ABSTRACT 
 
Title of Document: Atomic Layer Deposition Conformality and 
Process Optimization: Transitioning from 2-
Dimensional Planar Systems to 3-Dimensional 
Nano-structures 
  
 Erin D. Robertson Cleveland,  
Doctor of Philosophy, 2010 
  
Directed By: Professor Gary W. Rubloff,  
Department of Materials Science and 
Engineering and the Institute for Systems 
Research 
 
 
Conformal coatings are becoming increasingly important as technology heads 
towards the nanoscale.  The exceptional thickness control (atomic scale) and 
conformality (uniformity over nanoscale 3D features) of atomic layer deposition 
(ALD) has made it the process of choice for numerous applications found in 
microelectronics and nanotechnology with a wide variety of ALD processes and 
resulting materials.  While its benefits derive from self-limited saturating surface 
reactions of alternating gas precursors, process optimization for ALD conformality is 
  
often difficult as process parameters, such as dosage, purge, temperature and pressure 
are often interdependent with one another, especially within the confines of an ultra-
high aspect ratio nanopore.  Therefore, processes must be optimized to achieve self-
limiting saturated surfaces and avoid parasitic CVD-like reactions in order to 
maintain thickness control and achieve uniformity and conformality at the atomic 
level while preserving the desired materials? properties (electrical, optical, 
compositional, etc.).    
 This work investigates novel approaches to optimize ALD conformality when 
transitioning from a 2D planar system to a 3D ultra-high aspect ratio nanopore in the 
context of a cross-flow wafer-scale reactor used to highlight deviations from ideal 
ALD behavior.  Porous anodic alumina (PAA) is used as a versatile platform to 
analyze TiO2 ALD profiles via ex-situ SEM, EDS and TEM.  Results of TiO2 ALD 
illustrate enhanced growth rates that can occur when the precursors titanium 
tetraisopropoxide and ozone were used at minimal saturation doses for ALD and for 
considerably higher doses.  The results also demonstrate that ALD process recipes 
that achieve excellent across-wafer uniformity across full 100 mm wafers do not 
produce conformal films in ultra-high aspect ratio nanopores.  The results further 
demonstrate that conformality is determined by precursor dose, surface residence 
time, and purge time, creating large depletion gradients down the length of the 
nanopore.  Also, deposition of ALD films over sharp surface features are very 
uniform, and verified by profile evolution modeling.  This behavior, in contrast to that 
in high aspect ratio structures, suggests strongly that detailed dynamics, local flow 
conditions (e.g. viscous vs molecular), surface residence time, and ALD surface 
  
reaction kinetics play a complex role in determining ALD profiles for high aspect 
ratio features. 
  
 
ATOMIC LAYER DEPOSITION CONFORMALITY AND PROCESS 
OPTIMIZATION: TRANSITIONING FROM 2-DIMENSIONAL PLANAR 
SYSTEMS TO 3-DIMENSIONAL NANO-STRUCTURES 
 
By 
 
Erin Darcy Robertson Cleveland 
 
 
Dissertation submitted to the Faculty of the Graduate School of the  
University of Maryland, College Park, in partial fulfillment 
of the requirements for the degree of 
Doctor of Philosophy 
2010 
 
 
Advisory Committee: 
Professor Gary W. Rubloff, Chair 
Professor Ray Adomaitis 
Professor Robert M. Briber 
Professor Joonil Seog 
Professor John Cumings 
  
 
 
 
 
 
 
 
 
 
? Copyright by 
Erin Darcy Robertson 
2010 
 
 
 
 
 
 
 
 
ii 
 
Dedication 
This work is dedicated to my amazing husband, Ernie, and my mom and dad, 
who always believed in me and reminded me I could do this. Do you know 
something? 
 
