THESIS

                                       

 

 

 

 

UNI”NIH!!!”HUIIUIIIIHIHW

I 93 01565 6915

 

ilanAnv
Mlchlgan State
Unlverslty

”"" i — ‘.r.—‘_ AA AA.—

 

 

 

This is to certify that the
thesis entitled
"New Techniques for Attaching Small Superconducting Niobium
Leads to Current-Perpendicular-to-Plane Giant

Magnetoresistacfg Samples"
presente y

MATTHEW WAYNE OONK

has been accepted towards fulfillment
of the requirements for

M. S . degree in _hx11LLS__P '

 

we“; /m

Major professor

Date April 30, 1997

0-7639 MS U is an Afﬁrmative Action/Equal Opportunity Institution

 

 

 

 

PLACE N RETURN BOX to remove We checkout from your record.
To AVOID FlNEs return on or before dete due. l

     
   

DATE DUE DATE DUE DATE DUE

“mg” 7’ 1.3 .
ﬂ7e.s 44
MAGIC 2 t
59.2 6. 2 7475-3

 

 

 

 

 

 

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU Ie An Affirmative Action/Equal Opportunity Inetltwon
Wan-9.1

 

NEW TECHNIQUES FOR ATrACHmG SMALL SUPERCONDUCTING
NIOBIUM LEADS TO CURRENT-PERPENDICULAR-TO-PLANE GIANT
MAGNETORESISTANCE SAMPLES
By

Matthew W. Oonk

A THESIS
Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of
MASTER OF SCIENCE
Department of Physics and Astronomy

1997

ABSTRACT
NEW TECHNIQUES FOR ATTACHING SMALL SUPERCONDUCTING
NIOBIUM LEADS To CURRENT-PERPENDICULAR-TO—PLANE GIANT
MAGNETORESISTANCE SAMPLES
By

Matthew W. Oonk

Giant Magnetoresistance (GMR) is the term used to describe large
magnetoresistance behavior in magnetic multilayers. Studying multilayer samples in the
Current-Perpendicular—to-Plane (CPP) direction allows straightforward determination of
magnetic and spin-dependent parameters in the bulk and at interfaces while using
superconducting leads permits uniform current density and low contact resistance to
accurately measure CPP samples.

This thesis outlines a process for fabricating superconducting CPP samples with
small-scale Niobium (Nb) contacts. It demonstrates the advantages of measuring GMR
samples using micron-sized leads and outlines the basic steps for sample fabrication. Test
samples discussed demonstrate the feasibility of this process and show that predictable
results can be obtained. A discussion of current densities illustrates that further reduction
in the size of the top contacts will not destroy superconductivity in the samples. Finally an
example of a system whose study would beneﬁt from this technique is outlined: the

exchange-biased CPP spin-valve.

ACKNOWLEDGMENTS

I wish to thank my advisor Professor William P. Pratt Jr. for giving me the
opportunity to work on an interesting project and for always having the patience and time
to help me through the problems. I would also like thank Professor Jack Bass for giving
me the ﬁrst opportunity to work with the GMR group here at MSU.

I also am grateful to all the people who taught me all the things I needed to know
about fabrication tools and thin ﬁlm techniques. They always had an extra ﬁve minutes to
answer any question, ﬁx a problem, or point me in the right direction. This group includes
Reza Loloee, Wen-Chung Chiang, Baokang Bi, Larry Henry, Shih-ang (Amy) Hsu, Carl
Hoﬁ‘, and Mike Jaeger. Special thanks to Baokang for repairing the RIE and Reza for
ﬁxing everything else.

Also thanks to the rest of the group, Professor Peter Schroeder, Paul Holody,
Carole Vouille, Anne Schaefer/Reilly and Wen-Chung, for making the working
environment more enjoyable. Finally a special thanks to my wife, Angie, for making the
rest of my life more enjoyable.

The work of our group is funded by the U. S. National Science Foundation,
Michigan State University Center for Fundamental Materials Research, and the Ford

Motor Company. Their support is gratefully acknowledged.

TABLE OF CONTENTS

LIST OF

TABLES .......................................................................................................................... v
LIST OF

FIGURES ....................................................................................................................... vi
INTRODUCTION ........................................................................................................... 1
CHAPTER 1

GIANT MAGNETORESISTANCE, CPP AND NEW SAMPLE GEOMETRY .............. 2
CHAPTER 2

NEW SAMPLE FABRICATION ................................................................................... 15
CHAPTER 3

SAMPLES AND MEASUREMENTS ........................................................................... 26
CHAPTER 4

ISSUES AND FUTURE DIRECTIONS ........................................................................ 36
APPENDIX

TEST SAMPLES AND RESULTING FIXES ............................................................... 42
REFERENCES .............................................................................................................. 52

LIST OF TABLES

Table 3.1 - Physical Characteristics of Sample 820-6 .............................................. 28
Table 3.2 - Physical Characteristics of Sample 823-3 .............................................. 30
Table A] - SiO Films - Resistance Measurements .................................................. 47

LIST OF FIGURES

Figure 1.1: Magnetic Multilayers ........................................................................... 2
Figure 1.2: Resistance vs. Field for a Co/Cu Multilayer Sample at the First
Antiferromagnetic Coupling Peak ................................................................ 4
Figure 1.3: CPP vs. CIP ......................................................................................... 5
Figure 1.4: Present CPP Sample Geometry ............................................................. 7
Figure 1.5: Microfabricated Non-Superconducting CPP Sample ............................... 9
Figure 1.6: New Sample Geometry ........................................................................ 12
Figure 1.7: Current Transport in 3-Pillar Sample ..................................................... 13
Figure 2.1: Basic steps for new sample fabrication .................................................. 16
Figure 2.2: SiO Evaporation System. ..................................................................... 21
Figure 3.1: 2-Pillar 1mm Sample ............................................................................ 27
Figure 3.2: 3-Pillar 70um Sample ........................................................................... 27
Figure 3.3: Resistance vs. Current Plot for Sample 820-6 ......................................... 29
Figure 3.4: Resistance vs. Current Plot for Sample 823-3 ......................................... 31
Figure 3.5: Cross Section of Current Flow from a Nb Plane to a Nb Pillar ................ 33
Figure 4.1: CPP Exchange-Biased Spin-Valve with Small Current Leads ................. 41
Figure A. 1: Evaporator Crucible with Baﬁle ........................................................... 45
Figure A.2: SiO Test Sample ................................................................................. 46
Figure A.3: SEM Image of Holes in SiO Covered Sample ........................................ 49
Figure A.4: Lithography and SiO Test Sample ........................................................ SO

INTRODUCTION

For several years now, our group at Michigan State University has fabricated and
studied sputtered polycrystalline metallic multilayer samples. When ferromagnetic layers
are present in these multilayer structures, the system may exhibit an unusually large
resistance change on the application of a magnetic ﬁeld. This eﬂ‘ect has been named Giant
Magnetoresistance (GMR) and is useful as a study of an inhomogeneous, ordered,
physical system as well as having a variety of industrial sensing applications.

This thesis will attempt to introduce, develop and document a new method for
attaching Niobium top leads to the Giant Magnetoresistance samples studied with the
current ﬂowing perpendicular to the layers. The new method employs lithography, plasma
etching and evaporated insulating layers to make small, micron-scale top contacts. These
tools have several advantages which will be brieﬂy discussed and allow a much wider
range of sample types.

Unfortunately, implementation of these processes introduces many problems, and
consistent results have been difﬁcult to obtain. As a result, many of the results presented
here are for a small number of test samples and not a large group of successful measure-
ments. A very small amount of ﬁne-tuning is needed to complete this process work and
consistently produce GMR samples with this new procedure. It is therefore the goal of
this document to lay down the basic groundwork for ongoing process-development work

in this area and to deﬁne the progress and problems encountered to this point.

