LOW POWER MAGNETIC FIELD SENSORS UTILIZING JANUS MAGNETOELECTRIC NANOWIRES: FABRICATION AND CHARACTERIZATION

By

MATTHEW JEFFREY BAUER

A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2019

© 2019 Matthew Jeffrey Bauer

To my family

ACKNOWLEDGMENTS

This work represents the culmination of years of work and the contributions of many people. I wish to thank my parents, Jeffrey and Ellen Bauer for their support. I would like to thank the several graduate and undergraduate students who have came together to assist in the intensive, complex, and multidisciplinary task of fabricating novel nanowire based devices. In particular I wish to thank Dr. Xiao Wen, Prabal Tiwari,

John Varela, Andrew Thomas, Bridget Isenberg, and Andrea Faria. Additionally, I would like to thank the staff at NRF for the training on lab equipment utlilized in the fabrication process. I would like to thank Dr. Andrew and Dr. Arnold and the MIST Center for support and advice. Lastly, I would like to thank my advisory committee and the faculty and staff of the University of Florida MSE Dept.

4

TABLE OF CONTENTS

page

ACKNOWLEDGMENTS ...... 4

LIST OF TABLES ...... 7

LIST OF FIGURES ...... 8

ABSTRACT ...... 12

CHAPTER

1 MOTIVATION AND BACKGROUND ...... 14

1.1 Motivation ...... 14 1.2 Piezoelectrics ...... 15 1.3 Magnetoelectrics ...... 15 1.4 Nanowire Fabrication ...... 16 1.4.1 Electrospinning ...... 17 1.4.2 Sol-Gel Electrospinning ...... 20 1.5 Assembly ...... 21 1.6 Magnetic Field Sensors ...... 22 1.7 Dissertation Overview ...... 25

2 ALIGNED BARIUM TITANATE ...... 30

2.1 Experimental Process ...... 30 2.1.1 Materials ...... 32 2.1.2 Barium Titanate Sol–Gel Preparation ...... 32 2.1.3 Electrospinning Aligned Barium Titanate Fibers ...... 32 2.1.4 Characterization ...... 33 2.2 Results and Discussion ...... 34 2.3 Conclusion ...... 40

3 BARIUM TITANATE / COBALT FERRITE MAGNETIC FIELD SENSING ARRAYS ...... 50

3.1 Introduction ...... 50 3.2 Experimental ...... 55 3.2.1 Nanowire Fabrication ...... 55 3.2.2 AC Electrical Assembly ...... 57 3.2.3 Characterization ...... 58 3.3 Results ...... 61 3.3.1 Nanowire Fabrication ...... 61 3.3.2 AC Electrical Assembly ...... 63 3.3.3 Characterization ...... 65

5

3.4. Conclusions ...... 68

4 ULTRA LOW POWER CURRENT SENSOR UTILIZING MAGNETOELECTRIC NANOWIRES ...... 85

4.1 Introduction ...... 85 4.2 Design ...... 87 4.3 Fabrication ...... 89 4.3.1 Nanowire Synthesis ...... 89 4.2.3 Nanowire Array and Current Sensor Fabrication ...... 92 4.4 Current Sensor Characterization ...... 94 4.5 Conclusion ...... 95

5 CONCLUSIONS ...... 111

APPENDIX: DIELECTROPHORETIC FORCE SIMULATIONS ...... 113

LIST OF REFERENCES ...... 119

BIOGRAPHICAL SKETCH ...... 129

6

LIST OF TABLES

Table page

2-1 Agreement indices for the rietveld refinements on the barium titanate fibers ..... 41

7

LIST OF FIGURES

Figure page

1-1 Electrospinning droplet formation for biphasic magnetoelectric Janus fiber fabrication...... 27

1-2 Electrospinning diagram for biphasic magnetoelectric Janus fiber fabrication. ... 28

1-3 Diagrams illustrating the gradient of the electric field gradient showing the areas where the nanowires will be attracted to from a positive dielectrophoretic force ...... 29

2-1 Photograph of the rotating mandrel set‐up for the electrospinning of aligned nanofibers showing the as‐spun amorphous fibers aligned across the parallel copper wires on the rotating mandrel. (Photograph courtesy of the author) ...... 42

2-2 Scanning electron micrographs of uncalcined and calcined aligned barium titanate nanofibers...... 43

2-3 Fiber diameter distributions for uncalcined and calcined barium titanate fibers...... 44

2-4 A) X‐ray diffraction (XRD) spectra for barium titanate nanofibers calcined for 6 h at 750°C, 875°C, and 1000°C. B) XRD spectra for barium titanate nanofibers calcined at 875°C for 2, 4, and 6 h ...... 45

2-5 The weight percent of the A) tetragonal, B) cubic, and C) hexagonal phase of barium titanate as a function of calcining temperature, for a constant calcining time of 6 h...... 45

2-6 Raman spectra for barium titanate nanofibers calcined at various times and temperatures from 2 h to 6 h and 750°C to 1000°C...... 46

2-7 The natural logarithm of average grain size plotted versus reciprocal temperature for the tetragonal phase...... 47

2-8 A) Butterfly loop and B) phase plot of barium titanate nanofiber calcined at 750°C for 2 h, obtained with the assistance of Catherine Snyder...... 48

2-9 Phase angle scan of a barium titanate nanofiber calcined at 750°C for 2 h, revealing the domain structure of the fiber, the width of the scan shown is 1 μm...... 49

3-1 The magnetoelectric sensing element consisting of magnetoelectric nanowires bridging an electrode gap suspended above a substrate ...... 70

8

3-2 Schematic diagram of the lock-in magnetoelectric measurement setup and nanowire orientation within the setup...... 71

3-3 Schematic of the rotating magnetoelectric measurement setup...... 72

3-4 Scanning electron micrograph of Janus barium titanate - cobalt ferrite (BTO - CFO) nanowires post calcination...... 73

3-5 Graphs showing the effects of electrospinning field and calcination ramp rate on fiber diameter...... 74

3-6 X-ray diffraction spectra of the barium titanate / cobalt ferrite Janus nanowires post calcination showing peaks indicative of tetragonal barium titanate and spinel cobalt ferrite...... 75

3-7 Raman spectra of single phase barium titanate nanowires used to test whether barium carbonate (BCO) can be successfully removed with a dilute hydrochloric acid (HCl) wash...... 76

3-8 Mass magnetization curve indicating ferrimagnetism in the cobalt ferrite phase of unassembled Janus nanowires from the same batch as those used to fabricate the nanowire array...... 77

3-9 Successful assembly of barium titanate nanofibers in water, post barium carbonate removal with a dilute HCl wash and suspension using citric acid and adjusting the pH to around 9, at 5 kHz and 20 Vpp...... 78

3-10 Assembly of Janus nanofibers in various solvents at 5 kHz and 20 Vpp...... 79

3-11 Assembly of Janus nanowires in butanol at 5 kHz and 42 Vpp in the test array with a linear density of 19 NWs mm−1...... 80

3-12 Formation of upper electrical contacts across the nanowires ...... 81

3-13 Capacitance–voltage (C–V) measurement from the Janus nanowire array demonstrating ferroelectricity in the barium titanate phase. The test signal level was set to 100 mV at 750 kHz...... 82

3-14 Plot of the magnetoelectric coefficients of a row in the barium titanate / cobalt ferrite Janus nanowire array measured using the lock-in technique as a function of magnetic bias field at 200, 500, and 1000 Hz...... 83

3-15 Rotational magnetoelectric measurement results from a row of barium titanate/cobalt ferrite composite nanowires...... 84

4-1 Electrode design for the ultra low power current sensor...... 96

4-2 Flip chip design of the ultra low power current sensor...... 97

9

4-3 Circuit diagram of the magnetic field sensor...... 98

4-4 Process flow diagram for fabricating the current sensor...... 99

4-5 Scanning electron micrographs of lead zirconate titanate / nickel zinc ferrite nanowires before calcination (A,B), uncalcined, and post calcination (C,D), calcined...... 100

4-6 X-Ray Diffraction Spectra of the lead zirconate titanate (PZT) and nickel zinc ferrite (NZF) nanowires, showing the characteristic peaks of the tetragonal and inverse spinel structures, respectively...... 101

4-7 Raman spectra of lead zirconate titanate/nickel zinc ferrite demonstrating the removal of a lead oxide impurity phase with a dilute HCl wash...... 102

4-8 Mass magnetization curve indicating ferrimagnetism in the nickel zinc ferrite phase of unassembled Janus nanowires showing remnant magnetization in the nanowires...... 103

4-9 Effects of frequency on electrical assembly of lead zirconate titanate/nickel zinc ferrite nanowires ...... 104

4-10 Nanowires assembled across interdigitated electrodes post electroplating...... 105

4-11 Measured magnetoelectric coefficients as a function of frequency from 6 individual rows lead zirconate titanate / nickel zinc ferrite nanowire arrays at angles of approximately no inductive effects...... 106

4-12 Flip chip fabrication of of the current sensor IC...... 107

4-13 Calculated applied current through a current sense resistor and raw output voltage waveforms from the nanowire based current sensor at with applied currents at 200 Hz and 1 kHz...... 108

4-14 Fitted waveform amplitudes from the mains hum and applied signal...... 109

4-15 Noise density as a function of frequency of the current sensor as measured in a triple Faraday cage...... 110

A-1 Real part of the Claussius Mossotti Facter as a function of frequency in barium titanate predictive of positive dielectrophoresis in each of the solvents at high frequencies (2-20 kHz)...... 115

A-2 Real part of the Claussius Mossotti Facter as a function of frequency in cobalt ferrite predictive of weak positive dielectrophoresis with the low permittivity solvent butanol at high frequencies (10-20 kHz)...... 116

10

A-3 Real part of the Claussius Mossotti Facter as a function of frequency in lead zirconate titanate predictive of positive dielectrophoresis in each of the solvents at high frequencies (2-20 kHz)...... 117

A-4 Real part of the Claussius Mossotti Facter as a function of frequency in nickel zinc ferrite predictive of weak positive dielectrophoresis in butanol and 2- methoxyethanol at high frequencies (10-20 kHz)...... 118

11

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

LOW POWER MAGNETIC FIELD SENSORS UTILIZING JANUS MAGNETOELECTRIC NANOWIRES: FABRICATION AND CHARACTERIZATION

By

Matthew Jeffrey Bauer

May 2019

Chair: Jennifer Andrew Major: Material Science and Engineering

In this dissertation ultra-low power magnetic field sensing and current sensing discrete circuits utilizing magnetoelectric nanowires were fabricated. These sensors have applications in automobiles as rotational speed sensors and in electrical systems as current sensors. While active sensing elements such as Hall sensors function well in both applications their power and current requirements can be costly. The current draw in rotational speed sensors requires sufficiently large guage wire to accommodate the current draw increasing material cost. In complex systems where numerous current sensors may be necessary, decreasing the power loss incurred through current measurement becomes important. To decrease these costs, low power magnetic field sensors with the potential for miniaturization are desired. Magnetoelectrics, materials which exhibit an electrical polarization in response to an applied magnetic field could offer a solution as the voltage generated by the applied magnetic field could be used to measure the magnetic field without a power source needed for the sensing element.

Composite strain mediated magnetoelectrics were utilized as they generally generate larger voltages per unit length and applied field than their single-phase counterparts.

These composite magnetoelectrics are comprised of magnetostrictive and piezoelectric

12

materials which share an interface. A shape change induced on the magnetostrictive phase by an applied magnetic field results in a shape change and corresponding electrical polarization in the piezoelectric material which can then be used to measure

the magnitude of applied magnetic field. Composite strain mediated magnetoelectric nanowires are used here as the strain transfer between the phases is not clamped by

an underlying substrate allowing greater magnetoelectric coefficients/sensitivities to be

realized. Another advantage is that the optimal calcining temperature for

magnetoelectric performance can be used while maintaining CMOS compatibility as this

high temperature processing step occurs off substrate prior to assembly of the

nanowires into the final device.

13

CHAPTER 1 MOTIVATION AND BACKGROUND

1.1 Motivation

Piezoelectric, magnetostrictive, and magnetoelectric micro/nanomaterials have unique potential for applications in micro-electro-mechanical systems (MEMS), textiles, and other areas. Briefly, piezoelectric materials are useful for their unique ability to convert mechanical strain to electrical polarization and vice versa. Magnetostrictive materials couple magnetic and mechanical ordering, where a magnetization in these materials induces a shape change. When magnetostrictive and piezoelectric materials are coupled through a shared interface the strain can transfer between the two materials such that a magnetization in the magnetostrictive phase creates a strain in the material that when transfered to the piezoelectric induces an electrical polarization.

Thus, forming a magnetoelectric composite in which amagnetization applied magnetic field results in an electrical polarization and vice versa.

Applications for magnetoelectric materials involve converting magnetic to electrical energy for passive magnetic field and, by extension, electrical current sensing.

Clearly a major advantage of nanomaterials, in contrast to macroscale elements, is the ability for miniaturization. While thin films can also be miniaturized there are several advantages of nanomaterials over thin films, which make it worthwhile to establish methods to create nanomaterial based magnetoelectric devices.

The obvious advantage to the use of piezoelectrics and composite strain based magnetoelectrics in the form of nanomaterials over thin films is that it removes substrate clamping, where the shared interface between the functional material and the substrate reduces the strain necessary for magnetostriction or piezoelectricity. Another benefit to

14

using magnetoelectric nanomaterials is that the high temperature calcination of these materials can be performed off substrate prior to their incorporation into the final device.

This allows for the inclusion of underlying supporting circuitry, which might not be able to sustain such a high temperature processing step on chip. The potential for chip level integration of buffering and amplification circuitry further reduces the potential size of nanowire based sensors.

Here the focus will predominantly be on the fabrication of passive magnetic field

and electrical current sensors utilizing composite biphasic magnetoelectrics. This

dissertation will cover the fabrication of magnetoelectric nanowires, their assembly into

arrays, characterization of the magnetoelectric nanowires, incorporation of the arrays

into magnetic field and current sensing ICs, and the final device characterization.

1.2 Piezoelectrics

The piezoelectric and ferroelectric properties and high dielectric constant of

barium titanate have enabled its incorporation into a wide array of electrical applications

including actuators,1 medical imaging devices,2 ultrasound transducers,3 and high

dielectric capacitors.4 Nanoscale barium titanate has been considered a promising

material for high‐density ferroelectric random access memories.5–8 Additionally,

electrospun barium titanate nanofibers have been used to construct high‐performance

humidity sensors in place of barium titanate thin films due to their high surface area and

large aspect ratio.7,9

1.3 Magnetoelectrics

Magnetoelectrics are unique functional materials in which an applied magnetic

field can be used to control an electrical polarization.10,11 The voltage generated by a

magnetoelectric material can then be read to passively measure the magnetic field. As

15

such they can offer a lower power alternative to current magnetic field sensors or current sensors such as Hall Sensors. However, to be viable for use in magnetic field sensors, material systems and architectures need to be found which offer high magnetoelectric coefficients (dE/dH). Although single phase magnetoelectrics exist, they are comparatively rare, and have limited performance metrics.12 On the other hand,

composite magnetoelectrics are capable of producing greater magnetoelectric effects.

For magnetoelectrics the figure of merit is the magnetoelectric coefficient (α, which is

quantified as the magnitude of the electric field (dE) generated in a material in response

to an applied magnetic field (dH), α = dE/dH.10,11,13 Composite magnetoelectrics are

typically composed of magnetostrictive and piezoelectric phases which share an interface. When exposed to an applied magnetic field the magnetostrictive phase undergoes a shape change, which imparts a strain to the piezoelectric phase, thereby inducing an electrical polarization. When magnetoelectric composites are fabricated as thin films, the strain transfer between the magnetostrictive and piezoelectric phases is typically limited by the underlying substrate leading to a reduction in the magnetoelectric effect. 10,11,13,14 Less rigidly clamped 1-D magnetoelectric nanostructures could offer

increased magnetoelectric coefficients. Enhancements of up to a few orders of

magnitude seem feasible based on theoretical and scanning probe microscopy

measurements.15

1.4 Nanowire Fabrication

Nanowires and nanofibers may be fabricated by a variety of methods including

vapor-liquid-solid (VLS) deposition, top down lithography, and sol-gel electrospinning. In

VLS a vapor phase forms a liquid alloy with metal catalyst clusters and nucleates to

form the initial crystal. Further condensation and dissolution of the vapor phase leads to

16

the formation of nanowires as the liquid alloy is pushed upward.16,17 Nanowires can be

fabricated using top down lithography methods by forming a template for nanowire

growth, electrochemical deposition, and wet or otherwise chemically selective etching.18

Finally sol-gel electrospinning is a method by which a ceramic/polymer solution is

drawn, often from a syringe needle, into a nanofiber using a large electric field that is

applied between the solution and a counter electrode.19,20 It is unclear as of yet how

VLS or SLS could produce biphasic Janus morphology fibers and for this reason these

methods were ruled out. Top down lithography has the potential to produce such fibers,

however sol-gel electrospinning was chosen for this project as it is a more economical,

and thus more commercializable, method of nanofiber/nanowire fabrication.

1.4.1 Electrospinning

Electrospinning is an economical, scaleable, approach which uses an electric

field to draw nanofibers from polymer or sol-gel solutions. Generally planar liquid

surfaces are not amenable to electrospinning but an initial curvature of the liquid can be

accomplished in multiple ways21. These include extrusion through single or multiple

electrospinning needles or needleless electrospinning techniques. When using a needle

based electrospinning setup the droplet extruded from the needle creates the curvature

needed to concentrate charges needed to draw the fibers towards the grounded counter

electrode. Needleless electrospinning techniques include the use of various

mechanisms to create the initial curvature including baths containing polymer solution

atop a magnetic liquid used to create peaks and troughs, pumping a driver gas through

the electrospinning fluid, spikes on metal cylinders carrying the fluid, etc22. While

needleless electrospinning can produce nanofibers more efficiently22 these methods are

more challenging for prototyping than needle based electrospinning and synthesis of

17

biphasic fibers with this method would be difficult. Since the use of needles is relatively

scaleable and also allows for the extrusion of single phase solutions and two solutions

in a biphasic Janus morphology, a needle electrospinning setup was opted for (Figure

1-1).

