3 . .5: 3H“ fill
..U . . '1‘ II. ‘.E vtllll
. V II. II”
5... I. a. . . .vuttl T.-}Jxflml-lff§4clhu (I
I t t . a my ‘1 III - (Ix! -‘v I- t!
. . -nfl . .. lo 9.3:! $3.3:le can - Jr“

.- uv. . b ‘

in

.1 I. . «I vars,“-

IIRJ‘ ‘6!!! .0 lat... .

.hil".hll~lf. (.AJnclaoOHHhuu' -l W , l' I 01.1..“
I . ‘ 1|

    

    
                              

A, l.\ I
, ‘ . it!" 0|» . i I
i I O. I.‘ 1 l I 1 -." "‘ III a .‘K . n I.‘ Iti-
n 3.43.1: :1fl..r1h1...o._.nfh if! 139% I..- \r." . JIM!” vJI
Ir. 5"». clsrlLlllol) 13" - 1V0? . .-
\ .lodvf. "i.t\ncv.‘fifl4 v2.13, .‘Col ”All- ll..." lllk -. “Ia'...
. (AI! 13 r. truthl 3t ”‘71 u 3‘. .I‘Illt . I . -‘Av: 4
I “fir-1w}; flan slid.- . 41.33.22“! III :I .3}: \H 3:. v
. l :\ ‘5 | .. 1L . f .

I. 1L)... ,‘ ‘11]! r.‘

I .. :fi” u
It: 11 t t :1“ gill].
_ i w... H: A

      

    

 

v.1- -I..H
. "WWW; I! ”giadéu .. : -2 9. . ”A! ‘
} “91‘?! . I) I; . '94:. .‘n 3%....l A . I‘ .U -I-..0m. I 8.
- ’33 I)?!» 9.! .Jdflnmflflw I .1. $1.1 .uafierIH134Jv

‘ l

-T
'n‘.’ f”.

 

 

F J“ .
1 1' .
0 . O I I II I II 0 1 . . 0 U.
n -l t I . I .H
I- III O! u ‘ ~
V I - . D I 0‘.
t o I I b . .. ‘1’»!!! O] u
u: - 1.” juryiwr ' “'1’.
. I4: flu ‘14.)!“
y . I . . V
I H ‘1'

I

f
N

1
of

        
  

.HWa‘GJOou HI“ I. f .. I.II 1 11¢.” ‘

. -53.. N 01.0.! ’21 n1 4...; . .1.’
i . - .1 WHO: in“! 3.2.. .l.Mf.p1MU.'d...(/Founn.¢. .f. r. -
.‘I\~D.. 4‘ 1 l ...IH.O I. l‘. ‘Yln’ luv-lflo‘iI-fll,’ "HI

J) In!!! :51...” vllfiffllfil” (H.
I to... I 0 HM
‘3 Jfl-lb ”5|
.0 .2 . 'u

no
...
y K . :_ Al. 5‘.
. ~ 3. 090‘. I .
r .1 2%. an...” .1, 1| :5 l
. . .2 .Mflfialrflwfi. .33....5v m1}

Lust-r1

f in. G... t 94.50.." .. «3.4”er r haunmmummaqflrr

kHz...» (..,..ln...¢.lcxo VIRAJWJWIS-fil: .. . .nl . . 5..
Hanna]. .w. ,.n..r.l- v.1: 2...." 1.... .. 4 . v: w... rm
u-!4...‘., .J. 2...;.....fi 2.- . u .

 

F . 1 I,

l_m.uaxm.zr ‘I ‘ , - I .I .'-‘..‘ '1

 

 

 

This is to certify that the
dissertation entitled
Cha/Lacte/bézaLéon and Eliza/titan Ema/Lg y L044

Spec/Unabcopy on Mill and NiMo Supwflwttéceb

presented by

Sami Hu/se/én Mahmood

has been accepted towards fulfillment
of the requirements for

Ph.D. degree in PhyA&C/3>

X14 6....
() Major Earofessor
\

 

Date Septembe/L 9, 1986

MS U i: an Ajfirnum‘w Action/Equal Opportunity Institution 0-12771

 

 

MSU
LIBRARIES
.m—

RETURNING MATERIALS:

 

 

 

 

Place in book drop to
remove this checkout from
your record. FINES will
be charged if book is
returned after the date
stamped below.

 

 

 

CHARACTERIZATION AND ELECTRON ENERGY LOSS SPECTROSCOPY ON
' NIV AND NiMo SUPERLATTICES

BY

Sami Husein Mahmood

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Physics and Astronomy

1986

Au
:9

 

nu.

Ifllul

‘ .
nus

alv

3:54

ABSTRACT

CHARACTERIZATION AND ELECTRON ENERGY LOSS SPECTROSCOPY ON
NiV AND NiMo SUPERLATTICES

By

Sami Husein Mahmood

NiV superlattices with periods (It) ranging from 15 to
803, and NiMo superlattices with 1\ from 1A to 1103 were
studied using X-Ray Diffraction (XRD), Electron Diffraction
(ED), Energy-Dispersive X-Ray (EDX) microanalysis, and
Electron Energy Loss Spoectroscopy (EELS). Both of these
systems have sharp superlattice-to-"amorphous" (S-A)
transitions at about A-17K. Superlattices with A around
the S-A boundary were found to have large local variations
in the in-plane grain sizes. Except for ea few isolated
regions, the chemical composition of the samples were found
to be uniform. In samples prepared at Argonne National
Laboratory (ANL), most places studied with EELS showed
changes in the EELS spectrum with decreasing A . An
observed growth in a plasmon peak at ‘10ev in both NiV and
NiMo as A decreased down to 19A is attributed to
excitation of interface plasmons. Consistent with this
attribution, the peak height shrank in the "amorphous"
samples. The width of this peak is consistent with the
theory. The shift in this peak down to 9ev with decreasing
A in NiMo is not understood. Three other effects are also
not well understood, but appear to be related to the

presence of sharp interfaces in the samples: (a) the

Sami Husein Mahmood
dominent peak at ~zuev remained fixed in energy, indepen-
dent of .A, for NiMo superlattices, but shifted upward from
21ev to Zflev in Mill; charge transfer was ruled out as a
possible explanation for this shift; (b) the sharp V
2p-93d "white lines" grew in intensity with decreasing 11
down to 19K, and then shrank again in the "amorphous"
samples; (c) the Ni "white lines" remained constant in
intensity independent of 1\. Samples prepared at Michigan
State University (MSU) did not show similar changes in EELS
spectra with A. Their difference in behavior is
attributed to their different growth conditions, which
apparently produced much less perfect interfaces than for

the ANL samples.

To my wife, Lisa, and to my family.

iv

ACKNOWLEDGMENTS

I would like to thank my advisor, Professor Jack Bass,
for his continuous guidance and support. I am thankful to
Professors G. Bertsch, C. Foiles, S.D. Mahanti, and S.
Solin for many helpful discussions, comments, and sugges-
tions. The technical assistance of V. Shull in maintaining
the FE-STEM was invaluable. Thanks are also due to Dr.
I.K. Schuller and J. Slaughter for providing LHS with the
samples. I am indebted to many of my colleagues, especial-
1y X.H. Qian for his help in x-ray studies and computer-
related problems, and N. Kedarnath for continuously being a
helpful friend. I wish to acknowledge the financial
support.<3f the National Science Foundation in this work.
Finally, I would like to express my gratitude to Yarmouk
University for providing me with a full scholarship which

helped me complete my studies and made this work possible.

TABLE OF CONTENTS

LIST OF TABLESOOOOOOOOOOOO.OOOOOOOOOOOOOOOOOOOOOOOOOOOOViii

LIST OF FIGURESOOOOCOOOO0.0...00.0.0...OOOOOOOOOOOOOOOOOOIX

CHAPTER ONE

CHAPTER TWO

CHAPTER THREE

INTRODUCTIONOOOOOOOOOOO.0.00.00.00.000000001

1.1 Previous Work..........................2
1.1.1 Previous Work on Struc-
tural Properties of NiV
and NiMo Superlattices...........3
1.1.2 Previous Work on Optical
Studies and EELS.................S
1.1.2.A Plasmon Structure
in Simple Metals.........5
1.1.2.8 Plasmon and Core
Excitations in
Transition Metals........7
1.2 The Present Thesis...........,........1O
1.2.1 Characterization of
the Samples.....................1O
1.2.2 Electron Energy Loss
Spectroscopy (EELS).............12

THEORYOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO0.0.1“

2.1 Collective Excitations of a
Metal Foil............................1u
2.2 Interface Plasmons at Bi-
metallic Junctions....................19
2.3 Collective Excitations in
Metallic Superlattices................25
2.3.1 Collective and Surface
Excitations in the Local Limit..25
2.3.1.A Collective Excitations..25
2.3.1.8 Surface Excitations.....29
2.3.2 Collective Excitations
in the Non-local Limit..........32

SAMPLE PREPARATION AND CHARACTERIZATION...35
3.1 Sample Preparation....................3S

vi

3.2 Structural Properties of Super-

latticeSOOCOOOOOOOOOOIII...COO...0.0.037

3.2.1 8/28 X-Ray Diffraction

(XRD).........................

3.2.1.A One-Dimensional
Diffraction Theory

of Superlattices......

3.2.1.8 Experimental Results

and Analysis..........
3.2.2 Electron Diffraction (ED).....

3.2.2.A The Electron Diffrac-

tion Camera and

Bragg Reflection......

3.2.2.8 Selected Area Diffrac-
' tion (SAD) and Micro-

diffraction (MD).....

3.3 Energy-Dispersive X-Ray (EDX)
Microanalys1s...0.0.0.000...0...
3.3.1 Theoretical Background....

CHAPTER FOUR EXPERIMENTAL RESULTS AND ANALYSIS...

Introduction....................
EELS of Pure Metals and Oxides..
EELS of NiV SUperlattices.......
u.3.1 The Energy Range <100ev...
u.3.2 The Energy Range >500ev...

51::
O
LON-4

u.3.2.A V Core Edge at S13ev.
u.3.2.8 Ni Core Edge at ~850ev.117

A.“ EELS of NiMo Superlattices in

the Energy Range <100ev...........

CHAPTER FIVE SUMMARY AND CONCLUSION................

APPENDIXA The FE-STEM.O0.0.0.0....OOOOOOOOOOOOOO...

LIST OF REFERENCESOOOOOO00....OOOOOOOOOCIOOOOOOOOCOOO

vii

.037

..37

..uz
..50

..51

..53

..66
..69

o .80
..80
..81
..92
o 092
.106
.110
.120
.128
.133

.1145

 

 

2.4
.1 .A

. ~

«(a

 

LIST OF TABLES

Table Page

 

3.1 Fitting Parameters for Structure Factor Cal-
culations of the Superlattices......................H2

3.2 The d-Spacing and the Miller Indices of the
Ni Planes Responsible for the Observed Ref-
lectionSOOO0.0.0.0....OOOOOOOOOOOOOOOOOOOIOOO00.000055

3.3 The Experimental d-Spacings of the NiV Super-
lattice with 1A-19A, Along with the Indices
and Calculated d-Spacings of NiV Lattice
Planes that are Within 2% of the Observed Ones......59

".1 NiV Samples Prepared at ANL........ ...... ...........93

viii

3.3

3.4

3.5

306

LIST OF FIGURES

Cross-Sectional View of the Field Lines of the

(a) Symmetric and (b) Anti-symmetric Surface
Plasmon Modes in a Metallic Film.............

Dispersion of the Symmetric (wi) and Anti-
symmetric (w_) Surface Plasmons..............

Dispersion of the Interface Plasmon at a
Bi—Mg JunctionOOOOOOIIOOOOO0.000000000000IIOO

Schematic Diagram of an Infinite Superlattice.

Dispersion of Interface and Surface Plas-
mons in a Niv Superlattice...................

Dispersion Relation of Bulk Plasmons in
Metallic Superlattices.......................

XRD Spectra and Structure Factor of ANL
NiMo superlatticeSOOOO0.0.0.0....0.0...0.0...

XRD Spectra and Structure Factor of ANL
"iv superlatticeSOOOO00......OOOOOOOOOOOOOOOO

XRD Spectra in Two Different Pieces of the
ANL19xNiMosuperlatticeOOOOOOOOOOOOOOOOOOOO

The Behavior of the Average d-spacing in
NiMo superlatticeSOO0.0.0.0...OOOOOOOOOOOOOOO

The Behavior of the Average d-spacing in
Niv superlattices.-0.0.0.0...I.OOOOOOOOOOOOO.

SAD of (a) Crystalline Ni, and (b) Amorphous
Fezr FilmSOOOOIOOOOOO...OOOOOOOO0.00.0.0.0...

MD oraCPYStalline N1 FilmOOIOOIOOOOOOOOOOOO
SAD of ANL 19A NiV Superlattice..............

SAD in Three Different Regions of ANL 19A
Niv superlattice......C......................

ix

...23

...26

.0061

SAD in Three Different Regions of ANL 60A
Niv superlatticel...OIIOOOOOOOOOOOOO0.0...

SAD in Three different Regions of MSU 27A
Niv superlatticeOCCO0....0................

SAD and MD in Two Different Regions of MSU
153N1V sampleoo00.00.000.000...0.0.0.0000

pr Spectrum in a Region of MSU 15A NiV

sampleOOOOO...I......OOOOOOOOOOOOOOCOOIOO.

Ni-to-V Thickness Ratio in ANL 19A NiV
superlatticeOOOOOOOOOOOOOOIOOOOOOOOOOOC.O.

Topology of a Measured Piece (89) of ANL
19A Niv superlattice....OOOCCCOOOOOOOOOOOO

Ni-to-V Thickness Ratio in MSU 27A NiV
superlatticeOOOOOOOOOOO...OOOOOOOOOOOOO0.0

Mo-to-Ni Thickness Ratio in MSU 22A NiMo
superlatticeOOOOOOOO...OOOIOOOOOOOOOOOOOOO

Dielectric Function, Loss Function, and
EELS 0f AIOOOOOOOOOOOOOOOOOOO..00....0.0.0

Dielectric Function, Loss Function, and
EELS Of Mo....OOIOOOOOIOOOOOOOOOO0.0.00.00

Loss Function and EELS of V...................

Contaminated and Uncontaminated V EELS Spectra.

Loss Function and EELS of Ni..............

EELS of ANL 19A NiV Superlattice and NiV
pseudo-.allOYOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

EELS Of ANL NiV sampleSOQoooooooooooooooco

8F Image of a Nonuniform Region of MSU 27A
Niv superlattice....OCCCCCOOOOO00.0.0.0...

EELS of MSU NiV Samples...................

Schematic Diagram of L2’3 Core Excitations
in a Transition Metal.....................

112,3ExaltationSOfVooooooooooooooooooooo
Intensities of the V WLs in NiV Samples...

X

.062

..68

O .7”

..78

..82

..85

.087

..95

..97

.103

.105

L2’3 EXCitationS Of NiIOOOOOOOOOOOOOOOOOOO
EELS of ANL NiMo Samples..................
The Principle of the FE-STEM..............

Schematic Diagram of the Electron Optical
COlumn in the FE‘STEM.OOOOOOOIOOOOOOOOOOOO

xi

.119
.121

.13U

.135

 

 

I
I I I... I r . a.
. ..o .2 .ni. a: Qv A.» .o a
I . _ E E . .. :c a. i . 3. :3 a e n. 9 L
U "I n . 3. a. S e .J n.. ~ an 2 ..u . 1 a .1... 3. my.
5 r. “J. a: 5 u ad .3 {a x is D. an e P» o a\v Ow

 

CHAPTER ONE

INTRODUCTION

The production and perfection of artificial metallic
superlattices has recently been of great interest.
Evaporation [1] and sputtering [2] have been used to
prepare metallic superlattices. Although Molecular Beam
Epitaxy (MBE) was previously used primarily to prepare
semiconductor superlattices [3], it has also recently been
used to prepare single crystal metallic superlattices [A].

The interest in metallic superlattices stems in large
measure from the fact that it is possible to produce
superlattices with novel magnetic [A,5,6] and superconduct-
ing [7] properties. The possibility of applying them as
x-ray [8,9] and neutron [10] mirrors is another potentially
interesting application of these structures.

Recent experimental work on metallic multilayers, with
emphasis on superconducting systems, was reviewed by S.
Ruggiero [11 1. Several physical properties, such as
elastic, magnetic, superconducting, and transport proper-
ties were reviewed by I.K. Schuller and C. Falco [2]. To
our knowledge, there are no published Electron Energy Loss
Spectroscopy (EELS) studies of metallic superlattices.
Optical studies [12], x-ray photoemission (XPS), x-ray
absorption (XAS) [13,111], EELS [15,16], and synchrotron

radiation [17] were employed to obtain information about

2
the electronic structure of pure metals and alloys.

The Vacuum Generators (VG) model H8-501 Field Emission
Scanning Transmission Electron Microscope (FE-STEM;
permits the measurement of energy losses from 0 to 2Kev in
a single spectrum, with resolution <1ev. An electron beam
energy up to 100Kev can be used, making it possible to
measure transition metal samples that are ~HOOK-thick
without the need to correct for multiple scattering. The
beam size can be as small as SA, permitting measurements of
small particles. Energy Dispersive X-Ray (EDX) analysis
and electron diffraction facilities allow the characteriza-
tion of the small regions under investigation [18]. A cold
stage is available to minimize the migration of surface
contamination to the region under investigation.

In this thesis, we will present local EELS measure-
ments on NiV and NiMo superlattices in tune energy range
<100ev and >500ev. We also present 8/29 x-ray diffraction
(XRD), electron diffraction (ED), and energy-dispersive

x-ray (EDX) measurements.

1.1 Previous Work
In this section we discuss previous work in the
experimental areas related to the subject of this thesis.

Structural analysis for NiV and NiMo superlattices is

 

TA more detailed description of the FE-STEM system is
given in Appendix A.

prev
work

A:
A.»

Pb

c;‘.‘.
On.

Q~

Q-
bu
.b

Clscv

\ a.
O»
h.

the r

3
summarized. Optical and energy loss measurements are
reviewed for some pure metals and alloys. We also review
previous work on interface plasmons.

We focus on NiV and NiMo superlattices in the present
work 1Wn~ three reasons: First, samples of both superlat-
tices with a wide range of periods (xi) were kindly made
available by Dr. I.K. Schuller of Argonne National Labora-
tory (ANL); second, they were characterized by Schuller's
group using x-rays and found to have both long range
coherence perpendicular to the layers and fairly sharp
interfaces; third, the fact that these two superlattices
have a common element, Ni, allowed us to test the effect of
different electronegativities of the second constituent of
the superlattice (Mo or V) on the EELS data, as will be

discussed in conjunction with the experimental results.

L;1.1 Previous Work on Structural Properties of NiV
and NiMo Superlattices

 

In 1982, Khan et a1. [19] studied the structural,
elastic, and transport properties of NiMo superlattices
prepared by Sputtering on Mica and Sapphire substrates.
X~ray diffraction (XRD) in reflection geometry showed that
these superlattices are composed of layers of FCC Ni and
8CC Mo, oriented along the [111] and [110] directions,
respectively.

For .A>200A, two lines were observed, corresponding to

the (110) Plane of ”0 (dMo-2.22R) and the (111) plane of Ni

4U

{\

obse

supe

x-ra

 

nr~

9.

Q»
3

cs
AV
\H.

u

(dN1-2.03A). For A<16.6A, only one broad peak was
observed, reminiscent of an amorphous structure. For
superlattice periods in the range 130AZ_ A 116.6A, sharp
x-ray peaks were observed with satellites consistent with a
structure having well-defined layers. From the widths of
the x-ray lines, the coherence perpendicular to the layers
was estimated to be “1003.

As the superlattice period was varied from >100A down
to 16.6A, changes in the positions of the central lines
indicated a 2% expansion in the average d~spacing perpen-
dicular to the layers. For samples with 1A<16.6A, the
center of the broad "amorphous" peak showed a contraction
back to the average d-spacing corresponding to samples with
large A . In an attempt to find the source of this
expansion, two simple models were used to calculate the
structure factor of the lattice. In the first model, a
uniform expansion throughout the lattice was assumed. In
the second, the expansion was assumed to be due to a single
anomalous spacing at the interface. The intensity analysis
based on both models was found to be consistent with the
experimental intensities; hence, XRD spectra were insensi~
tive to the nature of expansion assumed in the models.

Khan et al. also analyzed the in-plane structure of
some equilayer thickness samples using XRD in transmission
geometry. They found that the samples were laterally

polycrystalline. Moreover, they found a slight (0.6%)

5

expansion in the Ni d-spacing, and a (0.5%) contraction in
the Mo d~spacing parallel to the layers. They also
performed reflection XRD in the small angle scattering
region and observed only odd order diffraction peaks, from
which they concluded that the composition modulation was
close to a square wave (i.e., equal layer thickness of the
two constituents with sharp interfaces).

In 1985, Homma et al. [20] studied the structural
properties of NiV superlattices, using x-ray and neutron
diffraction. The high~angle reflection peaks (around the
central peak) exhibited structural characteristics similar
to those of the NiMo system. From rocking curve measure~
ments around the first order (small angle) Bragg reflection
peak, they deduced a columnar structure for the growth of
the samples, with a coherence length “7AOA perpendicular to
the layers, and 50A in the plane. The in-plane grain size
of 50A estimated from transmission 8/28 scans near the V
(110) peak was consistent with the size estimated from the

rocking curve measurements.