Chapter 1 - Giant Magnetoresistance, CPP and New Sample Geometry

Giant Magnetoresistance (GMR)

Magnetoresistance is a phenomenon that arises in all metals. In the presence of a
magnetic ﬁeld the resistivity of a metal will increase due to the Lorentz force acting on the
moving conduction electrons in that applied ﬁeld. This “standard” magnetoresistance
increase is usually a fraction of a percent at room temperatures in very high magnetic
ﬁelds.

Giant Magnetoresistance (GMR) is the term used to describe the large
magnetoresistance seen in magnetic multilayers. A magnetic multilayer is simply a series
of alternating ferromagnetic (F) and nonmagnetic (N) metallic layers. If the spacing
between the F layers is chosen properly, magnetic coupling will align the domains in the F
metals so that the magnetization of alternating F layers is antiparallel in the absence of a

magnetic ﬁeld as seen in Figure 1.1a.

(a) (b)
Applied Field

 

 

 

 

Figure 1.1: Magnetic Multilayers

The GMR effect can be seen when we apply a magnetic ﬁeld in an in-plane
direction to these antiparallel multilayer samples and measure resistance changes. The
applied ﬁeld aligns the magnetic layers in the sample, as seen in Figure 1.1b, causing the
resistance to drop. At a sufﬁciently large applied ﬁeld, all the F layers are aligned in the
parallel direction and the resistance of the sample decreases no ﬁitther with applied ﬁeld.
Also, when the external ﬁeld is turned oﬁ‘, the F layers return to their antiparallel state and
the original peak value of the resistance is again reached. This resistance drop is
symmetric about the center with an equal change in either applied ﬁeld direction, and a
resistance versus applied ﬁeld curve is traced out in Figure 1.2.

There are several unusual attributes of this system worth examining. The ﬁrst
thing to note is that, unlike in normal metals, this magnetoresistance is a negative MR.
This means the resistivity decreases with applied magnetic ﬁeld. Unlike magneto-
resistance in normal metal samples, it is not the Lorentz Force on the conducting electrons
that changes the resistance. Rather, it is the spin dependent scattering of the electrons in
the bulk of the F metals and at the PIN interfaces for the parallel or antiparallel
magnetization states that causes the resistance change. In this case, when the F layers are
aligned parallel to each other, electrons of one spin direction experience less scattering and

cause the resistance to drop.

 

 

R/R(H=0)

 

 

 

 

 

H (kOe)

Figure 1.2: Resistance vs. Field for a Co/Cu Multilayer Sample at the First
Antiferromagnetic Coupling Peak

The other important aspect of these magnetic multilayers is the magnitude of the
MR change. The percent change of a sample’s resistance is much greater than those of
normal MR in metals, and therefore the term Giant Magnetoresistance is used. At liquid
Helium temperatures, applying a saturation ﬁeld can cause a resistivity decrease to 40% of
the zero-ﬁeld value and signiﬁcant changes in resistivity can be seen in ﬁelds of a' few
Oersteds. Although not as dramatic as cooled samples, room temperature resistivity
changes are still an order of magnitude or more larger than regular MR eﬂ‘ects. This

“Giant” aspect of GMR makes it a very desirable phenomenon to utilize in magnetic

sensing technology. As a result, GMR has become a very active area of study in industrial
and academic labs since its discovery by Albert Fert in 1988 [1]. For example, because of
their increased sensitivity, GMR read/write heads allow higher storage capacity on hard
disk drives and position sensing. IBM Almaden Research Labs have already broken
several storage density records for computer hard disk drives using a GMR “spin valve”
head, and many diﬁ‘erent companies now manufacture GMR heads for hard disk drives

and other magnetic sensing applications [2].

The CPP Geometry

Figure 1.3 below deﬁnes the two directions that current is passed in metallic
multilayers. GMR effects are seen in both directions of current ﬂow, and the Current in
Plane (CIP) directions is the popular choice among researchers in the ﬁeld of GMR.
Because sample thicknesses are usually orders of magnitude smaller than lengths, CIP
samples show much higher resistances (0.1 ohms or more) than CPP. samples, and CIP

looks to be the initial direction for industrial sensing applications.

 

 

Figure 1.3: CPP vs. CIP

The group at Michigan State primarily focuses on the Current Perpendicular to the
Plane (CPP) direction, where the transport theory is much simpler [3,4]. As a result, the
CPP directions allows us to quantify the important parameters that describe the spin
orientation-dependent scattering in the F layers and at the PIN interfaces that CIP would
not allow. Unfortunately, due to the fact that the samples are made of thin ﬁlms, the
sample resistance is quite small (around 10'7 ohms) and therefore more diﬁicult to measure

accurately.

CPP Sample Fabrication

The present CPP sample geometry is diagrammed in Figure 1.4. These samples
sandwich a magnetic multilayer between two Niobium cross strips which become super-
conducting when the sample is cooled to 4.2 K in liquid Helium. The strips only overlap
in one small area, and the use of superconductors allows uniform current in the sample
strips. Because the sample is much thinner than the width of the strips, current will pass
straight down through the sample with little current ﬂow in plane. Hence the area of the
multilayer sampled is simply the overlap of the two perpendicular Niobium strips. This
square is approximately 12-14 mm2 and is determined by scanning both leads in a
Dektak stylus proﬁle scanner.

It should be noted that the circular multilayer also has thin leads mnning to
rectangular contact areas. These thin leads allow measurement of the CIP resistance of

the sample.

 

 

 

 

Sample Area /
CIP Contact

Figure 1.4: Present CPP Sample Geometry

The samples are fabricated in MSU’s four-gun sputtering system. This system
contains four separate Simard Trimag sputtering guns placed at right angles from each
other in a circular vacuum chamber. The 8-16 substrates are mounted on a sample
positioning plate above the sputtering guns, and the plate is rotated to the proper gun
using a computer-controlled stepping motor. Each gun has a chimney which is opened
and closed via a vacuum feedthrough to ensure each metallic layer is not contaminated
with other metals as the substrate rotates past other sputtering guns. The geometry of
each metal layer is determined by stainless steel masks placed just below the samples, and
CPP samples have four mask settings which can be changed by rotating the disks to
diﬁ‘erent positions using a vacuum-compatible wobble stick [5]. The system is pumped to
better than 3x10'a Torr base pressure using a cryopump and liquid Nitrogen trap. The

chamber is then ﬁlled with 2-3 mTorr of puriﬁed Ar gas for sputtering.

All measurements made on the samples are performed at liquid Helium
temperatures (4.2 K) employing a sensitive four-probe technique which uses a
Superconducting Quantum Interference Device (SQUID) potentiometer circuit with the
sample resistance and a 95 u!) reference resistor. When current is sent through the
sample, the SQUID null detector feeds back current through the reference resistor until a
potentiometer balance is attained. This system allows measurements of resistances in the
picoohm range with great stability. Also, the system is designed to measure samples with
resistances in excess of 100m as long as the applied current is low enough. This ability
to measure higher resistances will become important as the Nb contacts are reduced to

micron sizes.

Small Scale CPP Measurements

Although Nb superconducting leads provide highly uniform current densities in
CPP samples, measurements are limited to 4.2 K. By abandoning the superconducting
leads and reducing the area of the contact pads, it is possible to study perpendicular
transport at higher temperatures and resistances that approach practical device operating
conditions. One major problem that arises from such CPP measurements is that, because
the samples are much thinner than they are wide and the contact pads have a ﬁnite
resistance, current densities vary over the width of the sample. Therefore the sample
areas must be made as small as possible to reduce the nonuniformity of the current. In an

attempt to study temperature dependence in CPP multilayers, others have used

various lithographic techniques to microfabricate multilayer samples and contacts with
varying degrees of success [6, 7, 8, 9].

The major problem with such measurements can be seen when we examine Figure
1.5 which shows the sample geometry used by Krebs et aI [ 6,7]. In this case the square
multilayer is sandwiched between two normal metal leads with a circular top contact that
is 1.2um in diameter. To prevent current from entering the voltage leads all the leads are

made as narrow as possible.

