ferroelectric microwave

8/18/2019 Ferroelectric Microwave

1/101

Abstract

Microwave materials have been widely used in a variety of applicationsranging from communication devices to military satellite services, and thestudy of materials properties at microwave frequencies and the developmentof functional microwave materials have always been among the most activeareas in solid-state physics, materials science, and electrical and electronicengineering. In recent years, the increasing requirements for the development

of high speed, high frequency circuits and systems require complete under-standing of the properties of materials function at microwave frequencies.Ferroelectric materials usually have high dielectric constants, and their

dielectric properties are temperature and electric field dependent. The changein permittivity as a function of electric field is the key to a wide range of applications. Ferroelectric materials can be used in fabrication capacitorsfor electronic industry because of their high dielectric constants, and thisis important in the trend toward miniaturization and high functionality of electronic products. The simple tunable passive component based on ferro-electric films is a varactor which can be made as a planar structure, and elec-trically tunable microwave integrated circuits using ferroelectric thin filmscan be developed. Therefore, it is very important to characterize the dielec-tric constant and tunability of ferroelectric thin films.

This thesis shows experimental results for growth, crystalline propertiesand microwave characterization of Na0.5K0.5NbO3 (NKN), AgTa0.5Nb0.5O3(ATN), Ba0.5Sr0.5TiO3 (BST) as well as AgTaO3 (ATO), AgNbO3 (ANO)thin films. The films were grown by Pulsed Laser Deposition (PLD) andrf-magnetron sputtering of a stoichiometric, high density, ceramic NKN,ATN, BST target onto single crystal LaAlO3(LAO), Al2O3 (sapphire), andNd:YAlO3, and amorphous glass substrates. By x-ray diffractometry, NKN,ATN, BST films on LAO substrates were found to grow epitaxially, whereasfilms on r-cut sapphire substrates were found to be preferentially (00l) ori-ented.

Coplanar waveguide interdigital capacitor (CPWIDC) structures werefabricated by standard photolithography processing and metal lift-off tech-nique. Microwave properties of the NKN/Sapphire and ATN/Sapphire withCPW structures were characterized using on-wafer microwave measurementtechnique. Measurement setup is composed of network analyzer, probe sta-


2/101

iv Abstract

tion, and microwave G-S-G probes. External electric field through the con-nection between network analyzer and power supply was applied to measurevoltage tunability. Measured S -parameter were used for the calculation of capacitance, loss tanδ , tunability and K-factor.

The NKN films interdigital capacitors with 2 µm finger gap on Nd:YAlO3

showed superior performance compared to ATN in the microwave range from1 to 40 GHz. Within this range, the voltage tunability (40V, 200 kV/cm)was about 29%, loss tangent ∼ 0.13, K -factor = tunability/tanδ from 152%@ 10GHz to 46% @ 40GHz.

The microwave performance of ATN film CPWIDC with 2 µm finger gapon sapphire substrate in the microwave range from 1 to 40 GHz showed thatfrequency dispersion is about 4.3%, voltage tunability was 4.7% @ 20GHzand 200 kV/cm, loss tangent ∼ 0.068 @ 20GHz, K -factor = tunability/tanδ is ranged from 124% @ 10GHz to 35% @ 40GHz.

The BST films CPWIDC with 2µm finger gap on Al2O3 substrate showedfrequency dispersion of capacitance in the microwave range from 1 to 40 GHz

about 17%, voltage tunability = 1 - C(40V)/C(0) ∼ 22.2%, loss tangent ∼0.137 @ 20GHz, and K -factor = tunability/tanδ from 281% @ 10GHz to95% @ 40GHz.

Key words: ferroelectrics, sodium potassium niobates, silver tantalateniobate, barium strontium titanate, thin films, pulsed laser deposition, rf-magnetron sputtering, coplanar waveguide, photolithography, microwave on-wafer measurement, scattering parameter measurement, tunability.


3/101

Preface

All the work presented in this thesis have been carried out at the divisionof Condensed Matter Physics (KMF), in the Department of Microelectron-ics and Applied Physics (MAP), School of Information and CommunicationTechnology (ICT), Royal Institute of Technology (KTH), in Stockholm, Swe-den. As a Ph.D. Student from June 2002 to September 2005, I had beensupported by the Swedish Foundation for Strategic Research (SSF).

This thesis is based on the following publications and manuscripts.

I. Jang-Yong Kim, Alexander M. Grishin, “Processing and on-wafer test of ferroelectric film microwave varactors” Appl. Phys. Lett., 88, 192905,2006.

II. Jang-Yong Kim, Alexander M. Grishin, “Microwave Properties of ATN(AgTa0.5Nb0.5O3) Thin Film Varactors on Various Substrates” 15th IEEE/ISAF Meeing. 2006; To appear in IEEE ISAF’06 proceedings.

III. S. Bonetti, J.-Y. Kim, S.I. Khartsev, A.M. Grishin, “Buried Tantalate-Niobate Microwave Varactors” 15th IEEE/ISAF Meeting. 2006; Toappear in IEEE ISAF’06 proceedings.

IV. Jang-Yong Kim, Alexander M. Grishin, “ AgTaO3 and AgNbO3 ThinFilms by Pulsed Laser Deposition” Thin Solid Films. Vol. 515, 2, pp615-618, 2006.

V. Jang-Yong Kim, Alexander M. Grishin, “Niobate-tantalate thin filmsmicrowave varactors” Thin Solid Films. Vol. 515, 2, pp 619-622, 2006.

VI. Jang-Yong Kim, Alexander M. Grishin, “ AgTa0.5Nb0.5O3 Thin FilmCoplanar Waveguide Microwave Capacitors” Integrated Ferroelectrics,vol. 77, pp 13-20, 2005.

VII. Alexander M. Grishin, Jang-Yong Kim, Sergey I. Khartsev, “Process-

ing and on-wafer measurements of ferroelectric interdigitated tunablemicrowave capacitors” Mat. Res. Soc. Proc. Vol. 811, D10.1.1, 2004.

VIII. Jang-Yong Kim, Alexander M. Grishin, “ Na0.5K0.5NbO3 Film Mi-crowave Varactors” Integrated Ferroelectrics , vol. 66, pp. 291-300,2004.


4/101

vi Preface

The following paper has been published, but is not included in this thesis.

IX. Jang-Yong Kim, Jung-Hyuk Koh, Jae-Sung Song, Alexander M. Gr-ishin, “Magnetically and electrically tunable devices using ferromag-netic / ferroelectric ceramics” Phys. Stat. Sol. B vol. 241, pp. 1714-

1717, 2004.

My supervisor, Prof. Alex Grishin, and our senior scientist, Dr. SergeyKhartsev, both throughout this thesis work have been involved in experi-mental and theoretical discussions. The ceramic targets and thin films of AgTaO3, AgNbO3, AgTa0.5Nb0.5O3 and Ba0.5Sr0.5TiO3 have been preparedby myself. The Na0.5K0.5NbO3 ceramic target and thin films were preparedby Dr. Sergey Khartsev. All the clean room processing have been done bymy own. I made all the microwave measurements and parameter calcula-tions with the technical support from Dr. Gunnar Malm. I wrote all themanuscripts.

The results in papers II and III have presented at 15th International Sym-posium on the Applications of Ferroelectrics (ISAF2006), NCSU, USA, July,2006. Papers IV and V were presented at the 13th International Congresson Thin Films, (ICTF 2005) Stockholm, Sweden, Jun 2005. Paper VI waspresented at the 17th International Symposium on Integrated Ferroelectrics(ISIF 2005) in Shanghai, China, April 2005. Paper VIII was presented atthe 16th International Symposium on Integrated Ferroelectrics (ISIF 2004)in Gyeongju, South Korea, April 2004.


5/101

Acknowledgement

First, I would like to thank my supervisor, Prof. Alex Grishin, for givingme opportunity to study and live in Sweden, and for his strong supportin improvement of my academical horizons and experimental proficiency.Professor encouraged me in my studies and inspired my self-confidence duringmy study. I also would like to thank Prof. Byung-Moo Moon, my formersupervisor of master course in Korea University, for his recommendation andgenerous mediation of studying in Sweden.

I would like to acknowledge some persons who have spent so many dayswith me in our division of Condensed Matter Physics: Dr. Sergey Khartsevfor his great advices and teaching me on target processing and thin filmdeposition, Dr. Sören Kahl for his interesting and fruitful discussions on PLDtechnique and friendship, Dr. Jung-Hyuk Koh, my senior at this work, forhis introducing me to electrical characterization and lithography, Dr. MatsBlomqvist, for all discussion on x-ray diffraction, lithography processing aswell as the life in Sweden, Dr. Gunnar Malm, researcher at EKT, for histeaching and advice on microwave on-wafer test.

Also thanks to all former and present members at KMF and Korean

members at EKT who have contributed to my study and life: Dr. AlvydasLisauskas, Dr. Jürgen Brünahl, Dr. Vasyl Denysenkov, Dr. Peter Johnsson,Dr. Akira Shibuya, Dr. Shin Ichi Aoqui, Kazufumi Omori, Dr. ZhigangZhang, Dr. Joo-Hyung Kim, Ramos Mays, Rickard Fors, Petra V. Johansson,Stefano Bonetti, Matteo Savoini, Dzmitry Dzibrou, Dr. Sang-Kwon Lee, Dr.Sang-Mo Koo, Hyung-Seok Lee.

