RECONFIGURABLE MICROWAVE ELECTRONICS USING ADDITIVE

MANUFACTURING AND MAGNETIC NANOMATERIALS

By

Yuxiao He

A DISSERTATION

Submitted to

Michigan State University

in partial fulﬁllment of the requirements

for the degree of

Electrical Engineering - Doctor of Philosophy

2020

ABSTRACT

RECONFIGURABLE MICROWAVE ELECTRONICS USING ADDITIVE

MANUFACTURING AND MAGNETIC NANOMATERIALS

By

Yuxiao He

This dissertation investigates how to design and fabricate low cost, lightweight, and re-

conﬁgurable RF and microwave electronics using additive manufacturing and nanomagnetic

materials. A key goal is to provide a speciﬁc methodology on obtaining high performance

RF electronics by taking advantage of the expanded design freedom, economic fabrication

process oﬀered by the additive manufacturing technology, as well as the unique microwave

properties of the magnetic nanomaterials. A number of commonly used RF and microwave

electronics are designed, fabricated, and measured to demonstrate that we have successfully

achieved such goal. The performance tradeoﬀs of the above two approaches during experi-

mental implementation are also discussed for each speciﬁc circuit and component category,

depending on the corresponding desired performance. This in turn provides us with practi-

cal and important comments regarding the limitations of our approaches. Therefore, we can

obtain an insightful guideline for achieving our key goal in terms of both design potentials

and implementation capabilities. The ﬁrst chapter addresses the current trend of smaller,

smarter, and cheaper electronics which motivates our work in this dissertation. The second

chapter provides the background on additive manufacturing technology and magnetic mate-

rials that were utilized as two separate approaches to achieve high performance and low cost

RF and microwave applications prior to this work. Based on such context, we are motivated

to propose a new thrust which pushes the boundaries beyond the current development of

both approaches and explores the new realm of combining them together. The third chapter

introduces several types of additive manufacturing techniques, and presents a number of

lightweight, reconﬁgurable, and high performance RF components that are developed using

one or two techniques and in a combined manner. Manufacturing factors that negatively

aﬀects the RF performance are analyzed and discussed with the corresponding technique

to alleviate such issue. In chapter four, we study the fundamental properties of the mag-

netic materials and explore the potential of using the magnetic nanomaterials fabricated by

novel chemical methods in microwave applications operating at frequencies higher than a

traditional range. The magnetic nanoparticle thin ﬁlms were fabricated using various com-

posites and were characterized for high frequency RF applications. Chapter ﬁve expands our

work by combining the additive manufacturing technology, magnetic nanoparticle ﬁlms, and

liquid crystal polymer substrate to develop low cost, miniaturized, and reconﬁgurable RF

components with compact and lightweight packages. The fabrication process is optimized

by considering the practical limitations from chapter two, and enhanced RF performances

are obtained in several applications. However, tradeoﬀs due to the magnetic materials are

also observed, which are then theoretically analyzed and veriﬁed by the measurement re-

sults. Our work has successfully demonstrated how the RF and microwave components and

circuits can be made as low cost, compact, and lightweight with enhanced performance using

additive manufacturing and magnetic nanomaterials. Future work will continue by explor-

ing new design capabilities and optimizing the fabrication process to further improve the

performance while reducing the overall fabrication cost.

To my parents, my wife, families and friends.

iv

TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii

LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

ix

CHAPTER 1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . .
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2 Problem Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3 Thesis Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4 Anticipated Research Contributions . . . . . . . . . . . . . . . . . . . . . . .
1.5 Dissertation Outline
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

CHAPTER 2 BACKGROUND . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1 Background of additive manufacturing technology . . . . . . . . . . . . . . .
2.1.1 General 3D printing process . . . . . . . . . . . . . . . . . . . . . . .
2.1.1.1 Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.1.2
3D Printing . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.1.3 Polishing . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.1.4 Metalization . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.2 Types of AM techniques . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.2.1
Stereolithography (SLA) . . . . . . . . . . . . . . . . . . . .
2.1.2.2 Fused Deposition Modeling (FDM) . . . . . . . . . . . . . .
2.1.2.3 Digital Light Processing (DLP) . . . . . . . . . . . . . . . .
2.1.2.4
. . . . . . . . . . . . . . . .
2.1.2.5 Polyjet
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.2.6 Aerosol Jet Printing (AJP)
. . . . . . . . . . . . . . . . . .
3D printing in RF and microwave applications . . . . . . . . . . . . .
2.2 Background of magnetic materials in RF applications . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . .
2.2.1.1 Circulators and isolators . . . . . . . . . . . . . . . . . . . .
2.2.1.2 Ferrite phase shifters . . . . . . . . . . . . . . . . . . . . . .
2.2.1.3 YIG ﬁlters
. . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1.4 Frequency selective limiter . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . .
2.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.2.2 RF devices using magnetic nanoparticle ﬁlms

Selective Laser Sintering (SLS)

2.1.3

2.2.1 Traditional microwave ferrite devices

1
1
2
4
4
5

7
8
8
8
9
9
10
11
11
12
12
13
13
15
16
18
19
19
20
20
22
23
24

CHAPTER 3 RF DEVICES USING ADDITIVE MANUFACTURING

3.1 Ultra-low loss lowpass ﬁlter using improved air substrate . . . . . . . . . . .
3.1.1 Low pass ﬁlter design . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.2 Fabrication process . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.3 Measurement results
. . . . . . . . . . . . . . . . . . . . . . . . . . .

TECHNOLOGY . . . . . . . . . . . . . . . . . . . . . . . . . . 25
25
27
27
29

v

3.2 Lego-like reconﬁgurable X-band cavity resonator

3.1.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
3.2.1 Reconﬁgurable rectangular cavity resonator design . . . . . . . . . . .
3.2.2 Fabrication process . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.3 Bi-material printed quasi yagi-uda antenna . . . . . . . . . . . . . . .
3.2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3 A Bi-material fully aerosol jet printed W -band quasi-yagi-uda antenna . . .
3.3.1 Antenna design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.2 Fabrication Process . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.3 Measurement setup and results
. . . . . . . . . . . . . . . . . . . . .
3.3.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.4 Aerosol jet printed 24 GHz quasi-yagi-uda antenna on a polyjet printed

cavity substrate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.1 The antenna design . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.2 Fabrication process . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.3
. . . . . . . . . . . . . . . . . .
3.4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Simulation and measurement results

30
31
32
34
36
38
39
39
40
42
44

45
45
46
49
50
52

CHAPTER 4 CHARACTERIZATION OF MAGNETIC NANOPARTI-

CLE FILM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

4.1 FMR Frequency and Magnetic Susceptibility in Magnetic

Nanoparticles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.1 FMR and magnetic susceptibility analysis
. . . . . . . . . . . . . . .
4.1.2 FMR measurement of ﬁlm particles . . . . . . . . . . . . . . . . . . .

4.2 Fabrication process of the magnetic nanoparticle thin ﬁlm and the basic

microwave device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.1 Layer-by-Layer Fabrication of CoFe2O4 Thin Films via Epoxy Curing
4.2.1.1 Epoxy coated slides
. . . . . . . . . . . . . . . . . . . . . .
4.2.1.2 Amine/ epoxy coated CoFe2O4 particles . . . . . . . . . . .
4.2.1.3
Layer-by-Layer Fabrication . . . . . . . . . . . . . . . . . .
4.2.2 Fabrication of CoFe2O4 Films via Solution Cast Method . . . . . . .
4.2.2.1 Epoxy coated slides and particles . . . . . . . . . . . . . . .
4.2.2.2 Film Casting . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.3 Basic microwave device . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.4
3-D polyjet printing technology . . . . . . . . . . . . . . . . . . . . .
4.3 Measurement results using the basic microwave device . . . . . . . . . . . . .
4.3.1 Measurement setup . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.2 Measurement Results . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

55
56
59

60
62
62
62
62
63
63
63
64
66
67
68
69
76

CHAPTER 5 PATCH ANTENNA ON NIFE2O4 NANOPARTICLE AIR

5.1 Antenna design and theoretical analysis . . . . . . . . . . . . . . . . . . . . .
5.1.1 Antenna on Air Substrate . . . . . . . . . . . . . . . . . . . . . . . .

SUBSTRATE . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
79
79

vi

5.1.2 Antenna on Magnetic Film Loaded Substrate
5.1.3

. . . . . . . . . . . . .
Impedance matching, realized gain analysis . . . . . . . . . . . . . . .
5.2 The characteristics of NiFe2O4 ﬁlms and Antenna Fabrication Process . . . .
5.2.1 The Characteristics of NiFe2O4 Films . . . . . . . . . . . . . . . . . .
5.2.2 NiFe2O4 Films Fabrication Process . . . . . . . . . . . . . . . . . . .
5.2.3 Antenna Fabrication and Assembly . . . . . . . . . . . . . . . . . . .
5.3 Antenna measurement results and analysis . . . . . . . . . . . . . . . . . . .
5.3.1 Antenna measurement setup . . . . . . . . . . . . . . . . . . . . . . .
5.3.2 Antenna on the self-biased magnetic substrate . . . . . . . . . . . . .
5.3.3 Antenna on the DC magnetic biased magnetic substrate
. . . . . . .

81
84
85
86
86
90
91
91
92
96
5.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

CHAPTER 6 TUNABLE BANDSTOP FILTER ON NANOMAGNETIC

THIN FILMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.1 Magnetic nanoparticles properties and bandstop ﬁlter design . . . . . . . . . 104
6.1.1 Magnetic Nanoparticles Properties
. . . . . . . . . . . . . . . . . . . 104
6.1.2 Bandstop Filter Design . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.2 Magnetic Films and Filter Fabrication . . . . . . . . . . . . . . . . . . . . . 106
6.2.1 Magnetic Film Loaded Air Cavity Fabrication . . . . . . . . . . . . . 106
6.2.2 Bandstop Filter Assembly . . . . . . . . . . . . . . . . . . . . . . . . 108
6.3 Filter Measurements and Discussions . . . . . . . . . . . . . . . . . . . . . . 108
6.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

CHAPTER 7 TUNABLE SPHERICAL CAVITY RESONATOR USING

MAGNETIC FILM . . . . . . . . . . . . . . . . . . . . . . . . . 114
7.1 Tunable Resonator Design and Theoretical Analysis . . . . . . . . . . . . . . 115
7.1.1
Spherical Cavity Resonator Design And Mode Analysis . . . . . . . . 115
7.1.2 Tunable Resonator Design . . . . . . . . . . . . . . . . . . . . . . . . 117
7.2 Resonator Fabrication Process . . . . . . . . . . . . . . . . . . . . . . . . . . 118
7.2.1 Magnetic Film Deposition And Characteristics . . . . . . . . . . . . . 118
7.2.2 Resonator Fabrication And Assembly . . . . . . . . . . . . . . . . . . 119
7.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
7.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

CHAPTER 8 SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126

vii

LIST OF TABLES

Table 3.1 Design parameters of the Lowpass ﬁlter. . . . . . . . . . . . . . . . . . . .

Table 3.2 Design parameters of the cavity resonator . . . . . . . . . . . . . . . . . .

27

33

Table 3.3 The measured unloaded qualify factor QL, simulated and measured tuned
frequencies (exclude P2, and Po is the unperturbed resonant frequency)

.

38

Table 3.4 The measured and designed dimensions of the key features of the 94 GHz

Quasi-Yagi-Uda antenna

. . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 3.5 Quasi-Yagi-Uda Antenna Design Parameters

. . . . . . . . . . . . . . . .

Table 3.6 Measured quasi-Yagi-Uda Antenna Design Parameters after curing process

Table 3.7 Quasi-Yagi-Uda Antenna Performance Parameters . . . . . . . . . . . . .

Table 4.1 Characterization of the selected magnetic nanoparticles.

. . . . . . . . . .

Table 4.2 Film thickness, types of materials and deposition methods of the mag-

netic ﬁlms

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 5.1 Measured performance metrics of S1–S3 and the air substrate case . . . .

44

48

52

52

56

64

94

Table 5.2 Comparison of the reference antennas (data format in the upper half:
before/after magnetic material loading, size reduction compared to the
patch without loading; lower half: size reduction compared to the λ

2 patch). 97

Table 5.3 Measured performance metrics of the antenna samples in under self-
biased and DC magnetic biased conditions (Date format: self-biased /
DC magnetic biased).

. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

99

Table 6.1 Performance metrics of the BSF before and after applying the 2500 Oe

of magnetic bias (Data format: before/after).

. . . . . . . . . . . . . . . . 113

Table 7.1 Measured unloaded quality factors and S parameters at the resonances

of the air-ﬁlled case and all tuning cases.

. . . . . . . . . . . . . . . . . . 124

Table 7.2 Comparison of the state-of-the-art tunable and 3D printed resonators.

. . 124

viii

LIST OF FIGURES

Figure 2.1 Form 3L SLA 3D printer [1]

. . . . . . . . . . . . . . . . . . . . . . . . .

Figure 2.2 Stratasys Fortys 450mc 3D printer [2] . . . . . . . . . . . . . . . . . . . .

Figure 2.3 Stratasys Connex Objet350 3D printer [2]

. . . . . . . . . . . . . . . . .

Figure 2.4 Optomec AJ 5X system [3] . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 2.5 Schematic diagram of phased array antenna module [4]

. . . . . . . . . .

Figure 2.6 Superconducting meander-line phase shifter [4] . . . . . . . . . . . . . . .

12

13

14

15

19

20

Figure 2.7 Simple MSW device with microstrip transducers. Depending on design,

this can provide time delay and ﬁltering functions [4]

. . . . . . . . . . .

21

Figure 2.8 Frequency response of an FSL, and input and output power spectra,
from multiple weak and strong signals, showing the frequency selectivity
inherent in YIG limiters [4]

. . . . . . . . . . . . . . . . . . . . . . . . .

Figure 3.1 Layout of the lowpass ﬁlter design . . . . . . . . . . . . . . . . . . . . . .

22

28

Figure 3.2 Schematic of low-pass ﬁlter implemented using combination of thin-ﬁlm

process and 3D printing . . . . . . . . . . . . . . . . . . . . . . . . . . .

29

Figure 3.3 Final assembled ﬁlter with all three layers combined together in a Lego-

like approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 3.4 Simulated and measured reﬂection ratio (S11) for low pass ﬁlter . . . . .

Figure 3.5 Simulated and measured reﬂection ratio (S21) for low pass ﬁlter . . . . .

Figure 3.6

(a).Simulated electric ﬁeld of the dominant TE101 mode; (b).Cross-
sectional view of the cavity; (c). Schematic view of the cavity.

. . . . . .

Figure 3.7 The assembling process of the cavity resonator . . . . . . . . . . . . . . .

Figure 3.8

(a) All the 6 posts are pulled up (no perturbation); (b) The center post
is pushed into the cavity to perturb the ﬁeld.

. . . . . . . . . . . . . . .

Figure 3.9 Measured and simulated S21 for all 7 frequencies

. . . . . . . . . . . . .

29

30

31

34

35

36

38

Figure 3.10 The design schematic of the W band antenna. All metal layers and the

dielectric substrate are Aerosol Jet printed . . . . . . . . . . . . . . . . .

40

ix

Figure 3.11 The illustration of the Aerosol Jet printing process

. . . . . . . . . . . .

41

Figure 3.12 (a) The surface proﬁle of the whole polyimide cross section (b) The
enlarged segment of the polyimide cross section (c) The top view of the
printed quasi-Yagi-Uda antenna . . . . . . . . . . . . . . . . . . . . . . .

Figure 3.13 The simulated and measured S11 of the printed quasi-Yagi-Uda antenna .

Figure 3.14 (a) The simulated 3D far ﬁeld radiation pattern at 94 GHz (b) The

simulated maximum realized gain V.S. frequency . . . . . . . . . . . . .

Figure 3.15 Top view of the quasi-Yagi-Uda antenna on the substrate

. . . . . . . .

Figure 3.16 Cross-sectional view of the quasi-Yagi-Uda antenna on the substrate . . .

Figure 3.17 (a) 3D printed substrate compared to the U.S. quarter coin, (b) Aerosol
Jet printed quasi-Yagi-Uda antenna on the VeroWhitePlus substrate
after curing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

43

44

44

47

48

49

Figure 3.18 (a) The Aerosol Jet printing system, (b) The quasi-Yagi-Uda antenna

with a 1.85 mm end launch connector . . . . . . . . . . . . . . . . . . . .

49

Figure 3.19 The simualted and measured return loss of the antenna fabricated by

AJP and polyjet technology . . . . . . . . . . . . . . . . . . . . . . . . .

50

Figure 3.20 The simualted and measured E plane radiation patterns of the antenna

fabricated by AJP and polyjet technology . . . . . . . . . . . . . . . . .

Figure 3.21 Maximum simulted realized gain versus frequency . . . . . . . . . . . . .

51

51

Figure 4.1 XRD pattern for CoFe2O4 (top) and MnFe2O4 (bottom), sizes calcu-

lated by Debye-Scherrer equation. . . . . . . . . . . . . . . . . . . . . . .

57

Figure 4.2 Temperature dependent magnetization for CoFe2O4 (top) and MnFe2O4
(bottom), zero-ﬁeld cooled (ZFC) susceptibility measurements were con-
ducted under an applied ﬁeld of 100 G with temperature range from 5
- 310 K.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 4.3 Selected FMR absorption proﬁles. . . . . . . . . . . . . . . . . . . . . . .

Figure 4.4 Correlation between resonance ﬁelds and magnetic susceptibilities. . . . .

58

59

60

Figure 4.5 FMR of nanoparticle ﬁlms. The intensity from CoFe2O4 nanoparticle

ﬁlm has been ampliﬁed 10 times from the original signals.

. . . . . . . .

61

Figure 4.6

Images of the MnFe2O4 sample LbL1 and CoFe2O4 samples SC3, SC6
and SC7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

64

x

Figure 4.7

(a) Microstrip line design (b) Microstrip line characterized with end-
launch connectors.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 4.8

(a) Overview of the basic microwave device (b) The microwave device
using additional two metal clamps and characterized by the RF end-
launch connectors.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 4.9 Electromagnetic ﬁelds illustration of the ﬂipped magnetic ﬁlm sample

on the microstrip line.

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 4.10 (a) Objet Connex 350 (b) Illustration of the 3-D polyjet technology [69].

Figure 4.11 Measurement setup process steps 1-4. . . . . . . . . . . . . . . . . . . . .

Figure 4.12 baseline: Mere glass substrate ﬂipped on top of the microstrip line.

Reference line: Only microstrip itself. . . . . . . . . . . . . . . . . . . . .

Figure 4.13 Group delay for all the solution cast CoFe2O4 samples and the baseline.

65

65

66

67

69

70

71

Figure 4.14 Solution cast CoFe2O4 sample sequences in terms of the ﬁlm thickness

and the measured group delay. . . . . . . . . . . . . . . . . . . . . . . . .

71

Figure 4.15 The relationship between the average group delay and the magnetic ﬁlm

thickness for selected solution cast CoFe2O4 samples and the baseline. . .

72

Figure 4.16 TGA curves for LbL and solution cast ﬁlms showing particle loading

densities of 90% and 65% for LbL and solution cast respectfully. . . . . .

73

Figure 4.17 Group delay for the MnFe2O4 samples LbL1 and LbL2 and CoFe2O4

sample LbL3.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

74

Figure 4.18 (a) S21 for all CoFe2O4 magnetic ﬁlms (b) The absolute power absorp-
tion percentage for all CoFe2O4 magnetic ﬁlms (c) S21 of LbL1 and
LbL2 compared to SC7 (d) The absolute power absorption percentage
LbL1 and LbL2 compared to SC7.

. . . . . . . . . . . . . . . . . . . . .

Figure 5.1

(a) The design parameters: Lp = 18.2 mm, Wp = 18.3 mm, WM S =
4.9 mm, LM S = 9.41 mm, Gp = 0.55 mm, H = 1.05 mm; (b) Fully
assembled antenna with auxiliary part.

. . . . . . . . . . . . . . . . . . .

Figure 5.2

(a) Patch antenna on the NiFe2O4 magnetic ﬁlm loaded substrate; (b)
The cross sectional view of the magnetic ﬁlm loaded antenna.

. . . . . .

Figure 5.3

(a) The 3D printed cavity with metalized surfaces; (b) The fabricated
NiFe2O4 magnetic ﬁlm. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

75

80

82

87

xi

Figure 5.4

(a)-(c): Sample S1, S2, and S3; (d) Illustration of edge and surface scan
using proﬁlometer; (e) Surface scan proﬁle; (f) Edge scan proﬁle. . . . . .

Figure 5.5

(a) The lego-like assembly process; (b) Fully assembled NiFe2O4 nanopar-
ticle ﬁlm loaded air substrate antenna

. . . . . . . . . . . . . . . . . . .

89

90

Figure 5.6 The radiation pattern measurement setup using SATIMO Starlab near

ﬁeld system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

91

Figure 5.7 The measured S11: S1–S3 and air substrate antenna, the simulated S11:

S2 and air substrate antenna.

. . . . . . . . . . . . . . . . . . . . . . . .

92

Figure 5.8 Measured radiation patterns for S1–S3 and air substrate antennas: (a)
E plane; (b) H plane. Measured resonant frequencies: 7.45 GHz (S1),
7.11 GHz (S2), 6.46 GHz (S3), 8.21 GHz (Air).

. . . . . . . . . . . . . .

Figure 5.9 Measured maximum realized gain and eﬃciency vs. frequency for S1–S3

and air substrate antennas: (a) maximum realized gain; (b) eﬃciency. . .

Figure 5.10 The antenna sample with permanent magnets on the back . . . . . . . .

95

96

98

Figure 5.11 Comparison of the measured S11 of the magnetic ﬁlms loaded antennas

under self-biased and DC magnetic biased conditions. . . . . . . . . . . .

99

Figure 5.12 The measured E and H planes radiation patterns of sample S1, S2 and
S3 under self-biased and DC magnetic biased conditions. S1:(a)(b),
S2:(c)(d), S3:(e)(f). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

Figure 6.1 Selected FMR absorption proﬁles of MnFe2O4 nanoparticles [? ].

. . . . 105

Figure 6.2 The design of the BSF on the air substrate: LM = 25.5 mm, WM 1 = 4
mm, WM 2 = 3.6 mm, LS1 = 7.3 mm, WS1 = 0.8 mm, LS2 = 7.2 mm,
WS2 = 2.3 mm, H = 1 mm, t = 120 µm - 300 µm.

. . . . . . . . . . . . 106

Figure 6.3 Fabricated sample: (a) 3D printed air cavity on the copper sheet; (b)

100 µm thick MnFe2O4 ﬁlm loaded air cavities.

. . . . . . . . . . . . . . 107

Figure 6.4 Fabricated sample: (a) The lego-like assembly process; (b) Fully assem-

bled microstrip BSF.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

Figure 6.5 Measured (a) insertion loss and (b) return loss of the ﬁlms and non-
ﬁlms loaded BSF. The corresponding ﬁlm thicknesses t of MnFe2O4,
NiFe2O4, ZnnFe2O4 and half-half mixed NiFe2O4 and ZnnFe2O4ﬁlms
are 100 µm, 30µm, 30 µm, and 30µm.

. . . . . . . . . . . . . . . . . . . 110

Figure 6.6 Measured insertion loss and reﬂection loss of all four ﬁlm cases under

the DC magnetic bias of 2500 Oe. . . . . . . . . . . . . . . . . . . . . . . 112

xii

Figure 7.1 The electrically excited air-ﬁlled cavity resonator with rotating rods

(magnetic ﬁlm is not shown for clarity). R = 16 mm, T = 3 mm.

. . . . 116

Figure 7.2 Magnetic ﬁeld distributions in the XY plane: (a) simulated total mag-

netic ﬁled; (b) normalized theoretical Hφ using (7.2).

. . . . . . . . . . . 117

Figure 7.3 Measured ﬁlm surface roughness and edge proﬁle.

Inset:

fabricated

NiFe2O4 ﬁlm on a 2 mil LCP host substrate. . . . . . . . . . . . . . . . . 118

Figure 7.4 Fabricated samples: (a) metalized cavity parts with magnetic ﬁlm placed:
(b) fully assembled cavity resonator; (c) enlarged picture of corrugated
rod sleeve; (d) rotation position at θ = 15° and θ = 30°. . . . . . . . . . . 119

Figure 7.5 Measured and simulated S parameters of the air-ﬁlled cavity resonator.
Sim ideal/resistive: without/with the resistive solder ring. The dimen-
sions of the solder ring are shown in the inset.

. . . . . . . . . . . . . . . 121

Figure 7.6 Measured S11 and S21 of 7 tuned resonances at the corresponding rota-

tion angle θ compared to the air-ﬁlled case.

. . . . . . . . . . . . . . . . 122

Figure 7.7

(a)Tuning step and cos θ V.S. θ; (b) Qu and fr V.S. θ . . . . . . . . . . . 123

xiii

CHAPTER 1

INTRODUCTION

1.1 Motivation

In recent years, we have witnessed the fast expansion of the consumer electronics market,

including smart electronics which can perform a number of jobs just beyond their basic

function, such as smart phones, smart watches, and smart home speakers. The additional

functions oﬀered by those smart electronics are mostly achieved wirelessly through wireless

communication network or sensors using delicate electronics modules that are integrated

within the device housing. The emerging trend of industrial design often dictates that those

devices have to be aesthetically appealing, such as sleek and slim, compact and conformal,

low-proﬁle and lightweight; while at the same time the functions of the devices cannot be

compromised or may even require more functions to be included in the next production

generation. Therefore, such trends pose a critical challenge to the design of electronics

that all the circuits and components have to be highly integrated and miniaturized, but

with maintained or better performance. On the other hand, the cost of massive production

for such highly integrated circuits will increase as better fabrication precision is needed,

especially at high frequency range. However, to balance the ﬁnal price of the product and

the proﬁt margin, it is pivotal to greatly reduce the overall fabrication cost through low-cost

and eﬃcient manufacturing techniques.

As one of the most important parts for those smart electronics to perform wirelessly, the

RF front end in such applications has to be compact, robust, and power eﬃcient. The fact

that many services or functions in the devices operate at diﬀerent frequencies demands that

the RF front end circuits sometimes may also need to be reconﬁgurable, since it is diﬃcult to

put one or more independent RF front ends that operate at every corresponding frequency

band into the already tight device housing. Especially with the advent of 5G, the cellphone

1

antennas for example, have to be compatible with the 4G LTE low band around 1 GHz and

the 5G high band around 5.9 GHz, so that the antennas have to be designed either with

multi-band characteristics or with reconﬁguration capabilities. As a result, it is advantageous

to have one set of RF front end that can be reconﬁgured to operate at two or more frequency

bands. Again, as we mentioned previously, the overall package of the RF circuits need to be

miniaturized and lightweight, and with the fabrication cost to be greatly reduced. Therefore,

to comply with the current trend and need of small and high performance electronics, we

are aiming to investigate new methodologies to design and fabricate low cost, lightweight,

and reconﬁgurable RF circuits and components.

1.2 Problem Formulation

To achieve the aforementioned goal, we propose to use additive manufacturing technology

and nanomagnetic thin ﬁlms. Additive manufacturing is well known for fast prototyping,

low cost, and environmental friendliness. For example, additive manufacturing can print

a whole piece of complex 3D structure without laborious cutting and modeling compared

to the traditional manufacturing process. Recently, this technology has attracted extensive

focus on applying it to RF and microwave applications. In this thesis, we target the design

of high performance RF circuits and components by taking advantage of the extra design

freedom oﬀered by the additive manufacturing, as well as fabricating the prototype with

compact and lightweight package in a low-cost manner. Moreover, the RF reconﬁgurability

also may be achieved through designing and fabricating special ﬁxtures of the RF circuits

which are usually hard and costly and realizable by traditional forging process. However,

the major RF reconﬁgurability mechanism we propose here is through the magnetic tuning

using magnetic nanoparticles thin ﬁlm, as opposed to the commonly used electronically

tuning approach. Nevertheless, some important properties of the magnetic nanoparticles

thin ﬁlm have to be ﬁrst studied in order to control this unique magnetic tuning mechanism,

which involves the fabrication and characterization of the magnetic nanoparticle ﬁlms. In

2

addition, since the relative permeability µr of magnetic nanoparticles is usually greater than
unity, i.e., µr (cid:29) 1, depending on the speciﬁc characteristics such as types of the composites,
ﬁlm thickness, ﬁll factor of the magnetic nanoparticles ﬁlms, thus we can utilize this property

to miniaturize the RF circuits, while exploring the potential performance improvement and

tradeoﬀs as well. Ultimately, we would like to investigate a way to combine both additive

manufacturing technology and the magnetic nanoparticle thin ﬁlms to design and fabricate

low cost, compact, and high performance RF circuits and components.