 

I- |‘_ 1pm __>| V-

Figure 1.5: Microfabricated Non-Superconducting CPP Sample

This type of CPP system allows measurements over a large temperature range, but
it has several problems. First, because no superconductors are used, the sample has non-
uniform current density, and current is concentrated on the left edge of the sample, as seen

in Figure 1.5. Therefore expressions must be derived for the current densities in various

10

samples, and GMR measurements have to be corrected accordingly. Secondly, the
situation becomes more complicated because the temperature dependence of resistivity is
diﬁ‘erent in the leads and the multilayer. As a result, the current density changes as the
sample is cooled, and MR curves therefore depend on the contact diameter. Also Gijs et
al, and Spallas et al have found that signiﬁcant contact resistances are observed between

the leads and the sample [8,9].

Advantages of Superconducting Microfabricated Samples

Lithographed CPP samples made with one or more small superconducting top
contacts have several unique properties. Micron-sized leads allow highly localized
resistance measurements of multilayer samples with uniform current density and low
contact resistance. Using multiple top contacts also allows study of the reproducibility of
measurements made at various places on an otherwise uniform sample. In addition
samples of varying thickness (wedge-shaped proﬁles) could allow CPP measurements for
a variety of layer thicknesses that have been grown under the same deposition parameters.
Finally, as the size of the contact area become comparable to ferromagnetic domain size,
novel magnetic effects could be observed.

In an effort to take advantage of these attributes of lithographed samples and not
abandon the uniform current characteristics of superconductor contacted samples, we
have developed a technique that uses lithographed Nb top contacts. Microfabricated
superconducting Nb samples have been explored in other areas of study and techniques

have been developed to study Josephson Junction samples as small as a micron with

ll

reliable characteristics [10]. If we can successfully adapt these techniques to fabricating

GMR samples, we will be able to explore a host of new effects, as outlined in the previous

paragraph.

New Sample Geometry

An example of a sample fabricated using new methods is shown on the following
page in Figure 1.6. This sample uses the same CPP structure as discussed before but
instead of one large top Nb lead, the new sample has several (in this case three) smaller
Niobium pillar contacts separated by electrically insulating materials. On top of these
small pillars there are larger Nb contact pads that make superconducting contact to the
small pillars in the sample. These pads are added to the top to allow easy connection of

soldered voltage and current leads in order to measure the sample resistance.

Advantages of the New Sample Geometry
The many advantages of using a combination of lithography and superconductors
have already been discussed. But, there are a few more beneﬁts that come from this new

sample geometry.

 

 

 

Figure 1.6: New CPP Sample Geometry

The ﬁrst advantage of the new sample geometry can be seen in Figure 1.7 which
shows a possible transport of current through the sample. With three individual pillars
current can be sent from one pillar to the other while voltage (V+, V.) is measured between
two contacts as shown. Since negligible current is ﬂowing between the bottom
superconductor and the V. contact, this V. contact is actually measuring the voltage of the
bottom superconductor. Other combinations of I and V will allow CPP measurements of
the other two pillars and connecting the I. and V. to the same pillar allows the
measurement of two in series. Thus the three pillar design allows several measurements
per sample and does not require that contact be made to the bottom superconducting
layer.

Because the new sample fabrication process uses lithography to control the
geometry of the top Niobium layer, no contact masks are used during the sputtering

process. This lack of masks will be particularly useﬁrl when epitaxial CPP sputtered

13

samples are made. Such epitaxial growth requires the ﬁrst Niobium layer to be deposited
with the substrate heated to about 1000 °C and the presence of masks would greatly
complicate this sample fabrication process. Epitaxial growth will allow the study of

magnetocrystalline effects that have not been explored in the CPP conﬁguration.

 

Figure 1.7: Current Transport in 3-Pillar Sample

There are many aspects of microfabricated GMR samples that have not yet been
well studied, particularly in the perpendicular direction. For example by microfabricating

CPP samples, interfacial and bulk properties of small samples could be studied. Overall,

14

developing a consistent process to make these samples would open up many new avenues

of study of CPP multilayers.

15

Chapter 2 - New Sample Fabrication

The basic steps for the fabrication of a three-pillar sample are outlined in Figure
2.1 on the following page. Changing the mask for the photoresist, and consequently the
geometry of the top Nb leads, does not affect the basic method of creation. Thus samples

of all sorts of geometries can be made using the same basic procedures.

Sputtering the Multilayer

The ﬁrst step of sample creation is sputtering a multilayer sandwiched between
two layers of Nb. Samples are DC sputtered onto Sapphire in the UHV compatible four-
gun system described before with or without masks according to need. The system is
pumped down to less than 2x10'8 Torr and then ﬁlled with Ar sputtering gas to a pressure
of 2.5 mTorr. The sample consists of a base of Nb of at least 250nm thickness which is
covered with a multilayer or single layer of metal which is topped with 20nm of copper.
This coating of copper is needed to act as a stop layer during the Reactive Ion Etching
(RIE) of the top Nb. The top layer of Nb is added last to the sample, and its thickness is
also 250nm. In later samples a 20nm capping layer of gold was placed on top of the Nb

to prevent oxidation of the Nb when the sample was removed from the sputtering system.

Photoresist Covering
The sample is removed from the sputtering system and is placed in a test tube with
isopropyl alcohol and agitated in a Branson ultrasonic cleaner for approximately one

minute.

A. Sputter Multilayer

 

 

 
     
 

'llll’;lllllllIIll:III}:litrel(I!!!IIIt!!!ltlllll’lllllllllltr;III!t/IrlllllrlttI'll/Illlllll ""1”"..uunu

.\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

B. Deposit Photoresist Pattern

C. Etch Niobium/Gold Layer

 

I 'll/llll VIII/ll.

x\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

D. Deposit SiO Insulating Layer

 

 

 
 
   

E. Remove Photoreist

 

 

Multilayer
Copper

  
  

 

Figure 2.1: Basic steps for new sample fabrication

17

The sample is blown dry with compressed Nitrogen and photoresist (PR) is spin
coated onto the top surface of the Nb. Covering the surface with a few drops of
photoresist and spinning at 4000 rpm for 30 seconds consistently produces photoresist
layer thicknesses between 1.3 and 1.6 urn. After being spun, the resist is baked for 20
minutes at 70°C to remove excess solvents from the resist.

The PR is then patterned by exposing it under a Mercury UV lamp (365nm) using
a black and white photo negative or iron oxide coated glass slide as a contact mask. The
smallest dimensions of the pattern are limited by the sharpness of the mask and the
wavelength of the UV light. Using photo negatives, dimensions as small as 2-4 pm can
be achieved; but if smaller dimensions are required, the photo negatives can be transferred
to the iron oxide ﬁlms on glass slides using projection mask photolithography and
chemical etching. Iron oxide, a UV absorbing material, allows smallest dimensions to be
as low as 0.5 urn. The sarnple is then developed in Shipley 452 developer for one minute
using de-ionized water as a stop. Aﬁer patterning the sample is again baked at 70°C for
20 minutes, and this is a hard-bake that strengthens the resist.

Although early samples did not have the advantage of a cleanroom, most samples
discussed in this paper were spun, baked and developed in a Class 1000 Microfabrication

room. This facility keeps dust from being imaged on or trapped in lithographed samples.

Etching the Top Nb Layer

The top layer of Nb is etched oﬂ‘ the multilayer using a Sulfur Hexaﬂuoride (SF5)

plasma in a 7” Plasma Therm Batchtop Plasma Processing System. Plasma etching or

18

Reactive Ion Etching (RIE) is an effective technique to perform controlled etching that
leaves the desired pattern of Nb protected by photoresist. In plasma etching various
gasses are placed in an RF ﬁeld to create a plasma that chemically reacts with materials in
the sample. By using different gasses and RF settings, it is possible to etch away one type
of layer while leaving other layers relatively unaffected by the plasma. Also, since RIE
etch rates are often slow, it is easy to control the amount of material taken off by simply
controlling the plasma time.

First, the samples are placed in the RIE system and pumped down to below
15mTorr and etched with an Oxygen plasma for 30 seconds (19 watts, 50 mTorr total
pressure is used in this low-power, low- pressure etch). This plasma serves as a
“descumming” process to eliminate any excess unexposed photoresist that may be left on
the sample after developing.