 
iii 
 
Acknowledgements 
I would like to thank all those who contributed to the development of my 
research and helped refine my research skills.   I especially want to thank my advisor, 
Dr. Gary Rubloff for providing me with the opportunity and encouragement to 
explore and develop my scientific creativity.  His guidance has helped me gain 
independence and a sense of confidence in my ability to do research.  I would also 
like to thank past, current, and honorary students of the Rubloff group, especially Dr. 
Susan Buckhout-White, Dr. Israel Perez, Dr. Laurent Lecordier, Dr. Sebastian 
Engelmann, and Parag Banerjee.  Thank you to Dr. Perez and Mr. Banerjee for 
insightful discussions about my research.  Thank you to Dr. Lecordier for taking the 
time to teach me in and out of the lab.  Thank you to Dr. Buckhout-White for being 
my hero and Dr. Engelmann for being a great roommate and role model. 
I want to acknowledge the Maryland Nanocenter Staff, FabLab staff, and 
NispLab staff.  I particularly want to thank Li-Chung (Larry) Lai of the NispLab for 
all of his support and guidance with SEM, EDS, and TEM training and image 
analysis.  I would also like to acknowledge Wonseok Hwang of Dr. Joonil Seog?s 
research group at the University of Maryland for his time in making AFM 
measurements.   
This work was supported in part by the Laboratroy for Physical Science 
(LPS), the UMD-NSF-MRSEC under grant DMR 05-20471, and the FabLab and 
NispLab facilities in the Maryland NanoCenter. 
 
 
iv 
 
Table of Contents 
Dedication..................................................................................................................... ii 
Acknowledgements......................................................................................................iii 
Table of Contents......................................................................................................... iv 
List of Tables .............................................................................................................viii 
List of Figures............................................................................................................viii 
Chapter 1 : Introduction................................................................................................ 1 
1.1 Atomic Layer Deposition (ALD)........................................................................ 1 
1.1.1 Background on ALD.................................................................................... 1 
1.1.2 ALD process sequence................................................................................. 3 
1.1.3 ALD process requirements .......................................................................... 6 
1.3 Porous Anodic Alumina (PAA)........................................................................ 13 
1.4 Motivation and Objectives................................................................................ 16 
1.5 Organization of the thesis ................................................................................. 20 
Chapter 2 : Experimental Setup .................................................................................. 22 
2.1 ALD Systems.................................................................................................... 24 
2.1.1 UHV-ALD Reactor.................................................................................... 24 
2.1.2 Beneq ALD Reactor................................................................................... 30 
2.2 Porous Anodic Alumina Templates.................................................................. 32 
2.2.1 Electropolished aluminum ......................................................................... 32 
2.2.2 Two-step anodization procedure................................................................ 33 
2.3 Spectroscopic Ellipsometry .............................................................................. 34 
 
 
v 
 
2.3.1 Operational Principles................................................................................ 34 
2.3.2 Sopra GES5 ellipsometer........................................................................... 38 
2.4 Scanning Electron Microscopy......................................................................... 39 
2.4.1 Operational principles................................................................................ 39 
2.4.2 Hitachi SU-70 SEM ................................................................................... 41 
2.4.3 Cross-sectional sample preparation ........................................................... 42 
2.5 Transmission Electron Microscopy .................................................................. 43 
2.5.1 Operational principles................................................................................ 43 
2.5.2 JEOL 2100F field emission TEM.............................................................. 45 
2.5.3 Nanotube sample preparation .................................................................... 45 
2.6 Energy Dispersive X-ray Spectroscopy............................................................ 46 
2.6.1 Operational principles................................................................................ 46 
2.6.2 Bruker EDS detector.................................................................................. 48 
2.6.3 Oxford EDS detector.................................................................................. 48 
2.7 Atomic Force Microscopy ................................................................................ 48 
2.7.1 Operation principles................................................................................... 48 
2.7.2 MFP-3D AFM............................................................................................ 51 
Chapter 3 :  Profile Evolution for Conformal ALD Over Roughened Nanotopography
..................................................................................................................................... 52 
3.1 Introduction....................................................................................................... 53 
3.2 Materials and Methods...................................................................................... 56 
3.3 Results & Discussion ........................................................................................ 57 
3.4 ALD-PAA Simulation ...................................................................................... 64 
 