In this setup a polymer or sol-gel solution is extruded through a positively

charged syringe needle directed toward a grounded counterelectride. As the solution is

extruded through the syringe needle it is held by surface tension as charge is induced

upon on the surface of the fluid by the electric field. Repulsion between between like

charges on the surface creates a force in opposition to surface tension. This charge

repulsion causes the solution to elongate into a conical shape referred to as the Taylor

cone. When the Coulombic repulsion becomes sufficiently strong to overcome the

surface tension a charged jet is emitted from the taylor cone and pulled toward the

grounded counter-electrode (Figure 1-2)23. Generally depictions of electrospinning jets

generally show a spiral shaped jet which is due to what is known as a bending

instability24. This bending instability is due to mutually repulsive electrostatic forces acting on a slightly perturbed segment of the jet which pushes the path of the electrospinning jet off it’s original axis of travel (Figure 1-2)23.

Here the major possible detrimental phenomena are branching

instabilities/formation of multiple jets, bead-forming instabilities, and melding in the

electrospun fibers. The formation of multiple jets and branching instabilities, where

secondary branches form off the main electrospinning jet, are especially problematic in

the electrospinning of Janus fibers as these additional jets can lead to single phase

fibers. The formation of multiple jets and branching is generally observed at higher

18

electric fields than are necessary for producing a single jet. As the electric potential is

increased on the jet, through increasing the voltage applied to the electrospinning

needle, undulations occur in the cross-sectional profile of the jet due to charge

repulsion, jets can then be emitted from these undulations similarly to how the initial jet

was formed from the Taylor cone. Therefore, branching can be reduced by reducing the

electric field used in electrospinning to that required to spin a singular jet24.

Beading instability results in the formation of physical beads along the

electrospinning jet. Rayleigh noted that in jets of fluid in air, capillary (surface) forces

render a cylindrical jet an unstable equilibrium and favors the disintegration of the jet

into droplets whose total surface area is less than that of the initial cylindrical jet25.

Electrical charges on the jet can help reduce the tendency for the solution to form

droplets26. The addition of polymer to the jet can also help to increase maintain the jet,

however bead on string like morphologies can occur. Increasing the viscosity of the

solution, through increasing the polymer concentration, and/or decreasing the surface

tension, through choice of solvent or addition of surfactant25, of the solution while

maintaining a surface charge on the jet can reduce or eliminate beading27.

Another detrimental phenomena which has to be considered in the tuning of

electrospinning parameters to allow for optimal fiber morphology is webbing. Webbing is

where multiple as spun fibers meld together to form webbed like structures due to

insufficient solvent evaporation before making contact with the grounded colletor plate

due to insufficient solvent evaporation. Factors which can decrease the presence of

webbing are electrospinning with high volatility solvents and low humidity.

19

1.4.2 Sol-Gel Electrospinning

In addition to the normal challenges of polymer electrospinning the inclusion of the ceramic precursors necessary to make piezoelectric and magnetostrictive fibers further complicate the process. In the process of adding the necessary ceramic precursors to the solution a sol is formed which gradually undergoes a sol-gel reaction.

Two sol-gel processes, or routes, of interest include the hydrolysis and condensation of metal alkoxides, and the gelation involving a concentration of metal-chelates. The steps involved in the sol-gel process involve (1) mixing, (2) gelation, (3) aging, (4) drying, (5) stabilization, and (6) densification.28

The sol-gel reaction which takes place before and during the electrospinning

process can impact the resultant nanowire morphology. The first use of

electrohydrodynamics to create submicron fibers in a ceramic system was

demonstrated in 2003 by Larson et al.29 in which it was proposed that the increase in air

sol interface due to the acceleration of the solution towards the counter electrode

affected the sol-gel transition. This possible gelation during the electrospinning process

can impact the viscoelastic properties and resultant morphology in the electrospun

fibers. The concentration of ceramic precursors as well as the chelating agents which

control this sol-gel reaction also can change the electrical properties of the solution,

which greatly impact the electrospinning process.30 In addition to the effects of the sol-

gel process on the viscoelastic and electrical properties which can impact

electrospinning, it can also impact solvent evaporation. As a gel is aged a porous

interconnected network28 is formed within the gel which can slow the solvent

evaporation needed to prevent webbing in electrospun fibers.31

20

1.5 Assembly

Once the nanowires have been fabricated they must be fabricated via the electrospinning process as chosen in section 1.4 additional steps are needed to incorporate them into devices. A main challenge in producing devices out of nanomaterials, is assembling them in such a way that maintains their structure and properties. Thus, AC electrical assembly was chosen as it is well suited to assembly of nanowires/particles. Although various types of bottom up fabrication techniques exist, including electrophoretic deposition55 and 3D printing methods,56 AC electrical assembly

utilizing the dielectrophoretic effect is particularly suited to producing arrays of

nanowires suspended across electrodes and its scalability has been previously

demonstrated to produce dense arrays of nanowires.57,58 In AC electrical assembly a

nanowire, or particle suspended in solution spontaneously forms a dipole, and

experiences a force along the gradient of the electric field, toward the electrode gap,

referred to as the dielectrophoretic force.59,60 Once near the electrode gap short range capacitive forces act to orient the nanowires across the gaps.57 Other forces present include dipole-dipole interactions, electrostatics, capillary forces, and AC- electroosmosis.57,61 These can cause repulsion or chaining between nearby nanowires,

adhesion to the substrate, disruption of nanowires upon drying, and a flow of solvent

around the nanowires, respectively, to varying extents depending on the assembly

parameters, such as the electrical and rheological properties of the nanowires and

solvent. Thus, these assembly parameters can be tuned to achieve improved nanowire

assembly.

As AC electrical assembly is highly dependent on the electrical properties,

namely conductivity and permittivity, of the solvent and particles used, electrical

21

assembly of the Janus nanowires was attempted in various solvents: water, ethanol, 2- methoxyethanol, and butanol. The motivation for this investigation was to obtain improved assembly guided by dielectrophoretic force and the real portion of the

Clausius Mossotti factor for a particle in solution at high frequencies.62

[ ] | |

𝐷𝐷𝐷𝐷𝐷𝐷 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 [ ] ( 𝐹𝐹 ∝ 𝜀𝜀 )𝑅𝑅𝑅𝑅/( 𝐾𝐾 ∇��𝐸𝐸���⃑ + 2 ) ∗ ∗ ∗ ∗ 𝑅𝑅𝑅𝑅 𝐾𝐾 ≈ 𝜀𝜀𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 − 𝜀𝜀𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 𝜀𝜀𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝜀𝜀𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 = + / (1-1) ∗ 𝜀𝜀 𝜀𝜀 𝑖𝑖𝑖𝑖 𝜔𝜔 Where is the dielectrophoretic force, is the dielectric constant of the

𝐷𝐷𝐷𝐷𝐷𝐷 medium or particle𝐹𝐹 being assembled, the conductivity𝜀𝜀 of the medium or particle, the

frequency of the applied AC electric field𝜎𝜎 , and | | the gradient of the magnitude of𝜔𝜔 the

electric field. A positive Claussius Mossotti factor,∇��𝐸𝐸��� ⃑Re[K]>0, gives the desired positive

dielectrophoretic force, an attraction of the nanowires toward the electric field gradient

maxima i.e., the edges of the electrodes (Figure 1-3). Detailed estimations of the

direction of the dielectrophoretic force

1.6 Magnetic Field Sensors

Above an overview of the methods of device fabrication for the magnetic field

sening arrays were covered. Here, an overview is given of how these arrays fit into the

current technology space. There are a number of active and passive magnetic field

sensing options for sensing DC and AC magnetic fields for applications including

position sensors, rotational speed sensors, and current sensors. As this is a rather large

scope of applications, this manuscript will focus on the implementation of the

magnetoelectric nanowire sensors developed here in rotational speed sensors, then

move to their incorporation in a proximity current sensing IC. First a brief background of

22

magnetic field sensors for rotational speed sensing will be given followed by where magnetoelectric sensors, and specifically where magnetoelectric sensors fabricated from magnetoelectric nanowires offer unique advantages.

Rotational speed sensors rely on the use of magnetic field sensors to detect the passing of a ferromagnetic gear tooth or gear teeth. The electrical output of the sensor is then used to measure the frequency of the passing of the gear teeth. Among the passive magnetic field sensors there are variable reluctance and Weigand effect sensors. Hall effect, magnetoresistive, anisotropic magnetoresistive (AMR), and giant magnetoresistive (GMR) sensors comprise the active sensing options.

Variable reluctance or inductive sensors are devices which rely on the changing flux through the sensor by the passing magnetic gear teeth to generate a voltage in a sensing coil via Faraday’s law. In general, these sensors offer low cost, moderate size, passive sensing, and good temperature stability. Disadvantages include air gaps between the sensor and gear tooth of less than 2 mm, variable signal strength, and loss of signal at zero shaft speed.

In Wiegand effect sensors, to allow for higher level pulsed signals at lower rotational speed, a magnetic alloy wire with a radial gradient magnetization which varies from the core to the surface of the wire is added within the coil. As the strength of the magnetic field applied to the magnetic alloy increases beyond a critical value, the resultant rapid switch in polarity of the magnetization of the Weigand element induces a self generated voltage in the pickup coil. However, this results in much more costly sensors and spikelike signal output.

23

Active sensors offer the ability to sense the flux level itself instead of essentially measuring changes in flux, this is due to the fact that they do not rely on the energy generated by the change in flux to generate the output signal. The main disadvantage of this is that they need varying levels of power supplied to the sensing element. This power consumption also means an increase in current to the sensor which requires thicker gauge wire to the sensor.

Hall effect sensors are among the more common active sensors. They rely on the Hall effect in which a magnetic flux imposes a force upon a current flowing through it. This then leads to a voltage perpendicular to the path of the original path of the current which is linearly proportional to the transverse component of the flux density passing through the sensing element. Hall effect sensors utilize bipolar semiconductor technology. Thus, they can be fabricated on chip along with signal processing circuitry.

This circuitry allows for amplification and buffering, temperature compensation, and other signal conditioning functions in one package; decreasing the cost. Other advantages include small size, and operation at zero speed. Disadvantages of this sensor include a maximum operating temperature of around 175 °C and an air gap no greater than 2.5 mm.

Magnetoresistive sensors are those that exhibit a change in resistance due to an external magnetic field. This can be achieved through an ordered arrangement of striped conductors within a high carrier mobility semiconductor, where an applied magnetic field changes the path of the electrons flowing through the sensor due to the

Lorentz force. AMR sensors rely on a phenomena noted by William Thompson, and again later by Lord Kelvin in 1857, in which a change in a the resistance of current

24

flowing in different directions within ferromagnetic metals is changed in relation to the direction of its magnetization.16 GMR sensors similarly measure the change in

resistance with respect to magnetization but use alternating layers of alternating

ferromagnetic and nonmagnetic layers to achieve larger changes in resistance.17

Generally these sensors offer operation at zero speed, the ability to sense the direction

of rotation, air gap operation up to 3 - 3.5 mm, on chip signal processing circuitry, and

temperature stability up to 150 - 200 °C. Disadvantages to these sensors include

medium size and cost, as well as being an active device which requires supply current

for the sensing element.

Magnetoelectric sensors utilizing magnetoelectric nanowires offer the potential

for a passive sensing element which could (1) allow for the integration of on chip signal

processing circuitry, (2) a large air gap, (3) low size and cost, and (4) good temperature

stability. While there are some disadvantages including loss of signal at zero speed, as

with all passive sensing elements, and lack of sense of rotational direction, these

sensors should be an optimal choice for a number of rotation speed sensing

applications.

1.7 Dissertation Overview

This dissertation presents a method to fabricate low power sensors using

magnetoeletric nanowires. Chapter 1 provided a background on the current state of the

art, as well as information on the synthesis and assembly methods. In Chapter 2 the

scaleable fabrication of aligned barium titanate nanofibers through electrospinning on a

rotating mandrel will be covered. This control of alignment could prove useful for the

implementation of piezoelectric fibers into useful transducers by providing a method to

fabricate patterned stuctures in situ. Additionally, in the process of fabricating and

25

characterizing aligned barium titanate fibers information about electrospinning fiber diameter distributions and the crystal structure of aligned barium titanate as a function of calcination conditions was learned which is useful for better understanding the following chapters. Chapter 3 covers the fabrication of biphasic Janus magnetoelectric barium titanate/cobalt ferrite nanowire magnetic field sensing arrays was covered. This

included the electrospinning and calcination to form the magnetoelectric nanowires, AC

electrical assembly to form the nanowire arrays, and the magnetoelectric

characterization of these arrays. In Chapter 4 the fabrication of ultra low power lead

zirconate titanate/nickel zinc ferrite nanowire based current sensors is presented.

Finally, Chapter 5 provides a summary and outlook of the research presented here.

26

Figure 1-1. Electrospinning setup where the two phases are flown through an electrically charged syringe needle to form a biphasic Janus fiber consisting of a cylinder where one hemisphere is comprised of the piezoelectric precursor solution and the other hemisphere is comprised of the magnetostrictive precursor phase.

27

Figure 1-2. Electrospinning diagram showing sol gel solution being flown through a positively charged syringe needle, charges accumulating on the surface of the solution pulling it into a conical shape referred to as a Taylor cone, emitted from the Taylor cone, and accelerated toward a grounded counter electrode.

28

Figure 1-3. Diagrams illustrating the gradient of the electric field gradient showing the areas where the nanowires will be attracted to from a positive dielectrophoretic force. (A) Diagram showing the physical electrode - gap - electrode structure of parallel electrodes across which the nanowires will be assembled. A COMSOL simulation showing the high electric field regions at the edges of the electrodes where the nanowires will be attracted in positive dielectrophoresis is given in (B) top and (C) side view.

29

CHAPTER 2 ALIGNED BARIUM TITANATE

2.1 Experimental Process

Electrospinning is a versatile method for producing nano- to micro-meter sized

fibers that are suitable for both rapid prototyping and large scale production.26∗

Electrospinning has been used to synthesize a wide array of nanoscale materials

including polymeric,27–29 ceramic,30–33 and even composite bi- and multi-phasic fibers

and particles.34–37 Fibers formed via eletrospinning are promising candidates for use in

many applications including filtration, catalysis, wound healing, and drug transport and

release.26 Traditionally, electrospinning results in the formation of randomly oriented

fibers. However, fiber alignment can be achieved by tailoring the geometry of the

counter electrode.29,38 In some applications, electrospinning a large mat of randomly

oriented fibers is sufficient, whereas for others alignment or a desired orientation in the

electrospun fibers may be desired.8,29,38 The piezoelectric and ferroelectric properties

and high dielectric constant of barium titanate have enabled its incorporation into a wide

array of electrical applications including actuators. This chapter demonstrates a route to

produce large quantities of uniaxially aligned barium titanate nanofibers in a manner

that can also be applied to a wider range of ceramic and metallic electrospun

nanofibers.1 medical imaging devices,2 ultrasound transducers,3 and high dielectric

capacitors.4 Nanoscale barium titanate has been considered a promising material for

high density ferroelectric random access memories.5–9 Additionally, electrospun barium

∗ Many aspects of this Chapter have been published and are adapted here with permission from “Structure–Property Relationships in Aligned Electrospun Barium Titanate Nanofibers”, by Bauer et al. in the Journal of the Amereican Ceramics Society volume 99, issue 12

30

titanate nanofibers have been used to construct high performance humidity sensors in place of barium titanate thin films due to their high surface area and large aspect ratio.7,9 A key figure of merit for barium titanate nanowires is their piezoelectric

coefficient, d, which is a measure of the magnitude of polarization that is generated in

response to an applied load. Given the push towards incorporating barium titanate and

similar perovskite nanofibers and nanowires into electrical devices, it is important to

develop methods to synthesize large quantities of these nanofibers. piezoelectric

coefficient is often reported, which measures the piezoelectric response along the axis

of the applied electric field.5,6,8 Control of their orientation is desirable since randomly

oriented fibers may not be best suited for use in microelectronics. While well aligned

barium titanate/polymer nanofibers have been previously synthesized with the use of

two parallel electrodes separated by an insulating air gap,6,8 the area of fibers which can

be produced with this method is limited by the distance between the electrodes and

their length. The method is further limited by a loss in alignment as subsequent fiber

layers are deposited.29 This has been overcome in polymer systems by parallel

electrodes spaced around a rotating mandrel.29 However, for ceramic nanofibers

compared to polymer fibers there is an additional challenge of maintaining alignment

during calcination. The as-spun fibers are amorphous thus a calcination step is required

to burn off the polymer binder and to form the final crystalline product.

This chapter demonstrates the first example of alignment in ceramic nanofibers

that persists during calcination by electrospinning onto a rotating mandrel. In this

chapter, two methods of maintaining fiber alignment throughout the calcination process

are explored. The first method involves the removal of the as-spun fibers using carbon

31

tape and the other involves the removal of sections of fibers by peeling them off the wire

drum as sheets. In this work the effects of other processing variables, including

calcining time and temperature on the structure and properties of the resultant

nanofibers will be examined.

2.1.1 Materials

Barium acetate, poly(vinyl pyrrolidone) (PVP, MW = 1300 000), acetic acid, and

titanium isopropoxide (≥97.0%) were obtained from Sigma Aldrich (St. Louis, MO).