The samples are left in the system and, immediately following the Oxygen de-
scum, the Sulfur Hexaﬂuoride process is run. At 125 watts of RF power and
approximately 25 volts of DC bias, the Sulfur Hexaﬂuoride etches Nb at a rate of
approximately 100-150nm/minute which means that a 250nm layer can be etched in
approximately 3 minutes. The Sulfur Hexaﬂoride reacts with the copper layer to form
various copper ﬂuoride compounds on top of the ﬁlm, and since these compounds do not
etch away, they serve as a shield for lower layers of the sample. Thus, using the copper
layer as a stop for the etch, it is permissible to run the plasma for a few extra minutes to
ensure the Nb is completely cleaned oﬁ‘ the sample. Although the gold capping layer does

not react with the SF5, at this higher power setting the gas will sputter oﬁ‘ the gold layer at

19

a rate of lO-20nm/minute. Thus by adding two minutes to the plasma time we can
effectively remove the gold as well as the Niobium.

It is important, though, that the process is not run too long because Sulfur
Hexaﬂuoride also etches photoresist, silicon and many metals. The plasma eats away at
silicon at a rate much faster than Nb so sapphire substrates are used for etched samples
instead of the usual (100) silicon. As for the photoresist, the process will remove
approximately 80-100nm/minute of photoresist at the same settings. This is not critical
since our PR layers are much thicker (a factor of 10 or so) than the Nb we want to etch
and thus are still relatively thick after both plasmas are run.

The other problem arising from over-etching the sample is the possibility of
undercutting. Although the top and bottom Nb layers, as well as any metals in the
multilayer, are all covered with a copper or photoresist layer, the plasma may attack them
from the side and eat away layers underneath. But, similar to the photoresist, most
horizontal dimensions are much larger than layer thickness so the eﬁ‘ect is normally
minuscule. Undercutting can really only become an important issue if pillar sizes are
shrunk down to sub- micron size.

The samples discussed in this thesis were all etched for between 5 and 10 minutes
in the Sulfur Hexaﬂuoride. This etch visibly takes oﬁ‘ all the Niobium (the top surface of
etched areas of the samples are copper colored), and Dektak scans of samples show that

all 250nm of Nb is removed.

20

Silicon Monoxide Deposition

After the sample is etched properly, a layer of Silicon Monoxide (SiO) is deposited
on top of the sample with the photoresist pattern still present. SiO ﬁlms of 100-1000nm
are excellent insulating thin ﬁlms; and when grown properly, they exhibit resistances in
the mega-ohm range for areas in the centimeter range. Unfortunately getting consistent
ﬁlms with no pinholing and good adherence is somewhat of a challenge. As is discussed in
the Appendix of this thesis, many methods were used in to deposit SiO with varying
degrees of success.

SiO is deposited onto the samples using an R. D. Mathis SO-22 SiO evaporation
source. The source heats the SiO to 1200°C using a current of up to 250A, and the vapor
then passes through several bafﬂes to a chimney. The bafﬂing system ensures the source
does not have a direct line of sight to the substrate because it has been found that spitting
of large particles by the source causes pinholing and thereby shorts.

The evaporation oven is mounted in a large, bell-jar vacuum system shown in
Figure 2.2. This system is designed as a resistance boat evaporator and the contacts have
been reconﬁgured to accommodate the SiO oven. The sample is mounted on a small
copper plate that is attached to a rotating motor directly above the chimney of the boat.
The sample is spun at a slow rate (20-30 RPM) to ensure even coverage. The whole
system is pumped using a turbo pump and can achieve base pressures of 4-6 X10‘6 Torr.

The current on the oven is slowly ramped up over a 5 minute period until the
desired current of 235-240 Amps is reached. During this period the shutter is closed and

the sample is not spinning. After the boat has been at the desired current for a few

21

minutes, the sample spinning motor is turned on and the shutter is open. The thickness of
the deposition is measured using a crystal ﬁlm thickness monitor (FTM) which is placed
just above the sample. With a current of 23 5-240 Amps and a source-sample distance of
18cm, a deposition rate of 35-80 A/sec is achieved. Thus to achieve the desired SiO
thickness of 3500-5000A, the shutter is left open for 60-100 seconds. This short
evaporation time is important because longer evaporation time could heat the sample and
melt the photoresist. If the photoresist edges are not sharp, it can be diﬁicult or
impossible to lift off after the evaporation. To prevent further substrate heating, the
samples are put in thermal contact with a copper plate; but if the evaporation time is too

long, even the copper plate will overheat.

  
 
   

Film Thickness
Monitor (FTM)

Spinning Motor
Shutter Control
Shutter

Sample on
Copper Plate

SiO Oven

Figure 2.2: SiO Evaporation System

22

Using the R D. Mathis SiO sources produces high quality ﬁlms provided the
underlying multilayer is clean and smooth. The Appendix will discuss the results of
resistance measurements made on ﬁlms made with the SiO oven and with an open source
evaporator. Test samples outlined in Chapter 3 will show that, with careﬁil monitoring of

the deposition parameters, the desired resistance characteristics can be achieved.

Removal of Photoresist, Ion Milling and Top Nb Contact Pad Deposition

The photoresist mask which served to protect the top Nb pillars during the RIB
process also prevents this same layer ﬁ'om being coated with insulating SiO. Once the SiO
layer is added to the sample, the photoresist is removed to create shafts or vias in the
insulating layer into which we sputter the top Nb contact pads. SiO evaporated on top of
the photoresist is taken away when the photoresist is removed, leaving exposed Nb.

To remove the photoresist from the sample, it is placed in a test tube of acetone
for approximately one minute. This step dissolves the bulk of the photoresist. To ensure
clean removal, the sample is usually “squirted” with acetone and blown dry with nitrogen
This step dislodges extra residue in the vias.

It is extremely important that the exposed Nb pillars be as clean as possible
before sputtering the contact pads. Any photoresist residue or other impure layers left on
the exposed areas will impede the superconducting contact between the top Nb pads and
the Nb pillars. To ensure total liftoﬁ‘, the sample is cleaned in the RTE using the same
descumming oxygen plasma as before. This thirty second etch cleans oﬁ‘ the remainder of

the photoresist left on the Nb.

23

Two methods have arisen to deal with the oxidized Niobium left after the
photoresist is cleaned oﬁ‘. For samples with no capping layer ion milling is an important
step that can ensure good superconducting contact. Upon exposure to the atmosphere,
Nb immediately forms an oxide which does not superconduct, and exposing the sample to
an oxygen plasma in the RIE undoubtedly compounds the problem. Although sputtered
Nb can penetrate some of this top layer, it does not break through all of it. Hence it is
necessary to ion mill the exposed Nb for a short amount of time to remove some of the
oxidized Nb before we sputter on the top contact pads.

Samples not capped with gold are put into a small ion milling chamber. This turbo
pumped system contains an ion mill and a small magnetron sputtering gun and can be
pumped down to 2x10"6 Torr which takes approximately four hours using only a small
turbo pump. After pumping down, a regulated ﬂow of Argon is bled into the chamber and
the ion mill is turned on. The sample is ion milled (500 volts acceleration and 5 mA beam
current) for 30-45 seconds. At the settings used, the ion mill etches at about 8-15 A/sec
depending on the material being milled.

It is equally important to keep the ion milling short in duration. Unlike RIB, ion
milling is not selective in which materials is etches off and ion milling too long could lead
to erosion of the SiO insulating layer. Ion milling could even knock oﬂ‘ SiO atoms and
deposit them on top of the Niobium pillars, making the surface even more impure.

Immediately following the ion milling, the argon ﬂow is increased and the top Nb
contact pads are sputtered on the sample through a contact mask. As mentioned before,

these large Nb top pads make it easier to attach leads to the sample. Because the chamber

24

has a base pressure of 2x10" Torr or better, is ﬁlled with puriﬁed argon, and the
sputtering gun takes only a few seconds to warm up, there is not suﬁcient time or oxygen
for a thick Nb205 layer to form. The contact pads are made to be approximately 2500A
thick, and this takes about 5 minutes at 200 Watts of power. The ﬁnished sample is
removed from the system following the sputtering of the top contact pads and checked at
room temperature with an ohmmeter to verify good contact.