To elaborate the above statements concisely, we formulate our problem in this thesis in

the following points:

• Develop novel RF circuits and components by taking advantage of extra design freedom

oﬀered by additive manufacturing technology.

• Investigate the fabrication process for speciﬁc RF applications using stand-alone or

combined additive manufacturing technologies to achieve satisfying performance.

• Characterize and fabricate the magnetic nanoparticle thin ﬁlms for high frequency RF

applications.

• Utilize the magnetic nanoparticle thin ﬁlms to achieve RF reconﬁgurability and circuit

miniaturization.

• Implement both additive manufacturing technology and magnetic nanoparticle ﬁlms

to obtain low cost, lightweight, miniaturized, and reconﬁgurable RF circuits and com-

ponents.

• Discuss the potential, limitation, and tradeoﬀ of applying both additive manufacturing

technology and magnetic nanoparticle ﬁlms onto RF applications

3

1.3 Thesis Statement

This dissertation consists of a series of studies that allow for the development of compact

and miniaturized, lightweight and reconﬁgurable RF components. The main contributions

of the present work include: investigation of using one or two additive manufacturing tech-

nologies and liquid crystal polymer substrate to develop novel RF circuits including ﬁlters,

transmission lines, resonators, and antennas; fabrication and characterization of magnetic

nanoparticle ﬁlms; the development of miniaturized and reconﬁgurable RF circuits with low

cost and lightweight packages combining both additive manufacturing technology: and mag-

netic nanoparticle thin ﬁlms, which provides an unique approach and expanded vision of

tackling the problem of smaller, smarter, and cheaper electronics.

Thesis Statement: By applying the additive manufacturing technology onto various RF

applications and characterizing the magnetic nanoparticle thin ﬁlms, we intend to propose

novel RF circuit design and eﬃcient fabrication process to signiﬁcantly reduce the overall

fabrication cost, and shrink the circuit package size while obtaining the RF reconﬁgurability.

1.4 Anticipated Research Contributions

This dissertation will directly address each of the research problems listed in Section 1.2.

Various novel RF components utilizing one or two additive manufacturing technologies were

successfully designed, fabricated and measured, pushing the boundary of applying additive

manufacturing technology into RF applications. By fabricating magnetic nanoparticle ﬁlms

using various composites and low cost chemical methods, we are able to characterize the ﬁlms

to study their important properties for high frequency RF and microwave applications. Based

on the above two foundations, we are capable of designing RF components with very low cost,

compact and miniaturized packaging, enhanced RF performance and reconﬁgurability. The

theoretical analysis are also carried out to provide some insightful understanding in terms

of the physics behavior, design potentials, and performance tradeoﬀs. Therefore, it paves

4

the way to implement such conﬁguration and technology in monolithic RF and microwave

applications.

1.5 Dissertation Outline

The remainder of this dissertation is organized as follows; in Chapter 2 the background

information on additive manufacturing technology in terms of its history original, current

development, types of implementation techniques, and especially the research on applying

it onto RF applications is provided. Similarly, we also discuss the usage of magnetic mate-

rials including ferrites and ferrites based magnetic thin ﬁlms in RF and microwave circuits

and analyze the potential and limitation of using magnetic nanoparticles thin ﬁlms in such

applications operating at much higher frequency.

In Chapter 3, we demonstrate a number of RF components that are designed by taking

advantage of the extra design freedom oﬀered by the additive manufacturing technology.

They are fabricated using one or two additive manufacturing technologies. The measure-

ment results show satisfying performance compared to the counterparts fabricated by the

traditional approaches, while the overall fabrication cost is signiﬁcantly reduced. Also, RF

reconﬁgurability can be achieved by designing special ﬁxtures in certain applications which

also exhibits the great potential of using additive manufacturing technology to implement

high performance RF circuits and components. Based on the experimental results, we also

discuss the limitations of such technology from a practical point of view, especially when it

is intended to be used in a high frequency range.

Chapter 4 ﬁrst studies several properties of the magnetic materials that are partic-

ularly important for RF and microwave applications operating at high frequencies. Then

we demonstrate the magnetic thin ﬁlms that are made with various composites through

two novel and low cost chemical methods. The characterization of the fabricated magnetic

nanoparticle ﬁlms proves the eﬃciency of the chemical methods thus enabling us to make

much thicker ﬁlms, and granting us the vast potential of using such ﬁlms to make miniatur-

5

ized and reconﬁgurable RF circuits and components.

Based on the foundations of the previous two chapters, in Chapter 5, we propose a

thrust of combining additive manufacturing and magnetic nanoparticle thin ﬁlms to imple-

ment high performance and reconﬁgurable RF components with low cost, lightweight, and

miniaturized packages. In this chapter, we demonstrate a number of RF components using

unique design conﬁgurations to achieve our goal, also the fabrication process is optimized

based on the practical considerations oﬀered by Chapter 3 and our extensive experiment

trials. Additionally, theoretical analysis is performed to provide physical explanation on the

measured results of our fabricated samples, thus we can obtain some insightful understand-

ing and design guideline of using such conﬁgurations. Chapter 6 demonstrates a bandstop

ﬁlter using the similar topology shown in Chapter 5, such that a tunable stop band can

be achieved. Chapter 5 and Chapter 6 showcases the tunable and high performance pla-

nar RF devices with low cost and light-weight packages using additive manufacturing and

magnetic materials. Chapter 7 presents a 3D printed spherical cavity resonator with ﬁne

tuning using magnetic materials. Additionaly, the theoretical analysis is carried out to un-

derstand the tuning mechanism, fabrication process, measurement results, and comparison

with the state-of-the-art. Finally, Chapter 8 draws the conclusions and summarizes the

contributions of this dissertation.

6

CHAPTER 2

BACKGROUND

Originally developed as a more eﬃcient way to build objects with complex shapes, additive

manufacturing technology has now become an alternative approach to manufacturing me-

chanic objects, compared to the conventional machining, casting, and forging process, where

materials are removed from a stock item or poured into a mold and shaped by means of dies,

press and hammers. Therefore, additive manufacturing uses less material and thus produces

less waste. Although it is commonly used in automobile and aerospace industries to make

mechanic parts, recently people have explored using such technology to design and fabricate

RF circuits and components. However, applying such technology in high frequency range

RF and microwave applications is still in its the early stages.

In the 1940s, ferrites were ﬁrst investigated to be used in microwave applications, espe-

cially for spinel and hexagonal ferrite families, and because of the intense interest from the

military in the 1950s and 1960s using ferrite phase shifters in phased array radar, research

on the use of ferrites materials in microwave applications underwent a vast transformation.

By the 1980s, conventional ferrites used in microwave applications reached its maturity. Due

to the excessive magnetic loss, these conventional ferrite materials found their applications

up to Ka-band. In the beginning of the 20-th century, people were able to synthesis ferrite

based thin ﬁlms, which enabled their applications operating at much higher frequency. Re-

cently, thin ﬁlms made of spinel magnetic nanoparticles were demonstrated in several RF

applications with operating frequencies at several GHz, which shows great potential in using

such ﬁlms in high frequency RF and millimeter wave applications.

7

2.1 Background of additive manufacturing technology

It is well known that additive manufacturing (AM) technology has a number of advantages

compared to the traditional fabrication process such as milling and machining. The AM

technology usually builds 3D objects in an additive layer-by-layer manner, which is why it is

often synonymously called 3D printing. The materials used for printing can be plastic, metal,

rubber, and possibly any material that would ﬁt the printer conﬁguration. In retrospect, the

AM technology was originally investigated by Hideo Kodama of Nagoya Municipal Industrial

Research Institute in 1981, where the three dimensional objects were printed using photo-

hardening thermoset polymer and then cured by UV exposure. A more close-to-modern 3D

printing technology was developed at Massachusetts Institute of Technology in 1993, where

the inkjet printer head was employed in a power bed processing. The founding of several

corporations based on this technology in 1993 demonstrates the commercialization of AM

technology. With decades of development, the AM technology has merged into several types

of techniques, each of them having its own unique way of executing the 3D printing process

which we are going to introduce in the following subsections.

2.1.1 General 3D printing process

Although diﬀerent AM techniques have their unique printing characteristics, the general 3D

printing process involves several common phases.

2.1.1.1 Modeling

First, the 3D object that is intended for printing has to be modeled in a computer-aided

design (CAD) software. Although other methods such as 3D scanner, plain digital camera,

and photogrammetry software may be used for the same purpose, CAD ﬁle is preferable as it

allows for modiﬁcation before printing is executed, thus it is more ﬂexible and use-friendly.

The format CAD ﬁle is commonly saved as stereolithography ﬁle format (STL) which stores

8

the data based on triangulations of the surface of CAD models. This type of ﬁle has relatively

large ﬁle size since it contains topology optimized parts and lattice structures due to the large

number of surfaces involved. In 2011, a new ﬁle format called Additive Manufacturing File

Format (AMF) was introduced where data is based on curved triangulations and thus is

more suitable for AM and has accordingly smaller ﬁle size. However, due to practical issues,

we are using the STL format CAD ﬁle since it is still widely used in most 3D printers.

2.1.1.2 3D Printing

The the veriﬁed error-free STL ﬁle is then imported to a special ”slider” software which

basically converts the 3D model into a stack-up of multiple thin layers, which provides a sort

of printing guide for the 3D printer since as we mentioned before, the 3D printing process

builds the objects in a layer-by-layer manner. Then the generated guide ﬁle is read by the

client 3D printer to execute the printing process. Since the actual speed of diﬀerent printers

may vary, the consumed printing time is diﬀerent for similar objects. In general, the lead

time of a 3D printing process is signiﬁcantly less than the conventional fabrication process,
for example it usually takes at most a day to 3D print a batch of ten 50 mm × 50 mm squares

in house whereas it may take several weeks or months for out-sourced foundry manufacture.

Hence, 3D printing is much more eﬃcient for fast prototype.

2.1.1.3 Polishing

After the printing process is complete, the object is taken out from the printing tray, and

depending on the types of 3D printing techniques, the objects may need to be cleaned to

eliminate the supporting materials in some structures like holes and bridges. Usually this

cleaning process can be easily done by ﬂashing away the supporting materials using water.

Most of the 3D printed objects have to thereafter undergo a polishing process to get better

surface ﬁnishing. Especially for RF and microwave applications, the surface roughness will

strongly aﬀect the surface resistance, and given that the skin depth at high frequency tends

9

to be small, poor surface roughness will result in the surface current being heavily attenuated.

In general, the surfaces of the printed objects can be polished using very ﬁne sandpapers or

machine sanders, however, more advanced approaches must be developed in order to further

improve the performance of the applications used at a much higher frequency range. It also

has to be pointed out that in some cases where the 3D printed surface is very thin, for

example a 3D printed dielectric substrate which only has several tenths of micrometers, the

polishing process handled by human hand or simple machine will most likely create inevitable

large concave area on the horizontal substrate surface, which may end up with even worse

negative eﬀects on the surface current and thus the RF performance. Therefore, in such a

scenario, (i.e., a very thin or fragile 3D printed object), we usually do not polish the surface

in order to maintain the geometric integrity of the 3D printed piece. Instead, alternative

methods may be applied to address the surface roughness to create a smoother surface ﬁnish,

such as coating the surface with another thin layer of material to ﬁll the saw-like gaps of the

rough surfaces, or applying special chemical etching solution onto the surface to eliminate

such saw-like gaps. When the surface roughness will not heavily aﬀect the RF performance,

no polishing is needed in this case to maintain the simplicity of the fabrication process.

2.1.1.4 Metalization

Ideally, 3D printing can print any material that is ﬁt for the printer, however, current

development only allows for using limited a number of materials, such as plastic, rubber,

and metal powder based ﬁlament. In general , for RF and microwave applications, researchers

often choose to ﬁrst print the circuit or component structures using non-metal materials, such

as plastic waveguides and plastic horn antennas, then after surface polishing, metalize the

needed surfaces of the objects. This metalization process can be performed either through

sputtering or electroplating, or combineinf both. For example, in a combined manner, the

target surfaces are ﬁst sputter with 0.5 - 1 µm of Titanium which serves as the seed/adhesion

layer, then the object is soaked into the electrolyte solution to electroplate the target surfaces

10

with desired metal, such as copper or aluminum. The thickness of the ﬁnal metal layer can be

controlled by adjusting the parameters of electroplating process including processing time,

current amplitude, and electrolyte density. As we previously stated, metal can also be 3D

printed using power based ﬁlament, or using nanoparticle based solution. However, the

conductivity of those materials becomes the most critical issue for the RF application usage.

For power based ﬁlament, the conductivity is currently 50% of copper at best, whereas

due to the aforementioned surface roughness issue and actual object geometry, the actual

conductivity will be lower. While for the nanoparticle based solution, special curing process

has to be carried out to get good conductivity. The best reported and most promising

conductivity is 80% silver using a heat curing process. However, such a curing process has

several practical limitations, because it acts a major barrier in cases where low thermal

temperature dielectric materials are used with the printed silver.

2.1.2 Types of AM techniques

For the past few decades, AM technology has merged into several types of techniques which

feature diﬀerent printing characteristics in term of materials, ways of constructing models,

and printing time. In the following, we will focus on some of the popular techniques and

give a brief description for each of them.

2.1.2.1 Stereolithography (SLA)

This technique builds objects in a pool of photopolymer resin. Using the sliced STL ﬁle,

the printer laser beam directing into the pool of resin traces the cross-section pattern of the

model for each sliced particular layer, then prints and cures layer-upon-layer until the whole

structure of the object is ﬁnished. Specialized material may be needed to add support to

some model features. Models can be machined and used as patterns for injection molding,

thermoforming or other casting processes. Figure 2.1 shows an SLA printer produced by

FormLabs.

11

Figure 2.1: Form 3L SLA 3D printer [1]

2.1.2.2 Fused Deposition Modeling (FDM)

Similar to SLA, FDM constructs objects layer by layer using thermoplastic from the very

bottom up. The so called thermoplastic transforms to liquid upon the application of heat

and solidiﬁes to a solid when cooled. First, the printer heats the theormplastic until its

melting point and extrudes it through a nozzle on a printing platform or tray, on a predeter-

mined pattern determined by the 3D model and Slicer software. Just like the SLA process,

the nozzles trace the cross-section pattern for each particular layer with the thermoplastic

material hardening prior to the application of the next layer. The process repeats until the

build or model is completed. In addition to thermoplastic, a printer may extrude support

materials too, which can be ﬂashed away using water. Figure 2.2 shows a professional grade

FDM printer develped by Stratasys Inc.

2.1.2.3 Digital Light Processing (DLP)

DLP is another popular 3D printing process that is widely used in the production of pro-

jectors. Larry Hornbeck of Texas Instruments, ﬁrst invented this type of technique in 1987.

Due to the fact that it utilizes digital micromirrors laid out on a semiconductor chip, DLP’s

12

Figure 2.2: Stratasys Fortys 450mc 3D printer [2]

applications include mobile phones, ﬁlm projectors, and many 3D printing applications.

2.1.2.4 Selective Laser Sintering (SLS)

Quite similar to SLA, SLS builds the objects in a layer-upon-layer manner using a laser as

power supply to sinter the power based plastic, metal, ceramic or glass, to form strong 3D

printed objects. However, no supporting material is needed for SLS as the build is supported

by unsintered material.

2.1.2.5 Polyjet

Developed by Stratasys Inc, Polyjet technology produces high quality 3D objects with more

realistic ﬁnishes compared to the aforementioned techniques. PolyJet builds by jetting layers

of liquid photopolymer at 27 micronmeters (0.00106”) layer by layer as a UV light cures

simultaneously. PolyJet can print rigid and ﬂexible materials in a single build to create over-

molded parts without tooling. The technology delivers parts with shore hardness ranging

13

from 30A-95A. Because PolyJet does not require hard tooling to deliver an over-molded part,

it is frequently used for prototypes requiring an elastomeric surface, such as grips or buttons,

or for testing material hardness. PolyJet can also print parts with a variety of ﬂexibilities,

from rubber-like parts to those that have a more rigid feel. The beneﬁts of polyjet include

the following:

1. Creating smooth, detailed prototypes that convey ﬁnal-product aesthetics.

2. Producing accurate molds, jigs, ﬁxtures and other manufacturing tools.

3. Achieving complex shapes, intricate details and delicate features.

4. Incorporating the widest variety of colors and materials into a single model for unbeat-

able eﬃciency.

The Polyjet printer we used is the Connex Objet350 from Stratasys, which is shown in Figure

2.3

Figure 2.3: Stratasys Connex Objet350 3D printer [2]

14

2.1.2.6 Aerosol Jet Printing (AJP)

Unlike the above techniques, AJP achieves the 3D printing process by directly printing

the objects using a dense aerosol of materials, such as silver nanoparticles and polyimide

materials. First, the printing materials are usually diluted with water or other solution,

then they are atomized using either pneumatic atomizer or ultrasonic atomizer. Then the

atomized materials are mixed with carrier gas, for example Nitrogen gas, to create a dense

aerosol into the printing nozzle. The sheath gas injected into the nozzle helps to shape

and focus the aerosol, delivering high resolution of the printed trace.

In addition, some

advanced AJP printed, for example the Optomec AJ 5X system shown in Figure 2.4, the

printing platform can be rotated during the printing process. Along with the X,Y,Z moving

directions of the printing head, such system oﬀers higher ﬂexibility that allows for printing

extremely complex objects with much higher precision. The advantages of AJP technique

Figure 2.4: Optomec AJ 5X system [3]

are the following:

1. Feature resolutions from 10 microns to millimeters.

2. Conformal printing on non-planar and 3D surfaces.

15

3. Print interconnects and active/passive components.

4. Print using commercially-available materials and bio-materials Print bio-materials.

5. Print on plastic, ceramic, and metallic substrates.

6. Scalable to support high-volume production requirements.

Because of the above advantages, the AJP technique is ideally suited for 3D Printed

Electronics applications such as fully printed antennas, sensors, and Molded Interconnect

Devices (MIDs).

2.1.3

3D printing in RF and microwave applications

Since 2005, researchers all around the world have worked on applying 3D printing technology

into various RF applications. Among these, SLA, Polyjet, and AJP techniques are the most

widely used 3D printing technologies.

In general, this work involves the characterization

of the materials used for 3D printing, fabrication of lightweight, low cost RF components

compared to the traditional counterparts, novel design for ultra-low loss, compact passive

RF circuits and sensors, integration of active and passive components using 3D printed

interconnects and packages.

One of the important purposes of applying 3D printing technology in RF application is

to reduce the cost and weight of the fabricated circuits or packages while obtaining close or

the same performance of the conventional counterparts. For this purpose, people have found

that the fabrication process usually play a pivotal role. Except AJP, all the other 3D printing

techniques employed in RF applications rely on the printed dielectrics like plastic to form

the structural object for the RF circuits and components. For example, a waveguide is ﬁrst

realized by 3D printing the pipe-like structure using either SLA or Polyjet, then the inner

surfaces of the structure have to be metalized to actually achieve the RF functionality of the

waveguide. During this process, the surface roughness of the inner walls of the waveguide

which is dictated by the resolution of the printer heavily aﬀects the RF performance of

16

the waveguide, i.e., the transmission and reﬂection coeﬃcient. Depending on the actual

metalization process, the geometric shape of the waveguide may be aﬀected, for example

warpage due to the heat during the metal sputtering process, which will cause the wave

propagation characteristics to be changed such as the modes of the wave and the cutoﬀ

frequency of the waveguide. For this reason, researchers have experimented using a number

of approaches to address this issue.

As for the AJP, although great potential can be realized with higher resolution and

printing capabilities, the most critical problem which hinders its applicability into production

realization still comes from the fabrication process. Unlike the SLA, FDM, and Polyjet which

can print a relatively large object in a short amount of time in a layer-by-layer fashion, AJP

features tens of micrometers of ultra high resolution single printed trace, it is unrealistic to

print bulk geometry using AJP, but rather, to print extreme compact planar RF circuits,

such as transmission lines, interconnects, and millimeter wave antennas and power dividers.

Also, due to the nature of aerosol, the issue of overspray is inevitable which poses a major

problem when for example, the printed circuit is down to 20 µm and thus is close to the

AJP printing resolution range. On the other hand, the surface roughness still exists however

is more controllable since such surface roughness is actually determined by the extent of

overlapping between adjacent printed traces. Therefore, a careful study and experiment

in terms of the amount of materials fed to the printer, the printing speed, the clearance

between the print head and the printing platform, and so on has to be carried out to ﬁnd

the best conﬁguration of the system for achieving the best surface roughness. Besides, since

AJP can easily print the metal using the uncured silver nanoparticles, it is desirable to ﬁnd

the best curing process to achieve high conductivity to ensure good RF performance of the

printed circuitry. Currently, two major curing methods, UV exposure and heat curing, are

tested to achieve moderate to high conductivity ranging from 20% - 80%, however, practical

limitations may prevent them from achieving the best tested conductivity. For example,

when the silver nanoparticles are printed on top of some dielectrics with low glass transition

17

temperature, the heating method which usually requires very high curing temperature will

be less eﬀective.

Therefore, an important research direction is how to minimize the negative eﬀect caused

by the fabrication process on RF performance through better utilizing 3D printing technolo-

gies to propose new design, and combining other eﬀective techniques to further improve the

overall RF performance. It will be interesting and challenging to combine one or more 3D

printing techniques to obtain higher printing capabilities, in particular for some designs that

are theoretically beneﬁcial but hard to realize using conventional fabrication process or just

a single 3D technique. By doing so lower fabrication cost and better RF performance are

expected to be achieved compared to single 3D printing technique strategy.

2.2 Background of magnetic materials in RF applications

It is widely known that magnetic materials in RF applications are usually referred to mi-

crowave ferrite devices because of extensive research that was done from the 1940s to 1980s.

Microwave ferrite devices permit the control of microwave propagation by a static or switch-

able DC-magnetic ﬁeld. The devices can be reciprocal or nonreciprocal, linear or nonlinear,

and their development requires a knowledge of magnetic materials, electromagnetic theory,

and microwave circuit theory. Unlike a magnetic metal, a ferrite is a magnetic dielectric that

allows an electromagnetic wave to penetrate the ferrite, thereby permitting an interaction

between the wave and magnetization within the ferrite. However, due to the excessive mag-

netic loss at high frequency and the vast development of silicon technology, the traditional

ferrite materials are less used in the >600 MHz high frequency applications. Since 2004,

ferrite based magnetic thin ﬁlms have been developed and characterized to have high de-

magnetization which enables them to be used at GHz range even under self-biased condition.

This implies the magnetic nanoparticle thin ﬁlms have great potential in RF and millimeter

wave applications.

18

2.2.1 Traditional microwave ferrite devices

2.2.1.1 Circulators and isolators

In microwave systems that use a single antenna aperture for both sending and receiving,

the function of circulators is primarily to facilitate the routing of outgoing and incoming

signals to the transmitter or receiver as appropriate. This is illustrated in Figure 2.5, which

shows a schematic diagram of a phased-array antenna module using a switchable circulator

in addition to an ordinary (nonswitchable) circulator. The module also contains two T/R

switches. Their function can also be satisﬁed by two further circulators, but switches are

generally preferred in these locations because of cost and size considerations. Circulators

and isolators are made possible by the nonreciprocal character of the microwave behavior of

ferrites (and certain other materials).

Figure 2.5: Schematic diagram of phased array antenna module [4]

In microwave systems that use separate antenna apertures for sending and receiving, non-

reciprocal devices are required in order to control the voltage standing-wave ratio (VSWR)

seen by the high-power ampliﬁer at the output of the transmitter, and also to control undesir-

able reﬂections from the receive antenna array. In this case, isolators rather than circulators

are required. These isolators can be made by connecting a matched load to one of the ports

of a circulator, but under some conditions, a “dedicated” isolator design may be preferable.

19

2.2.1.2 Ferrite phase shifters

Ferrite phase shifters generally take advantage of the ability to control the permeability of

a waveguiding medium to vary the phase velocity of a microwave signal passing through

it. In most cases, the change in permeability changes the phase velocity and, therefore, the

insertion phase. In a few cases, such as in the rotary ﬁeld phase shifter, the orientation of

the internal magnetization controls the phase shift. A meander line can be used to enhance

the phase shift by generating a circularly polarized RF magnetic ﬁeld around the center

conductor, as shown in Figure 2.6. The main use of ferrite phase shifters is in phased-array

antenna systems. The direction of a beam radiated by an array of elements in a linear array

can be steered by inserting a diﬀerent phase shift in each radiators feed. A phase control

range of only 360o is required, but that full 360o range is required regardless of the scan

angle.

Figure 2.6: Superconducting meander-line phase shifter [4]

2.2.1.3 YIG ﬁlters

In Grenoble, France, in 1956, Bertaut and his colleagues synthesized the rare-earth iron

garnets and analyzed their magnetic characteristics. These ﬁndings were quickly replicated

20

and reﬁned by Geller and Gilleo at BTL, and then expanded to form a base of crystallographic

and magnetic data for most of the known garnet compounds, the most important of which

was yttriumiron garnet (YIG). By the end of the decade, both the spinel and garnet systems

were becoming available to microwave device engineers. Relying on the FMR of the single

crystal YIG, electronically tunable microwave ﬁlters can be achieved by integrating the

YIG sphere into the ﬁlter package. Basically, because of the narrow linewidth of YIG, i.e.,

∆H<0.5 Oe, yields the potential unloaded quality factor Q = f /(γ∆H) >5000 at 10 GHz,

where γ = 2.8 MHz/Oe is the gyromagnetic ratio. The negative permeability shown by

ferrite materials at frequencies higher than the FMR frequency enables wave propagation

in structures with arbitrary small cross section over limited frequency ranges. These are

referred to as Magnetodielectric Surface Waves (MSW) since the electric-ﬁeld contribution

is negligibly small. The theory and experiments were performed specially on planar epitaxial

YIG ﬁlms spaced from conducting ground planes, an example of YIG ﬁlm based MSG ﬁlter

is shown in Figure 2.7.

Figure 2.7: Simple MSW device with microstrip transducers. Depending on design, this can
provide time delay and ﬁltering functions [4]

21

2.2.1.4 Frequency selective limiter

In a high-signal density environment, receivers and signal processors are easily overloaded in

this situation. An ordinary diode limiter placed in front of a wide-band receiver to limit the

high-level signals, will attenuate all signals by the same amount. A frequency selective limiter

(FSL) is deﬁned as one that can attenuate strong signals without attenuating other weaker

signals present simultaneously, as illustrated in Figure 2.8. Most microwave FSLs utilize the

Figure 2.8: Frequency response of an FSL, and input and output power spectra, from multiple
weak and strong signals, showing the frequency selectivity inherent in YIG limiters [4]

frequency-selective nature of a magnetized ferrite. Above a critical RF magnetic-ﬁeld level,

the spin precession angle can increase no further, and coupling to higher order spin waves

begins to grow exponentially. Energy is eﬃciently coupled to spin waves at approximately

one-half the signal frequency and then into lattice vibrations (heat) in the ferrite. The

critical ﬁeld strength is proportional to the spinwave linewidth of the ferrite, and because

of the narrow linewidth of YIG, it can be used in high power FSLs severing as the receiver

protector to prevent burnout. Ferrite FSLs have been achieved in a variety of transmission

22

line structures, for example stripline using single crystal YIG ﬁlm, while another type of FSL

is based on the nonlinear excitation of spin waves in structures that can propagate MSW.

2.2.2 RF devices using magnetic nanoparticle ﬁlms

The traditional ferrite materials have magnetic loss due to the lower FMR frequency which

caps them to be used at RF applications at very high frequencies, especially for antennas.