 
vi 
 
3.5 Conclusion ........................................................................................................ 70 
Chapter 4 :  TiO2 ALD Thin Film Growth Behavior Across 2D Planar Wafers Using 
TTIP and Ozone.......................................................................................................... 72 
4.1 Introduction....................................................................................................... 73 
4.2 Experimental Setup........................................................................................... 76 
4.3 Results............................................................................................................... 78 
4.4 Discussion......................................................................................................... 86 
4.5 Conclusion ........................................................................................................ 90 
Chapter 5 : Metrology of ALD Conformality in 3D Ultra-high Aspect Ratio 
Nanopores ................................................................................................................... 91 
5.1 Introduction....................................................................................................... 92 
5.2 Materials and Methods...................................................................................... 94 
5.3 Results............................................................................................................... 97 
5.3.1 EDS Inferred Thickness............................................................................. 97 
5.3.2 TEM Nanotube Metrology....................................................................... 103 
5.4 Discussion....................................................................................................... 105 
5.4 Conclusions..................................................................................................... 107 
Chapter 6: ALD Conformality and Process Optimization in 3D Ultrahigh Aspect 
Ratio Nanopores........................................................................................................ 108 
6.1 Introduction..................................................................................................... 109 
6.2 Experimental Setup......................................................................................... 112 
6.3 Results............................................................................................................. 113 
6.4 Discussion....................................................................................................... 122 
 
 
vii 
 
6.5 Conclusions..................................................................................................... 125 
Chapter 7: Conclusions and Future Work................................................................. 127 
7.1 Conclusions..................................................................................................... 127 
7.2 Future Work.................................................................................................... 130 
Appendices................................................................................................................ 133 
Appendix I.  Substrate Temperature Calibration .................................................. 133 
Appendix II.  Reactant Vapor Pressure................................................................. 135 
Appendix III. Trimethylaluminum (TMA) Gas Line Delivery System................ 136 
Appendix IV.  Calibration of Ozone Dose............................................................ 137 
Appendix V. HfO2 ALD Nanotube Fabrication[44]............................................. 138 
References................................................................................................................. 141 
 
 
 
viii 
 
List of Tables 
Table 1.   Short list of materials deposited by ALD.  Puurunen listed more than 75 
elements and 125 compounds that could be deposited by ALD in his 2005 
review.[4]................................................................................................................ 3 
Table 2. Bond lengths (?) and angles (degrees) associated with the titanium 
tetraisopropoxide molecule shown in Figure 38.[55] .......................................... 88 
 
List of Figures 
Figure 1.  Linear correlation between cycle number and deposited film thickness in 
ZrO2 ALD at 275 ?C.  Demonstrates the unprecedented thickness control ALD has 
to offer atomic layer by atomic layer.[7]................................................................ 2 
Figure 2.  TEM image of 210 ? hafnium aluminate film deposited by ALD in a high 
aspect ratio SiO2 trench.[7].................................................................................... 2 
Figure 3. Schematic of the Al2O3 ALD process sequence illustrating the four cyclic 
process steps of 1) metal exposure, 2) purge, 3) oxidizing exposure, and 4) purge.
................................................................................................................................. 5 
Figure 4.  Schematic of the possible behavior for the ALD growth rate versus 
temperature displaying the ?ideal? ALD window.[63].......................................... 7 
Figure 5.  Al2O3 deposition rate as a function of H2O exposure time and purge time.  
Saturating self-limited reactions are demonstrated by a plateau in thte growth 
 