Because of the unreliability of the ion mill to properly clean the surface while
keeping the SiO untouched, a new method has been developed. With the recent addition
of a ﬁfth sputtering gun to the UHV sputtering chamber, it is now possible to deposit a
capping layer of gold to the surface of the Niobium immediately after the top Nb layer is
deposited. Gold does not oxidize in the atmosphere and therefore protects the Nb. Also, a
thin layer of gold in contact with the Nb will go superconducting at liquid Helium
temperatures by the proximity eﬁ‘ect. Therefore a superconducting contact can be
achieved without the use of an ion mill.

In this case, the samples are put back in the large four-gun system, pumped
overnight to a pressure of 3x10" Torr, and Nb is sputtered through contact masks to
make the contact pads. Using the large system gives the advantage of lower base pressure
and therefore higher quality Nb. This Nb has a higher superconducting critical current and
therefore allows more current to be passed through the vias during measurement.

The next chapter will discuss measurements made on a few samples fabricated
using this entire process. Because of the complex nature of the fabrication and our

relative inexperience with many of the tools involved, a few steps were tested individually

25

before the full combination of fabrication techniques was applied to a sample. These step-
by—step tests served to deﬁne and adjust sample and deposition parameters referred to in

this chapter and are discussed with some detail in the appendix.

26

Chapter 3 -Samples and Measurements

This section will discuss two samples made using the process described in Chapter 2.
. Both samples are gold-capped samples that were not ion milled before deposition of top
Niobium contact pads. In these samples the multilayer is replaced by a 20 or 40 nm layer
of cobalt (capped with the 20 nm of copper.) This single layer system is easy to study and
allows us to ignore the complexity of any multilayer eﬂ‘ects in our early measurements.

All other parameters are as discussed in Chapter 2.

Details of Sample Physical Parameters

Two types of sample were created for ﬁnal measurement: A two-pillar l-mm-
square contact sample and a 3-pillar 75 umx75 um-square sample. In both cases the
Nb/Co/Cu/Nb multilayer was sputter deposited using the CPP 1mm-wide stainless steel
contact masks.

The two-pillar sample is outlined in Figure 3.1. This sample has two large
lithography holes 1mm. wide placed across the multilayer strip. The multilayer CPP area
is determined by the size of the hole made. This sample has three top contact pads
sputtered on top of the insulator with a middle pad that does not make any connection to a
pillar. This is used to test the integrity of the SiO layer

The second type of sample is a 3-pillar sample exactly the same as the ones
outlined in Chapter Two. The nominal pillar dimensions are 75x75 pm, but the actual area

may be smaller due to a little undercutting. The geometry is shown in Figure 3.2, but it is

27

important to note that for easy illustration purposes the dimensions of the pillar have been

exaggerated so it is visible in the drawing.

 
 

Special Solder Contact "
Top Nb Contact Pad

    

Figure 3.1 : 2-Pillar 1mm Sample

 

Figure 3.2 : 3-Pillar 75pm Sample

28

2-Pillar Sample : 820-6

The actual physical characteristics of 2-pillar sample number 820-6 are
summarized in Table 3.1. The size of the hole is determined by Dektak scanning the pillar
aﬁer liﬁoﬁ‘ (Step E in Figure 2.1) and taking an average of the width of the hole where
the SiO is only 250A or less off the bottom of the depression and averaging three scans.
These scans of the vias are in good agreement with measurements of the PR/Au/Nb pillars
left on the sample after the RIE stage (Step C in Figure 2.1). Since we scan the samples
prior to Nb pad deposition to ensure good liﬁoﬂ‘, measuring the vias seems the logical
method for determining sample area. SiO thickness is also the average of three scans of
the ﬁlm edge at different parts of the sample. All errors are the largest diﬁ‘erence between
the mean value and an individual scan.

The SiO-layer resistance measurement is also included in this table. The SiO
resistance is determined at room and liquid He temperatures by attaching leads to the
center contact and the base of the multilayer and measuring with both an Ohmmeter and 4-
probe resistance. The area measured is approximately 1.6 mm2 and, in the case of this
sample, resistances were too high to measure at liquid He temperatures. The SiO ﬁlm

resistance of this sample goes up when it is cooled.

Table 3.1 - Physical Characteristics of Sample 820-6

Lithography Hole Length Pillar 1: 1021i12 um Pillar 2: 1034118 urn
Lithography Hole Width: Pillar 1: 1041i23 um Pillar 2: 1024i15 um
Sample Area: Pillar 1: 1.06i0.04 mm2 Pillar 2: 1.0mm mm2
SiO thickness: 5149i183 A

Resistance of SiO layer Room Temp: 200i20 KS2 Liquid He Temp: > 200 M!)

29

Figure 3.3 shows a graph of resistance vs. current at 4.2°K for sample 820-6. The

two curves represent the resistance of two l-pillar-to-base measurement and a 2-pillar in

series measurement. Recalling the current transport discussion Chapter 1, we see that

one-pillar measurements can be made on this 2-pillar sample if the current is sent from one

pillar to a special solder contact on top of the multilayer (shown in Figure 3.1) and voltage

is measured between the pillar and the other pillar. The mean resistances for Pillars 1 and

2 are 8.10:h0.12 n!) and 8.28:1:0.05nﬂ respectively while in series the resistance is

16.65:|:0. 12 n9. The sum of the two pillar resistance measurement is consistent with the

measurement of the series resistance as it should be and sample shows stable resistance

measurements over a range of 1-100 mA.

 

l6

14

12

Resistance (rt-ohms)

10

Iljlrllrrllrlrﬁm"

 

 

 

 

 

 

 

 

 

 

 

I ?

 

 

 

 

 

 

u 2 Pillars in Series
0 Pillar 1
0 Pillar 2
X Pillar 1 + Pillar 2

 

 

,Minlrnr1rlalrmrlm4‘

 

10'2
Current (Amps)

ﬂ
q

Figure 3.3 Resistance vs. Current Plot for Sample 820-6

30

3—Pillar Sample: 823-3

Unfortunately, due to a number of problems with liftoff and the Reactive Ion
Etching system, no three-pillar samples have been constructed that have all three pillar
contacts working. The photoresist liftoﬂ‘ process and oxygen plasma have left residue on
most three-pillar samples that leads to a very high resistances. Repairs to the RIE will
most likely solve this problem (see the Appendix) and Sample 823 -3, the only 3-pillar
sample run on the newly repaired RIE, has shown promising results in at least one pillar
that cleanly lifted off. Table 3.2 summarizes its physical characteristics. Area is also

determined by scanning the hole in both directions three times.

Table 3.2 - Physical Characteristics of Sample 823-3
Lithography Hole Area: Pillar 1: 73.713 um x 74.1:31 mm = (5.46 10.44) x 10'9 m2

SiO thickness: 4884i63 A

Figure 3.4 shows a graph of resistance vs. current at 4.2 K for sample 823-3. The
mean resistances for Pillar 1 was 1333:0003 u!) and the readings are consistent ﬁ'om
SOuA to SOmA. Measurements at 7 SmA and lOOmA cannot be made in the SQUID
system on this small a sample because it was not designed to handle voltages at such high

CUITCIIIS.

31

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1.4

C :

A b r

1.35 I. .j

3 I _ ii _,tLl trii t1 -

g 2 ti ;

v 1.3 r -:

9 : :

E 1.25 E. .;
1.2 '

lo" 10‘ 10'3 10'2 10 ‘

Current(Amps)

Figure 3.4: Resistance vs. Current Plot for Sample 823 -3

Analysis of Resistance Measurements

Using independent measurements on Cobalt made previously by this group, we
will now try to look at the conductance per unit area or AR. of our samples. AK which is
simply the total resistance times the sample area is the quantity of interest in CPP-GMR
measurements because measurements all involve a wide range of areas.