Recently, the ferrite based magnetic nanoparticle based thin ﬁlms have been developed and

characterized to have high magnetization saturation, which results in the FMR frequency

can be as high as several GHz. This implies such magnetic thin ﬁlms can be used in RF and

millimeter wave applications. In the past decade, researchers have used such thin ﬁlms in

antennas and ﬁlters to achieve circuit miniaturization and performance enhancement. For

example, under self-biased conditions, it has been demonstrated that the patch antennas can

be ﬁrst fabricated on a regular PCB substrate, then multiple layers of magnetic ﬁlms are

deposited on top of the antenna acting as the superstrate, which can shift down the resonant

frequency as well as increase the impedance bandwidth and eﬃciency. The thickness of single

layer ﬁlm is typically 2 µm, and each ﬁlm layer is constructed as the magnetic nanoparticles

are sandwiched in between two protective polymer thin layers. Depending on how much

magnetic ﬁlm layers are deposited, the resonant frequency can be shifted down to tens of

MHz, while the antenna eﬃciency can be enhanced by at most 40%. However, the magnetic

ﬁlm layer deposition process is rather complicated which prevents such method from making

thick enough magnetic ﬁlms, thus large resonant frequency shift is hard to achieve. In other

words, the potential of miniaturizing the overall package cannot be realized. On the other

hand, by applying the DC magnetic bias, the FMR frequency of the magnetic ﬁlms can

be tuned. Using the similar superstrate conﬁguration, a transmission line based tunable

bandstop ﬁlter was demonstrated to achieve a relatively wide tuning range by applying

moderate DC magnetic bias. Nevertheless, the reported bandstop ﬁlter shows large insertion

loss within the operation frequency range, which again due to the magnetic loss of the ﬁlm

23

and the dielectric loss of the PCB substrate.

Therefore, to better utilize the magnetic nanoparticle ﬁlms in high frequency RF appli-

cations, three important issues have to be addressed.

1. Fabricate magnetic ﬁlms using low cost and eﬀective method to have large thickness,

good ﬁll factor, and acceptable surface roughness

2. Characterize the fabricated ﬁlms to determine important properties of the ﬁlms includ-

ing FMR frequency, magnetization saturation, and the relative permeability

3. Propose better design to integrate the magnetic ﬁlms into RF applications to improve

the RF performance while obtain compact, robust, and minimized overall package

2.3 Summary

This chapter discussed the backgrounds of both additive manufacturing and magnetic ma-

terials in RF and microwave applications. We reviewed the history and development of

both technologies, introduced several types of the AM techniques and the ferrite microwave

devices. We also addressed their advantages and challenges when applied to RF devices.

For additive manufacturing technology, we would like to propose new design for mi-

crowave devices by utilizing the expanded printing capabilities and extra design freedom

through combining two or more additive manufacturing and other low cost techniques, thus

a cheap, robust, and compact RF components can be realized. For the magnetic materials,

we are targeting the fabricating and characterizing of thicker magnetic ﬁlms and proposing

new design of miniaturized, reconﬁgurable RF devices by combining the magnetic materials

with the additive manufacturing technology.

24

RF DEVICES USING ADDITIVE MANUFACTURING TECHNOLOGY

CHAPTER 3

In this chapter, we proﬁled a number of RF devices designed and fabricated using additive

manufacturing technology with one and two 3D printing techniques to achieve comparable

performance with traditional counterparts but lower fabrication cost. Also, by incorporat-

ing with other low cost techniques, we are able to improve the RF performance and get

reconﬁgurability for certain devices.

3.1 Ultra-low loss lowpass ﬁlter using improved air substrate

We know that RF circuits such as ﬁlters, antennas fabricated on regular dielectric PCB

boards suﬀer from dielectric loss, which will cause series associated problems, such as the

reduced overall quality factor, higher insertion loss, and degraded antenna eﬃciency. One of

the eﬀective ways to address such problems is by suspending the planar circuits above the

ground plane such that the circuits can be treated fabricated on top of an air substrate. Due

to the ultra-low loss of this air substrate, the aforementioned problems can be alleviated

by minimizing the dielectric loss. AM technology oﬀers the great ﬂexibility for achieving

such air substrate since it can directly print out the ﬁxtures from the design concept that is

normally hard or costly to achieve using conventional techniques. For example, one of the

recent strategies involves printing the cavity or well-like substrate structure using dielectric,

for example plastic, and metalizing the inner walls afterwards, then the circuits pattern

is 3D printed in a trench manner on another piece of planar dielectric boards, such that

the pattern is “etched” out on it. The circuits pattern is metalized using sputtering and

an electroplating process. Finally, these two pieces are snapped together using a lego-like

self-alignment structure face to face to form the air substrate.

However, in such conﬁgurations, three issues may hinder the ﬁnal performance of the cir-

25

cuits. First, the eﬀective dielectric loss tangent is still partially aﬀected by the top dielectric

piece where the circuit pattern is “etched” onto. Especially when the top dielectric piece oc-

cupies a greater portion of the air substrate, the eﬀective loss tangent is deviated from zero,

but close to that of the dielectric piece; and the eﬀective dielectric constant is close to that

of the top dielectric piece instead of unity. Higher dielectric loss tangent will increase the

loss, while the eﬀective dielectric constant deviating from unity will make the circuit design

less accurate, otherwise additional characterization process of the air substrate needs to be

performed. Secondly, the metalization process may cause warpage of both the bottom and

top dielectric pieces which will cause the thickness of the air substrate to be non-uniform,

which will negatively aﬀect the wave propagation characteristics within the air substrate.

For example, for a simple microstrip transmission line, the non-uniform air substrate will

cause mismatch of the line since the characteristic impedance of the transmission line is

related to the substrate thickness. Third, from a practical point of view, very ﬁne ﬁnishes

of both dielectric pieces have to be achieved in order to reduce the surface resistance on the

metal parts, i.e., the ground plane and the circuits pattern. Nevertheless, it would still be

less comparable to the quality of the regular PCB board in terms of the surface roughness.

Therefore, we proposed a new design using both AM technology and thin membrane

process. Basically, a very thin layer of LCP substrate is suspended on top of the metalized

cavity ground plane to form the air substrate. Because the LCP is very thin, typically as

thin as 1 mil, the eﬀective dielectric loss tangent is very close to zero, and the eﬀective

dielectric constant creates a very close unity. Due to this improvement, we can ensure the

actual fabricated circuit performance is as close as possible to the designed one as we assume

the eﬀective dielectric constant of the air substrate is unity. Also, the circuits pattern can

be easily etched out on the LCP substrate using standard photolithography process with-

out worrying about the surface roughness issue encountered during the 3D printing process.

Additionally, the engineered coeﬃcient of thermal expansion (CTE) of LCP substrate is the

same as copper, the fabricated circuits can have better heat and power handling capabili-

26

ties when compared to the “etched” pattern on the dielectric boards which generally have

diﬀerent CTE from copper.

3.1.1 Low pass ﬁlter design

A common method to realize high frequency low pass ﬁlters are stepped-impedances using

microstrip transmission line structures [5]. Here a ﬁve element maximally ﬂat low-pass ﬁlter

design with a cutoﬀ frequency of 3 GHz, impedance of 50 Ω, and 35 dB of insertion loss at

4 GHz was designed using the stepped-impedance design procedure. Finite-element method

(FEM), with the aid of ANSYS HFSS (High Frequency Structure Simulator), was used to

optimize the design layout of the complete structure. Figure 3.1 shows the schematic of

top view of the ﬁlter implemented using the stepped impedance design. The dimensions of

various microstrip lines used for designing the ﬁlter are shown in Table 3.1. For the foil,

thin LCP, from Rogers Corporation, with a thickness of 0.1 mm and copper thickness of

18 µm was used. Air cavity thickness between the ground conductor and the LCP layer is

1.524 mm. This conﬁguration provides an eﬀective dielectric constant value of 1.06. Holes in

the LCP layer are also introduced to achieve alignment between the diﬀerent layers during

assembly. The ﬁlter is connected to 50 Ω transmission lines and in turn to coaxial cables at

the input and output ports. These were included in the simulations.

Table 3.1: Design parameters of the Lowpass ﬁlter.

Parameter
mm

A

B

C

D

E

F

68.87

0.96

3.83

14.83

55.07

115.62

3.1.2 Fabrication process

Figure 3.2 shows the schematic of the low-pass ﬁlter implemented using a combination of

thin-ﬁlm process and 3D printing. The top supporting structure (I) and cavity for ground

plane (III) are fabricated using a 3D printer; the ﬁlter was patterned on LCP with no ground

27

Figure 3.1: Layout of the lowpass ﬁlter design

plane (II) using standard optical lithography. The structures were stacked together to form

a ﬁlter as shown in Figure 3.3. The spacer in the 3D printed cavity (III) creates an air

gap between the LCP and the ground plane. All 3D printed components are printed using

a professional-grade commercially available 3D printer (Objet Connex350) using a photo-

polymer resin. The material used here for 3D printing is called verowhite which has a

dielectric constant r= 2.8 at the frequency of interest and a loss tangent (tan δ) of 0.04.

After 3D printing the top and the bottom layers, both were blanket coated on one side.

First a 50 nm of Titanium (Ti) is sputtered followed by a 1 µm thick copper (Cu) layer.

Ti is used here to promote adhesion between the verowhite and the copper. The low pass

ﬁlter is patterned on a thin LCP (Roger 3850) substrate using standard lithography process.

LCP is used here as a supporting material and thus Cu on back side (ground plane) has to

be removed. First a thin layer of positive resist S1813 is coated on LCP followed by UV

exposure and development. Then, Cu is selectively etched using sodium persulphate : water

(1:4) solution at 40o C. During this etching process, copper on the back side of LCP is also

etched. A CO2 laser is then used to open alignment holes in order to assemble the ﬁlter

structure. For opening of the holes using the laser, holes in the Cu layer were simultaneously

patterned during etching of the ﬁlter structure. Copper acts as the shadow mask for the

28

laser to etch through the LCP layer.

Figure 3.2: Schematic of low-pass ﬁlter implemented using combination of thin-ﬁlm process
and 3D printing

Figure 3.3: Final assembled ﬁlter with all three layers combined together in a Lego-like
approach

3.1.3 Measurement results

Measurements for the transmission and reﬂection coeﬃcients were done using an Agilent

N5227A PNA network analyzer. Figures 3.4 and 3.5 show the simulated and measured

results for reﬂection (S11) and transmission (S21) ratios, respectively. The measured results

29

match closely over a wide frequency range. Measured isolation of 32 dB is achieved at 4 GHz

as opposed to design value of 35 dB. There is a small diﬀerence in cutoﬀ frequency which

can be contributed to some discrepancy in the fabricated cavity size. The gap between the

LCP and the ground plane is slightly lower than the designed value. In addition, the extra

loss in the measured results as opposed to simulations can be attributed to loss from the

electrically conductive silver adhesive used for connecting coax cable connectors. Overall the

ﬁlter performs well and shows equivalent or better characteristics than some of the previously

demonstrated LPF using other technologies to obtain low-loss substrates [6][7].

Figure 3.4: Simulated and measured reﬂection ratio (S11) for low pass ﬁlter

3.1.4 Conclusion

Taking advantage of additive manufacturing and thin ﬁlm process, a new air substrate ﬁlter is

demonstrated. The ﬁlter provides low-loss performance and ease to manufacture. Simulation

and measured results match closely. Slight loss in the measured results is noticed which is

largely due to the attachment of coax based connector using electrically conductive silver

adhesive. This technique allows the fabrication of novel RF structures that would be diﬃcult

to realize using conventional fabrication approaches.

30

Figure 3.5: Simulated and measured reﬂection ratio (S21) for low pass ﬁlter

3.2 Lego-like reconﬁgurable X-band cavity resonator

Traditional microwave waveguides as well as the cavity resonators are bulky and heavy since

they are made of thick metal (brass), and are especially expensive to build. In [8] an X-band

high Q rectangular resonator was built using a commercially available 3D Polyjet printer

called Object Eden 250, in which the feature resolution was improved to be 16 µm per layer
and the surface roughness is measured to be 0.5 µm ± 0.1 µm. The measured resonator

with 0.3% frequency shift yielded a quality factor of 214 at 10.26 GHz, indicating a good

implementation of low loss and low cost RF devices using polyjet printing.

Here we demonstrate a low cost, low loss, lightweight reconﬁgurable cavity resonator by

utilizing the 3D Polyjet printing technique in combination with a novel lego-like assembling

process and tuning posts structure. The cavity is achieved by assembling two 3D Polyjet

printed half sections through a simple lego-like process to achieve a good alignment when

assembled. The specially designed structure on the cavity allows the 6 lego-like posts to

perturb the ﬁeld inside the resonator, thus tuning the resonant frequency. Furthermore,

the conductive silver paste is applied at the edges to help mechanically seal the cavity.

31

The simulation of the cavity resonator was carried out using ANSYS HFSS. The details of

the cavity design, its fabrication and its experimental characterization are presented in this

section.

3.2.1 Reconﬁgurable rectangular cavity resonator design

A X-band rectangular cavity resonator is designed to operate at the TE101 dominant mode.

The corresponding resonant frequency is determined by [5]:

(cid:115)(cid:18) mπ

(cid:19)2

a

(cid:18) nπ

(cid:19)2

+

b

(cid:18) lπ

(cid:19)2

d

+

(3.1)

fT E101

=

c
√
µrr

2π

where r and µr are the relative permittivity and the relative permeability, respectively.

Here, µr = r = 1 since the cavity is ﬁlled with air. The mode indices of TE101 mode are

m = l = 1, and n = 0. a = 20 mm, b = 10 mm, and d = 40 mm are the width, height and

length of the cavity correspondingly. Therefore the fundamental resonant frequency can be

calculated as 8.39 GHz. Also, the theoretical total unloaded quality factor Q0 theo is given

as

1

Q0 theo

=

1
Qc

+

1
Qd

(3.2)

where Qc and Qd are the unloaded Q’s of the cavity with lossy conducting walls and a

lossy dielectric ﬁlling, respectively. By [5], both Qc and Qd can be found as

Qc =

(kad)3bη
2π2Rs

1

(2l2a3b + 2bd3 + l2a3d + ad3)

and

Qd =

1

tanδ

(3.3)

(3.4)

Since the cavity is ﬁlled with air, which can be treated as lossless, thus 1/Qd = tanδ = 0.
Assuming the cavity walls are metalized using copper, whose conductivity is σ = 5.813 ×
107S/m, and the permeability of the air is µ0 = 4π × 10−7 H/m, then we can determine the
surface resistance Rs to be

(cid:115)

Rs =

= 0.0239 Ω

(3.5)

πfT E101

µ0

σ

32

Note the permittivity of the air is 0 = 1/(36π) × 10−9F/m, hence the wave number k

and the wave impedance η in (3) can be easily calculated as

and

η =

= 120π Ω

k = 2πfT E101

√

µ00 = 175.6204

(cid:114) µ0

0

(3.6)

(3.7)

Therefore, by plugging (3.5), (3.6) and (3.7) in (3.3), and setting a = 20 mm, b = 10

mm, d = 40 mm and l = 1, we can ﬁnally calculate Qc = 7301. Recall that 1/Qd = 0, thus

the theoretical total unloaded quality factor of the caity is Q0 theo = Qc = 7301.

To excite and detect the TE101 mode inside the cavity, two coaxial cables with character-

istic impedance of 50 Ω are used to electrically couple energy. For the purpose of maximizing

the energy transfer, the feed probes of the coaxial cable should be parallel to the electric

ﬁeld lines of the excited mode. The feed position should be near the maximum of the electric

ﬁeld which in this case is λg/4, where λg is the guided wavelength. But for better impedance

matching, such position can be slightly varied. Finite-element method (FEM) with the aid of

ANSYS High Frequency Structure Simulator (HFSS), was used to optimize the feed positions

as well as the probe length of the coaxial cable inside the cavity. The detailed parameters

are provided in Table 3.2. Figure 3.6 shows the simulated electric ﬁeld of the TE101 mode

excited by the coaxial cables given the optimized feed positions and probe length.

Table 3.2: Design parameters of the cavity resonator

Dimensions (mm)

a
d
Feed lo
Base d

20
40
7.1
2.78

10
10

b
T
Post d 1.46
hCu
0.003

To perturb the ﬁeld inside the cavity resonator, either the approach of material pertur-

bations or shape perturbations is commonly used. Here we utilize the shape perturbation

method by using the lego-like metal posts, so that the resonant frequency can be tuned.

33

Figure 3.6: (a).Simulated electric ﬁeld of the dominant TE101 mode; (b).Cross-sectional view
of the cavity; (c). Schematic view of the cavity.

Various positions of the tuning posts will result in diﬀerent resonant frequency shifts. A

total of 6 tuning posts are used. Therefore, when one of the posts is pushed into the cavity

to perturb the ﬁeld, other posts are pulled up to prevent perturbation. In order to make

a good contact between the tuning post and the cavity wall, the base of the tuning post is

made larger in diameter than that of the post. However, such base requires a same size dent

at the top of the cavity wall, so that when one post is pulled up not to perturb the ﬁeld,

the base will merge into the top cavity wall to restore its ﬂatness. Otherwise, the resonant

frequency shift will be created by not only the tuning post, but also the bases of the 5 posts

sticking out into the cavity. The schematic view and detailed design parameters are shown

in Figure 3.6(c) and Table 3.2.

3.2.2 Fabrication process

The 3D Polyjet printer used here is a professional-grade commercially available printer, Objet

Connex 350 that uses a photo-polymer resin. It provides a ﬁne feature resolution as well as

34

high quality surface ﬁnish with rapid prototyping. The material used here for printing is

called ”VeroWhitePlus” with a relative permittivity r of about 2.8 and a loss tangent tanδ

of 0.04. The fabrication process is clean and cost-eﬀective, combined with a high accuracy

resolution, making the 3D Polyjet printing techniques suitable for the fabrication of RF

circuits.

The cavity cap and base were printed by the 3D Polyjet printer. A layer of 50 nm

Titanium (Ti) is sputtered ﬁrst followed by another layer of 1 µm thick copper (Cu). Ti

is used to promote adhesion between the VeroWhitePlus and the copper. The two cavity

pieces are also electroplated with copper to increase the coverage and the metal thickness

to approximately 6 µm. Figure 3.7 shows the assembling process of the cavity resonator

with 6 tuning posts. First, the 6 tuning posts are assembled through the holes on the cavity

cap, then the cavity cap stacks on top of the cavity base through the lego-like process. To

help better seal the cavity and make a good contact between the coaxial cable and cavity

wall, silver paste is applied to the four side walls of the cavity as well as around the coaxial

cables, and about 4 - 5 hours is needed for the silver paste to dry and become conductive.

Fig. 3.8(a) shows the cavity resonator fed by two coaxial cables and with all 6 tuning posts

pulled up (no perturbations). Fig. 3.8(b) shows the cavity with one tuning post pushed into

the cavity to perturb the ﬁeld inside the resonator.

Figure 3.7: The assembling process of the cavity resonator

35

Figure 3.8: (a) All the 6 posts are pulled up (no perturbation); (b) The center post is pushed
into the cavity to perturb the ﬁeld.

3.2.3 Bi-material printed quasi yagi-uda antenna

All EM-simulations were performed using Ansys HFSS. The measurement was carried out

using a Keysight N5227A Performance Network Analyzer from 8 - 10.3 GHz. An SOLT cali-

bration was performed up to the edge of the coaxial cable connectors, and all measurements

are carried out at room temperature.

The measured and simulated results are shown in Fig. 4. The simulated unperturbed

dominant resonant frequency of the TE101 mode is 8.32 GHz, which is a little bit diﬀerent

from the 8.39 GHz calculated from (3.1). That is because the probes of the coaxial cables

themselves also inevitably serve as a perturbation of the ﬁeld. From Fig. 4, it can been seen

that the measured unperturbed dominant frequency is 8.31 GHz, which is 0.1% of frequency

shift relative to the simulated one. This small diﬀerence is due to fabrication tolerance.

Additionally, we can see the measured frequency tuning range is from 8.31 GHz - 10.12

GHz, which gives 1.81 GHz of tuning bandwidth.

By observing the unperturbed resonant frequency fr0 = 8.31 GHz and the 3 dB band-
width ∆f3dB = 0.04 GHz from Fig. 4, the loaded quality factor QL can be calculated

as:

QL =

fr0

∆f3dB

≈ 208

(3.8)

The external quality factor Qe as well as the unloaded quality factor Q0 can be respec-

tively determined by:

36

and

S21(dB) = 20log10

QL
Qe

1
Q0

=

1
QL

− 1
Qe

(3.9)

(3.10)

From Figure 3.9, S21(fr0) = -10.68 dB, thus the unloaded quality factor can be calculated
as Q0 = 391. However, it is relatively lower than the simulated one. Also, except no obvious

peak was observed in the measured ”P1” curve, all the other 5 measured curves show the

corresponding tuned resonant frequencies which agree very well with the simulation. The

diﬀerences between the simulated and measured results are due to the following reasons:

(a). The weak coupling between the two coaxial cables during the measurement, (b). The

two VeroWhitePlus cavity pieces were not polished after they were printed (certain surface
roughness exists, ∼5 µm, that was not included in the simulation); thus, surface roughness

increases loss, (c). There are fabrication tolerances both for the cavity and the posts, (d).

The conductivity of the silver paste is not as good as that of metal at these freqencies.

Deembedding the data up to the front end of the coaxial cables will be carried out in the

future. Table 3.3 lists all the simulated and measured tuned resonant frequencies by pushing

one tuning post into the cavity each time, which shows a very close matching between the

simulation and measurement. The average unloaded quality factor is calculated to be 169,

and the average frequency shift is 1.05%. Furthermore, the simulated unperturbed unloaded

cavity quality factor is 3418 assuming perfectly smooth walls. However, if one considers the

average measured surface roughness of 5µm, then the simulated unperturbed quality factor

reduces to 3351. The lower measured values can be attributed to the overall waviness of

the printed material, as well as the lossy nature of the silver epoxy used to seal the two

cavity pieces. The overall result demonstrates a good combination of 3D Polyjet printing

techniques and the novel RF circuits.

37

Figure 3.9: Measured and simulated S21 for all 7 frequencies

Table 3.3: The measured unloaded qualify factor QL, simulated and measured tuned fre-
quencies (exclude P2, and Po is the unperturbed resonant frequency)

Metrics
Sim/Meas freqs (GHz)
Measured Unload Q0
Metrics
Sim/Meas freqs (GHz)
Measured Unload Q0
Metrics
Sim/Meas freqs (GHz)
Measured Unload Q0

P0
8.32/8.31
391
P3
8.45/8.43
132
P6
8.34/8.33
276

P1
9.72/10.12
91
P4
8.57/8.66
111

P2
8.96/9.06
72
P5
8.47/8.42
107

3.2.4 Conclusion

This section presents a combination of lego-like tuning posts and a cavity resonator structure

using 3D Polyjet printing techniques. The fabricated X-band reconﬁgurable cavity resonator

shows the dominant resonant frequency of 8.31 GHz with 0.1% frequency shift and an unload

qualify factor of 391. By using the tuning posts, the ﬁeld inside the cavity is eﬀectively

perturbed resulting in 1.81 GHz of tuning bandwidth. These results provide an avenue for

fabricating low cost and good performance RF components using the 3D Polyjet printing

techniques. Further performance improvement can be achieved by addressing the substrate

waviness and roughness issue, as well as using sealing materials with higher conductivity.

38

3.3 A Bi-material fully aerosol jet printed W -band quasi-yagi-uda

antenna

The planar quasi-Yagi-Uda antenna is chosen to achieve the end-ﬁre radiation pattern be-

cause of its merits of high gain, low proﬁle and compatibility with the standard IC fabrication.

The design of the antenna was based on the standard cylindrical-wire element Yagi-Uda an-

tenna in air which features one reﬂector, one driver and two directors. The initial dimensions

were ﬁrst obtained for maximum directivity and then were scaled to the designed dimen-

sions on the polyimide substrate [9]. Given that the dielectric constant of the polyimide is

r = 3.5 [10], we can determine the eﬀective dielectric constant of the air-polyimide media to

√
be ef f = 2.6 which yields the dimension scaling factor to be 1/

2.6 = 0.62. The equivalent

widths of the microstrip elements were also obtained by mapping from the diameter of the

cylindrical-wire elements.

3.3.1 Antenna design

Figure 3.10 shows the schematic of the antenna, in which the ground plane was utilized as

the pseudo reﬂector instead of adding an independent reﬂector element. The two dipole

arms are separately Aerosol Jet printed on both sides of the polyimide substrate, and are

connected to the microstrip line and the ground plane, respectively, such that the dipole is

directly fed by the microstrip line. Thus, the traditionally used microstrip balun for feeding

the dipole is avoided. In particular, given the high antenna input impedance at the driver

element, a quarter wavelength transformer is designed to match with the 50 Ω microstrip

line. Moreover, a vialess Coplanar Waveguide (CPW) to microstrip transition is designed as

part of the feeding network for W-band probing, as can be seen in Fig. 1.

39

Figure 3.10: The design schematic of the W band antenna. All metal layers and the dielectric
substrate are Aerosol Jet printed

3.3.2 Fabrication Process

Compared to the traditional photolithography in which the antenna pattern is etched out of

the copper cladding sheet, the Aerosol Jet printing technology utilizes the atomized liquid

material, such as silver ink or polyimide dielectric, to directly print the pattern on the sub-

strate, oﬀering a more economic, green and rapid fabrication process. The basic illustration

of the Aerosol Jet printing process is shown in Figure 3.11. First the liquid material is

atomized to create dense aerosol droplets with diameter between 1-5 µm; then the aerosol

is transported using an inert gas to the deposition head where the aerosol is focused by

an annular sheath gas, creating the jetted aerosol material. The feature resolution of the

Aerosol Jet ranges from 5 µm to 1 mm depending on the size of the nozzle, the amount of

the feeding material, and the sheath gas pressure.

Here we use the commercially available Optomec Aerosol Jet 5X printer.

Its 5 axes

degree of motion freedom allows for printing electronics onto complex 3D surfaces. The

printing preparations begin by generating the CAD toolpath ﬁle by specifying the trace

pattern, trace width, and the trace overlapping percentage for suﬃcient coverage. A 4 mil

Rogers 3850 HT LCP host substrate was cleaned with alcohol after removing the copper

cladding. For maximum dimension accuracy while considering shortening the printing cycle,

the 100 µm nozzle was used to print the top antenna layer since the elements dimensions are

crucial to the antenna performance. Whereas the ground plane with one driver arm and the

polyimide substrate were printed using the 150 µm and 300 µm nozzles, respectively, since

40

Figure 3.11: The illustration of the Aerosol Jet printing process

less accuracy is required and certain thickness needs to be achieved quickly. In addition,

the printer parameters, such as the amount of feeding material and the sheath gas pressure,

were also adjusted to obtain the same trace width as speciﬁed in the toolpath. After the

preparation, the host substrate was loaded onto the stage for printing. Note that the desired

metal or polyimide thickness was achieved through stacking multiple layers. Particularly, 2

layers of silver ink were ﬁrst printed on the LCP forming the ground plane with one driver

arm, which was then cured on the hot plate in nitrogen under 180oC for 6 hours to sinter the

silver nanoparticles, resulting in the measured ground plane thickness of about 3 µm. Next,

the polyimide substrate and the antenna elements were sequentially printed above the ground

plane. Speciﬁcally, 20 layers of polyimide and 6 layers of the silver ink were printed using

the aforementioned diﬀerent nozzles respectively. As a result, the polyimide substrate was

sandwiched in between the ground plane and the antenna layer. Finally, the whole sample

was cured in nitrogen under 280oC for 5 hours to solidify the polyimide and a subsequent

180oC for 2 hours to sinter the silver nanoparticles of the antenna layer. Figure 3.12 shows

41

the printed antenna sample under the microscope and the measured surface proﬁle of the

polyimide substrate. As can be seen, the surface roughness of the polyimide is about 0.5 µm

and the average substrate thickness is about 22 µm. The slope of the surface proﬁle curve

is due to the fact that the sample was placed on the non-horizontal plane right after the

printing, causing the uncured polyimide to ﬂow to one side.

Compared to fabricating the substrate using the polyjet technology, three major ad-

vantages of using AJP technology with the polyimide are highlighted: (a) the polyimide

substrate is smoother and remains in shape, as opposed to the “potato-chipped” polymer

substrate, after the curing process; (b) the polyimide substrate can be as thin as several
micrometers whereas the polymer substrate is limited to ≥100 µm; (c) the loss tangent of

the polyimide substrate is signiﬁcantly less than that of the polymer substrate, making it

better for W-band and high frequency circuits and components.