I wish to acknowledge Prof. Mikael Östling, Dean of School of Infor-mation and Communication Technology, for his friendly support of officialwork. I also acknowledge Zandra Lundberg, secretary of MAP, for her kindhelp of practical things.

Personally, I am very grateful to Dr. Sang-Ho Yun, Research Associate of Material Physics in KTH, for his professional advice of the academic sourceand my view of life.

I very much appreciate Dr. Kun-Hwa Park, who was a chief of KarolinskaUniversitetssjukhuset, for his invisible support and care.


6/101

viii Acknowledgement

I also appreciate Dr. Julia Hahn, who is headmaster of Koreanska SkolanStockholm for giving me opportunities to teach Korean language for Swedishadults and adopted Korean people. I am really thanks to my young & oldstudents in Koreanska Skolan Stockholm.

I would like to thanks to Choong-Il Cho, who is pastor of Immanuel-

skyrkan Stockholm Korean fellowship, for giving me religious guidance andspiritual awakening. I also would like to thank my lovely secondary & highschool students in Immanuelskyrkan Stockholm and all the members of Ko-rean Students and researchers Association in Sweden (KOSAS).

I sincerely would like to express my gratitude to Jung-Min Lee, Direc-tor of Korea Trade-Investment Promotion Agency (KOTRA) in Stockholm,for giving me a new position and experience of new area as a trade andinvestment consultant.

Many thanks to my new & close business partner, Jonas Hubbe, Manag-ing Director and Björn Linder, Customer Relations of Sinterteknik AB, Kelly

Rao, Business Development and Marketing Manager of Elfa Group AB, In-gele Hallström, Managing Director of Sensei Scandinavia, Kristina Wiik andLars Forsen, the owner of Alternativ Föradling AB, Jan Kinsten of JanecoKonsult & Eva Trade AB, Tony Kvarnström, Managing Director of AuroraIT Systems AB, Fredrik Branberg, Treatment Manager and Ulrik Alexan-dersson, Customer Responsible of SITA Sverige AB, Daniel Broman, Headof Marketing of Smart Microfiber System AB, Jarmo Penttinen, Project En-gineer of SKANSKA AB, Stefan Erhag and Annelie Lindstedt, Advocate of Delphi & Co, Sigvard Beck-Friis, Head of Establishment Services and RealEstate Project and Dr. Ciro Vasquez, Business Development Manager(ICT& IVSS) of Invest in Sweden Agency,

Finally, I sincerely would like to express my deepest gratitude to my affec-tionate father Ki-Bong Kim, who was the Senior Committeeman of AuditingSupervisory Committee in Korean Institute of Certified Public Accountants(KICPA), for his continuous assistance and everlasting love. I also wouldlike to emphasize gratitude to my mother Soon-Ja Choi for her unboundedaffection. I also thank my dear wife Mi-Jung Lee for her durable and moralsupport throughout the study.

Especially, I sincerely dedicate this thesis to my beloved son, Joon-Suh Kim , and new baby in mother’s womb, for supplying me permanent andincredible energy to accomplish my duties.

Jang-Yong Kim

Stockholm, SwedenNovember, 2006


7/101

Contents

Abstract iii

Preface v

Acknowledgement vii

1 Introduction 1

2 Dielectric Materials for Microwaves 5

2.1 The Function of Dielectrics . . . . . . . . . . . . . . . . . . . 52.2 Frequency Dispersion of Dielectrics . . . . . . . . . . . . . . . 62.3 Ferroelectric Materials . . . . . . . . . . . . . . . . . . . . . . 8

2.3.1 Ferroelectricity . . . . . . . . . . . . . . . . . . . . . . 82.3.2 Voltage Tunability . . . . . . . . . . . . . . . . . . . . 9

2.4 Ferroelectric Materials for Microwaves . . . . . . . . . . . . . 102.4.1 Silver Tantalum Niobium Oxide . . . . . . . . . . . . . 10

2.4.2 Sodium Potassium Niobium Oxide . . . . . . . . . . . 122.4.3 Barium Strontium Titanium Oxide . . . . . . . . . . . 14

3 Material processing & properties, Device design & fabrication 15

3.1 Ceramic Target Processing . . . . . . . . . . . . . . . . . . . . 153.2 Thin Film Growth Technique . . . . . . . . . . . . . . . . . . 16

3.2.1 Pulsed Laser Deposition . . . . . . . . . . . . . . . . . 163.2.2 RF Sputtering . . . . . . . . . . . . . . . . . . . . . . 19

3.3 Films Crystalline Properties using X-ray Diffraction . . . . . 203.3.1 X-ray Diffraction . . . . . . . . . . . . . . . . . . . . . 20

3.3.2 θ − 2θ scan . . . . . . . . . . . . . . . . . . . . . . . . 213.3.3 ω-scan (Rocking curve) . . . . . . . . . . . . . . . . . 223.3.4 φ-scan . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.4 Photolithography and Lift-off Technique . . . . . . . . . . . . 253.5 Coplanar Waveguide . . . . . . . . . . . . . . . . . . . . . . . 27


8/101

x CONTENTS

4 Microwave On-wafer Characterization for Ferroelectrics 31

4.1 Microwave Network . . . . . . . . . . . . . . . . . . . . . . . . 324.1.1 Concept of Microwave Network . . . . . . . . . . . . . 324.1.2 Impedance and Admittance Matrix . . . . . . . . . . . 324.1.3 Scattering Parameters . . . . . . . . . . . . . . . . . . 33

4.2 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344.2.1 SOLT calibration . . . . . . . . . . . . . . . . . . . . . 344.2.2 TRL calibration . . . . . . . . . . . . . . . . . . . . . 354.2.3 LRM calibration . . . . . . . . . . . . . . . . . . . . . 36

4.3 De-embedding Analysis . . . . . . . . . . . . . . . . . . . . . 364.4 Experimental results and Discussion . . . . . . . . . . . . . . 384.5 New design of Buried Interdigital Capacitors . . . . . . . . . . 41

4.5.1 Device Fabrication . . . . . . . . . . . . . . . . . . . . 414.5.2 Measurement results . . . . . . . . . . . . . . . . . . . 424.5.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . 43

5 Conclusion and future work 45

Bibliography 47

Appendix I - Schematic of Calibration Substrate 53

Appendix II - Equation of S, Y, Z, ABCD parameter 55

Appendix III - Practical guide for PLD processing 57

Appendix IV - Quick guide for On-wafer RF Measurement 63

Abbreviations 75

Papers 77


9/101

Chapter 1

Introduction

Microwave communication has turned out to be one of the most rapidlygrowing fields of our information society. Wireless communication based onindividual handsets has revealed breathtaking growth rates with a dramatic

increase of transferred date rates and number of subscribers. In particu-lar, wireless multimedia applications are considered to enforce a dramaticincrease of wireless system capability in the near future.

In contrast to optical communication links, where the data rate capabil-ity appears to be as infinite from the point of view of physical limitations,microwave communication systems are intrinsically limited by the availabil-ity of bandwidth. Therefore, modern systems have to utilize the availablebandwidth most efficiently. In addition, the cost issue is rather important,for example due to strong competition between different service providers inmobile communication.

Current system specifications including those for the 3rd generation mo-

bile telephone systems are based on the available performance of the currentdevices and subsystems. However, novel devices with improved performanceare likely to increase the capacity of the systems or may become even anecessity in order to operate a system properly under any circumstances.Therefore, any new material and device development leading to a perfor-mance increase or/and to a cost reduction is worth to be investigated.

Ferroelectric materials usually have high dielectric constants, and theirdielectric properties are temperature and electric field dependent. For manyyears, although ferroelectric materials are used in a wide variety of devices,mainly the pyroelectric and piezoelectric properties of ferroelectrics were

utilized. This situation has been changed during the past few years.The application of ferroelectric materials in random access memories is

expected to replace magnetic core memories and magnetic bubble memo-ries. Besides, ferroelectric materials can be used in fabricating capacitorsfor electric industry because of their high dielectric constants, and this is


10/101

2 Introduction

important in the trend toward miniaturization and high functionalities of electronic products. Furthermore, ferroelectric materials can be used for thedevelopment of tunable microwave devices, which have potential in satel-lite, terrestrial communications, and other microwave applications where theworking frequencies are higher than the useful range of si-based devices [1, 2].

One of the recent ferroelectric technology has been focused on Ferroelec-tric Reflectarray Antenna (FRA) researched by NASA. Ferroelectric materialhas overcome some drawbacks of MMIC-based phase shifters such as highersignal loss, only limited power handling, and expensive to manufacture.

The ferroelectric reflectarray antenna is competitive with gimbaled par-abolic dish antennas while eliminating the mechanics and vibration. It alsoprovides the electronic beam steering of an MMIC direct-radiating arraywhile being more efficient and less expensive to produce.

NASA’s prototype FRA consists of 616 microstrip antenna radiators in-tegrated with 616 ferroelectric phase shifters on quartz and lanthanum alu-

minate substrates, respectively. Each phase shifter can be independentlycontrolled from a PC, using the same algorithm that conventional phase-array antennas use. The entire array of 616 elements can be updated in mil-liseconds. The FRA provides beam steering of greater than plus-or-minus40 degrees, and throughput of more than 100 Mbits/sec.

The FRA’s phase shifters are made of ferroelectric materials in this case,barium strontium titanate. Although they contain no iron, they offer manybenefits of ferrite materials, such as high-resolution scanning, while steeringbeams faster and consuming less power.

These ferroelectric materials are dielectric (non-conducting), so the phaseshifters draw virtually no current. Only the FRA’s single amplifier and PCcontroller draw power. The power from the amplifier is reflected across allelements required to form a particular beam.