1.1.2 Previous Work on Optical Studies and EELS
1.1.2.A Plasmon Structure in Simple Metals
Since the work of Pines and Bohm [21,22] in the early
19508, electron energy losses experienced by fast electrons
passing through material foils have received great inter-

est, both from the theoretical [23-25] and experimental

 

”71

2.
Z

O c
v‘.‘
A .u

6
[26] points of view. The interpretation by Pines and Bohm
[22] that some of the energy losses are due to plasma
oscillations or "plasmons" was strengthened by the observa~
tion of plasmon excitations in simple metals such as Al
[26], Na and K [27] and in semiconductors and semimetals
[28]. The plasmon energies of the above mentioned
materials were found to be in excellent agreement with the

classical free electron plasma energy [29]:

(1.1)

 

where n is the density of the free electrons, e is the
electronic charge, and m is the electronic mass.

Stern and Ferrell [30] predicted the existence of an
interface plasmon localized at the interface between two
materials A and 8.

This plasmon energy is given by:

(1.2)

where wA and ”8 are the energies of the plasmons of the two
materials.

This interface plasmon mode was first observed in
Bi-Mg bilayer films by Miller and Axelrod in 1965 [31].
They failed, however, to observe such a plasmon mode in
Al-Mg bilayer films, which they attributed to blurring of
the interface between A1 and Mg. Their calculations showed
a negligible diffusion between Bi auui Mg, and a 2AA

diffusion length between A1 and Mg during the measuring

 

 

I
I . e . a u c. n g P. ”Aw 1|:
. . . .. . a” . . I. . a. I S n. 3 e 9 S n. e M. V. MW: .1 n. 0U
rug «C a» O r». 3.. e a a a. a. u-» u u {a a. to s A» a. e n 0%.
4/. AC 3 .. . at a... a c x -n .. . . an a .n h... a {A .n... Ti H. .. i
rIL r- L o ow r..- n .1. m n.- n. e .I. m . ..l mu :u T In.- s u. a due

6
[26] points of view. The interpretation by Pines and Bohm
[22] that some of the energy losses are due to plasma
oscillations or "plasmons" was strengthened by the observa-
tion of plasmon excitations in simple metals such as Al
[26], Na auni K [27] and in semiconductors and semimetals
[28]. The plasmon energies of the above mentioned
materials were found to be in excellent agreement with the

classical free electron plasma energy [29]:

(1.1)

 

where n is the density of the free electrons, e is the
electronic charge, and m is the electronic mass.

Stern and Ferrell [30] predicted the existence of an
interface plasmon localized at the interface between two
materials A and 8.

This plasmon energy is given by:

(1.2)

where wA and ”8 are the energies of the plasmons of the two
materials.

This interface plasmon mode was first observed in
Bi-Mg bilayer films by Miller and Axelrod in 1965 [31].
They failed, however, to observe such a plasmon mode in
Al-Mg bilayer films, which they attributed to blurring of
the interface between Al and Mg. Their calculations showed
a negligible diffusion between Bi and Mg, and a 211A

diffusion length between Al and Mg during the measuring

L. r. g». Q»
fig. a as.

U in,“ O 0
3 tv C n

 

 

 

7
time of 50 seconds. These calculations seem to support
their interpretation. Interfame plasmons were later
observed in several other systems, such as Ag-Mg [32],
Al~Cd and Ai—Cds [33]. and A1~Ag [3A].

The properties [35] and the dispersion [36] of these
plasmons have been studied theoretically. In the theory
chapter, we discuss in greater detail the interface
plasmons at metallic bilayer Junctions, as well as in

superlattices.

1.1.2.8 Plasmon and Core Excitations in Transi
tion Metals

 

Transition metal energy loss spectra are more complex
than the previously discussed simple metals, due to the
complexity of their electronic structures. Since there was
no single theory until the late 19703 that described and
interpreted the various peaks that occur in transition
metal spectra, experimental work was the only source to
shed light upon the dielectric properties of these metals.
Optical [37,38] as well as EELS [39,40] techniques were
used extensively to study these metals.

In the beginning of the transition metal series, the
spectra are relatively simpler than those at the end of the
series, where the number of peaks in the spectra increases.
Thus, while there is agreement between the various experi-
mental groups on the assignment of peaks in the EELS of

metals imi‘the beginning of the transition metal series,

 

assig

H

I)

r?
10
l

 

8
such as 11, 17, and Mo, there is some disagreement on the
assignment of some peaks in the more complex spectra of
metals as Fe, Ni, and Cu.

The peaks in the V spectra at ‘11 and 22ev were
assigned to volume plasmons [u1-u3], the first due to the
collective excitation of the us electrons [AU], and the
second due to the excitation of all (As+3d) electrons. The
two peaks in Mo Spectra at ~10.5 and 2u.5ev were also
assigned to volume plasmons [37,uo,u5], the first due to SS
electrons and the second due to (53+ud) electrons.

The Ni spectrum is more complex, with several peaks.
Most investigators [38,”1,u6] assigned the peaks at ~8 and
~20ev to volume plasmons. The ~27ev peak was assigned to
interband transitions [A1]. Others [“6] calculated the
optical constants from their EELS and found that the
optical absorption coefficient p, and the imaginary part of
the dielectric constant,1£2, are weaker for the 27ev peak
than for the 20ev peak, from which they concluded that the
assignment of the 27ev peak to an interband transition is
not justified.

As will be discussed in conjunction with the data,
Ksendzov [“7] suggested a model for interpreting the energy
loss spectra in which all three peaks in Ni, as well as the
two previously discussed peaks in V and Mo, are due to
volume plasmons.

The peaks in the energy range 35~70ev were assigned to

Qv

:u

 

an.

Rd.

‘1.

0‘

t;

u

6c
AV

9
single particle transitions from the 3p shell (M2,3) in N1
and V, and to transitions from the up shell (N2’3) in Mo
spectra. These assignments are consistent with atomic
energy level calculations [AB].

Excitations of core~level electrons, as in the 2p-$3d
transitions, provide information about the unoccupied
density of states (DOS) in the conduction bands of solids,
particularly the d-bands [49].

More than a decade ago, a superposition model [50] was
suggested to describe the DOS of binary alloys. This model
is based on charge transfer between the two constituents.
The model was found to describe the experimental DOS in
Ni~Rh alloys [51]. Good agreement between the theory and
experiment was obtained by assuming charge transfer from Ni
to Rh. On the other hand, XAS studies of Ni-Cu alloys
indicated no change in the number of Ni d-holes despite the
decline in the saturation magnetic moment of tuna samples
with increasing Cu concentration [52].

Recently, high-resolution EELS data involving the 2p
to 3d transitions were reported [53]. Strong deviations
from the single~particle transition probability calcula~
tions were found. Such discrepancies include [5“]:

1) The deviation of the intensity ratio of the P3/2
(L3) and P1/2 (L2) transitions from the statistical ratio
of 2:1 [55]. This deviation was also found by Leapman et

a1. [“9].

39:1

 

excha

 

P
a.»
6,9
n3 9» p
I1 I a
all a
s

10

2) The deviation of the line shape from that expected
for the DOS above the Fermi level, in that the intensity is
suppressed near the threshold and enhanced further above
it.

3) The deviation of the apparent spin-orbit splitting
(inferred from the separation of the L2 and L3 peak maxima)
from that observed in 2p core-line photoemission.

.A theory for XAS of the beginning of the transition
metal series (below Cr) was presented [5”]. First princi-
ples band structure and atomic data were used as input for
the theory, and the spectra for Ca, Ti, and V were calcu-
lated and found to be in good agreement with the experi~
nmntal data. This theory is applicable to metals with
d-band widths much larger than the d-d electron Coulomb and
exchange interactions. For the three metals considered,
the typical band width (“6ev) is much larger than the d~d
electron interactions (”1ev) [5A]; hence, the d-d electron

interactions can be ignored.

1.2 The Present Thesis

1.2.1 Characterization of the Samples

This project started when we received a few NiV
samples from I.K. Schuller of ANL. Since the samples were
prepared several months before we received them, it was
important to check the structural characteristics of these

samples. Three different techniques were used:

11

1) XRD: As mentioned in Section 1.1.1, Khan et al.
studied the structural properties of N11“) superlattices,
and found a 2% expansion in the average d-spacing perpen-
dicular to the layers as (A decreased down to “17A, and a
contraction back to approximately the bulk value as A
decreased down to “15A. We studied the structural charac-
teristics of NiMo and NiV samples to investigate any
structural changes of the samples with time.

2) EDX microanalysis: Since XRD studies do not provide
information about the local composition of the samples, EDX
microanalysis was carried out to check the uniformity in
the composition of the samples. The EDX analysis is
important to separate effects of contamination, or changes
in the relative concentrations of the two constituents,
from possible superlattice or interfacial effects.

3) Electron Diffraction: Selected Area Diffraction
(SAD) and microdiffraction (MD) studies of the samples
using a 100-Kev electron beam provide information about the
in-plane structure of the samples. We studied the electron
diffraction patterns of the superlattices and the pure
metals: we estimated the crystallite sizes (i.e., the
in-plane coherence) from the widths of the diffraction
peaks, and 1H3 made qualitative conclusions about the

in-plane structure.

12

1.2.2 Electron Energy Loss Spectroscopy (EELS)

In this thesis, we focus on three major effects:

1) The interfaces: As discussed above, energy loss
spectra of bimetallic junctions show new resonances
attributable to interface plasmons. NiV and NiMo superlat-
tices with a range of periods that cross the superlattice-
"amorphous" border provide a useful system for studying
interface plasmons. The usefulness of these systems stems
from two reasons:

a) the availability of superlattices with different
periods provides samples with different numbers of inter-
faces per unit volume; thus, we would expect to observe
interface resonances with intensities that scale with the
number of interfaces per unit volume (i.e., with 1\).

b) The availability of samples that cross the crystal-
line: (layered)-"amorphous" (nonlayered) border provide an
internal check for the interpretation of these resonances
being due to interfaces, since, as we go from layered to
nonlayered structures across the border, we would expect a
sudden (“w“) in the intensity of resonances attributed to
interfaces.

2) The effect of layering: The additional periodicity
in the z direction (perpendicular to the layers) introduces
bands of collective excitations of the superlattice due to
coupling of the interface plasmons, as will be discussed in

Chapter 2. If splitting of the plasmon bands is larger

13
than the instrumental resolution, it should be possible to
see several interface excitations in the EELS spectra. The
effect of layer thickness on these bands can also be
investigated.

3) Charge transfer: When two metallic slabs are
brought into contact, charge is transferred from the metal
with the lower in: that with the higher Pauling electro-
negativity [56]. Since charge transfer occurs in the
vicinity of the interface between the two metals (’10A)
[57] we would expect to see a large effect in superlattices
with small periods. Thus, the positions of the plasmon
peaks in EELS spectra possibly reveals information about
the charge transfer between the two metals. In addition,
charge transfer from one metal to another in a superlattice
consisting of two transition metals (e.g., from V to Ni in
NiV superlattices) leaves more d-holes in one (the V) and
fewer holes 1J1 the other (Ni). Thus, EELS in the L2,3
energy range (>500ev) provide information about charge
transfer between the two metals. To check for charge
transfer, we measured: (a) Two systems in the energy range
<100ev; NiV, in which the two constituents have different
Pauling electronegativities (1.8 for Ni and 1.6 for V). and
NiMo where the two constituents have essentially equal

electronegativities; (b) The V and Ni L2,3 WLs.

CHAPTER TWO

THEORY

In this chapter, we first review the theory of bulk
and surface plasmons in a metal foil. Second, we review
the theory of interface plasmons at bimetallic junctions.
Third, we review the theory of bulk and interface plasmon
bands in metallic superlattices, and we discuss the effect
of terminating the superlattice, and calculate the surface
plasmon dispersion relation. We also apply the appropriate
dispersion relations to NiV and NiMo superlattices and

calculate the plasmon bands and the surface plasmons.

2.1 Collective Excitations of a Metal Foil [15,58,59]
The dielectric function [15] of a free electron gas,

assumed real for simplicity, is given by:

if? 1
£(w,?) = 1 '- ~— (2.1.a)
w2 1 - g (Y.Vk/w)2

 

The bulk plasmon frequency is determined by the condition
e-O. Hence, the dispersion relation of bulk plasmons in

simple metals is given by:

w2 - w2 + 82k2 (2.2.a)

32 - 3 v2 (2.2.0)
Here wp is the bulk plasmon frequency given by equation 1.1

1H

an

is

n u
n1 n . A“...
a. to . .
.1 a»

1m

 

0»

iv

an.

 

II

D. S . . . ,
F. a #u
e o n .1 5 n a» h.» h
.nfl. . h.» 9 Hi 0...
“I ..d a x .n . e F m

b - ad. A: 6 M NH“ ON..- s mu.»

V L A‘

15
and V} is the Fermi velocity. In a simple local theory, 8
is set equal to zero. The dielectric constant in equation

2.1.a then becomes:

2
W

e(w)-1- 423 (2.1.b)
N

which will be assumed for all metals, unless otherwise
mentioned.

In the long wavelength limit (retardation effects are
ignored), the collective excitations produce a macroscopic
electric field which can be derived from an electrostatic
potential f(Y,t). The electrostatic potential satisfies
Laplace's equation everywhere outside the metal:

72f(Y,t) - 0 . (2.3.a)
and the following relation inside the film:

¢(w)72r‘(i’,t) - 0 (2.3.13)
where only one Fourier component of f(Y,t) is considered.

Because of translational invariance parallel to the
two surfaces of the film, the two~dimensional wave vector F
parallel to the surfaces is a good quantum number of the
excitations. The electrostatic potential everywhere can
then be written as:

f(f,t) - f(z) exp(ikx-iwt) (2.”)
where the wave vector V is assumed parallel to the x~axis
for convenience.

For a: metal film with thickness d, two types of

elementary excitations which generate a macroscopic field

«(a
pit.»

RU

ah.

F.

ca

lo

fa

3U.”

anti:

2 .
at

RU

16
exist. One type conSists of the symmetric and antisym-
metric superposition of the surface plasmons at the two
surfaces, as illustrated in Figure 2.1 (from reference
[30]). The second consists of bulk plasmons. Now we
briefly discuss each type of excitation.

(1) Surface plasmons: A metal-vacuum interface
supports a surface plasmon with a wave vector independent
frequency ws given by:

£(ws) . -1 (2.5)
This relation can be readily obtained from the boundary
conditions at the metal-vacuum interface (2:0). Combining
equations 2.1.b and 2.5, we obtain ws . wp/21/2, as
originally derived by Ritchie [59]. These excitations are
localized at the surface, and take the form:

f(z) “ e-kz (2.6)

For a metal film with a finite thickness d, the two
surface modes couple to produce the symmetric (wi) and
antisymmetric (w-) modes illustrated in Figure 2.1. The

dispersion relation for these modes satisfies the relations

£(w_) - -coth (Eg) (2.7.a)
€(w+) = -tanh (£3) (2.7.b)

from which we obtain:

'kd

1 e )1/2 (2.8)

t
t p 2

This dispersion relation is plotted in Figure 2.2. The two

a)Symmetric mode.

 

 

1V+1V—:
I o 1 o

1
[A 1m:

Figure 2.1 Cross-Sectional View of the Field Lines of the
(a) Symmetric and (b) Anti‘symmetric Surface Plasmon Modes
in a Metallic Film

18

 

 

 

 

 

‘3 l l 1 I l
14 - -
(0+
12 - 4
F; — -—¢—‘-'——.—.-:
v ” ’
>- a - x/ (1).. .1
c9 /’
m ./
I!
m /
4 "l _
I
2 I
I
0 [_ l l l l l
0.0 0.5 1.0 1.5 2.0 2.5 3.0
kd
Figure 2.2 Dispersion of the Symmetric (wi) and Anti-

symmetric (w_) Surface Plasmons

19
branches approach ws for large kd, since in this limit the
plasmon wavelengths are sufficiently small that the two
surface modes decouple [30].

(ii) Bulk plasmons: A amtal film also supports bulk
plasmons of longitudinal character with frequencies for
which £(w)-0. In such a case, Laplace's equation does not
hold in the film. The boundary conditions require that the
electrostatic potential vanishes identically outside the
film, so that the excitations do not produce a macroscopic
field outside the film. Thus, the electrostatic potential
inside the film has the form:

f(z) ‘ sin (2%!) (2.9)

and one has standing wave resonances inside the film, the
frequencies of which are identical to those for an infinite

dielectric.

2.2 Interface Plasmons at Bimetallic Junctions
[30.60.35.361

Using an argument similar to that followed in the
preceeding section, Stern and Ferrell [30] predicted the
existence of interface plasmons localized at a metal-metal
or metal-dielectric interfaceu The eigenfrequencies of
such modes are given by the relation:

£1(w) + 62(w) - 0 (2.10)
This equation is the generalized form of equation 2.5, and

gives the constant interface frequency:

 

 

_. . .I . u . a» a» n.“ |.—1. .
W e , n . Ah» ”0 AV .\ Q 9 U .l t ‘1: e
D” d 1. w a 1; a. a. a» a. ...v n. or. a n
.n n U a C e .7. . . 9 .1 a... a. n: a x e.
a .3 C .I n. H a. .C so V 0 Ti. Me L1.
..u

20

 

(2.11)

where w1 and w2 are the bulk plasmon frequencies of metal 1
and 2, respectively.

The hydrodynamic model, including dispersion of the
bulk plasmon given by equation 2.2 was later used to
calculate the dispersion relation of these modes. The
following analysis will closely follow the analysis of

references [35,36]:

0.
N

)a' - 8217(1-7..D’) - (3)17} - o (2.12)

(

N

dt
where E3 is given by equation 2.2.b, U is the macroscopic
electron displacement vector from eqilibrium, and 62-63
defines the displacement potential 3.

The bulk plasmon frequency is assumed to have the form
given 1J1 equation 2.2 (the dispersion for small wave
vectors), and for later convenience, we adopt the following
definitions:

82p2 - wp2 + 82k2 - w2 w2 <wb2 (2.13.a)

82q2 - w2 - wp2 — 32k2 w2 >wb2 (2.13.6)
In these two energy regions, both g and f satisfy the

relation:

N

2

(—‘ll - k2)(——d ~12)g(z) - o (2.111)
2 2
dz dz
2 p2 for w2< w:
where ‘ j -

 

Av

n.
0.

\

:he 0

Hhera
and

f and g are related by:

21

2
f(z) - E(82-2- - 82k2 + w2)g(z) (2.15)
e 2
dz
In the following discussion, we let metal 1 occupy
the space z<0, and metal 2 z>0, and subscripts 1 and 2
refer to these metals. We also assume ”b1<wb2' In the
energy range w§1< w2< ”:2’ the general solution for
equation 2.1M is:
kz
g1- Acos(q1z +¢ ) + 8e (2.16.a)
‘P Z _
32- Ce 2 + De “2 (2.16.b)

The electrostatic potential inside the metals is derived

from equation 2.15.
conditions,

eigensolutions satisfy

coscp + sinqp

where

2-

2
L(w,k) - w1q1(w2

and

2 2 2 2 2
K(w,k) - kw (w2 w1) p2w2(w

The condition K(w,k)-O

(w$+

exponentially decaying

that starts at [g

in metal 1. This mode

Then applying the appropriate boundary

it was found in reference [36] that the

the relation:

0 (2.17)

w2)(w:+ wf- 2w2) (2.18.a)
2 2 2 2_ 2
w1)(2w w1 w2)

(2.18.6)

corresponds to an interface mode

2 1/2
w2)] .

mode in metal 2 and a pure cosine

'This branch represents an

was first observed by Miller and

sol
fouf

A. S a:
11. n. .6;
a 3 M.”

O .. a i

 

Axe

 

Q

As
the

22
Axelrod at a Bi-Mg interface, as discussed in Chapter 1.
Equation 2.17 can be solved, and Figure 2.3 shows the
solution for several values ofcp in the first and
fourth quadrants for the Bi-Mg system. Those curves were

calculated using the simplified solution of equation 2.17

 

 

for the primary branch which has the value w1_2 at k-O.
This solution takes the form:
2
w + w
w2 - -l-3-3'(1 + c1k + c2k2) (2.19.a)
[2(w: - w$)]1/2cos¢
c1 - 2 2 (2.19.b)
[w2 cosm / 82 + w1 sin¢ / 81]
(w:- 3w$)c$
c2 - (2.19.0)

2 2
2(w1+ w2)

As shown in Figure 2.3, the modes are clustered close to
the ¢-0 solution, and each mode arises from a relatively
narrow plasmon band.

Jewsbury [2.3] has provided a rough estimate of the
conditions for observing an interface plasmon using the
following argument. For k below the cut-off wave vector
kc, the predominent excitation is the collective excita-
tion. Above kc, single-particle excitations become
important, and plasmons decay into electron-hole pairs
(Landau damping). The other cut-off wave vector kD, the

Debye cut-off, represents the upper limit for the existence

23

 

 

 

 

 

 

 

2° 1 l 1 I 1
18 - ¢-—ma -4
--/o
is - Lo". -1
A—sfi-W/o
14 1- ’ ‘P-m -
>
3 12 — 4"" -
>- 10 - -
CD
a . - -
I:
“J B _ _
4 - —
2 -' .—
0 l l l 1 i
0.0 0.2 0.4 0.3 0.0 1.0 1.2
—1
k (A )

Figure 2.3 Dispersion of the Interface Plasmon at a Bi-Mg
Junction

 

\/

2A
of bulk plasmons. These cut-off parameters are given by:

W

 

F1 r
s
201‘18
kD- 1.26kF- rs (2.21)

Using the dispersion relation of bulk plasmon in small k
region given by equation 2.2, one can roughly estimate
those systems where we can observe interface plasmons. The
dispersion relation of the bulk plasmon combined with
equation 2.20 gives:

w2 i w

1_2 (2.22)

2 2 2
1+ 0.6w1- 1.6w1

This relation shows that we should observe an interface

plasmon with frequency w anywhere between w (the lower

1’2

plasmon of the two metals) and 1.26 w

1

1. For higher w1_2,

the plasmon will suffer from Landau damping. Beyond kD,
the interface plasmons are forbidden.