3.3.3 Measurement setup and results

In order to justify the printing quality, the printed dimensions were measured under the

microscope and compared with the designed values, as shown in Table 3.4. We can see

that the dimensions of the printed key antenna features are almost exactly the same as the

designed values, and the maximum percentage diﬀerence between the measured and designed

dimensions is 3% indicating a highly accurate fabrication process using the AJP technology.

Due to the small antenna size, the four-point conductivity measurement of the silver ink was

not performed. Under the curing process speciﬁed in previous section, the printed silver can

achieve 40% of the bulk conductivity of the solid silver.

The return loss of the antenna sample was measured using the Cascade Industry 150

µm pitch G-S-G waveguide Inﬁnity probe on a MPI TS150-THZ on-wafer probe station

utilizing a Keysight N5227A Performance Network Analyzer, which was calibrated with

the LRRM method up to the probe tips from 75 - 110 GHz. The full-wave simulations

were carried out using ANSYS HFSS. Figure 3.13 compares the measured and simulated

42

(a)

(b)

(c)

Figure 3.12: (a) The surface proﬁle of the whole polyimide cross section (b) The enlarged
segment of the polyimide cross section (c) The top view of the printed quasi-Yagi-Uda
antenna

S11, from which we can see that a good agreement is achieved. Speciﬁcally, the measured

resonant frequency is 92.7 GHz with 20 dB of return loss, whereas the simulated return

loss at 94 GHz is 31.6 dB given the conductivity of the printed silver is 40% of the solid

silver. Furthermore, the measured and simulated 10 dB impedance bandwidth is 12.5 GHz

and 8.1 GHz, respectively. Wider measured impedance bandwidth can be attributed to the

tradeoﬀ between the radiation eﬃciency and the impedance matching. The far ﬁeld radiation

was not measured due to the equipment limitations; as a reference, the simulated radiation

pattern as well as the realized gain versus frequency are provided in Figure 3.13. As shown in

Figure 3.14, with 40% of the bulk conductivity for the printed silver, the simulated maximum

realized gain is 7.65 dBi at 94 GHz and the 3 dB gain bandwidth is 9 GHz, suggesting that

a good end-ﬁre radiation with decent gain is obtained from this design.

43

Table 3.4: The measured and designed dimensions of the key features of the 94 GHz Quasi-
Yagi-Uda antenna

Features

Substrate

Dir 1, Dir 2
Drv
G1, G2
Microstrip
QT
CPWG

Max diﬀ. %

Dimensions (mm)

1.3 (W) × 7.7 (L)

Designed
× 0.023 (H)
0.03 × 0.96
0.03 × 1.02
0.065 × 2.2
0.09 × 0.78

0.88

0.015 (G), 0.04 (S)

1.3 (W) × 7.6 (L)

Measured
× 0.022 (H)
0.03 × 0.95
0.03 × 1.01
0.063 × 2.1
0.09 × 0.78
0.015, 0.04

0.87

0.45(L)

0.44

3 %

Figure 3.13: The simulated and measured S11 of the printed quasi-Yagi-Uda antenna

(a)

(b)

Figure 3.14: (a) The simulated 3D far ﬁeld radiation pattern at 94 GHz (b) The simulated
maximum realized gain V.S. frequency

3.3.4 Conclusion

This work demonstrates for the ﬁrst time a bi-material, fully Aerosol Jet printed W-band

quasi-Yagi-Uda antenna operating at 94 GHz. A 4 mil LCP served as the host substrate,

where silver ink nanoparticles were used to print the ground plane and the top metal antenna

44

7580859095100105110frequency (GHz)-35-30-25-20-15-10-50S11 (dB)SimMeas859095100Frequency (GHz)-4048Realized gain (dBi)Max realized gainlayer with the same thickness of 3 µm, and polyimide materials were deposited to form the

22 µm dielectric substrate. Various nozzle sizes were utilized for high quality geometry

resolution and fast prototyping. The measured S11 matches well with the simulation results,

suggesting a successfully fabricated antenna and paving the way for fabricating RF circuits

and components in W-band or even higher frequencies band fully using the AJP technology.

Future works involve improving the surface quality and the silver ink conductivity, as well

as measuring the antenna gain.

3.4 Aerosol jet printed 24 GHz quasi-yagi-uda antenna on a poly-

jet printed cavity substrate

The purpose of this section is to demonstrate a low loss, high gain and good eﬃciency

quasi-Yagi-Uda antenna by utilizing the Aerosol Jet printing technology and 3D Polyjet

printing technique. The substrate is printed by the commercially available 3D printer using

VeroWhitePlus material, and the antenna structure is mostly printed using silver ink. Full

EM wave simulation was carried out using ANSYS HFSS. The details of the antenna design,

fabrication process and the measurement results are presented.

3.4.1 The antenna design

Similar to the quasi-yagi-uda antenna shown in the previous section, two directors are used

here. But here to feed the antenna properly, a microstrip balun is used to convert the signal
carried by the transmission line into two equal magnitude but opposite polarity (180◦ phase

shift) signals that feed two arms of the driven element. Since the input impedance is largely

aﬀected by the lengths of the driven element, directors and the spacing of the directors, these

parameters are intentionally optimized to match the 50 Ω microstrip transmission line. In

the measurement, a 1.85 mm end-launch connector whose characteristic impedance is 50 Ω

is used to connect to the microstrip line.

45

The antenna and the feed structure are Aerosol Jet printed on top of a 3D printed

substrate fabricated with the Polyjet technology. In a traditional design, to have a relatively

low loss, a localized back-side etching process is utilized to create a cavity by etching out

the silicon substrate underneath the antenna, and another piece of SiO2 is used as the

support substrate for the antenna and the feed network. In this design, we take advantage

of additive manufacturing technology to 3D print a single piece of substrate which integrates

the cavity into the back side of the substrate, thus simplifying the fabrication process and

minimizing the prototyping time. Moreover, the dielectric constant r and loss tangent tan δ

of the material used for printing are 2.8 and 0.04, respectively, hence providing a merit of

relatively low loss compared to silicon and SiO2. Figure 3.15 and Figure 3.16 illustrate

the design schematic of the end-ﬁre quasi Yagi-Uda antenna on the 3D printed substrate.

From the ﬁgures, we can see that the total thickness of the 3D printed substrate is 300 µm.

Considering the practical fabrication resolution limit as well as the mechanical stability, the

distance between the top of the cavity and the surface of the substrate (“hSub-hCav”) is

chosen to be 150 µm. Other parameters, for example the size of the cavity (“xCav”, “yCav”

and “hCav”), can also aﬀect the performance of the antenna. Particularly, the size of the
cavity should not be less than the area occupied by the antenna (5 mm × 5.6 mm), otherwise

the gain of the antenna will be signiﬁcantly deteriorated. By taking the fabrication resolution

limit into account, we performed an EM-simulation using Ansys HFSS to optimize the design

parameters, and the results are reported in Table 3.5.

3.4.2 Fabrication process

Figure 3.17(a) shows the fabricated sample of the 3D printed substrate compared to a U.S.

quarter coin. Note that the two vias with diameter of 1.98 mm on the substrate are for

connecting the 1.85 mm end launch connector to perform measurements. First, the substrate

is printed using an industrial-grade commercially available 3D printer called Object Connex

350 with a photo-polymer resin. The material used here is called ”VeroWhitePlus” which

46

Figure 3.15: Top view of the quasi-Yagi-Uda antenna on the substrate

has a dielectric constant r of 2.8 and a loss tangent tan δ of 0.04. Then a 1 µm layer of

copper is sputtered on the backside of the substrate. The area of the metalized ground plane
is 10 mm × 6 mm which is isolated by a 3D printed special cover.

The design pattern of the antenna as well as the feed network is ﬁrst processed by

AutoCAD to generate the toolpath for the Aerosol Jet printer nozzle. Then the pattern

is printed on the opposite side of the 3D printed substrate where the ground plane is. By

adjusting the pressure of the sheath gas ﬂow, the width of one single toolpath is ﬁxed at 10

µm. As illustrated in the design schematci, the width of the driven element, two directors

and microstrip balun is 0.2 mm, whereas the microstrip transmission line has a width of 0.84

mm. The average thickness of the printed lines is 4 µm. Compared to the commonly used

47

Figure 3.16: Cross-sectional view of the quasi-Yagi-Uda antenna on the substrate

Table 3.5: Quasi-Yagi-Uda Antenna Design Parameters

dir 1
drv l
d1
xCav
bal l
wSig
xGND
s
hSub
hAg

Dimensions (mm)
dir 2
wAnt
d2
yCav
bal g
lSig
yGND
d Via
hCav

4.35
2.25
2.2
9.5
1.497
0.84
10
0.31
0.3
0.003

4.4
0.2
1.84
14.5
0.85
8.656
6
1.98
0.15

photolithographic processes which demand a strict clean room environment, a big advantage

of Aerosol Jet printing technology is that the entire process is accomplished in a single

step without wasting materials and resources, hence lowering the overall cost and relaxing

the environmental requirements. Finally, after the whole pattern is printed, the sample

was ﬁrst baked under 65 celsius for 1.5 hour, and then was cured using 1 minute of Argon

and 1 minute of plasma which contains 97% of Argon and 3% of SF6. Then the silver

ink can achieve a conductivity of about 10% of that of solid silver, which is calculated to
be 5.13×106 Siemens/m. Figure 3.17 (b) shows the cured quasi-Yagi-Uda antenna sample.

Note the darkened ground plane beneath the antenna is due to the curing process. Figure

3.18(a) shows the Aerosol Jet printing system as well as the process of printing the antenna

on the 3D printed VeroWhitePlus substrate. Figure 3.18(b) shows the Aerosol Jet printed

quasi-Yagi-Uda antenna with a 1.85 mm end launch connector.

48

Figure 3.17: (a) 3D printed substrate compared to the U.S. quarter coin, (b) Aerosol Jet
printed quasi-Yagi-Uda antenna on the VeroWhitePlus substrate after curing

Figure 3.18: (a) The Aerosol Jet printing system, (b) The quasi-Yagi-Uda antenna with a
1.85 mm end launch connector

3.4.3 Simulation and measurement results

The EM-simulation was performed using Ansys HFSS. Since the silver ink used by the

Aerosol Jet achieves 10% of conductivity of the solid silver, the conductivity of the silver ink
was set to be 5.13×106 Siemens/m in the simulation. The measurement was performed with

a Keysight N5227A Performance Network Analyzer. One 1.85 mm end launch connector was

used to connect to the microstrip feed line of the antenna. The measurement was calibrated

with the SOL method up to the end launch connector.

Figure 3.19 compares the simulated and measured return loss over a frequency range of

20 - 30 GHz. At 24 GHz, the simulated return loss is 26.6 dB, and the measured return loss

is about 37.4 dB at 26.3 GHz. The diﬀerence is due to the fabrication accuracy, since the

resonance frequency is highly sensitive to the dimensions of the antenna, and the connector.

49

Also because of the low conductivity of the printed silver, the antenna is more lossy which

give lower impedance matching. Table 3.6 provides the measured dimensions of several

critical antenna parameters after the curing process. It can be seen that the length of the

driven element “drv l” reduces to 2.28 mm compared to the designed 2.45 mm, which also

explains why the measured resonance frequency increases to 25.8 GHz. The simulated and

measured E plane radiation patterns displayed in Figure 3.20 exhibits a good agreement.

Simulated maximum realized gain versus frequency are presented in Figure 3.21. We can see

that simulated the 3 dB bandwidth is more than 10 GHz. Table 3.7 provides the summary

of the simulated parameters in terms of the antenna performance. We can see that there is

a tradeoﬀ between the 3 dB bandwidth and the maximum gain; wider bandwidth will result

in lower maximum gain.

Figure 3.19: The simualted and measured return loss of the antenna fabricated by AJP and
polyjet technology

3.4.4 Conclusion

In this section, a compact, low proﬁle 24 GHz quasi-Yagi-Uda antenna was developed using

additive manufacturing techniques. The antenna is optimized to be impedance matched with

50

Figure 3.20: The simualted and measured E plane radiation patterns of the antenna fabri-
cated by AJP and polyjet technology

Figure 3.21: Maximum simulted realized gain versus frequency

a 50 Ω microstrip transmission line. Good return loss and end-ﬁre radiation patterns are

achieved. The simulation and the measurements match closely. Moreover, the combination

of Aerosol Jet printing and Polyjet printing technologies is demonstrated for the ﬁrst time at

mm-wave frequencies, which facilitates low cost and fast prototyping. It shows the advantage

of additive manufacturing technology that allows for highly eﬃcient fabrication of novel RF

structures.

51

Table 3.6: Measured quasi-Yagi-Uda Antenna Design Parameters after curing process

dir1 l
drv l
d1
bal l
wSig

Dimensions (mm)
dir2 l
4.02
2.15
wAnt
d2
1.31
bal g
1.47
0.79
lSig

4.03
0.18
1.63
0.84
8.58

Table 3.7: Quasi-Yagi-Uda Antenna Performance Parameters

Simulated performance parameters at 24 GHz

S11
Maximum Gain
Radiation eﬃciency
3 dB bandwidth
Front-to-back ratio

26.4 dB
5.75 dBi
47.7%
>10 GHz
4.6

3.5 Summary

In this chapter we demonstrated a number of RF devices that have very good RF performance

as well as low cost. These devices are either fabricated used one 3D printing techniques but

achieved reconﬁgurability, or printed two diﬀerent materials at one single device to obtain

showcase fully printed capability, or fabricated by combining two 3D printing techniques to

display printing ﬂexibility. The above results show great potential of utilizing 3D printing

technologies of designing and fabricating RF circuits and devices.

52

CHARACTERIZATION OF MAGNETIC NANOPARTICLE FILM

CHAPTER 4

With the advent of the 5G wireless communication system, wireless applications utilizing

the radio frequency (RF) and millimeter wave bands will become more widespread [11].

However, the signals generated by other nearby applications, especially occupying the same

or adjacent bands, can be treated as an RF interference or jammer to the local receiver.

Given the scarce bandwidth, the RF interference can get more critical which degrades the

quality of service [12]. However, if such a high power jammer is above a certain power

threshold, some power-sensitive electronic components in the RF front end can be adversely

aﬀected or even damaged. In many practical RF systems, the input power threshold for the

low noise ampliﬁers (LNAs) is around 30 dBm to prevent the LNA from being damaged due

to a high collector current density in the transistor [13]. Power limiters are also employed to

protect the LNAs and receivers by blocking the high power interferences [14][15]. Given the

narrow band nature, these high power interfering signals can still damage the components

since the limiters are not frequency selective.

To track down the narrow band high power interference signals, a device called Fre-

quency Selective Limiter (FSL) has been proposed which is able to capture and suppress

the large power interference signal even if it has a very narrow bandwidth. The FSL can

be achieved through many approaches such as using paramagnetic [16], parametrics [17]

and switched-multiplexers [18][19]; however, these FSLs either suﬀer from severe distortion

or require sophisticated topology and manufacturing which limits their integration capabil-

ity (especially for monolithic circuits) and increases their cost. It has been reported that

the magnetic ferrite materials and nanoparticles oﬀer an intrinsic non-linear signal property

that is widely explored in microwave and millimeter-wave applications. Above a critical RF

power level, the non-linear spin wave excitation will cause the magnetic loss to the RF signal.

53

In light of this, one promising solution is the FSL utilizing the magnetic ferrite materials

[20][21]. This type of FSL can selectively track down and attenuate more than one large

power interference signals while maintaining the dynamic range of the signals of interest, or

with little attenuation. In light of this, the FSL becomes a passive circuit without additional

complicated circuitry, so that the overall size as well as the cost is greatly reduced.

Recently, FSLs using magnetic thin ﬁlms or spheres have been developed with sin-

gle crystal Yttrium Iron Garnet (YIG) and scandium doped Barium Hexaferrite materials

[21][22][23][24]. These ﬁlms or spheres are produced by using a sol-gel method which is sim-

ple and cost-eﬀective. Nevertheless, such a method is unable to produce signiﬁcant amounts

of magnetic materials in a solid state conﬁguration. For example, the achieved scandium

doped Barium Hexaferrite ﬁlm thickness on the alumina (99.6%, Al2O3) substrate is only

10 µm [23], which is too thin to show any signiﬁcant FMR eﬀect. Given the relatively large

actual size, these FSLs are hard to be integrated into RF integrated circuits.

In this work, a broad range of magnetic ferrite nanoparticles has been chosen to explore

their fundamental magnetic properties and the fabrication process. Particularly, Cobalt

Ferrite (CoFe2O4) and Manganese Ferrite (MnFe2O4) nanoparticles are deposited on a soda-

lime glass substrate to form the magnetic ﬁlms with thicknesses ranging from 2.5 µm to 125

µm, which is much thicker than the 100 nm CoFe2O4 ﬁlms in some reported works [25][26].

These magnetic nanoparticles show a much higher saturation magnetization than that of

YIG. Two types of methods, the Layer-by-Layer (LbL) process and the Solution Cast (SC)

method, are employed to fabricate the magnetic ﬁlms. These methods are-to the best of our

knowledge-implemented for the ﬁrst time to develop relatively thick magnetic ﬁlms on large

area substrates without the usage of expensive vacuum deposition systems. By integrating

the magnetic ﬁlms with a microstrip transmission line structure, we created a basic microwave

device to investigate the properties of the magnetic thin ﬁlms in terms of the ﬁlm thickness,

types of the materials and deposition methods. A 3-D polyjet printer is used to enable the

ﬂexibility and fast prototyping to fabricate the microstrip line, as well its assemblies for

54

measurement purposes, so the overall cost of the fabrication can be reduced.

4.1 FMR Frequency and Magnetic Susceptibility in Magnetic

Nanoparticles

The unique non-linear property of the magnetic ferrite materials can be attributed to the

Ferromagnetic Resonance (FMR) phenomenon, which represents the absorption capability

of a magnetic material to electromagnetic waves of some frequency. The small-signal theory

of FMR was developed by Kittel [67]. This intrinsic property gives rise to the use of the

magnetic materials in the FSL to achieve the unique performance aforementioned.

Note that diﬀerent types of magnetic materials may display distinct FMR eﬀect, so in

this work, very diverse magnetic metal oxides nanoparticles have been chosen to study the

determining factors to the frequency of FMR. These nanoparticles were synthesized via

aminolytic reactions with ﬁne controls of sizes and metal ion concentrations, which included
MnO, ZnFe2O4, MnFe2O4, ZnxMn1−xFe2O4, GdxMnFe2−xO4, HoxMnFe2−xO4, CoFe2O4,
HoCoFe2−xO4 and NdCoFe2−xO4 (0≤x≤1), as shown in Table 4.1. The exact chemical
compositions of these nanoparticles are still being studied. Figure 4.1 shows the X-ray

diﬀraction (XRD) patterns collected with a Bruker X-Ray Diﬀractometer over a 15o-85o 2θ

range. Particle sizes were determined from the average peak broadening of the ﬁve strongest

Bragg peaks using the commercial program IGOR and the Debye-Scherrer equation. Fig-

ure 4.2 shows the temperature dependent magnetic moment of these nanoparticles. Magnetic

measurements were performed using a Quantum Design MPMS-5S SQUID magnetometer.

Zero-ﬁeld cooled (ZFC) susceptibility measurements were conducted under an applied ﬁeld

of 100 G with temperature ranging from 5 - 310 K.

Spinel ferrite nanoparticles have displayed FMR under magnetic ﬁelds. After forming

nanoparticle ﬁlms, the FMR peak position and proﬁle do not display any diﬀerence from

the original nanoparticles. The nanoparticles and their ﬁlms have shown diﬀerent FMR

absorption proﬁles with diﬀerent chemical composition as shown in Figure 4.3, which were

55

determined under a constant excitation wavelength, f = 9.88 GHz. The resonance frequency

increases with the increase of nanoparticle size. However, the peak width is independent of

the nanoparticle size at least in the 4 - 10 nm size range that we studied. FMR spectroscopies

were acquired with a Bruker EMX EPR spectrometer with a center ﬁeld of 5125.0 G and a

sweeping width of 9750.0 G under a resonance frequency of 9.88 GHz at room temperature.

Table 4.1: Characterization of the selected magnetic nanoparticles.

Sample names

MnO 20nm
ZnFe2O4 6nm
CoFe2O4 9nm

Species

MnO

ZnFe2O4
CoFe2O4

CoFe2O4 Ho2.8 6nm HoCoFe2−xO4
CoFe2O4 Ho3.3 6nm HoCoFe2−xO4
CoFe2O4 Ho5.9 6nm HoCoFe2−xO4
CoFe2O4 Nd2.5 6nm NdCoFe2−xO4
CoFe2O4 Nd3.5 6nm NdCoFe2−xO4

MnFe2O4 7nm
MnFe2O4 6nm

MnFe2O4
MnFe2O4

MnFe2O4 Gd14 6nm GdxMnFe2−xO4
MnFe2O4 Gd4 6nm GdxMnFe2−xO4
MnFe2O4 Ho6 6nm HoxMnFe2−xO4
MnFe2O4 Zn14 4nm ZnxMn1−xFe2O4
MnFe2O4 Zn57 4nm ZnxMn1−xFe2O4

Metal substitution

Particle
(mol/mol, %) from ICP-MS size (nm)

-
-
-

Ho(III) 2.8% for M(III)
Ho(III) 3.3% for M(III)
Ho(III) 5.9% for M(III)
Ho(III) 2.5% for M(III)
Ho(III) 3.5% for M(III)

-
-

Gd(III) 14% for M(III)
Gd(III) 4% for M(III)
Ho(III) 6% for M(III)
Zn(II) 14% for M(II)
Zn(II) 57% for M(II)

20
6
9
6
6
6
6
6
7
6
6
6
6
4
4

4.1.1 FMR and magnetic susceptibility analysis

The nanoparticles shown in Table 4.1 are ferromagnetic spinel ferrites except for the MnO

nanoparticles. All these nanoparticles have displayed FMR under various magnetic ﬁelds ap-

plied, although they have shown diﬀerent FMR absorption proﬁles as we can see in Figure 4.3.

With careful examinations of the Electron Paramagnetic Resonance (EPR) absorption pro-

ﬁles and the magnetic susceptibility measurements, the change of resonance frequency can be

simply correlated to the change in the magnetic susceptibility of these nanoparticles, despite

the diﬀerence among all various chemical compositions and possible spin orders. As shown in

Figure 4.4, the applied ﬁelds for the maximum FMR absorption under a constant excitation

56

Figure 4.1: XRD pattern for CoFe2O4 (top) and MnFe2O4 (bottom), sizes calculated by
Debye-Scherrer equation.

wavelength, f = 9.88 GHz were plotted versus the room temperature magnetic susceptibil-

ities. Qualitatively, a trend could be addressed among these nanoparticles: with a larger

magnetic susceptibility, a lower applied ﬁeld is required to reach the resonance frequency at

f = 9.88 GHz.

57

Figure 4.2: Temperature dependent magnetization for CoFe2O4 (top) and MnFe2O4 (bot-
tom), zero-ﬁeld cooled (ZFC) susceptibility measurements were conducted under an applied
ﬁeld of 100 G with temperature range from 5 - 310 K.

FMR is really an excitation of the nanoparticles between two energy states. With a

higher magnetic susceptibility, the degree of magnetization would be larger in response to

an applied magnetic ﬁeld, and thus cause a larger energy diﬀerence between two energy

states. This results in a smaller applied ﬁeld in EPR measurements to achieve the ﬁxed

resonance frequency. In our EPR studies, this FMR excitation can be achieved by sweeping

the applied magnetic ﬁeld under a ﬁxed frequency. Also this excitation can occur with a

sweeping frequency under a ﬁxed applied magnetic ﬁeld as in microwave components of a

58

wireless device. The trend in Figure 4.4 suggests that the FMR frequency increases with

increasing magnetic susceptibility of the materials. FMR absorption proﬁles can also be

aﬀected by more physical factors, i.e. spin-orbit coupling, g-value shift and damping factor

of the ferromagnetic materials.

It is worthy to note that two diﬀerent slopes could be

found based on the diﬀerence of the FMR line widths in Figure 4.4. A more detailed study

within certain types of particles and with the elimination of other factors has addressed the

qualitative correlation between the susceptibilities and FMR frequency [28].

Figure 4.3: Selected FMR absorption proﬁles.

4.1.2 FMR measurement of ﬁlm particles

For RF applications, the nanoparticles have been cast into ﬁlms with a layer-by-layer method

with a thickness of a few micrometers or a solution casting method with a ﬁlm thickness of

several ten to hundreds of micrometers. The nanoparticles and organic components that form

the ﬁlm were collected from the complete ﬁlms to measure the FMR absorptions. The results

shown in Figure 4.5 are 7 nm MnFe2O4 nanoparticles and 8 nm CoFe2O4 nanoparticles, with

a 10 times magniﬁcation of the original signal. Compared to CoFe2O4, MnFe2O4 tends to

gain narrower absorption peak proﬁles with a stronger intensity which is easier to observe

59

Figure 4.4: Correlation between resonance ﬁelds and magnetic susceptibilities.

during the measurements. The resonance intensity of MnFe2O4 nanoparticles is 100 times

stronger than the same amount of CoFe2O4 nanoparticles. As shown in Figure 4.5, the
strong spin-orbit coupling arising from Co2+ could broaden the FMR absorption proﬁle and

thus CoFe2O4 tends to be a lot less sensitive than MnFe2O4 in FMR spectrum. In all, such

comparison exhibits the stronger signal response for MnFe2O4 nanoparticles over the weak

signal response of CoFe2O4 nanoparticles. It requires 100x CoFe2O4 to achieve a similar

RF response to the MnFe2O4 nanoparticles. Thus, from a device performance perspective,

it is better to have manganese ferrite ﬁlms to be integrated into RF devices such as FSL,

to have narrower and sharper signal absorption proﬁles given the same amount of magnetic

nanoparticles in the ﬁlms.

4.2 Fabrication process of the magnetic nanoparticle thin ﬁlm and

the basic microwave device

The magnetic nanoparticle ﬁlms used for the characterization here are fabricated on the
1 inch × 1 inch soda lime glass slides of 1 mm thickness. Two types of magnetic ma-

terial nanoparticles-CoFe2O4 and MnFe2O4-are deposited on the glass substrate through

60

Figure 4.5: FMR of nanoparticle ﬁlms. The intensity from CoFe2O4 nanoparticle ﬁlm has
been ampliﬁed 10 times from the original signals.

either the LbL process or the SC method. Details of these two kinds of deposition meth-

ods are introduced and compared. The LbL ﬁlm fabrication technique has been developed

for decades and proven to produce high-quality nanoparticle thin ﬁlms with high loading

density; whereas the SC method has been pretty common in industrial manufacturing for

ﬁlms with thickness in micro-meter range. In order to achieve comparable FMR absorptions,

diﬀerent ﬁlm fabrication methods were applied onto diﬀerent particles. MnFe2O4 ﬁlms were

built up through LbL to maintain uniform layer structures with thickness not exceeding 5

µm. CoFe2O4 ﬁlms were made with the SC method to ensure that a large quantity of par-

ticles could be deposited onto one ﬁlm surface. It is worth noting that the thickness of SC

CoFe2O4 ﬁlms was more than 10 times the thickness of LbL MnFe2O4 ﬁlms; which provides

the compensation of the FMR sensitivity from the particle nature. For the purpose of the

characterization, we designed and 3-D printed a simple tapered microstrip line as the test

bed. Together with the 3-D printed clamping and supporting structures, we built a basic

microwave device using the magnetic ﬁlm.

61

4.2.1 Layer-by-Layer Fabrication of CoFe2O4 Thin Films via Epoxy Curing

The CoFe2O4 nanoparticles were chemically bound to the glass substrate using the SN 2

nucleophilic addition between the surface-coated amine groups and epoxy groups. Hence two

sets of surface modiﬁed particles were prepared, namely, 3-aminotriethoxypropylsilane coated

particles and (3-glycidyloxypropyl) trimethoxysilane coated particles. The glass slides were

surface modiﬁed with (3-glycidyloxypropyl) trimethoxysilane prior to particle attachment.