 
ix 
 
rate once the reactant doses are at or above the saturation level.  Self-limiting 
purge times indicate reactions are not physisorbed on the surface due to excess 
precursor or parasitic-CVD reactions.[7].............................................................. 9 
Figure 6.  Schematic of the deposition rate as a function of purge time.  The figure 
displays the potential behaviors found in ALD..................................................... 10 
Figure 7.  Influence of the process pressures found in a laminar viscous flow regime 
on (a) across wafer uniformity and growth rates, as well as (b) conformality with 
in a high aspect ratio nanostructure.[65]............................................................. 12 
Figure 8.  PAA formed in 0.3M of oxalic acid at 40V, (a) PAA 3D Shcematic with 
labeled metrics, (b) Top down SEM displaying honeycomb structure and (c) 
cross-sectional view demonstrating straight, parallel nanopores adhered to an 
underlying aluminum substrate.[44] .................................................................... 15 
Figure 9.  Ragone plot detailing the state of energy storage technologies by 
contrasting their power density versus energy density......................................... 17 
Figure 10.  Custom built UHV ALD reactor ? a) ALD platform with load-lock and 
ALD chamber. b) Embedded mini ALD reactor in UHV chamber with cap in the 
down position. c) Schematic of ALD chamber...................................................... 25 
Figure 11 ? Cross-section of mini-reactor demonstrating gas flow direction. .......... 26 
Figure 12 ? View of the 100 mm diameter substrate heater with visualization of flows 
across the wafer. ................................................................................................... 26 
Figure 13.  Schematic of gas delivery setup.  TTIP is introduced in volume V for 1 sec 
and the released through the outlet ALD valve.  Description of TMA delivery line 
can be found in Appendix III................................................................................. 29 
 
 
x 
 
Figure 14.  Commercial Beneq ALD reactor ? (a) Cold wall vacuum chamber with 
serpentine IR heaters surrounding an embedded mini ALD reaction chamber (b) 
the mini ALD reaction chamber demonstrating the cross-flow direction of gaseous 
precursors. ............................................................................................................ 31 
Figure 15.  Ellipsometry measures the change of polarization of a linear polarized 
light when reflected on a surface.  Upon interference with the film, the reflected 
light is elliptically polarized.  Film properties, thickness, and refractive index are 
determined by the measure in phase change, ?, and amplitude change, ?, of the 
reflected light.[83] ................................................................................................ 36 
Figure 16.  Sopra GES5 spectroscopic ellipsometer configuration. .......................... 38 
Figure 17.  Schematic of an SEM setup.[87].............................................................. 41 
Figure 18.  Schematic of SEM sample preparation for AAO cross-sections.............. 42 
Figure 19.  Schematic of the path an electron beam takes within the evacuated 
column of a TEM.[88]........................................................................................... 44 
Figure 20.  Characteristic x-ray production............................................................... 47 
Figure 21. Schematic of the diminished electron beam interaction within a TEM 
sample versus an equivalent SEM sample.[90] .................................................... 48 
Figure 22.  SEM image of the AFM Si tip at the free end of a Si cantilever.[92] ...... 50 
Figure 23.  Schematic of an AFM operational setup.[92].......................................... 50 
Figure 24.  A removed PAA template imaged in SEM: (a) top-down view of the 
hexagonal nano-dimple array and (b) cross-section view of the scalloped Al 
surface, emphasizing the peak structure above the nano-dimples........................ 54 
Figure 25.  Progression of TiO2 ALD profiles across a scalloped Al surface formed 
 
 
xi 
 
using phosphoric anodizing acid imaged in SEM after (a) 0 cycles, (b) 1000 
cycles (~55 nm), (c) 2000 cycles (~110 nm), and (d) 3000 cycles (~165 nm).  (e) 
SEM image of a scalloped Al surface using oxalic anodizing acid coated with 733 
cycles (~40 nm) of TiO2.  Top-down images (d and e) display the same formation 
of triangular ALD surface profiles for the same ratio of equivalent thickness (t) to 
interpore spacing distance (Dint) for each anodizing acid. ................................. 59 
Figure 26.  A delaminated conformal TiO2 ALD film imaged in SEM: (a) tilted top-
down view of merged growth fronts influenced by the hexagonal pattern of the 
removed underlying surface and (b) tilted bottom-up view of the film where 
dashed arrows in the cross-section indicate where growth fronts merged and 
laterally moved due to film broadening at the peaks (solid black arrows) and 
causing the disappearance of the nano-dimples (solid white arrows). ................ 61 
Figure 27.  (a) AFM contour map of a 2000 cycle (~ 110 nm) TiO2 ALD film on a 
scalloped Al surface dictating the [111] and [121] directions with peaks A, B, C, 
and D (denoted by x) across the template.  (b) AFM line scans across the 121 
direction after 0 cycles, 1000 (~ 55 nm), 2000 cycles (~110 nm), and 3000 cycles 
(~165 nm) of deposited TiO2 ALD (arrows denote locations of peaks). The 
profiles are numbered 0-3 for intervals of 1000 cycles each................................ 63 
Figure 28.  Geometric 3D extrusion model: (a) five equally spaced data points on the 
surface of a sharp peak. (b) surface normals are projected out and the data points 
for the new surface are placed along them using the prescribed growth rate; (c) 
the new data is fit with a ?best fit? line; (d) the ?best fit? line is used to generate 
more data which is equally spaced in the x-direction........................................... 66 
 