IN) the NiV and NiMo systems, Ni has a plasmon with
energy ”8ev, V a plasmon with energy ”11ev, and Mo a
plasmon with energy ~10.5ev. Equation 2.22 shows that
interface plasmons at “9.5ev for both NiV and NiMo systems
are in the allowed energy regime. Higher plasmons (at

>20ev) are outside the spectral region 61/5250, where

interface plasmons may exist [58].

25

2.3 Collective Excitations in Metallic Superlattices

For a metallic film there are two types of plasmons:
one type 143 the symmetric and anti-symmetric surface
plasmons (Figures 2.1 and 2.2); the second type consists of
bulk plasmons propagating in the plane of the film. For a
superlattice, bands of the first type of plasmons
can be obtained assuming a local theory. Bands of the
second type require nonlocal theory (8&0 in equation 2.2).
Nonlocal effects yield extra splittings in the interface

plasmon bands obtained in the local limit.

2.3.1 Collective and Surface Excitations in the Local
Limit [58,61]

In tniis section, we make use of the local theory to
investigate interface plasmon modes of metallic superlat-
tices. In this limit, we recall that 8 in equation 2.2 is
set equal to zero, and the dielectric function of the
metals is given by equation 2.1.b. We first study the
collective excitations of an infinite superlattice; second,
we study the surface plasmons localized at the metal-vacuum

interface of a semi-infinite superlattice.

2.3.1.A Collective Excitations
We first consider an infinite superlattice illustrated
:Ui Figure 2.u. As discussed in Section 2.1, an isolated

slab of metal 1 or 2 supports standing wave resonances

(with frequencies for which 61(w) or 52(w) vanishes) which

26

A «1\\\\.\\\I§_I 1]"

B 2 d

A K\\\\\\\\
A \_\\\\\\X

B

 

 

 

 

Figure 2." Schematic Diagram of an Infinite Superlattice

 

 

 

 

van:

Ever

I
' I.
In

I
—

(D

I.)
(J
(D

(I
—

 

27

generate a macroscopic field confined within the slab.
Thus, in a superlattice such as that in Figure 2.A, each
slab possesses bulk plasmons as in the isolated slab. In
addition, there are collective excitations of the whole
structure with frequencies w such that neither metal has a
vanishing dielectric constant. INT such a case, the
electrostatic potential satisfies Laplace's equation
everywhere. Combining equations 2.3.a and 2.11, f(z) is
found to satisfy the differential equation:

2 ,
(Lg — k2) f(z) - o (2.23)

dz
Because of the periodicity in the z-direction, f(z) must
satisfy the proper boundary conditions at each interface.
Hence the solution must form a Bloch wave, with respect to
translations in the z-direction. Thus, f(z) takes the
form:

iqz

f(z) - e Uq(z) (2.2U)

and for any integer n, Uq(z) satisfies the relation:

Uq(z + nzl) - Uq(z) (2.25)

Thus, a general solution of equation 2.23 within the

slab of metal 1 at nA_<_z_<_nA.+d1 is:

Uq(z) _ e-iq(z-n1\)[A1ek(z-nA )+ AZe-k(z-n1\)] (2.26)
So that:
f(z) eian [A ek(z-nA.) + A e-k(z-n1\)] (2.27.a)

1 2

28
Similarly, in the slab of metal 2 at nA.+d1gzg(n+1)A
we have:

eian ek(z-nA -c1,)+

1 e"k(Z"flA '01)

f(Z) [B 32 ] (2.27.b)
We then apply the proper boundary conditions to these
solutions, i.e., the continuity of the electrostatic
potential and the normal component of the electric dis
placement vector across the interface between the two
metals at z-nA. and z-nA.+d1. We thus obtain four equations
involving the coefficients in equation 2.27. Then
setting the appropriate uxu determinent equal to zero leads

to the implicit dispersion relation:

6
6
[1+(-l)2]sinh kd1sinh kd2+22l[cosh kd1cosh kd2-cos q)\]=0
‘2 2 (2.28)

where the explicit reference to the frequency dependence
of the dielectric constant is dropped in equation 2.28 for

convenience. This equation has the solution:

 

6
1/
-l - -c(k,q)+[02(k,q) - 1] 2 (2.29)
62 "
where:
cosh kd cosh kd - cos qu
c - 1 2 (2 30)
sinh kd1 sinh kd2 '

For real solution of the eigen frequencies, we must have
c11. Hence, collective excitations occur in the frequency
region for which 61/5230. This is the spectral region
where surface plasmons may exist at the interface. The

collective excitations discussed above may be viewed as a

 

su;

 

area

only

dowr

 

 

 

29
superposition of surface plasmons which, for finite
layer thickness, couple to form normal modes of the whole

structure. Notice that in the limit kd2>>1, and a finite

kd equation 2.29 reduces to the dispersion relations of

1’

an isolated metal slab given in equation 2.7, as expected.
Equation 2.29 can be solved for the eigenfrequencies

and hence we obtain the explicit dispersion relation:

2 2 2 2
{w1+ w2 w1 w2 cosh k(d1 d2) cos qA 1/2 1/2

(2.31)

 

cosh kAJ : cos qA

The general solution 2.31 consists of two bands with a gap

This gap closes when d ad where we have

1-2' 1 2’

only one band with width (w2- w1) at kA =0, and narrows

around w

down to zero width for large k)\. Figure 2.5 shows the
dispersion relation of collective superlattice excitations

in the NiV superlattice with d1- d This dispersion

2.
relation is not affected by A .

2;}.1.8 Surface Excitations

We now consider the semi-infinite superlattice
occupying the half-space z>0, while the half-space z<0 is
vacuum. We are also interested in modes localized in the
vicinity of the interface at z-O. We obtain for such modes
a solution identical to that in equation 2.27, with q
replaced by LB, which implies that the solution decays
exponentially in the superlattice. In the region z<0, the

solution has the form:

 

ENERGY (ev)

FiEur
am)

 

3O

 

    

 

 

 

‘2 I I I I I
11 Collective mode -
10 -
9 _
3‘ 8 -
E, 7 1— _
_ __‘ Surface mode
>- 6 ‘ ‘~ ~ __________ -
(D _______
CE 5 — ..
%
UJ 4 ' ]
3 — —1
2 — _
1 '— —1
0 l l l l I
0 1 2 3 4 5 6
M

Figure 2.5 Dispersion of Interface and Surface Plasmons in
a NiV Superlattice

 

 

KI

Comb

013p

 

 

31

f(z) - C exp{+kz} (2.32)
Upon applying the boundary conditions at z'nA. and
z-n/i+d1 we obtain four equation involving the four
coefficients of equation 2.27. We also apply the appro-
priate boundary conditions at z-O and obtain two other
equations involving the coefficients in equations 2.26 and
2.32. We combine these six equations and obtain only three
equations involving A1 and A2. Those three equations are

then combined to obtain the following implicit dispersion

relations:

-kd
e‘fiA' - e 2[cosh kd1+ P1 sinh kd,] (2.33)
+kd
e‘BA' - e 2[cosh kd1 + P2 sinh kd1] (2.3a)
where:
P - ----— , P - -—-——-——
1 61(1 62) 2 61(1+€2)

Combining equations 2.33 and 2.3a we obtain the implicit
dispersion relation for the surface mode:

kd2 -kd2
2 cosh kdISinh kd2+ sinh kd1[P2e -P1e ]=0 (2.35)

Equation 2.35 gives two solutions for w. For d -d one

1 2’
solution is constant and has the value w the other

1-2’
starts at a frequency w1> w > w1/21/2, and approaches the

value of the surface mode of the metal in contact with
O

vacuum. Figure 2.5 also shows the surface modes for a NiV

superlattice with the Ni layer being in contact with the

‘Va

 

 

QGPSiC

32

vacuum at z-O.

2.3.2 Collective Excitations in the Non-local Limit

 

Eliasson et al. [62] used the hydrodynamic approach,
including non-local effects, to investigate the collective
excitations generated by coupling of bulk plasmons in
metallic superlattices. The equation of motion for the
density fluctuation n(z) was set up, and the differential
equations obeyed by n(z) were derived in both materials of

the superlattice. In the regime w Swgw the solution was

1 2
assumed to be:
Aeiqz + 8e”iqz ,Oizgd1
n(z) 3 (2036)
Ce”Pz + DePz ,d <z<1l

1

where q and p are given in equation 2.13. Bloch's condi-
tion (equation 2.27) was then used, and the continuity of
n(z) and the current density jz at the interfaces were
applied at z=0 and z-d1; thus, a set of equations for the
coefficients A, 8, C, and D were obtained. The dispersion
relation was then obtained from the zeros of the uxu
determinent of the coefficients.
Figure 2.6.a shows the dispersion relation of the bulk
superlattice plasmons calculated by Eliasson et al.,
2 2

assuming w2- 2w1, d2- 5d1, d1-100A, and the equilibrium

number density of electrons n1- 2.33x1017cm-3. This dis-

persion relation is very nearly that for the NiV system,

 

 

L4

10 I

 

OJ

R

[Igur
Metal]

33

a) b)

 

 

 

 

 

 

ENERGWGV)

 

 

 

 

 

0.0 0.5 1.0 0.0 0.5 110
kd1

Figure 2.6 Dispersion Relation of Bulk Plasmons in
Metallic Superlattices

30

since w2- 2 however the assumed n is much smaller than

w2

V Ni’ 1

for NiV. The large difference between the layer thicknesses
of the two constituents results in large band gaps; as the
layer thicknesses approach the same value, the band gaps
narrow down. Figure 2.6.b (estimated from figure 2 in ref.
[62]) shows qualitatively the dispersion relation of the
same system as in Figure 2.6.a, except with equal layer
thicknesses of the two constituents. The band gaps at k”0
are “0.2ev. We would not be able to resolve such small gaps
in EELS measurements with energy resolution ~0.7ev. Thus we
expect to observe a single plasmon peak :ui NiV having a

width of ~3ev, the difference between the bulk plasmons for

Ni and V.

 

64a

..
.1

:a

s

a:
a5

0»
n v
A,»
a:

 

 

stab.

CHAPTER THREE

SAMPLE PREPARATION AND CHARACTERIZATION

3.1 Sample Preparation

ANL sample preparation by sputtering was described
elsewhere [2,19]. Briefly, a target made of the desired
material i4; placed in the vacuum system. The system is
diffusion pumped to a base pressure of ‘10’7torr. An inert
gas (Ar) is then admitted to the chamber. A high voltage
is applied between the target and an anode to ionize and
accelerate the ions to strike the target and produce a beam
of sputtered atoms. Deposition rates are controlled by
stabilizing the power input to the target.

For the preparation of superlattices, two or more
sputtering guns are used. The guns are well separated to
prevent overlap of the two atom beams. Substrates (such as
Mica, Sapphire, NaCl, Si, etc.) are held on a temperature
controlled, heated, rotating platform which moves them from
one beam of atoms to the other. The layer thickness is
controlled by the deposition rate and the rotational Speed
of the platform, and monitored by a quartz single crystal
film thickness monitor.

A high degree of texture or epitaxy (i.e., coherent
growth in which each deposited atomic layer is in registry
with the previous layers)in the samples depends critically

on the substrate temperature, the Ar pressure, the deposi-

35

36
tion rates, and the target-to-substrate distance. Typical
conditions used for the preparation of ANL samples were:
10-15mtorr Ar pressure; 10-30A/sec. deposion rates; 8in.
target-to-substrate distance: and room temperature.

The MSU sputtering system and conditions are different
than those of ANL in the following ways. ‘The system is
cryo-pumped down to ~10'8torr, which produces a "cleaner"
deposition atmosphere. Masks are used and enable the
preparation of different samples with different periods and
possibly different constituents (from conductors to
insulators) due to the availability of four different guns,
some of which are fed with an RF power supply. The
platform with the substrates oscillates between the guns,
and the time a substrate spends in a given beam of atoms
can be controlled; thus, the layer thickness essentially
can be controlled without having to change the deposition
rate.

For preparing MSU samples, the Ar pressure, the
substrate temperature, and the deposition rates were chosen
to be similar to those used for preparing ANL samples. The
target-to-substrate distance used at MSU was Min. Longer
target-to-substrate distances usually moderate the energy
of the beam of atoms, so that the atoms arrive at the
substrate with energy that is too low to disturb the

epitaxy.

37
3.2 Structural Properties of Superlattices

3.2.1 0/20 X-Ray Diffraction (XRD)

XRD is widely used to study the structural properties
of both semiconductor [63] and metallic [6A] superlattices.
In reflection geometry, the momentum transfer q is
perpendicular to the surface of the sample (and hence to
the layers): thus, structural information perpendicular to
the layers is obtained.

Since the ANL samples studied in this thesis were
prepared several months, or even a year, before the time of
our EELS measurements, we thought it was important to
remeasure the 9/28 XRD characteristics of these samples, to
check that the interfacial information obtained by Schul-
ler's group [19] was still appropriate. We later extended
the structural characterization to samples prepared at MSU.
To analyze the data, we adopted the simple one-dimensional
diffraction model previously used [65,19] to fit the
experimental data. This model assumes sharp interface
boundaries and ignores effects of strains at the inter-

faces.

3.1.1.A One-Dimensional Diffraction Theory
of Superlattices

As a first approximation, we assume a perfect super-
lattice of constituents A and 8. The separation of the

atomic planes (the d-spacing) of these materials is d1 in

material A and d2 in material 8, as measured in the bulk.

ar

“(U

fly

as.

.U n
1. 3H
nun

090$

 

 

\I‘J b u 5 (v

,0 £ n. F. a. U

r L (H. HI a 04 an
1 t Fla 1 H.155 “U.

and?

38
The layer thicknesses of the two materials is assumed
equal, as is the case with our samples. The number of
atomic planes of the two constituents in their layers is N1
and N2.
The period (1\) of the superlattice is then given by:

A- d1 N1 + d2 N2 (3.1)

and the average d-spacing over the superlattice [19] is

given by:

5 - ---- (3.2)

X-rays falling on the sample will be reflected from
the planes of the two materials with planar scattering
powers F1 and F2, where the planar scattering power is
given by:

F - fa (3.3)
Here f is the atomic form factor and a is the atomic planar
density.

The form factor can be obtained from tabulated results
[66], and the atomic planar density can be calculated from
the lattice parameter a and the structure of the lattice.
Furthermore, it is known that in NiV and NiMo the BCC
lattices grow in the (110) direction, and the FCC grow in
the (111) direction [20]. The planar density of the BCC

structure is:

 

(3.u.a)

and that of an FCC structure is:

 

 

 

Sivi

”he:

iv;

55

31a:

39

u
aFCC ' 1/2 2 (3°"°b)
3 a

An incident x-ray beam with wave vector kg and wave
length A scatters from the one-dimensional lattice with
wave vector I. In the process, it transfers momentum 3 to
the lattice in the z direction. Since the scattering is

elastic, i.e.,IFJJFt we obtain:
q - anili—Q‘ (3.5)

The position evof the reflection peak of order n from
the lattice is given by Bragg's law:

2D sine - n A (3.6)
where D is a characteristic length that is determined by
the periodicity of the system. Hence, the position in
q-Space of that reflection can be obtained from equations
3.5 and 3.6:

27th
0 - D (3.7)

 

The scattering amplitude [3.5] from the lattice is
given by:

S(n) -,§ Fp.exp{i;b.3} (3.8)
where Fp is the scattering power from the pth plane as
given in equation 3.3, and {p is the position of that
plane.

Since the system is periodic in the z direction, the
contribution to the scattering amplitude of the pth plane
is equal to the contribution of the (N1+N2+p)th plane.

Applying this boundary condition to equation 3.8, and

 

Q.
n:

 

 

A0
keeping in mind that 3 is parallel to £5, so that Yb is
simply the z-coordinate of the pth plane, we obtain:
exp{i(N1 d1+N2 d2) 0} - 1 (3.9)
From which we obtain:

. 27ft!
q A

 

(3.10)

Equations 3.7 and 3.10 show that the period A' defines
the periodicity of the system, and hence D in equation 3.7
is simply A .

The d-spacing corresponding to the nth reflection is

then given by:
d - - (3.11)

From equation 3.2 for the ayemage d-spacing of the
lattice, and equation 3.11 for the d-spacing corresponding
to the nth reflection, we notice that there is always a
reflection corresponding to the average d-spacing CH? the
lattice, and that the order of that reflection is equal to
the sum of the atomic planes of one period in the superlat-
tice. In the model chosen, the position in q-space of the
main line (that corresponding to the average d-spacing) is
fixed, independent of the period of the superlattice, and
is given by equation 3.10.

In using equation 3.8 to calculate the scattering
amplitude from the crystal, we must bear in mind that the
dimension of the crystal in the z direction is finite.

Hence, we need to add up contributions from a finite number

1..

«V
q ..

‘5

t

PI‘

 

 

I! i I c a o firm

a
e n
a: kdie e n1 0
1 . .4 in“: a .18
K. c 6 i. c FM. Li

A1

of planes. The finite size of the stack of atomic planes
in the z direction (the coherence length) results in
broadening of the diffraction peaks. This length can be
viewed as the length of the crystal in the z direction over
which reflections from the atomic planes are almost in
phase, and hence add up constructively. In using equation
3.8, contributions from planes in a total of NT periods
will be added; hence, the coherence lengui is NTA.. The
scattered intensity as a function of q is then given by:

sin(qd1N1/2)

I(q)"1F1 + Fee
sin(qd1/2)

iqA l2

 

 

sin(qd2N2/2)|2 x
sin(qd2/2)

sin2(01\NT/2)

 

2 (3.12)
sin (qA /2)

To determine the planar scattering powers F1 and F2, we
tried various alternative fits to tabulated x-ray data
[66] for Ni, v, and Mo in the range 1.5 5 q 5 6.58'1, which
covers the range of q in XRD spectra of the superlattices.
The best fits were obtained with a exponential function of
the form:

F - Ac.-Bq

(3.13)
The fitting parameters A and 8 were evaluated. The various
parameters (other than A.) needed for the Spectral calcula-

tion are given in Table 3.1.

'0 “ v
a. e s
a w. n.
Cy x fiu.
n1... 0. Q.

we S

TU
An»
Q.
.1)»:
RV

 

O 1
HM A; .r a: c .
n. K e JV H a 1n“.
“1U b n ‘ o ‘19
w)! an“; O 6
‘Il .1. .1 It
e A VI A .ITIi ”Vt

H2

Table 3.1 Fitting Parameters for Structure Factor Calcula-
tions of the Superlattices

Parameter

d(X) A 8(A)

 

 

 

Material

Ni 2. 3 6.0 0.1uu
V 2.1A ”.0 0.153
Mo 2. 2 6.5 0.120

The value of A for Mo was found to be a little too
small to fit the experimental XRD spectra for NiMo super-
lattices, especially for A150A. The value (8.0) was found
to be nuuni better, and the resulting spectra agree very

well with the experiment.

3.2.1.8 Experimental Results and Analysis

 

Figure 3.1 shows typical experimental spectra for NiMo
superlattices prepared at ANL, together with the calculated
lines normalized such that the intensity of the strongest
peak is equal to that of the corresponding peak in the
experimental spectra. The calculated and experimental
spectra were superimposed for comparison. Figure 3.2 shows
the Spectra for ANL NiV superlattices with the calculated
spectra. In both systems we observe the same pattern of
evolution, which agrees with the calculation. Moreover, as
A becomes smaller than "15A, the spectrum shows no sharp
lines, but rather a broad peak, reminiscent of the
diffraction pattern of an amorphous structure. The width

of the "amorphous" Spectra indicate that the coherence

 

 

 

 

 

 

 

 

 

 

I l

.75
‘E
D
8
S

A-aoX
>.
t
w
z I
m 1
.- ‘g 1 1
E

[1.193

I
—/1\— l
M143.
I l
35 40 50

45
201099.)

Figure 3.1 XRD Spectra and Structure Factor of ANL NiMo
Superlattices

 

 

 

(Arb. Unns)

 

 

INTENSITY
>
(‘1.
01
)0

 

I
|
l
L .2
W

l l I l
33 4o 42 44 4s 48 50
9 .

20mg)

 

 

 

 

Figure 3.2 XRD Spectra and Structure Factor of ANL NiV
Superlattices

 

 

 

(Arb. Units)

 

 

INTENSITY
>
311
m
)0

 

 

I
I
I
1 i...

W

l l l l

33 4o 42 44 46 48 50
29(Deg.)

 

 

 

 

Figure 3.2 XRD Spectra and Structure Factor of ANL NiV
Superlattices

HS
length is ~20K. The transition from a layered structure to
a nonlayered "amorphous" one seems to be fairly sharp
(within “1 atomic plane).

The superlattices with 1&“193 are not very uniform.
X-ray spectra on different parts of the same sample show
that the diffraction pattern from some places on the sample
shows a single sharp peak (the main peak corresponding to
the average d-spacing), on top of a broad, amorphous-like
peak. This indicates the existence of disorder coexisting
with layered structure in the same sample. The weak signal
and the possibly diffuse interfaces in these regions of the
sample do not allow the detection of any satellite around
the main peak. Figure 3.3 shows the diffraction pattern
from two different pieces of the same NiMo sample of period
“1911.