4.2.1.1 Epoxy coated slides

(3-glycidyloxypropyl) trimethoxysilane ethanol solution was prepared (1% v/v) and the glass

slide was dipped into the solution and allowed to soak for 30 minutes. After the slide was

removed from the solution, it was air dried and ready to be further coated with particles.

4.2.1.2 Amine/ epoxy coated CoFe2O4 particles

The CoFe2O4 nanoparticles were synthesized and followed with NaOH stripping ensuring

that the particle surface is bare without other capping agent: a 3-Aminotriethoxypropylsilane

ethanol solution was prepared (1% v/v) and bare particles were added to it (1 mg/ 1mL).

The solution was stirred overnight and the particles were collected through a magnet and

washed with ethanol three times to remove any excess coating agent. The particles were

then re-suspended into ethanol and agitated before being used. The same procedure was

followed for epoxy coated particles, but the (3-glycidyloxypropyl)trimethoxysilane ethanol

solution was used.

4.2.1.3 Layer-by-Layer Fabrication

The glass slide was dipped into an amine coated particle solution and then an epoxy coated

particle solution. Each dipping step was set to 30 minutes and followed with rinsing and

air-drying. With the alternating dipping from these two solutions, a ﬁlm with a certain

62

thickness could be made. The ﬁlm was then protected by spin-coating a layer of Torlon

TF4000 polymer onto the surface to both chemically and physically protect the underlying

ﬁlm.

4.2.2 Fabrication of CoFe2O4 Films via Solution Cast Method

4.2.2.1 Epoxy coated slides and particles

A (3-glycidyloxypropyl)-trimethoxysilane ethanol solution was prepared (1% v/v) and the

glass slides and particles were dipped/soaked into the solution individually. The glass slide

was removed from the solution after 12 hours and air-dried. The particles were collected

through a magnet after stirring overnight and washed with ethanol for three times to remove

any excess coating agent. The coated particles were suspended in an appropriate amount of

dichloromethane.

4.2.2.2 Film Casting

Approximately 10-20% (w/w) poly(methyl methacrylate) PMMA was added to the particle

DCM solution and stirred for approximately 30 minutes before the solution was cast onto the

substrate. With a syringe the nanoparticle solution was applied to the prepared substrate

evenly, allowing the solution to ﬂood above the substrate. The ﬁlm was air-dried and stored

in a desiccator until treated with Torlon protective coating. The ﬁlm was then protected

by spin-coating a layer of Torlon TF4000 polymer onto the surface to both chemically and

physically protects the underlying ﬁlm.

The details of the fabricated magnetic ﬁlms are summarized in Table 4.2. Figure 4.6 shows

the images of a few samples. Speciﬁcally, sample LbL1 is made of MnFe2O4 nanoparticles

by the LbL process, whereas samples SC3, 39I and SC7 are deposited using the CoFe2O4

nanoparticles through the SC method. The ﬁlm thicknesses of LbL2 and SC3 are 3 µm and

40 µm respectively. As we can see, both LbL1 and SC3 have evenly distributed nanoparticles

63

and the color diﬀerence is mostly because of the materials nature. However, there is a large

void in SC6 and a condensed area in SC7, which are due to the fabrication defects. The

measured ﬁlm thicknesses of the smooth area in both samples are 60 µm and 125 µm,

respectively.

Table 4.2: Film thickness, types of materials and deposition methods of the magnetic ﬁlms

MnFe2O4

3µm

SC
LbL LbL1 LbL2 LbL3

-

-

CoFe2O4

5.4µm 20µm 30µm 40µm 50µm
SC4

SC1

SC3

SC2

60µm

125µm

SC5/SC6

SC7

-

-

-

-

-

-

(a) LbL1

(b) SC3

(c) SC6

(d) SC7

Figure 4.6: Images of the MnFe2O4 sample LbL1 and CoFe2O4 samples SC3, SC6 and SC7.

4.2.3 Basic microwave device

To characterize the magnetic ﬁlms, it is necessary to ﬁt them into a test bed so that we

can observe their eﬀect. Here we ﬁrst design a microstrip transmission line with a wide-to-

64

narrow line taper transition as the core test bed. Particularly, most of the center area of the

substrate is designed to be thin to let the substrate thickness be as comparable as possible to

the magnetic ﬁlm thickness as shown in Figure 4.7(a). The thick areas featured with the vias

at the end of the substrate are used to maintain the substrate rigidity as well as facilitate the

RF measurements by using the end-launch RF connectors, as illustrated in Figure 4.7(b).

Then we ﬂip the magnetic ﬁlm on top of the microstrip line.

In order to ensure a good

(a)

(b)

Figure 4.7: (a) Microstrip line design (b) Microstrip line characterized with end-launch
connectors.

contact between the ﬁlm surface and the microstrip line, we incorporate them using two 3-D

printed ﬁxtures. Figure 4.8(a) illustrates the overview of the whole basic microwave device.

In practice, we also use two metal clamps for additional mechanic stability and applying

more pressure, as shown in Figure 4.8(b). The transmitted electromagnetic wave along the

(a)

(b)

Figure 4.8: (a) Overview of the basic microwave device (b) The microwave device using
additional two metal clamps and characterized by the RF end-launch connectors.

65

microstrip line is therefore mostly conﬁned within the region that is near the signal trace.

The presence of the magnetic nanoparticle ﬁlm on top of the microstrip line now serves as

the material perturbation to the RF magnetic ﬁeld, aﬀecting the scattering parameters of

this two-port network, which is illustrated in Figure 4.9. As such, the inﬂuence due to the

existence of the magnetic nanoparticles can be seen from the changes of the corresponding

group delay and insertion loss.

Figure 4.9: Electromagnetic ﬁelds illustration of the ﬂipped magnetic ﬁlm sample on the
microstrip line.

4.2.4

3-D polyjet printing technology

The microstrip substrate and the basic microwave device assemblies are 3-D printed. The

3-D polyjet technology is advantageous for its low cost and fast prototyping, especially since

it allows for fabricating complex structures that are diﬃcult and expensive to manufacture

by using the traditional micro-machining and photo-lithography methods. A commercially

available Objet Connex 350 3-D polyjet printer is used, with a resolution of 100 µm in the

X/Y axes and 30 µm in Z axis as shown in Figure 4.10. The printer head can move in X and

Y axes patterning the planar geometry, and the thickness proﬁle can be generated by moving

along the Z axis. The material used to print the substrate is a photopolymer resin named

VeroWhitePlus, with a dielectric constant of 2.8 and loss tangent of 0.04 [29]. The material

is instantly cured by the UV light integrated on the printer head. Temporary supporting

material can be easily ﬂushed away after the printing is done. After the substrate is printed

and cleaned, the microstrip line is patterned through a copper sputtering process using a

66

3-D printed mask which avoids the expensive cleanroom and photo-lithography fabrication

process. Especially the 3-D printed assemblies would be diﬃcult and costly to manufacture

traditionally.

(a)

(b)

Figure 4.10: (a) Objet Connex 350 (b) Illustration of the 3-D polyjet technology [69].

4.3 Measurement results using the basic microwave device

As discussed in Figure 4.9 of Section III, due to the existence of the magnetic nanoparticles

in the ﬁlm, the RF magnetic ﬁeld of the microstrip line is perturbed, which is reﬂected by

the changes of the group delay and insertion loss compared to the case without the magnetic

ﬁlm. Furthermore, magnetic ﬁlms with various thicknesses which diﬀer in the amounts

of magnetic nanoparticles, deposited using diﬀerent types of materials and methods are

supposed to display distinct amount of group delay. For example, although the relative

permeability of diﬀerent types of magnetic nanoparticles may vary, given the same type

of magnetic nanoparticles and fabrication method, thicker ﬁlms are supposed to result in

higher group delay. Considering the fabrication limits in practice, the optimum combination

of the fabrication method, types of materials and ﬁlms thickness is needed to obtain the

desired ﬁlm properties in a time and cost eﬃcient manner. Therefore, by examining the

extracted group delays of all the samples, we can get some insights into the relationship of

67

the magnetic ﬁlm thickness, types of magnetic material, deposition methods and the amount

of group delay. Ultimately, our goal is to characterize the magnetic material by determining

the relative permeability which enables us to fabricate the monolithic microwave devices and

circuits utilizing such materials.

4.3.1 Measurement setup

Since the magnetic ﬁlm sample mainly consists of a layer of magnetic nanoparticles and a

1 mm thick soda-lime glass substrate, in order to justify the observation results in terms

of the group delay, we establish the observation baseline by measuring the group delay of a

mere soda-lime glass substrate with the same size of the magnetic ﬁlm sample, but without

the magnetic nanoparticles, ﬂipped on top of the microstrip line, as illustrated in the upper

picture of step 2 in Figure 4.11. Therefore, the only diﬀerence between the baseline case and

the magnetic ﬁlm sample case is whether the magnetic nanoparticles exist or not. In other

words, assuming that all the environmental factors are the same, the diﬀerence of the group

delays between the baseline case and the magnetic ﬁlm sample case can be only attributed

to the existence of the magnetic nanoparticles. The measurement setup has four steps as

illustrated in Figure 4.11: a) tightly connect two coaxial cables to the end-launch connectors

of the microstrip line (note that the microstrip line substrate should be ﬂat without any

twist since the VeroWhitePlus substrate is thin and ﬂexible); b) place the microstrip line on

the ﬁxture platform; c) establish the observation baseline by ﬂipping the glass slide on top

of the microstrip line in step 2, and enclose the whole sample with the ﬁxture cap as well as

two clamps in step 3 such that the RF structure is stable and ready to measure; d) measure

the scattering parameters using the portable Keysight Vector Network Analyzer (VNA) and

record the data. After measuring the baseline case, we take oﬀ the clamps and the ﬁxture

cap (while keeping the connection between the VNA and the microstrip line), replace the

mere glass slide with the magnetic ﬁlm sample (glass slide + magnetic nanoparticle ﬁlm)

and repeat steps 3 and 4 to obtain the data of the magnetic ﬁlm case.

68

Figure 4.11: Measurement setup process steps 1-4.

4.3.2 Measurement Results

The scattering parameters can be measured for the above basic microwave device, where S21

denotes the ratio of the transmitted signal over the incident signal between the two ports of

the device. The group delay that is associated with the relative permeability is a measure

of the phase response of the transmitted signal. From the magnetic ﬁeld distribution of Fig-

ure 4.9, we can see the thicker and denser the magnetic ﬁlm, the more the RF magnetic ﬁeld

is conﬁned within the magnetic ﬁlm region, thus causing a stronger phase delay. Therefore

the measured group delay is a good indicator of the eﬀect due to the magnetic ﬁlm. From

the measured scattering parameters, the group delay can be extracted from the S21 using

the following equation:

Group delay = − dφ
dω

(4.1)

where the phase φ = rad(S21), and ω denotes the angular frequency. Although it is suﬃcient

to have the baseline established, another reference line in which no glass slide is ﬂipped on

top of the microstrip line is also provided. Speciﬁcally, if step 2 in Figure 4.11 is skipped,

69

then the measured result reﬂects the group delay of the microstrip line itself, which can be

used as the reference line of the established baseline, as shown in Figure 4.12. We can see

from Figure 4.12 that the “baseline” has a higher group delay than the Reference line, and

the average group delay value for both cases is 350ps and 337ps, respectively, indicating that

the presence of the mere glass slide introduces an extra group delay than the plain microstrip

line as expected. Therefore, the baseline measurement and the whole microwave device setup

are proven to be eﬀective and correct.

Figure 4.12: baseline: Mere glass substrate ﬂipped on top of the microstrip line. Reference
line: Only microstrip itself.

Figure 4.13 shows the measured group delay for all the Cobalt Ferrite samples deposited

against the baseline. As we can see, a higher group delay is displayed for the Cobalt Ferrite

samples than the baseline. And within all the Cobalt Ferrite samples, we can observe diﬀerent

values of the group delay. For example, the average group delay for the sample SC1 is 354ps,

while that of the sample SC3 is 357ps. Since all samples except sample LbL3, are deposited

using the same type of material (Cobalt Ferrite) and method (Solution Cast), the primary

reason for the diﬀerence of the group delay is the magnetic ﬁlm thickness. According to

the nature of the SC method, higher quantity of the magnetic nanoparticles is supposed to

exist in the thicker magnetic ﬁlm, thus causing higher group delay compared to the baseline

70

012345678Frequency (GHz)325330335340345350355360365Group delay (ps)BaseReferenceor thinner magnetic ﬁlms. To verify this proposition, the samples are ﬁrst ordered in an

ascending manner in terms of the ﬁlm thickness from Table 4.2; such ﬁlm thickness sequence

is then compared to the measured group delay sequence from Figure 4.13 which is ordered

in a similar fashion as the ﬁlm thickness sequence. The comparison result is illustrated in

Figure 4.14. We can see that the sequence based on the ﬁlm thickness can be mostly veriﬁed

Figure 4.13: Group delay for all the solution cast CoFe2O4 samples and the baseline.

Figure 4.14: Solution cast CoFe2O4 sample sequences in terms of the ﬁlm thickness and the
measured group delay.

by the measured group delay in Figure 4.13. However, SC6 and SC7 are supposed to exhibit

the second and ﬁrst most group delay among all the samples, as opposed to about the same

amounts of group delay as sample SC4. As observed previously in Figure 4.6, the large void

in SC6 and the condensed area in SC7 that occurs during the deposition process reduces the

amount of the group delay, which is supposed to be higher if the magnetic nanoparticles are

71

012345678Frequency (GHz)340345350355360365370375Group delay (ps)BaseSC1SC2SC3SC4SC5SC6SC7LbL3evenly distributed. And the reason of SC5<SC2 from the group delay, which is contrary to

the ﬁlm thickness based result, may be attributed to the fabrication variation resulting in

more magnetic nanoparticles actually deposited in SC5 than SC2. Therefore, based on the

veriﬁed sequence of SC1<SC3<SC4<SC5, the relationship between the average group delay

and the magnetic ﬁlm thickness is plotted in Figure 4.15, where the measured group delay is

averaged in the range of 1-8 GHz for each sample. From Figure 4.15 we can see that within

our expectation, the thicker the magnetic ﬁlm, the higher group delay. More importantly,

the relationship between the above two quantities is almost linear, indicating that the good

eﬀectiveness of the SC method and the group delay can be potentially even higher if we

continue increasing the magnetic ﬁlm thickness (assuming an almost uniform nanoparticle

distribution with no major fabrication defects).

Figure 4.15: The relationship between the average group delay and the magnetic ﬁlm
thickness for selected solution cast CoFe2O4 samples and the baseline.

On the other hand, sample LbL3 is fabricated by depositing the same CoFe2O4 nanopar-

ticles via the LbL process. Hence it can be used to study the eﬀects of the diﬀerent deposition

methods. Again, it can be observed from Figure 4.13 that SC3<LbL3<SC4 in terms of the

diﬀerences of group delay. However, Table 4.2 shows the ﬁlm thickness of LbL3 is only 5.4

µm as opposed to the ﬁlm thickness of SC3 (40 µm) and SC4 (50 µm), suggesting the LbL

72

010203040506070Magnetic film thickness (m)348350352354356358360362364Average group delay (ps)BaseSC1SC3SC4SC5MeasurementLinear fittingprocess is better than the SC method at producing magnetic ﬁlm that results in higher group

delay given the same ﬁlm thickness. In fact, through the TGA experiment on the ﬁlm ﬁll

factor, the LbL process results in 90% weight percent nanoparticles, while that of the SC

method is about 68%, as shown in Figure 4.16. However, it takes much longer to deposit

the same ﬁlm thickness using the LbL process than the SC method. Speciﬁcally, using the

LbL process, it usually takes 4-5 days to achieve a 5-6 µm ﬁlm thickness; for the SC method,

approximately 3-4 hours are needed to fabricate a 50-60 µm ﬁlm, including 1-2 hours of sub-

strate preparation and 1-2 hours of ﬁlm deposition. Therefore, given no restrictions on the

ﬁlm thickness and targeting the same level of group delay, the SC method is more eﬃcient

in producing much thicker ﬁlms if a shorter fabrication cycle is preferred.

Figure 4.16: TGA curves for LbL and solution cast ﬁlms showing particle loading densities
of 90% and 65% for LbL and solution cast respectfully.

To determine the eﬀects of diﬀerent types of magnetic materials, MnFe2O4 samples LbL1

and LbL2 are compared with CoFe2O4 sample LbL3 in Figure 4.17 since they are all made

via the LbL process (both samples have uniform nanoparticle distribution). First, it can be

seen that the average group delay for both LbL1 and LbL2 is around 365 ps given only 0.1

µm variance in terms of the ﬁlm thickness, while LbL3 introduces 357 ps of group delay in

73

Figure 4.17: Group delay for the MnFe2O4 samples LbL1 and LbL2 and CoFe2O4 sample
LbL3.

average which is less than that of LbL1 and LbL2. As shown in Table 4.2, LbL3 is even

more than twice of the ﬁlm thickness of LbL1 and LbL2, indicating that manganese ferrite

nanoparticles have stronger ability than cobalt ferrite nanoparticles in increasing the group

delay.

Based on the above comparisons on fabrication methods and types of magnetic materials,

we conclude that the SC method is faster in fabricating magnetic ﬁlms and the manganese

ferrite nanoparticles exhibit stronger eﬀect in increasing the group delay. Thus, we can

claim that for the purpose of making the magnetic ﬁlm achieve a certain level of group delay

increase, the most time eﬃcient method is to fabricate such ﬁlm using manganese ferrites

via SC. Such a combination also has the potential for further improvements. If the ﬁll factor

of the ﬁlms made by the SC method can be improved from 68% to 90% or comparable to

the LbL process, then-for the same group delay-the required ﬁlm thickness to be achieved

through the SC method can be reduced and, thus, less fabrication time will be needed. In

other words, given the same ﬁlm thickness and fabrication time, the ﬁlm will introduce higher

group delay by improving the ﬁll factor of the SC method.

Figure 4.18 (a) and (c) shows the S21 and S11 measurements for all the samples, which

74

012345678Frequency (GHz)340345350355360365370375Group delay (ps)BaseLbL1LbL2LbL3(a)

(b)

(c)

(d)

Figure 4.18: (a) S21 for all CoFe2O4 magnetic ﬁlms (b) The absolute power absorption
percentage for all CoFe2O4 magnetic ﬁlms (c) S21 of LbL1 and LbL2 compared to SC7 (d)
The absolute power absorption percentage LbL1 and LbL2 compared to SC7.

can be used to determine the transmitted signal power absorption rate solely due to the

presence of the magnetic ﬁlms. The power absorption is a very important property to know

due to the FMR, which can be utilized in microwave devices, like FSLs, to determine the

attenuation level to the unwanted interference signals. We ﬁrst calculate the overall signal

power loss percentage P L of all cases by using the following equation:

P L = 1 − |S21|2 − |S11|2

(4.2)

If we denote P L0 as the calculated power loss of the “Reference” case where only the

75

02468Frequency (GHz)-40-35-30-25-20-15-10-50S21 and S11 (dB)BaseSC1SC2SC3SC4SC5SC6SC7LbL3S21S1102468Frequency (GHz)0246810Absolute power absorption (%)ReferenceBaseSC1SC2SC3SC4SC5SC6SC7LbL302468Frequency (GHz)-40-35-30-25-20-15-10-50S21 and S11 (dB)BaseLbL1LbL2LbL3S21S1102468Frequency (GHz)0246810Absolute power absorption (%)ReferenceBaseLbL1LbL2LbL3microstrip line exists (without soda lime glass slide or magnetic ﬁlms on top), then the abso-

lute power absorption rate P Lr due to the magnetic samples can be obtained by subtracting

P L0 from the calculated P L’s of all the magnetic samples, that is:

P Lr = P L − P L0

(4.3)

Therefore, the P Lr’s of all the magnetic samples including the “Base” case against the

frequency are depicted in Figure 4.18 (b) and (d). We can see that the power absorption

increases with the frequency and compared to the “Base” case, where merely the glass

slide exists, the magnetic samples absorb more transmitted energy and peak around 7.5

GHz. Additionally, we can observe the similar trend shown in Figure 4.14, such that thicker

magnetic ﬁlms absorb more power, because more magnetic nanoparticles existing in the

thicker ﬁlm convert the absorbed power into loss. We also determine the power absorbed

by the magnetic nanoparticles P Ln by eliminating the eﬀect of the glass slide, i.e., P Ln =
P Lr − P LBase. For example, the P Ln of sample SC7 is 5.9% at 7.5 GHz, which corresponds
to 0.26 dB of total loss and 0.07 dB/cm of loss given the 4 cm length of the microstrip line.

Thus, from the above perspective, we further prove the conclusion about the eﬀectiveness of

the proposed fabrication methods and the fact that the added line loss due to the magnetic

nanoparticles is negligible.

4.4 Summary

This chapter presents the study of various magnetic nanoparticle based thin ﬁlms. We pro-

pose new methods of fabricating magnetic ﬁlms, namely the Layer-by-Layer process and the

Solution Cast method. The magnetic ﬁlms are made of CoFe2O4 or MnFe2O4 nanoparticles

with a ﬁlm thickness ranging from 2.5 µm - 125 µm. By exploring the relationship among

diﬀerent types of the magnetic materials, deposition methods, magnetic ﬁlm thickness and

the amount of group delay, it is evident that: (1) there exists a positive linear relationship

between the magnetic ﬁlm thickness and the amount of group delay, in other words, thicker

76

magnetic ﬁlm introduces higher group delay while the added loss is negligible assuming a

uniform ﬁll factor; (2) the LbL process produces more condensed magnetic ﬁlms than the

SC method but is less time eﬃcient than the latter in producing thicker ﬁlms, both methods

are proven to be eﬀective to fabricate thick magnetic ﬁlms so that higher group delay can be

introduced; and (3) given the same ﬁlm thickness and deposition method, MnFe2O4 nanopar-

ticles have a stronger ability in increasing the group delay than the CoFe2O4 nanoparticles.

Given the aforementioned, we can conclude that it is most time eﬃcient to fabricate magnetic

nanoparticle ﬁlms using manganese ferrites via the SC method to achieve a certain level of

group delay increase, which can be further enhanced by improving the ﬁll factor of the SC

method. The SC method can be easily implemented in wafer scale or board size fabrica-

tion processes. The above conclusions give the guidelines and pave the way for fabricating

monolithic microwave integrated circuits and devices utilizing such magnetic ﬁlms.

77

PATCH ANTENNA ON NIFE2O4 NANOPARTICLE AIR SUBSTRATE

CHAPTER 5

The rapid development of modern wireless communication systems has led to a continuously

growing demand for compact, eﬃcient and wideband antennas [31]. So far, various techniques

have been investigated and employed for antenna miniaturization, such as shorting pins [32],

fractal shape [33], dielectric loading [34], and using electromagnetic band gap structures or

metamaterials [35][36]. However, these techniques generally either sacriﬁce bandwidth and

eﬃciency or are hard to fabricate and integrate with monolithic circuits.

Another type of miniaturization approach is loading the antenna with bulk ferrite materi-

als or magnetodielectric materials as substrate/superstrate which exhibit relative permittiv-

ity r and relative permeability µr both greater than unity (r > 1 and µr > 1), so that the

antenna can be miniaturized by the factor of n =

√
rµr [37]–[40]. Normally a large biasing

magnetic ﬁeld is needed for bulk ferrite materials to operate at a high frequency, or their

lossy nature prevents them to be used at frequencies larger than 1 GHz under self-biased

conditions [41], [42]. Magnetic thin ﬁlms, however, have been reported for achieving a de-

cent resonant frequency shift (2 – 71 MHz), gain and bandwidth enhancement for patch and

annular ring antennas operating higher than 1 GHz under self-biased conditions [43]–[45].

However, the magnetic ﬁlm is rather thin (2 – 10 µm), thus it would be of great interest

to fabricate and utilize thick magnetic ﬁlms for substantial antenna miniaturization with

improved performance.

In our previous work, we have proposed low cost and eﬀective chemical methods of

fabricating thick magnetic ﬁlms using various material composite nanoparticles, which were

utilized to develop several RF circuits with improved performance [46]–[48]. Recently in [48],

a 30 µm MnFe2O4 nanoparticle ﬁlm loaded rectangular patch antenna achieves a frequency

shift of 0.5 GHz, a gain enhancement of 0.5 dB and an improved bandwidth by 0.14 GHz,

78

which is promising in antenna miniaturization.

In this communication, we report for the ﬁrst time much thicker Nickel-ferrite (NiFe2O4)

nanoparticle ﬁlms ranging from 875 – 900 µm within the fully 3D printed air cavity. The

loaded patch antennas exhibit a signiﬁcant resonant frequency shift of 0.75 – 1.75 GHz

with an enhanced bandwidth of 0.16 – 0.25 GHz relative to the nonmagnetic counterpart.

This large frequency shift indicates a 9.3–21.4% of size reduction. We also obtain enhanced

antenna eﬃciency and gain when good ﬁlm quality is achieved. To the best of the authors’

knowledge, this work demonstrates the thickest magnetic ﬁlms fabricated for RF and antenna

applications using a simple and cost eﬀective chemical method, and the largest resonant

frequency shift with enhanced antenna performance which leads to a great potential in low

cost, miniaturized RF circuits and antennas.

5.1 Antenna design and theoretical analysis

5.1.1 Antenna on Air Substrate

The microstrip rectangular patch antenna is designed to resonate at 8.14 GHz, and the

inset-fed topology is used to match the antenna input impedance to a 50 Ω microstrip line.
To minimize the dielectric loss, the antenna pattern is etched on top of a 2 mil (≈ 0.05

mm) Liquid Crystal Polymer (LCP) membrane which is suspended over a 1 mm deep 3D

printed metalized cavity, thus forming an air substrate antenna, as shown in Fig.1(a). The

LCP has a dielectric constant r = 3.14, and a loss tangent tan δ = 0.0025 [49], and since it

only occupies a small portion of the air substrate, the eﬀective dielectric constant of the air
substrate is very close to unity (r sub ≈ 1) and the loss tangent is close to zero. In fact, the
plastic auxiliary part shown in Fig.1(b) slightly increases the r sub, therefore, r sub = 1.1

is chosen in our design. The design parameters are labeled and provided in Fig.1.

Such an antenna conﬁguration is beneﬁcial over an antenna on a regular PCB board with
r sub (cid:29) 1, because the proposed design can enhance the antenna eﬃciency by minimizing

79

(a)

(b)

Figure 5.1: (a) The design parameters: Lp = 18.2 mm, Wp = 18.3 mm, WM S = 4.9 mm,
LM S = 9.41 mm, Gp = 0.55 mm, H = 1.05 mm; (b) Fully assembled antenna with auxiliary
part.

the dielectric loss and improving the space wave radiation. This can be easily understood

using the lossy resonator cavity model for the patch antenna, where the TM010 mode is the

fundamental mode for a broadside radiation pattern of the antenna. As we know, the total

quality factor Q of the patch antenna can be decomposed as [74]:

(cid:20) 1

Q =

+

1

Qsw

+

1
Qc

+

1
Qd

Qsp

(cid:21)−1

(5.1)

where Qsp denote the desired space wave radiation quality factor, while Qsw, Qc, and Qd

represent the dissipated power quality factors due to the TM0 mode surface wave, conductor

loss and dielectric loss, respectively. And the antenna eﬃciency can be thereby deﬁned as

[74]:

ηrad =

Q
Qsp

=

1 + Qsp

(cid:16) 1

1

Qsw

(cid:17)

+ 1
Qc

+ 1
Qd

(5.2)

It is obvious from (5.2) that the antenna eﬃciency can be enhanced by decreasing the

space-wave quality factor Qsp and increasing the loss related quality factors Qsw, Qc, and

Qd. Again, according to [74], the quality factors Qsp, Qc, and Qd are given as:

Qsp =

2ω0 r sub

hsub Gt/Lpatch

K

Qd =

1

tan δ

80

(5.3)

(5.4)

where hsub is the thickness of the substrate, σ is the bulk conductivity of the metal conductor,
Lpatch

Gt/Lpatch is the total conductance per unit length of the radiating aperture, and K =

4

for a rectangular patch antenna operating in the dominant TM010 mode.