For CPP samples it is simple to predict both the zero ﬁeld resistance and the GMR
of a sample using a basic model that adds the resistances of each layer and interface in
series. For a single layer of Cobalt, assuming the Copper and Gold are superconducting
by the proximity eﬂ‘ect, we can write the AR. as [5]:

ARt=2ARNbICo+pCotCo (3-1)
Here peak;o is simply the resistivity of our sputtered Cobalt layer times the thickness. The

ﬁrst term is the contribution of the two Nb/Co interfaces on each side of the Co ﬁlm.

32

From previous independent measurements of Nb/Co/Nb ﬁlms, we know that
2ARWC°= 6i] a2 m2 and that pa(, = 68:1: 10 n!) m [5, 11]. Therefore by substituting into
(3.1), we get AR. = 7.36d:1.2 in m2 for our 20nm Co samples.

Now we look at our two measured samples. We have seen that the resistance
measurements have very small uncertainties and are stable over a large range of current.
Therefore, as is the case with present CPP samples being made, the area contributes the
primary error in our measurements. Therefore we shall look at AK IR and compare it to
our Dektak data.

For 820-6 we have resistances for Pillars l and 2 as 8.10i0. 12 n!) and
8.28:1:0.05nQ and this corresponds to pillar areas of 0.91:h. 15 mm2 and 0.89115 mmz.
These numbers are a little lower than the approximate 1mm2 Dektak size, but the
measured values are within the large tolerances of the predicted areas.

For 823-3 we have a resistance for Pillars 1 of 1.333i0.003 it!) and this
corresponds to a pillar area of (5.51i0.88) x10'9m2. This number is very close to the

(5.46 $0.44) x 10'9 m2 measured by the Dektak for this sample area.

Current Density Issues

It is important to look at current densities as the sample size decreases. As the
sample area decreases higher current densities may drive the sample pillars into the normal
regime although none of the samples measured thus far has had any problems
superconducting at 100 mA or less. We will take a brief moment to explore critical

currents and look at values and size limits.

33

Consider the geometrical situation whose cross section is represented in Figure
3.5. Here we have made our conduction contact circular instead of square, and we have
idealized the top contact pads as large circular planes. In this case a circular pillar of
radius R is connected to a top contact plane of a set thickness d which we assume is much
larger than the penetration depth 2.

When potential is put across the top Nb superconducting plane current will only
ﬂow along edge of the plane to as deep as the penetration depth A. Also, by symmetry the
current will ﬂow radially inward toward the pillar and then down. To explore this
problem we will look at two ﬂow densities. We will deﬁne in-plane ﬂow as J I ,and ﬂow
through the via as, Ji. We will look at their relative magnitudes and explore possible

number for each.

 

 

142m

Figure 3.5: Cross Section of Current Flow from a Nb Plane to a Nb Pillar

34

Assuming that the ﬂow along the superconductor is radial and to a depth of 71. and
assuming that ﬂow in the superconducting pillar is uniform over the same depth, we can
deﬁne current densities for input current i at a radius r from the center as: J I = i/21rr).
and Ji=i/1r(R2 —(R-A)2). For R>>A, which is the case here, we have Ji=i/21rRk. So at
r=R, J I = JL as it should to conserve current ﬂow, and J I is at its maximum. Thus we
should look at 1...... = i/ZrtRk.

By approximating our pillar in sample 823-3 as a circular shaft of diameter 75 pm
(this is the closest the parallel current gets to the center and thus maximizes J I and also
underestimates the area), we get Jm = 2.12 x 105A/cm2 at SOmA. This is well below the
critical current density for sputtered Niobium in our cryopumped system which has been
measured as Jc ~ 5x106 A/cm2 [12]. On the other hand, at it’s maximum density in a 1 um
radius pillar the current density is approximately equal to Jum = (i(Amps))(1.6x108)/ cmz.
Thus a current in excess of 30 mA will likely cause Jm to be greater than Jc for a l-um—
radius pillar. Therefore it appears we must be careful when we go to smaller areas to
ensure that J“...x does not exceed Jc.

This approximation is for Nb in the presence of no magnetic ﬁeld and measuring
some GMR samples can require ﬁelds well in excess of 100 gauss. These high ﬁelds
could push the critical current down signiﬁcantly. Also, it should be noted that Niobium
sputtered in the ion milling system is of lower quality and has a lower critical current.
Thus, the need for magnetic ﬁelds further compounds the issue of critical current

dependence.

35

However, most of our samples will not experience any critical current problems for
two reasons. First, many or our early GMR studies with l-um-scale Nb contacts will use
perrnalloy F-layers which only require 10-20 0e ﬁelds. Secondly, we must keep in mind
that the resistance of our sample scales like the inverse of the area and our SQUID
measuring system has a limiting sample voltage of around 100 nV. Thus we must have
(AR.)J’i SIOOnV where J ’i is the perpendicular current density in the multilayer. So for
our 75 umx75um samples we obtain J ’15] .4x103 A/cmz, a value that is independent of area
A. Ifthe pillar diameter of this sample were reduced to a 2pm diameter, we would have
Jms 7x105 A/cm2 which is well below 1,. Hence, we anticipate no dimculty with reduced

superconducting critical currents because it is the measuring system that limits our value

for i.

36
Chapter 4 - Issues and Future Directions

Developing a Consistent System

As stated in the introduction, this is work is just the foundation for the
development of a consistent process. Numerous problems (many of which are
documented in the Appendix) have already been solved, and present work is being done to
update a number of other steps:

The installation of an 810 boat into the small ion mill system will be completed
soon. This system will have a longer source-sample distance, better control of current and
a much lower base pressure. The latter may improve the adherence of SiO ﬁlms to the
sample. This system will also allow ion milling prior to evaporation to clean and roughen
the sample. This system can also monitor the ﬁlm thicknesses more accurately.

Numerous repairs have been done to the RIE which have solved the plasma
problems (see Appendix on RIE etching.)

The addition of a mask aligner to the cleanroom will allow easier and more reliable
methods to make small, sharp samples. This system has better masking and exposure
control.

With the addition of all these new processes and the progress that has already been
made, sample fabrication should become routine in a short time. When this process is
implemented, it will allow study of some interesting systems. The system most likely to be

studied ﬁrst is the microfabricated CPP Exchange-Biased Spin-Valve.

37

The Spin Valve

A diagram of a CPP Exchange-Biased Spin-Valve can be seen in Figure 4.1. This
system consist of 4 layers wedged between the Niobium leads. The layers are, in order, an
antiferromagnetic layer (usually iron-manganese), a ferromagnetic layer, a nonmagnetic
spacer layer and a top ferromagnetic layer. The antiferromagnetic layer serves as a
“pinning” layer and holds the magnetization of its neighboring ferromagnetic layer in one
direction. With the proper post-deposition annealing, the bottom ferromagnetic layer
magnetization is pinned and will not rotate in reasonable ﬁelds. The magnetization of the
top “free” layer rotates with the ﬁeld and can be aligned parallel or antiparallel to the
magnetization of the pinned layer causing an MR change. The CPP spin-valve, which has
already been fabricated and studied here at MSU, has several unique properties that make
it an ideal sample for rnicrofabrication. [13].

Using the new process, lithographed top Nb contacts can be placed on top of the
sample that have lateral dimensions much less than those of the actual spin-valve. Thus
the CPP current ﬂow will occur in a highly localized region of the spin-valve structure

First, the spin valve is a low-ﬁeld system that exhibits high MR. A suitably thin
sample has a saturation ﬁeld on the order of 10 Oersteds, and this low ﬁeld will not
signiﬁcantly lower the superconducting critical current values discussed in Chapter 3.
Also the layers in a spin valve are thin and small in number and this makes the device

easier to grow.