Furthermore, the position of the main line in the
diffraction patterns was found to move down in q-space as
A decreased down to 19X. For zigj53, the center of the
broad peak moved back to approximately the position
corresponding to the samples with large A.. This effect
had been observed previously in the NiMo system [19].
Figures 3.4 and 3.5 show the experimental values of (1/5)
vs. A. for NiMo and NiV superlattices, respectively. The
positions of Bragg peaks from the substrates were used to
calibrate the q-axis.

Figure 3.h shows that, for large A , 1/3 reaches the

H6

 

NIMo
4:193

 

 

INTENSITY
(Arb Umts)

 

 

 

40 1 4'5 r I T t 50
29 (Deg.)

 

V I U '1' l' I I I

Figure 3.3 XRD Spectra in Two Different Pieces of the ANL
193 NiMo Superlattice

u?

 

1/6 (A I

l l l I I l I l I I I I
0-‘73 + KHAN 0t :1. T
0 47s A m. SAMPLEstwn was.”
' El wsu mstom was.)
00‘7‘ -
0.472_ ___ ___ __ __ __ __ __:
0.470 + 4’ é ‘
0.458 % + é -I
0.406 + a
4.
0.404 Lg _
:I-I +
0.482 ..
.1 +-L J l i l l l J. I l l

 

 

(L4BC
0 10 20 30 4O 50 60 70 80 90 100 110 120 130

A (A)

Figure 3.” The Behavior of the Average d-spacing in NiMo
Superlattices

118

 

0'“ I .. I I I I I I I I I I I
0,490 -+ HOMHA 0t 01. “
A “‘4. SAMES (M NEASJ
0 . 400 [3 H50 SAMPLES (0m «543.1
0.405 A _I
A 0.484 _
°$
0.482 .I
no _I&_._._.___ _ ___—__i
\
0.470 + -I
0.476 gfilé d
0.474 .—
o.472 J L l l I l l L l L l l

 

 

 

0

Figure 3.5
Superlattices

10 20 30 40 50 60 70 BO 90 100 110 120 130

Am

The Behavior of the Average d-spacing in NiV

"9
calculated value (dashed lines) using eqation 3.2. As A.
decreases, 1/3 decreases until .A‘19X, when the difference
between large A and small AK values of 1/5 reaches :21, in
agreement with the results of reference [19]. For A ”1&3,
the center of the broad peak moves back to approximately
the calculated value.

Figure 3.5 shows that the NiV system exhibits a
similar behavior. The change in 1/3, however, is “1%, half
of the effect observed in the NiMo system. For A -153, the
peak is broad and the uncertainty in locating its center is
large.

These experimentally observed expansions in the
average d-spacings for NiMo and NiV as .A decreases have no
analogue in the calculation. In addition to this expansion
perpendicular to the layers, a slight expansion in the Ni
spacing and a comparable contraction in the Mo spacing
parallel to the planes have been reported [19].

The coherence lengths of these systems were crudely
estimated from the experimental line widths of the central
Bragg peaks of the superlattices. The instrumental width
was estimated to be equal to the width of a sharp Mica
line. The line shapes of both the Bragg peaks and the
instrumental width were assumed to be Lorentzian; thus, in
correcting for the instrumental width, we simply subtracted
the instrumental width from the width of the central Bragg

peak of the superlattices. This procedure gave a coherence

50
length of “3003 for both the NiV and NiMo systems prepared
at ANL. This result is in qualitative agreement with the
coherence lengths of ~100K for NiMo system obtained by Khan
et al. [19], and of ~7003 for NiV system obtained by Homma
et al. [20]. MSU-prepared samples that showed superlattice

lines had a coherence length of 100-2003.

3.2.2 Electron Diffraction (ED)

The structure of materials can be divided into two
broad categories: amorphous and crystalline, according to
the degree of long-range order of the atoms in the mater-
ial. A material with no ordered structure is considered
amorphous, while a material in which the same structure
repeats over the entire macrosCOpic sample is said to be a
single crystal. Between the two extremes we find the
polycrystalline materials in which the atoms are arranged
in small crystallites that are usually randomly oriented in
the bulk of the sample [18].

A portion of the electrons passing through thin films
of matter are scattered elastically by the atoms. As in
the case of x-rays, crystal planes diffract these electrons
in well-defined directions [68] governed by Bragg's Lawg
giving strong, narrow peaks, while amorphous materials
scatter the electrons in a wide range of angles, giving a
weak, broad peak.

The wave length of 100 Kev electrons is ”0.037A, which

51
gives an angular diffraction of the electron beam from a
typical crystal plane of “IOmrad (0.50) for d"23. This
deflection angle is about two orders of magnitude smaller
than in the case of XRD, which accounts for the poorer
resolution of ED. 1n: ED experiments, we observe more
reflections than in the case of XRD, since the Ewald sphere
fbr electrons has a larger radius, and intersects more

points in the reciprocal space of the sample [69].

3.2.2.A The Electron Diffraction Camera and
Bragg Reflection

 

 

A parallel electron beam falling on a crystal at an
angle 9 with lattice planes separated by a distance d
diffracts through an angle 26 from the undeviated beam. If
a detector is located a distance L from the sample, a
diffraction spot corresponding to the lattice spacing d
wild. be detected at a distance R-LtanZB from the central
spot of the undeviated beam. Using Bragg's Law, along with
the fact that, for electron scattering, the angle 6 is
small (sinB-tanB-B), we obtain a formula for the d-spacing
of the lattice planes:

A L

d - R

(3.1”)

L is the camera length, and A is the wave length of the
electron beam.
If the beam falls on a region of the sample that

contains more than one set of lattice planes with different

52

d-spacings, a diffraction pattern (DP) of several spots
will be observed. If the region of the sample contains
very small crystallites that are randomly oriented, rings
with different radii R will be observed. Each ring with a
radius R corresponds to reflection from a set of planes
with a d-spacing that is related to B through equation
3.1”. The widths of the rings are determined by the sizes
of the crystallites. Crystallites as small as “103 still
give fairly sharp rings. Amorphous materials, in contrast,
give a broad, diffuse ring associated with the closest
distance of approach of the atoms [69]; a weak second-
order ring associated with the second nearest neighbor
correlation might also be observed. For accurate measure-
ments of the radii of the rings, it is best to use a
microdensitometer, or the line scan facility available in
the FE-STEM. The line scan facility provides a horizontal
scan of the beam across the diffraction pattern, and gives
peaks corresponding to the rings. The radii of the
diffraction rings are then measured to within <21.
Knowledge of the camera length L as well as of the radius
of the diffraction ring allows the calculation of the
d-spacing of the set of planes giving rise to the diffrac-
tion ring under consideration.

The combination of the electron gun, the sample
holder, and the detector constitutes the electron diffrac-

tion camera. )\ L is called the camera constant. The

53
camera length in the FE-STEM can be varied from 10 to

5000cm.

3.2.2.B Selected Area Diffraction (SAD) and
Microdiffraction (MD)

In the selected area diffraction (SAD) mode, a
parallel beam of electrons falls on the sample, and a
diffraction pattern (DP) from a.certain area of the sample
is viewed on the CRT screen. The investigated area of the
sample can be controlled by means of the SAD aperture. An
area as small as “102 can be measured by choosing a 25p
aperture. The resolution of the DP is controlled by the
size of the collector aperture (CA). A 150p or 500
collector aperture gives a reasonably high resolution DP
with momentum resolution ”0.1A“. Figure 3.6 shows the DP
of a 3003 thick FCC polycrystalline Phi film prepared at
MSU, and of an amorphous Fle film prepared at ANL. Both
films were sputtered on NaCl substrates. Table 3.2 shows
the Miller indices of the Ni planes responsible for the
cmserved diffraction rings. The camera constant of
2.1u5Acm was obtained by matching the calculated and the
observed d-spacings of the (111) and (200) Ni lattice

planes.

5H

SAD

a)

b)

 

Figure 3.6 SAD of (a) Crystalline Ni, and (b) Amorphous
Fle Films

55

Table 3.2 The d-spacing and the Miller Incices of the Ni
Planes Responsible for the observed Reflections

 

1 of Reflections Qexp.(3) £5; icaic.£§l
1 2.043:0.024 111 2.032
2 1.751:0.018 200 1.760
3 1.24310.009 220 1.245
n 1.059;0.007 311 1.061
5 0.867:0.015 400 0.880
6 0.79110.010 331 0.808

420 0.787
7 0.715:0.006 422 0.719
8 o.670;0.oo7 511 0.677
333 0.677
9 0.617:0.006 440 0.622
10 0.584:0.005 531 0.595
442 0.587
600 0.587

In the Microdiffraction (MD) mode, a convergent beam
of small spot size (“103) falls on the sample, and gives a
DP from very small regions of the sample. The resolution
of the DH) pattern can be controlled by the sizes of the
objective aperture (0A) and CA. Because the beam converges
at the spot under investigation, a range of incident angles
exist simultaneously, which leads to broader diffraction
spots and hence to poorer angular resolution than in the
case of SAD. Figure 3.7 shows a microdiffraction pattern
on a region of the Ni film. An estimate of the crystallite
size of the Ni was obtained by moving the beam across the
sample until the MD pattern changed. The estimate of ”1003
was in qualitative agreement with the crystallite size of
~50A estimated from the width of the first diffraction peak
in the SAD pattern.

In the MD mode, a double-tilt goniometer specimen

56

MD

 

Figure 3.7 MD of a Crystalline Ni Film

57

stage is used to obtain a recognizable diffraction pattern
of the portion of the sample under investigation. Tilting
the sample moves the portion of the sample of interest away
from the beam. This problem can be solved by tilting the
sample while in imaging mode, where we can shift the beam
back to the point of interest. This procedure, however, is
lengthy, especially if the microcrystallites are (303 in
size. An additional difficulty in using MD to study the
structure of such microcrystallites is that the drift per
minute of the beam can be larger than the size of the
crystallites. In this section we will be mainly concerned
with using SAD to study the in-plane structure of NiV
superlattices of total thickness <IOOOA. The NiMo samples
were not characterized by ED since these samples were thick
("1u).

To investigate local variations in structure, we
measured pieces of the 19K sample grown on NaCl substrate
at ANL with total thickness ~1000A. The DP differed from
one place in the sample to another. Ihi some places, we
obtained a DP with several sharp and well resolved rings.
Figure 3.8 shows the diffraction pattern that id; typical
for such regions in the film. The d-Spacings were calcu-
lated from the peaks in the line scan. Those d-spacings
are shown in Table 3.3, together with the indices and
calculated d-spacings for those lattice planes of pure Ni

and V that are within 2% of the experimental d-Spacings.

Diffraction pattern

Line scan

Figure 3.8

58

 

SAD of ANL 19A NiV Superlattice

59

Table 3.3 The Experimental d-spacings of the NiV Superlat-
tice with 1i-19A, Along with the Indices and Calculated
d-spacings of NiV Lattice Planes that are within 2% of the
Observed Ones

 

 

# of Reflection gexp.(x) metal(hkl) icaic.£§l
1 2.145:0.027 V(110) 2.135
2 1.262:0.018 Ni(220) 1.245
3 1.073:0.013 Ni(311) 1.061

V(220) 1.068
n 0.802:0.007 Ni(331) 0.808
Ni(h20) 0.787
v(312) 0.807
5 0.715:0.006 Ni(422) 0.719
V(330) 0.712
6 0.613-0.588 Ni(u40) 0.622
v(uzz) 0.616
v(510,u) 0.592
Ni(531) 0.595
Ni(600,u) 0.587

The diffraction ring correSponding to the Ni (220)
reflection shows an azimuthal nonuniformity of the intensi-
ty, InLhfli means that in-plane crystal orientation is
preferred in certain directions. The sixth peak in Figure
3.8 is broad, apparently due to the combination of several
Ni and V peaks. From the widths of the peaks we estimated
the crystallite size to be “RDA. Figure 3.8 also shows
several other weak reflections which were not presented in
Table 3.3. The shoulder between the first and second peaks
corresponds to the V (200) reflection, and the hint of a
peak between the third and fourth peaks corresponds to the
weak Ni (”00) reflection. The tail of the pattern shows a
very broad, weak peak corresponding to a combination of
higher order Ni and V reflections. Those reflections are

not presented in Table 3.3 because they were weak and

60
dominated by the tails of other peaks, and hence their
,positions have uncertainty >21.

Figure 3.9 shows the diffraction patterns of three
other regions of the sample. We notice that the diffrac-
tion peaks are broader, indicating smaller crystallites.
The crystallite size in cases c and d was estimated to be
~10A. The DP in case a shows a structure that is between
that of crystalline and amorphous, with an apparent
crystallite size of ”7A. DPs b and c in Figure 3.9 show
additional examples of crystal orientation preference
(e.g., in b the intensity of the Ni (220) reflection is
comparable to that of the V (110) in the direction of the
line scan). The intensities of the Ni (311) and V (220)
reflections are very small, which means that there are very
few crystallites with those planes oriented in the
direction of the line scan. The above results show large
variations in the structure and the in-plane grain sizes in
different regions of the sample.

The A-6OA sample showed a more nearly uniform
structure with in-plane grain size ~203. The rings still
showed anisotropic intensity distribution which indicated
preferential crystal orientation in certain directions.
Figure 3.10 shows the DP in three different regions of the
sample. The peaks correspond to the reflections given in
Table 3.3 for the 193 sample. The ~0.6% expansion in Ni

spacing for the thinnest samples observed by Khan et al.

61

a)

b)

c)

 

Figure 3.9 SAD in Three Different Regions of ANL 19A NiV
Superlattice

62

a)

b)

c)

 

Figure 3.10 SAD in Three Different Regions of ANL 60A NiV
Superlattice

63
[19] was not resolved in our ED experiments and hence, to
within the experimental uncertainty of <21, the d-spacings
for large and small A samples are the same.

We also studied the structure of samples prepared at
MSU. The A.-ZTA sample showed large structural variations.
Most.<if the regions showed broad diffraction peaks, from
which we estimated the grain size to be (103. The DP in
some regions indicated an amorphous structure. In one
region, the diffraction peaks were sharp, indicating a
grain size of ”303. Figure 3.11 compares the structure in
three different regions of the film. Figure 3.11.c shows a
stronger peak corresponding to the V (200) reflection than
in any other sample. This pattern was isotropic, indicat-
ing that there was no preferential in-plane orientation of
the crystallites.

At different Spots, the 40-153 sample prepared at MSU
showed either an amorphous structure or wide diffraction
peaks indicative of crystallite sizes ~103. MD in crystals
line and amorphous regions was obtained; the regions where
the SAD looked amorphous gave an amorphous MD structure (a
wide diffuse ring), while the crystalline regions gave a MD
pattern that was a superposition of amorphous and crystal-
line structure, as indicated by hints of spots superimposed
on the diffuse ring. Figure 3.12 shows typical SAD and
microdiffraction patterns on two different regions of the

film.

a)

b)

c)

 

Figure 3.11 SAD in Three different Regions of MSU 27K NiV
Superlattice

a)

b)

Figure 3.12
NiV Sample

65

 

SAD and MD in Two Different Regions of MSU 153

66
The above results indicate that samples with large A
seem to have a relatively uniform structure with in-plane
grain size of ”20A. The reflections observed in such
samples correspond, within experimental uncertainty, to
reflections obtained from pure materials. For the 19A
sample, we observed variations in the grain sizes from ”ADA
down to ‘7A. The in-plane structure in most of the regions
of the 27K inn) sample did not show crystallite sizes as
large as those of ANL. The MSU-153 sample showed small
grain sizes euui amorphous structures in various regions.
Microdiffraction at different spots of the sample confirmed
the amorphous nature of certain regions of the film. These
results suggest that different sample preparation
conditions (such as the target-to-substance distance) used
in preparing ANL and MSU samples affect the crystallinity

of the samples.

3.3 EnergyeDispersive X—Ray (EDX) Microanalysis

The development and improvement of the FE-STEM
electron microscope system, with high accelerating voltages
(Z100Kev), and high probe current densities (“109Amp./cm2)
together with small probe size ("53) made it possible to
use EDX microanalysis to study the composition of local
regions ("503) of thin film samples. A thin film is
required to minimize effects such as beam-broadening in the

sample, absorption or florescence [70]. For NiV and NiMo

67
samples we found that samples with thicknesses (10003 obey

the thin-film criterion originally adopted by Tixier and

Philibert:
p/p)gpec csc(a) pt < 0.1 (3.15)
where u/p)gpec is the mass absorption coefficient of the

characteristic x-ray line of material A in the specimen, a
is the angle between the plane of the sample and the axis
of the detector, p is the density of the sample, and t is
the thickness of the sample.

When a small volume of the sample is irradiated with
an electron beam, electrons are knocked out of deep core
levels. In relaxing, the atom emits the characteristic
x-ray lines having energies equal to the transitions of
electrons from the conduction band to the core levels. An
EDX detector (e.g., Lithium-drifted silicon, ~3-5mm thick
and with area ‘10-30mm2) placed Just beneath the sample
collects the radiation and converts it into pulses of
current proportional to the energies of the incident x-ray
photons. These pulses are amplified, digitized euui fed
into a multichannel analyzer and/or a computer memory.
Thus the EDX spectrometer simultaneously displays an entire
spectrum in a chosen energy range from O to 10, 20, H0 or
80Kev. Figure 3.13 shows an example of such a spectrum.

The energy range 1-10Kev is particularly useful for
microanalysis, since it contains K0 lines of elements with

atomic number 2 from 11 to 32, La lines of 2 from 30 to 80,

68

 

INTENSITY

 

 

NV A NI K"
1 A=15
V Ka Jr
J, o
a O
E I o 00
E ‘ o
D :0
e 1- .
S
I oo 0o
I 00 0
° ° NI K
o V KB 0 B
.21 . 1
o 8 0 ° 6%
b - o 23% 3 ° 0:
! mg !:0 g ! io v a g g 2 o ‘
0.0 2.0 4-0 6.0 8.0 10.0

ENERGY (Kev)

Figure 3.13 EDX Spectrum in a Region of MSU 158 NiV Sample

69
and M lines of 2:62.

The x-ray detector is inserted through a side port
(using bellows) into the specimen chamber. The detecting
crystal and its pre-amplifier should be cooled to liquid
Nitrogen temperature to maintain optimum resolution and
signal-to-noise ratio. An exposed end of this cooled
crystal would quickly become coated with ice and other
condensed vapors from the environment; thus a (“7.50)
beryllium "window" is used to protect.Iniis exposed end.
The protecting window absorbs x-rays with energy <1 Kev, so
that characteristic x-ray emission lines of materials
lighter than in: [71] cannot be detected by such Si (Li)

detectors.

3.3.1 Theoretical Background [70]

 

The characteristic x-ray lines of materials in a
sample can be obtained in a single scan. The intensity
ratio IA/IB of x-ray lines of material A and B can be
calculated after subtracting the background contribution to
the peaks. The ratio of the mass concentration of the two

materials is then given by:

 

c I
A A
—-—-1< _ (3.16)
CB AB IB
and
6K
B B
KAB'eK (3.17)

70
where KA is the x-ray generation constant of the charac-

teristic x-ray line of material A, and (A is the detector
efficiency to that line.

Equation 3.16 is applied to a multilayer structure as
follows. The shape of the beam in the film is approximated
as cylindrical with a cross sectional area S. The mass of

metal A in the irradiated volume is:

InA - tASpA (3.18)

where tA is the total thickness of metal A in the super-
lattice. Hence, the concentration ratio of metals A and B
in the superlattice (which is equal to the mass ratio for

our samples with equal layer thicknesses of the two

materials) is given by:

C t p
5-5 - 5-9-35 (3.19)
B B B

Combining equations 3.16 and 3.19 gives the thickness

ratio of metals A and B as:

t p I
A B A

t ' [KAB I (3.20)
B pA B

Equation 3.20 shows that the ratio of the thicknesses of
the two metals in the superlattice can be determined once
the ratio of the integrated intensities of the appropriate
characteristic x-ray lines of the two metals is determined.
The multiplicative constants are tabulated in reference
[70], and the densities of the materials are tabulated in

any standard Solid State text [72]. Equation 3.20 can be

71
reduced to a simple form for NiV samples, which relates the
thickness ratio of Ni/V to the integrated intensity ratio

of their Ka lines:

A l

. 0.85 (3021)

cflcf
HIH

E N
v v

To avoid x-ray absorption in the grid bars, especially
when a Cu-folded grid was used, and to maximize the
characteristic peak-to-background ratio (e.g., see Zaluzec
in ref. [70]), the double-tilt goniometer was used to tilt
the sample relative to the detector axis. Since the V line
is absorbed more than the Iii line in (mi grid bars, the
sample was tilted and measured at intervals of (5° tilt
until the maximum ratio of the V line to the Ni line was
obtained. The absorption effect in the grid bars was less
significant when a one-sided grid was used, so tnuat most
measurements were taken at tilt angles (-20°,+20°) to face
the detector. A check was made, however, by measuring
around this tilt angle, just to be sure that there were no
absorption artifacts. In some cases, VH3 found that the
region of the sample under investigation should be tilted
(-AO°,+AO°) to avoid absorption effects.

Since each piece of a sample of ~1mm2 area contains
~105112, it is impractical to measure the composition of the
sample at regions separated by ”in. Ii) a measurement was
made at the center of each block between the grid bars.