2.1.2 Barium Titanate Sol–Gel Preparation

The barium titanate precursor solution was prepared in a 1:1 molar ratio of

barium to titanium by first dissolving 2.55 g barium acetate in 8 mL of acetic acid

followed after 2 h by the dropwise addition of 2.95 mL titanium isopropoxide.36,39 A

polymer solution was simultaneously prepared, consisting of 1.1 g PVP (MW = 1 300

000) in 10 mL ethanol and was also stirred for 2 h. Then, the barium titanate solution

was added to the polymer solution and allowed to stir for 10 min before electrospinning.

2.1.3 Electrospinning Aligned Barium Titanate Fibers

To electrospin aligned nanofibers, a counter electrode consisting of parallel

copper wires was constructed on a rotating mandrel (Fig. 2-1). The copper wires had a

separation of 1 cm and the drum was rotated by a DC electric motor at a rate of 21 rpm,

determined by the rotational speed of the DC motor and the pulley diameters.

For electrospinning, the barium titanate precursor solution was fed through a

stainless steel needle (20 gage) with a flow rate of 2 mL/h. The needle was connected

to a high voltage power supply, with an applied voltage of 13 kV, and positioned normal

to the surface of the electrically grounded rotating copper wire drum, with a separation

distance of 16 cm. After electrospinning, two methods were tested for the removal of as‐

32

spun fibers from the copper wire mandrel for calcination: removal by peeling the fibers off as sheets or by adhering the aligned fibers to a piece of carbon tape. The as‐spun fibers were calcined at 750°C, 875°C, and 1000°C for 2, 4, or 6 h.

2.1.4 Characterization

The viscosity of five barium titanate precursor solutions was measured using a

Brookfield DV‐II+ Pro rotational viscometer with a spindle speed of 20 rpm. The average viscosity of the as‐prepared precursor solutions was 137 cP with a standard deviation of

14 cP.

The nanofiber samples were imaged with a FEI Phillips (Hillsboro, OR) XL40

FEG scanning electron microscope (SEM). Fiber diameters were measured using

ImageJ (NIH, Bethesda, MD), where diameter measurements were taken on at least

150 fibers per sample. The measured fiber diameters were then fit to multiple normal distributions using normal mixEM40 in Rstudio (Boston, MA). A fit for multiple log normal

distributions was found by first taking the log transform of the diameter measurements,

fitting the data with normal mixEM, and converting the normal fits of the log transformed

data back into log normal distributions. The crystal structure of the nanofibers was

determined via X‐ray diffraction (XRD) (PANalytical X'Pert Powder, Alemelo,

Netherlands) with subsequent Rietveld refinement in PANalytical Highscore Plus.

Additional crystallographic structure information was obtained using Raman

spectroscopy; Raman spectra were obtained using a Renishaw Invia Raman

microscope (Hoffman Estates, IL) with a 633 nm laser and a 20× objective lens.

Piezoelectric force microscopy (PFM) was performed using a Park Systems XE

70 atomic force microscope (Santa Clara, CA) modified with a function generator

33

(Agilent, 33210A, Santa Rosa, CA). PFM samples were prepared by electrospinning directly onto platinum‐coated silicon wafers with a layer of titanium dioxide between the platinum and silicon to promote adhesion of the thin film, after which they were calcined for 2 h at 750°C. The d33 value of a barium titanate nanofiber calcined at 750°C for 2 h

was obtained via linear regression on the positive voltage sweep using 20 cycles.

2.2 Results and Discussion

After electrospinning onto the copper wire drum, SEM images were obtained to

verify alignment and to observe the morphology of the as‐spun fibers (Figure 2-2). After

the alignment of the as‐spun fibers was verified, the fibers were removed from the

copper wire mandrel using two different methods. In the first method, fibers were

removed by adhering them to carbon tape. The carbon tape was then placed on a

ceramic dish, and was calcined in air at 750°C for 2 h, resulting in the oxidation and

removal of the carbon tape from the fibers (Figure 2-2(B)). The adhesive on the carbon

tape restricted the shrinking of the fiber as the PVP was removed, leading to the fibers

being pulled apart lengthwise, breaking to form aligned nanowires instead of longer

nanofibers (Figure 2-2(B–C)).

The second method of maintaining alignment during calcination involved

removing the fibers by peeling off sections of sheets from the copper wire drum. Using

this method, the fibers were free to shrink along their length during calcination. Figure 2-

2(C–D) show that this sheet removal method maintained alignment and resulted in

uniaxially aligned nanofibers, after calcination at 750°C and 1000°C, respectively.

Figure 2-3(A-B) shows the fiber diameters distributions for both calcined and

uncalcined fibers, respectively, revealing a bimodal distribution of fiber diameters. This

34

multimodal distribution has been previously observed in both polymer41,42 and sol–gel

electrospinning43 and has been attributed to branching, the formation of secondary

polymer/sol–gel jets from the primary jet, during the electrospinning process. This

branching phenomena is due to an instability of the circular cross section of the fiber at

high potentials. This instability causes the formation of cusps from which secondary jets

can be emitted.41 Though previously such multimodal fiber distributions have been

modeled using multiple normal distributions,42 it was found that the diameters of the

electrospun fibers were better fit by multiple log normal distributions. The origin of this

log normal fiber diameter distribution can be related to solution inhomogeneities that are

present during electrospinning. In electrospinning, a polymer or sol–gel solution is

placed in a syringe that is connected to a high voltage power supply, which leads to the

formation of an elongated droplet at the syringe tip, referred to as the Taylor Cone. The

radius of the Taylor cone has been shown to be a determining factor of the droplet size

in electrospraying,44 and it has also been predicted that it is related to the final jet

diameter for electrospinning.24 Additionally, the radius of the Taylor cone is proportional

to the viscosity of the fluid,45 where the viscosity of the polymer is related exponentially

to its concentration, as described by the Martin equation, which varies randomly

throughout the solution.46 These random variations then give rise to a viscosity that is

log normally distributed, resulting in a Taylor cone radius and fiber diameters that also

exhibit a log normal distribution (Figure 2-3).

The fiber diameter distributions from both normal and log normal models were

graphically compared to the experimental cumulative distribution in Figure 2-3(C–D).

The fit of each model distribution was also quantitatively evaluated using the

35

Kolmogorov–Smirnov (K–S) test to determine if the model distribution is a reasonable fit for the experimental results.47 Based on the K–S test, the multiple normal distributions

could be rejected for both the uncalcined (P = 1.60 × 10−4) and calcined (P = 2.60 ×

10−14) fibers. The multiple log normal distributions could not be rejected with the K–S

test with a 95% confidence level, thus the multimodal log normal distribution model was

accepted. The multiple log normal distribution for the uncalcined fibers can be defined

as follows: the first distribution had a weight of 0.58, a mean diameter of 232 ± 117 nm,

while the second had a weight of 0.42, a mean of diameter of 463 ± 58 nm. For the

calcined multiple log normal distribution, the first distribution had a weight of 0.447, a

mean diameter of 180 ± 98 nm, while the second had a weight of 0.553, a mean

diameter of 302 ± 47 nm.

The crystallographic properties of the calcined nanofibers were analyzed using

XRD. The XRD spectra are shown in Figure 2-4, revealing peaks characteristic of

perovskite barium titanate.5,8,48 Rietveld refinement indicated the presence of the

tetragonal perovskite structure for all calcination conditions. However, Figure 2-4 shows

a lack of observable peak splitting in the (200) peak at 45°, which can be attributed to

peak broadening in the nanocrystalline samples.48–50 The Rietveld refinement also

indicated the presence of a cubic and hexagonal phase of barium titanate, in addition to

the presence of barium carbonate (witherite). The metastable hexagonal phase has

been previously shown to coexist with the tetragonal phase of barium titanate in

nanoparticles. The phase transformation between the hexagonal and tetragonal phases

is a reconstructive phase transformation, and thus characterized by sluggish kinetics51.

Therefore, as anticipated, the amount of hexagonal phase present in the sample

36

decreases with increasing calcination temperature, along with a corresponding increase in the amount of tetragonal phase present, when calcination time is kept constant

(Figure 2-5). It is important to note, that all samples had a barium carbonate impurity phase, which varied from 10–14 wt% depending on the calcination temperature. Across all calcination times, temperature was negatively (R = −0.621) and significantly (P =

0.0005) correlated with the wt% of the intermediate phase as more of the recombination transformation occurred. Furthermore, the percent of the tetragonal phase was positively (R = 0.522) and significantly (P = 0.005) correlated with increasing calcination temperature and attributed to the more complete transformation of the metastable hexagonal phase at higher temperatures, where more energy was available to promote the kinetically sluggish transformation. Similarly, longer calcination times also corresponded to an increasing wt% of the tetragonal phase, and a decrease in the amount of hexagonal phase present. However, it was observed that calcination time has a lesser impact on the phases present compared with temperature for the temperatures and times tested. Lastly, both cubic barium titanate and barium carbonate decreased in weight percent with increasing calcination times and temperatures. Rietveld refinement results are given in Table 2-1, where three samples were synthesized and analyzed under each calcination condition. Rwp is a measurement of the distance between the

observed and fitted spectra, Rexp is a measure of the data quality and essentially the

expected Rwp for an ideal fit, and the reduced chi‐square statistic = / gives 2 2 2 𝑉𝑉 𝑤𝑤𝑤𝑤 𝑒𝑒𝑒𝑒𝑒𝑒 an indication of the quality of the Rietveld refinement on the spectra.𝜒𝜒 For𝑅𝑅 example,𝑅𝑅 a

< 1 indicates the spectra may be over fitted, >> 1 indicates a poor fit, and 1 2 2 2 𝑣𝑣 𝑣𝑣 𝑉𝑉 signifies𝜒𝜒 a good fit. 𝜒𝜒 𝜒𝜒 ≈

37

Raman spectroscopy was also employed to further confirm the presence and transformation between the metastable hexagonal and tetragonal phases of barium titanate. Figures 2-6(A) and (B) show the evolution of normalized Raman spectra as a function of calcination temperature and time, respectively. Peaks that are attributed to tetragonal (T),51–54 cubic (C),55–57 and hexagonal (H)51,52,58 barium titanate as well as

barium carbonate (witherite, W)58 are labeled within the spectra. Figure 2-6(A) reveals that the tetragonal (T) peak intensities increase with increasing temperature, alongside a corresponding decrease in the intensities of the metastable hexagonal (H) phase.

However, Figure 2-6(B) shows the evolution of Raman spectra at calcination times of 2,

4, and 6 h, for a constant calcination temperature of 875°C. It is apparent here that the calcination temperature has a greater effect on the transformation from the hexagonal to the tetragonal phase, as also observed in the XRD data.

It is anticipated that barium titanate nanofibers with the highest fraction of the tetragonal phase will result in the best ferroelectric and piezoelectric properties.

Therefore, the effects of calcination on the grain size of the tetragonal barium titanate phase were also determined by relating peak broadening to the average grain size.

Since strain can also contribute to peak broadening, strain was also considered in the

Rietveld refinement, however, no significant strain contributions to peak broadening were found. Figure 2-7 shows a plot of the natural logarithm of grain size of the barium titanate nanofibers as a function of inverse temperature. From Figure 2-7, the activation energy of grain growth of tetragonal barium titanate was calculated to be 2.641 ± .377 eV/(atom K). This fits in to a trend previously observed in barium titanate nanofibers by

38

Peng et al. where the activation energy of grain growth was found to decrease with decreasing fiber diameter.59

Lastly, the piezoelectric properties of the electrospun barium titanate nanofibers

were determined using PFM. Figure 2-8(A) shows the butterfly loop, with corresponding

phase plot Figure 2-8(B) demonstrating a 180° phase shift, obtained from PFM analysis

on a barium titanate nanofiber calcined at 750°C for 2 h, confirming piezoelectric

behavior. The d33 value, extracted via linear regression on Figure 2-8(A) was found to

be 15.5 pm/V with a standard deviation of 0.1 pm/V, in a method established by Zhou et

al.60 The standard deviation reported here refers to that of the fitting parameter based

on the 20 repeated scans used to obtain the butterfly loop that is shown in Figure 2-

8(A). The value reported here is also in good agreement, with a 20 pm/V value

previously reported for piezoelectric barium titanate nanofibers.61 However, it is

important to note that there are several biases present in PFM measurements, including

the effect of surface charges which act upon the body of the PFM cantilever, in addition

to the forces that are acting solely on the tip.62 For cases where domains are more

randomly oriented and relatively small in comparison to the cantilever dimensions, this

bias is reduced, if not eliminated.62 Figure 2-9 shows the phase angle scan of a barium

titanate nanowire, revealing that the as‐calcined nanofibers exhibit a random distribution

of domains, that are much larger than the cantilever tip. It is anticipated that all

calcination conditions would result in the formation of a random domain structure,

requiring subsequent poling steps. Although one‐dimensional piezoelectrics cannot

easily be poled using conventional techniques, methods have been developed where

arrays of piezoelectric nanowires are assembled across electrodes, whereby electrical

39

contacts can be made across both ends of the nanowire so as to apply a voltage along the length of the wire.63 Additionally, a corona discharge poling have been

demonstrated as a means to pole piezoelectric barium titanate nanofibers/nanowires.

63,64

2.3 Conclusion

In this chapter, a method for obtaining aligned barium titanate nanofibers using a

rotating copper wire drum collector, where the alignment was maintained during

calcination by one of two removal methods. A distribution model was quantitatively tested that appropriately fit the measured diameters of the barium titanate nanofibers.

Additionally, the links between processing conditions and phase evolution in electrospun barium titanate nanofibers were presented. By increasing the calcination time and temperature, the formation of a tetragonal barium titanate phase was promoted, leading to the formation of barium titanate nanofibers with enhanced piezoelectric properties. The use of a rotating drum collector, alongside optimized calcination conditions hold potential for the large‐scale synthesis of electrospun ceramic nanofibers with enhanced properties, greatly increasing their ability to be incorporated into novel devices and applications.

40

Table 2-1. Agreement indices for the rietveld refinements on the barium titanate fibers. Agreement indices for the Rietveld refinements performed.Rwt and Rexp are the weighted and expected R-factors, respectively and is the reduced chi- square statistic. 2 𝑣𝑣 Temp (°C) Time (h) Trial Rexp (%) Rwt (%)𝜒𝜒 750 2 1 5.03 5.99 1.422 𝑉𝑉 2 6.12 6.50 1.13𝜒𝜒 3 6.49 7.08 1.19 4 1 6.42 7.95 1.54 2 5.76 6.13 1.13 3 6.87 7.45 1.18 6 1 4.89 5.05 1.07 2 6.22 7.52 1.46 3 6.64 7.56 1.30 875 2 1 5.09 6.00 1.39 2 5.88 6.22 1.12 3 6.60 7.16 1.18 4 1 4.95 6.19 1.57 2 6.14 7.08 1.33 3 6.75 7.44 1.21 6 1 5.13 5.88 1.31 2 6.29 7.36 1.37 3 6.61 7.03 1.13 1000 2 1 6.52 8.47 1.69 2 6.65 8.34 1.57 3 6.86 7.68 1.25 4 1 4.92 6.34 1.66 2 6.31 7.28 1.33 3 6.69 7.69 1.32 6 1 6.42 7.95 1.54 2 6.54 7.94 1.48 3 6.81 8.97 1.73

41

Figure 2-1. Photograph of the rotating mandrel set‐up for the electrospinning of aligned nanofibers showing the as‐spun amorphous fibers aligned across the parallel copper wires on the rotating mandrel. (Photograph courtesy of the author)

42

Figure 2-2 Scanning electron micrographs of A) uncalcined, aligned nanofibers B) aligned barium titanate nanofibers, calcined at 750°C for 2 h after removal via the carbon tape method, and aligned nanofibers calcined at C) 750°C for 2 h and D) 1000°C for 2 h after removal by the peeling off method.

43

Figure 2-3. Fiber diameter distributions for A) uncalcined barium titanate fibers and B) calcined barium titanate fibers with the log normal probability density function given by the black line. Experimental cumulative distribution of the C) uncalcined and D) calcined barium titanate fibers and along with fitted bimodal normal and log normal distributions.

44

Figure 2-4. A) X‐ray diffraction (XRD) spectra for barium titanate nanofibers calcined for 6 h at 750°C, 875°C, and 1000°C. B) XRD spectra for barium titanate nanofibers calcined at 875°C for 2, 4, and 6 h

Figure 2-5. The weight percent of the A) tetragonal, B) cubic, and C) hexagonal phase of barium titanate as a function of calcining temperature, for a constant calcining time of 6 h.

45

Figure 2-6. A) Raman spectra for barium titanate nanofibers calcined for 6 h at 750°C, 875°C, and 1000°C and B) Raman spectra for barium titanate nanofibers calcined at 875°C for 2, 4, and 6 h. Peaks are labeled as being attributed to tetragonal barium titanate (T), cubic barium titanate (C), hexagonal barium titanate (H), or barium carbonate (witherite, W).

46

Figure 2-7. The natural logarithm of average grain size plotted versus reciprocal temperature for the tetragonal phase. The slope from these plots was used to calculate the activation energy for grain growth in the barium titanate nanofibers.

47

Figure 2-8. A) Butterfly loop and B) phase plot of barium titanate nanofiber calcined at 750°C for 2 h, obtained with the assistance of Catherine Snyder.

48

Figure 2-9. Phase angle scan of a barium titanate nanofiber calcined at 750°C for 2 h, revealing the domain structure of the fiber, the width of the scan shown is 1 μm.