The FRA phase shifters also incur less signal loss than MMIC phaseshifters-5 dB for the FRA vs. 8 dB for the MMIC. The higher the signalloss per phase shifter, the less power the antenna can radiate. This lossultimately limits array sensitivity and efficiency. Thus, the FRA’s low-lossphase shifters support larger apertures with a lower signal loss.

The FRA also costs about 90 percent less to produce. Since its smallestcomponents are over ten times larger than those of the MMIC-based an-tenna, they can be lithographically printed using simple design rules. TheFRA antenna elements also have fewer layers, and are therefore easier toconstruct. As a result, NASA’s patented 19GHz ferroelectric reflectarray

antenna is clearly a breakthrough in low-cost space communications.

By using ferroelectric thin films, electrically tunable microwave inte-grated circuits can be developed. Therefore, it is very important to char-acterize the dielectric constant and tunability of ferroelectric thin films at


11/101

3

microwave frequencies [3, 4].Novel tantalate-niobate oxide ferroelectrics promise enhanced microwave

properties because of previous researches as follows.NaxK1−xNbO3 (NKN) thin films on various oxide substrate possess ex-

cellent dielectric and piezoelectric characteristics therefore they have been in-

tensively studied by Cho for the varactor, non-volatile memory and acousto-electric device applications [25, 26, 27].

AgTaxNb1−xO3 thin films also have been studied by Koh for varactorapplications [16, 17] because ATN ceramics shows low loss tangent, highdielectric permittivity and very flat dielectric response in a wide frequencyrange from RF to 100GHz [9]. But all the previous researches in film sta-tus were mainly focused on low frequency dielectric properties up to 1 Mhzand intrinsic microwave NKN and ATN properties were not thorough by ex-plored. Furthermore, the study of AgNbO3 (ANO), AgTaO3 (ATO) seem tobe fundamentals to understand the features of the solid solution ATN system.

In this thesis, I would like to introduce mainly niobate-tantalate filmswhich can be good candidate for microwave devices. I also would like to in-troduce Ba1−xSrxTiO3 (BST) films, which already have been widely studied,for the comparison. Brief description of other chapters are as follows:

Chapter 2 introduce general overview of dielectric and ferroelectric materi-als specially for microwave applications. Brief concept of ferroelectricsas well as dielectrics, frequency dispersion of dielectric response andnovel tantalate-niobate ferroelectric materials are described in thischapter.

Chapter 3 explains whole device processing, e.g. target fabrication, thin

film deposition processing, lithography and metal lift-off processing.Material growth properties of thin films and a feature of CoplanarWaveguide structure are also demonstrated.

Chapter 4 shows microwave network analysis for on-wafer structure. Aconcept of microwave network, calibration technique, on-wafer mea-surement procedure and de-embedding analysis are explained as wellas experimental data.

Chapter 5 Summary of experimental results, discussions and future work.


12/101


13/101

Chapter 2

Dielectric Materials for

Microwaves

Ferroelectric materials have been widely used in various applications. High

dielectric coefficients over a wide temperature and frequency range are usedas dielectrics in integrated or SMD capacitors. The large piezoelectric effectis applied in a variety of electromechanical sensors, actuators and transduc-ers. Infrared sensors need a high pyroelectric coefficient which is availablewith this class of material. The significant non-linearities in electromechan-ical behavior, field tunable permittivities and refractive indices, and elec-trostrictive effects open up a broad area of more different application. In thischapter, a short overview of dielectric and ferroelectric materials and theirproperties. Especially for microwave applications, novel tantalate-niobate aswell as barium strontium titanate ferroelectric materials are described.

2.1 The Function of Dielectrics

Dielectrics are insulating materials that are used technically because of theirproperty of electrical polarization to modify the dielectric function of thevacuum, e.g. to increase the capacity of storing the electric charge. Inmaterials, electric field is presented by electrical displacement

D = ε0(1 + χe)E = ε0εrE

which is the sum of external electric field E (vacuum contribution) and po-larization

P = ε0χeE

which is characterizes the number of electrical dipoles in 1 cm3. Here ε0 isdielectric permittivity in vacuum, χe is electrical susceptibility, εr is relative


14/101

6 Dielectric Materials for Microwaves

Figure 2.1: Frequency dependence of permittivity for a hypothetical dielec-

tric [5].

permittivity

If χe or εr themselves are field-dependent, e.g. become reduced for highelectric fields, tunable dielectrics are achieved.

2.2 Frequency Dispersion of Dielectrics

The ability of a dielectric material to store electric energy under the influenceof an electric field results from the field-induced separation and alignment of

electric charges. Polarization occurs when the field causes a separation of thepositive and negative charges in the material. The larger the dipole momentarm of this charge separation in the direction of a field and the larger thenumber of these dipoles, the higher the material’s dielectric permittivity.Typical behavior of permittivity as a function of frequency is show in figure2.1. Moving charge causes a frequency dependent phase shift between appliedfield and charge displacement. To express this mathematically, the relativedielectric permittivity is written as a complex value:

εr = ε

r + iε

r .

The loss tangent is defined as

tan δ = ε

r

εr.

The permittivity of material is related to a variety of physical phenom-ena that contribute to the polarization of a dielectric material. In the low


15/101

2.2 Frequency Dispersion of Dielectrics 7

Figure 2.2: Electrical polarization of dielectric. (a) Polarization of lineardielectric, (b) typical hysteresis loop for ferroelectric material, (c) Nonlineardielectric [6].

frequency range, ε

is dominated by the influence of ion conductivity. Thevariation of permittivity in microwave range is mainly caused by dipolar re-laxation, and the absorption peaks in the infrared region and above is mainlydue to atomic and electronic polarizations.

• Electronic polarization occurs in neutral atoms when an electricfield displaces the nucleus with respect to the surrounding electrons.This induced dipole effect occurs in all materials, including air, but isusually very small compared to other polarization mechanisms sincethe moment arms of these dipoles are very short, usually a fraction of the size of an atom.

• Atomic polarization occurs when adjacent positive and negative ionsstretch under an applied electric field.

• Dipolar polarization Dipolar polarization is due to the dipole mo-ment arising from different electronegativity of atoms in a molecule. If


16/101


the electric field oscillates slowly, the time taken by the electric field tochange direction is longer than the response time of the dipole, there-fore the dielectric polarization keeps in phase with the electric field.

• Ionic polarization is similar to atomic polarization but involves the

shifting of ionic species under the influence of the field. This shift canbe considerable and can lead to very high dielectric constant, up toseveral thousand.

All materials exhibit electronic polarization because they all have atoms,whereas atomic and ionic polarization require specific types of structuresto be present. Purely electronic polarization would result in low dielectricconstants, perhaps up to 2-4. Atomic and especially, ionic polarization areresponsible for much larger dielectric constants.

2.3 Ferroelectric Materials

2.3.1 Ferroelectricity

As shown in figure 2.2, compared to nonlinear dielectrics, ferroelectric ma-terials display a hysteresis effect of polarization with an applied field. Thehysteresis loop is caused by the existence of permanent electric dipoles in thematerial. When the external electric field is initially increased from zero, thepolarization increases as more of the dipoles are lined up. When the field isstrong enough, all dipoles are lined up with the field, so the material is ina saturation state. Until this stage non-linear dielectrics and ferroelectricsbehave similarly. Then, when the applied electric field decreases from thesaturation point, the polarization also decreases. However, when the exter-

nal electric field reached zero, the polarization in ferroelctrics does not reachzero. Ferroelectric material remains to be electrically polarized in the ab-sence of external electric field. The polarization at zero field is called theremanent polarization (Pr). When the direction of the electric field is re-versed, the polarization decreases. When the reverse field reaches a certainvalue, called the coercive field , the polarization becomes zero. By furtherincreasing the field in this reverse direction, the reverse saturation can bereached. When the field is decreased from the saturation point, the sequence

just reverses itself.For ferroelectric material, there exists a particular temperature called

the Curie temperature (Tc). Ferroelectricity can be maintained only below

the Curie temperature as shown in figure 2.3 (a). When the temperature ishigher than Tc, a ferroelectric material is in its paraelectric state. Dielectricpermittivity drastically increases in the vicinity of Curie temperature.

ε = ε0

1 +

C

T − T c


17/101

2.3 Ferroelectric Materials 9

Figure 2.3: (a) Schematical plots of temperature dependent polarizationP(T) and polarizability χ(T). (b) Schematical plots of electric field depen-

dent hysteresis of polarization P(E) and polarizability χ(E).

Here C is the curie constant.Ferroelectric materials have great application potential in developing

smart electromagnetic materials, structures, and devices, including minia-ture capacitors, electrically tunable capacitors, filters and phase shifters inrecent years. Microwave ferroelectrics are still under intensive investigation.

2.3.2 Voltage Tunability

The dielectric permittivity ε of ferroelectric materials can be presented as aderivative of the field dependent polarization as shown in figure 2.3 (b).

ε = 1 + dP

dE

Therefore, ε can be varied by applying a dc electric field, and this propertycan be used in the development of electrically tunable electric devices. Thechange of ε because of the external electric field is often described by thetunability, as defined below:

Tunability[%] = εr0 − εre

εr0× 100

Where εr0 is the dielectric permittivity of the ferroelectric material with-out external dc electric field, and εre is the dielectric permittivity when thematerial is biased by an external dc electric field.