Other spots in the same block were measured if any evidence

72

of unusual behavior was found in that region. Measurements
in AREA mode, where a region of the sample ”511 across is
measured, and measurements using SPOT mode (<SOA), showed
no difference except in a very few regions where different

compositions were found.
Figure 3.111 shows the Ni/V thickness ratio obtained
from two different pieces (BB and B9) of the NiV sample
with A -19K and total thickness “1000A grown on rock salt.
The horizontal axis represents the index of the block, from
which the measurement(s) was (were) obtained. Figure 3.15
shows the topology of piece B9. together with the indices
of the measured block. Figure 3.1“ shows that, except for
few places in the sample, the layer thickness ratio of Ni
to V was essentially constant. The average over 25 places
on the sample B9 was 1.14:0.11. Two places in the same
block (#1) showed different thickness ratios. One gave a
thickness ratio of 1.5 10.1 (i.e., Ni-rich region), and the
other of 1.23:0.1. Measurements as a function of tilt
angle indicated that the difference in the thickness ratio
in the two measured places was not an artifact of
absorption. In the next block (#2), the thickness ratio

dropped to 1.0:0.1.

The piece of the sample in the block labelled (7-21)
in Figure 3.15 was measured at five different places. The
composition analysis at four different spots showed a

thickness ratio of 1.16, consistent with the average across

 

"

 

73

 

 

 

 

3-0 ' I I I I I I I I I I
1.8 — A as ..
NV [j 33
1.6 - .1
é Fe Fe Cl
1 4 - I I 1 ..
i 2 A A A 4
> "' A
.1.) El % 1% A % A A A A
\. i 0 - 8 A _
Z
t I
0.3 - _
0.5 - d
04 — a
02 — . a
0.0 1 1 1 1 1 1 L 1 1 1
0 2 4 6 8 10 12 14 16 18 20 22

n (Bock #)

Figure 3.111 Ni-to-V Thickness Ratio in ANL 19R NiV
Superlattice

711

 

 

 

 

 

Figure 3.15 Topology of a Measured Piece (B9) of ANL 19A
NiV Superlattice

75
the mmmle. At the fifth spot (a dark inclusion), the
thHMmess ratio was 0.9 (i.e., V-rich). More than 50
meamwements in that region showed a lower thickness ratio

in that inclusion than anywhere else in that region.
Certain regions of the sample contained contaminants such

as Fe and Cl. These contaminants can be detected by EDX.

Other contaminants, such as O and C, cannot be detected by

this technique. Energy Loss Spectroscopy is a useful

technique for detecting such low-Z contaminents, as will be

discussed later. The Fe contamination found in some

regions, such as block #10 was barely detectable. The CI

contamination found in other places was due to undissolved
pieces of the NaCl substrate.

Figure 3.1" also shows the thickness ratio for piece
88 of the 19R sample with a total thickness of 10001. The
thickness ratio in seven different regions of the sample
showed ani essentially uniform composition with an average

inatio (of 1.15:0.1, consistent with most measurements in

piece B9.

To study the local variations in the composition of

samples prepared at MSU, we measured a piece of the

SODA-thick, 27A NiV sample (Figure 3.16). Except in one
region where the thickness of the film was highly nonuni-
forum the composition was essentially uniform with an
The nonuniform region gave a large

average of 1 £710.06.

vaT'iatzion in the composition at different places. Two

76

 

2'0 I I I I I I I I
1.8 -' % _
hflv
1.6 - é .—
1.4 — é _
A §
4-,
‘2. L0-— _
z
4.’
0.8 '- —-1
03 L _
0.4 I- _
0.2 _- _
00 1 1 1 1 1 1 1 1 1

 

 

 

0 i 2 3 14 s s :7 s s 10
n (Block #)

Figure 3.16 Ni-to-v Thickness Ratio in MSU 27R NiV
Superlatt ice

77
thick places (>5003) gave a Ni/V thickness ratio of
1 .25:0.06 and 1.35:0.011. Two thin places ((5003) gave a

thickness ratio of 1.88:0.13 and 1.53:0.09. The uniform

region of the film in the same block gave a thickness ratio
of 1.19:0.011. Absorption effects in the grid bars cannot
be responsible for these variations, since we used 8 Be

grid (which is transparent to Ni and V x-ray lines), and a

tilt angle of (~20°,+20°).

We also measured a ~10003-thick, 223 NiMo sample
prepared at MSU on NaCl. The multiplicative factor for the
Mo/Ni thickness ratio in equation 3.21 is 3.3. The
composition If) 26 different places was 1.11:0.05 (Figure
3.17). This sample did not show the large composition
variations observed in the MSU NiV sample or in piece B9 of
the ANL samples.

Other pieces of other samples, such as the ANL 153 NiV
sample of total thickness “11: grown on Mica were studied
using x-ray microanalysis. In such samples, however, thin
regions along the edge of the sample were studied, so that
there is In: information about the composition of the
huerior region of these samples. Besides, the x-ray
spectra on such samples were detected using a fixed tilt
angm (~20°,+ZO°). The Ni-to-V thickness ratio in the 153

sample was found to oscillate between 0.86 and 1.00 in 16

different places.

For most of the remaining samples, emission EDX was

th/tNI

2.0
1.8
1.8
1.4
1.2
1.0
0.8
0.8
0.4
0.2

0.0

Figure
Superlattice

78

 

 

 

 

I l I l I I I V I I I I I
' NiMo ‘
AAAAAAA AA AA Aéééééééééé
.. A A .1

i 1 I I l J I l l L l L L
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28

n (Block #)

3.17 Mo-to-Ni Thickness Ratio in MSU 223 NiMo

79

not used for quantitative purposes, but rather for occa-

sional checking for contaminents or to identify the sample

and the regions to be studied by other techniques, such as

EELS.

CHAPTER FOUR

EXPERIMENTAL RESULTS AND ANALYSIS

u.1 Introduction

In this chapter, we first present the characteristic

Elechwannergy Loss Spectra (EELS) for the pure metals Al,

V, Ni and Mo, and compare our data with optical and other

EELS data to check the reliability of our measuring

techniques and the calibration of the system. Second, we

present EELS data on NiV superlattices in the energy range

<100ev, and then in the higher energy range (>500ev) in the
(2p-—93d) transitions. We

vicinity of both Ni and V L2,3
A.

examine how the spectra vary with superlattice period
and compare the intensities of the various lines with those

mm“ a NiV "pseudo-alloy" consisting of a superposition of

and v spectra. This study will

appropriately weighted Ni
be presented for samples prepared at ANL. Then we discuss
and inter-

data fwn~ NiV superlattices prepared at MSU,

compare the data for the two sets of superlattices. Third,

in the energy region <100ev for NiMo

we present data
discuss the results, and

superlattices prepared at ANL,
then compare them with results for NiMo samples prepared at

In NiMo, the Ni L2’3 lines were too weak to be used

MSU.
The Mo L2,3 lines start at

quantitative analysis.

for'
wriich is above the effective range (<2000ev) of our

ZSZOev,

electron spectrometer.
8o

81
11.2 EELS of Pure Metals and Oxides
As mentioned in Section 1.1.2, several metals have
been found to exhibit plasmons at energies given by the
classical free electron plasma energy wp. The dielectric
function of such metals is given by the classical, free

electron model [13]:

2
W
9 (1m)

 

w(w + igo)
where w is the free electron plasma energy given by

equation 1.1 and go is the damping constant for the plasma

oscillation.

Equation “.1 shows that, for small damping (go<<wp),

the real and imaginary parts of the dielectric function are

given by:
2
W
‘1”) '1' ‘5 (14.2.a)
W
2
W 80
62(w) - 4-3—— (u.2.b)
W

This result shows that both 61 and 62 are smooth
functions of w. 61 intersects the w-axis at wp, and 62 is

very small at wp.

The bulk loss function Im(:€l) measured in EELS
w

[151 exhibits a peak at wp, which has the value 32.

W o

ImC—L) exhibits a peak at 4-. This is a surface
6+1 21/2

plasmon as predicted by Ritchie [59].

Figure 14.1 shows the optically determined values of

 

 

 

82

 

 

 

 

 

 

 

 

 

  

 

 

7 I I I I I I I I r I I I I
----- Epsilon i
5 _ Epsilonz ‘
3 '- -
N 1 ' __ _______ -
w" I,
/
-1 - /,/ _
/
-3 .. // _.
/
/
5 /
/
I I II I I I I I I I I I I
I F I I I I I I I I I I I
----- OPTICAL DATA
EELS (MSU,
as - A l
[I
>_ 20 T I} _
I— h
5': "
DJ l'
'- II
2 l
H 10 '- 'l _.
[I
II
5 - 1‘ _
\
\
o I_./I l\\+ _ L._l

 

 

o 2 4 s a 10 12 i4 16 18 20 22 24 as as
ENERGY (ev)

Figure 14.1 'Dielectric Function, Loss Function, and EELS of
Al

83

61,62 mullm(€%l for Al [73]. These functions are nicely
consistent with the free electron dielectric model for Al.

EdgureiLI also shows our EELS measurements for a hoot—
thick Al film, normalized and superimposed on the optical

loss humtion. The Al film was sputtered on a freshly-

cfleaved NaCl substrate. (hn~ EELS data were not corrected

for the energy distribution of the beam (the half width of

the electron beam was <1ev). Thus, we notice that the

~15.2ev obtained by EELS is somewhat wider than

plasmon at
Deconvoluting

that obtained from the optical measurements.

the primary beam width from our measured spectrum, assuming

Lorentzian line shape for both the plasmon and the

a
in good agreement with the optical

primary beam, results

~7ev is the surface plasmon of

measurements. The peak at

Al, shifted downward due to an oxide layer at the surface

[7”].
Transition metal EELS spectra are more complnmnmd

than Imany other metals, because of different types of

collective excitations that can occur in these metals [1414],
and because of the onset of interband transitions at low

energies. Ksendzov [147] calculated the dielectric function

of the 3d metals,

acting on the core electrons,

taking into consideration the field
and obtained a formula for

the effective plasma frequency. He suggested that
electrons in the 3d transition metal series can be hybrid-
In the case of Sc and Ti,

ized into bands as follows.

84

there is one band, giving rise to one plasmon in these

metals. In V, Cr, and Mn, there are two bands, the

conduction and the first 3d valence bands, giving two

plasmons in the loss spectra, one corresponding to the

conduction electrons (3 electrons/atom), and the other

corresponding to the conduction and valence electrons

together. In Fe, Co, Ni and Cu, there are three bands, the

conduction band with one electron/atom, and two 3d bands;

thus these metals exhibit three different plasmons. The

first plasmon correSponds to the conduction electrons (one

electron/atom), the second corresponds to the electrons in

both the conduction and the second (unfilled) d-band, and

the third corresponds to all (us+3d) electrons in the three

bands. Ksendzov compared the results of this analysis with

the existing EELS data of the transition metals, and found

good agreement. With some reservations, the same argument

could be applied to the 14d transition metals, due to the

chemical and structural similarities of the 3d and lId

transition metals.

Figure lI.2 shows the optical constants [1le of Mo. 62

can be viewed as a monotonically decreasing function,

complicated by interband transitions which start below 1ev.
Our EELS measurements at regions of the Mosample with

thicknesses (500R are also shown in Figure lI.2 on the same

energy scale as the optical constants. The Mo sample was

prepared by Jet thinning a 3mm diameter disc, cut from a

85

 

'3
O
_‘m

:§—’—-——

“x:

\

|
I
I
|
I

l

|
I
I

 

 

 

.... Optical

— Lynch & Swan

(Arb. Units)

INTENSITY

 

 

 

 

ENERGY (ev)

Figure 18.2 Dielectric Function, Loss Function, and EELS of
Mo

86

ZSII-thick, high purity Mo foil. For Jet thinning, we used
a solution of 20% H280” and 80% Methanol at 0°C. The loss
function, as calculated from the optical constants, and the
reflection EELS of Lynch and Swan [140] are also shown in
Figure 14.2 for comparison. Our data agree very well with
the optical data below 30ev, in both the locations and
widths of the loss peaks. The prominent loss in Lynch and
Swan's data, however, is about 1.5ev lower. This is
probably due to the existence of a surface loss at "20ev
that is comparable in intensity to the prominent loss at
“24.5ev and unresolved from it. The peak at '10ev in the
latter study is higher in intensity than the peak in the
optical or our measurements. This can also be explained as
due to the higher probability of exciting the surface
plasmon [38] at 9.uev [37] when using reflection EELS. The
broad peak centered at u8ev in our data is the up (N23)
transition. This peak consists of a large number of
multiplets which are unresolved due to solid state band
widths and to finite transition lifetime [75]. The peak at
“36ev in the optical data is more pronounced due to a
stronger coupling constant in the optical than in the EELS
measurements, which results in a stronger multiplet

intensity at this energy.
Figure )4.3 shows our EELS data at thin (<SOOX) regions
around the cracked edges of a pure V film. The film was

prepared by sputtering on a Mica substrate, and had a total

INTENSHY

87

 

(Arb. Units)

 

 

 

\I
~——- Opficm
/""
.\ / .V _— Misell & Atkins
/ . '

I, : ‘\

; f I ....... “asst;

I 1 3

I, :s .\

/ .

I .'- \\

’ ' ‘ v M

I : -
/ ,5 2“ 2,3
l, ‘3' \\\\ l l

“I ‘ ‘1‘ \\ bfififfih. V M1

'4- ' a. \ ”v . l
I &o\\_,.l ‘3...
I V- v R‘

I m; .M.

0 20 4O 60 80

Figure “.3

ENERGY(eV)

Loss Function and EELS of V

88

thickness of “In. Figure u.3 also shows the 60Kev trans-
mission EELS of Misell and Atkins (their films were 250-
7003 thick, prepared by evaporation on Mica substrates),
and the optical data from reference [In]. Our data agree
well with those of Misell and Atkins, so far as the
positions of the various peaks are concerned. Our data,
however, show a sharper prominent loss and a deeper
shoulder at the position of the ~10ev plasmon. Presumably,
the differences are due to differences in the beam energy
(60 vs. IOOKev), contamination, and instrumental resolu-
tion. The optical measurements do not agree with either
measurement, and may have suffered from surface contamina-
tion of the sample, as suggested by the authors. of ref.
[37]. The broad peak centered at ~A8ev is due to the 3p
(M2’3) transitions. The arrow at ~37ev in Figure u.3 marks
the threshold of the M2’3 transitions. The arrow at ~66ev
marks the position of the 38 (M1) transition. The M1 edge
was not observed in our measurements, but was observed by
Misell and Atkins. This might suggest that they used a
larger collector aperture in their measurements, since the
nondipolar transitions cannot be observed in transmission
with zero momentum transfer. The use of a large collector
aperture results in poorer resolution, which might also
explain their wider plasmon peaks.

To study the effect of contamination, small pieces of

the V film were taken off the substrate and exposed to air

89

for several days before the time of measurement. The
surface contamination of the cracked edges tn) be studied
was thus uncontrolled. .A few pieces were introduced into
the microscope in a folded Cu-grid. EELS on some places
show characteristic spectra that are different from the
uncontaminated V spectra below 35ev. The spectra of
contaminated V exhibited a peak at ~12ev with intensity
comparable to that of the prominent peak at ‘26ev peak.
High energy EELS scans (>500ev) showed an oxide core
structure at ‘530ev in places where the EELS structure
differed from that for uncontaminated places on the sample.
Our EELS data on contaminated pieces of V film differ
considerably from the optical data of Weaver et al [37] in
both the shape and positions of the peaks below 35ev. This
suggests that the nature of contamination in the two
experiments is different. This is not surprising, since
the contamination of both samples was uncontrolled and V
has several oxides [76]. Figure u.u compares EELS spectra
on uncontaminated and contaminated regions of the V film.

Figure u.5 shows our measurements on a 3003 Ni film,
prepared by sputtering onto a Mica substrate, along with
the measurements of Misell and Atkins, and optical measure-
ments from reference [1“]. The agreement between our
measurements and the others is fairly good below 3Sev. The
optical data, however, show a sharp peak at ~IOev, which is

observed as a shoulder or a slope change in transmission

EELS INTENSITY

Figure A."

(Arb. Units)

90

 

Z}
<

f” a

I ;‘. ‘3’" ~.

I . a

. ‘PW :.

. = V M 2.3
I - '-, I
II .' ".
4» o ’ 0‘ ’5‘. ’*K.
4’ he, 1' ‘ .M’. .3
J. ‘ .‘
v ' g "1“ Contaminated

“‘3 ‘ ‘1‘ Uncontaminated

' M

: ' M
O
FFIIIFFTFIIIrTrIIrrfIIIIIITIIIfIIIrIIIrIIIIIIIIrIf

-5 15 35 55 75 95
ENERGY(ev)

 

 

 

Contaminated and Uncontaminated V EELS Spectra

EELS INTENSITY

91

 

Optical

NI
.a-g —— Misell & Atkins

(Arb. Units)

 

 

 

 

IIIIIIIIIIIIIIIIIUIIIIIIIIIIIIIIIIIIIIIITIIVI

0 15 35 55 75
ENERGY (ev)

Figure “.5 Loss Function and EELS of Ni

92
EELS measurements. The reason for the higher intensity of
this feature in the optical data is not clear. The step at

~67ev is the 3p (M2,3) transition.

is; EELS of NiV Superlattices

EELS measurements were performed in two different
energy regions. In the energy region <100ev, the spectra
show single particle interband transitions, volume,
surface, and, perhaps, interface collective excitations, as
well as transitions from 3p shells. In the energy range
>200ev, we find the L2,3 transitions of Ni and V, as well
as occasional evidence of K-edges of oxygen, <uarbon, and

other contaminants.

iaisl The Energy Range <100ev

In this energy range, several features of interest
appear. Volume and surface plasmons, as well as possible
interface plasmons, occur at energies <25evq Interband
transitions also occur in this energy range, introducing
additional complications to the spectra. Low-lying M2’3
core excitations occur at energies between 35 and 70ev.

The spectra in the core edge region are insensitive to
the atomic environment, and hence should be just a simple
superposition of the individual metal spectra [77]. In the
energy range <35ev, however, we found that the spectra

deviated significantly from simple superposition, as will

93
be discussed in detail later in this chapter.
Several samples prepared in ANL on different sub-
strates were studied. Those samples are listed in Table

“.1.

Table “.1 NiV Samples Prepared at ANL

 

Sample # A(X) Substrate Total thickness(D)
I 15 Mica 1p
II 19 Mica 1p
III 19 Mica 1p
IV 19 NaCl 1000K
v 19 Mica Ioooi
VI 35 Glass 1p
VII no NaCl 7001
VIII 60 NaCl 750K
IX 80 NaCl 680%
x 80 Mica 1p

Our first reliable spectra on NiV superlattices were
obtained on a piece of the 19K sample #IV, which we studied
at 13 different places. Ten of these spectra exhibited a
peak at ~10-11ev, and three showed only a shoulder or a
slope change at ~11-12ev. In eleven of these spectra, the
prominent peak position was at 2“.0:0.“ev. Two of the
spectra had the prominent loss at 21.6ev, with a peak at
10ev.

One piece of sample #II (a) and two pieces of sample
#III (b and c) were studied to check the effect of sub-
strate and total sample thickness on the spectra. Most of
the spectra exhibited a peak at “10-11ev; however, in piece
(c) it appeared at ~11.5ev. In piece (a), the position of

the prominent loss fluctuated between 21.5 and 23.5ev. In

9“

piece (b), the position of this peak was at “2“.0:0.“ev.
In piece (0). four places were studied. Three of the
spectra showed the prominent loss at 2“.0:0.2ev, while the
fourth showed the prominent peak at 22.0ev, and a slope
change at ~11ev. A different piece of sample #IV was
studied in seven different places. In six places, the
spectra exhibited a peak at 23.510.5ev, and a peak at
10.610.“ev. At one place, the spectrum showed a shoulder
at '11ev, and a prominent peak at 23.8ev. Still a third
piece of sample #IV was studied at five different places to
further check the results and test for reproducflxfldty.
The prominent peak was found to be at 23.810.6ev in all
spectra. Four of these five spectra exhibited a peak at
11.“:0.“ev, while the fifth Spectrum showed only a shoulder
at 11.6ev.

The above results from six different pieces of three
different samples with period 193 showed that most of the
spectra exhibited a peak at 11.010.8ev. The prominent peak
in the spectra of five of the six pieces studied was at
2“_+_1ev. In one piece, however, it seemed questionable
whether a simple average usefully defines the position of
the prominent peak, since the position did not seem to be
centered at any given energy in the range 21.5-23.5ev. As
illustrated in Figure “.6, the dominant spectra of the 19K
NiV superlattices, consisting of 11 and 2“ev peaks, cannot

be obtained by a simple superposition of pure Ni and V

95

 

SJ.

 

>- ‘i
I:
sir»
m5” 1
I- .4
29
.<:»
mv
A .
u:

I”

J’ ‘a
’4’. 1”! ‘43.. 1‘
.' .f a 2\ v NI
.' ' 1'. a" 2, 3
\V NI M
.o' a “x. 2.3
f“. 1 7% 1
. X ‘5
.3 '~. - \
.' "\ ‘Ww‘y‘t. -\
" \, A: 19 X /'
~1-
_ /
Pseudo Alloy

NIV

 

 

'5

Figure “.6
pseudo-alloy

JIITIITIIIIITIIIIIIIIIIIIIITITIIIIIIIIIIIIIII

15 35

ENERGY (ev)

55

EELS of ANL

75

19K NiV Superlattice and NiV

96

spectra.lu3 form a "pseudo-alloy." The pseudo-alloy
spectrum shown in Figure “.6 was constructed by adding a
pure V spectrum to a pure Ni spectrum with intensity at the
peak position of half that of the V spectrum. The factor
(1/2) was obtained from comparing the values of the loss
functions for Ni and V at the peak positions, as evaluated
from transmission EELS measurements [“6].