49

CHAPTER 3 BARIUM TITANATE / COBALT FERRITE MAGNETIC FIELD SENSING ARRAYS

The fabrication and characterization of the first magnetoelectric sensors utilizing arrays of Janus magnetoelectric composite nanowires composed of barium titanate and cobalt ferrite are presented.1∗ By utilizing magnetoelectric nanowires suspended across

electrodes above the substrate, substrate clamping is reduced when compared to

layered thin-film architectures; this results in enhanced magnetoelectric coupling. Janus

magnetoelectric nanowires are fabricated by sol–gel electrospinning, and their length is

controlled through the electrospinning and calcination conditions. Using a directed

nanomanufacturing approach, the nanowires are then assembled onto pre-patterned

metal electrodes on a silicon substrate using dielectrophoresis. Using this process,

functional magnetic field sensors are formed by connecting many nanowires in parallel.

The observed magnetic field sensitivity from the parallel array of nanowires is

0.514 ± .027 mV Oe−1 at 1 kHz, which translates to a magnetoelectric coefficient of

514 ± 27 mV cm−1 Oe−1.

3.1 Introduction

Magnetoelectrics are unique functional materials in which an applied magnetic

field can be used to control an electrical polarization10. As such they can offer a lower

power alternative to current magnetic field sensors such as Hall Sensors. However, to

be viable for use in magnetic field sensors, material systems and architectures need to

be found which offer high magnetoelectric coefficients (dE/dH). Although single phase

∗ Many aspects of this Chapter have been published and are adapted here with permission from “Magnetic field sensors using arrays of electrospun magnetoelectric Janus nanowires”, by Bauer et al. under a Creative Commons Attribution 4.0 International License http://creativecommons.org/licenses/by/4.0/

50

magnetoelectrics exist, they are comparatively rare12. On the other hand, composite

magnetoelectrics are capable of producing greater magnetoelectric effects. For

magnetoelectrics the figure of merit is the magnetoelectric coefficient (αV), which is

quantified as the magnitude of the electric field (dE) generated in a material in response

10 to an applied magnetic field (dH), αV = dE/dH . Composite magnetoelectrics are

typically composed of magnetostrictive and piezoelectric phases which share an

interface. When exposed to an applied magnetic field the magnetostrictive phase

undergoes a shape change, which imparts a strain to the piezoelectric phase, thereby

inducing an electrical polarization. When magnetoelectric composites are fabricated as

thin films, the strain transfer between the magnetostrictive and piezoelectric phases is

typically limited by the underlying substrate leading to a reduction in the magnetoelectric

effect10,11,13,14. Less rigidly clamped 1-D magnetoelectric nanostructures could offer

increased magnetoelectric coefficients. Enhancements of up to a few orders of

magnitude seem feasible based on theoretical and scanning probe microscopy

measurements15.

However, to make use of magnetoelectric nanowires demands

nanomanufacturing processes that enable (a) the synthesis of nanowires with controlled

length, (b) the ability to direct the assembly of these nanowires into ordered

arrangements while avoiding substrate clamping effects, and (c) the ability to make

suitable electrical connections to one or more nanowires.

Here, this section seeks both to demonstrate suitable nanomanufacturing methods and to confirm the increased magnetoelectric coefficients 1-D structures offer by fabricating magnetic field sensors consisting of an array of magnetoelectric biphasic

51

fibers suspended across electrodes (Figure 3-1). Specifically the barium titanate and

cobalt ferrite system was selected for the 1-D magnetoelectric, as it has previously been

shown to have a significant magnetoelectric effect in bulk65–70 and thin film form71. A

bilayer, Janus, morphology was chosen in order to promote the bending mode in the

magnetoelectric, which will likely allow for greater strain in the nanowire while

suspended between the electrodes.

As mentioned in Chapter 1, several methods exist to fabricate 1-D

magnetoelectrics including sol-gel electrospinning36,72–75 , hydrothermal synthesis76, and

various chemical and physical vapor deposition processes77. In this work sol-gel

electrospinning was selected as it has been previously shown to be capable of

producing magnetoelectrics with a wide range of compositions and various

connectivities including fibers with Janus36,74,75, core shell73,78, and randomly dispersed

morphologies72. Sol−gel electrospinning was also chosen due to its scalability and its

relatively low cost. Additionally, when combined with a bottom up assembly technique,

the high temperature calcination step can be performed off substrate making it CMOS

compatible and feasible for application in a manufacturing setting26,79.

Sol-gel electrospinning is a method by which a ceramic/polymer solution is

drawn, often from a syringe needle, into a nanofiber using a large electric field that is

applied between the solution and a counter electrode19,80. While the sol-gel solution is

extruded into a droplet at the tip of the syringe needle, a surface charge forms on the

droplet due to the applied voltage. As the solution accumulates a surface charge it is

pulled toward the counter-electrode by the electric field in a shape referred to as a

Taylor cone19,80. When the surface charge on the solution overcomes surface tension, a

52

charged jet is emitted from the Taylor cone. This jet is then accelerated toward the counter-electrode by the applied electric field, during which time the solvent evaporates and hydrolysis and condensation of the precursors occurs19,80. After electrospinning, the

as-spun amorphous fibers undergo a high temperature calcination step to burn off the

polymer and to crystallize the ceramics.

For the assembly of 1-D magnetoelectrics into devices, it is desirable to have

discrete nanowires. Therefore, a controlled salt calcination method was developed,

where an increased temperature ramp rate was employed to break up the as- electrospun continuous nanofibers into shorter nanowires and the salt prevented agglomeration during calcination, resulting in the formation of discrete nanowires81. This

is similar to a method the method utilized in Chapter 2 for the fabrication of barium

titanate nanowires that leveraged the tension caused by the shrinkage of the as-spun fibers during the calcination step, while clamped by carbon tape, to break up the nanowires radially82. For assembly a nanowire slightly longer than the electrode gap is

desirable since the nanowire must bridge the electrode gap for electrical connections to

be made (Figure 3-1) and nanowires which are too long may quickly settle out of

solution. As such this work sought to find electrospinning and calcination parameters

which could provide control of the nanowire lengths. The two parameters that are

focused on here are the electrospinning voltage, as a means to control the as-spun fiber

diameter and calcination ramp rate. Though these are not the only parameters which

could control nanowire length, they can be readily applied to other systems.

One of the challenges in producing devices out of nanomaterials, is assembling

them in such a way that maintains their structure and properties. Thus, AC electrical

53

assembly was chosen as it is well suited to assembly of nanowires/particles as

mentioned in Chapter 1. Although various types of bottom up fabrication techniques exist, including electrophoretic deposition83 and 3D printing methods84, AC electrical

assembly utilizing the dielectrophoretic effect is particularly suited to producing arrays of

nanowires suspended across electrodes (Figure 3-1) and its scalability has been

previously demonstrated to produce dense arrays of nanowire85,86. In AC electrical assembly a nanowire, or particle suspended in solution forms a dipole in response to an applied electric field, and experiences a force along the gradient of the electric field, toward the electrode gap, referred to as the dielectrophoretic force85,87. Once near the

electrode gap short range capacitive forces act to orient the nanowires across the

gaps85. Other forces present include dipole-dipole interactions, electrostatics, capillary

forces, and AC-electroosmosis85,87. These can cause repulsion or chaining between

nearby nanowires, adhesion to the substrate, disruption of nanowires upon drying, and

a flow of solvent around the nanowires, respectively, to varying extents depending on

the assembly parameters, such as the electrical and rheological properties of the

nanowires and solvent. Thus, these assembly parameters can be tuned to achieve

improved nanowire assembly.

As AC electrical assembly is highly dependent on the electrical properties, namely conductivity and permittivity, of the solvent and particles used, electrical assembly of the Janus nanowires was attempted in various solvents: water, ethanol, 2- methoxyethanol, and butanol. The motivation for this investigation was to obtain improved assembly guided by the real portion of the Clausius Mossotti factor as

54

covered in Chapter 1 Eq. 1-1. Post assembly upper electrical contacts were created using lithography and electroplating.

For characterization of the magnetoelectric nanowires and assembled arrays it was desireable to verify ferrimagnetism in the cobalt ferrite phase88, ferroelectricity in the barium titanate phase, and finally quantify the magnetoelectric coefficient; this was accomplished through vibrating sample magnetometry (VSM), capacitance voltage (C–

V) measurements, and direct magnetoelectric measurements respectively. A direct magnetoelectric measurement that records an electrical response to an applied magnetic field89,90 , were focused on as the main characterization technique for

magnetoelectricity as this is the mechanism by which the array can be used for passive

magnetic field sensing.

In this chapter the fabrication of the first passive magnetic field sensors using 1-D

magnetoelectric nanostructures through scalable nanomanufacturing methods was

detailed. The presented fabrication techniques are readily applicable to a wide range of

magnetoelectric systems allowing their extension to many magnetoelectric-based

devices.

3.2 Experimental

3.2.1 Nanowire Fabrication

First barium titanate and cobalt ferrite sol gel precursor solutions were prepared.

A barium titanate ceramic solution was prepared by dissolving 0.4246 g barium acetate

in 3 ml acetic acid at 80 °C under constant stirring, followed by cooling to room

temperature. After 1 h 0.493 ml of titanium isopropoxide was added. Simultaneously, a

polymer solution was prepared by dissolving 0.4 g polyvinylpyrrolidone in 3 ml ethanol

under constant stirring. After an additional hour, the ceramic solution was added

55

dropwise to the polymer solution under constant stirring. Similarly, a cobalt ferrite

ceramic solution was prepared by dissolving 0.484 g cobalt nitrate hexahydrate and

1.342 g ferric nitrate nonahydrate in 2 ml of acetic acid and 0.75 ml ethanol. After stirring

for 1 h 0.412 ml acetylacetone was added. Simultaneously a polymer solution was

prepared by dissolving 0.4 g polyvinylpyrrolidone in 3 ml ethanol. After an additional

hour the ceramic solution was added dropwise to the polymer solution under constant

stirring.

Both solutions were co-electrospun side by side to form Janus nanofibers, with

one semi-cylinder of the fiber consisting of cobalt ferrite and the other barium titanate16.

The ceramic/polymer nanofibers were then calcined; burning off the polymer and

breaking them along their length, and crystallizing the amorphous as-spun ceramic at

1100 °C. To prevent sintering together of nearby nanowires a salt calcination was performed in sodium chloride. The effect of electrospinning voltage and calcination ramp rate on fiber diameter were studied. After salt calcination the salt and nanowires were immersed in a beaker of water, wherein the salt dissolved. After which a permanent magnet was used to attract the nanowires to the bottom of the beaker and the excess water was decanted. The nanowires were placed in dialysis tubing, and dialyzed in deionized water to remove the salt. Post calcination the nanowires were imaged via scanning electron microscopy, and their crystal structure was analyzed via

X-ray diffraction with Rietveld Refinement. A dilute HCl wash was used to remove any barium carbonate from the surface of nanowires, verification of the effectiveness of this technique, as tested on barium titanate, was obtained via Raman spectroscopy.

56

3.2.2 AC Electrical Assembly

Electrical assembly of the Janus nanowires was attempted in water, ethanol, 2-

methoxyethanol, and butanol. For assembly in water, citric acid was added to the solution such that the pH of the solution was around 9 to achieve a more stable suspension of the nanowires. For the ethanol, 2-methoxyethanol, and butanol solutions

the nanowires were placed a centrifuge tube and nearly all of the water was similarly

decanted while the nanowires were held in place with a permanent magnet. The

nanowires were then dried in a vacuum oven, the respective solvent was added, and

the solvent/nanowire solution was sonicated and vortexed.

During the electrical assembly a droplet of the nanowire solvent solution was

placed over the electrode array. For the initial assemblies of nanowires in parallel

across interdigitated electrodes to tune assembly parameters, a function generator was

used which could supply a sinusoidal voltage of 20 volts peak to peak (Vpp), and the

the frequencies were swept from 100 Hz–10 MHz. Since 5 kHz seemed to promote

positive dielectrophoresis, this frequency was used with an applied voltage of 20 Vpp

across the electrode gaps, to compare the assemblies using each of the solvents. Once

a suitable solvent for assembly was found, test arrays were fabricated which allowed

multiple rows of nanowires to be measured from a single assembly. When performing

assembly on these test arrays a pulse generator which was capable of producing higher

voltages was implemented, and these assemblies were performed with 42 Vpp at 5 kHz.

After nanowire assembly, upper electrode contacts were formed with the nanowires via

spin coating AZ1512 photoresist, optical lithography and removal of photoresist from the

ends of the nanowires, electroplating copper, and stripping of the remaining photoresist.

57

3.2.3 Characterization

To verify ferrimagnetism in the barium titanate/cobalt ferrite nanowires, the mass

magnetization curve, M–H, was obtained using a VSM. For the M-H curve the field was swept from 22 to −22 kOe, then swept back up to 22 kOe. This M–H curve was not obtained from the assembled nanowires on the array but in nanowires fabricated in the same batch as those assembled.

To verify ferroelectricity in the barium titanate/cobalt ferrite nanowires, C–V curves were measured using a Hewlett Packard 4294 A Impedance Analyzer by probing the response of an assembled row of nanowires. The C-V bias voltage was swept from

−30 to +30 V and from −30 to +30 V with an AC source voltage of 100 mV and frequency of 750 kHz. A low AC source voltage was desired for the nanowire C-V curves so that the AC signal would have a minimal effect on the polarization state of the ferroelectric barium titanate. For comparison, an empty electrode substrate (no nanowires) was measured with the same AC source voltage.

A lock-in measurement setup90, depicted in Figure 3-2 was used to measure the

magnetoelectric effect in the nanowire arrays as a function of bias field and frequency.

As shown in Figure 3-2b, the magnetic fields were applied transverse to the long axis of

the nanowires. The DC bias field HDC was cycled from 0 kOe up to 8 kOe, then from

0 kOe down to -8 kOe using a GMW 3473-70 electromagnet with a 5 second dwell at

each bias field value. The DC bias field was measured using a Hall probe and

Microsense Gaussmeter, which also provided closed-loop control of the current

delivered to the electromagnet. A small AC field at 200, 500, or 1000 Hz was applied via

Helmholtz coils placed at the electromagnet poles. The Helmholtz coils were powered

using a Hewlett Packard 33120 A function generator and a Pyramid PB-101 amplifier.

58

For each frequency a 1 Oe AC field amplitude was targeted, but the actual AC field amplitude varied with the DC bias field by ~20% due to changes in the reluctance of the electromagnet. Consequently, the actual applied AC field was measured by a second

Hall probe connected to a Lakeshore 425 Gaussmeter for use in calculating the magnetoelectric coefficient. The analog output of the AC field signal from the

Gaussmeter was amplified by a Stanford Research Systems SR560 low-noise

preamplifier and served as the phase reference for the lock-in measurements. The

voltage from the nanowire array was measured using two Stanford Research Systems

SRS830 lock-in amplifiers phase locked at 0 and 90 degrees relative to the AC

magnetic field.

The voltage waveform produced by the nanowire array is potentially a

combination of the magnetoelectric response as well as an electromagnetic induction.

For a sinusoidal excitation, these two constituent signals would each be sinusoids at the

same frequency as the applied AC magnetic field, but they would be expected to have

differing phases (relative to the AC magnetic field). The total measured signal can thus

be expressed in phasor form as = + , where a tilde over a variable is

𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀 𝐼𝐼𝐼𝐼𝐼𝐼 used to indicate that it is a phasor.𝑉𝑉� Consequently,𝑉𝑉� 𝑉𝑉� efforts were made in the experimental

setup to minimize the induction signal by orienting the wires connecting to the nanowire

array in a manner so as to minimize the cross-sectional area of any loop. As will be shown in a second experiment, the induction voltage is relatively small (<15%) compared to the magnetoelectric signal. The magnetoelectric coefficient was calculated as

| | = , = 10 | | (3-1) 𝑉𝑉�𝑀𝑀𝑀𝑀 𝛼𝛼𝑉𝑉 𝐿𝐿∙ 𝐻𝐻𝐴𝐴𝐴𝐴 𝑤𝑤ℎ𝑒𝑒𝑒𝑒𝑒𝑒 𝐿𝐿 𝜇𝜇𝜇𝜇

59

The nanowires generate sufficiently large voltage to allow direct measurement of the output voltage using an oscilloscope. To obtain additional zero-bias magnetoelectric measurements at additional frequencies up to 10 kHz, a second magnetoelectric measurement setup was constructed as shown in Figure 3-3. Here, an AC magnetic field (ranging from 1 to 5 Oe) was applied to the nanowire array via Helmholtz coils driven by a function generator. As before, the magnetic fields were applied perpendicular to the long axis of the nanowires. The AC field was measured by a

Lakeshore Hall Probe, and the output voltage is measured via a Tektronix DPO2004

Oscilloscope. In the setup, the nanowire array substrate is placed on a stage that facilitates physical rotation of the substrate in order to explore the influence of unwanted electromagnetic induction in the output signal. The induction arises due to the finite cross-sectional area determined by the substrate electrode geometry and connecting wires.

With a voltage of 20 Vpp @ 10 kHz applied to the Helmholtz coils, the nanowire

array is rotated 360° in 30° increments, where the rotation angle is defined as φ. During

this rotation, the fields acting on the nanowires remain in the same direction, but the

electromagnetic induction signal varies both constructively and deconstructively. The

observations from this measurement are subsequently fit to a cosine wave, which is

then used to determine the angle φmin,ind at which inductive effects have approximately no

contribution to the output voltage of the array. Here the angles at which induction has

the maximum additive and subtractive contribution were also determined. To minimize

inductive effects on the magnetoelectric measurements the substrate was positioned to

the angle found above of φmin,ind while obtaining the magnetoelectric coefficient vs

60

frequency spectra as detailed below. Additionally to examine the maximum destructive and constructive contributions of induction, from the array and electrical connections, to the effective magnetoelectric coefficient spectra were also taken at angles φmin,ind ± 90°.

