolution at 
the interface is further investigated in the 
context of elemental diffusion behavior and 
thermal cycles. Remelting at the interface 
facilitated the migration of γ-stabilizer 
elements from the ASS side to the FSS side, 
promoting martensitic transformation 
during cooling to ambient temperature. 
 
Chapter 3. Influence of compositional variation on the microstructure of austenitic stainless steel-ferritic 
stainless steel bimetallic component fabricated by dual-wire arc additive manufacturing 
Fig. 1 Macroscopic overview and relevant hardness profile 
along the building direction of the bimetallic structure. 
 
 
In this chapter, a compositionally graded bimetallic component of austenitic stainless steel (ASS) and ferritic 
stainless steel (FSS) is successfully fabricated using the dual-wire arc additive manufacturing (D-WAAM) 
process. The functionally graded material (FGM) can be divided into three distinct regions (R2–R4), each with 
unique microstructural morphologies. Along the building direction, the nickel (Ni) content gradually decreased, 
leading to a reduction in the fraction of γ-austenite and an increase in the martensite start temperature from 
below to above room temperature (Fig.2). The 
compositional variation resulted in the 
transformation of the matrix from γ-austenite 
to α’-martensite, accompanied by an increase in 
dislocation density and KAM values. 
Consequently, the hardness profile showed a 
consistent rise from R4 to R3. In R2, the presence 
of large-sized δ-ferrite is associated with 
significant variability in hardness. 
 
Chapter 4. Influence of compositional variation on the mechanical properties of austenitic stainless steel-
ferritic stainless steel bimetallic component fabricated by dual-wire arc additive manufacturing 
This chapter investigates the mechanical behavior and deformation characteristics of ASS-FSS bimetallic 
components fabricated using the dual-wire arc additive manufacturing (D-WAAM) process. Two types of 
samples were analyzed: directly deposited and compositionally graded. Vertical specimens (VD, VG) exhibited 
microstructure transitions from ductile ASS region to hard interfacial/graded region and soft FSS region, with 
strain concentration consistently localized in the soft FSS region, leading to relatively low strength. Horizontal 
specimens (HD, HG) showed distinct microstructure distributions. In HD, the microstructure ranged from soft 
R2 to hard R3 and ductile R4, with strain concentration initiating at soft R2 and ductile R4 before propagating 
to hard R3. In HG, hard R2 and R3 prevented significant strain concentration, resulting in higher strength but 
lower elongation compared to HD. Layer-by-layer tensile tests revealed a rise-then-fall trend in strength and a 
continuous decline in elongation along the building direction, closely associated with the microstructure 
evolution. 
Chapter 5. Optimization on the microstructure and properties of as-WAAMed austenitic stainless steel-
ferritic stainless steel bimetallic component via post-process heat treatment 
Fig. 2 Microstructure distribution of the FGM region along 
the building direction, 
 
 
This chapter investigates the optimization of FSS-ASS bimetallic components fabricated via dual-wire arc 
additive manufacturing (D-WAAM) through post-process heat treatment (PPHT) at 600°C for 4 hours. The 
PPHT significantly improved the mechanical properties and microstructural uniformity of the functionally 
graded material (FGM) region. Hardness across the FGM became more uniform, with martensite softening 
observed after tempering. Microstructural analysis revealed the transformation of as-quenched α’-martensite 
into tempered martensite, with carbide precipitation in R4, while retained γ-austenite and δ-ferrite were 
preserved. Tensile testing showed enhanced mechanical performance, with ultimate tensile strength (UTS) 
increasing from 852MPa to 915MPa and elongation improving significantly from 20% to 27% for horizontal 
specimens. Vertical specimens 
also exhibited slight 
improvements due to residual 
stress release (Fig.3). Fracture 
morphology analysis confirmed 
pronounced necking and typical 
dimple structures, indicating 
improved toughness PPHT.  
 
Chapter 6 Conclusions 
Chapter 6 summarized the results of each chapter. This study elucidates the microstructural evolution and 
mechanical behavior of ASS-FSS bimetallic structures fabricated via D-WAAM. The interfacial formation 
mechanisms were analyzed, and effective strategies for controlling microstructure and mechanical properties 
were proposed. These findings provide a foundation for optimizing compositional gradients and processing 
parameters, contributing to the advancement of multi-material additive manufacturing. 
 
Fig. 3 The tensile curves and tensile properties of the PPHT specimens 
compared with as-WAAMed specimens. 


Not a member yet