18/101


The dielectric loss of a ferroelectric material is also dependent on the ap-plied dc electric field. Generally speaking, the dielectric loss under appliedexternal dc electric field. Experiment show that a ferroelectric material withhigher loss tangent usually has larger tunability. As loss tangent of a ma-terial is an important factor affecting the performances of electric circuit,

in the development of electrically tunable ferroelectric microwave devices, afigure of merit K (K -factor), defined by K=Tunability/tanδ , is often used toindicate the quality of ferroelectric material. Usually, in the calculation of K , the loss tangent at zero external dc electric field is used.

2.4 Ferroelectric Materials for Microwaves

Thin film ferroelectrics are receiving increased attention because of their po-tential in producing tunable RF and microwave circuits. Tunable circuits,including filters, matching networks and phase shifters offer the flexibilityto adapt to changes in operating conditions, such as frequency, impedance

environment or RF drive level, a property that is highly sought after in thewireless communications industry. Components using alternative technolo-gies, such as GaAs varactor diodes, tend to have low Q values, implying highRF loss, and generally do not handle high RF power levels. MEMS-baseddevices have unproven reliability at this point, often have difficult biasingrequirements and have stringent packaging needs. This development encour-ages the investigations of electronic materials, in particular the developmentof new dielectric materials with specific dielectric properties that are de-manded by electronics engineers. Incidentally, ferroelectrics also have theproperty of piezoelectricity (stress induced polarization) and pyroelectricity(temperature induced polarization), but those properties are not important

to this discussion. Table 2.1 presents properties of the most known ferro-electric materials together with novel tantalate-noibate ferroelectrics whichare the main object of our study.

2.4.1 Silver Tantalum Niobium Oxide

Ag(Ta,Nb)O3, perovskite solid solution of AgNbO3 and AgTaO3, has beenintensively studied in ceramic bulk form [10], [11] due to its excellent mi-crowave properties. But AgTaxNb1−xO3 has not been widely investigatedin thin film form. One of the reasons is considerable difficulty to fabricateAgTaxNb1−xO3 ceramics of good quality, due to the high Ag volatility thatcauses the decomposition of Ag-based compounds at high temperatures. In

ceramic bulk form, AgTaxNb1−xO3 shows much less than 3 × 10−3 for losstangent and higher than 200 for dielectric permittivity from 100 K to 900 Kin a wide temperature range. AgTaxNb1−xO3 ceramics show an interestingdependence of its physical properties such as a sequence of phase transitionsand anomalies of dielectric permittivity as a function of x (refer to figure


19/101

2.4 Ferroelectric Materials for Microwaves 11

Table 2.1: Selection of ferroelectric materials.[12, 13, 15, 23]

Material Chemical formula Tc[◦C ] εr Pr [µC/cm2]

Barium titanate BaTiO3 135 500 26Lead titanate PbTiO3 490 17 50

Lead zirconate ti-tanate

Pb(Zr0.52Ti0.48)O3 350 26 10

Lithium niobate LiNbO3 1200 15 71

Barium strontium ti-tanate

Ba1−xSrxTiO3 350 600 4

Sodium PotassiumNiobate

NaxK1−xNbO3 400 500 12

Silver Tantalum Nio-bate

Ag(Ta, Nb)O3 370 400 0.1

Figure 2.4: The temperature dependence of the dielectric permittivity of various Ag(Ta, Nb)O3 ceramics at a frequency of 1 MHz.


20/101


2.4). Therefore as decreasing the Ta composition, phase transition tempera-ture between ferro (M1)-antiferro (M2) moves to the room temperature anddielectric permittivity increases. (Refer to figure 2.4, 2.5)

With regard to particular microwave applications, a new microwave ma-terial that is based on a perovskite-type AgTaxNb1−xO3 (ATN) solid solution

i.e., the compounds AgNbO3 and AgTaO3 meets the majority of commer-cial requirements, especially because of its high permittivity and moderatedielectric losses.

Dielectric studies of the AgTaxNb1−xO3 have shown weak indications of ferroelectricity at temperatures < 348 K [7] and < 180 K [9], respectively.High static permittivity (εr0) of the AgNbO3 originates from the high fre-quency relaxation motion of niobium ions [8]. Dielectric properties of thesolid solutions and their dependence on temperature and composition werestudied by Kania [10]. He revealed that the permittivity at room tempera-ture decreases as the tantalum content increases and that the anomalies in

the temperature dependence of permittivity can be associated with phasetransitions. The existence of a low-temperature ferroelectric phase has al-ready been confirmed for a AgTaxNb1−xO3 solid solution with x < 0.8 [10].

Volkov et al.[9] made important contributions to the understanding of AgTaxNb1−xO3 solid solutions by performing dielectric spectroscopy on thisbulk ceramic solid solution with different compositions. In regard to theapplication of AgTaxNb1−xO3 as a microwave material, the main conclusionfrom their work is the comprehension regarding the absence of any dielectricdispersion in the microwave-frequency range (up to 30 GHz). Lithium-dopedAg(Nb,Ta)O3 ceramic materials also has been investigated by Sakabe etal.[14] for the improvement of ferroelectric response and dielectric properties.

In case of silver tantalate niobium films, there were only few reports[15, 16, 17, 20]. ATN films deposited on various substrates have shown gooddielectric properties: εr ∼ 300, low loss tan δ < 0.004 and tunability ∼ 6 %at 1 MHz and 100 kV/cm. However, the applicability of this material for mi-crowave tunable devices remains questionable due to the lack of experimentalresults of dielectric properties in microwave frequency ranges.

2.4.2 Sodium Potassium Niobium Oxide

Ferroelectric sodium potassium niobium oxide, NaxK1−xNbO3 (NKN), is thecontinuous solid solution of KNbO3 and NaNbO3, and is perovskite struc-tured for x < 0.97. Bulk ceramic NKN exhibits a complicated phase diagram

with several structural phase transitions [29]. The dielectric properties andphase transitions of NaxK1−xNbO3 as well as KNbO3 and NaNbO3 ceramicswere reported few decades ago [21, 22].

NKN films on various oxide substrate fabricated by pulsed laser depo-sition [23, 24] and RF magnetron sputtering [30] showed preferential c-axis


21/101

2.4 Ferroelectric Materials for Microwaves 13

Figure 2.5: Phase diagram for the Ag(Ta,Nb)O3 solid system. The

boundaries between the different structured phases are shown: C-cubic, T-tetragonal, O-orthorhombic, R-rhombohedral, M1-ferro, M2-antiferro, M3-orthorhombic symmetry in rhombic orientation, M4-monoclinic distortionsof the pseudoperovskite unit cell


22/101


orientation with self assembling phenomena, high dielectric permittivity andhigh tunability. Because of its excellent crystallinity and electric properties,NKN films have been also intensively studied for varactors [25], memory de-vice [26], acoustoelectric device [27] and microwave device applications [28].But intrinsic microwave NKN properties were not thorough by explored.

2.4.3 Barium Strontium Titanium Oxide

BST, with its chemical formula of Ba1−xSrxTiO3, can be thought of as a solidsolution of barium titanate, BaTiO3, and strontium titanate, SrTiO3. Byadjusting the Ba/Sr ratio and other film growth parameters, precise controlcan be exercised over the dielectric constant, tunability, and quality factorQ of the resulting capacitors.

Most ferroelectrics including BST will, when the temperature reachesthe Curie temperature, Tc, undergo a phase transition from the ferroelectricto the paraelectric state. This is an important property. Below the Curietemperature, the material is in the ferroelectric phase, its lattice structureis tetragonal and its polarization response to an applied electric field is hys-teretic. When the Curie temperature is reached the material undergoes atransition to the paraelectric phase, the crystal lattice changes from tetrag-onal to cubic and it displays the highly desirable non-hysteretic electric fieldresponse. Another important phenomenon related to this phase change is astrong temperature dependence of the relative permittivity in the neighbor-hood of Tc.

In thin BST films, the material remains in the paraelectric state, its po-larization response is non-hysteretic and the temperature dependence of εris greatly suppressed. The BST films dielectric constant is generally in the200 to 300 range, yielding high capacitance densities and therefore relatively

small-sized capacitors. Film thicknesses in this range are also compatiblewith standard thin film processing, meaning that integration of BST capaci-tors with other thin film components is possible. For all these reasons, muchemphasis has been placed on thin barium strontium titanate, Ba1−xSrxTiO3,films deposited by various deposition techniques and such issues as composi-tional grading [31, 32], oxygen vacancies [33], strain and stress effects [34, 35],composites [36], film thickness and grain size effects[37], doping effect [38, 39],annealing [40], Ba/Sr composition effect [41], buffer layers [35], electricaldegradation [42] including breakdown voltage, power handling [43] and com-ponent integration [44] and orientational dependence [45]. The research forapplication device, e.g. filter [46], oscillator [47], capacitive switches [48],

phase shifter [49] also have been progressed actively.Although a lot of research reports for BST have been published, we fab-ricated our own BST ceramic target for film deposition and grown our ownBST films for more accurate comparison between BST and niobates mi-crowave performance.


23/101

Chapter 3

Material processing and

properties,

Device design and fabrication

The deposition of thin film functional layers on different substrates is anessential step in many area of modern high technology. Considering broadspectrum of applications it is obvious that there cannot be only one perfectdeposition method which can be applied in all field. In this chapter, we arefocusing on PLD and RF sputtering for thin film deposition using ceramictarget which is produced by ourselves. Device structures fabrication is de-scribed by optical lithography and metal lift-off processing. Material growthproperties of thin films and a feature of Coplanar Waveguide structure arealso demonstrated.