A spectra of the zero-bias magnetoelectric coefficient at frequencies ranging from

20 Hz to 10 kHz was obtained using applied AC magnetic fields between 1 and 5 Oe. At

a given frequency, the voltage output of the nanowire array and the Hall probe output

were measured for 5 different voltages applied to the Helmholtz coil (6, 7, 8, 9, and

10 Vpp). The magnetoelectric coefficient at that frequency was found by performing a

linear regression to the voltage amplitude vs. field amplitude data. This process was

automated via computer control of the instruments and subsequent data analysis with

Python and PyVISA.

3.3 Results

3.3.1 Nanowire Fabrication

Janus barium titanate and cobalt ferrite magnetoelectric nanofibers were formed via sol gel electrospinning and utilized a rapid calcination ramp rate to shrink the fibers radially, creating tension to break them into shorter nanowires. A scanning electron microscope image of calcined nanowires is shown in Figure 3-4a, revealing that the electrospinning and subsequent calcinations utilizing fast ramp rates were successful in producing barium titanate/cobalt ferrite Janus nanowires. An image of a single Janus nanowire is shown in Figure 3-4b as here it is easier to distinguish the two distinct

halves of the wire. The effects of electrospinning voltage and calcination ramp rate on

nanowire length were investigated, as for the subsequent electrical assembly step

nanowires sufficiently long to span the electrode gap but not so large as to settle quickly

out of solution are desireable. It was believed that increasing the calcination ramp rate

61

could decrease nanowire length as the resultant faster polymer burnoff should lead to the breakup of the fibers into shorter nanowires. This is supported by Figure 3-5a which shows a decrease in nanowire lengths from 29.06 ± 19.34 μm to 19.34 ± 6.08 μm when the calcination ramp rate was increased from 10 to 25 °C min−1. It was also

hypothesized that as-spun nanofibers with larger diameters would in turn produce

longer nanowires. The electrospinning voltage can be readily tuned to control the fiber

diameter, where higher electrospinning voltages result in smaller diameter fibers. This is

because an increase in applied field produces a larger elongating force on the fiber jet

during electrospinning, leading to smaller diameter nanofibers, which given the same

calcination ramp rate would form similar aspect ratio, and thus shorter nanowires.

Figures 3-5b, c demonstrate that a decrease in electrospinning field from 2 to

1.83 kV cm−1 resulted in longer nanowires, increasing the length from 29.06 ± 19.34 μm

with 2 kV cm−1 to 77.43 ± 46.11 μm with 1.83 kV cm−1, and larger diameter fibers with

similar aspect ratios. This is also supported by the positive correlation coefficient

between nanowire length and diameter as shown in Figure 3-5c of R = 0.604 with a p-

value of p < 0.01 (>99% confidence level). It is also important to note the

heteroskedasticity in the Figure 3-5c; i.e., that an increasing nanowire diameter is

positively correlated with increased nanowire length, but also increased length variation.

To verify the crystal structure of the calcined barium titanate and cobalt ferrite

nanowires first X-ray diffraction (Figure 3-6) was performed, then subsequently

analyzed the results using Rietveld refinement. From this it was found that the fibers to

be comprised of 62 wt.% tetragonal barium titanate, P4mm, and 38 wt.% spinel cobalt

ferrite, Fd-3m. The agreement indices of the refinement, Rexpected = 5.19, Rweighted = 5.49,

62

and χ2 = 1.26, indicate that these are indeed the structures present and that there are

low levels of or no crystalline impurities. Though XRD showed no impurities, a barium

carbonate surface impurities in as-calcined barium titanate nanowires formed via sol gel

electrospinning has been previously observed in Chapter 291. The removal of this

impurity via acid treatment with dilute HCl in single phase barium titanate nanowires

was tested using Raman spectroscopy; single phase wires were used for this test so that any signal from the cobalt ferrite phase would not obscure the barium carbonate peaks. In Figure 3-7 it can be seen that the barium titanate phase contains barium carbonate peaks which are no longer present after acid treatment, showing that a dilute

HCl treatment can remove the barium carbonate impurity from the as-calcined wires.

3.3.2 AC Electrical Assembly

To assemble the magnetoelectric Janus nanowires across parallel electrodes to

form devices the nanowires were dispersed them in solution and utilized an AC

electrical assembly technique85. To achieve successful electrical assembly, a number of

electrical assembly parameters including nanowire surface treatments, solvent, and

frequency of the applied electric field were varied. Initial assembly attempts of Janus

nanowires in water were unsuccessful, which was attributed to the presence of low

dielectric constant barium carbonate on the surface of the nanowires. While the removal

of barium carbonate promoted strong positive dielectrophoresis in single phase barium

titanate nanowires in water, Figure 3-8, it did not sufficiently improve assembly in the

Janus nanowires. This suggests that the relatively low permittivity of cobalt ferrite is

decreasing the dielectrophoretic force to a large extent. Thus, decreasing the permittivity of the solvent relative to the nanowires could promote assembly, performing assembly in ethanol, 2-methoxyethanol, and butanol. Assembly in each of these

63

solvents showed good positive dielectrophoresis (Figure 3-9), however the ethanol evaporated rather rapidly, not allowing much time for assembly to occur. As such, butanol was selected for the subsequent assemblies which led to good assembly in the test arrays which allowed individual nanowire rows to be measured (Figure 3-10). In an attempt to optimize the frequency for electrical assembly was swept the frequency from

100 Hz–10 MHz while observing the nanowires in solution. It appeared that a frequency of around 5 kHz performed well for the nanowires in the lower dielectric constant solvents, hence this frequency was used for assembly of the nanowires across the electrodes in the test array. Figure 3-10 shows successfully assembled nanowires across the test electrodes using 42 volts peak to peak @ 5 kHz in butanol post barium carbonate removal. The linear density of the as assembled nanowires was found to be approximately 19 NWs/mm across 27 rows of nanowires. After assembly upper electrical contacts were formed with the nanowires via lithography and deposition of upper electrical contacts via copper electroplating as shown in Figure 3-11.

The resultant linear density of the nanowire assembly post electrode deposition decreased to 3.6 NWs/mm. During deposition of the upper electrical contacts, this decrease in nanowire density can be attributed to the fact that some nanowires were not sufficiently long enough to be covered by the upper contacts, and loss of adhesion in the upper electrodes. While this density was sufficient to test the concept of using these nanowires to form a passive magnetoelectric magnetic field sensors and should allow for miniaturization, methods to produce a higher density of assembly will be the focus on ongoing research, to allow fabrication of devices with even smaller footprints. A higher density assembly could likely be achieved through the use of alternative solvents

64

or further tuning of the assembly frequency, or the implementation of a microfluidic channel92. To increase the proportion of nanowires which remain assembled across the

electrodes through the process of depositing upper electrical contacts the proportion of

the nanowires long enough to be covered by the electrodes could be increased. To do

so the electrospinning parameter space could be further explored to increase the

average length of the electrospun nanowires while maintaining a low variance in their

distribution. To help adhesion of the upper contacts the surface of the wafer could be

cleaned via carbon dioxide snow cleaning93 after photolithography and before electrode

deposition, between steps 2 and 3 in Figure 3-12.

3.3.3 Characterization

A sample of calcined nanowires (not assembled, but from the same batch used

for the assembly process) was measured by a VSM to confirm a ferrimagnetic

response. Figure 3-13 shows the hysteresis loop for the randomly oriented nanowires,

showing a saturation of 60 emu/g, a remanence of ~18 emu/g, and a coercivity of

~0.8 kOe. Figure 3-14 shows the C–V curves for an assembled row of nanowires. The

data for the nanowires shows a hysteretic behavior for the up sweep and down sweep,

confirming both electrical connectivity and a ferroelectric response due to the barium

titanate in the wires. For comparison the measured C–V curve for an empty

substrate did not exhibit a ferroelectric response.

Magnetoelectric measurements were performed (a) using the lock-in

technique90 with a DC bias field up to 8 kOe provided by an electromagnet and (b) using

only an AC field (zero DC bias) provided by a Helmholtz coil pair in conjunction with a

rotational measurement setup to explore and quantify the effect of electromagnetic

induction. In the rotational setup the nanowire array and electrical connections could be

65

rotated 360° within a set of Helmholtz coils to find the angle of no induction, which occurs when the loop of wire formed by the nanowire array and electrical connections is orthogonal to the applied field.

The magnetoelectric coefficients of a row of nanowires in the nanowire array were measured by the lock-in technique90, shown in Figure 3-2, as a function of bias

field at 200, 500, and 1000 Hz. Here, for all three frequencies, a zero bias

magnetoelectric effect is observed. Where previously observed, this effect has been

attributed to remanent magnetization in the magnetic phase94–97, which is consistent

with the remanence observed in the magnetic hysteresis loops of the nanowires. This

phenomenon can be thought of as a “self-biasing” effect98–101 , in that the remanent

magnetization of the wires creates an internal demagnetizing field that provides a bias

field to the magnetic phase.

Figure 3-14 also reveals a general increase in magnetoelectric coefficient with

frequency, which has been previously reported and can be attributed to changes in the

dielectric constant of the constituent phases as a function of frequency94,102. This likely

leads to a reduction in charge leakage through the magnetostrictive cobalt ferrite phase

with increasing frequency.

To obtain additional magnetoelectric measurements at zero-bias, a rotational

measurement setup to eliminate inductive effects was constructed and the zero-bias

nanowire array response was directly measured by an oscilloscope. This setup allows

for the angle of the array with respect to the applied magnetic field to be adjusted

systematically in order to explore the influence of electromagnetic induction in the

measurement. Specifically, this setup was used to (1) find the angle at which the normal

66

of the array and electrical connections were orthogonal to the applied magnetic field,

eliminating induction, and perform magnetoelectric measurements at this angle (2)

measure the effective magnetoelectric coefficients at ±90° from this angle, where

induction either maximally constructively or destructively interfered with the voltage

generated through the magnetoelectric effect. Figure 3-15a shows the the voltage

output of the nanowire array as a function of the angle of the nanowire array with

respect to a 1 kHz applied AC field. Here, it can be seen that there is an inductive

contribution which is angle dependent and an angle independent magnetoelectric

response. As expected the inductive contribution is sinusoidal and can be fit to find the

angle of zero inductive contribution φmin,ind = 177°. The results of direct

magnetoelectric measurement spectra can be seen in Figure 3-15b which compare the

effective magnetoelectric coefficients as measured at φmin,ind, an angle with induction

additive to the magnetoelectric effect φmin,ind + 90° and an angle at which induction

which opposes the magnetoelectric effect φmin,ind −90°. It is observed that the

maximum inductive contribution of the nanowire array is much smaller than the

magnetoelectric effect, 15% when rotated to an angle at which the maximum inductive

contribution occurred at 1 kHz.

Figure 3-15b shows the measured device response from 20 Hz to 10 kHz. Again,

a general increase in magnetoelectric effect with respect to increasing frequency is

found, and there is good agreement with the lock-in measurements (denoted by the red

squares). For the case of minimal induction, magnetoelectric coefficients with zero bias

at 200 Hz, 500 Hz, and 1 kHz were found to increase from 152 ± 13 to 267 ± 14 to

67

514 ± 27 mV cm−1 Oe−1. These values correspond to device sensitivities of 0.152 ± 0.013

to 0.267 ± 0.014 to 0.514 ± 0.027 mV Oe−1, at 200 Hz, 500 Hz, and 1 kHz, respectively.

3.4 Discussion

After completing the magnetoelectric measurements the magnetoelectric

coefficients from the nanowire rows in the arrays were compared to those found in the

literature for the barium titanate/cobalt ferrite system in thin film71 and bulk103 structures.

Here the nanowires were found to exhibit significantly larger magnetoelectric coefficients (514 ± 27 mV cm−1 Oe−1 at 1 kHz) compared to values previously observed

with this material system. For example, bulk barium titanate / cobalt ferrite systems

have been shown to demonstrate magnetoelectric coefficients of 50 mV cm−1 Oe−1 at 60

kHz65 Zhang et al. reported 104 mV cm−1 Oe−1 at 1 kHz for BTO-CFO single crystal thin

films71. The increased magnetoelectric coefficient observed in the nanowire arrays in

comparison to thin films seems reasonable based on theoretical study of substrate

clamping effects104 and PFM measurements made by Xie et al. on PZT-CFO

fibers15 and a comparison to PZT-CFO thin films94.

3.4. Conclusions

In conclusion, in this chapter a magnetic field sensor was successfully fabricated

via the assembly of arrays of magnetoelectric nanowires using methods that are readily

scalable, economical, and CMOS compatible. It was demonstrated that magnetoelectric

nanowires with controllable lengths can be prepared by tuning both the electrospinning

and calcination conditions and that dielectrophoretic assembly methods allow the

fabrication of functional arrays of magnetoelectric nanowires. The direct magnetoelectric

effect from the nanowire array was measured using the lock-in technique at 200, 500,

and 1000 Hz with bias fields between −8 and 8 kOe. The magnetoelectric coefficients of

68

the assembled nanowire arrays at zero-bias was also measured from 20 Hz to 10 kHz and the observed magnetoelectric coefficients were notably greater than bulk and thin film systems comprised of the same magnetoelectric material.

Additional improvements can be made to the electrospinning, electrical assembly, and upper electrode formation parameters to maintain reliable device fabrication in increasingly smaller electrode footprints, allowing even smaller devices to be manufactured. Furthermore, as all processing steps on the wafer were designed to be performed at low temperature these devices can be readily integrated with on chip signal processing components with the potential to further reduce device size.

69

Figure 3-1: The magnetoelectric sensing element consisting of magnetoelectric nanowires bridging an electrode gap suspended above a substrate

70

Figure 3-2. A) Schematic diagram of the lock-in magnetoelectric measurement setup. Here a large DC bias H-field can be applied to the nanowire arrays alongside an AC H-field. Lock-in amplifiers, phase-locked to the magnetic field signal are used to measure the voltage from the nanowire array. The low noise preamplifier was used to amplify the Hall probe signal to help maintain phase lock to the field. B) Schematic of the nanowire array showing the direction of the applied AC and DC fields in relation to the nanowire axis.

71

Figure 3-3. Schematic of the rotating magnetoelectric measurement setup, where the angle of the array with respect to the applied magnetic field can be adjusted to explore the effects of induction on the measured magnetoelectric coefficient.

72

Figure 3-4. A) Scanning electron micrograph of Janus barium titanate - cobalt ferrite (BTO - CFO) nanowires post salt calcination. These nanowires were electrospun at 2 kV cm-1 and calcined for 8 h at 1100 °C with a ramp rate of 10 °C min-1. B) Scanning electron micrograph of a single Janus BTO - CFO nanowire post assembly across the nanowire arrays, the two phases of the Janus nanowire are labelled here as BTO and CFO solely for illustrative purposes as in the micrograph it is uncertain which section of the nanowire corresponds to each phase.

73

Figure 3-5. A) Histogram and fitted lognormal distributions for nanowires electrospun at 2 kV cm-1 and 1.83 kV cm-1 with a constant calcination ramp rate of 10 °C min-1 showing an increasing nanowire length with decreasing electrospinning voltage. B) Histogram and fitted lognormal distributions for nanowires calcined at 25 °C min-1 and 10 °C min-1 with an electrospinning voltage of 2 kV cm-1 showing a decreasing nanowire length with increasing calcination temperature. C) Nanowire lengths and voltages from nanowires electrospun at 2 kV cm-1 and 1.83 kV cm-1 with a constant calcination ramp rate of 10 °C min-1 showing a positive correlation between nanowire diameter and length.

74

Figure 3-6. X-ray diffraction spectra of the barium titanate / cobalt ferrite Janus nanowires post calcination showing peaks indicative of tetragonal barium titanate and spinel cobalt ferrite.

75

Figure 3-7. Raman spectra of single phase barium titanate nanowires used to test whether barium carbonate (BCO) can be successfully removed with a dilute hydrochloric acid (HCl) wash. Absence of the barium carbonate peaks in the as HCl treated sampled demonstrates that this wash was successful in removing BCO impurities.

76

Figure 3-8. Mass magnetization curve indicating ferrimagnetism in the cobalt ferrite phase of unassembled Janus nanowires from the same batch as those used to fabricate the nanowire array.

77

Figure 3-9. Successful assembly of barium titanate nanofibers in water, post barium carbonate removal with a dilute HCl wash and suspension using citric acid and adjusting the pH to around 9, at 5 kHz and 20 Vpp.

78

Figure 3-10. Assembly of Janus nanofibers in A) ethanol, B) 2-methoxyethanol, and C) butanol at 5 kHz and 20 Vpp. While positive dielectrophoresis was observed with all three solvents, ethanol evaporated quickly leaving less time for assembly and butanol appeared to produce slightly better assembly than 2- methoxyethanol.

79

Figure 3-11. Assembly of Janus nanowires in butanol at 5 kHz and 42 Vpp in the test array with a linear density of 19 NWs mm−1.

80

Figure 3-12. Formation of upper electrical contacts across the nanowires via A) patterning of Ti/Cu electrode patterns via sputtering and lift-off; followed by spin coating blanket layer of LOR resist; and assembly of nanowires; B) spin coating and patterning AZ1512 photoresist to expose ends of the nanowires; C) electroplating copper to make electrical contacts with nanowire; and D) stripping of the remaining photoresist.

81

Figure 3-13. Capacitance–voltage (C–V) measurement from the Janus nanowire array demonstrating ferroelectricity in the barium titanate phase. The test signal level was set to 100 mV at 750 kHz.

82

Figure 3-14. Plot of the magnetoelectric coefficients of a row in the barium titanate / cobalt ferrite Janus nanowire array measured using the lock-in technique as a function of magnetic bias field at 200, 500, and 1000 Hz.

83

Figure 3-15. A) Effect of angle of the Janus nanowire array angle on the observed voltage generated by the array with a 4.9 Oe applied AC field at 1 kHz, as well as the fitted sine function used to find the angle of zero induction φ0. B) Measured magnetoelectric coefficients as a function of frequency from the barium titanate / cobalt ferrite nanowire arrays at angles of approximately no inductive effects, φmin,ind = 177°, the angle of maximum constructive induction, φmin,ind + 90°, and maximum destructive induction, φmin,ind + 90°.