 
xii 
 
Figure 29.  (a-d) Scalloped Al surface imaged in SEM after TiO2 ALD deposition of 0 
cycles, 1000 cycles (~55nm), 2000 cycles (~110nm), 3000 cycles (~165nm); (e-h) 
simulation of surface evolution by extruding a model of the initial surface at same 
rate as the deposition per cycle for ALD TiO2..................................................... 67 
Figure 30.  Cross-sectional surface profiles as a function of film thickness: (a) SEM 
image of three discontinuous 1000 cycle TiO2 ALD films along the 111 direction 
of a scalloped Al surface and (b) modeled profiles of TiO2 ALD films along the 
111 and 121 directions of a scalloped Al surface................................................. 68 
Figure 31.  Normalized surface area as a function of ALD film thickness................. 69 
Figure 32.  Film thicknesses as a function of number of ALD cycles as determined by 
ellipsometry for two different reaction temperature, 160 ?C and 200 ?C.  Linear 
regression demonstrates a growth rate of ~0.65 ?/cycle for the 200 ?C case and 
that the growth never really took off for the 160 ?C case..................................... 79 
Figure 33.  Representation of pressure within the mini-reactor during three different 
runs: (a) single ?pulse of TTIP, (b) 3 ?pulses of TTIP in parallel and (c) 3 
?pulses of TTIP in series. ..................................................................................... 80 
Figure 34.  Growth rate as a function of the number of TTIP molecules when doses 
are introduced in parallel or in series.  Demonstrates that an increase in 
molecules leads to higher growth rates and subsequent film densities, however, 
TTIP doses introduced in series have lower growth rates and densities as opposed 
to their in parallel counterparts............................................................................ 83 
Figure 35.  Growth rates as a function of TTIP purge time for two different TTIP 
 
 
xiii 
 
doses and an over-saturating dose of ozone (1.3 Torr-sec) for 250 cycles at a 
reaction temperature of 200 ?C. ........................................................................... 85 
Figure 36. Growth rate profiles of TiO2 ALD films after 250 cycles deposited on 
100mm wafers. ...................................................................................................... 86 
Figure 37.  Schematic of the structures of a free TTIP molecule, where the black ball 
is the Ti atom, the dark grey balls are oxygen, the light grey balls are carbon, and 
the white balls are hydrogen.  Their associated bond lengths and angles are 
reported in Table 3.[55] ....................................................................................... 88 
Figure 38.  (a) SEM image displaying a cleaved TiO2 coated AAO template epoxied 
to a silicon substrate demonstrating the position of an EDS line scan position and 
(b) the corresponding averaged Ti Intensity curves versus the length of the EDS 
line scan for the tilts 5, 0, and -5 degrees............................................................. 97 
Figure 39. Schematic of how the EDS inferred thickness method was derived. (a) 
Demonstrates an electron beam of width ? scans several nanotubes of length (L) 
at position x.  (b) Demonstrates the electron beam of area ?2 spanning several 
nanopores within the porous AAO template with an interpore spacing of Dint.  (c) 
Demonstrates film thickness at position x is a function of the inner radius at 
position x............................................................................................................. 102 
Figure 40.  Graph of EDS inferred film thickness as a function of pore length.  Curve 
displays a notable gradient down the length of the pore with film thickness 
measuring around ~7nm near the bottom of the poor. ....................................... 103 
Figure 41.  TEM images of HfO2 nanotubes made by deposition of HfO2 ALD into 
 