Figure 3.1: An illustration of Ceramic Target processing

3.1 Ceramic Target Processing

Ceramic targets for thin film deposition have been prepared by standardsintering solid state reaction method using basic oxide materials, such as


24/101

16 Material processing & properties, Device design & fabrication

Table 3.1: Sintering Conditions of Ceramic Target

Parameter AgTaO3 AgNbO3 BaSrTiO3

1st Sintering

(Calcination for BST) 900 ◦C/8h 900 ◦C/4h 1100 ◦C/2h

2nd Sintering 1030 ◦C/3h 1030 ◦C/3h 1200 ◦C/6min.

3rd Sintering 1100 ◦C/3h 1060 ◦C/3h 1400 ◦C/18min.

Ag2O, Ta2O5, and Nb2O5. Pure powder have been ground and sinteredthree times totally to make fine crystalline structure. Cold isostatic pressure(CIP) of 3kbar for 5min was employed to the sample in the hydrostaticconditions. Duration and temperature of target sintering are presented in

table 3.1. Crystalline structure was studied using X-ray diffraction techniqueas shown in figure 3.7.

3.2 Thin Film Growth Technique

3.2.1 Pulsed Laser Deposition

The Pulsed Laser Deposition (PLD) method has been widely used to makeepitaxial films of various oxide materials. Although PLD technique is unableto make uniform film coating on large scale wafers and for mass production,it still have appeal to many researchers for the fabrication of high-qualityfilms of complex oxide materials.

In general, the method of pulsed laser deposition is simple. Only fewparameters need to be controlled during the process. Targets used in PLDare small compared with other targets used in sputtering techniques. It isquite easy to produce multi-layer film composed of two or more materials.Besides, by controlling the number of pulses, a fine control of film thicknesscan be achieved. Thus a fast response in exploiting new material systemis a unique feature of PLD among other deposition methods. The mostimportant feature of PLD is that the stoichiometry of the target can beretained in the deposited films. This is the result of an extremely highheating rate of the target surface due to pulsed laser irradiation. It leads to

the congruent evaporation of the target irrespective to the evaporating pointof the constituent elements or compounds of the target. And because of the high heating rate of the ablated materials, laser deposition of crystallinefilm demands a much lower temperature than other mentioned film growthtechniques. For this reason the semiconductor and the underlying integrated


25/101

3.2 Thin Film Growth Technique 17

Figure 3.2: A Schematic of Pulsed Laser Deposition system

Figure 3.3: A picture of Pulsed Laser Deposition system


26/101


circuit can refrain from thermal degradation.The advantages of PLD technique are as followings [50]:

• PLD can be useful to deposit pure elements and multi-component com-pounds.

• The stoichiometry of the target material can be reproduced in the film.

• Deposition rate is relatively high compared with other physical depo-sition methods.

• The oxide films can be annealed in-situ directly after deposition.

• The system setup is rather simple and more economical than otherdeposition technique.

• The instantaneous deposition flux can be tremendous and can be var-ied (e.g. target-substrate distance; laser fluency; target temperature)

independently of either the average growth rate (laser repetition rate)or the kinetic energy of the ablated species (ambient gas mass andpressure; laser fluency).

In our experiment with ATN, a KrF excimer laser (Lambda Physik-300 )at a wavelength of 248 nm and having a pulse duration of 25 ns was used asthe irradiation source, and the pulse energy at the rotating target was set400 mJ at 15 Hz repetition rate of laser pulse in 20 minutes. The depositionof thin film was performed in the vacuum chamber with 450 mTorr oxygenatmosphere at a substrate temperature of 650 ◦C, followed by an anneal atthe same temperature in 500 Torr oxygen for 30 min. The deposition rate

was about 20 nm/min. A brief description of the processing parameter is intable 3.2 and practical guide for PLD processing is described in AppendixIII.

Table 3.2: Film Deposition Conditions by PLD

Process NaKNbO3 AgTaNbO3 BaSrTiO3

Vacuum (Torr) 5×10−7 5×10−7 5×10−7

Substrate Temperature ( ◦C) 650 650 700

Energy of Laser (mJ/cm2

) 350 400 250Oxygen pressure (mTorr) 400 450 10

Repetition rate (Hz) 25 15 15


27/101

3.2 Thin Film Growth Technique 19

Figure 3.4: Schematic of the deposition chamber and of the off-axis RFmagnetron sputtering system

3.2.2 RF Sputtering

In its simplest representation, the phenomenon of sputtering consists of ma-terial erosion from a target on an atomic scale, and the formation of a thinlayer of the extracted material on a suitable substrate. The process is ini-tiated in a glow discharge produced in a vacuum chamber under pressure-controlled gas flow. Target erosion occurs due to energetic particle bom-bardment by either reactive or non-reactive ions produced in the discharge.

The off-axis radio frequency magnetron sputtering technique consists of a target, which is a plate of a stoichiometric mixture of the material to grow,and of a substrate placed on a grounded sample holder positioned at 90◦ of the target off-axis configuration. The glow discharge is initiated by applyingpower to the target in a controlled gas atmosphere, and is constituted of a partially ionized gas of ions, electrons, and neutral species. The ejectedmaterial diffuses until it reaches and nucleates on the substrate. The durationof this process controls the thin film thickness. The crystalline growth of thinfilms on single crystal substrates, with a well defined orientation, definesepitaxy .

The use of a radio frequency (RF) generator is essential to maintain thedischarge and to avoid charge build-up when sputtering insulating materialssuch as PZT. The presence of a matching network between the rf generatorand the target is necessary in order to optimize the power dissipation in thedischarge. Magnets are used to enhanced the sputtering rate, by increasing


28/101


Figure 3.5: A picture of RF magnetron sputtering system; Leybold-HeraeusDeposition System

the ionizing effect of electrons magnetically trapped in the vicinity of thetarget (magnetron sputtering). Their use provides the advantage of trappingnot only electrons, but also charged species at the target, so that they donot hit the substrate, with an improvement of the film quality.

In our experiment, having a base pressure lower than 1 × 10−6 Torr anAr − O2 (3:1) gas mixture built up a total pressure of 60 mTorr inside thechamber. During growth the substrate temperature was 650 ◦C and the

rf power 60W. Changing the distance between target and substrate, thedeposition rate was settled to 5.5 and 6.4 Å/s for thinner and thicker films,respectively. Following deposition the films were post-annealed in situ at540 ◦C in 700 Torr oxygen pressure for 10 min and then slowly cooled toroom temperature [30].

3.3 Films Crystalline Properties using X-ray Dif-

fraction

3.3.1 X-ray Diffraction

Bragg diffraction occurs a electromagnetic waves interact with a regularstructure whose repeat distance is about the same as the wavelength. Thephenomenon is common in the natural world, and occurs across a broadrange of scales. For example, light can be diffracted by a grating havingscribed lines spaced on the order of a few thousand angstroms, about the


29/101

3.3 Films Crystalline Prop erties using X-ray Diffraction 21

2

C

B

A d

Figure 3.6: Geometry for Bragg diffraction of x-rays by a crystal [51].

wavelength of light. It happens that X-rays have wavelengths on the orderof a few angstroms, the same as typical interatomic distances in crystallinesolids. That means X-rays can be diffracted from minerals which, by defini-tion, are crystalline and have regularly repeating atomic structures.

When certain geometric requirements are met, X-rays scattered from acrystalline solid can constructively interfere, producing a diffracted beam.In 1912, W. L. Bragg recognized a predictable relationship among severalfactors.

1. The distance between similar atomic planes in a mineral (the inter-atomic spacing) which we call the d -spacing and measure in angstroms.

2. The angle of diffraction which we call the θ angle and measure indegrees. For practical reasons the diffractometer measures the anglebetween the incident and reflected beams. Not surprisingly, we call themeasured angle ‘2-θ’.

3. The wavelength of the incident X-radiation, symbolized by the Greekletter λ and, in our case of CuK α radiation, equal to 1.54 Å.

3.3.2 θ − 2θ scan

In XRD characterization, the θ − 2θ scan is ordinary operation mode. Itgives us the information about crystalline orientation of the material in the


30/101


growth direction (also called "out-of-plane"direction) and distances betweenlattice planes parallel to the film surface. In this mode the incident angleθ is varied and the detector angle for the diffracted beam is simultaneouslymoved uniformly at 2θ. To collect maximal information, a θ − 2θ scan isusually performed over a wide angular range, say at least 20◦< 2θ < 100◦.

Figure 3.7 (a) shows a typical θ−2θ diffraction pattern of PLD depositedAg0.5Ta0.5NbO3 thin film on a single crystal LaAlO3 (001) substrate in therange 20◦< 2θ < 120◦. X-ray diffraction (XRD) measurement was carriedout using an x-ray powder diffractometer (Siemens D5000 ) with filteredCuK α radiation source (λ=1.54056 Å) to evaluate crystalline properties of ATN films.

ATN films grown on the LaAlO3 (001) single crystal show exclusivec (001)-axis orientation due to the epitaxial relation between perovskite ATNfilm and LaAlO3 substrate. ATN films on r -cut sapphire have strong pref-erential (00l) orientation: the intensities of ATN (001) and (110) reflectionsare in the ratio of I ATN(001) : I ATN(110) = 2.30 compared to 0.06 in ATN

ceramics [16]. As shown in [18], [19], there is a pseudo-epitaxial relationbetween the perovskite structure and r -cut sapphire:

(001) perovskite (0112) Al2O3,[010] perovskite [421] Al2O3.