84

CHAPTER 4 ULTRA LOW POWER CURRENT SENSOR UTILIZING MAGNETOELECTRIC NANOWIRES

The fabrication and characterization of an ultra low power current sensor utilizing

magnetoelectric lead zirconate titanate / nickel zinc ferrite nanowires is presented. Low-

cost, low-temperature, and post-CMOS-compatible fabrication methods were utilized in

the fabrication of a nanowire array. This array was electrically and mechanically flip-chip

bonded onto a printed circuit board (PCB) having a current trace. The PCB also

contained a low-power amplification circuit to provide buffering and a gain of 10.

Characterization of the sensor up to 70 mA showed a sensitivity of 3.24 mV/mA,

sensitivity error of 1.16%, nonlinearity of 4%, noise floor of < 2 mA, and noise density of

8.4 nA Hz-½ at 1kHz. Lastly the only power consumption necessary for device operation was the power required for a low power op amp, 0.225 mW.

4.1 Introduction

Typically, current sensors depend either on measurement of the voltage across a

current-sense resistor or sensing the magnetic field created from current flowing through a wire or current trace. The former offers low cost, small size, and simplicity105,

but often requires a tradeoff between using small resistor values to avoid unwanted

power loss, versus requiring large amplification gain to generate sufficiently large

voltages for downstream electronics. Additionally, when using a current-sense resistor,

the rest of the circuit is also generally not isolated from the electronics detecting the

current flow.

In contrast to resistor-based methods, non-intrusive methods rely on detecting

the magnetic field generated by a current carrying wire. This can include Faradaic

induction (e.g., a pickup coil or current transformer), or direct measurement of the stray

85

magnetic field, commonly using Hall effect sensors or magnetoresistive sensors.

However, sensors using Faraday’s law of induction such as Rogowski coils are large and not easily incorporated with signal conditioning circuitry in a single chip package.

Hall sensors and magnetoresistive sensors can be integrated with on chip signal processing circuitry, however both of these magnetic field sensors require power for field measurement.

Magnetoelectrics, materials that exhibit an electric polarization in response to an applied magnetic field offer the potential to create compact, ultra-low-power current sensors with on chip signal processing capability, overcoming the limitations of existing current sensing technologies. To date, the magnetoelectric materials with the greatest magnetoelectric coefficients (dE/dH) are biphasic strain mediated magnetoelectric composites comprising a magnetostrictive phase and a piezoelectric phase. The mechanism for voltage generation in these materials is based on strain transfer between the magnetostrictive phase, which exhibits a strain in response to an applied magnetic field, and a piezoelectric, which when strained generates a voltage. Since this is a strain-based phenomena, thin-film multilayer magnetoelectric structures fabricated on semiconductor substrates are limited by mechanical clamping from the underlying substrate. In contrast, unclamped freestanding nanomaterials have been predicted to exhibit orders of magnitude greater magnetoelectric effect, and this has been experimentally confirmed through indirect magnetoelectric measurements72. In Chapter

3 methods to create functional devices, including magnetic field sensors, from these

nanomaterials and in addition direct measurements of their magnetoelectric coefficients

were reported106,107.

86

Among strain based magnetoelectrics those containing lead zirconate titanate exhibit comparably large magnetoelectric coefficients due to the high piezoelectric coefficient of the lead zirconate phase. Nickel zinc ferrite and lead zirconate titanate have previously been electrospun separately and in conjunction to form biphasic fibers and have previously demonstrated high magnetoelectric coefficients as a composite107–

110. For these reasons the lead zirconate titanate / nickel zinc ferrite system was chosen

for the ultra low power current sensor.

Here, an ultra low power, non-invasive, current sensor utilizing magnetoelectric

nanowires is fabricated and characterized. This chapter discusses the design and

fabrication and characterization of a magnetoelectric nanowire-based current sensor.

Lastly, key performance metrics of these current sensors will be presented. The

methods developed here can be readily expanded to fabricate a wide range of nanowire

based devices.

4.2 Design

In the current sensor developed here, the fundamental sensing element consists

of magnetoelectric nanowires bridging an electrode gap suspended above a substrate

as previously shown in Figure 3-1. When the nanowire is exposed to the magnetic field

generated by a current carrying wire the magnetostrictive phase undergoes a shape

change, resulting in a strain that is then transferred to the piezoelectric phase, thereby

creating a voltage across the electrodes. This voltage can then be used to determine

the magnetic field strength and thus the current flowing through the wire. The larger the

magnetoelectric effect, the greater the voltage generated for a given magnetic field, and

the greater sensitivity of the magnetoelectric sensing element before amplification.

87

Thus, the larger magnetoelectric effect of the nanowires as compared to thin films could offer a greater sensitivity device.

In a magnetic-field-based current sensor, either a magnetic core can be incorporated or a core-less design can be used. With a magnetic core design, the current carrying wire of interest is surrounded by a magnetic core with a gap where the sensing element is placed. The main advantage to this design is that it enhances the signal from the current source while minimizing the effects of extraneous magnetic fields. Disadvantages of a design employing a magnetic core include higher parasitic inductance on the current trace, potential heating of the magnetic material, higher material costs, and more complex fabrication. On the other hand, core-less designs lack the magnetic core and the sensing element is simply placed nearby the current carrying wire. The main disadvantage to the core-less design is that extraneous magnetic fields are not filtered out, however it is a simpler design to prototype and generally has a lower cost of manufacture. Due to these considerations, a core-less design was used for the initial prototype of the magnetoelectric nanowire current sensor.

Since the magnetic field generated by a current carrying element decreases with distance, it is advantageous to place the magnetic field sensing element close to the current element of interest. To position the magnetoelectic nanowires closer to the current, a flip-chip approach is used, where a nanowire array is flip-chip-bonded to a printed circuit board (PCB) with a current trace running parallel to the axis of the nanowires. A schematic diagram of the nanowire array containing magnetoelectric nanowires self-assembled onto patterned electrodes is given in Figure 4-1. The flip chip

design is shown in Figure 4-2. Additionally, the signal from the nanowires is buffered via

88

a low power voltage mode amplifier (Analog Devices AD8541) with a gain of 10 and a

10 MΩ resistor in parallel with the nanowires to create a DC bias path for the terminal and control the low frequency roll off (Figure 4-3).

4.3 Fabrication

The fabrication process is broken down into three major steps:(1) synthesis of

the lead zirconate titanate/nickel zinc ferrite magnetoelectric nanowires forming the

sensing element (2) formation of the sensing element via assembly and electrical

connection of the nanowires across electrodes and (3) integration of the

magnetoelectric nanowire arrays into PCB boards with a current carrying trace and a

discrete amplification circuit to buffer the signal and provide the desired gain of 10. In

each of these fabrication steps different characterization techniques are used to validate

the success of each step. A process flow diagram is given in Figure 4-4.

4.3.1 Nanowire Synthesis

First, lead zirconate titanate and nickel zinc ferrite sol gel precursor solutions were prepared. A lead zirconate titanate ceramic solution was prepared by dissolving 2 g lead acetate trihydrate in 2.5 ml ethanol and 2.5 ml acetic acid under constant stirring.

After 1 h 0.31 ml of titanium isopropoxide and 0.82 ml zirconium n-butoxide were added.

Simultaneously, a polymer solution was prepared by dissolving 0.72 g

polyvinylpyrrolidone in 2.61 ml ethanol and 2.18 ml dimethylformamide. After an

additional hour of stirring, the ceramic solution was added dropwise to the polymer

solution under constant stirring. Similarly, a nickel zinc ferrite solution was prepared by

dissolving 0.980 g ferric nitrate, 0.226 g nickel nitrate, and 0.322 g zinc nitrate in 4.5 ml

ethanol under constant stirring. After 1 h 0.81 ml of acetylacetone was added.

Simultaneously a polymer solution was prepared by dissolving 0.72 g

89

polyvinylpyrrolidone in 2.61 ml of ethanol and 2.18 ml of dimethylformamide. After stirring for an additional hour, the ceramic solution was added dropwise to the polymer solution under constant stirring107.

Both solutions were co-electrospun from side by side syringe needles to form

Janus nanofibers, with one semi-cylinder of the fiber consisting of lead zirconate titanate

and the other consisting of nickel zinc ferrite. The ceramic/polymer nanofibers were then

calcined at 700°C; burning off the polymer and breaking them along their length,

forming nanowires. The high temperature heat treatment also crystallized the

amorphous as-spun ceramic. Post calcination the nanowires were imaged via scanning

electron microscopy to verify Janus morphology (Figure 4-5). Subsequently their crystal

structure was analyzed via X-ray diffraction (Figure 4-6) with Rietveld Refinement.

From the Rietveld refinement inverse spinel nickel zinc ferrite was observed.

Evidence of tetragonal and rhombohedral lead zirconate titanate phases was also

found, indicating that the lead zirconate titanate was near the morphotropic phase

boundary111. Crystalline impurities were not observed beyond trace amounts. The wt. %

ratio of lead zirconate titanate to nickel zinc ferrite was found to be 69:31. Specifically

the estimated weight percents of inverse spinel nickel zinc ferrite, tetragonal lead

zirconate titanate, and rhombohedral lead zirconate titanate were estimated to be 31 wt.

%, 47 wt. %, and 22 wt. %, respectively. The key agreement indices were =

𝑅𝑅𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 5.02, = 6.91, = = = 1.90 2 . Briefly, Rexpected is a measure of the 𝑅𝑅𝑤𝑤𝑤𝑤 2 2 𝑅𝑅𝑤𝑤𝑤𝑤 𝐺𝐺𝐺𝐺𝐺𝐺 𝑅𝑅𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 𝜒𝜒 noisiness of the measured spectra, Rwp is the difference between the experimentally

measured profile and a simulated spectra from the fitted crystal structures weighted

more heavily around the peaks, and goodness of fit (GOF) is a measure of how well the

90

spectra was fit from the refinement. The reasonably low Rexpected indicates a quality scan while a low Rwp and GOF indicates that the spectra was fit well. These factors coupled with the dissimilarity of the peaks in the lead zirconate titanate and nickel zinc ferrite structures indicate that the estimated relative wt. % ratios of these two compounds are accurate. Due to an overlap in many of the strong peaks in the rhombohedral and tetragonal lead zirconate titanate phases the estimated relative weight percent of these phases may differ more from their actual values. Additionally, it has been argued that the rhombohedral phase in PZT may be more accurately modelled with a monoclinic phase112. In the analysis a rhombohedral phase was chosen

as a larger chi square value was obtained using a monoclinic phase, = 1.96, 2 however due to the nonlinearity of the Rietveld refinement process it is𝜒𝜒 difficult to

assess whether or not this was by chance. As such the debate on the nature of the

morphotropic phase boundary should be further explored with other techniques,

possibly three dimensional rotation electron diffraction113. However, Rietveld refinement

still provides evidence that the PZT is at the morphotropic phase boundary, and thus the

resultant device should benefit from the enhancement of the piezoelectric properties of

PZT at this morphotropic phase boundary.

Raman spectroscopy was employed to investigate any surface impurities of the

nanowires. Surface impurities can greatly impact electrical assembly by changing the

electrical properties of the material interfacing with the surrounding solution, and

amorphous passive films may form on the surface of certain ceramics and metals. While

these compounds might not be identified using X-ray diffraction, From Raman spectroscopy on the lead zirconate titanate/nickel zinc ferrite nanowires (Figure 4-7) an

91

impurity phase was found which was attributed to lead oxide114. A dilute HCl wash was

thus used to remove any lead oxide from the surface of nanowires prior to assembly

(Figure 4-7).

To verify ferrimagnetism in the lead zirconate titanate/nickel zinc ferrite

nanowires, the mass magnetization curve, M-H, was obtained using a vibrating sample

magnetometer (VSM). For the M-H curve the field was swept from 22 kOe to -22 kOe,

then swept back up to 22 kOe at room temperature. Figure 4-8 shows the hysteresis

loop for the randomly oriented nanowires, as well as an inset from -0.2 kOe to 0.2 kOe.

The maximum magnetization observed in the M-H curve was found to be 8 emu/g, with

a remanence 0.6 emu/g, and the coercivity 50 Oe.

4.2.3 Nanowire Array and Current Sensor Fabrication

The interdigitated electrodes diagrammed in Figure 4-1, consisting of a seed layer of 200 Angstroms of titanium and 600 Angstroms of gold, were fabricated via liftoff on a silicon wafer with a 500 nm thermal oxide layer. Subsequently the substrate was covered with 210 nm of lift-off resist (LOR). The calcined and HCl washed lead zirconate titanate/nickel zinc ferrite nanowires were then assembled across the electrodes utilizing the dielectrophoretic force. In this process the wires were suspended in butanol, and this butanol/nanowire solution was pipetted onto the electrode array while an AC voltage was applied across the pairs of electrodes. To investigate the effects of frequency on the dielectrophoretic force driving assembly, assembly was performed at 200 Hz, 2 kHz, and 20 kHz.

Assembly at the lowest frequency, 200 Hz, resulted in negative dielectrophoresis where the nanowires were repelled from the highest electric field gradient and deposited across the middle of the electrodes (Figure 4-9A). Assembly at the highest frequency,

92

20 kHz, resulted in positive dielectrophoresis where the nanowires were attracted to the regions of the highest electric field gradient and assembled across the interdigitated electrodes (Figure 4-9C). At the intermediate frequency of 2 kHz both positive and negative dielectrophoresis was observed (Figure 4-9B).

To form electrical connections to the nanowires from the underlying electrodes

photolithography and electroplating were used. 2 microns of AZ1512 positive

photoresist was deposited. A pre-exposure soft bake was performed at 112 C for 2

minutes. Following this, the wafer was exposed with a dose of 220 mJ, with a mask

designed to expose the edges of the electrodes where the ends of the nanowires are

aligned post electrical assembly. Post exposure the photoresist was developed for 75

seconds in AZ300MIF developer, followed by a deionized water rinse and nitrogen blow

dry, leaving the ends of the nanowires exposed (Figure 4-10).

For copper electroplating over the ends of the nanowires to make electrical

contact the substrate was submerged in a copper(II) sulfate solution along with a small

copper sheet anode. A potential of +9 V was applied to the copper anode while the

substrate was held at 0 V. During the electroplating process the current was limited to

0.2 A.

Separately a test array was fabricated as was used in Chapter 3 for the

measurement of individual rows of barium titanate / cobalt ferrite arrays. The lead

zirconate titanate/nickel zinc ferrite nanowires were assembled across this array in the

same manner as was outlined above in butanol at 20 kHz. The results of the zero bias

measurements utilizing the rotational measurement setup described in Chapter 3 is

given in Figure 4-11. Here magnetoelectric coefficients with zero bias of 485 ± 93 mV

93

cm-1 Oe-1, 1114 ± 132 mV cm-1 Oe-1, and 2404 ± 678 mV cm-1 Oe-1, were observed at

200 Hz, 500 Hz, and 1 kHz, respectively. These translate to an array sensitivities of

0.485 ± 0.093 mV Oe-1, 1.114 ± 0.132 mV Oe-1, and 2.404 ± 0.678 mV Oe-1 at 200 Hz,

500 Hz, and 1 kHz.

After the nanowire array was fabricated it was flip chip epoxied onto a PCB board

above a current carrying trace. Discrete circuit components were then soldered onto the

other side of the PCB to buffer and amplify the signal from the nanowires. The final

device is shown in Figure 4-12.

4.4 Current Sensor Characterization

To characterize the current sensor device, currents from 12 to 70 mA were

applied through the current trace using a Hewlett Packard 33120 function generator and

Pyramid PB-101 amplifier amplifier with a resistor in series with the device to allow for

the estimation of applied current through the voltage drop across the resistor. The

voltage waveforms from the sensor output and the voltage drop across the resistor were

then obtained using a Tektronix DPO2004 Oscilloscope (Figure 4-13). It can be seen

here that there is a strong signal in the sensor output at the frequency of the applied

current but also a 60 Hz signal from extraneous sources. This extraneous 60 Hz signal

could be reduced in comparison to the signal of interest by incorporating a magnetic

core in the sensor design as previously noted. The magnitude of the voltage output from

the current sensor due to the applied and extraneous signal was fitted using Python and

the Broyden–Fletcher–Goldfarb–Shanno (BFGS) method and given in Figure 4-14.

From this the sensitivity was calculated to be 3.24 mV/mA, the sensitivity error was

1.16%, and the nonlinearity was found to be 4%. From the residual of the fits the noise

was calculated to be below 2 mA RMS. Noise density measurements from 0 to 800 Hz

94

were made using a SR780 dynamic signal analyzer with a triple faraday cage setup

Figure 4-15. It can be seen here that even with the triple Faraday cage the 60 Hz noise or mains hum was the main contribution to the noise as well as multiples of this frequency. The noise density at 1 kHz was also measured and found to be 8.4 nA Hz-½.

4.5 Conclusion

Here, the integration of an array of lead zirconate titanate/nickel zinc ferrite

nanowires into a functional current sensing discrete circuit was demonstrated in a flip

chip design. This device offers ultra low power current sensing capability with low

sensitivity error, low noise, and moderate nonlinearity in a small package. Further

improvements on the design could include on chip signal processing circuitry as the

fabrication steps used in the fabrication of the nanowire arrays are low temperature and

CMOS compatible.

95

Figure 4-1: Electrode design for the current sensor (A) zoomed in view showing the electrode gaps and parallel electrodes (B) single array view where the top and bottom pads can be connected to read out the electrical signal from the nanowires (C) wafer level view showing the electrical contacts at the bottom of the wafer for AC electrical assembly.