 
xiv 
 
porous AAO membranes followed by membrane dissolution;  a) Group of HfO2 
nanotubes demonstrating the nail-head top of the nanotube and thinning of wall 
thickness down the length of the nanotube and b) a single HfO2 nanotube broken 
midway down the full nanotube with an inset of the selected area diffraction 
pattern dictating that the films are amorphous.[44] .......................................... 104 
Figure 42.  TEM image of a TiO2 nanotube with wall thickness of ~7 nm near the 
bottom and ~9 nm about 0.5 ?m up the length of the nanotube. ........................ 105 
Figure 43.  Highly uniform films of TiO2 ALD thin films deposited across 2D planar 
wafers for 250 cycles for three different dosing schemes.  Demonstrates increased 
film thicknesses, yet maintained film uniformities. ............................................. 115 
Figure 44. a) TiO2 thickness profiles down the length of a 10 ?m deep nanopore for 
increasing doses introduced in parallel and b) their corresponding conformalities 
as a function of the number of TTIP molecules introduced in parallel for three 
different regions within the nanopore: 0-8.5 ?m, 0-5 ?m, and 5-8.5 ?m........... 117 
Figure 45.  Highly uniform films of TiO2 ALD thin films deposited across 2D planar 
wafers using an over-dose of ozone and a TTIP dose of 3 ?pulses (1.6 ?mols) in 
parallel followed by various purge lengths. ....................................................... 119 
Figure 46.  TiO2 thickness profiles down the length of a 10 ?m deep PAA nanopore 
for different purge times after 3 ?pulses introduced in parallel.  Profiles are 
associated with the planar depositions found in Figure 48................................ 119 
Figure 47. Conformality determined for thickness profiles found in Figure 49 as a 
function of the TTIP purge time following a dose of 3 ?pulses introduced in 
 
 
xv 
 
parallel................................................................................................................ 120 
Figure 48.  TiO2 thickness profiles down the length of a 3D ultra-high aspect ratio 
nanopore with an aspect ratio of 111 for different dosing schemes.  Profiles are 
associated with the planar depositions found in Figure 43................................ 121 
Figure 49.  Conformality determined for thickness profiles found in Figure 44, 46 and 
48 as a function of the number of TTIP molecules introduce into the system for 
various dosing schemes and purge times............................................................ 122 
 
 
1 
 
Chapter 1 : Introduction 
1.1 Atomic Layer Deposition (ALD) 
1.1.1 Background on ALD 
ALD is a gas phase deposition technique which utilizes self-limiting chemistries 
in order to produce thin solid films down to the atomic scale.    ALD is widely accepted 
as a means to produce high quality films with thickness control (Figure 1) over large 
surface areas with excellent uniformity and unprecedented conformality (Figure 2).[1-6]  
These key attributes are inherent to the self-limiting nature of the ALD surface reactions 
where it employs alternating reactant exposures of two gaseous precursors which are 
sequentially pulsed creating a saturated chemisorbed surface in order to achieve film 
growth one sub-monolayer at a time. [3, 4, 6] The sequential behavior of the precursors 
also helps avoid the formation of gas-phase particles that sometimes plague other gas 
phase deposition techniques in order to create thin, continuous, pin-hole free films which 
are critical to the engineering of the physical and electronic properties essential to the 
nano-community. 
 
 
2 
 
 
 
Figure 1.  Linear correlation between cycle 
number and deposited film thickness in ZrO2 
ALD at 275 ?C.  Demonstrates the 
unprecedented thickness control ALD has to 
offer atomic layer by atomic layer.[7] 
Figure 2.  TEM image of 210 ? hafnium 
aluminate film deposited by ALD in a 
high aspect ratio SiO2 trench.[7] 
 