ATN films on amorphous glass substrate are polycrystalline.

Figure 3.7 (b) shows NKN, ATN and BST films properties fabricatedon sapphire substrate. All films are single phase and strongly c(001)-axisoriented. The intensities of ATN (001) and (110) Bragg reflections are inthe ratio of I BST(001) : I BST(110) = 1.41 compared to 0.10 in BST ceramics.

On the other hand, NKN film on sapphire single crystal substrate possessessuperior crystalline quality since it has exclusive c (001)-axis orientation.Figure 3.7 (c) and (d) shows ATO and ANO films properties as well as

targets. ATO and ANO films on sapphire are single phase with a perovskitestructure though a small fraction of Ag2Ta4O11 phase was observed in ATOceramic target in the bottom panel of figure 3.7 (c). In ANO/ LaAlO3 film,the intensities of ANO (001) and (110) reflections are in the ratio of I ANO(001): I ANO(110) = 0.19, whereas it is only 0.015 for ANO ceramic target. In caseof ATO/LaAlO3 film, the ratio of I ATO(001) : I ATO(110) was 0.07 compared to0.23 in ATO ceramic target in the bottom panel of figure 3.7 (d). Therefore,we conclude on r -cut sapphire ANO and ATO films grow preferentially (001)and (110) oriented, respectively.

3.3.3 ω-scan (Rocking curve)

An ω-scan, or rocking curve, is performed to obtain information about themisorientation of crystallites in thin film layer. The 2θ angle is fixed corre-


31/101

3.3 Films Crystalline Prop erties using X-ray Diffraction 23

Figure 3.7: X-ray diffraction θ−2θ scan of (a) ATN film on sapphire, LaAlO3,Glass substrate (b) NKN, ATN, BST films on sapphire substrate, (c) ATOfilm on sapphire and LaAlO3, ATO target, (d) ANO film on sapphire andLaAlO3, ANO target


32/101


Figure 3.8: Schematic diagram of ω-scan and φ-scan: ω is the tilting angleof the sample, 2θ the angle of the detector position, and φ the rotation of the sample perpendicular to its surface [51].

Figure 3.9: XRD rocking curve (ω-scan) and FWHM of the (002) reflectionsfrom Ag0.5Ta0.5NbO3 film and LaAlO3 substrate. The inset shows φ-scan of the off-normal (103) planes.


33/101

3.4 Photolithography and Lift-off Technique 25

sponding to the maximum of a Bragg peak and the incident angle is variedω around the θ angle, as shown in figure 3.8.

Even though a thin film may be single phase with almost perfect ori-entation in the growth direction, there can still be small deviations in theorientation of the different grains that the film consists of. The standard

way of comparing rocking curves is to measure the full width at half max-imum (FWHM). The narrower FWHM, the smaller deviation of the grainorientations regarding the normal to the film surface. Figure 3.9 shows therocking curve for the (002) Bragg reflections of ATN/LaAlO3 measured bydiffractometer, where θ + ω is designated just by θ as often is done. Narrowrocking curve with the full width at half maximum (FWHM) is 0.240◦ forATN (002) Bragg reflections.

3.3.4 φ-scan

In a φ-scan, the sample is rotated around its surface normal, having theincident beam and detector positions fixed. From the scan, information of in-plane orientation of grains, and relative in-plane orientation of grains com-pared to the substrates, can be collected (texture). Figure 3.9 shows a full360◦ φ-scan of the oblique (103) planes of the ATN film and LaAlO3 sub-strate. A fourfold symmetry is observed, where the film and substrate peaksappeared at the same φ angles indicating strong in-plane grains orientation.

The inset in figure 3.9 shows φ-scan of the off-normal (103) planes mea-sured at oblique geometry: θsam = 56.95◦ 2θdet = 77.03◦ for ATN (103) andθsam = 58.51

◦ 2θdet = 79.94◦ for LaAlO3 (103). These features together

with the coincidence of φ-scan of ATN (103) and LaAlO3 (103) reflectionundoubtedly indicate epitaxial quality of ATN film on LaAlO3 substrate.

3.4 Photolithography and Lift-off Technique

Photolithography is a reproduce process used in semiconductor device fab-rication to transfer a pattern from a photo-mask to the surface of a waferor substrate using light. Developed originally for reproducing engravingsand photographs, photolithography was found ideal in the 1960s for mass-producing integrated circuits [57].

A layer of photoresist – a chemical that hardens when exposed to light –is applied on top of the metal layer. The photoresist is selectively hardenedby illuminating it in specific places. For this purpose a transparent platewith patterns printed on it, called a mask, is used together with an illumi-

nation source to shine light on specific parts of the photoresist. Then, thephotoresist that was not exposed to light and the metal underneath is etchedaway with a chemical treatment. Finally, the hardened photoresist is etchedusing a different chemical treatment, and all that remains is a layer of metalin the same shape as the mask.


34/101


Figure 3.10: Schematic view of the coplanar waveguide interdigital capacitor.

Projection exposure tools, which are now used routinely in the semicon-ductor industry, have continually improved over the past several decades tosatisfy the insatiable demand for reduced feature size, increased chip size, im-proved reliability and production yield, and lower overall cost. High numeri-

cal aperture lenses, short-wavelength light sources, and complex photoresistchemistry have been developed to achieve fabrication of fine patterns overfairly large areas. Research and development efforts in recent years havebeen directed at improving the resolution and depth of focus of the pho-tolithographic process.

To make simplification of clean room processing, metal lift-off techniquewas employed for device structure. The pattern for microwave character-ization consists of a coplanar waveguide interdigital capacitor (CPWIDC)structure shown schematically in figure 3.10. Brief description of clean roomprocessing for an image reversal lithography described as follow[52]:

Cleaning To remove atmospheric dust and contamination, film sample wascleaned for 5 minutes in an ultrasonic bath of acetone, then rinsed iniso-propanol and de-ionized (DI) water, and finally dried with nitrogen.

Dehydration Bake Cleaned sample should be pre-baked at 110 ◦C for 2minutes for drying. Otherwise, photoresist will be coated with watervapor.

HMDS The sample was placed in an HMDS (Hexamethyldisilazane) envi-ronment for 2 minutes to make good adhesion between film and pho-toresist layer.

Spin coating Clariant image reversal photoresist (AZ 5214E) was used forcoating layer. Spin speed depends on desired thickness. 4000 RPM for30 seconds gives an approximately 1.4 µm thick uniform layer.

Soft bake Soft baking removes most of the remaining solvent from the pho-toresist film, thereby densifying it. Softbake time and temperature also


35/101

3.5 Coplanar Waveguide 27

influences adhesion, photospeed, and dimensional control of the pho-toresist. Our samples were baked on a hot-plate at 100◦C for 1 minute.

Exposure Exposure depends on resist thickness, lamp intensity and manyother variables. The pattern was defined using a Karl Süss MA6/BA6

mask aligner with UV-light exposure for 2 seconds (35–45 mJ/cm2

).Post bake The most critical parameter of the IR-process is reversal-bake

temperature, once optimised it must be kept constant within ± 1 ◦Cto maintain a consistent process. This temperature also has to beoptimised individually. In any case it will fall within the range from115 to 125 ◦C. If IR-temperature is chosen too high (> 130 ◦C) theresist will thermally cross-link also in the unexposed areas, giving nopattern. In our case, 120 ◦C for 1 minute on hot plate made bestresults.

Flood exposure The flood exposure is absolutely uncritical as long as suf-

ficient energy is applied to make the unexposed areas soluble. 300mJ/cm2 was a good choice for us, but 150-500 mJ/cm2 will have nomajor influence on the performance.

Developing The exposed resist was developed in a solution for 40 seconds.The solution composed of developer (Microposit 351) and DI waterwith the ratio of 1:5.

The electrodes consisted of a thin layer of Cr (10 nm) covered by a thickerlayer of Au (500 nm). The chrome layer is acting as an adhesive layer betweenthe ferroelectric thin film and the gold contact. Gold is chosen because of its high conductivity and oxidation resistance. The metal deposition was

performed by electron-beam evaporation.In the lift-off process, the sample was put in an ultrasonic acetone bath

for about 10 minutes, thereafter rinsed in iso-propanol (about 2 minutes) andDI water, and finally dried with nitrogen to remove the resist and unwantedmetal.

3.5 Coplanar Waveguide

Coplanar Waveguide (CPW) is a type of planar transmission line used inmicrowave integrated circuits (MICs) as well as in monolithic microwaveintegrated circuits (MMICs). The unique feature of this transmission line is

that it is uniplanar in construction, which implies that all of the conductorsare on the same side of the substrate. This attribute simplifies manufacturingand allows fast and inexpensive characterization using on-wafer techniques.

CPW fabricated on a dielectric substrate was first demonstrated by C.P.Wen [61] over three decades ago. The basic structure of CPW is illustrated


36/101


Figure 3.11: Typical Coplanar Waveguide: (a) structural dimensions, (b)electromagnetic field distribution. The solid lines represent electric field andthe dashed lines represent magnetic field components.

in figure 3.11 where the arrangement is assumed to be symmetrical with stripwidth w and equal longitudinal gap s. The side conductors are ultimatelygrounded, theoretically at infinity. As originally conceived Wen’s design for-mulations for CPW demanded an infinitely thick substrate, and this prac-tical difficulty doubtlessly impeded its implementation in any general sense.This was in spite of the fact that CPW has the following advantages overmicrostrip:

• Easier grounding of surface-mounted components.