Because of r sub ≈ 1 and tan δ (cid:28) 0.0025 of the air substrate, we can easily observe
from (7.3) and (7.4) that the Qsp and the Qd of the proposed air substrate antenna are

signiﬁcantly decreased and increased, respectively, compared to the antennas on lossy high

dielectric substrates. Meanwhile, both Qc and Qsw are not aﬀected as they are not related to

r sub and tan δ, which for the sake of brevit is not discussed here, instead, they are further

discussed in the next section of magnetic ﬁlm loading. Although the antenna eﬃciency can

be enhanced with the air substrate, the physical dimensions of the antenna are inevitably

larger than its counterparts on high dielectric substrates. Thus, it is of great interest to

miniaturize the air substrate antenna while maintaining or enhancing the performance.

5.1.2 Antenna on Magnetic Film Loaded Substrate

As mentioned in Section I, one of the eﬀective miniaturization methods is through loading the

antenna with high permittivity or high permeability materials as a superstrate to enhance

the eﬃciency while reducing the antenna size. The eﬃciency enhancement principle can

be explained using the transmission line equivalent model, as the superstrate acting as an

impedance transformer to improve the coupling between the antenna and the free space

[51]. The authors in [44], [45], and [52] demonstrated respectively that by loading magnetic

ﬁlms and dielectric as superstrate, the gain of the antenna can be enhanced. However, the

superstrate conﬁguration inevitably increases the vertical height of the overall package since

it essentially adds additional top layers to the original circuit, although the horizontal size

is reduced. Thus it is hard to integrate superstrate loaded antennas into ICs, especially in

the compact hand held devices.

In our work, by taking advantage of the ﬂexibility of the 3D printing technique and

the low cost chemical methods [46], we managed to deposit signiﬁcant amounts of NiFe2O4

81

nanoparticles within the air cavity such that the magnetic ﬁlm thickness is very close to

that of the air substrate, thereby creating an antenna on magnetic ﬁlm loaded substrate, as

shown in Fig. 2 (the auxiliary part is not shown). Note that the magnetic ﬁlm thickness

can be adjusted through our chemical deposition process, which grants us another degree of

freedom to control the antenna performance.

(a)

(b)

Figure 5.2: (a) Patch antenna on the NiFe2O4 magnetic ﬁlm loaded substrate; (b) The cross
sectional view of the magnetic ﬁlm loaded antenna.

Because of the loaded magnetic ﬁlm, the antenna substrate now has a µr sub > 1, and a

magnetic loss tangent tan δµ which depends on the inhomogeneity of the ﬁlm, the nanopar-

ticle ﬁll factor, and the ﬁlm surface roughness[53]. Therefore, (5.1) and (??) can be modiﬁed

(cid:21)−1

(5.5)

(5.6)

as:

and

(cid:20) 1

Q =

+

1

Qsw

+

1
Qc

+

1
Qd

+

1
Qm

Qsp

ηrad =

1 + Qsp

(cid:16) 1

Qsw

1
+ 1
Qc

(cid:17)

+ 1
Qd

+ 1
Qm

where the magnetic loss quality factor Qm = 1

tan δµ

. Since the dielectric loss tangent tan δ

of the NiFe2O4 nanoparticle ﬁlms is very low [46], [54], we consider that the dielectric Qd is

not aﬀected. Using the transmission line model, the total conductance Gt can be determined

as [74]:

Gt =

Wpatch
120π

k

(cid:21)

(khsub)2

(cid:20)

1 − 1
24

82

(5.7)

where k = ω0

√
µr sub r sub. Substituting (5.7) into (7.3), we ﬁnd out that Qsp is inversely

proportional to µr sub, which implies the Qsp is reduced by loading the magnetic ﬁlm, thus

more energy is radiated out. Besides, the conductor loss is reduced as Qc is increased based

on [74]:

Qc = hsub

(cid:112)πf0 σ µr sub

(5.8)

The surface-wave Qsw can be related to the space-wave Qsp when only the surface-wave

loss is considered, which is expressed as [55]:

Qsw = Qsp

(cid:32)

(cid:33)

erad
r
1 − erad

r

1

(cid:16) 3π

(cid:17)

4

(5.9)

(5.10)

(cid:17)3

1 − 1
n1

(cid:16)

where the eﬃcency erad

r

can be well approximated by [55]:

erad
r =

1 + µr sub(k0hsub)

c1

, and so is the function erad
1−erad

where c1 and n1 are constants. We can see from (5.9) that increasing µr sub leads to reduced
r ∈ [0, 1]. Since
erad
r
both Qsp and erad
1−erad
Qsw, which means the surface wave loss is increased.

are reduced with the magnetic ﬁlm loading, we obtain the reduced

in (5.8) as it is positively related to erad

r

r

r

r

At this point, the decrease of Qsp and Qsw and the increase of Qc resulted from the

magnetic ﬁlm loading still leave the behavior of the overall Q and ηrad ambiguous. From

[56], the bandwidth BW of a patch antenna printed on a magnetodielectric substrate is

approximated as:

BW ≈

(cid:113) µr sub

r sub
√

96

2(cid:2)4 + 17

√

hsub
λ0

µr sub r sub

(cid:3)

(5.11)

which shows that increasing µr sub yields broader BW . And due to the following relationship

[55]:

FBW =

VSWR − 1
√
VSWR
Q

(5.12)

where FBW = BW /f0, we can conclude that loading the magnetic ﬁlm increases the BW

and decreases the overall antenna quality factor Q.

83

Therefore, based on (5.2), we can summarize that the antenna eﬃciency ηrad can be

enhanced by loading the magnetic ﬁlm when the reduction of Q is less than the reduction

of Qsp. However, if the ﬁlm quality is poor, the excessive magnetic loss and surface loss will

lead to more reduction of Q than that of Qsp, resulting in a deteriorated antenna eﬃciency.

While the BW is enhanced due to the increased µr sub and/or loss.

5.1.3

Impedance matching, realized gain analysis

For magnetic ﬁlms that have(cid:112)µr sub/r sub ≈ 1, the impedance matching of the antenna can

be improved. Just like the superstrate, thinking in terms of the transmission line equivalent

model, the magnetic ﬁlms serves as an impedance transformer which helps to enhance the

coupling between the antenna and the air. Therefore, it will result in better input reﬂection

coeﬃcient Γin of the antenna.

With that in mind and let D denote the directivity of the antenna, then we can derive

the realized gain of the antenna on the magnetic ﬁlm loaded air substrate as

(cid:16)
1 − |Γin|2(cid:17)

Greal = ηrad

D

(5.13)

As we discussed in the last section, the radiation eﬃciency ηrad of the proposed antenna

will reduce assuming no advantageous frequency dispersion, however, it does not necessarily

imply that the realized gain of the antenna will also hereby decrease. Since we may be able to

obtain better impedance matching, the reﬂected input power of the antenna can be lowered,

thus theoretically we can have higher realized gain. But if the radiation eﬃciency is too

low, such that its negative eﬀect outweighs the contribution of the improved matching, the

realized gain will still go down. Again, our experiment results reveal that the magnetic ﬁlm

loaded antenna obtained better realized gain in certain case, but such gain can drop down

due to far worse radiation eﬃciency, as we will see in the following section.

84

5.2 The characteristics of NiFe2O4 ﬁlms and Antenna Fabrication

Process

Traditionally, bulk ferrite materials with high permeability and low loss are used in RF and

microwave applications under 600 MHz, including in antenna miniaturization [39], [40]. At

frequencies >600 MHz, the magnetic loss of these materials increases rapidly with FMR being

the most signiﬁcant loss which hinders these materials to be used in GHz frequency range.

Therefore, the FMR frequency caps the frequency range in which the antenna substrates can

operate on.

Recently, research has shown that the magnetic ﬁlm laminates, for example NiCo-ferrite

ﬁlms, under self-biased condition exhibiting high magnetization saturation and large in-plane

and out-plane anisotropy diﬀerence can have FMR frequency in several GHz, and were suc-

cessfully utilized for antenna miniaturization [43]–[45]. For such applications, two of the most

important parameters are the static relative permeability of the magnetic substrate µr sub

and the FMR frequency of the material fFMR. It is well known that these two quantities

are inversely related to each through Snoek’s law. For most bulk magnetic materials, it can

be described as

(5.14)
where γ ≈2.8 MHz/Oe which is the gyromagnetic constant, and 4πMs is the magnetization
saturation. For thin magnetic ﬁlms with in-plane anisotropy, this relation can be modiﬁed

γ 4πMs

(µr sub − 1)fFMR =

2
3

as

(µr sub − 1)f 2

FMR = (γ 4πMs)2

(5.15)

which shows the µr sub are allowed to be larger given the same FMR frequency.

It has

to be pointed out both (5.14) and (5.15) are estimated results in certain cases.

In fact,

µr sub and fFMR are dependent on the eﬀective anisotropy ﬁeld of the material, which, in

turn, depends on the material composition, as well as the manufacturing process, treatment,

small admixtures, etc. In addition, the relative permeability of the magnetic materials can

be modeled as a complex format using the Lorentzian dispersion rule at low frequencies. In

85

our paper, for simplicity, we only consider the static relative permeability µr sub since the

Lorentzian dispersion rule is not applicable at GHz range.

5.2.1 The Characteristics of NiFe2O4 Films

In our recent work, magnetic ﬁlms made of various composites that were fabricated using ef-

fective and low cost chemical methods have been characterized to exhibit high magnetization

saturation and controllable ﬁlm thickness[46]. These magnetic ﬁlms have been demonstrated

in RF circuits and antennas to show good performance[47], [48]. In this communication, we

report for the ﬁrst time, the 875–900 µm thick NiFe2O4 nanoparticle ﬁlms, which to the

best of the author’s knowledge, are the thickest magnetic ﬁlms fabricated for RF and an-

tenna applications using a simple chemical fabrication method. The NiFe2O4 ﬁlm has a

magnetization saturation 4πMs = 40.2 kG, and a linewdith ∆H = 850 G [47].

The LandauLifshitz-Gilbert (L-L-G) equation is used to describe the magnetodynamics.

By deﬁning α as the Gilbert damping factor, Happl is the applied ﬁeld, and Hk is the in-

plane anisotropy ﬁeld, we have ωappl = γµ0Happl, ωk = γµ0Hk, and ωm = γµ0Ms. Then

the in-plane susceptibility spectra of the magnetic ﬁlms is given by

(cid:0)ωm + ωk + ωappl + jωα(cid:1)
(cid:0)ωk + ωappl

−ω2 + jωαωm + ωm

(cid:1)

χ(ω) =

ωm

(5.16)

5.2.2 NiFe2O4 Films Fabrication Process

First, the cavity used for NiFe2O4 nanoparticles deposition was 3D printed using polyjet

technology, which was widely used for RF and microwave applications due to its fast pro-

totyping, low cost, and environmentally friendly advantages. We used a professional grade

commercially available Connex Objet350 polyjet printer which features a resolution of 100

µm in the X/Y axes and 30 µm in Z axis, and the cavity was printed using a ABS plastic

like photosenstive polymer resin called VeroWhitePlus [57]. The detailed information about

the polyjet technology can be referred in previous chapters.

86

After the cavity was printed, it was cleaned by ﬂushing away the unnecessary supporting

material and was further sanded for smooth surfaces. The inner surfaces of the cavity have

to be metalized to provide the RF ground plane for the patch antenna. We ﬁrst sputtered

0.5 µm of titanium as the seed layer to promote the adhesion between the VeroWhitePlus

surface and the following copper layer, then an approximate 5 µm of copper was sputtered

and electroplated. Based on our experiment trails, we found out the chemical solution that

was later used to deposit the NiFe2O4 nanoparticles was very corrosive which may etch

through the thin metal layer. Therefore, in our actual sample as shown in Figure 5.3a, a 50

µm thick copper tape was placed on top of the electroplated copper layer. Note that the

ground plane and the two edges where the SMA connectors were meant to be soldered on

was covered by the copper tape, and additional small portion of solder was applied to ensure

the positioning as well as the electrical connectivity.

(a)

(b)

Figure 5.3: (a) The 3D printed cavity with metalized surfaces; (b) The fabricated NiFe2O4
magnetic ﬁlm.

The NiFe2O4 nanoparticles ﬁlm was then deposited using the low-cost chemical method

called Solution Cast which was proposed in Chapter 4. Speciﬁcally, the nanoparticles were

ﬁrst surface treated using (3-glycidyloxypropyl)-trimethoxysilane ethanol solution, and were

then collected through a magnet to mix with 10-20% (w/w) poly(methyl methacrylate)

87

PMMA and an appropriate amount of dichloromethane. This nanoparticle based solution

was applied evenly within the 1 mm deep cavity by a syringe. After it was air-dried and

stored in a desiccator, a layer of Torlon TF4000 polymer was spin-coated onto the surface to

both chemically and physically protect and stabilize the underlying ﬁlm. Note the thickness

of the ﬁlm can be controlled through the amount of NiFe2O4 nanoparticles being deposited

within the cavity. The size of the magnetic nanoparticles can also be controlled through

managing either time or temperature of the reaction solution. Figure 5.3b shows one of the

fabricated NiFe2O4 ﬁlm loaded air cavity in the experiment test, which has the estimated

ﬁlm thickness of 0.9 mm.

Figs. 5.4(a)-(c) show the three fabricated magnetic ﬁlm samples. We can observe that all

three ﬁlms display a certain degree of inhomogeneity. Sample 2 (S2) shows the best surface

uniformity with the least number of cracks, whereas sample 1 (S1) has more ﬁlm cracks.

And the ﬁlm surface of sample 3 (S3) is shattered with a concave area in the center. The

proﬁlometry data of the ﬁlm edge and center, as indicated in Fig. 5.4(d), are provided in Fig.

5.4(e) and (f), respectively. The corresponding root-mean-square (rms) surface roughness of

S1–S3 is 13, 3, and 31 µm, respectively. Since the cavity depth is 1 mm, then by evaluating

the edge–ﬁlm surface diﬀerence, we can estimate the corresponding ﬁlm thickness of S1–S3

to be 875, 900, and 900 µm, respectively.

In fact, multiple proﬁle scan had been done for the above two areas, although the actual

data are somewhat varying, qualitatively the observation results are the same, i.e., the surface

roughness from the most to the least: S2>S3>S1 and the vertical diﬀerence from the largest

to the smallest: S1>S2>S3. Moreover, some surface areas of the ﬁlm samples accumulate

signiﬁcantly more nanoparticles than the surrounding areas creating a heavily rough proﬁle,

which may potentially damage the probe of the proﬁlometer, thus the surface roughness data

of these areas and the edge-edge cross section scan data are not available. Despite this, it

will not dramatically aﬀect our observation results. Since the variance of the edge vertical

diﬀerences is less than 30 µm based on multiple scan, our above ﬁlm thickness estimation is

88

(a)

(c)

(e)

(b)

(d)

(f)

Figure 5.4: (a)-(c): Sample S1, S2, and S3; (d) Illustration of edge and surface scan using
proﬁlometer; (e) Surface scan proﬁle; (f) Edge scan proﬁle.

still valid.

Based on our analysis, we would expect the ﬁlm roughness and thickness to aﬀect our

antenna performance. Particuarly the cracks of ﬁlm will introduce additional energy loss.

Since it is a magnetic ﬁlm loaded air substrate, the ﬁlm thickness will also aﬀect the substrate

eﬀective relative permeability µr sub. In addition, one of the important ﬁgures of the ﬁlm

quality is the ﬁll factor of the NiFe2O4 nanoparticles which evaluates the concentration of

89

(a)

(b)

Figure 5.5: (a) The lego-like assembly process; (b) Fully assembled NiFe2O4 nanoparticle
ﬁlm loaded air substrate antenna

the nanoparticles in a unit volume, however, it may require destructive measurement on the

ﬁlm. The ﬁll factor along with the above ﬁlm quantities will determine µr sub, fFMR, and

tan δµ etc, and eventually aﬀect our antenna performance.

5.2.3 Antenna Fabrication and Assembly

The designed rectangular patch antenna was patterned on top of a 2 mil Rogers LCP through

standard photolithography process, then four holes of each side were punched through near

the two edges of the antenna pattern. We also 3D printed an auxiliary part which serves

for the two purposes: (1) Evenly stretch and mechanically stabilize the LCP on top of the

ﬁlm loaded air cavity; (2) Facilitate the SMA connector assembly to the antenna. Finally,

through the lego-like process shown in Figure 5.5(a), we can achieve antenna on the NiFe2O4

nanoparticle ﬁlm loaded air substrate. The fully assembled antenna with SMA connector

that is ready for measurement is shown in Figure 5.5(b). We can see the overall antenn

package is very compact and light-weight, and the whole fabrication process is very low cost

with very little wastes.

90

5.3 Antenna measurement results and analysis

5.3.1 Antenna measurement setup

The reﬂection coeﬃcient of the antenna was measured using the Keysight N5227A Perfor-

mance Network Analyzer, which was calibrated using the SOL method up to the SMA con-

nector. And the antenna radiation pattern measurement was conducted using the SATIMO

Starlab near ﬁeld system, in which the near-to-far ﬁeld transformation was performed through

the built in software. The system is calibrated up to the cable connector using the stan-

dard horn antenna customized for the system. Figure 5.6 shows the anechoic chamber

measurement setup. We measured the antenna samples on both air substrate and NiFe2O4

nanoparticle ﬁlms loaded substrate conﬁgurations. Particularly for the later conﬁguration,

three antenna samples were measured in two scenarios: (1) The magnetic ﬁlms under self-

biased condition; (2) A DC magnetic bias was applied perpendicular to the patch plane using

permanent magnets. The corresponding measurement results are shown in the following two

sections, respectively.

Figure 5.6: The radiation pattern measurement setup using SATIMO Starlab near ﬁeld
system

91

5.3.2 Antenna on the self-biased magnetic substrate

The measured S11 of S1–S3 and the air substrate antennas are plotted in Fig. 5.7. The S11

of S1, S2 and the air substrate antennas are simulated using FEKO and HFSS, respectively.

Since S1 and S2 have better ﬁlm quality than S3, the ﬁlm is assumed to be homogeneous

and ﬂat with 875 and 900 µm thickness, respectively, and by closely matching the simulated

resonant frequency (S1: 6.39 GHz, S2: 7.04 GHz) to the measured value (S1: 6.46 GHz, S2:

7.11 GHz), we can estimate the µr sub = 15, and tan δµ = 0.04.

Figure 5.7: The measured S11: S1–S3 and air substrate antenna, the simulated S11: S2 and
air substrate antenna.

As shown in Fig. 5.7, for the air substrate case, the measured S11=-13.2 dB at the

resonance of 8.21 GHz, which agrees well with the simulated result. The small diﬀerences are

due to the fabrication error, soldering loss, and the uneven surface of the LCP membrane. By

loading the NiFe2O4 nanoparticle ﬁlms in S1–S3, we observe a down shift of the resonance.

Speciﬁcally, compared to the measured resonant frequency of the air substrate case, the

resonance down shift of S1–S3 is 0.75 GHz, 1.09 GHz, and 1.75 GHz, respectively. To

the best of the authors’ knowledge, the results show the largest frequency shifts that have

ever been reported in the literature for antenna applications using the magnetic ﬁlms as

92

substrate/superstrate, which leads to signiﬁcant antenna miniaturization.

In addition, the impedance matching and the BW are also enhanced for S1–S3. For

example, at the resonance, S2 has a S11= -18.5 dB and a BW = 0.51 GHz, which is 5.3

dB better and 0.4 GHz wider than those of the air substrate antenna. This result conﬁrms

the conclusion in Section II.C, that loading the magnetic ﬁlms results in an enhanced BW

because of the increase of µr sub. This BW enhancement is also partially due to the loss of

the ﬁlms, but is minor compared to the µr sub factor.

The measured radiation patterns of S1–S3 and the air substrate antenna are depicted in

Fig. 5.8. The measured realized gain and radiation eﬃciency of the air substrate antenna

at the resonance is 8.2 dBi and 85%, respectively, as opposed to the simulated values of 9.2

dBi and 87.4%. The diﬀerence is due to the impedance mismatch, unevenness of the LCP

and copper tape loss. While for S1–S3, the measured realized gain is 8.5 dBi, 8.8 dBi, and

6.4 dBi, respectively, with the corresponding eﬃciency of 90%, 91%, and 68%. And we can

observe that the broadside patterns of S1–S3 are mostly retained with the cross polarizations

less than -20 dB.

Compared to the air substrate case, S1/S2 enhances the gain and eﬃciency by 0.3 dB/0.6

dB and 5%/6%, respectively. Such eﬃciency improvement is explained in Section II.C that

the magnetic ﬁlm helps enhance the space wave radiation by lowering Qsp, while the good ﬁlm

quality of S1/S2 ensures minimum magnetic and surface wave loss; thus by (5.6) we obtain

enhanced radiation eﬃciency. In contrast, when the ﬁlm quality is poor, which is the case

of S3, the corresponding eﬃciency is reduced by 17%. This is evidenced by Fig. 5.4(a)–(c)

where S3 has more cracks and more than twice the surface roughness than S1 and S2, so that

the ﬁlm related loss completely counteracts the improved space wave radiation. However,

with better ﬁlm quality, we can expect more enhancement of eﬃciency, BW , and gain while

obtaining signiﬁcant antenna miniaturization. The reason that S3 has a larger frequency

shift than S2 given the same 900 µm ﬁlm thickness is because the magnetic nanoparticle

ﬁll factor of S3 is higher than that of S2, which results in a larger µr sub in S3. A detailed

93

comparison in terms of the aforementioned measured metrics of S1–S3 and the air substrate

case is presented in Table 5.2. Fig. 5.9 compares the measured maximum realized gain and

eﬃciency over the frequency band for S1–S3 and the air substrate antennas. We can see

that with good ﬁlm quality, the 3 dB gain bandwidth of S1 and S2 are signiﬁcantly wider

(approximately doubled) than that of the air substrate antenna, and a similar trend can be

also observed on the eﬃciency from Fig. 8(b). Again, these results prove that the magnetic

ﬁlm can help improve the space wave radiation.

Table 5.1: Measured performance metrics of S1–S3 and the air substrate case

Samples

Terms
Resonance fr (GHz)
Resonant shift
S11 (dB)
BW (GHz)
FBW (BW/fr)
Max. realized gain (dBi)
Eﬃciency ηrad
Film thickness (µm)
rms surf. roughness (µm)

Air

S1

S2

S3

8.21

–

7.11

-18.5
0.51

7.45
6.46
9.3% 13.4% 21.3%
-14.2
-14.2
-13.3
0.11
0.28
0.27
1.3% 3.6% 7.1% 4.3%
6.4
8.2
8.8
85% 90% 91%
68%
900
900
3
31

875
13

8.5

–
–

Table 5.2 compares the presented antennas with the recently published antennas using the

similar magnetic material loading conﬁguration or other antenna miniaturization techniques.

As we can see, the fabricated magnetic ﬁlm allows the proposed antennas to operate at a

much higher GHz range compared to the traditional magnetodielectric composite substrate

used in [38]. Given the patch antenna setting, because of the 875 µm ﬁlm thickness of S1

which is much thicker than the 8 µm thick ﬁlm used in [45] or the 30 µm thick ﬁlm used in

[48], S1 thereby has a much larger resonance shift, enhanced BW , eﬃciency and gain. Again,

we expect more of such performance enhancement and size reduction when good ﬁlm quality

is achieved. Since it is known that the resonant frequency is inversely proportional to the

antenna size/length, thus the resonant frequency shift can be interpreted as the antenna size

reduction. Although other miniaturization techniques employed in [32], [33], [36] can lead to

larger size reduction, one or more of their performance metrics are negatively aﬀected, and

94

(a)

(b)

Figure 5.8: Measured radiation patterns for S1–S3 and air substrate antennas: (a) E plane;
(b) H plane. Measured resonant frequencies: 7.45 GHz (S1), 7.11 GHz (S2), 6.46 GHz (S3),
8.21 GHz (Air).

95

(a)

(b)

Figure 5.9: Measured maximum realized gain and eﬃciency vs. frequency for S1–S3 and air
substrate antennas: (a) maximum realized gain; (b) eﬃciency.

some of them are relatively diﬃcult to fabricate.

Therefore, based on our results, we can conclude that by using the proposed magnetic ﬁlm

loaded antenna, we can signiﬁcantly miniaturize the antenna without comprising the antenna

performance, instead, enhancing the antenna’s bandwidth, eﬃciency, and gain at the same

time which is usually not possible using traditional techniques. Especially in applications

where high antenna gain and wide bandwidth are preferred over certain miniaturization, this

work provides a novel way to satisfy such requirement and poses many advantages than the

conventional designs.

5.3.3 Antenna on the DC magnetic biased magnetic substrate

One of the special characteristics of the magnetic material is the ferromagnetic resonance

phenomenon, which represents the absorption capability of a magnetic material to electro-

magnetic waves of some frequency. The small-signal theory of FMR was developed by Kittel.

This intrinsic property gives rise to the use of the magnetic materials in many tunable RF

and microwave devices to achieve the unique performance. The FMR and the magnetic

susceptibility characterization of the similar nanoparticles to our NiFe2O4 nanoparticles are

96

Table 5.2: Comparison of the reference antennas (data format in the upper half: before/after
magnetic material loading, size reduction compared to the patch without loading; lower half:
size reduction compared to the λ

2 patch).

Ref.

[38]

[44]

[45]

[48]

S1

S2

S3

fr & fr shift
size reduction
100/98 MHz
2 MHz / 2%

BW &
FBW

ηrad
(%)

Gain
(dBi)

N.G./2 MHz

N.G./2%

N.G. N.G./0

1.72/1.671 GHz
49 MHz / 2.8%

5/7 MHz
0.3/0.4%

N.G.

0.6/1.3

2.147/2.097 GHz
50 MHz / 2.3%

18/39 MHz

0.8/1.9%

41/65

1.3/1.7

8.3/7.8 GHz

500 MHz / 6%

240/340 MHz

2.9/4.3%

N.G.

6.5/7

8.21/7.45 GHz
760 MHz / 9.3%

110/270 MHz

1.3/3.6%

8.21/7.11 GHz

110/510 MHz

1100 MHz / 13.4%

1.3/7.1%

8.21/6.46 GHz

110/280 MHz

1750 MHz / 21.3%

85 /
90

85 /
91

85 /
68
ηrad
(%)

77.2

8.2/8.5

8.2/8.8

8.2/6.4

Gain
(dBi)

5.2

9

-4.9

N.G.

N.G.

19.8

-3.9

Ref.

[32] DSPA
Rogers 5880

fr &

size reduction

2.153 GHz

62%

[32] DSPA

2.455 GHz

FR4

62%

[33] Fractal

patch

[36] Patch
on MTM

5.2 GHz

38%

250 MHz

85%

1.3/4.3%
BW &
FBW

9.9 MHz

0.46%

37 MHz

1.5%

21 MHz

0.4%

2 MHz
0.83%

referenced in Chapter 4.

In the antenna applications, the DC magnetic bias is usually applied either through

97

permanent magnets or electromagnetic circuits embedded inside the substrate. The latter

approach usually requires costly and complicated fabrication process. By applying the mag-

netic bias, the antenna can be granted many beneﬁcial properties such as resonant frequency

tuning, the change of polarization sense, etc. Here we simply used permanent magnets as

the DC magnetic bias source, trying to explore the corresponding eﬀects. As we can see in
Figure 5.10, four 7 mm × 7 mm cylinder Neodymium permanent magnets were put together

at the back of the antenna, and a double sided tape was placed in between to ensure the

magnets adhere to the antenna. The ﬁeld strength provided by the magnets is measured as

975 Gauss. The measured S11 and the radiation patterns are compared against the scenario

of self-biased case, which are shown in Figure 5.11 and Figure 5.12(a)–(f), respectively. Note

that the antenna samples used in this measurement are not the same samples measured in

Fig. 5.7.

Figure 5.10: The antenna sample with permanent magnets on the back

From Figure 5.11, we can obviously see that the DC magnetic bias causes signiﬁcant

changes to the antenna performance for all three samples, among which S3 experienced

the most, i.e., the eﬃciency was reduced 23.1%, the maximum realized gain was lowered

by 2.7 dB, and the level S11 dropped down by 32.4 dB. It is interesting to note that for

S3, even the impedance matching was improved signiﬁcantly, the ﬁnal realized gain was

still reduced. This is because the heavily degraded eﬃciency counteracted such impedance

98

Figure 5.11: Comparison of the measured S11 of the magnetic ﬁlms loaded antennas under
self-biased and DC magnetic biased conditions.