Comparison between the characteristic EELS of the 193
sample euui the pseudo-alloy spectrum shown in Figure “.6
suggests that the growth of the low energy peak and the
upward shift of the prominent peak in the 193 sample might
be due to interfaces and/or interactions between the
layers. In order to investigate these possibilities
further, we studied several other samples with different
periods. Figure “.7 shows typical spectra for all the
measured samples from ANL, which we run: discuss in more
detail.

We started with the 803-period sample, which, because
of its large A., we expected to be most like the pseudo-
alloy. Indeed, in a piece of sample #X the loss spectra
exhibited a shoulder or a slope change at 11-12ev, and a
prominent loss at 20.511.0ev, with a hint of an additional
shoulder at ~25-26ev. The loss peak at ~IIev in this
sample was considerably lower in intensity than the peak in
most of the measurements on the 193 sample, and much

closer to the pseudo-alloy spectrunn The main peak was

EELS INTENSITY
(Arb. Units)

97

 

NIV

t.- \ ‘Q
.r 1" \ 1
c 3 ..r’ ‘ \W
J 2.. Q’ .A a. \
”:5,“ (~‘—/ ‘2 33’. . :‘W#M\“ \—
'J i . \xwuw ' “7 ‘\b) I
J "v "Ra, 5‘
. . . , F I.“ - in...
. ~ . In. I‘M.“-
_. \‘f‘x a-st \ - *-
" .Uax \ (a)
h J a '1 “1.x
1 'v \. \..
I 1 5‘”, 2 \V\ A 8 19 R
J ‘ I A .1.
\V JJ -‘ .. \
. I " ‘x. “a.
P. .. .J-r" 1‘ ‘ _‘KKJ A8 19 x
I \J .1. \. 2" A‘\¢"—\_.
I," “x '- -"
o .. ‘.
t ,r .5 A=353

’5 3 ’ ‘ A=80X
o \

NiV Pseudm

IIIlIITIIjITTIIIIITTIIIIIITI'IIIIITIIIIIIITT—TIIIT

-5 15 35 55 75 95
ENERGY (ev)

 

 

 

Figure “.7 EELS of ANL NiV Samples

98
also at ~21ev like the pseudo-alloy. The shoulder at ~26ev
is probably the remnant of the pure Ni 26ev peak which is
also noticed in the pseudo-alloy spectrum.

We next investigated a piece of the BSA-period sample
#VI, which was taken off the glass substrate. The spectra
exhibited a shoulder or a small peak at about 10.011.0ev
and a prominent peak at “22ev. The peak at ~10ev was
weaker than that of the 193 sample, but seemed to be
stronger than the 80A sample, as indicated in Figure “.7 by
the deeper shoulder of the feature in the 353 sample. To
cross-check, another piece of the BOX sample #X was
measured, and the results were consistent with the measure-
ments on the first piece.

So far it seemed that the 11ev peak grew in intensity
as A decreased down to 193. It seemed a natural step to
next measure samples with periods (193. Studying such
samples is essential for the interpretation of the observed
evolution of the “11ev peak, and the shift in energy of the
prominent peak since, as noted in Section 3.2.1.8, the 153
sample looked amorphous when examined by XRD both by Khan
et al. [19] and later by us. Thus, if the observed growth
of the ~10ev peak was not due to interfaces, we would
expect this peak to be stronger in the 153 sample than in
the 193 sample. We would also expect the upward shift in
the prominent peak position to increase in the 15A sample

if this shift was not related to the structure of the

99
samples. So, we cracked a small piece of sample #I with
15K period and transferred it to the microscope in a folded
Cu grid. EELS data showed a slope change at ~11-12ev and
the prominent peak at 22.011.0ev. We went back and studied
another piece of the 193 sample #IV. The spectra still
showed a peak at “10.5ev, comparable in intensity in: the
prominent peak at ~22.5ev. Then, a different piece of the
ISA sample was measured. The spectra were consistent with
the previous measurements. At one place, however, out of
the 15 places measured in the two pieces, the spectrum
showed a peak at ~11ev, comparable in intensity to the
prominent peak at 21ev. We again went back and studied a
different piece of the 193 sample (#III) to make sure that
there were no measuring artifacts or other unknown effects.
Measurements at four different places on this sample gave a
strong peak at 10.“ and a prominent loss at 2“ev, consis-
tent with the previous measurements on this sample. No
significant contamination effects were observed in any of
the above measurements. Thus the evolution of the peak at
11ev was tentatively attributed to interface plasmons. The
interpretation seemed reasonable, since for large A
samples, the number of interfaces per unit volume is small,
so that we expect to find a smaller interface peak relative
to the prominent bulk loss peak at ~21ev. As the period
decreases down to 193, the number of interfaces per unit

volume increases, and we expect to have a stronger

100
interface peak.

The structural properties of the 15X NiV samples are
not yet completely clear. Khan et al. [19] suggested that
at such small periods there might be agglomeration in the
Ni layers or alloying in the samples. The typical EELS
data seem to be consistent with this suggestion, since, in
either case, layering of the two metals would not occur,
and the EELS data would not be expected to show interface
effects. The one spectrum that showed a peak at ”11ev in
this sample could then be interpreted as having been
acquired from a local region which had reasonable layering.

We later repeated measurements on other pieces of the
various ANL samples, and, with an exception noted below,
found results consistent with those described above. The
largest variations in the data were observed in the 193
samples. These variations might be attributed to the fact
that the sharp transition to an "amorphous" structure
occurs Just below 193, and since this transition occurs
within a layer thickness change of about one atomic layer,
one would expect to find large local variations in 193
samples. These variations would "average out" as a
coherent structure IJI the x-ray measurements due to the
large beam size, but would appear in local measurements
with the small (<503) electron beam size used for EELS.

We subsequently measured thin samples (310003) with a

wide range of periods that were newly prepared at ANL so as

101
in) provide larger thin areas for EELS measurements. The
new samples were #V, VII, VIII, and IX (see Table “.1).
The measurements on these samples gave results consistent
with the previous results. The 193 sample gave a peak at
10.910.9ev and a peak at 2“:1.1ev; the “OK sample gave a
shoulder at 11.1:0.3ev and a peak at 22.110.3ev; and the
60A sample gave a shoulder at 11.“:0.“ev and a peak at
22.010.6ev. Figure “.7 shows that the intensity of the
”11ev peak grew from a shoulder for large period samples,
to a peak comparable in intensity to the prominent peak for
193 samples. It also shows that the prominent loss peak
shifted up to “23-2“ev in the 193 sample, and back to
~22.5ev for the 153 sample. The spectrum at one place of
the 153 sample that showed a strong peak at “11ev had a
prominent loss peak at 21ev.

As noted above, these results suggest that the peak at
~11ev is due to interface plasmons. The source of the
shift of the prominent peak is less clear. we initially
tentatively attributed this shift to charge transfer
between the two metals. To check this possibility, we
carried out measurements in the Ni and V L2,3 core regions
(SOOSESIOOOev). In the next section, we will discuss the
results in this energy range, and investigate their
connections with the data in the low energy range (<100ev).

To see how general was the behavior found in ANL NiV

samples, we then prepared several samples at MSU. In the

102
first run, we prepared four samples. The sputtering
conditions were set to obtain samples with periods varying
from 15 to 303. All of these samples were prepared on
freshly-cleaved NaCl substrates, and had a total thickness
<of about 5003» X-ray structural analysis indicated that
none of these samples exhibited the long range structural
coherence seen in ANL samples having A3193. All EELS
measurements on these samples showed similar characteristic
spectra, with a shoulder or a slope change at ‘11-12ev, and
a prominent peak at '22-23ev. In the second run, we
prepared 5003-thick samples on NaCl substrates with periods
20 and 263. X-ray structural analysis of these samples
showed long range structural coherence perpendicular to the
layers, and the x-ray determined periods were ~22 and 273,
respectively. The characteristic EELS on these samples
were indistinguishable from those of the previously
measured MSU samples, and of the 153 sample prepared in
ANL, with a shoulder at '11ev, and a prominent peak at
~22-23ev. In the 273 sample we observed one 2p size,
nonuniform region, as shown in Figure “.8. EDX analysis at
various spots of this region indicated a nonuniform
composition across the region. The thin region (the light
region in the bright field image) showed a Ni to V ratio of
~1.9 (Ni-rich region). EELS data in the energy range
<100ev in this thin region showed a slope change at “11.5ev

and a prominent peak at “2“ev. Core edge measurements at

103

 

Figure “.8 BF Image of a Nonuniform Region of MSU 273 NiV
Superlattice

10“

'500ev showed a clear indication of oxidation of the
sample. This last result again shows the lack of correla—
tion between the strength of the peak at about 11ev and
oxidation of NiV superlattice samples, since this was the
strongest case of oxidation of any sample, and yet we did
not observe a strong peak at '11ev. Further characteriza-
tion of this thin region of the film using electron
diffraction to study the structure parallel to the layers
indicated that the region was "amorphous" (as discussed in
a previous chapter). Figure “.9 shows EELS data on various
samples prepared at MSU, together with the data of the 153
sample prepared at ANL. These results show clearly that
all the spectra have almost identical shape and positions
of the loss peaks. The fact that EELS measurements on all
the MSU samples were nearly identical suggests that those
samples do not support an interface plasmon, maybe due to
blurring of the interfaces (as suggested by Miller and
Axelrod according to their results on Al-Bi bilayer films).

In view of the differences between the spectra for the
samples prepared at MSU and those from ANL, Ina measured
another piece of the 193 sample (#III) at 8 different
places to check the variability of the data on different
Pieces of the sample, and at different times. This time,
five of the measured places showed a peak at ~9ev, and the
Prominent peak at ~21.5ev, and three showed a shoulder at

~10ev and a prominent loss at “22ev. None showed the peaks

EELS INTENSITY

105

 

.-.-.. NIV

v.1 ' '1. I ' NI M2'3

. . '4. .5 _P '5‘ ‘Y'a . .
’3‘? of L A,A._,...~,"*'r'-\.;_ I
' a
. . 3‘, »""-’- “a “$34?th
. a s o? . fl.
3 . '. .-" f 1, 1. MM
= ~‘ -. I .1- 1- ‘ A-15RIANI.)
5 r-.5’ f 1
. - ' 3" ’J-‘éfi'x’ ‘\ w
9 :r .- .. ' \ ‘ . ‘ I
2 : .J P. ‘- W- ...‘o‘. A-15 X(M SU)
7' J - .2 ~. 7!. - A
r -. . ., ., “up —.
LI" a! "11’: ' KS “‘5'-
'w "l - a a .
.2 .2: ‘._ «luau/“\hw,’ A-223iMSU)
3 'U' " '9‘.
F, '. 5" ‘ - Mfi
.4 ./.';. 2‘

.- 1; _ A-273IMSU)

K -
‘ Way’s-1h"

 

A=303IMSUI

 

 

IIIIIIIIIIIIIIIIIIIIIIIIIIIIII1IIjIIIIIIIIII1

15 35 55 75
ENERGY (ev)

Figure “.9 EELS of MSU NiV Samples

106

at 11 and 2“ev seen in all previous studies of ANL 193
samples. Again, there was no correlation between these
variations and the contamination or variations in the
composition of the sample in the different measured places.
The variations in the previously measured 193 samples could
not be correlated to contamination or to composition
variations either. All the spectra on the last piece of
the 193 sample showed a shoulder at ~5-6ev that might be
assigned to a surface plasmon. This feature was not
observed in previous measurements.

To summarize, the dominent behavior of EELS data of
NiV superlattices as A decreased down to 193 showed an
increase in intensity of the 11ev peak and a shift of the
21ev peak up to 2“ev. The 11ev peak was attributed to
interface plasmons. The shift of the 21ev peak is not
understood. The EELS of "amorphous" samples showed spectra
similar to those of the pseudo-alloy. MSU samples showed
spectra similar to those of the "amorphous" samples, which
suggests that the interfaces are not as sharp as in ANL
samples; hence, interface effects do not exist in such

samples.

“.3.2 The Energy Range >500ev
In this energy range, long acquisition times are
reQuired for a strong signal. Large, thin areas are thus

Preferred in these measurements, so that most of the

 

107
measurements were made on thin ((10003) NiV samples.

In this energy range, EELS has the general form of a
smoothly falling "background" which has been found experi-
mentally to satisfy the relation:

I(E) - A.E"" (“.3)

where I(E) is the EELS intensity at energy loss E, and A
and r are fitting parameters [70]. This background comes
from different kinds of excitations, such as valence
electron excitations to the vacuum, multiple plasmon
losses, and tails of features at lower energies. On this.
background is superimposed the inner shell transitions.
These transitions are used to identify the elements in the
Sample, since each core edge transition is a unique
Pl" operty of the excited atom. This identification of
ma terials is important, especially in the case of low 2
materials (Z<11) to which regular EDX is insensitive. In
aC‘lciition, study of the edges reveals information about the
electronic structure of the materials under investigation
I: 78]. For quantitative analysis of a core edge, it is
important to strip the rapidly falling background. Thus,
we wrote a program that uses the least-squares method to
£1 t ‘100ev of the pre-edge background, evaluate the
pal“ameters A and r, and then subtract the intensity given
by equation “.3 from the experimental spectrum. The core
edSes of transition metals show two sharp lines, called

"1'1 ite lines (WL), near the transition threshold [“9].

108

Those lines arise from transitions from 3P3/2 (L3) and from
3P1/2 (L2) states to the unoccupied part of the 3d or “d
band. The two lines are separated in energy by the
spin-orbit splitting of the 2p states which is ~7ev for V
and ~17ev for Ni. The WLs are superimposed on steps that
correSpond to transitions from 2p states to the s-band.
Thus, the L2’3 transition in Cu, which has a filled d band,
is simply a double step as shown in reference [“9].
According to the (2J+1) degeneracy of the initial 2p
states, we deduce that the ratio of the L3 to L2 transi-
tions should be 2:1, because of the statistical weights of
“ and 2 that the P3/2 and P1/2 states have [55]. This
statistical ratio is observed in the case of Cu, but large
deviations from this ratio are found in transition metals
[“9].

To isolate the d-hole contribution to the edge
structure, we extrapolated each step back to time peak of
the WL, and drew a line perpendicular to the horizontal
axis. This procedure is possible in the case of Ni, since
the two steps of Ni are observable in the experimental
spectrum. For V, we adopted an alternative procedure
described Just below. The WL intensities as well as the
height of the step used in our analysis are shown in Figure
“.10. This analysis procedure was used for simplicity,
since comparison between the results obtained in this

fashion and results obtained by the more rigorous procedure

109

INTEN SITY(Arb.UnIts)

 

 

 

ENERGY(ev)

Figure “.10 Schematic Diagram of L2,3 Core Excitations in
a Transition Metal

110

of assuming an arctangent line shape [79] for the 2p
transition to the vacuum showed no significant difference
:hi the intensity analysis of the WLs. The threshold for
the transitions to the vacuum (i.e., the step cn* the
inflection point of the arctangent) were aligned with the
WL maxima because the Fermi energy position is nearly
coincident with the WL peaks [80,81]. For the V spectra,
where the two WLs overlap, we estimated the height of the
first step by measuring the total height after the WLs, and
assuming the statistical ratio 2:1 for the heights of the
steps. This assumption is not perfectly accurate, but
should be sufficient for our current purposes, since we use
the same criterion for analyzing all the V core spectra,
and we are interested primarily in differences in the
spectra as a function of .A. The WL intensities were
measured relative to the step height after the WLs, as will
be briefly discussed in the next section.

We first discuss the V core edge :hi pure II and NiV
superlattices, and then the Ni core edge in pure Ni and NiV

superlattices.

“.3.2.A V Core Edge at 513ev
Several spectra were collected for pieces torn off a
pure V film of total thickness “1p grown on a Mica sub-
strate. The pieces of the film were transferred into the

microscope in.ai Cu-folded grid within ~10 minutes after

111
tearing, so as to minimize the effect of oxidation at the
freshly torn edges to be used for EELS. We also studied
pieces of'II film that were left at room temperature for
several days in order to study the effect of oxidation on
the spectra. Then we proceeded to study the superlattices.

The areas under the WLs were measured using a plani-
meter. The height (counts/ev) of the step after the WLs
(at ~5“0ev) was used as a standard for the intensities of
the WLs. The WL intensity relative to the step height was
calculated. The ratio of the intensities of the WLs was
found almost constant (”1) for all V and NiV samples;
hence, in our analysis below, we used the sum of the
integrated intensities (12’3-I2+I3) of the two WLs normal-
ized to the step height.

For pure, uncontaminated V, the intensity of the WLs
was 13.3:3.1, and the ratio I3/I2 was 1.06:0.12. For
contaminated V, the WL intensity was 1“.5:1.0 and the ratio
was 1.06:0.10. These results showed that oxidation has
little effect (“10%) on the WL intensities, and their
ratio. We compared our results with those (intensity of
11.7:0.5, and ratio of 1.18:0.0“) obtained from the EELS
Atlas [82], and found fair agreement. The data of Fink et
al. [53] show a lower intensity (9.“:0.“) and a higher
ratio (1.37:0.05) than either our data or that of the
atlas. The data of Fink et al. stops at <535ev, in which

case the measured height of the step might be a little too

112
high due to the tail of the L2 WL, which results in a lower
intensity and a higher ratio.

To compare the results for a large 1\ with those for
pure V, we first measured a 603 NiV sample with total
thickness “7003, prepared at ANL. This sample showed a WL
intensity of 17.“:1.6, somewhat larger than for pure
V, and a ratio of 1.01:0.07, in good agreement with pure V.
One of the spectra showed a hint of oxygen K-edge struc-
ture, and had a WL intensity of 15.1:0.8 and a ratio of
1.11:0.0“, both consistent with the data for contaminated
v. New pieces of the ANL 10003-thick ANL 193 sample grown
on NaCl substrate (#IV) were then measured to check for
any systematic change with .A. The core spectra in
uncontaminated regions showed a WL intensity of 21.711.“
and a ratio of 0.99:0.05. This suggests a strong depen-
dence of the WL intensities on A . In the regions where we
found evidence of oxidation, the WL intensity was 18.1:2.1,
and the ratio was 1.13:0.16. Thus, oxidation cannot be
responsible for the increase of the WL intensity from '13
for pure V up to ~21 for the 193 sample. As shown in
Figures “.11 and “.12, the ANL samples show an increase in
intensity of the WLs as we decrease the superlattice period
down to 193. The horizontal lines in Figure “.11 represent
a standard peak height to compare the intensities of the
different spectra.

Initially two possible effects were considered to

EELSINTENSHY

 

h”

(Arb. Units)
I
i.
b‘ ,

> '
u _
5 i
>0

L12 V L 2.3

.1
'-

 

 

 

a ‘ : l x“
A=603
1&1», [I
.........
Pure V
50'0"“5'1'6 """ 5' '2'b""'5'56""'546"r550

Figure “.11

ENERGY(ev)

L2'3 Excitations of V

20

15

123 /h

10

Figure “.12

 

 

 

 

 

ZS ANL.&M*1£B
C] MSUISMNEEB
I I AJ. 1 I I I I L
0 10 20 30 40 50 60 70 80 90 100
Intensities of the V WLs in NiV Samples

115

interpret the results: (1) interface effects, in which case
we expect a smaller WL intensity for the "amorphous" sample
with A.-153, and the effect on the Ni WLs is unclear; and
(2) charge transfer, in which case we expect a bigger WL
intensity for A.-153 and also a reduction in the WL
intensity of the Ni WLs--which we did not observe, as will
be discussed in the next subsection. To see how the WL
intensity changed as we crossed the superlattice-"amor-
phous" boundary, we measured pieces of the 15 and 303
samples prepared at MSU, neither of which exhibited long
range coherence in the x-ray structural analysis. The 153
sample gave a WL intensity of 16.“:1.0 and a ratio of
1.1“:0.1. The 303 sample gave an intensity of 16.2:2.7 and
a ratio of 1.09:0.07. The data did not show any clear
indication of sample contamination.

To see if the WL behavior of thin 193 ANL NiV samples
was reproduced in thicker samples, we measured several
pieces of the “in thick 193 sample #III. The data showed
an intensity of 15.“:2.5 and a ratio of 1.0“:0.12 in the
uncontaminated regions, and an intensity of 17.0:2.3 and a
ratio of 0.99:0.05 in the contaminated regions. These
results show a lower intensity than what we would have
expected from the 10003-thick, 193 sample. The data also
exhibited large variations in the intensity of the WLs,
which extended from that for pure V up to that for the

10003-thick, 193 sample, which might be correlated with

 

116

structural nonuniformity of the sample. The pieces of
sample #III studied here were also studied in the energy
range <100ev, and showed a considerably lower peak position
than in sample #IV, as discussed in the previous section.
So, there seemed to be a correlation between the prominent
peak position in the plasmon excitation energy region, and
the WL intensity. To investigate this possibility, we went
back and analyzed three spectra taken at threer<iifferent
regions of a different piece of sample #III. In the first
region, we found a prominent peak at 22.2ev and a WL
intensity of 1“.6:1.0, in the second, 23ev and 17.3:1.0,
and in the third, 2“.2ev and 20.011.0. In the regions of
the 10003-thick, 193 sample #IV where we obtained spectra
for both E<100ev and E>500ev, we found that most of the
measured spectra showed a prominent main loss 123.5ev, and
a WL intensity ~20. This apparent correlation between the
prominent loss position and the WL intensity was only
discovered near the end of our studies, and thus was not
studied in detail for all measured samples.