96

Figure 4-2: Flip chip design of the ultra low power current sensor (A) the front of the PCB board containing the amplification and buffering circuitry, (B) back of PCB board showing the current trace from which the nanowires will measure the current through magnetic field measurement, (C) diagram illustrating the flip chip bonding of the nanowires on the side of the current trace, (D) diagram illustrating the scale of the designed sensor.

97

Figure 4-3: Circuit diagram of the magnetic field sensor.

98

Figure 4-4: Process flow diagram for fabricating the current sensor.

99

Figure 4-5: Scanning electron micrographs of lead zirconate titanate / nickel zinc ferrite nanowires before calcination (A,B), uncalcined, and post calcination (C,D), calcined.

100

Figure 4-6: X-Ray Diffraction Spectra of the lead zirconate titanate (PZT) and nickel zinc ferrite (NZF) nanowires, showing the characteristic peaks of the tetragonal and inverse spinel structures, respectively.

101

Figure 4-7: Raman spectra of lead zirconate titanate/nickel zinc ferrite demonstrating the removal of a lead oxide impurity phase with a dilute HCl wash.

102

Figure 4-8: Mass magnetization curve indicating ferrimagnetism in the nickel zinc ferrite phase of unassembled Janus nanowires from the same batch as those used to fabricate the nanowire array showing remnant magnetization in the nanowires.

103

Figure 4-9: Effects of frequency on electrical assembly of lead zirconate titanate/nickel zinc ferrite nanowires where (A) negative dielectrophoresis (nDEP) was observed at 200 Hz, (B) a mix of nDEP and positive dielectrophoresis (pDEP) was observed at 2 kHz and (C-D) pDEP was observed at 20 kHz.

104

Figure 4-10: Nanowires assembled across interdigitated electrodes post electroplating.

105

Figure 4-11. Measured magnetoelectric coefficients as a function of frequency from 6 individual rows lead zirconate titanate / nickel zinc ferrite nanowire arrays at angles of approximately no inductive effects.

106

Figure 4-12: Flip chip fabrication of of the current sensor IC, (A) electrode array containing the assembled nanowires, (B) top of the IC showing the buffering and amplification circuitry, (C) bottom view of the device showing the wafer containing the nanowire array flip chip bonded to the PCB board, (D) side view of the current sensor.

107

Figure 4-13: Calculated applied current through a current sense resistor and raw output voltage waveforms from the nanowire based current sensor at with applied currents of (A) 24 mA peak at 200 Hz, (B) 71 mA peak at 200 Hz, (C) 24 mA peak at 1 kHz, and (C) 71 mA peak at 1 kHz.

108

Figure 4-14: Fitted waveform amplitudes from the mains hum and applied signal where (A) the 60 Hz extraneous signal remains constant and (B) the sensor output voltage at the frequency of the applied current varies linearly with applied voltage.

109

Figure 4-15: Noise density as a function of frequency of the current sensor as measured in a triple Faraday cage.

110

CHAPTER 5 CONCLUSIONS

In this dissertation the fabrication of novel ultra low power magnetic field sensing and current sensing ICs utilizing magnetoelectric nanowires with applications in automobiles and electrical systems was demonstrated. The main advantage to using composite strain mediated magnetoelectric nanowires is that the strain was not clamped by an underlying substrate allowing greater magnetoelectric coefficients/sensitivities to be realized. Another advantage is that the optimal calcining temperature for magnetoelectric performance can be used while maintaining CMOS compatibility as this high temperature processing step occurs off substrate prior to assembly of the nanowires into the final device. These devices as of yet had not been actualized due to challenges in the synthesis and fabrication process.

The first obstacle was the conceptualization of a device architecture and fabrication process flow which allowed for the incorporation of magnetoelectric nanowire based device. This was achieved through a combination of electrospinning with a fast calcination ramp rate to break the as spun fibers into shorter nanowires, the fabrication of parallel electrodes and subsequent electrical assembly across said electrodes, and connection of the nanowires to the underlying electrodes through photolithography and electroplating. Each of these steps required optimization to produce the final device.

The electrospinning process and tuning of electrospinning parameters was presented and discussed. Substrate fabrication and assembly of the nanowires onto the substrate via dielectrophoresis was covered. The final fabrication step of making electrical contacts with the nanowires is discussed. In each step the optimization/selection of processing parameters is covered. Reliable electrospinning

111

solutions of barium titanate, barium titanate/cobalt ferrite, and lead zirconate titanate/nickel zinc ferrite nanowires are given in Chapters 2, 3, and 4 respectively. It was found that both magnetoelectric systems covered here suffered from surface impurities which hindered electrical assembly detectable with Raman spectroscopy which could be removed with a dilute HCl wash. After removal of these surface impurities the nanowires could be reliably assembled at 20 kHz in a low permittivity solvent including ethanol, isopropanol, 2-methoxyethanol, and butanol.

The first direct measurement of the magnetoelectric coefficient in magnetoelectric nanowires was obtained using the lock in technique and in an induction minimizing rotational setup. The results of these measurements demonstrate the predicted increase in magnetoelectric effect through the elimination of substrate clamping present in thin films as discussed in Chapter 2. The first ultra-low power current sensors utilizing magnetoelectric nanowires were assembled and characterized in Chapter 3. These current sensors exhibit low power consumption, low sensitivity error, 1.16%, and reasonable nonlinearity, 4%.

Future Work. In the interest of practical implementation ongoing research should focus on the ruggedization of the magnetic field sensors, as well as their performance in the specific application of rotational motor speed sensors. Effects of temperature and thermal cycling on the magnetoelectric effect will also be studied. Optimal ratios of the piezoelectric to magnetostrictive materials in the lead zirconate titanate/nickel zinc ferrite system is currently being studied. Additionally, DC magnetic field sensing utilizing magnetoelectric nanowires could also be explored.

112

APPENDIX DIELECTROPHORETIC FORCE SIMULATIONS

In Chapter 1 a simulation of the electric field gradient across parallel electrodes was given showing the high field areas were located at the edges of the electrodes

Figure 1-3. The dielectrophoretic force was given in Equation 1-1 where the real part of the Claussius Mossotti factor determines the direction of the dielectrophoretic force.

This factor is dependent on the dielectric constant and conductivity of the particle/nanowire being assembled and the medium in which it is assembled as well as the frequency of the AC electric field used in the assembly. A positive Claussius

Mossotti factor indicates positive dielectrophoresis or an attraction of the particles/nanowires to the high electric field gradient region, and a negative Claussius

Mossotti factor indicates the particles/nanowires will experience a force towards the low electric field gradient regions. Here a positive dielectrophoretic force is desired as this force will attract the ends of the nanowires toward the edges of the electrode gap and result in the desired device structure of the nanowires bridging the gap. Chapters 2 and

3 experimentally explore the use of various solvents and assembly frequencies on the dielectrophoretic force as motivated by simulations of the Claussius Mossotti factor here.

Permittivity and conductivity values were obtained from various literature on material systems and microstructures reasonably close to the barium titanate115,116, cobalt ferrite117, lead zirconate titanate118,119, and nickel zinc ferrite120 assembled in this dissertation. The permittivity and conductivity of common solvents including water121,122, isopropanol123, ethanol121,122, butanol121,122, and 2-methoxyethanol124 were also gathered from literature.

113

The real part of the Claussius Mossotti factor is given for barium titanate, cobalt ferrite, lead zirconate titanate, and nickel zinc ferrite in Figure 6-1, Figure 6-2, Figure 6-

3, and Figure 6-4, respectively. Recall a positive Claussius Mossotti factor is desired as it is indicative of positive dielectrophoresis and thus a force on the nanowires directing them across the high field regions at the edges on the electrode gaps. This desired positive Claussius Mossotti factor is present in both piezoelectrics, barium titanate and lead zirconate titanate, in all solvents at high frequencies. Thus, assembly in water appears possible with these materials. The magnetostrictive cobalt ferrite and nickel zinc titanate, however, are predicted to experience weak negative dielectrophoresis in water at these frequencies. As such, a lower permittivity solvent such as butanol may be necessary.

114

Figure A-1: Real part of the Claussius Mossotti Facter as a function of frequency in barium titanate predictive of positive dielectrophoresis in each of the solvents at high frequencies (2-20 kHz).

115

Figure A-2: Real part of the Claussius Mossotti Facter as a function of frequency in cobalt ferrite predictive of weak positive dielectrophoresis with the low permittivity solvent butanol at high frequencies (10-20 kHz).

116

Figure A-3: Real part of the Claussius Mossotti Facter as a function of frequency in lead zirconate titanate predictive of positive dielectrophoresis in each of the solvents at high frequencies (2-20 kHz).

117

Figure A-4: Real part of the Claussius Mossotti Facter as a function of frequency in nickel zinc ferrite predictive of weak positive dielectrophoresis in butanol and 2-methoxyethanol at high frequencies (10-20 kHz).

118

LIST OF REFERENCES

1. Zhu, X. & Meng, Z. Operational principle, fabrication and displacement characteristics of a functionally gradient piezoelectric ceramic actuator. Sensors and Actuators A: Physical 48, 169–176 (1995).

2. Shung, K. K., Cannata, J. M. & Zhou, Q. F. Piezoelectric materials for high frequency medical imaging applications: A review. J. Electroceramics 19, 141– 147 (2007).

3. Mason, W. P. & Wick, R. F. A barium titanate transducer capable of large motion at an ultrasonic frequency. J. Acoust. Soc. Am. 23, (1951).

4. Chen, L. S., Fu, S. L. & Huang, K. D. Barium titanate-based capacitors buried into ceramic substrates. Japanese J. Appl. Physics, Part 2 Lett. 37, (1998).

5. Yuh, J., Nino, J. C. & Sigmund, W. M. Synthesis of barium titanate (BaTiO3) nanofibers via electrospinning. Mater. Lett. 59, 3645–3647 (2005).

6. Mimura, K., Moriya, M., Sakamoto, W. & Yogo, T. Synthesis of BaTiO3 nanoparticle/poly(2-hydroxyethyl methacrylate) hybrid nanofibers via electrospinning. Compos. Sci. Technol. 70, 492–497 (2010).

7. He, Y. et al. Humidity sensing properties of BaTiO3 nanofiber prepared via electrospinning. Sensors Actuators B Chem. 146, 98–102 (2010).

8. McCann, J. T., Chen, J. I. L., Li, D., Ye, Z.-G. & Xia, Y. Electrospinning of polycrystalline barium titanate nanofibers with controllable morphology and alignment. Chem. Phys. Lett. 424, 162–166 (2006).

9. Wang, L., He, Y., Hu, J., Qi, Q. & Zhang, T. DC humidity sensing properties of BaTiO3 nanofiber sensors with different electrode materials. Sensors Actuators B Chem. 153, 460–464 (2011).

10. Nan, C. W., Bichurin, M. I., Dong, S., Viehland, D. & Srinivasan, G. Multiferroic magnetoelectric composites: Historical perspective, status, and future directions. J. Appl. Phys. 103, (2008).

11. Bichurin, M. I., Petrov, V. M. & Srinivasan, G. Low-frequency magnetoelectric effects in ferrite–piezoelectric nanostructures. J. Magn. Magn. Mater. 321, 846– 849 (2009).

12. Spaldin, N. A. & Fiebig, M. Materials science. The renaissance of magnetoelectric multiferroics. Science (80-. ). 309, 391–392 (2005).

13. Yan, L. et al. Review of magnetoelectric perovskite–spinel self-assembled nano- composite thin films. J. Mater. Sci. 44, 5080–5094 (2009).

119

14. Ma, Y. G., Cheng, W. N., Ning, M. & Ong, C. K. Magnetoelectric effect in epitaxial Pb(Zr0.52Ti0.48)O3∕La0.7Sr0.3MnO3 composite thin film. Appl. Phys. Lett. 90, 152911 (2007).

15. Xie, S.-H., Liu, Y.-Y. & Li, J.-Y. Synthesis, microstructures, and magnetoelectric couplings of electrospun multiferroic nanofibers. Front. Phys. 7, 399–407 (2011).

16. Wu, Y. & Yang, P. Direct Observation of Vapor−Liquid−Solid Nanowire Growth. J. Am. Chem. Soc. 123, 3165–3166 (2001).

17. Berrier, S., Li, D., Solomon, V., Bauer, M. & Li, C. Synthesis of Silicon Oxide Nanowires by Chemical Vapor Deposition. Microsc. Microanal. 20, 1990–1991 (2014).

18. Qin, L., Park, S., Huang, L. & Mirkin, C. A. On-wire lithography. Science (80-. ). 309, 113–115 (2005).

19. Taylor, G. Electrically Driven Jets. Proc. R. Soc. A Math. Phys. Eng. Sci. 313, 453–475 (1969).

20. Doshi, J. & Reneker, D. H. Electrospinning process and applications of electrospun fibers. J. Electrostat. 35, 151–160 (1995).

21. Teo, W.-E., Inai, R. & Ramakrishna, S. Technological advances in electrospinning of nanofibers. Sci. Technol. Adv. Mater. 12, 13002 (2011).

22. Yu, M. et al. Recent advances in needleless electrospinning of ultrathin fibers: From academia to industrial production. Macromol. Mater. Eng. 302, 1700002 (2017).

23. Reneker, D. H. & Yarin, A. L. Electrospinning jets and polymer nanofibers. Polymer (Guildf). 49, 2387–2425 (2008).

24. Reneker, D. H., Yarin, A. L., Fong, H. & Koombhongse, S. Bending instability of electrically charged liquid jets of polymer solutions in electrospinning. J. Appl. Phys. 87, 4531–4547 (2000).

25. Amariei, N., Manea, L. R., Bertea, A. P., Bertea, A. & Popa, A. The influence of polymer solution on the properties of electrospun 3D nanostructures. in IOP Conference Series: Materials Science and Engineering 209, 12092 (IOP Publishing, 2017).

26. Wendorff, J. H., Agarwal, S. & Greiner, A. Electrospinning: materials, processing, and applications. (John Wiley & Sons, 2012).

27. Koski, a., Yim, K. & Shivkumar, S. Effect of molecular weight on fibrous PVA produced by electrospinning. Mater. Lett. 58, 493–497 (2004).

120

28. Gupta, P., Elkins, C., Long, T. E. & Wilkes, G. L. Electrospinning of linear homopolymers of poly(methyl methacrylate): exploring relationships between fiber formation, viscosity, molecular weight and concentration in a good solvent. Polymer (Guildf). 46, 4799–4810 (2005).

29. Katta, P., Alessandro, M., Ramsier, R. D. & Chase, G. G. Continuous Electrospinning of Aligned Polymer Nanofibers onto a Wire Drum Collector. Nano Lett. 4, 2215–2218 (2004).

30. Dai, Y., Liu, W., Formo, E., Sun, Y. & Xia, Y. Ceramic nanofibers fabricated by electrospinning and their applications in catalysis, environmental science, and energy technology. Polym. Adv. Technol. 22, 326–338 (2011).

31. Li, D. & Xia, Y. Direct Fabrication of Composite and Ceramic Hollow Nanofibers by Electrospinning. Nano Lett. 4, 933–938 (2004).

32. Baji, A. et al. One-dimensional multiferroic bismuth ferrite fibers obtained by electrospinning techniques. Nanotechnology 22, 235702 (2011).

33. Li, D. & Xia, Y. Fabrication of Titania Nanofibers by Electrospinning. Nano Lett. 3, 555–560 (2003).

34. Roh, K., Martin, D. C. & Lahann, J. Biphasic Janus particles with nanoscale anisotropy. Nat. Mater. 4, 759–63 (2005).

35. Gupta, P. & Wilkes, G. L. Some investigations on the fiber formation by utilizing a side-by-side bicomponent electrospinning approach. Polymer (Guildf). 44, 6353– 6359 (2003).

36. Starr, J. D. & Andrew, J. S. Janus-type bi-phasic functional nanofibers. Chem. Commun. (Camb). 49, 4151–3 (2013).

37. Andrew, J. S., Starr, J. D. & Budi, M. a. K. Prospects for nanostructured multiferroic composite materials. Scr. Mater. 74, 38–43 (2014).

38. Li, D., Wang, Y., Xia, Y. & Uni, V. Electrospinning of Polymeric and Ceramic Nanofibers as Uniaxially Aligned Arrays. (2003).

39. Starr, J. D. & Andrew, J. S. A route to synthesize multifunctional tri-phasic nanofibers. J. Mater. Chem. C 1, 2529 (2013).

40. Benaglia, T., Chauveau, D., Hunter, D. & Young, D. mixtools: An R Package for Analyzing Finite Mixture Models. Journal of Statistical Software. J. Stat. Softw. 32, 1–29 (2009).

41. Yarin, a. L., Kataphinan, W. & Reneker, D. H. Branching in electrospinning of nanofibers. J. Appl. Phys. 98, 64501 (2005).

121

42. Malašauskiene, J. & Milašius, R. Mathematical analysis of the diameter distribution of Electrospun nanofibres. Fibres Text. East. Eur. 83, 45–48 (2010).

43. Gevorkyan, A. G. et al. Branching effect and morphology control in electrospun PbZr0.52Ti0.48O3 nanofibers. J. Mater. Res. 29, 1721–1729 (2014).

44. Deitzel, J. ., Kleinmeyer, J., Harris, D. & Beck Tan, N. . The effect of processing variables on the morphology of electrospun nanofibers and textiles. Polymer (Guildf). 42, 261–272 (2001).

45. Collins, R. T., Jones, J. J., Harris, M. T. & Basaran, O. a. Electrohydrodynamic tip streaming and emission of charged drops from liquid cones. Nat. Phys. 4, 149– 154 (2008).

46. Malkin, A. Y. & Isayev, A. I. Rheology - Concepts, Methods, & Applications. ChemTec Publishing (ChemTec Publishing, 2006).