Due to its ability to deposit films with excellent uniformity and conformality over 
complex topographies with unprecedented thickness control, ALD has emerged as a key 
technology used in the down scaling of semiconductor device dimensions such as the 
sub-nm gate dielectrics found in transistor and high density memory devices, such as 
metal-insulator-metal (MIM) stacks fabricated in trench and high aspect ratio dynamic 
random access memory (DRAM) structures.[3, 8-14]  Recently, ALD has expanded its 
range of applications beyond the field of conventional semiconductor technology and has 
pervaded in various fields, such as optoelectronics [15, 16], catalysis [17-19], energy 
devices [20-22], and micro-electro-mechanical systems (MEMS) [23-27].  ALD has also 
been a key driving force behind nanotechnology where complex 3D structures can be 
filled or coated to produce a wide range of nanowires [5, 28, 29], nanotubes [5, 21, 22, 
 
 
3 
 
29-31], and nanoparticles [32-39].  In order to support all of these advancing 
technologies, suitable precursors have been developed in order to deposit a wide variety 
of oxides, nitrides, semiconductor, and elemental materials (Table 1).[1, 4, 40] 
Oxides Nitrides Sulfides Semiconductors Elements 
Al2O3 [4, 41-
43] 
TiN[21] SrS GaAs Si 
In2O3 NbN CaS Si Ge 
ZrO2 [27, 42] TaN BaS InAs Cu[37] 
HfO2 [44-47] Ta3N5 ZnS InP Mo 
Ta2O5[48, 49] MoN  GaP W[50-53] 
SiO2[31] WN  InGaP Ta 
TiO2[43, 54-
57] 
BN   Ti 
SnO2[58]    Pt 
Nb2O5    Ru[39] 
ZnO[59]    Ni 
RuO[60]    Pd[38, 61] 
Table 1.   Short list of materials deposited by ALD.  Puurunen listed more than 
75 elements and 125 compounds that could be deposited by ALD in his 2005 
review.[4] 
 
1.1.2 ALD process sequence 
ALD is similar to chemical vapor deposition (CVD) in that they are both based on 
the chemisorption surface reactions, i.e., chemisorbed, where adsorbate species will 
chemically react with the atoms on the surface creating strong chemical bonds.[4]  ALD 
differs from CVD in that precursors are sequentially introduced in a cyclic fashion to a 
substrate?s surface rather than flown simultaneously, therefore, initiating self-limited 
surface reactions between a chemisorbed metal containing precursor and an adsorbed 
 
 
4 
 
nucleophilic reactant such as an -OH surface specie in order to deposit only one (or less 
than one) monolayer per cycle.[4, 7, 62]  To illustrate the basic concept of the ALD 
process sequence, we can consider the prototypical example of Al2O3 deposition resulting 
from the sequential exposures of the metal precursor Trimethylaluminum (Al(CH3)3) 
(TMA) and water.  During the first TMA half-cycle, TMA molecules are exposed to the 
surface and will adsorb on the active ligands present on the surface, in the case of TMA 
and water these are hydroxyl ?OH* groups.  A thermally activated reaction will occur 
splitting the (Al(CH3)3) molecules on the surface.  The CH3 ligands will combine with the 
hydroxyl surface groups and form a volatile organic compound, such as methane CH4, 
which will desorb from the surface.  The remaining part of the molecule chemisorbs to 
the surface chemically bonding the aluminum atom with either one or two oxygen atoms 
on the surface in order to deposit a stochiometric Al2O3 film.  This reaction is referred to 
as a ligand exchange reaction.[4]  When all the hydroxyl active sites that are physically 
accessible to the (Al(CH3)3) molecules then the surface is considered saturated and the 
half cycle is complete.  Excess TMA molecules that could not react on the surface as well 
as desorbed reaction by-products are purged out of the reactor.  For the 2nd half cycle, for 
this case water, water molecules will bind to the methyl groups through another ligand 
exchange reaction causing the release of more CH4 molecules.  The reaction steps can be 
seen in Figure 3 along with the reaction mechanisms displayed below. 
First half reaction: Al-(OH*) + Al(CH3)3 barb2right Al-O-Al(CH3)2* + CH4 
Second half reaction: Al-O-Al(CH3)2* + H2O barb2right Al-OH* + CH4 
Where the asterisks denote the active surface species[63] 
 