To compare the above results with those for samples
with long range coherence prepared at MSU, the 5003-thick,
223 and 273 samples were measured. The 223 sample had a WL
intensity of 12.9:1.o and a ratio of 1.13:0.07. The 273
sample had a WL intensity of 12.5:1.0 and a ratio of
1.19:0.08. The WL intensities of those two samples were

similar to those for pure V. The prominent peak positions

117
of the low loss Spectra of those samples were ~22.5ev.
These results support the previously noted correlation
between time prominent peak position and the intensity of
the WLs.

In summary, the intensity of the V WLs increased as A
decreased down to 193, and a correlation between the upward
shift of the dominant plasmon peak and the WL intensity was
observed. MSU samples showed WL intensities close to that
for bulk V. These results suggest that the intensity
increase of the WLs is an interface effect rather than a

charge transfer effect.

“.3.2.B Ni Core Edge at ~850ev

Ni core-edge spectra are weak and require long
acquisition times (“20 minutes) to obtain a signal with
fairly low noise level. This acquisition time was often
unavailable because the electron beam was unstable over
such periods of time. Alternative techniques were also
employed, such as repeatedly acquiring spectra using
shorter acquisition times, and adding them in) give a
stronger signal. This procedure, however, takes more than
an hour to acquire one spectrum with a fairly low noise
level, which limits the amount of data that can be acquired
in a given working period of time.

Measurements on a pure Ni, 3003-thick film prepared

on a NaCl substrate at MSU, and on the cracked edges of a

118

pure Ni, ‘1p-thick film grown on Mica at ANL, showed a WL
intensity of 6.510.? and a WL ratio of 3.5:0.7. The data
of Fink et al. [53], as analyzed by us, showed an intensity
of 6.0:0.3 and a ratio of 3.9:0.3. We also analyzed the
data of Leapman et al. [“9] and found an intensity of
8.“:0.“ and a ratio of 3.6:0.2. These results show that
our data agree well with Fink et al., and with the ratio of
Leapman et al. The higher WL intensity of Leapman et al.
might be chm: to surface oxidation, since their film was
only ~903 thick, and hence, even a thin (“103) oxide layer
on the surfaces of the film can have a significant effect
on the intensity of the WLs. In contrast, the samples of
Fink et al. were 600-10003 thick, and were prepared under a
better vacuum.

We measured SUperlattices with periods 193 (samples
#111 and #IV prepared at ANL) and the 303 and 153 samples
prepared at MSU. All of these superlattices showed a WL
intensity between 6.5 and 8.0, and a ratio ~3.0. Although
there were differences in the intensities or ratios from
one sample to another, they fell within the uncertainty of
the data. Figure “.13 shows typical Ni spectra of various
samples.

If charge transfer occurs from V to Ni layers, the Ni
empty d-states would be expected to fill up and hence
reduce the WL intensity considerably. No such reduction in

the intensity was observed. These results leave the charge

EELS INTENSITY

(Arb. Units)

 

835

119

 

  

Nl L 2,,

.I
I a
' '. 0. a N a. ”3.3-1.. 5". "s.
a . . a

 

A=153

.0 I» e-
» 1* o
v + other ”If" +5...th e
006’" +.+ u. t + * 30"
+ O O

A=193

x xx 1‘ x "
xxx 1‘
{Xxx-(- at “In,“ x xi!“ xx
x 1‘ 3 Xxx x

X

 

" Pure Ni

VIII]iIrIIUIIIIIITIIIIIlIIIIllilillIIIII'IIIIIIIIIIIIiITIIIIlrilITT!T*I

855 875 895 915
ENERGY (ev)

Figure “.13 L2,3 Excitations of Ni

 

 

120
transfer model unable to explain all of our experimental

results.

u.u EELS of NiMo Superlattices in the Energy Range <100ev

Since Ni and Mo have essentially the same Pauling
electronegativity (“1.8) [57], this system was chosen to
investigate the relation between the upward shift in energy
of the prominent peak in NiV system and the difference in
electronegativity (i.e., possible charge transfer) between
Ni and V (the electronegativity of V is ~1.6). Thus, if
the shift of the prominent peak was related to charge
transfer, we expected no such shift to occur in NiMo
system. On the other hand, if the growth of the low loss
peak was due to interfaces, we expected such growth in NiMo
samples that showed layering in XRD studies.

Several NiMo superlattices were prepared on Mica at
ANL. The total thickness of each sample was ”1p. The
periods of the different samples ranged from A.-1u to 1103.
They thus crossed the superlattice-"amorphous" boundary
which occurs somewhere between 19 and 1&3.

A small piece (“1mm) from the middle of the large 1i
(1103) sample was measured first. EELS spectra were
obtained at twelve different spots around the cracked edges
of the sample. The data showed a small peak at '10.310.3
ev, and a prominent peak at '2u.h10.2ev (Figure H.1u). The

broad Mo N2,3 peak is centered at ~l16ev, and the Ni M2,3

EELSINTENSHW

(Arb. Units)

. _ _j- 1‘. I, . - inf-g (b )
. . . fl," ‘.- «LP-.N.
’.-' ' ~r: " - - - ’\u.
’1 I, \A' Mm
~ .— i . Ma (.1
g g - , aefifiqk
‘b 1"...“ 2 Fm ‘—
' - .’ .-' . ‘. . '\.
‘ A I r \. .r‘
" ' ‘- ' N“
J - - u- a
. ‘ P .- J x
.. . 3 -- we "‘5 -u-_
.-" ‘. .-' "' ‘2 \z 5““ A 193
\ e A" 2‘ n. \, .1. .
. '_ ¢ —J“ r ‘s ,. .W—‘a NM
d I as P.J ...- J . fl \h’.\ A: 3 0 x
. " '\
..\// 1“ WM“
J - :. W“ A- 110 X M
h, 2V!

121

 

_- . - _ . 9.. S“, .\ M
.J 'Jh_ .5 g I.

 

NlMo

)"fo
0 .V. s

‘ -"’ " ' M N

. Q. I g .P. a o

,f v ,3 3 2. 3

I‘. I. :0 s. r. l

o . ‘. Q' ‘ I. N' M 2 3
- . k . MA ,
‘ ... 13" , ‘9’." “x,

. z. .1 WI 1. ‘fi.’\ 1

' '\

.545.- 's .- fi-L'ivfgp N
o g I l

NiMo Pseudo-Alloy

 

 

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIHIHIIHIIIIIIl—I‘r‘r‘n
5 15 35 55 7

95
ENERGY (ev)

Figure 11.111 EELS of ANL. NiMo Samples

122
step is at “66.5ev. Figure H.1h shows that the positions
of the peaks as well as the strength of the low energy peak
are similar to those of a NiMo pseudo-alloy constructed in
a manner similar to that for the NiV pseudo-alloy.

A piece of the A -19R sample was then measured at 9
different places around its edges. The spectra showed 51
peak at ~9.1:O.Uev which was stronger than the correspond-
ing peak in the pseudo-alloy spectrum, and a prominent peak
at 2h.3:0.1ev (Figure n.1u). The position of the prominent
peak in tune 193 sample was the same as for both the
pseudo-alloy and the 1103 sample, but the low energy peak
was shifted down by '1ev.

‘ha check an intermediate value of A , we measured a
piece of the A -303 sample. The data at nine different
places around the edges of this piece exhibited a peak at
9.5:0.6ev and a prominent peak at 2H.3:0.2ev. As shown in
Figure n.1u, the relative intensity of the low energy peak
for the 303 sample was between those for the 19 and 1103
samples. A piece of the sample with 1H3 period, which
showed EH1 "amorphous" structure in XRD measurements, was
then measured at 17 different places around the edges. At
15 different places we observed a weak shoulder in the EELS
spectra at 9.8:p.uev and a prominent peak at 2n.u:0.2ev.
At two other places, the spectra showed a larger peak at
9.110.1ev, with the prominent peak still at 2u.u:p.2ev.

Figure H.1H contains one spectrum (a) representative of the

123
15 spectra with only a weak peak or shoulder at 9.8ev, and
one (b) of the two spectra that showed a stronger peak at
”9.1ev.

To test the reproducibility of the data, we studied a
different piece of the 19K sample at 25 different places.
At 17 of these places, the spectra exhibited a strong peak
at 9.210.7ev, and a prominent loss peak at 2A.3:p.2ev.
Those two peaks were of comparable heights. The spectra at
four places showed a shoulder at 10.u10.uev and a prominent
peak at 2h.u10.2ev. At the remaining four places, the
spectra showed a strong peak at 9.110.3ev and a prominent
loss at 23.210.3ev. Thus, 8A; of the measured spectma
showed a peak at about 9.2ev, in excellent agreement with
the previous measurements on the first piece of this
sample. The downward shift in energy of ~1ev of the
prominent peak in 16% of the new measurements had not been
observed anywhere in the previously measured samples with
the various periods. The appearance of a shoulder at
10.uev rather than a strong peak in 16% of these measure-
ments was also not observed in the previous measurements on
the first piece of this sample. These last spectra are
essentially the same as the spectra of the 1&3 sample, and
are tentatively attributed to intermixing of the two metals
in the sample at these places. The statistics, however,
show that the majority of the spectra agree very well with

previous measurements.

12”

A different piece of the 1103 sample was then measured
to cross check the previous results. Measurements on 15
different places showed a small peak at 10.510.3ev and a
prominent peak at 2u.510.1ev. Those results were in
excellent agreement with the previous measurements on this
sample.

The above results show the following for the ANL
samples:

1) The prominent peak stayed fixed at 2u.u:0.2ev for
all superlattices regardless of their periods, except in
“10% of the measurements on the 193 sample where we
observed a ~1ev downward shift.

2) The low energy peak generally grew in intensity as
the period decreased down to 19K, below which it went back
down to an intensity equal to that of the large period
sample.

3) The position of the low energy peak shifted from
10.3ev for the large period sample down to 9.1ev for the
193 sample. The position of this peak was ‘9.8ev in those
1AA sample spectra that showed only a small peak or a
shoulder, but was 9.1ev for those spectra that showed a
stronger peak.

Although the prominent peak positions in NiV and NiMo
systems are consistent with the interpretation of the
upward shift in NiV system as due to charge transfer, this

model does not explain the fact that in the 153 NiV sample

125
the prominent peak went back to the position of that for
large period samples.

The growth of the ~1Oev peak as A decreased from 1103
down to 19K is consistent with this peak being due to
interface plasmons, as discussed in conjunction with the
NiV data. When we went further down to 1AA period, most of
the sample regions were not layered, as indicated by the
XRD data on this sample. Thus, we did not expect to
observe the strong peak at '9ev, which we observed in most
of the spectra of the 193 sample. The two places on the
sample where we observed a stronger peak at 9.1ev can be
thought of as local places where well-defined layers give
rise to a stronger low energy peak.

The shift in the low energy peak from 10.3 down to
9.1ev might be due to interface coupling. It could also
be, however, that this shift is an artifact of the super-
position of the ~9ev and the ”Zuev peaks. Upon superposi-
tion, the tail of the prominent peak pulls the apparent
position of the 9ev peak to higher energies; the stronger
is the 9ev peak is, the smaller is the upward shift in its
position.

Since there was a sharp transition in both the EELS
and the XRD data in going from 193 to 1H3 periods, we
prepared at MSU samples with periods ~203 and samples with
periods “1&3 on Mica and NaCl substrates. Each sample had

a total thickness of ~10003. Structural x-ray analysis of

126
the nominally 203 samples indicated that they actually had
periods of 233 (for the sample on Mica), and 223 (for the
sample on NaCl). The two samples which were intended to
have a period of ~1113 showed an amorphous XRD pattern, as
anticipated.

A piece of the 2&3 sample on Mica was taken off the
substrate and transferred to the microscope in a Cu folded
grid. Thin regions ((10003 thick) around the edges were
measured at 10 different places. EELS spectra at six of
these places showed a shoulder at 11.810.2ev and a promi-
nent peak at 23.310.2ev. At four places, the spectra
showed a strong peak at 10.810.7ev, and a prominent loss a
2H.3:0.1ev. The peak at 10.8ev was not as strong as those
of the 193 sample prepared in ANL.

We floated a piece of the 223 sample on water and
picked it up on a T1 grid. The sample was measured at six
different places. At two of these places we observed a
strong peak at 10.710.2ev and a prominent peak at
2u.2:0.3ev. The other four spectra exhibited a small peak
or a shoulder at 11.110.2ev and a prominent peak at
2M.8:0.5ev.

From the measurements on pieces of both samples
prepared at MSU, we see that the position of the prominent
peak is in excellent agreement with the data of the samples
prepared in ANL. The postion of the low energy peak is,

however, considerably higher than that of the previous

127
measurements.

The results CH1 these two MSU samples showed the
following: The position of the low-energy peak in those
spectra that showed a strong peak was the same in both
samples, while in those spectra that showed a small peak or
a shoulder; it was lower in the sample prepared on NaCl
than in the sample prepared on Mica. As in the previous
measurements, the shoulder (or small peak) positions are
higher than those of stronger peaks.

The above data show that all samples prepared at ANL
and at MSU exhibited a prominent peak at “Zuev, which
suggests that the position of this peak is not affected by
the structural properties of the sample. The growth of the
low energy peak is strongly correlated with A and with the
structural coherence of the samples. The shift in the
position of the low energy peak is also correlated with the
strength of the peak (peaks with higher intensity are lower
in energy). The low energy peaks for MSU samples are
higher in energy (“1ev) than those for ANL samples, which
means that the downward shift of this peak is sample and

structure dependent.

 

 

CHAPTER FIVE

SUMMARY AND CONCLUSION

This thesis consists of characterization of NiMo and
NiV superlattices and measurements of their EELS. To
characterize the samples, we used three different tech-
niques:

a) EDX to study the local composition of the samples
at different places.

b) XRD to study the structural characteristics of the
samples perpendicular to the layers.

c) SAD and MD to study the in-plane structural
characteristics of the samples.

EELS measurements were carried out in two energy
regions: <100ev to study bulk, surface, and interface
plasmons, and >500ev to study core-level excitations.

We studied NiV samples with periods ranging from 15 to
803 prepared an: ANL and with periods from 15 to 303
prepared at MSU. We also measured NiMo samples with
periods from 1n to 1103 prepared at ANL and with periods of
“203 prepared at MSU. In both systems, the samples crossed
the superlattice-to—"amorphous" boundary below 193.

a) EDX generally showed a uniform Ni-to-V or Ni-to-Mo
ratio, although occasionally we found variations in the
ratio of the two constituents.

b) In agreement with previous work, XRD data showed a

128

 

129
slight expansion in the average d-spacing perpendicular to
the layers as .A decreased down to ~193, and a sudden
contraction for the "amorphous" samples back to the
expected bulk value.

c) SAD and MD showed local variations, especially in
samples with A. near the superlattice-"amorphous" boundary.

XRD and electron diffraction measurements on MSU
samples showed smaller coherence lengths perpendicular to
the layers and smaller crystallite sizes parallel to the
planes than for the ANL samples.

EELS data on ANL samples with large A were similar to
those for a NiV pseudo-alloy, corresponding in: an appro-
priately-weighted sum of spectra for pure Ni and pure V.
As A decreased to ~193, three effects were observed in the
typical data on ANL samples:

a) An 11ev shoulder in the spectra for large A
samples developed into a peak which grew with decreasing A
until it was comparable in height to the main peak located
at ~Zuev.

b) The main peak shifted upward from ~21 to ~211ev.

c) The normalized V White Line (WL) intensity increas-
ed by about 50%.

The EELS of the 193 samples showed substantial
local variations; in some places, instead of the behavior
defined above, both the low energy peaks and the WL

intensity were similar to those for the pseudo-alloy. EELS

 

 

130

data on the 153 "amorphous" samples were generally similar
to those of the pseudo-alloy; however, in one place the
spectra showed a strong peak at ‘11ev, just as in the 193
sample. When both low energy peaks and WLs were measured,
a correlation was observed between the upward shift of the
position of the prominent plasmon peak and the increase in
the V 1“. intensity. The Ni WL intensity did not show a
significant change with Al.

All the bum] NiV samples had EELS spectra similar to
those for the pseudo-alloy, regardless of A or whether the
sample showed superlattice or "amorphous" x-ray structure.

As in the case of NiV, ANL NiMo samples with large A
had EELS spectra similar to those of a NiMo pseudo-alloy,
with a small peak at ~10ev and a main peak at ~23ev. As A
decreased down to 193, the ~10ev peak typically grew in
height and shifted downward to ”9ev. The prominent peak
remained fixed at ~211ev. Typical spectra of the 1A3 sample
were similar to those of the pseudo-alloyu .Again, some
variations in spectra were observed; at some places in the
19A sample the spectra were similar to the pseudo-alloy,
and at a few places in the 133 sample the spectra were like
typical spectra for the 193 sample.

EELS measurements on MSU NiMo samples of nominal
period “203 showed no consistent behavior; rather, the
spectra were distributed between typical spectra of the ANL

19 and 143 samples.

 

131

In both NiMo and NiV systems, the growth of the ”10ev
peak is attributed to excitations of interface plasmons.
Although charge transfer provides a potential explanation
for the increase in intensity of the V WLs as it decreases,
and possibly also for the upward shift of the (“21ev)
plasmon peak, it explains neither the fact that the
intensity of the Ni WLs did not decrease nor the fact that
the prominent peak shifts down when we cross the superlat-
tice-"amorphous" boundary. The downward shift of the “10ev
peak in ANL NiMo samples is attributed to a strong inter-
face plasmon at “9ev which dominates the original weak
plasmon peak in the pseudo-alloy spectrum.

In summary, in going from superlattice to "amorphous",
the ANL samples typically show unusual behavior, tHDth in
structure and in EELS measurements. The samples with A.
near the superlatticeé"amorphous" border show strong local
variations from the typical behavior. Those variations are
not observed on a macroscopic scale in XRD studies, but are
observed 131 the local studies ayailable in the FE-STEM.
Thus we conclude that, on a microscopic scale, the samples
with A703 are not uniform. The FE-STEM is a powerful
tool to probe local variations and nonuniformities in such
samples.

NiV samples prepared at MSU did not support interface
plasmons, since the spectra of all samples were similar to

those of the amorphous ANL sample. Thus, the interfaces of

132
MSU NiV samples are apparently not as sharp as those from
ANL. MSU NiMo samples, on the other hand, showed at
different places behavior similar to both the crystalline
and amorphous ANL samples, which suggests that MSU NiMo
samples have local regions with reasonably sharp inter-

faces.

APPENDIX

 

Appendix A
The FE-STEM
The Field Emission Scanning Transmission Electron
Microscope is an instrument dedicated for near-atomic
resolution. In 1971, A.V. Crewe and his collaborators [83]
first realized the concept, and later constructed the ultra
high vacuum (UHV) instrument, in addition to the electron
source. The first instrument was the size of a small
suitcase, and was easily evacuated to UHV. To change the
sample, however, required that the instrument be taken
apart, and hence the instrument was impractical for the
study of several samples within a period of a day. The

principle of Crewe's microscope is illustrated in Figure

A.1.

A.1 The VG H8501 FE-STEM [8U]

 

UHV FE-STEMs are now commercially available. The
optical column of the Vacuum Generators (VG) model H8501
FE-STEM instrument used in the present work is shown in
Figure A.2. This instrument consists of two basic units;
(a) the electron optical column; and (b) the electronics
console. In this section these two basic units are briefly
reviewed; for more details, the reader is referred to the

V0 Operation manual.

133

 

13a

Focused
Electron beam Display CRT

Generator

 
 
   
   
    

 

 

 

 

sample

Elastically Scattered
‘r’ Electrons

 

Signal __

E::3 Amplifier

 

 

 

 

 

 

Inelastically __,
Scattered and
Non-interacted
Electrons

 
     

Detectors

 

/

Spectrometer

Figure A.1 The Principle of the FE-STEM

135

WON SPEC! Roman AND
BRIGHT FIE—DD DETECTOR

 

 

 

 

 

 

 

 

 

 

 

 

 

 

DETRACUON 5m
/
mommosmcron- _‘E—E . 3---- mmwrm
. '- - OUTGASSING noun
MICBODIITRACHON AND BEAM TILT l | -- '
COIL WY ' .
!
osmcnv: SHGMATOR\ _
_mmcmmummmm
DMONAPERTURE
‘ | , ma“ . M

 

‘ SCAN AND ALIGNMDJT
' COIL ASSEMBLY

commsnmmmcuu
0011.555th \

 

3
a
!