• Simpler fabrication with low cost.

• Reduced dispersion (for small geometries).

• Decreased radiation losses.

• Easy to construct shunt and series element.

• Reduced cross talk between lines.

• Couplers having higher directivities.

• Availability of closed-form expressions for the characteristic impedance.

• Photolithographically defined structures with relatively low dependenceon substrate thickness.

The current density on the CPW signal line is not uniform across itssurface. Here, current crowds to the edges of the signal line, even more so

than in a microstrip line of comparable width. The edges of the line arerougher than the conductor’s surface. Compared to a similar microstripline, a CPW line with rough edges will have greater insertion loss. CPW’sadvantage over microstrip is that the CPW line width is independent of the line’s impedance, made possible by adjusting the CPW’s slot width. A


37/101

3.5 Coplanar Waveguide 29

wider line provides more area, lowering the conductor’s high-frequency loss.As a results, CPW circuits can be made denser than conventional microstripcircuits. These, as well as several other advantages, make CPW ideally suitedfor MIC as well as MMIC applications.


38/101


39/101

Chapter 4

Microwave On-wafer

Characterization for

Ferroelectrics

Modeling the microwave dielectric properties of ferroelectric materials, and inparticular the physical mechanisms underlying the temperature, electric fieldand frequency dependencies of εr and tan δ , have been discussed extensivelysince the late 1950s. A phenomenological model of the permittivity andlosses of ferroelectrics has been developed by Vendik [58] and subsequentlydiscussed by Gevorgian [59]. It is not our intention to discuss these modelshere.

In general, the change of dielectric constant with frequency is small in themicrowave frequency range. The losses in a ferroelectric crystal or film aremore difficult to analyze as they originate from different sources. As a rule of

thumb, the loss tangent normally increases with frequency and applied field,and the losses in a thin film are likely to be higher than in a bulk crystal of the same material [60].

Figure 4.1: A two-port network with voltages, currents, input wave a andoutput wave b.


40/101

32 Microwave On-wafer Characterization for Ferroelectrics

Figure 4.2: SOLT standards for CPW structure.

4.1 Microwave Network

4.1.1 Concept of Microwave Network

The basic concept of microwave network is developed from the transmissionline theory, and is a powerful tool in microwave engineering. Microwavenetwork method studies the responses of a microwave structure to externalsignals, and it is a complement to the microwave field theory that analyzesthe field distribution inside the microwave structure. In the network ap-proach, we do not care the distributions of electromagnetic fields within amicrowave structure, and we are only interested in how the microwave struc-ture responds to external microwave signals.

Two sets of physical parameters are often used in network analysis. Asshown in figure 4.1, one set of parameters are voltage V (or normalizedvoltage v) and current I (or normalized current i) the other set of parameters

are the input wave a (the wave going into the network) and the output waveb (the wave coming out of the network). Different network parameters areused for different sets of physical parameters. For example, impedance andadmittance matrixes are used to describe the relationship between voltageand current, while scattering parameters are used to describe the relationshipbetween the input waves and output waves.

4.1.2 Impedance and Admittance Matrix

For two port network shown in figure 4.1, we have

[V ] = [Z ][I ]

[I ] = [Y ][V ]

where [V ] is the unnormalized voltage, [I ] is unnormalized current.


41/101

4.1 Microwave Network 33

Figure 4.3: A picture of on-wafer measurement setup.

The impedance and admittance matrix are

[Z] =

Z 11 Z 12Z 21 Z 22

[Y] =

Y 11 Y 12Y 21 Y 22

Z 11 is the input impedance at port 1 when the other port is open.Z 12 is the transition impedance from port 2 to port 1 when port 1 is open.Y 11 is the input admittance at port 1 when the other port is shorted.Y 12 is the transition admittance from port 2 to port 1 when port 1 is shorted.

From above equations, we can get the relationship between [Y ] and [Z ]

[Z ][Y ] = 1

4.1.3 Scattering Parameters

The responses of a network to external circuits can also be described by theinput and output microwave waves. The input waves at port 1 and port 2are denoted as a1 and a2 respectively, and output waves from port 1 and port2 are denoted as b1 and b2 respectively. These parameters may be voltageor current, and in most case, we do not distinguish whether they are voltageor current. The relationships between the input wave [a] and output wave[b] are often described by scattering parameters [S ]:

[b] = [S ][a]


42/101


DUT

Figure 4.4: Three sets of device structures used for de-embedding the probepads and interconnects.

[S] = S 11 S 12S 21 S 22

When port 1 is connected to a source and the other port is connected to

a matching load, the reflection coefficient at port 1 is :

Γ11 = S 11 = b1a1

When port 2 is connected to a source and port 2 is connected to a match-ing load, the transmission coefficient from port 2 to port 1 is :

T 2→1 = S 12 = b1

a2

4.2 Calibration

Generally speaking, calibration transfers the quality of the calibration methodto the quality of the RF test system. In other words, any error in the cali-brated measurement will be due to the calibration method or standards, nottoe the test system’s imperfections. A variety of calibration methods areavailable to the test engineer. This section introduces only two main cali-bration methods that were used for our experiments. More practical guidefor calibration technique with network analyzer is described in Appendix IV.

4.2.1 SOLT calibration

SOLT (Shorts, Opens, Loads, and Thrus) calibration relies on well-knownstandards, all defined along the same reference plane(see figure 4.2). With


43/101

4.2 Calibration 35

SOLT, there is a direct connection between knowledge of the standards’precise RF characteristics and the accuracy of the calibration. Well-knownstandards bring forth a better SOLT calibration.

Although popular, the SOLT method has disadvantages, the principleone being accuracy of the standards. The better the standards are known,

the better the calibration. All the standards used in SOLT are direct stan-dards. Even small deviations from ideal can lead to large errors, manifestedin regions of the Smith chart far from the calibration standard [63]. Fur-thermore, accurately characterizing the SOLT standards becomes laboriousat frequencies above 20GHz [64]. The distributed nature of the standardscomplicates generating high-frequency models of the open, short, load. Asmentioned, the ideal thru connects the two ports perfectly. Because thisis rarely possible, applying offset delay and offset loss are the best ways toimprove the thru.

4.2.2 TRL calibration

TRL (Thru-Reflect-Line) calibration is a powerful method because it is basedon a transmission line standard [65]. A short transmission line uses as thethru in figure 4.2. When the offset delay is set to zero, the thru’s midpointsets the electrical reference plane. In general, the line lengths leading to theDUT should be the same length as the thru.

The reflect standard can be either an open or a short. Only the sign if its reflection coefficient need be know. The reflect does not have to be aperfect open or short, although the best results have a reflection coefficientclose to 1. The phase of reflect should be generally known to within +/-

90. When selecting a reflect, keep in mind an open operates over a broaderbandwidth than a short. Be sure to use the same reflect standard for bothports. When its phase is well known, the reflect can be used to define theelectrical reference plane instead of the thru.

The one precision standard is the line. Its characteristic impedance Z 0sets the reference impedance for the entire RF test system. For good RFmeasurements, the line’s impedance should be precisely defined; otherwise,normalization is required. For example, say the characteristic impedance of the line is not 50Ω but instead 48Ω. Because the measurements are referencedto the characteristic impedance of the line, the measured S-parameters mustbe normalized from 48Ω to 50Ω.

TRL sets the Z 0 of test system to be equal to the impedance of the linestandard. However, tolerances in the conductor etch fabrication can resultin the line’s impedance being slightly different than Z 0. Since the DUT mea-surements are defined in terms of Z 0, error will result. The TRL’s inabilityto compensate for imperfect transmission lines is its principal weakness.


44/101


4.2.3 LRM calibration

LRM (Line-Reflect-Match) is very similar in form of TRL (Thru-Reflect-Line) [66]. The line and reflect are analogous to the thru and reflect standardsin TRL, the difference being that LRM uses a precision match (or load) todefine the system’s characteristic impedance Z

0. Again, either an open or a

short can serve as the reflect. As in SOLT, the load must be well defined;otherwise, the calibration sensitivity is degraded. The line standard sets theelectrical reference plane.

Contrary to SOLT, neither the open’s capacitance nor the short’s induc-tance need to be known before calibration due to the match involved [ 67].Using the same reflect and load standards to calibrate both ports gives themost accurate results. Upon completion of an LRM calibration, the refer-ence plane for the two ports is at the center of the line. The LRM’s uniqueaspect, the match, is also its weakness and parasitic inductance corrupts theload. The inductance of the interconnecting pads at either end of the matchresults in LRM calibration error. When this inductance is accounted for, the

accuracy of LRM is comparable to TRL. To minimize inductance, ensure theconductor traces leading to the match are short in length.

When a precise, well-defined load is available, the LRM method has abe constructed, TRL has the advantage. To trade off their strengths, SOLTcan be performed at lower frequencies and TRL at higher frequencies. LRMfunctions as a combination of the two. It generally has better accuracy thanthe SOLT method, even when using the same standards to perform bothmethods.

4.3 De-embedding Analysis

In microwave measurements, it is important to accurately account for theparasitic elements because the measured S -parameter contains the charac-teristics of the device under test coaxial lines and the probing pads. Theimpedance of the transmission metal lines causes additional loss in the struc-ture. The loss due to the line impedance has to be separated from the di-electric loss in the device. De-embedding must be performed to discover theRF performance of the on-wafer DUT. Preparing to characterize a DUT ona wafer is a two-step process. First, a VNA calibration is performed, settingthe reference plane at the probe tips. Then, a set of dummy test structuresis measured. These are used to move the electrical reference plane from theprobe tips to the DUT plane.