Table 5.3: Measured performance metrics of the antenna samples in under self-biased and
DC magnetic biased conditions (Date format: self-biased / DC magnetic biased).

Samples

Terms
Res. freq.fr (GHz)
S11 (dB)
BW (GHz)
FBW (BW/fr)
Max. r. gain (dBi)
Eﬃciency (self)
Eﬃciency (DC)

S1

S2

S3

6.46/ 6.58
7.45/ 7.55
-14.2/ -47.6
-14.2/ -37.1
0.27/ 0.45
0.28/ 0.45
3.6%/ 6.0% 5.1%/ 6.3% 4.3%/ 6.8 %

7.11/ 7.17
-21.0/ -43.8
0.36/ 0.45

7.2/ 7.0
71.5%
65.6%

5.4/ 4.6
52.3%
42.1%

5.1/ 2.4
48.1%
25.0%

matching improvement completely, resulting in an overall negative eﬀect to the antenna

gain, which can be again explained by (5.13). Similar observation can be found for both

S1 and S2. This is due to the DC magnetic bias excites the ferromagnetic resonance of the

NiFe2O4 nanoparticles creating excessive magnetic loss of the magnetic ﬁlm, such that part

of the accepted power by the antenna was absorbed by the magnetic nanoparticles, instead

of getting reﬂected back. Therefore, strictly speaking, applying DC magnetic bias does not

essentially improve the physical impedance matching at the port of the antenna; but rather,

99

(a)

(b)

(c)

(d)

(e)

(f)

Figure 5.12: The measured E and H planes radiation patterns of sample S1, S2 and S3 under
self-biased and DC magnetic biased conditions. S1:(a)(b), S2:(c)(d), S3:(e)(f).

100

raises the magnetic loss of the magnetic ﬁlm which reduces the associated magnetic Qm,

causing the overall antenna eﬃciency to be degraded. Therefore, the more the eﬃciency is

degraded, the stronger the magnetic loss is excited. Due to this increased lossy eﬀect, the

total antenna Q is reduced. Then given the relationship provided in (5.12), the bandwidth

is thereby increased which is what we see from the measurement results.

Furthermore, since we know higher susceptibility of the magnetic ﬁlm gives rise to

stronger FMR eﬀect, which in turn is determined by the ﬁlm thickness and especially the ﬁll

factor of the ﬁlm; thus based on our experiment results we can conclude that not only the

ﬁlm thicknesses follows the order S3–S1 from the most to the least, but also the ﬁll factors

spans from the highest to the lowest.

With the above knowledge, we know that the antenna eﬃciency can be enhanced using

the air substrate conﬁguration and the antenna can be miniaturized by loading the mag-

netic ﬁlm within the air substrate to obtain much more compact package compared to the

traditional superstrate loading approach, while bandwidth and impedance matching can be

improved. Tradeoﬀ is expected between the antenna miniaturization level and antenna ef-

ﬁciency; however, by carefully choosing the magnetic ﬁlm thickness and improving the ﬁlm

quality, one can obtain both improved eﬃciency and realized gain. By applying the DC

magnetic bias, the antenna performance can be aﬀected, noticeably, the degraded radia-

tion eﬃciency and realized gain. However, for applications which requires signiﬁcant loss

characteristics instead of radiation at certain frequency, e.g. ﬁlters, the magnetic loss from

self-biased and DC magnetic biased conditions are actually advantageous.

5.4 Summary

In this chapter, we designed, fabricated and measured the antennas on the compact NiFe2O4

nanoparticle ﬁlms loaded air substrates. First, we analyzed the eﬃciency improvement pro-

vided by the antenna on air substrate conﬁguration, and further discussed the antenna on

magnetic ﬁlm loaded air substrate conﬁguration in terms of antenna miniaturization, radia-

101

tion eﬃciency, bandwidth and several other performance metrics. Then we used 3D printing

technology and eﬃcient chemical synthesis methods to fabricate antenna samples of both

conﬁgurations with a compact, low cost, and light-weight package. All four antenna samples

were measured, especially the magnetic ﬁlms loaded antennas were tested in both self-biased

and DC magnetic biased conditions to further investigate the antenna performance. The

measurement results show very good antenna performance and were analyzed theoretically

to gain additional insightful understanding, which provides a useful guide to using such

magnetic material and conﬁgurations to design miniaturized RF and microwave applications

with enhanced performance.

102

TUNABLE BANDSTOP FILTER ON NANOMAGNETIC THIN FILMS

CHAPTER 6

With the rapid development of the wireless communication technologies, the RF bandwidth

scarcity problem is becoming more critical. On the other hand, to better utilize the already

crowded frequency spectrum, one of the promising methods is to use frequency-tunable

microwave devices. On the other hand, the widespread usage of wireless devices poses serious

RF interference problems such that the unwanted signals in a speciﬁc frequency range have

to be suppressed. This can be done generally by bandpass ﬁlters, however, in some cases

when the interference signal is particularly strong or a limited number of close frequencies

have to be kept apart, one or more bandstop ﬁlters will be more eﬃcient than a bandpass

ﬁlter which discriminates against all frequencies outside the pass band [58]. Therefore, the

demand for broadband tunability, low cost and compact bandstop ﬁlters (BSFs) has been

continuously increasing in modern wireless communication and radar systems.

Traditionally, research on technologies in achieving ﬁlter tunability were mainly done on

the bandpass ﬁlters, which includes electronic tuning through microelectromechanic system

varactors or ferroelectric varactors [59][60], mechanical tuning by piezoelectric-transducer

controlling[61], magnetoelectric coupling based E- and H-ﬁeld tuning [62], and magnetic

tuning using ferrite based thin ﬁlm/slab loaded ﬁlters [4][63][64]. Among the above four

categories, the magnetically tunable ﬁlter is attractive because of its compactness, low cost

and fast tuning mechanism. Most recently, in [65], a slim ribbon shaped BSF integrated with

a 240 µm thick FeGaB/Al2O3 multilayer ﬁlm was proposed showing a good tunability of 3.5

GHz by applying a rather modest magnetic bias ﬁeld less than 400 Oe, and the stopband

rejection was over 28 dB with a 3 dB in band reﬂection loss.

In this section, a microstrip BSF was ﬁrst fabricated through a low cost thin ﬁlm and

lego-like process as shown in [29][66]. The very low loss air substrate contributes to a 41 dB

insertion loss at 8.5 GHz and a 1 dB in band reﬂection loss. Using the low cost Solution Cast

103

method, various types of magnetic material nanoparticles-MnFe2O4, ZnFe2O4, NiFe2O4,

and half-half NiFe2O4 and ZnFe2O4 mixed-were deposited within the air cavity to form the

magnetic ﬁlm loaded air substrate. Due to the ﬂexibility of the 3D printing technology, the

magnetic ﬁlm thickness can be controlled, such that the fabricated ﬁlm thickness ranges from

30 µm to 100 µm, which gave the corresponding self-resonances of the ﬁlm integrated BSFs

from 7.6 GHz to 8.3 GHz. Moreover, by applying a vertical DC magnetic bias of 2500 Oe,

we can observe a frequency shift eﬀect, for example from 7.6 GHz to 9.1 GHz. The proposed

magnetic ﬁlm integrated BSF demonstrates not only good in band insertion and reﬂection

loss by taking advantage of the 3D printed air substrate, but also a very compact size for

working on multiple frequencies. Therefore, this process shows a great deal of potential for

the fabrication of low cost and compact magnetically tunable microwave devices such as

ﬁlters, resonators and antennas.

6.1 Magnetic nanoparticles properties and bandstop ﬁlter design

6.1.1 Magnetic Nanoparticles Properties

It is known that the unique non-linear property of the magnetic materials can be attributed

to the Ferromagnetic Resonance (FMR) phenomenon, which represents the absorption fre-

quency of a magnetic material to certain electromagnetic waves.

in [67], the small signal

theory of FMR was developed. This intrinsic property gives rise to the use of magnetic

materials in many tunable and frequency selective microwave devices. Note that diﬀerent

types of magnetic materials may display a distinct FMR eﬀect which is due to the magnetic

susceptibilities of the materials. Since FMR is really an excitation of the nanoparticles be-

tween two energy states, with a higher magnetic susceptibility, the degree of magnetization

would be larger in response to an applied magnetic ﬁeld, and thus cause a larger energy dif-

ference between two energy states. In this work various magnetic nanoparticles—MnFe2O4,

NiFe2O4, and ZnFe2O4—were studied for their FMR properties by carrying out the Elec-

104

tron Paramagnetic Resonance (EPR) measurement. For example, Fig. 6.1 shows the FMR

absorption ﬁles of MnFe2O4 nanoparticles with particle size of 6 nm and 7 nm [46]. At 9.88

GHz, the FMR of the MnFe2O4 nanoparticles ﬁlm appears around 3300 Gauss of EPR ﬁeld,

and exhibits a 1400 Gauss of linewidth. Similar FMR proﬁles were obtained for NiFe2O4

and ZnFe2O4 nanoparticles, and the FMRs were around 3200 Gauss and 3400 Gauss with a

linewidth of 850 Gauss and 1000 Gauss respectively. The relatively narrow linewidth implies

these magnetic materials have good potential in RF applications.

Figure 6.1: Selected FMR absorption proﬁles of MnFe2O4 nanoparticles [? ].

6.1.2 Bandstop Filter Design

The BSF is designed to operate at the center frequency of 8.5 GHz with a 3 GHz bandwidth,

and to have a rejection level higher than 30 dB at the center frequency. It is implemented

using the microstrip open-circuited stubs and is optimized by incorporating the redundant

unit elements in the design, so that a wide stopband and steeper rejection characteristics can

be obtained [68]. Then we take advantage of the 3D printing technology to create a low loss

air substrate such that the ﬁlter is suspended above the ground plane to minimize the in band

105

reﬂection loss of the BSF. The 3D printed lego-like structure facilitates the ﬁlter assembly.

Moreover, compared to the separate ﬁlm fabrication and post ﬁlm-ﬁlter integration in [63],

we can directly deposit the magnetic nanoparticles within the air cavity to form the magnetic

ﬁlm, so that the ﬁlm is integrated simultaneously as a magnetic ﬁlm loaded air substrate. In

[63], the yttrium iron garnet (YIG) material was either deposited on the gadolinium gallium

garnet (GGG) substrate or sandwiched between two GGG substrates to fabricate a stand-

alone magnetic ﬁlm; then the ﬁlm was ﬂipped on top of the ﬁlter surface as a superstrate to

enable magnetic tuning, and additional clamping structures maybe needed for mechanical

stability. Fig.6.2 illustrates the topology and design parameters of the BSF. The detailed

assembly and ﬁlm fabrication will be discussed in Section III.

Figure 6.2: The design of the BSF on the air substrate: LM = 25.5 mm, WM 1 = 4 mm,
WM 2 = 3.6 mm, LS1 = 7.3 mm, WS1 = 0.8 mm, LS2 = 7.2 mm, WS2 = 2.3 mm, H = 1
mm, t = 120 µm - 300 µm.

6.2 Magnetic Films and Filter Fabrication

6.2.1 Magnetic Film Loaded Air Cavity Fabrication

A square well with a certain depth of either 30 µm or 100 µm was 3D printed on top of the

copper sheet creating an air cavity, as shown in Fig. 6.3a. A photopolymer resin material

called VeroWhitePlus and a commercially available Connex Objet350 3D printer , which

features a resolution of 100 µm in the X/Y axes and 30 µm in Z axis [? ], were used.

106

(a)

(b)

Figure 6.3: Fabricated sample: (a) 3D printed air cavity on the copper sheet; (b) 100 µm
thick MnFe2O4 ﬁlm loaded air cavities.

In the next step one of the following magnetic nanoparticles: MnFe2O4, NiFe2O4, ZnFe2O4,

and half-half mixed NiFe2O4 and ZnFe2O4, were deposited within the cavity through the SC

method. Speciﬁcally, the nanoparticles were ﬁrst surface treated using (3-glycidyloxypropyl)-

trimethoxysilane ethanol solution, and were then collected through a magnet to mix with 10-

20% (w/w) poly(methyl methacrylate) PMMA and an appropriate amount of dichloromethane.

This nanoparticle based solution was applied evenly within the cavity by a syringe. After it

was air-dried and stored in a desiccator, a layer of Torlon TF4000 polymer was spin-coated

onto the surface to both chemically and physically protect and stabilize the underlying ﬁlm.

Note that by controlling the depth of the 3D printed cavity, we can have the ﬁlm in a variety

of thicknesses; here we have 30 µm and 100 µm. The size of the magnetic nanoparticles

can be also controlled through managing either time or temperature of the reaction solution.

Fig. 6.3b shows a fabricated MnFe2O4 ﬁlm loaded air cavity which has a ﬁlm thickness of

100 µm.

107

6.2.2 Bandstop Filter Assembly

The microstrip BSF design as shown in Fig. 6.2 was patterned on top of a Rogers Ultralam

3850HT LCP substrate through photolithography. The LCP substrate has a dielectric con-

stant of 3.14, a loss tangent is 0.002 [? ], and the thickness is chosen as 2 mil to minimize the

dielectric loss. We also 3D printed two other auxiliary structures A1 and A2 for the follow-

ing two purposes: (1) Evenly stretch and mechanically stabilize the LCP on top of the ﬁlm

loaded air cavity; (2) Facilitate the RF connector assembly for the ﬁlter. Finally, through

the lego-like process shown in Fig. 6.4a, we can achieve an air suspended BSF loaded with

the magnetic ﬁlm where the air substrate is 1 mm thick. Due to the fact that the thickness of

the LCP is only 2 mil, which is very thin compared to that of the air substrate, the eﬀective

dielectric constant of the air substrate is slightly larger than 1. Fig. 6.4b which shows the

fully assembled BSF with two SMA connectors that is ready for measurement.

(a)

(b)

Figure 6.4: Fabricated sample: (a) The lego-like assembly process; (b) Fully assembled
microstrip BSF.

6.3 Filter Measurements and Discussions

The S-parameters of the ﬁlters were measured using a Keysight N5227A Performance Net-

work Analyzer which was calibrated with the SOLT method up to the RF connectors. The

108

S21 and S11 denote the insertion and the reﬂection loss of the ﬁlter, respectively, and are

shown in Fig. 6.5a - 6.5b. For the non-magnetic ﬁlm loaded air substrate BSF, both the

measured insertion and reﬂection loss agree with the simulations. Due to the low loss nature

of the air substrate, we can see the steep rejection characteristic with 41.1 dB of insertion loss

at 8.5 GHz, and the minimum 1.6 dB of in band reﬂection loss. Also, the out-of-band inser-

tion loss can be as low as 0.33 dB and the reﬂection loss is better than 12 dB. When loaded

with the magnetic ﬁlms, the self-resonances of the ﬁlters shift downward which can be as low

as 7.6 GHz which corresponds to 0.9 GHz of frequency shift due to the 30 µm ZnFe2O4 ﬁlm.

Similar results can be obtained for the other ﬁlms. In addition, we can observe that for all

the magnetic ﬁlm loaded cases, the inband insertion loss is higher than 30 dB, and the reﬂec-

tion loss can be as good as 0.7 dB. The out-of-band insertion and reﬂection loss are also very

good. This frequency shift eﬀect due to the magnetic ﬁlm loading is because the magnetic

ﬁlm has a relative permeability µr > 1, which serves as a perturbation to the air substrate.

The electromagnetic wavelength λ =

√

f

1
rµr

at a ﬁxed frequency in the air substrate reduces

when loaded with the magnetic ﬁlm, causing the ﬁlter to be electrically larger. Therefore,

the center frequency of the magnetic ﬁlm loaded BSF shifts lower. Moreover, we notice that

the frequency shifts of both the 30 µm ZnFe2O4 and NiFe2O4 ﬁlms are very close to that

of the 100 µm MnFe2O4 ﬁlm. This indicates that the ZnFe2O4 and NiFe2O4 nanoparticles

have higher magnetic susceptibilities than the same amount of MnFe2O4 nanoparticles. The

above results imply that we can miniaturize such BSF or other circuits on the similar air

substrate package that is operating at a ﬁxed frequency by just loading the magnetic ﬁlms;

and the thicker the magnetic ﬁlms or the higher magnetic susceptibility of the nanoparticles,

the smaller the ﬁlter size. Again, because the magnetic ﬁlm is simply integrated with the

substrate and the ﬁlm thickness can be easily controlled by the 3D printed air cavity, the

overall package of the BSF is very compact and the fabrication is very low cost.

109

(a)

(b)

Figure 6.5: Measured (a) insertion loss and (b) return loss of the ﬁlms and non-ﬁlms loaded
BSF. The corresponding ﬁlm thicknesses t of MnFe2O4, NiFe2O4, ZnnFe2O4 and half-half
mixed NiFe2O4 and ZnnFe2O4ﬁlms are 100 µm, 30µm, 30 µm, and 30µm.

To further explore the eﬀect of the loaded magnetic ﬁlms in terms of the material types

and ﬁlm thicknesses, we applied an external DC magnetic bias ﬁeld of 2500 Oe to the ﬁlter.

The DC magnetic bias was applied vertically to the plane of the ﬁlm loaded ﬁlter by placing

the ﬁlter between the two poles of the electromagnets. The results are shown in Fig. 6.6a -

6.6d. A shift of the operation frequency can be observed in all four ﬁlm cases. For example

in Fig. 6.6a, the stopband center frequency of the MnFe2O4 ﬁlm case increased from 7.8

GHz to 8.7 GHz when 2500 Oe of the magnetic bias was applied, while the 10 dB bandwidth

shrinks from 3.3 GHz to 2.5 GHz. Similar results can be observed for the NiFe2O4 ﬁlm

case. While keeping the same ﬁlm thickness of 30 µm, NiFe2O4 ﬁlms resulted in even larger

resonant frequency shift of 1.5 GHz (from 7.6 GHz to 9.1 GHz) at 2500 Oe of DC magnetic

bias, as we can see from Fig. 6.6e. This is due to the higher magnetic susceptibility in

NiFe2O4 nanoparticles, which implies that a smaller magnetic bias ﬁeld is needed for the

same amount of tuning when compared to the other ﬁlms. We can also see that the rejection

at the center resonance increases with the magnetic bias strength for NiFe2O4 and the

mixed ﬁlms; however it remains relatively unchanged for the MnFe2O4 and ZnFe2O4 ﬁlms.

For instance in Fig. 6.6e, the rejection at the resonant frequency increases from 30 dB to

46 dB as the magnetic bias is at 2500 Oe. This again is due to the larger magnetization

110

saturation of the NiFe2O4 nanoparticles. Table 6.1 provides the comparison of the center

resonant frequency, the rejection at the center resonant frequency, the 10 dB bandwidth,

and the magnetic ﬁlm thicknesses of the BSF before and after applying the 2500 Oe of the

magnetic bias.

111

(a)

(c)

(e)

(g)

(b)

(d)

(f)

(h)

Figure 6.6: Measured insertion loss and reﬂection loss of all four ﬁlm cases under the DC
magnetic bias of 2500 Oe.

112

Table 6.1: Performance metrics of the BSF before and after applying the 2500 Oe of magnetic
bias (Data format: before/after).

Films

Terms
Ctr.freq (GHz)
Rej.level (dB)
BW (GHz)
Film thickness

MnFe2O4 ZnFe2O4 NiFe2O4 Mixed

7.8/8.7
38/36
3.1/2.5
100 µm

8.1/9
33/32
2.7/2.1
30 µm

7.6/9.1
30/46
3.3/2.5
30 µm

7.6/8.7
35/40
3.3/2.6
30 µm

6.4 Summary

In this Chapter a magnetic ﬁlm loaded microstrip BSF was developed and characterized.

Various types of magnetic ﬁlms with diﬀerent thicknesses were deposited within a 3D printed

cavity well using the low cost SC method. The self resonances of the magnetic ﬁlms loaded

BSFs ranges from 7.6 GHz to 8.3 GHz. In addition, by applying a vertical DC magnetic

bias of 2500 Oe, the NiFe2O4 ﬁlm loaded BSF exhibits a center frequency shift of 1.5 GHz,

as well as 16 dB more rejection at the center frequency. This work demonstrated a low

cost and compact magnetically tunable BSF which has good potential in RF and microwave

applications.

113

TUNABLE SPHERICAL CAVITY RESONATOR USING MAGNETIC FILM

CHAPTER 7

Tunable resonators are attractive due to the performance agility to meet diﬀerent regional

industry standards which makes it a key role in RF front ends. Traditionally, most of

the tunable resonators were developed to be electronically tunable based on RF MEMS or

ferroelectric tuned elements [70]–[71], which are usually diﬃcult and expensive to fabricate.

Recently, the authors in [72] proposed a magnetically tunable ferrite slab loaded substrate

integrated waveguide cavity resonator which achieved a tuning range of more than 6% and an

unloaded Q (quality factor) more than 200. Such tuning often requires an external magnetic

bias circuitry which makes the system large and complex. And the above tunable resonators

usually have relatively large tuning step which is ineﬀective when ﬁne tuning is desired.

Lately, a spherical cavity resonator was demonstrated to have an unloaded Q of 3728 using

3D printing technology [73]. Thus, it is possible to develop high Q ﬁne–tunable resonator

with cost eﬀective fabrication process and simple topology.

In this paper, a 3D printed X-band spherical cavity resonator with very ﬁne tuning

using a magnetic nanoparticle ﬁlm is proposed and characterized for the ﬁrst time. The

polyjet 3D printing technology facilitates the fabrication of light-weight spherical cavity

assemblies for tuning. A magnetic nanoparticle ﬁlm deposited on a LCP host substrate is

placed at the center of the 3D printed cavity where the magnetic ﬁelds are the strongest.

By mechanically rotating the ﬁlms the resonant frequency can be tuned accordingly due to

the perturbed magnetic ﬁelds. The measurement results show a frequency tuning range of

70 MHz with the average and minimum tuning step of 11.7 MHz and 4.4 MHz, respectively.

The corresponding unloaded quality factors ranges from 539 to 732, which is at the expense

of a narrow tuning range. However, the tuning range can be increased by improving the

magnetic nanoparticle ﬁll factor in the ﬁlm and choosing a ferrite material with a higher

114

µr. This work shows a cost eﬀective and light-weight high Q spherical cavity resonator with

very ﬁne tuning based on a simple design, which can be used in designing high performance

ﬁlters and other microwave structures.

7.1 Tunable Resonator Design and Theoretical Analysis

7.1.1 Spherical Cavity Resonator Design And Mode Analysis

The spherical cavity is known for its extremely high Q compared to its rectangular and

cylindrical counterparts, as it does not posses any sharp corners or edges, thus its volume

and surface area are better utilized by the interior ﬁelds. Depending on the orientation of

the designated coordinate system, the ﬁeld conﬁguration can be treated as TEr and/or TMr

mode; and based on the ﬁeld distributions, the exact mode can be determined. The resonant

frequency of the lowest-order TMr

m11 mode of the air-ﬁlled spherical cavity as shown in Fig.

7.1 can be found as [74]:

(fr)T M r

m11 =

ζ(cid:48)
√
2πR

11

µ

, m = 0 or 1

(7.1)

where R = 16 mm, ζ(cid:48)
which yields (fr)T M r

011 = (fr)T M r

11 = 2.744, µ = 4π × 10−7 H/m, and  = 1/(36π) × 10−9 F/m,
111 (odd) = 8.18 GHz, corresponding to

111 (even) = (fr)T M r

three degenerate modes. Thus the ﬁrst step is to determine which mode(s) is (are) excited

speciﬁcally.

The spherical cavity is designed to be electrically excited by two probes oriented along

the Z axis, as indicated in Fig. 7.1. We can thereby plot the magnitude of the total H

(magnetic) ﬁeld in the XY plane as shown in Fig. 5.2a. It can be seen that the total H

ﬁeld is symmetrical which is invariant of the azimuth angle φ. By examining all the H
components (r, θ, φ) of three degeneracies and letting the elevation angle θ = 90°, we ﬁnd

out that the observed H ﬁeld distribution corresponds solely to the Hφ component of TMr

011

mode, which we can conclude is the only excited mode within the spherical cavity. To verify

115

Figure 7.1: The electrically excited air-ﬁlled cavity resonator with rotating rods (magnetic
ﬁlm is not shown for clarity). R = 16 mm, T = 3 mm.

the above conclusion, the Hφ component of TMr

011 mode can be written as [74]:

(cid:16)

(cid:17)

Hφ =

C
µ r

ˆJ1

2.744

r
R

sin θ

(7.2)

where C is constant, r is the radial coordinate of the spherical cavity. Since Hr = Hθ = 0
011 mode, therefore in the XY plane where θ = 90°, the total magnetic ﬁeld
for the TMr

(cid:1), which shows that the magnitude of the total H ﬁeld in

is Htotal = Hφ = C
µr

ˆJ1

(cid:0)2.744 r

R

the XY plane is independent of φ and varies with r following the special spherical Bessel

function ˆJ1(2.744 r

R ), as can be seen from Fig. 5.2b. This mode analysis lays the foundation

of implementing the magnetic tuning mechanism, as shown in the next section.

116

(a)

(b)

Figure 7.2: Magnetic ﬁeld distributions in the XY plane: (a) simulated total magnetic ﬁled;
(b) normalized theoretical Hφ using (7.2).

7.1.2 Tunable Resonator Design

The perturbation theory states that the resonant frequency of the cavity can be decreased

by increases in the permittivity  and permeability µ of the ﬁlled material which interacts

with the internal ﬁelds [75]. Hence, the frequency tuning in this work is achieved through

perturbing the H ﬁeld inside the cavity using the magnetic ﬁlm.

As shown in Fig. 7.2a and (7.2), the total H ﬁeld of the dominant TMr

011 has the maxi-

mum value in the XY plane, thus the magnetic ﬁlm will result in the most resonant frequency

tuning eﬀect when it is placed in the XY plane. Moreover, the planar shape and size of the

magnetic ﬁlm, as well as the ﬁlm thickness also determine the tuning range. Obviously, a

circular shaped magnetic ﬁlm will have the strongest tuning eﬀect compared to other planar

shaped ﬁlms, as it has the full interaction with the total H ﬁeld in the XY plane.

In practice, given the fabrication limitation, a rectangular magnetic ﬁlm is suspended
in the XY plane (θ = 90°) by adhering to the ends of two dielectric rods protruding from

the cavity, as shown in Fig. 7.1. By mechanically rotating the rods along the X axis, the

magnetic ﬁlm is rotated in an angle of θ with respect to Z axis. In this case, the magnetic

ﬁlm has less interaction with the H ﬁeld whose strength is also decreased with smaller θ.

117

Therefore, the resonant frequency is tuned accordingly with the rotation of the ﬁlm, and the

very ﬁne tuning step can be achieved by accurately controlling the rotation angle θ, whereas

the tuning range is dictated by the geometric and magnetic properties of the nanoparticle
ﬁlm. Note that the mechanical tuning range is restricted by θ ∈ [0°, 90°], as the Hφ in (7.2)
varies sinusoidally with respect to θ given a ﬁxed r, such that the frequency tuning dynamic
repeats in a period of 90°.

7.2 Resonator Fabrication Process

7.2.1 Magnetic Film Deposition And Characteristics

Using a cost eﬀective Solution Cast chemical method in [46], the magnetic ﬁlms made of
NiFe2O4 nanoparticles are fabricated on a 25 mm × 15 mm × 2 mil rectangular LCP host
substrate, as shown in the inset of Fig. 7.3. The measured average ﬁlm surface roughness is

1.5 µm. By measuring the diﬀerence between the ﬁlm and LCP surface, the ﬁlm thickness

is estimated to be 35 µm, as indicated in Fig. 7.3.