47. Massey Jr., F. J. The Kolmogorov-Smirnov Test for Goodness of Fit. J. Am. Stat. Assoc. 46, 68–78 (1951).

48. O’Brien, S., Brus, L. & Murray, C. B. Synthesis of monodisperse nanoparticles of barium titanate: Toward a generalized strategy of oxide nanoparticle synthesis. J. Am. Chem. Soc. 123, 12085–12086 (2001).

49. Dutta, P. K. & Gregg, J. R. Hydrothermal synthesis of tetragonal barium titanate (BaTiO3). Chem. Mater. 4, 843–846 (1992).

50. Xu, H., Gao, L. & Guo, J. Preparation and characterizations of tetragonal barium titanate powders by hydrothermal method. J. Eur. Ceram. Soc. 22, 1163–1170 (2002).

51. Yashima, M. et al. Size effect on the crystal structure of barium titanate nanoparticles. J. Appl. Phys. 98, 0–8 (2005).

52. Nyutu, E., Chen, C., Dutta, P. & Suib, S. L. Effect of microwave frequency on hydrothermal synthesis of nanocrystalline tetragonal barium titanate. J. Phys. Chem. C 112, 9659–9667 (2008).

53. Hoshina, T., Kakemoto, H., Tsurumi, T., Wada, S. & Yashima, M. Size and temperature induced phase transition behaviors of barium titanate nanoparticles. J. Appl. Phys. 99, 1–9 (2006).

54. Rimai, L., Parsons, J. L., Hickmott, J. T. & Nakamura, T. Raman spectrum of long-wavelength phonons in tetragonal barium titanate. Phys. Rev. 168, 623 (1968).

55. Robins, L. H. et al. Investigation of the structure of barium titanate thin films by Raman spectroscopy. J. Appl. Phys. 76, 7487–7498 (1994).

122

56. Sanjurjo, J. A., Katiyar, R. S. & Porto, S. P. S. Temperature dependence of dipolar modes in ferroelectric BaTi O 3 by infrared studies. Phys. Rev. B 22, 2396 (1980).

57. Fontana, M. P. & Lambert, M. Linear disorder and temperature dependence of Raman scattering in BaTiO 3. Solid State Commun. 10, 1–4 (1972).

58. Vinothini, V., Singh, P. & Balasubramanian, M. Synthesis of barium titanate nanopowder using polymeric precursor method. Ceram. Int. 32, 99–103 (2006).

59. Chen, Z. & Peng, F. Normal and Abnormal Grain Growths in BaTiO 3 Fibers. J. Am. Ceram. Soc. 97, 2755–2761 (2014).

60. Zhou, Z., Tang, H. & Sodano, H. A. Vertically aligned arrays of BaTiO3 Nanowires. ACS Appl. Mater. Interfaces 5, 11894–11899 (2013).

61. Zhuang, Y. et al. Fabrication and Piezoelectric Property of BaTiO 3 Nanofibers. J. Am. Ceram. Soc. (2014). doi:10.1111/jace.13028

62. Huey, B. D. et al. The importance of distributed loading and cantilever angle in piezo-force microscopy. J. Electroceramics 13, 287–291 (2004).

63. Zhou, Z., Bowland, C. C., Malakooti, M. H., Tang, H. & Sodano, H. A. Lead-free 0.5Ba(Zr 0.2 Ti 0.8 )O 3 –0.5(Ba 0.7 Ca 0.3 )TiO 3 nanowires for energy harvesting. Nanoscale 8, 5098–5105 (2016).

64. Bowland, C. C. et al. Scalable synthesis of morphotropic phase boundary lead zirconium titanate nanowires for energy harvesting. Nanoscale 26, 5098–5105 (2014).

65. Van Run, A. M. J. G., Terrell, D. R. & Scholing, J. H. An in situ grown eutectic magnetoelectric composite material. J. Mater. Sci. 9, 1710–1714 (1974).

66. Ren, S. Q. et al. BaTiO3/CoFe2O4 particulate composites with large high frequency magnetoelectric response. J. Mater. Sci. 40, 4375–4378 (2005).

67. Nie, J., Xu, G., Yang, Y. & Cheng, C. Strong magnetoelectric coupling in CoFe2O4–BaTiO3 composites prepared by molten-salt synthesis method. Mater. Chem. Phys. 115, 400–403 (2009).

68. Botello-Zubiate, M. E., Bueno-Baqués, D., De Frutos Vaquerizo, J., Fuentes Cobas, L. E. & Matutes-Aquino, J. A. Magnetoelectric Measurements by Two Different Methods of Cobalt Ferrite-Barium Titanate Composites. Ferroelectrics 338, 247–253 (2006).

123

69. Botello-Zubiate, M. E., Bueno-Baqués, D., Vaquerizo, J. D. E. F., Fuentes Cobas, L. E. & Matutes-Aquino, J. A. SYNTHESIS AND MAGNETOELECTRIC CHARACTERIZATION OF COBALT FERRITE—BARIUM TITANATE COMPOSITES USING A NEW PULSED MAGNETIC FIELD METHOD. Integr. Ferroelectr. 83, 33–40 (2006).

70. Mazumder, S. & Bhattacharyya, G. S. Synthesis and characterization of in situ grown magnetoelectric composites in the BaO–TiO–FeO–CoO system. Ceram. Int. 30, 389–392 (2004).

71. Zhang, Y., Deng, C., Ma, J., Lin, Y. & Nan, C.-W. Enhancement in magnetoelectric response in CoFe2O4–BaTiO3 heterostructure. Appl. Phys. Lett. 92, 62911 (2008).

72. Xie, S. H. et al. Multiferroic CoFe2O4–Pb(Zr0.52Ti0.48)O3 nanofibers by electrospinning. Appl. Phys. Lett. 92, 62901 (2008).

73. Xie, Y. et al. Synthesis of Multiferroic Pb(Zr0.52Ti0.48)O3–CoFe2O4 Core–Shell Nanofibers by Coaxial Electrospinning. Nanosci. Nanotechnol. Lett. 5, 546–551 (2013).

74. Starr, J. D., Budi, M. a. K. & Andrew, J. S. Processing-Property Relationships in Electrospun Janus-Type Biphasic Ceramic Nanofibers. J. Am. Ceram. Soc. 98, 12–19 (2015).

75. Budi, M. A. K., Glass, E. B., Rudawski, N. G. & Andrew, J. S. Exchange bias in bismuth ferrite/cobalt ferrite Janus nanofibers. J. Mater. Chem. 5, 8586–8592 (2017).

76. Liu, B., Hu, B. & Du, Z. Hydrothermal synthesis and magnetic properties of single- crystalline BiFeO3 nanowires. Chem. Commun. 47, 8166–8168 (2011).

77. Zou, Y., Chen, Z.-G., Huang, Y., Drennan, J. & Zou, J. Au-catalyzed and catalyst- free growth of one-dimensional Bi2Se3 nanostructures. 2014 Conference on Optoelectronic and Microelectronic Materials & Devices 11– 14 (2014). doi:10.1109/COMMAD.2014.7038638

78. Corral-Flores, V., Bueno-Baqués, D. & Ziolo, R. F. Synthesis and characterization of novel CoFe2O4–BaTiO3 multiferroic core–shell-type nanostructures. Acta Mater. 58, 764–769 (2010).

79. Kulkarni, G. U., Kiruthika, S., Gupta, R. & Rao, K. D. M. Towards low cost materials and methods for transparent electrodes. Curr. Opin. Chem. Eng. 8, 60– 68 (2015).

80. Doshi, J. & Reneker, D. H. Electrospinning process and applications of electrospun fibers. Conf. Rec. 1993 IEEE Ind. Appl. Conf. Twenty-Eighth IAS Annu. Meet. (1993). doi:10.1109/IAS.1993.299067

124

81. Li, D. et al. Hard magnetic FePt nanoparticles by salt-matrix annealing. J. Appl. Phys. 99, 08E911 (2006).

82. Bauer, M. J. et al. Structure-Property Relationships in Aligned Electrospun Barium Titanate Nanofibers. J. Am. Ceram. Soc. 99, 3902–3908 (2016).

83. Sarkar, P. & Nicholson, P. S. Electrophoretic Deposition (EPD): Mechanisms, Kinetics, and Application to Ceramics. J. Am. Ceram. Soc. 79, 1987–2002 (1996).

84. Derby, B. Inkjet Printing of Functional and Structural Materials: Fluid Property Requirements, Feature Stability, and Resolution. Annu. Rev. Mater. Res. 40, 395–414 (2010).

85. Li, M. et al. Bottom-up assembly of large-area nanowire resonator arrays. Nat. Nanotechnol. 3, 88–92 (2008).

86. Freer, E. M., Grachev, O., Duan, X., Martin, S. & Stumbo, D. P. High-yield self- limiting single-nanowire assembly with dielectrophoresis. Nat. Nanotechnol. 5, 525–30 (2010).

87. Ramos, A., Morgan, H., Green, N. G. & Castellanos, A. Ac electrokinetics: a review of forces in microelectrode structures. J. Phys. D Appl. Phys. 31, 2338– 2353 (1998).

88. Gul, I. H. & Maqsood, A. Structural, magnetic and electrical properties of cobalt ferrites prepared by the sol–gel route. J. Alloys Compd. 465, 227–231 (2008).

89. Srinivasan, G., Rasmussen, E. T., Levin, B. J. & Hayes, R. Magnetoelectric effects in bilayers and multilayers of magnetostrictive and piezoelectric perovskite oxides. Phys. Rev. B Condens. Matter 65, 28 (2002).

90. Duong, G. V, Groessinger, R., Schoenhart, M. & Bueno-Basques, D. The lock-in technique for studying magnetoelectric effect. J. Magn. Magn. Mater. 316, 390– 393 (2007).

91. Jones, T. B. Electromechanics of Particles by Thomas B. Jones. (2017).

92. Albrecht, D. R., Sah, R. L. & Bhatia, S. N. Geometric and Material Determinants of Patterning Efficiency by Dielectrophoresis. Biophys. J. 87, 2131–2147 (2004).

93. Sherman, R. in Developments in Surface Contamination and Cleaning 695–716 (Elsevier, 2016).

94. Wan, J. G. et al. Magnetoelectric CoFe2O4–Pb(Zr,Ti)O3 composite thin films derived by a sol-gel process. Appl. Phys. Lett. 86, 122501 (2005).

125

95. McDannald, A. et al. Magnetoelectric coupling in solution derived 3-0 type PbZr0. 52Ti0. 48O3: xCoFe2O4 nanocomposite films. Appl. Phys. Lett. 102, 122905 (2013).

96. Shi, M. et al. Magnetoelectric properties of CoFe 2 O 4–Pb (Zr 0.52 Ti 0.48) O 3 multilayered composite film via sol–gel method. J. Mater. Sci. 46, 4710–4714 (2011).

97. Liu, X.-M., Fu, S.-Y. & Huang, C.-J. Synthesis and magnetic characterization of novel CoFe2O4–BiFeO3 nanocomposites. Mater. Sci. Eng. B 121, 255–260 (2005).

98. Zhou, Y., Chul Yang, S., Apo, D. J., Maurya, D. & Priya, S. Tunable self-biased magnetoelectric response in homogenous laminates. Appl. Phys. Lett. 101, 232905 (2012).

99. Zhou, Y., Apo, D. J. & Priya, S. Dual-phase self-biased magnetoelectric energy harvester. Appl. Phys. Lett. 103, 192909 (2013).

100. Li, M., Wang, Z., Wang, Y., Li, J. & Viehland, D. Giant magnetoelectric effect in self-biased laminates under zero magnetic field. Appl. Phys. Lett. 102, 82404 (2013).

101. Nan, T., Hui, Y., Rinaldi, M. & Sun, N. X. Self-biased 215MHz magnetoelectric NEMS resonator for ultra-sensitive DC magnetic field detection. Sci. Rep. 3, 1985 (2013).

102. Srinivasan, G. et al. Magnetoelectric bilayer and multilayer structures of magnetostrictive and piezoelectric oxides. Phys. Rev. B 64, 214408 (2001).

103. Van Den Boomgaard, J., Terrell, D. R. & Born, R. A. J. An in situ grown eutectic magnetoelectric composite material. J. Mater. Sci. 9, 1705–1709 (1974).

104. Zhang, C. L., Chen, W. Q., Xie, S. H., Yang, J. S. & Li, J. Y. The magnetoelectric effects in multiferroic composite nanofibers. Appl. Phys. Lett. 94, 102907 (2009).

105. Ziegler, S., Woodward, R. C., Iu, H. H.-C. & Borle, L. J. Current Sensing Techniques: A Review. IEEE Sens. J. 9, 354–376 (2009).

106. Bauer, M. J., Wen, X., Tiwari, P., Arnold, D. P. & Andrew, J. S. Magnetic field sensors using arrays of electrospun magnetoelectric Janus nanowires. Microsystems Nanoeng. 4, 37 (2018).

107. Bauer, M., Arnold, D. & Andrew, J. Arrays of Janus-Type Magnetoelectric Nanowires for Passive Magnetic Field Sensing. 2018 IEEE 18th International Conference on Nanotechnology (IEEE-NANO) 1–3 (2018). doi:10.1109/NANO.2018.8626229

126

108. Ma, J., Hu, J., Li, Z. & Nan, C. W. Recent progress in multiferroic magnetoelectric composites: From bulk to thin films. Adv. Mater. 23, 1062–1087 (2011).

109. Wu, W. et al. Lead zirconate titanate nanowire textile nanogenerator for wearable energy-harvesting and self-powered devices. ACS Nano 6, 6231–6235 (2012).

110. Fu, J. et al. Wire-in-tube structure fabricated by single capillary electrospinning via nanoscale Kirkendall effect: the case of nickel-zinc ferrite. Nanoscale 5, 12551– 12557 (2013).

111. Lucuţa, P. G., Teodorescu, V. & Vasiliu, F. SEM, SAED, and TEM investigations of domain structure in PZT ceramics at morphotropic phase boundary. Appl. Phys. A Mater. Sci. Process. 37, 237–242 (1985).

112. Ragini, Ranjan, R., Mishra, S. K. & Pandey, D. Room temperature structure of Pb(ZrxTi1−xO3) around the morphotropic phase boundary region: A Rietveld study. J. Appl. Phys. 92, 3266–3274 (2002).

113. Wan, W., Sun, J., Su, J., Hovmöller, S. & Zou, X. Three-dimensional rotation electron diffraction: software RED for automated data collection and data processing. J. Appl. Crystallogr. 46, 1863–1873 (2013).

114. Thibeau, R., Brown, C. & Heidersbach, R. Characterization of oxidation product films on lead in aqueous media by in situ raman spectroscopy. (RHODE ISLAND UNIV KINGSTON DEPT OF OCEAN ENGINEERING, 1979).

115. Mahajan, R. P., Patankar, K. K., Kothale, M. B. & Patil, S. A. Conductivity, dielectric behaviour and magnetoelectric effect in copper ferrite-barium titanate composites. Bull. Mater. Sci. 23, 273–279 (2000).

116. Kumazawa, H. & Masuda, K. Fabrication of barium titanate thin films with a high dielectric constant by a sol–gel technique. Thin Solid Films 353, 144–148 (1999).

117. Arunkumar, A., Vanidha, D., Oudayakumar, K., Rajagopan, S. & Kannan, R. Metallic magnetism and change of conductivity in the nano to bulk transition of cobalt ferrite. J. Appl. Phys. 114, 183905 (2013).

118. Randall, C. A., Kim, N., Kucera, J., Cao, W. & Shrout, T. R. Intrinsic and extrinsic size effects in fine‐grained morphotropic‐phase‐boundary lead zirconate titanate ceramics. J. Am. Ceram. Soc. 81, 677–688 (1998).

119. Gerson, R. & Jaffe, H. Electrical conductivity in lead titanate zirconate ceramics. J. Phys. Chem. Solids 24, 979–984 (1963).

120. Mohan, G. R., Ravinder, D., Reddy, A. V. R. & Boyanov, B. S. Dielectric properties of polycrystalline mixed nickel–zinc ferrites. Mater. Lett. 40, 39–45 (1999).

127

121. Washburn, E. W. & West, C. J. International critical tables of numerical data, physics, chemistry and technology. (National Academies, 1926).

122. Mohsen-Nia, M., Amiri, H. & Jazi, B. Dielectric constants of water, methanol, ethanol, butanol and acetone: measurement and computational study. J. Solution Chem. 39, 701–708 (2010).

123. Shell Chemicals. Technical Datasheet for Isopropyl Alcohol, Product Code S1111, CAS Registry Number 67-63-0.

124. Wohlfarth, C. in Static Dielectric Constants of Pure Liquids and Binary Liquid Mixtures 45 (Springer, 2015).

128

BIOGRAPHICAL SKETCH

Matthew Jeffrey Bauer was born on August 18, 1991 to Jeffrey and Ellen Bauer in Knox, Pennsylvania, United States. He attended Knox elementary school until the age of 7 with his brother Micheal J. Bauer before the family moved to Clarion

Pennsylvania and he began attending Clarion elementary school and subsequently

Clarion area high school.

After high school Matt as enrolled in Clarion University of Pennsylvania with a

Full Board of Governers Scholarship to persue a Bachelors degree Physics and

Mathematics dual degrees with minors in Nanotechnology and Honors. During his time at Clarion University and a research experience for undergraduate program in Penn

State, Matt researched silica nanowire synthesis via VLS, carbon coatings for dental implants, carbon nanodiamond purification and functionalization, and bottom up assembly techniques.

Upon graduating summa cum laude he entered into the University of Florida, completed his Masters Degree in Materials Science and engineering where he also researched magnetoelectric nanowires and device fabrication using said nanowires.

Upon being awarded his masters, Matt continued on to pursue his PhD in Materials

Science and Engineering at the University of Florida, graduating obtaining his degree in the spring of 2019.

129