   

External Magnetic Field

Figure 4.1 : CPP Exchange-Biased Spin-Valve with Small Current Leads

39

Most importantly, the spin valve is a popular device in industry because it has a
low switching ﬁeld. Spin valve GMR heads are becoming the industry standard in the
hard disk arena, and many groups are exploring microfabricated read heads. By making
localized CPP measurements on these devices we could gain insight into how these
devices work at a basic level as well as study some unique effects. For example, with the
addition of small CPP contacts, domain wall movement in the free layer would be
observable in these devices by gradually increasing the ﬁeld. This type of measurement

could explore the fundamental nature of the ferromagnetic layers in the system.

Conclusions

Although some adjustments need to be made to existing equipment, we have
demonstrated both the feasibility and the advantages of making CPP samples with micron-
sized superconducting contacts. These samples open up a host of new avenues for study

by simply modifying existing CPP structures.

Giant Magnetoresistant (GMR) multilayer samples have many device applications
and are a unique study of an inhomogeneous system. These samples show large changes
in resistivity in the presence of an applied ﬁeld. Studying these multilayers in the CPP
direction allows determination of important electron-scattering parameters in the bulk of

the layers and at the interfaces.

We demonstrated the many advantages of small scale superconducting

measurements in Chapter 1. Small samples approaching ferromagnetic domain sizes are

40

useful in studying unique small-scale eﬂ‘ects as well as allowing multiple measurements per
multilayer sample. By continuing the use of superconducting contacts, we do limit
ourselves to measurements at temperatures less than 4.2 K, but we also maintain uniform
current densities and small contact resistance not seen in other microfabricated CPP

samples.

The measurements outlined in Chapter 3 show that stable and predictable results
can be achieved and that the reduction in size does not limit the current enough to make
stable micron-scale superconducting measurements impossible. These samples show no
current dependence and agree, within error, with theoretical calculations of AR. Also, as
calculations Showed, currents are only limited by the voltage maxima of our highly-
sensitive SQUID measuring system and not by superconducting critical current even as we

go to micron sized contacts.

Finally we outlined a simple system, the exchange biased spin valve, that could be
studied using the new sample fabrication techniques. This system, which is popular for
sensing applications, has never been studied in the CPP direction at the micron-level. By
putting small contacts on the spin valve, magnetic domain-wall movement could be
studied in the free layers of the system and this could lead to greater understanding of the

fundamentals of this unique device.

Problems with the process are outlined in the Appendix and, to some extent, in
Chapter 2. These problems are all easily corrected with future modiﬁcations to the

systems involved in fabrication. Most of these modiﬁcations are already underway on

41

various deposition and etching tools, and these repairs will lead to a consistent fabrication

process for small-scale superconducting CPP samples.

Appendix

42

Appendix - Test Samples and Resulting Fixes

Desirable Parameters in a Complete Sample

In order to be able to measure the CPP resistivity of the multilayer with
conﬁdence, a number of things must go right. The desired result of all the processing
done on the top Nb layer is to have a superconducting pillar of well established area that is
in intimate contact with the top of the sample. The top contact pads must also be in
intimate contact with the Nb pillars, and they have be large enough to allow easy voltage
and current connections that are electrically isolated from neighboring contacts. Although
this seems simple enough, we must have control over several things simultaneously for this
to happen.

First and foremost we need the SiO insulating layer to be a reliable electrical
insulator. After the sputtering of top contact pads for soldering, it is the SiO layer that
prevents this Niobium ﬁ'om making contact with the exposed Copper top of the
multilayer. Ifthe SiO layer fails in any way to electrically insulate, then the sample has no
deﬁnable geometry and resistivity determinations become impossible. Therefore we
should try to have a ﬁlm that over the entire contact pad area has a resistance of more than
1000 times the pillar resistance.

The second major measurement issue lies in the superconducting paths.
Remembering that current densities must be uniform in CPP measurements, it is important
that there be a continuous superconducting current ﬂow from the top contact pads
through to the top surface of the multilayer. This superconducting contact must also be

stable at reasonably high current densities for GMR measurements. So the Nb-Nb or Nb-

.—

43

Au-Nb interface on the top of the sample has to be ﬁee of any impurities such as
photoresist scum or Nb205 layers.

Finally the removal of layers in the RE must be complete and must not affect
underlying layers. It is essential that the RIE removal of the unwanted top Niobium layer
be complete or else the area of the top Nb pillars becomes undeﬁned. To have a well
deﬁned geometry, the etching must also leave the photoresist-protected pillars and not
affect their size or shape. The top copper layer must protect from the etch, leaving the
underlying multilayer unaffected by the SFG.

What follows is a brief item-by-item description of test samples and problems they
isolated. Much of this section is also more qualitative than quantitative in nature and only
a few numbers have been included where they were deemed useful. On the other hand,
this is where, by far, most of the work has gone and therefore it bears some documenting.

(i.e. It is intended to make sure that history does not repeat itself.)

Silicon Monoxide Deposition

When properly grown, a thin layer of SiO (30-40 nm) serves as an excellent
electrically insulating layer with resistances in the MD range. The problem with SiO is
that if it is evaporated incorrectly it has a tendency to produce a large number of pinholes
and free silicon, making reproducible results difﬁcult to obtain. Therefore developing a
process for growing consistent SiO insulating ﬁlms has been a major part of the

development work. Present samples use a commercially made boat, but older samples

were made using a small evaporator with a baﬁling system. Because many test samples

 

44

were made using this system, I have included a brief description of this evaporator and the
early problems with silicon monoxide layers.

SiO sublimes in a vacuum at temperatures around 1100°C and high quality films
can be quickly deposited at temperatures between 1200°C and 1300°C. Because silicon
monoxide sublimes instead of melting or evaporating, the source is subjected to spitting
which causes pinholes and defects in the ﬁlm. As a result, high quality SiO ﬁlms which
have high resistances and no pinholing cannot be achieved by evaporating with an open
source. Others have found that using a boat or chimney design that eliminates the direct
line of sight between the source and the sample will greatly reduce the spitting and
pinholing that lead to lower quality ﬁlms [14].

SiO was originally deposited onto the samples using a Luxel Radak I evaporator.
This evaporator heats a small (1” High and 0.4” in diameter) Alumina crucible to
temperatures of up to 1500°C using simple radiating coils that surround the crucible. The
power controller system for the evaporator contains both a temperature and a current
controller that allow precise setting of temperature and current limits. By setting the
temperature and limiting the current to 2 or 3 amps, the source can be slowly heated for
15 minutes under a closed shutter until the proper deposition conditions are met.

The evaporator is mounted in the same chamber as the ion mill described in
Chapter 2 . As mentioned before the entire system can be pumped down to pressures
below 10‘6 Torr using the turbo pump. Iflower pressure or faster pumping is necessary,
the small nitrogen trap on the system can be ﬁlled to help remove water vapor. For these

particular samples, the nitrogen trap was not used.

45

At ﬁrst ﬁlms were deposited directly from the open crucible. These ﬁlms were of
very poor quality, having low electrical resistance and poor adherence to the substrate.
To minimize pinholing in our system, we constructed a small bafﬂe that is placed inside the
ceramic crucible and contains only indirect paths from source to substrate. A diagram of

the bafﬂe can be seen in Figure A. 1.

I/ .
\_I"\l \ \ |L[\
I F I ,I I

 

Figure A.1: Evaporator Crucible with Bafﬂe

46

The baﬁle is machined out of graphite and then coated with boron nitride paint to
make it chemically inert. The painted bafﬂe is then baked dry in the evaporator for 10
minutes at 1290 °C. With the bafﬂe in place the evaporator will deposit SiO onto
substrates at a rate of 12-17 A/sec at a temperatures between 1230-1275 °C and a
pressure of 3-5x10'5 Torr. Unfortunately, even with the baﬁle in place the results are
inconsistent and diﬁcult to control, and, as a result, the evaporator was abandoned in
favor of a commercially produced boat.

Figure A.2 shows a basic sample used to test SiO. This sample has two Niobium
strips sandwiching a SiO ﬁlm. Samples were grown on both silicon and sapphire

substrates.