We have modified de-embedding method, developed for active componentin reference [68, 69], to our CPW structures to extract capacitance and losstan δ of device under test (DUT). Our basic assumption is that the test struc-ture can be corresponded to the equivalent circuit shown in figure 4.4-4.5.


45/101

4.3 De-embedding Analysis 37

Figure 4.5: The equivalent circuit of Coplanar Waveguide Interdigital Ca-pacitors.

All the Y -parameters can be calculated by standard conversion equations

between two-port network parameters [53] from the measured S -parameters.The admittances y1 and y2 represent the pad parasitics between signal andground line. z1 and z2 can be expressed by the metal interconnect seriesimpedance between port 1 and port 2. De-embedding start by measuringthe open structure. Referring to the equivalent circuit, the admittances y1and y2 are found by measuring Y -parameter.

y1 = Y 11open + Y 12open

y2 = Y 22open + Y 12open

Using y1 and y2, de-embed the shunt admittance from the measurement,

resulting inY A = Y meas −

y1 0

0 y2

Next, the series impedances z1, z2 can be calculated from the measuredY -parameters of thru and device.

z1 = 1

2

−1

Y 12Thru+

1

Y 11Thru − y1−

1

Y 22Thru − y2

z2 = 1

2

−1

Y 12Thru−

1

Y 11Thru − y1+

1

Y 22Thru − y2

Converting the Y -parameter matrix (Y A) to the Z -parameter matrix(Z A), and subtract z1 and z2 using

Z B = Z A −

z1 0

0 z2


46/101


Table 4.1: Comparison of various CPWIDC on sapphire substrate

Parameter NKN ATN BST ANO

Frequency dispersion of capacitance [%] 34 4.3 12 13Capacitance C @ 20GHz [pF] 0.19 0.11 0.15 0.045

Tunability = ∆C/C @ 40V, 20GHz [%] 10 4.7 26 4.6

Loss tanδ @ 20GHz 0.21 0.06 0.14 0.10

K-factor = tunability/tan δ @ 40V, 20GHz[%]

50 76 306 45

Parasitic Capacitance C @ 20GHz [fF] 40 27 30 23

Parasitic Loss tanδ @ 20GHz 0.16 0.08 0.11 0.012

Interconnect impedance Z Ω @ 20GHz Re z 1.9 1.0 1.2 1.5

Interconnect impedance Z Ω @ 20GHz Im z 6.9 6.2 6.6 7.1

Transform Z -parameter matrix (Z B) into Y -parameter matrix (Y B) toget the DUTs Y -parameters. This method works particularly well with smallperiphery devices.

Y DUT = Y B

Capacitance and loss tanδ of device under test (DUT) can be calculated

byC DUT =

1

2πf

× Im

Y 12device × Y 12thruY 12device − Y 12thru

Tanδ DUT = Re ((Y 12device × Y 12thru)/(Y 12device − Y 12thru))

Im ((Y 12device × Y 12thru)/(Y 12device − Y 12thru))

4.4 Experimental results and Discussion

Results of de-embedding calculations are presented in figure 4.6 and com-

piled in table 4.1. Varactors on NKN film exhibit moderate tunability definedas [1- C(40V)/C(0V)]×100% and K-factor = tunability/tanδ whereas sufferhigh loss tanδ and parasitic ground-signal-ground capacitance Cp. BST filmvaractors show superior tunability and K-factor as high as 26 % and 306%, respectively, at 20 GHz and planar electric field of 200 kV/cm. To the


47/101

4.4 Experimental results and Discussion 39

Figure 4.6: Frequency dependence of the capacitance, loss tan δ , tunabil-ity, K -factor for (a) BST/Al2O3, (b) NKN/Al2O3, (c) ATN/Al2O3, (d)ANO/Al2O3, 2 µm gap CPWIDC at various dc bias voltages.


48/101


Figure 4.7: Microwave properties of 2 µm gap CPWIDC fabricated directlyon to the sapphire (Al2O3-0112, r -cut) single crystal. Capacitance C : DUTis the upper line, ground-signal-ground parasitic is the lower line; loss tan δ :DUT is the lower line, ground-signal-ground parasitic is the upper line. Allthe data recorded at 0, 10, 20, 30, and 40 V which overlap with each otherare shown.

best of our knowledge these results are comparable with the state-of-the-artmicrowave parameters published so far: at 6 GHz loss tanδ ∼ 0.06 and 0.07,

tunability ∼ 24 % and 15 % at 175 kV/cm in our BST varactor and pub-lished in [70], respectively.

AgTa0.5Nb0.5O3/Al2O3 varactors show superior loss properties and uniqueflat frequency dependence. At 20 GHz tanδ = 0.06, very low parasitic ca-pacitance and frequency dispersion as low as 3.4 % in whole frequency rangefrom 1 to 40 GHz. ATN varactors integrated on LaAlO3 substrates (high-k substrates with ε ∼ 25) show spurious resonances at high frequency (∼32.5 GHz)due to unperfect calibration since the standard calibration kit isfabricated on the substrate material different to the device under test. ATNCapacitors on the Corning 7059 glass (amorphous) possess low loss and flat

dispersion though demonstrates reduced capacitance and tunability.The reliability of the de-embedding technique is clearly proved by two

important observations. First, for all the samples calculated parasitic ca-pacitance Cp and complex interconnect impedance z appeared to be voltageindependent (Re z and Im z vs. f plots are similar to figure 4.7). Second,


49/101

4.5 New design of Buried Interdigital Capacitors 41

Figure 4.8: Microwave properties of 2 µm gap coplanar interdigital capacitors(a) fabricated on top of 2.2 µm thick ATN film (b) buried inside 2.5 µm thickATN film.

characteristics of CPWIDC structures fabricated directly on sapphire crystal

do not show, as expected, any frequency dispersion (figure 4.7).In conclusion, developed de-embedding technique was demonstrated ap-

plicable for on-wafer tunable microwave devices characterization and givestrustworthy ferroelectric characteristics not corrupted by parasitic impedanceof coplanar transmission line. AgTa0.5Nb0.5O3 film varactors demonstratelower dispersion and loss compared to (BaSr)TiO3 devices. However, tun-ability of ATN devices remains to be improved. This drawback could beovercome by improving the crystalline films qualities and varying Ta-to-Nbfilms composition.

4.5 New design of Buried Interdigital Capacitors4.5.1 Device Fabrication

The fabrication procedure of the buried coplanar waveguides is illustratedin figure 4.9. A 2 µm thick photoresist layer was spin coated on the surface


50/101


Figure 4.9: Schematic of the fabrication of buried interdigital capacitors: (a)the pattern is transferred from the mask to the photoresist by UV expositionand developing, (b) the patterned photoresist acts as a thick mask (2 µm)

for the Ar-ion milling of the ATN film, (c) electron gun evaporated gold isdeposited on the sample, and after lift-off process, (d) buried IDC are left.

of ATN film, in order to execute an i-line photolithographic process. Thepatterned photoresist acted then as mask to the Ar-IBE (ion beam etching)step which defined 400 nm deep grooves. Subsequently, 10 nm thick Cr and390 nm thick Au metallic layers were evaporated by electron gun on thesample. The photoresist was removed by lift-off standard procedure, usingresist remover in ultrasonic bath. Finally, in order to planarize surfaceseventually not perfectly flat, Ar-IBE can be employed again. High Au-to-ATN selectivity of the Ar-IBE process permits a low loss of the film. Apart

from grooves fabrication by etching process, all the others steps were followedalso for the fabrication of IDCs on top of ATN. 2.5 µm thick ATN films werechosen for buried devices.

4.5.2 Measurement results

In figure 4.8 (a), the microwave performance of ATN film Buried CPWIDCwith 2 µm finger gap on sapphire substrate in the microwave range from 1 to40 GHz showed that frequency dispersion is about 7.3%, voltage tunabilitywas 3.0% @ 20GHz and 200 kV/cm, loss tangent ∼ 0.07 @ 20GHz, K -factor= tunability/tanδ is 45% @ 20GHz.

On the other hand, Topped ATN CPWIDC with same size, same sub-strate and same frequency range in figure 4.8 (b) showed that frequencydispersion is about 3.3%, voltage tunability was 2.0% @ 20GHz and 200kV/cm, loss tangent ∼ 0.07 @ 20GHz, K -factor = tunability/tanδ is 31% @20GHz.


51/101

4.5 New design of Buried Interdigital Capacitors 43

4.5.3 Conclusion

Buried structures showed higher values of capacitance and tunability, similarlosses thus 50 % increased K -factor. Highly polarized ferroelectric which fillsthe gap between the fingers of interdigital capacitors causes the enhancedperformance. The new design offers a further advantage: being the electrodesinside the film, there is no need of any other step in order to planarize thesurface. Since planarization could be necessary in integration context, buriedstructures seems to be preferable also regarding this matter.


52/101


53/101

Chapter 5

Conclusion and future work

Stoichiometric perovskite ferroelectric Silver Tantalate Niobate (AgTa0.5Nb0.5O3,ATN), Sodium Potasium Niobate (Na0.5K0.5NbO3, NKN) and Barium Stron-tium Titanate (Ba0.5Sr0.5TiO3, BST) ceramic targets were sintered by mixed

oxide solid state reaction. X-ray diffraction showed they are single phase andslightly textured because of the cold isostatic pressure was employed. Fab-ricated ta

ferroelectric microwave

Documents