Figure 7.3: Measured ﬁlm surface roughness and edge proﬁle. Inset:
ﬁlm on a 2 mil LCP host substrate.

fabricated NiFe2O4

118

7.2.2 Resonator Fabrication And Assembly

Two hemispherical resonator parts are 3D printed using a photopolymer resin material called

VeroWhitePlus and a commercially available Connex Objet350 polyjet printer, which has a

resolution of 100 µm in the X/Y axes and 35 µm in Z axis [66]. Then the inner surfaces

of the cavities are metalized with a layer of 15 µm thick copper using electroplating, as

shown in Fig. 7.4a. The 3 mm thick cavity wall helps the cavity to remain mechanically

in shape during the metalization process. In addition, as mentioned earlier, two 3D printed

dielectric rods are used to suspend and rotate the ﬁlm to achieve the frequency tuning.
Note the magnetic ﬁlm is adhered to the ends of the rods using two 1 mm × 1 mm tapes.

The two cavity parts are then assembled through a lego-like process. An amount of silver

epoxy is applied to the interfaces of the two hemispherical parts to ensure reliable electrical

connection, and two SMA connectors are used to characterize the cavity. The whole package

as depicted in Fig. 7.4b is 60 g in total weight, which is very light-weight.

Figure 7.4: Fabricated samples: (a) metalized cavity parts with magnetic ﬁlm placed: (b)
fully assembled cavity resonator; (c) enlarged picture of corrugated rod sleeve; (d) rotation
position at θ = 15° and θ = 30°.

In order to precisely rotate the ﬁlm, a corrugated rod sleeve with a rotation lock pin is

119

designed, as displayed in Fig. 7.4c–d. The corrugated sleeve has 24 evenly spaced slots on
the circumference, which gives 15° per slot. For each rotation, a lock pin is used to pass

through the slot and the hole in the rod to stabilize the ﬁlm. Fig. 7.4d shows the ﬁlm is
rotated at θ = 15° and 30°, respectively.

7.3 Results and Analysis

The S-parameters of the cavity resonator are measured using a Keysight N5227A Perfor-

mance Network Analyzer which is calibrated with the SOLT method up to the SMA connec-

tors. From Fig. 7.5 we can observe that the measured S parameters matches very well with

the simulated results for the air-ﬁlled cavity resonators. The measured resonance at 8.05

GHz has a S11=-24.2 dB and S21=-0.71 dB, which is slightly diﬀerent from the correspond-

ing simulated results of 8.06 GHz resonant frequency with S11=-22.7 dB and S21=-0.14 dB.

The minor diﬀerence is due to the cavity surface roughness, and mainly due to the resistance

of the solder applied at the conjunction of the probe and the inner cavity surface. This re-

sistance is estimated to be 14.5 Ω by simulating the solder ring as shown in the inset of Fig.

7.5, such that the simulated S21 and S11 at 8.05 GHz are -0.71 dB and -23.3 dB, respectively,

which matches closely with the measured results.

The measured results of the tunable resonator are shown in Fig. 6, where the ﬁlm is
rotated by θ ∈ [0°, 90°] per 15°. The simulation with the same ﬁlm size and µr = 15 was
carried out using HFSS. The plot is not shown for clarity and the key simulated data are

provided in the following discussion. From Fig. 6, We can see the resonant frequency shifts
downward with the increased rotation angle θ. When θ = 0°, the resonant frequency is at

8.028 GHz which is 25 MHz lower than that of the air-ﬁlled cavity. This is because the 2 mil

LCP has a dielectric constant of 3.14 [49], thus slightly perturbs the electrical ﬁeld in the XZ
plane (θ = 0°) where the electrical ﬁeld is maximum. As θ increases, large magnetic pertur-

bation resulted from the magnetic ﬁlm dominates, leading to a smaller resonant frequency.
Particularly at θ = 90°, the resonant frequency shifts down to 7.958 GHz which indicates

120

Figure 7.5: Measured and simulated S parameters of the air-ﬁlled cavity resonator. Sim
ideal/resistive: without/with the resistive solder ring. The dimensions of the solder ring are
shown in the inset.

the maximum magnetic perturbation eﬀect is obtained. In addition, the maximum tuning
step is 26.3 MHz when θ turns from 0° to 15°, whereas the minimum step is 4.4 MHz as θ
∂Hφ
increases from 75° to 90°. This can be attributed to the cos θ dependence of the
∂θ which
can be derived from (7.2); as θ getting closer to 90°, the change of the Hφ becomes smaller,
thus less perturbation is obtained, and vice versa. Fig. 7.7a shows that the trend of tuning

step roughly follows the curve of cos θ. The diﬀerence is because the cos θ dependence is

given a circular shape ﬁlm fully interacting with the H ﬁeld. However, the rectangular ﬁlm

used here has less such interaction, thus smaller the tuning step change. While the measured

tuning step matches with the simulation closely, the minor diﬀerence is due to the ﬂexibility

of the ﬁlm and not being ﬂat during the rotation.

In addition, we observe the second high resonance starts to appear when θ ∈ [15°, 75°],
then disappear at θ = 90°. This resonance becomes less obvious as θ becomes larger. This

phenomenon is because of the disturbed coupling between the two excitation probes due to

the rotation of the magnetic ﬁlm which causes the excited wave travel through two diﬀerent
path resulting in two resonances; whereas at θ = 0° and 90°, probes are normally coupled so

121

that the wave travels in a single patch and thus only the fundamental resonance is formed.

Note that the small deterioration of S21 with the ﬁlm loading compared to the air-ﬁlled case

is due to the magnetic loss of the ﬁlm.

Figure 7.6: Measured S11 and S21 of 7 tuned resonances at the corresponding rotation angle
θ compared to the air-ﬁlled case.

The unloaded quality factor Qu of the cavity resonator can be calculated as [75]:

u = Q−1
Q−1

L − Q−1

e

(7.3)

where the loaded quality factor QL and the external quality factor can be determined by

[75]:

QL =

fr

∆f3dB

and Qe =

QL

10S21(dB)/20

(7.4)

where S21(dB) takes the value at the resonant frequency fr. Fig. 7.7b shows the closely

matched simulated and measured tuned resonant frequency fr, and the measured Qu is 83

122

(a)

(b)

Figure 7.7: (a)Tuning step and cos θ V.S. θ; (b) Qu and fr V.S. θ

in average less than the simulated values, which is due to the surface roughness and the

magnetic loss of the ﬁlm that are not considered in the simulation. The data of measured

Qu, fr, and the S parameters of all tuning cases are included in Table 7.1. We observe that

the air-ﬁlled cavity has a Qu of 1188 which is slightly less than the simulated value of 1211
based on Fig. 7.5. As for the ﬁlm loaded cavity, the Qu has a maximum of 732 at θ = 0°,

and as θ increase, the ﬁlm interacts more with the H ﬁeld resulting in more energy loss from
the ﬁlm, therefore a reduced Qu with a minimum of 539 at θ = 90°. This trend of Qu can

be also observed from Fig. 7.7b. Note the above Qus are not far away from those extracted
from a weakly coupled such resonator (air: 1393, θ = 0°: 764), which are not shown given

the limited space. In addition, the Qu can be further improved by polishing the 3D printed

cavity surface [73] and reducing the magnetic ﬁlm roughness loss.

Table 2 shows the state-of-the-art tunable and 3D printed resonators. We can see that

the proposed resonator has a high Q and ﬁne tuning step, which also eliminates the need for

external biasing network. However, a rotor may be needed to mechanically rotate the ﬁlm

in real practice. The tuning range can be increased by improving the magnetic nanoparticle

ﬁll factor and choosing a higher µr ferrite material.

123

Table 7.1: Measured unloaded quality factors and S parameters at the resonances of the
air-ﬁlled case and all tuning cases.

Metric
Meas. Q
fr (GHz)

S21 / S11 (dB)

Metric
Meas. Q
fr (GHz)

Air
1188
8.053

θ = 0°
732
8.028

θ = 15°

641
8.001

θ = 30°

583
7.993

-0.7 / -19.4

θ = 45°

-1.5 / -22.6

θ = 60°

-2.0 / -19.8

θ = 75°

-2.0 / -19.1

θ = 90°

670
7.975

654
7.971

555
7.962

539
7.958

S21 / S11 (dB)

-1.9 / -19.1

-1.9 / -18.9

-2.1 / -19.1

-2.1 / -17.0

Max./min./average tuning step (MHz): 26.3 / 4.4 / 11.7

Table 7.2: Comparison of the state-of-the-art tunable and 3D printed resonators.

Ref.

[70]

[72]

[73]

[66]

Type

RF MEMS

Tunability
4–6 GHz

Bias

0–50 V

0.11 GHz step

per switch

Qu

300–500

Ferrite loaded

SIW

12.4–13.2 GHz
0.18 GHz step

0-0.45 T

289–360

3D printed
sph. cavity

10.07 GHz

Not for tuning

N.A.

3728

Rect. cavity w.

tuning posts

8.21–10.12 GHz
0.53 GHz step

Not needed

72–391

This
work

Sph. cavity w.

7.958–8.028 GHz

mag. ﬁlm

11.7 MHz step

Not needed 539–732

7.4 Summary

In this chapter, a tunable spherical cavity resonator is achieved by perturbing the magnetic

ﬁeld using a 35 µm NiFe2O4 magnetic ﬁlm. The resonant frequency can be tuned from 7.958

GHz to 8.028 GHz, with the minimum and average tuning step of 4.4 MHz and 11.7 MHz.

The measured Qu ranges from 539–732. It demonstrates a light-weight and high Q cavity

resonator with ﬁne tuning based on a simple design and cost eﬀective fabrication process.

124

CHAPTER 8

SUMMARY

In this work, the application of using additive manufacturing technology and magnetic nano-

materials in the RF devices and components is explored. First, several types of RF devices

including ﬁlter, resonator, and antennas are fabricated using the polyjet or Aerosol jet print-

ing technology, and the combined process. Good performances such as low insertion loss,

wideband tunability, and high gain are achieved respectively, which shows the advantage of

fabricating low cost and high performance RF and millimeter wave devices using additive

manufacturing technology. Second, two economical and eﬃcient methods are proposed to

fabricate magnetic ﬁlms using various types of magnetic nanoparticle composites. These

ﬁlms are characterized to demonstrate good magnetic properties, such as high magnetiza-

tion saturation and high relative permeability. Then by utilizing the ﬂexibility oﬀered by the

additive manufacturing technology, we are able to design and fabricate novel reconﬁgurable

devices using these magnetic ﬁlms. Due to the high permeability, we can achieve a good

amount of size miniaturization; and with good magnetization saturation, we can tune the

ﬁlter and cavity resonator either in a wideband manner or ﬁne tuning. Also by carefully

choosing the antenna type and the ﬁlm position, we can obtain enhanced antenna gain and

eﬃciency along with the size reduction, which is not realizable using the transnational meth-

ods. This work paves the way for using additive manufacturing technology and low cost

magnetic materials to design and fabricate novel reconﬁgurable RF devices in a low cost and

light-weight package.

125

BIBLIOGRAPHY

126

BIBLIOGRAPHY

https://formlabs.com

https://www.stratasys.com

https://optomec.com/

J. D. Adam, L. E. Davis, G. F. Dionne, E. F. Schloemann and S. N. Stitzer, ”Ferrite
devices and materials,” in IEEE Transactions on Microwave Theory and Techniques,
vol. 50, no. 3, pp. 721–737, March 2002.

[1]

[2]

[3]

[4]

[5] David M. Pozar, Microwave Engineering, 4th ed. New Jersey: John Wiley & Sons,

Inc., 2012. ch.8, pp.422–426.

[6]

Liu, L., et al., “Compact harmonic ﬁlter design and fabrication using IPD technol-
ogy,”IEEE Transactions on Components and Packaging Technologies, Vol. 30, No. 4,
556–562, 2007.

[7] Kim, D. W., et al., “High performance RF passive integration on a Si smart substrate

for wireless applications,” Etri Journal, Vol. 25, No. 2, 65–72, 2003.

[8]

Fan Cai, Wasif Tanveer Khan, John Papapolymerou. “A Low Loss X-band Filter Using
3-D Poly jet Technology,” in IEEE MTT-S Int. Microwave Symp. ,Phoenix, AZ, May
2015 pp. 1–4.

[9] Constantine A. Balanis, Antenna Theory Analysis and Design, 4th Edition. New Jer-

sey: John Wiley & Sons, Inc., 2016. ch.10, pp.559–580.

[10] F. Cai et al., ”Low-Loss 3-D Multilayer Transmission Lines and Interconnects Fabri-
cated by Additive Manufacturing Technologies,” in IEEE Transactions on Microwave
Theory and Techniques, vol. 64, no. 10, pp. 3208–3216, Oct. 2016.

[11] C. X. Wang et al.,“Cellular architecture and key technologies for 5G wireless commu-
nication networks,” in IEEE Communications Magazine, vol. 52, no. 2, pp. 122–130,
February 2014.

[12] J. P. Magalhes, J. M. N. Vieira, R. Gmez-Garca and N. B. Carvalho, “Bio-Inspired
Hybrid Filter Bank for Software-Deﬁned Radio Receivers,” in IEEE Transactions on
Microwave Theory and Techniques, vol. 61, no. 4, pp. 1455–1466, April 2013.

127

[13] T. K. Thrivikraman, A. Madan, and J. D. Cressler, “On the large-signal robustness
of SiGe HBT LNAs for high-frequency wireless applications,” in Proc. IEEE Radio
Wireless Symp. (RWS), Jan. 2010, pp. 364–367.

[14] M. van Wanum and F. E. van Vliet, “A 58-dBm S-band limiter in standard 0.25-µm
BiCMOS technology,” IEEE Trans. Microw. Theory Techn., vol. 61, no. 8, pp.
3034–3042, Aug. 2013.

[15] J. Givernaud et al., “Microwave power limiting devices based on the semiconduc-
tormetal transition in vanadiumdioxide thin ﬁlms,” IEEE Trans. Microw. Theory
Techn., vol. 58, no. 9, pp. 2352–2361, Sep. 2010.

[16] T. R. Billeter, “EPR frequency-selective limiter,” Proc. IEEE, vol. 56, no. 3, pp.

370–371, Mar. 1968.

[17] F. Ramirez, R. Melville, A. Suarez, and J. S. Kenney, “Nonlinear analysis and design
of frequency selective limiters based on parametric circuits,” in IEEE MTT-S Int.
Microw. Symp. Dig., Jun. 2008, pp. 947–950.

[18] J. C. S. Chieh and J. Rowland, “X-band frequency selective limiter using absorp-
tive notch ﬁlters embedded multiplexers,” 2017 IEEE MTT-S International Microwave
Symposium (IMS), Honololu, HI, USA, 2017, pp. 1919–1922.

[19] E. J. Naglich and A. C. Guyette, “Frequency-Selective Limiters Utilizing Contiguous-
Channel Double Multiplexer Topology,” in IEEE Transactions on Microwave Theory
and Techniques, vol. 64, no. 9, pp. 2871–2882, Sept. 2016.

[20] J. D. Adam and F. Winter, “Magnetostatic Wave Frequency Selective Limiters,” in

IEEE Transactions on Magnetics, vol. 49, no. 3, pp. 956–962, March 2013.

[21] J. D. Adam. “Mitigate the Interference: Nonlinear Frequency Selective Ferrite De-

vices,” IEEE Microwave Magazine, Vol. 15, No. 6, pp. 45–56, Sep–Oct. 2014.

[22] F. K. H. Gellersen, F. M. Klichowski, A. G. C. Rusche and A. F. Jacob, “Barium
hexaferrite ﬁlms prepared by means of the sol-gel method for self-biased nonreciprocal
microwave devices,” 2016 German Microwave Conference (GeMiC), Bochum, 2016,
pp. 221–224.

[23] F. K. H. Gellersen, J. Peschel, A. Ochsenfarth and A. F. Jacob, “Scandium-doped
barium hexaferrite thin-ﬁlms for nonreciprocal satellite components,” 2017 Topical
Workshop on Internet of Space (TWIOS), Phoenix, AZ, 2017, pp. 1–4.

[24] A. L. Geiler and V. Harris. “Atom Magnetism,” IEEE Microwave Magazine, Vol 15,

No. 6, pp. 66–72, Sep–Oct. 2014

128

[25] M. A. Morton, C.R. Vestal, J. Papapolymerou, Z.J. Zhang. “Fabrication and Char-
acterization of Spinel Magnetic Nanoparticle Thin Film Transmission Lines,” 33rd
European Microwave Conference, 2003,Volume: 3, pp. 1307–1309, 7–9 Oct. 2003.

[26] Y. He, S. Pavlidis, W. Y. Chen, E. Drew, Z. J. Zhang and J. Papapolymerou, “Fab-
rication and characterization of CPW transmission lines with CoFe2O4 nanomagnetic
thin ﬁlms,” 2017 IEEE Radio and Wireless Symposium (RWS), Phoenix, AZ, 2017,
pp. 72–75.

[27] C. Kittel, “Interpretation of the anomalous Larmor frequencies in ferromagnetic reso-

nance experiment,” Phys. Rev., vol. 71, no. 4, pp. 270–271, 1947

[28] Chen, W.Y. 2018 “Aminolytic synthesis and ferromagnetic resonance of cobalt and
manganese based spinel ferrite nanoparticle”, Doctorate of Philosophy in Chemistry,
Georgia Institute of Technology, Atlanta, GA.

[29] L. Hernandez, Y. He, A. Kaur, J. Papapolymerou and P. Chahal, “Low-loss RF ﬁlter
through a combination of additive manufacturing and thin-ﬁlm process,” 2017 IEEE
Radio and Wireless Symposium (RWS), Phoenix, AZ, 2017, pp. 114–116.

[30] http://www.stratasys.com/polyjet-technology

[31] Gokhan Mumcu, Thomas Weller, “Small Antennas and Miniaturization Techniques,”
Antenna Engineering Handbook, J. L. Volakis, 5th ed. New York: McGraw Hill, 2018,
ch.10.

[32] Z. Shao and Y. P. Zhang, “Diﬀerential Shorted Patch Antennas,” in IEEE Trans.

Antennas Propag., vol. 67, no. 7, pp. 4438–4444, July 2019.

[33] J. P. Gianvittorio and Y. Rahmat-Samii, “Fractal antennas: a novel antenna minia-
turization technique, and applications,” in IEEE Antennas Propag. Mag., vol. 44, no.
1, pp. 20–36, Feb. 2002.

[34] D. H. Schaubert, D. M. Pozar and A. Adrian, “Eﬀect of microstrip antenna substrate
thickness and permittivity: comparison of theories with experiment,” in IEEE Trans.
Antennas Propag., vol. 37, no. 6, pp. 677–682, June 1989.

[35] D. Sievenpiper, Lijun Zhang, R. F. J. Broas, N. G. Alexopolous and E. Yablonovitch,
“High-impedance electromagnetic surfaces with a forbidden frequency band,” in IEEE
Trans. Microw. Theory Techn., vol. 47, no. 11, pp. 2059–2074, Nov. 1999.

[36] K. Buell, H. Mosallaei and K. Sarabandi, “A substrate for small patch antennas pro-
viding tunable miniaturization factors,” in IEEE Trans. Microw. Theory Techn., vol.
54, no. 1, pp. 135–146, Jan. 2006.

129

[37] A. D. Brown, J. L. Volakis, L. C. Kempel and Y. Botros,“Patch antennas on ferromag-
netic substrates,” in IEEE Trans. Antennas Propag., vol. 47, no. 1, pp. 26–32, Jan.
1999.

[38] R. V. Petrov, A. S. Tatarenko, S. Pandey, G. Srinivasan, J. V. Mantese, and R. Azade-
gan, “Miniature antenna based on magnetoelectric composites,” in Electron. Lett., vol.
44, no. 8, pp. 506–507, 10 April, 2008.

[39] J. Lee, Y. Hong, S. Bae, G. S. Abo, W. Seong and G. Kim, “Miniature Long-Term
Evolution (LTE) MIMO Ferrite Antenna,” in IEEE Antennas Wireless Propag. Lett.,
vol. 10, pp. 603–606, 2011.

[40] J. Lee, J. Lee, K. Min and Y. Cheon, “Miniaturized Antennas With Reduced Hand
Eﬀects in Mobile Phones Using Magneto-Dielectric Material,” IEEE Antennas Wireless
Propag. Lett., vol. 13, pp. 935–938, 2014.

[41] P. J. Rainville and F. J. Harackiewicz, “Magnetic tuning of a microstrip patch antenna
fabricated on a ferrite ﬁlm,” in Microwave and Guided Wave Letters, vol. 2, no. 12,
pp. 483–485, Dec. 1992.

[42] H. How, P. Rainville, F. Harackiewicz and C. Vittoria, “Radiation frequencies of ferrite

patch antennas,” in Electron. Lett., vol. 28, no. 15, pp. 1405–1406, 16 July 1992.

[43] G. M. Yang et al., “Tunable miniaturized patch antennas with self-biased multilayer
magnetic ﬁlms,” in IEEE Trans. Antennas Propag., vol. 57, pp. 2190–2193, July 2009.

[44] G. Yang et al., “Planar Annular Ring Antennas With Multilayer Self-Biased NiCo-
Ferrite Films Loading,” in IEEE Trans. Antennas Propag., vol. 58, no. 3, pp. 648–655,
March 2010.

[45] G. M. Yang et al., “Loading eﬀects of self-biased magnetic ﬁlms on patch antennas
with substrate/superstrate sandwich structure,” in IET Microw. Antennas Propag.,
vol. 4, no. 9, pp. 1172–1181, September 2010.

[46] Y. He, W. Chen, E. Drew, Z. J. Zhang and J. Papapolymerou, “Fabrication and
Characterization of CoFe2O4 and MnFe2O4 Nanomagnetic Thin Films for RF Appli-
cations,” IEEE Trans. Compon. Packag. Manuf. Technol., vol. 9, no. 5, pp. 973–983,
May 2019.

[47] Y. He, E. Drew, Z. J. Zhang, T. Hogan and J. Papapolymerou, ”Compact BandStop
Filter Utilizing Low Cost Solution Cast Nanomagnetic Thin Films,” 2019 IEEE MTT-
S International Microwave Symposium (IMS), Boston, MA, 2019, pp. 1118–1121.

130

[48] Y. He and J. Papapolymerou, “Compact Microstrip Patch Antenna Utilizing Low Cost
Solution Cast Nanomagnetic Thin Film,” 2019 IEEE International Symposium on
Antennas and Propagation & USNC/URSI National Radio Science Meeting, Atlanta,
GA, 2019, pp. 1531–1532.

[49]

[Online] http://www.rogerscorp.com/documents/730/index.aspx

[50] Constantine A. Balanis, Antenna Theory: Analysis and Design, 4th ed. New Jersey:

John Wiley & Sons, Inc., 2016. ch.14, pp.788–815

[51] D. Jackson and N. Alexopoulos, “Gain enhancement methods for printed circuit an-
tennas,” in IEEE Trans. Antennas Propag., vol. 33, no. 9, pp. 976–987, September
1985.

[52] J. M. Edwards and G. M. Rebeiz, “High-Eﬃciency Elliptical Slot Antennas With
Quartz Superstrates for Silicon RFICs,” in IEEE Trans. Antennas Propag., vol. 60,
no. 11, pp. 5010–5020, Nov. 2012.

[53] P. M. Ikonen, K. N. Rozanov, A. V. Osipov, P. Alitalo, and S. A.Tretyakov, “Magne-
todielectric substrate in antenna miniaturization: Potential and limitations,” in IEEE
Trans. Antennas Propag., vol. 54, pp. 3391–3399, Nov. 2006.

[54] N. Popandian, P. Balay, A. Narayanasamy, “Electrical conductivity and dielectric be-
haviour of nanocrystalline NiFe2O4 spinel,” in J. Phys.: Condens. Matter 14, 3221,
2002.

[55] David R. Jackson, “Microstrip antennas,” Antenna Engineering Handbook, J. L. Volakis,

5th ed. New York: McGraw Hill, 2018, ch.7.

[56] R. C. Hansen and M. Burke, “Antennas with magneto-dielectrics,” Microw. Opt. Tech.

Lett., vol. 26, no. 2, pp. 75–78, Jul. 2000.

[57] M. I. M. Ghazali, S. Karuppuswami, A. Kaur and P. Chahal, “3-D Printed Air Sub-
strates for the Design and Fabrication of RF Components,” in IEEE Trans. Compon.
Packag. Manuf. Technol., vol. 7, no. 6, pp. 982–989, June 2017.

[58] L. Young, G. L. Matthaei and E. M. T. Jones, “Microwave Bandstop Filters with
Narrow Stop Bands,” PGMTT National Symposium Digest, Washington, DC, USA,
1962, pp. 46-51.

[59] G. M. Rebeiz, K. Entesari, I. C.Reines, S. J. Park, M. A. El-Tanani, A. Grichener, and
A. R.Brown, “Tuning in to RF MEMS,” IEEE Microw. Mag., Focused Issue Feature,
pp. 5072, Oct. 2009.

131

[60] J. Papapolymerou, C. Lugo, Z. Zhao, X. Wang, and A. Hunt, “A miniature low-loss
slow-wave tunable ferroelectric bandpass ﬁlter from 11-14 GHz,” in IEEE MTT-S Int.
Microw. Symp. Dig., San Francisco, CA, USA, Jun. 2006, pp. 556-559.

[61] T. Y. Yun and K. Chang, “Piezoelectric-transducer-controlled tunable microwave cir-

cuits,” IEEE Trans. Microw. Theory Tech., vol. 50, pp. 1303-1310, May 2002.

[62] A. S. Tatarenko, V. Gheevarughese, and G. Srinivasan, “Magnetoelectric microwave

bandpass ﬁlter,” Electron. Lett., vol. 42, pp. 540-541, Apr. 2006.

[63] G. Yang, J. Wu, J. Lou, M. Liu and N. X. Sun, “Low-Loss Magnetically Tunable
Bandpass Filters With YIG Films,” in IEEE Transactions on Magnetics, vol. 49, no.
9, pp. 5063–5068, Sept. 2013.

[64] S. Adhikari, S. Hemour, A. Ghiotto and K. Wu, “Magnetically Tunable Ferrite-Loaded
Half-Mode Substrate Integrated Waveguide,” in IEEE Microwave and Wireless Com-
ponents Letters, vol. 25, no. 3, pp. 172–174, March 2015.

[65] Y. He et al., “Integrated Tunable Bandstop Filter Using Self-Biased FeGaB/Al2O3
Multilayer Thin Film,” in IEEE Transactions on Magnetics, vol. 54, no. 9, pp. 1–4,
Sept. 2018.

[66] Y. He, P. Chahal and J. Papapolymerou, “A lego-like reconﬁgurable cavity using 3-D
Polyjet technology,” 2017 IEEE MTT-S International Microwave Symposium (IMS),
Honololu, HI, 2017, pp. 525–528.

[67] C. Kittel, “Interpretation of the anomalous Larmor frequencies in ferromagnetic reso-

nance experiment,” Phys. Rev., vol. 71, no. 4, pp. 270-271, 1947.

[68] Jia-Sheng Hong, M. J. Lancaster, Microstrip Filters for RF/Microwave Applications,

John Wiley & Sons, 2001.

[69] http://www.stratasys.com/polyjet-technology

[70] S. Park, I. Reines, C. Patel and G. M. Rebeiz, “High-Q RF-MEMS 4–6 GHz Tunable
Evanescent-Mode Cavity Filter,” in IEEE Trans. Microw. Theory Techn., vol. 58, no.
2, pp. 381–389, Feb. 2010.

[71] S. Courreges, Y. Li, Z. Zhao, K. Choi, A. Hunt and J. Papapolymerou, “Ferroelectric
Tunable Bandpass Filters for Ka-Band Applications,” 2008 38th European Microwave
Conference, Amsterdam, 2008, pp. 55–58.

[72] S. Adhikari, Y. Ban and K. Wu, “Magnetically Tunable Ferrite Loaded Substrate
Integrated Waveguide Cavity Resonator,” in IEEE Microw. Wireless Compon. Lett.,
vol. 21, no. 3, pp. 139–141, March 2011.

132

[73] C. Guo et al., “Shaping and Slotting High-Q Spherical Resonators for Suppression of
Higher Order Modes,” 2019 IEEE MTT-S International Microwave Symposium (IMS),
Boston, MA, USA, 2019, pp. 1205–1208.

[74] Constantine A. Balanis, Advanced engineering electromagnetics, 4th ed. New Jersey:

John Wiley & Sons, Inc., 2016. ch.10, pp. 561–568.

[75] David M. Pozar, Microwave Engineering, 4th ed. New Jersey: John Wiley & Sons,

Inc., 2012. ch.10, pp. 305–308.

133