 

   
  
  
  

Top Nlobium Contact __~

/SiOInsulator\

I

Bottom Nrobium Strip \

Substrate \

 

 

 

 

Figure A2 : SiO Test Sarrple

To produce more consistent results on future samples, it became necessary to

purchase a commercial fumace to make SiO ﬁlms. A smaller version of the oven

47

described in Chapter 2 was used at ﬁrst but later the system was switched to the larger
80-22. This fumace is much larger and requires high current leads but it does consistently
produce high quality ﬁhns with a higher depostion rate. Films measured with all the

sources are outlined in Table A.1

Table A.l - SiO Films - Resistance Measurements

 

Sample #’s Source Thickness lmm-square 1mm -square
Resistance Resistance
At Room At 4.2 K
Temp

011096-020296 Open Evaporator l400-3300A 70-500 (2 0* - 10 (I,

021396-081496 Bafﬂe Evaporator 2100-5600 A 1509-101“) 0* - 20 Mo
099096412696 Small SiO Oven 600-5100 A 3009—1001(5) 20 (22001149
112696-Present Large SiO Oven 3000-7000A IKQ-IMQ loco->200Mo

 

* Samples showed superconducting pinholes with critical currents in the mA range.

Known and Possible Problems With the SiO Films

Several problems existed in past ﬁlms that required some adjustment of deposition
parameters, and they consist of:

Pinholing and Spitting: Many samples had serious problems with unseen holes on
their surface that led to poor resistance of the ﬁlm. Most of this was reduced with the

addition of the commercial oven.

48

Poor Adherence: This continues to be a problem with some ﬁlms, although most
of the SiO covered by the contact pads appears to be protected from cracking and peeling.
The SiO ovens should be used at a base pressure of better than 5x106 Torr, and this is
often difﬁcult to obtain in the large evaporation system. The addition of a boat in the ion
mill chamber should solve this problem by allowing a lower base pressure. Also the ﬁxing
of the RIB (discussed later) will help to de-scum the surface better.

Surface or Trapped Dirt: Figure A.3 show an SEM of a top Nb strip deposited on
an SiO ﬁlm. The large holes in the system appear to run all the way to the substrate.
(There is no white conductor seen at the bottom of the holes.) This probably means that
trapped gasses and dust particles may gather on the substrate during the initial deposition
and escape during one of the next pump down cycles. During SiO deposition, samples are
heated and subjected to a low pressure simultaneously, and this could help trapped
particles migrate to the surface. These large holes on the SiO are coated inside during
ﬁnal Nb sputtering, shorting out the sample layers. Many of the samples which had
bottom contacts sputtered in the small system had this particular problem; and after the
samples are lithographed, they cannot be cleaned with solvents until the liftoﬂ‘ occurs.
This was never seen as an issue but the addition of the cleanroom may have helped to
some extent. Holes may still arise occasionally but they appear to be avoidable with more
careﬁil substrate cleaning and deposition techniques (such as using the large, cryopumped

system).

49

 

FIGURE A.3: SEM Image of Holes in SiO Covered Samples

Lithography and Superconduction Test Samples
Figure A.4 shows the samples used to test SiO ﬁlms that have been lithographed.
This sample has two top strips with one maldng contact through a via and the other testing

the integrity of the SiO. Lithography holes are about 90m x 90 um square.

Lithography and Superconduction Problems

A few lithography and superconductor lead problems have arisen:

Clean liftoﬁ‘: This problem has mostly to do with poor SiO oven control and bad
RIE etching. Ifthe RIB etch is having diﬂiculty etching the Nb top layer, it may take

several extra minutes to clean it away. This eats away at the photoresist and can cause it

50

to be only 400 or 500 nm thick at the end of etching. If a 500 nm ﬁlm of SiO is then
deposited on top, the sample will not liftoff properly. Also if the SiO deposition rate is

too slow, the sample will heat causing the resist will ﬂow and liftoﬁ‘ fails to occur.

 

   
  
  
  
  

Top Niobium Contact -
Lithography Via \1 . '
SiO Insulator \

Bottom Niobium Strip. ”

 

 

Substrate \ .

 

 

Figure A.4 : Lithography and SiO Test Sample

Clean Nb under the PR after liftoff: It is critical that the Nb be completely free of
PR after liftoff. This problem has also been remedied with RIE improvements.
Low Tc Nb: This has only been an issue on a couple of occasions in the ion mill

system and is never a problem in the large UHV sputtering system.

The RIE
Many of the liftoﬁ‘ and etching problems occurred due to Nitrogen and Air leaks in

the RIE. These large leaks have been repaired by Baokang Bi, and the system has

 

51

improved dramatically. The SF6 plasma now etches at twice the rate it had previously.
Oxygen plasmas appear to clean better as well.

Still the SF6 etch can be a problem. Some sample require high power (130 Watts
or more) to begin etching the Nb, even if there is not a gold cap. This may be due to
residue or oxides that need sputtering off, but it still poses a hazard to photoresist on the
sample if higher power is used for long periods of time. A harmless, yet puzzling problem
has also popped up recently. Occasionally the RE leaves thin Nb residue on the
multilayer even with extra etching time. This thin layer does not appear to superconduct
and is barely thick enough to be visible on the Dektak so it is not seen as a major problem.

Overall the repair of the RE has aided sample production immensely. This tool is

very integral to the process for several reasons at several steps.

References

 

[1]

[2]

[3]

[4]
[5]

[6]

[7]

[3]

[9]

[10]

[11]

[12]

[13]

[14]

52

References

M.N. Baibich, J .M. Broto, A. Fert, F. Nguyen Van Dau, F. Petroff, P. Etienne, G.
Creuzet, A Friederich, and J. Chazelas, Phys. Rev. Lett. 61, 2472 (1988).

For overview of spin valve head performances see The Era of Magnetoresistive
Heads by Ed Grochowski, on IBM Almaden’s Web Site
(http://www.almaden.ibm.com/storage/oem/tech/eraheads.htm).

SF. Lee, W. P. Pratt, Jr., Q. Yang, P. Holody, R. Loloee, P. A. Schroeder, and J.
Bass, J. Magn. Magn. Mater. 118, L1 (1993).

T. Valet and A. Fert, Phys. Rev. B 48, 7099 (1993).
SF. Lee, Q. Yang, P. Holody, R. Loloee, J. H. Hethrington, S. Mahmood, B.

Ikegami, K. Vigen, L. L. Henry, P. A. Schroeder, W. P. Pratt, Jr., and J. Bass,
Phys. Rev. B 52, 426 (1995).

J. J. Krebs, W. Vavra, G. A. Prinz, S. F. Cheng, and A. Fink, J. Appl. Phys. 79,
6084 (1996).

W. Vavra, S. F. Cheng, A. Fink, J. J. Krebs, and G. A. Prinz, Appl. Phys. Lett.
66, 2579 (1995).

M. A. M. Gijs, S. K. J. Lenczowski, and J. B. Giesbers, Phys. Rev. Lett. 70, 3343
(1993)

J. P. Spallas, Y. Huai, S. Vernon, B. Fuchs, B. Law, D. R. Kania, D. Kroes, M.
Thomas, D. O’Kane, and Z. C. H. Tan, EEE Trans. on Magn. 32, 4710(1996).

Baokang Bi, K. Wan, Wenxing Zhang, Siyuan Han, and James Lukens, EE Trans.
on Apl. Supercnd., l, 145(1991).

C. Fierz, S.-F. Lee, W. P. Pratt, Jr., P. A. Schroeder, and J. Bass, J. Phys.
Condens. Matter. 2, 1224(1991).

C. D. Nelson, W. P. Pratt Jr., S. G. Kim and P. M. Duxbury, IWCC Conference,
(World Scientiﬁc, Amsterdam), 1994

S. Steenwyk, W. P. Pratt Jr., S. Y. Hsu, W.-C. Chiang, AC. Shaefer, R. Loloee
and J. Bass, J. Magn. Magn. Mat. (In Press).

E. D. Blevis, Silicon Monoxide - Properties and Evaporation Techniques, R. D.
Mathis Company, 1995.