 

 

 

 

 

 

 

GUN ISOLATION VALVE
HIGH BRIGHTNES ‘- - ‘-
F'm EMISSION GUN\
_7 I 5}: _ll

 

 

Wild

Figure A.2 Schematic Diagram of the

Electron Optical
Column in the FE-STEM

m‘RON SPECXROm AND
BRIGHT rm DEX ECTOR

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

prnumnowamnm
,/
ANNULARDARKFIEDDEI'ECTOR; é : ',-__- mummm
- .unmummrumnv
mmmxmwamnumnuomumrnm ‘1 '"
omLusammx .
!
tnmcmmsmmunm“ _
_omunwtumxnm:
mnmunumananms
‘ . mm , M
o P-_-- '
J
F l gannxnaucmman
: , . comnsmmmx
l 3
l
I
, .
l
. l
_| man
comnmamsncmxnmauocun :
mannmtcomasmmmx.2‘~ :
1 GUNBmJflmNVMNE
I
mcnnmcxnnss ~--
nunzmsmouamw\
If #11

 

 

‘ iii

 

Figure A.2 Schematic Diagram of the Electron Optical
Column in the FE-STEM

136

a. The Electron Optical Column: This contains the
electron gun at the bottom, and the apertures, detectors,
etc. above it, as shown in Figure A.2. The UHV column
consists of two vacuum sections that are pumped using
different mechanisms, and can be isolated from each other
by means of the gun isolation valve. Those two vacuum
sections are (1) the gun chamber, maintained at pressure
P“1O'11mbar, and pumped by an ion pump/Ti sublimation pump,
and (2) the specimen chamber, maintained at P<10"9
mbar, and pumped by a diffusion pump/Ti sublimation pump.
During operation the Ti sublimation pump is turned off.
The diffusion pump is backed by a mechanical pump, and
cooled by a liquid Nitrogen trap that also works as a
cryopump.

The electron gun consists of a single crystal Tungsten
wire sharpened to ai‘tip of curvature "10003 [18]. It is
mounted on a support wire that can be used for occasional
cleaning of the tip by passing a short-duration current
through it. The field emission current of the tip can be
monitored by applying an appropriate voltage between the
tip and the first "extraction" anode. This voltage is
typically (H.5KV, and can give a source brightness up to
109Amp./cm2 steradianm .A second anode accelerates the
electrons. The voltage between the tip and the second
anode can be varied either continuously from 5-3OKV, or

from 20-100KV in steps of ZOKV. Since the real emitting

 

137
region of the tip is ~1003, this gun assembly can produce
an electron beam of 2503 in diameter.

The beam size can be further reduced to <53 using the
double condenser lens system, C1 and C2. Those two
electromagnetic lenses are external to the vacuum, and are
water-cooled. In selected area diffraction mode, they are
used to focus the diffraction pattern. In the low magnifi-
cation mode, they are used to focus the beam on the sample.

The third lens is a high excitation objective lens.
It is partially inside the vacuum, and is water-cooled. It
is used to focus the beam on the sample in the normal
imaging mode.

The imperfect axial symmetry of the magnetic lenses
give rise to astigmatism [70]. Hence, stigmators are
needed for high resolution imaging. In the FE-STEM there
are two octopole stigmators. The gun stigmator is an
8-pole electromagnetic stigmator used for stigmating the
gun/condenser system. This stigmator is important in low
magnification and in selected area diffraction mode. The
objective stigmator is an 8-pole electrostatic stigmator
used to correct astigmatism in the objective lens region.
This is important for high resolution imaging.

The electron beam can be scanned before it strikes the
sample. The scanning assembly coils produce a double
deflection of the electron beam that permits the scanned

beam to be rocked about the objective aperture in the

138

imaging mode, or about the selected area diffraction (SAD)
aperture in the SAD mode. These coils are situated between
the objective and the condenser lenses, and are external to
the vacuum. The beam can also be scanned after it leaves
the sample, using the microdiffraction (Grigson) scan coils
situated above the sample stage. This coil assembly
consists of two sets of double deflection coils. One is
used to produce the microdiffraction pattern, and the other
to align the bright field image. The coil assembly is
bellowssealed, and must be taken out during specimen
changing.

Four different sets of apertures are fitted to the
column. Each aperture mechanism is a bellows-sealed
mechanism with micrometer adjustment for fine control of
the aperture position in the horizontal plane. The
positions of three of these mechanisms are shown in Figure
A.2. The diffraction aperture is used to define an area on
the sample from which a Selected Area Diffraction (SAD)
pattern is obtained. The smallest area that can in:
measured is ~0.3u, achieved with a 10p aperture. Larger
areas can be measured using 50, 100, <n~ 250p apertures.
The objective aperture defines the convergence angle of the
beam onto the specimen in normal imaging, cn' in the
microdiffraction mode. A combination of sizes of this
aperture and the detector (collector) aperture defines the

resolution in these two modes. The smallest size of the

 

139

objective aperture (‘10u) is needed to obtain a high
resolution microdiffraction pattern. In the normal imaging
mode, the 1009 objective aperture is used. When collecting
energy loss spectra in the core region (51500ev), the
signal is weak, and a high intensity signal is needed. In
this case, a large (>100u) objective aperture is required,
in addition to a large (>1mm) collector aperture. This
choice, however, results in loss of energy resolution in
the energy loss spectra.

The collector aperture defines the acceptance angle of
the electron spectrometer. Small apertures are used for
high resolution signals.

The virtual objective aperture (VOA) is optically
conjugate to the objective aperture, and is situated
between the electron gun and the double condenser lens
system. Together with the SAD aperture, it is used instead
<3f the real objective aperture for x-ray microanalysis.
This eliminates the huge Cu x-ray line coming from the 0A
blade after being illuminated by stray electrons.

On top of the column a 90° magnetic sector electron
spectrometer is fitted. This is suitable for the study of
energy losses up to 2Kev. The exit slit width is adjust-
able by means of a bellows-sealed micrometer drive. The
spectrometer entrance angle is defined by the collector
aperture size. The spectrometer output is detected by the

bright field detector. The energy resolution is defined by

 

130
the collector aperture and the slit width.

The bright field detector consists of a scintillator/
photomultiplier/pre-amplifier assembly situated just after
the exit sli¢;<3f the spectrometer. It collects the
electrons after being analyzed by the spectrometer, and
hence provides the video signal for the bright field image,
energy loss spectra, selected area, and microdiffraction.

The annular dark field detector consists of a scintil-
lator and a light guide inside the vacuum, and a photo-
multiplier/preamplifier assembly external to the vacuum.
'It has a hole in it so that the forward scattered beam with
scattering angle (NOmrad. will pass through. The annular
dark field detector collects electrons suffering large
angle elastic scattering (i.e., Bragg reflection from
crystal planes).

An energy dispersive x-ray (EDX) spectrometer is
fitted for microanalysis. The detector is Lithium-drifted
Silicon (SiLAJ. It is continuously maintained at liquid
Nitrogen temperature, since appreciable periods of warm-up
result in changing the detector characteristics [85].
Specimens are loaded into the top entry stage by means of
an air lock, which can be isolated from the specimen
chamber and pumped using the roughing and then the diffu-
sion pump. A cold stage is fitted to pass cold Nitrogen
gas to the specimen area, lowering the specimen temperature

to about 150K. This is essential to minimize the migration

 

1H1
of surface contaminents to the spot being measured.

b. The Electronics Console: This console is modular in
design, and in many cases can be controlled externally by
computer. It contains the necessary control and display
modules. Mechanical shift and tilt controls are located in
the desk top attached to the console. These controls are
necessary for moving from one region to another on the
sample while measuring, and also for diffraction studies.
There are three monitors in the console. Two are normally
used for bright-field and dark-field emission, and the
third is used with a polaroid camera to take pictures of
images or diffraction patterns or energy loss spectra from
the monitor screen directly. In the remaining part of this
section, some functions in the electronics console are
discussed. For further details, the reader is referred to
the V0 Microscope operation manual [8“].

1. The Time Base Fwdule: This unit generates smooth
line and frame scan wave forms. It controls both the
electron beam in the column and the displays. Line and
frame speeds are controlled by front panel switches. TV
rate scanning may be selected by a pushbutton control. The
following modes are controlled by this module:

LOCATE: This facility provides a square spot marker on
all three monitors. Its position in AREA mode determines
the position of the beam in the SPOT and SPECTRUM modes,

and the Y-position of a line scan in LINE mode. The LOCATE

 

132
is selected by means of a push-button, and the position of
the marker is varied using front panel X and Y dials.

AREA scan: At all available speeds, this is used in
imaging and diffraction modes.

LINE scan: Ii single horizontal line is scanned with
the video signal producing a vertical deflection on the
displays in a manner similar to the function of a micro-
densitometer. The speed of the line scan is determined by
the frame speed, and the position by the Y-position of the
LOCATE marker. This feature is important for electron
diffraction analysis in cases where the crystallites are
small and the diffraction rings (in polycrystalline
materials) are run: very sharp. Thus, locating the LOCATE
umrker at the central spot and going to LINE scan mode
using the push-button control produces a set of peaks
surrounding an intense central peak (as will be discussed
later).

SPECTRUM: The beam is static on the specimen, but a
line scan drives the displays. This line scan is usually
used to scan the output of the bright field detector, and
the resulting video signal is used to produce a vertical
deflection of the line on the display. This mode is used
to optimize the control settings for energy loss measure-
ments, and thus, an energy loss spectrum from the region of
the sample marked by the LOCATE spot is displayed on the

monitor screen being used for bright field signal

 

det

pos

mea

bea

the

Sig

the

t0

the

ex.

 

COT

COT

DU:

Cor

1H3
detection.

SPOT: The beam is static on the specimen at the
position of the LOCATE marker. This mode is used for
measuring regions of the sample as small as the size of the
beam.

REDUCED AREA: This reduces the area scanned on both
the sample and the displays. This mode gives an improved
signal/noise ratio. It is also used to measure parts of
the sample displayed on the monitor.

RECORD: The two (left and right) RECORD push-buttons
to the left of the monitor screens select which image on
the monitors is simultaneously displayed on the photograph-
ic monitor. The display and beam scans can be driven by

external analog or digital scan wave forms.

 

2. Magnification-Diffraction Module: This module
contains two identical and independent magnification
controls for normal imaging. Those may be selected by the
push-buttons TM and M2. The push-button SA selects the
condenser lenses for SAD. The camera length [18] is varied
by switched and continuous fine controls. The MICRO
push-button selects the microdiffraction mode. The
condenser lenses settings in this mode are similar to those
for normal imaging. The ROTATION dial continuously rotates
the scan on the sample through 360°; this allows the
observation of image at any angle» It also allows the

pick-up of asymmetry in the ring diffraction pattern when

INC

3212

si‘

CO

931

er

 

1AA
the visually more clear line scan is used.

3. Alignment and Shift Module: This module contains
two basic units; viz., the alignment controls and the
analog spectrometer controls.

Alignment: This unit drives the alignment coils
situated in the scan coil assembly. They are used for
co-aligning the optical axes of the gun and objective lens.

Energy analyzer: This unit controls the current in the
energy loss spectrometer.

SCAN WIDTH Switch: It controls the energy range of the
energy loss spectrum. A push-button allows the OFFSET
facility, which enables one to acquire a spectrum that
starts at a certain energy that can be chosen using dial
controls. Thais feature is essential for acquiring core
edge spectra with reasonable resolution where a scan width
<250ev is selected in the core edge region.

u. Lens Stabilizer Module: This unit provides front
panel controls for the three lenses and the two stigmators.
Pushbutton controls are used to switch the lenses and
stigmators on and off. Dial controls are used to obtain
high accuracy settings for the lenses and the stigmators.
Fine tuning of these dial controls is the determining
factor in obtaining high resolution images and diffraction

patterns.

11111

1m

135

REFERENCES

[1] K. Behrndt in Techniques of Materials Preparation
and Handling, R. Bunshah, ed.. (Interscience, New York,
1968).

[2] I. Schuller, and C. Falco in VLSI Electronics,
vol. A, in Einsbruch, ed.. (Academic Press, New York,
1982).

[3] Molecular Beam Epitaxy and Heterostructures, L.
Chang and K. Ploog, eds., (Martinus Nijhoff, Dordrecht,
1985).

[11] J. Kwo, M. Gyorgy, D. McWhan, M. Hong, F. Di
Salvo, C. Vettier, and J. Bower, Phys. Rev. Lett., 55,
1302, (1985).

[5] I. Schuller, and M. Grimsditch, J. Appl. Phys.,
5_5_, 21:91 (19811).

[6] E. Gyorgy, J. Dillon, D. McWhan, L. Rupp, Jr., L.
Testardi, and P. Flanders, Phys. Rev. Lett., 15, 57 (1980).

[7] C. Chun, G. Zheng, J. Vicent, and I. Schuller,
Phys. Rev. B, 22, A915 (198“).

[8] E. Spiller, App. Opt., 12, 2333 (1976).

[9] T. Barbee, in Low Energy X-Ray Diagnosis, D.
Attwood, and B. Henke, eds., 57, 131, (AIP, New York,
1981).

[10] A. Saxena, and B. Schoenborn, Acta Cryst., A33,

805 (1977).

 

5V
rv

9‘.
e
C

Rev

“J

’15

71.

1A6

[11] S. Ruggiero, Proceedings of the International
Conference (”1 Superlattices, Microstructures, and Micro-
devices, Champaign, IL (1984), and refs. therein.

[12] H. Ehrenreich, H. Philipp and D. Olechna, Phys.
Rev., L21, 2u69 (1963).

[13] Optical Properties and Electronic Structure of
Metals and Alloys, F. Abeles, ed.. (North-Holland, Amster-
dam, 1966).

[in] .1. Weaver, Ch Krafka, D. Lynch, and E. Koch,
Optical Properties of Metals, part I (1981).

[15] H. Raether, Excitation of Plasmons and Interband
Transitions by Electrons, (Springer-Verlag, Berlin, 1980).

[16] T. Chen, E. Wolf, A. Giorgi, Phys. Rev. B, 22,
6036 (1983).

[17] J. Weaver, D. Lynch, and C. Olson, Phys. Rev. B,
1, 11311 (1973).

[18] I. Watt, The Principles and Practice of Electron

Microscopy. (Cambridge Univ. Press, Cambridge, 1985).

 

[19] M. Khan, C. Chun, G. Felcher, M. Grimsditch, A.
Kueny, C. Falco, and I. Schuller, Phys. Rev. B, 21, 7186
(1982).

[20] H. Homma, Y. Lepetre, J. Murduck, I. Schuller,
and C. Majkrzak, SPIE Proc., 56;, 150 (1985).

[21] D. Pines, and D. Bohm, Phys. Rev., 82, 625

(1951).

 

Q4

’\

Pre

P31

Qu-

 

 

.1117

[22] D. Pines, and D. Bohm, Phys. Rev., 22, 338
(1952).

[23] D. Pines, in Solid State Physics, (Academic
Press, Inc., New York, 1955).

[2A] R. Ferrell, Phys. Rev., 131, 553 (1956).

[25] D. Gabor, Phil. Mag., L, 1 (1956).

[26] V. Ageev, L. Belabanova, and M. Bredov, Sov.
Phys.-Sol. St., 2, 2578 (1961), and refs. therein.

[27] J. Swan, Phys. Rev., 122, A1U67 (196k).

[28] 0. Sueoka, J. Phys. Soc. Japan, 2Q, 2203 (1965).

[29] D. Pines, Elementary Excitations in Solids, (W.
Benjamin, Inc., New York, 196“).

[30] E. Stern, and R. Ferrell, Phys. Rev., 129, 130
(1960).

[31] W. Miller, Jr., and N. Axelrod, Sol. St. Comm.,
1. 133 (1965).

[32] G. Kovacs, and G. Scott, Phys. Rev. B, Lév 1297
(1977).

[33] L. Brillson, Phys. Rev. Lett., 28, 245 (1977).

[3A] T. Lopez-Rios, F. Abeles, and 0. Vuye, J.
Physique, 22, 635 (1978).

[35] P. Jewsbury, and P. Summerside, J. Phys. F: Met.
Phys., 1g, 6H5 (1980).

[36] P. Summerside, J. Phys. F: Met. Phys., 2, 16u3

(1979).

I
i" I

 

 

1H8
[37] J. Weaver, D. Lynch, and C. Olson, Phys. Rev. B,
19. 501 (1973).
- [38] R. Vehse, and E. Arakawa, Phys. Rev., 122, 695

(1969).

[39] L. Feldkamp, M. Stearns, and S. Shinozaki, Phys.
Rev. B, 2Q, 1310 (1979), and refs. therein.

[HO] M. Lynch, and J. Swan, Aust. J. Phys., 21, 811
(1968).

[“1] D. Misell, and A. Atkins, Phil. Mag., 21, 95
(1973).

[AZ] W. Simmons, and E. Scheidner, J. Appl. Phys.,
£3. 693 (1972).

[A3] H. Kolodziej, J. Wesolowski, and B. Rozenfeld,
Acta Physica Polonica, 231, 751 (1975).

[an] M. Krsunskii, Ya. Genkin, and V. Zavodinskii,
Sov. Phys.'Sol. St., L2, 2u63 (1971).

[AS] H. Kolodziej, J. Olesik, and U. Madalkiewicz,
Acta Physica Polonica, 222, uu3 (1981).

[116] C. Wehenkel, and B. Gauthe, Phys. Stat. Sol.
(D), 63, 515 (197“).

[N7] Ya. Ksendzov, Phys. Stat. Sol. (D), 22, H15
(1979).

[H8] A. Bearden, and A. Burr, Rev. Mod. Phys., 22,
125 (1967).

[A9] R. Leapman, L. Grunes, and P. Fejes, Phys. Rev.

 

 

 

(D

(I)

(I)

199

[50] R. Kleber, z. Physik, 223, 301 (1973).

[51] R. Kleber, Z. Physik, 222, 309 (1973).

[52] B. Cordts, D. Pease, and L. Azaroff, Phys. Rev.
B, 2;, H692 (1979).

[53] J. Fink, Th. Muller-Heinzerling, B. Scheerer, W.
Speier, F. Hellebrecht, J. Fuggell, J. Zaanen, and G.
Sawatzky, Phys. Rev. B, 22, 9899 (1985).

[5H] J. Zaanen, A. Sawatzky, J. Fink, W. Speier, and

J. Fuggell, Phys. Rev. B, 22, “905 (1985).

[55] Herzberg, Atomic Spectra and Atomic Structure,

 

160 (Prentice-Hall, Inc., New York, 1937).

[56] L. Pauling, The Nature of the Chemical Bond, 3rd

 

ed.. (Cornell Univ. Press, Ithaca, 1960).

[57] M.P. Das, and N. Nafari, Sol. St. Comm., _§, 29
(1986).

[58] R. Camley, and 0. Mills, Phys. Rev. B, 22, 1695
(1989).

[59] R. Ritchie, Phys. Rev., 1.2- 87a (1957).

[60] P. Jewsbury, J. Phys. F: Met. Phys., 12, 2377
(1982).

[61] G. Giuliani, J. Quinn, and R. Wallis, J. de
Physique, Colloque 22, C5‘285 (1983).

[62] G. Eliasson, (L Giuliani, .J. Quinn, and R4
Wallis, Phys. Rev. B, 22, 1905 (1986).

[63] A. Segmuller, and A.E. Blakeslee, J. Appl.

Cryst., g, 19 (1973).

W a

\Jl

 

150

[6A] D.B. McWhan, M. Gurvitch, J.M. Rowell, and L.R.
Walker, J. Appl. Phys., 22 (7). 3886 (1983).

[65] I.K. Schuller, Phys. Rev. Lett., 22, 1597
(1980).

[66] ‘B. Warren, X-Ray Diffraction, (Addison-Wesley,
Reading, Mass., 1969).

[67] C. Kittel, Introduction to Solid State Physics,
5th ed., 58 (John Wiley and Sons, Inc., New York, 1976).

[68] R.W. James, X-Ray Crystallography, 59 (John
Wiley and Sons, Inc., New York, 1953).

[69] A. Guinier, X-RaygDiffraction, (W.H. Freeman and

 

Co., San Francisco, 1963).

[70] Introduction to Analytical Electron Microscopy,

 

J. Hren, J. Goldstein, and D. Joy, eds., (Plenum Press, New
York, 1979).

[71] G. Thomas, and M. Goringe, Transmission Electrom

 

Microscopy'<af Materials, (John Wiley and Sons, Inc., New
York, 1979).

[72] N. Ashcroft, and N. Mermin, Solid State Physics,
(Saunders College HRW, Philadelphia, 1976).

[73] E. Shiles, T. Sasaki, M. Inokuti, and D. Smith,
Phys. Rev. B., 22, 1612 (1980).

[7a] P. Schmuser, Sol. St. Comm., 2, U1 (196A).

[75] L. Davis, and L. Feldkamp, Sol. St. Comm., 12,

1113 (1976).

 

 

1°

151

[76] See the oxidation states of V in, e.g., Handbook
of Chemistry and Physics, (CRC Press, Inc., Bokaraton,
Florida).

[77] W. Gudat, and C. Kunz, Phys. Stat. Sol. (b), 22,
“33 (1972).

[78] G. Rossi, R. Jaeger, J. Stohr, T. Kendelewcz,
and I. Lindau, Phys. Rev. B, 21, 515" (1983), and refs.
therein.

[79] F. Lytle, J. Catal., 51_3, 376 (1976).

[80] D. Pease, and L. Azaroff, J. Appl. Phys., 22,
6605 (1979).

[81] F. Szmulowicz, and D. Pease, Phys. Rev. B, 17,
33B1 (1978).

[82] <3. Ahn, and O. Krivanek, EELS Atlas, (Gatan,

 

Inc., Warrendale, PA, 1983).

[83] A.V. Crewe, Phil. Trans: R. Soc. Ser. B, 221, 61
(1971).

[8A] V.G. Microscope Operating Manual

[85] R.W. Fink, in Proceedings of a Workshop on
Energy Dispersive X-Ray Spectrometry, K.F. Heinrich, D.E.
Newbury, R.L. Myklebust, and C.E. Fiori, eds., NBS SP6OA

(Government Printing Office, Washington, 1981).