 
5 
 
 
Figure 3. Schematic of the Al2O3 ALD process sequence illustrating the four 
cyclic process steps of 1) metal exposure, 2) purge, 3) oxidizing exposure, and 4) 
purge. 
 The fundamental aspect of ALD is that after the completion of the 2 half 
reactions, the surface is functionalized again to react with TMA, therefore depositing a 
new layer by repeating the cycle.  This four step sequence constitutes a full reaction cycle 
in which a fixed amount of material is deposited, referred to as the ?growth rate? 
(thickness/cycle).[4]  Ideally, if sufficient doses are applied to saturate the surface and 
purge times are adequate to avoid parasitic CVD-like reactions, one monolayer of film is 
deposited.[58]  However, most processes are limited by the steric hindrance of the metal 
containing precursor due to their relatively large size, therefore, blocking active surface 
sites from being chemisorbed by other metal precursor molecules.[4, 64]  Thus, only a 
fraction of a monolayer (1/2ML) is actually deposited.[2]  In the case of Al2O3 the growth 
rate is ~0.1 nm/cycle.[63] 
 
 
6 
 
1.1.3 ALD process requirements 
Precise thickness control, excellent uniformity and unprecedented conformality 
are deemed the golden standards by which an ALD process is measured and are highly 
dependent upon the self-limiting behavior of the ALD process.  However, a number of 
interdependent parameters can highly influence the performance of the process.  This 
section will discuss only a few of the key requirements and challenges often found during 
ALD processing. 
1.1.3a Precursors 
Precursor chemistry plays a vital role in order to achieve ALD.[1]  Several 
requirements are taken into consideration when choosing a chemical precursor. (1) They 
must be volatile to produce a high enough vapor pressure (at least above 0.1 Torr) so they 
can easily be transported into the reaction zone of the process chamber.  However, many 
precursors are low vapor liquids or even solids at room temperature and therefore require 
higher temperatures to raise their vapor pressure.[1, 4]  (2) Precursors must be thermally 
stable in that they do not decompose in the source or on the substrate.[1, 4]  Reactions 
due to the decomposition of reactants are likely to produce non-self limited reaction with 
higher growth rates and possible film contamination of C, H, N or halide atoms 
depending on the precursor used.  (3) Precursors also must react rapidly and aggressively 
with surface groups from the previous exposure and not with its own adsorbates in order 
to achieve a saturated self-limiting surface in a short time (usually < 1s).[1]   
Different precursors have been employed for the deposition of an oxide ALD 
chemistry including metal halides and several types of organometallics for the metal 
 
 
7 
 
containing precursor and water, O2 plasma, and ozone for some of the more common 
oxidizers.[1, 4]  Halides have been the most widely studied of all the metal containing 
precursors, however, their main drawback lies in the corrosive nature of the HCl by-
product that is often produced causing etching of the deposited films and the encasing 
reactor.[1]  Several organometallic precursors have been investigated including metal 
alkyls, amides, alkoxides, and cyclopentadienyls, however, growth rates are usually low 
due to the steric hindrance associated with the large ligands attached to the molecules.[1, 
4] 
1.1.3b Temperature window  
ALD is a thermally activated process in order to produce self-limiting reactions 
on the surface of a substrate.[65]  Therefore, a temperature process window can be 
identified based on the constant growth rate of the ALD process (Figure 4).  Outside the 
ALD window, low temperatures are likely to exhibit low growth rates due to the fact that 
the reactant was unable to overcome the activation energy necessary to  
chemisorb on the surface.  In 
some cases high growth rates 
may occur as a result of 
precursor condensation 
ultimately leading to a CVD 
process that can physisorb on the 
surface if precursors are not 
purged properly.  This is 
 
Figure 4.  Schematic of the possible behavior for 
the ALD growth rate versus temperature 
displaying the ?ideal? ALD window.[63] 
 
 
8 
 
particularly troubling when dealing with water, as it is a highly polar molecule that tends 
to stick to the surface in multilayers.  However, high temperatures can lead to the 
decomposition of the ALD precursor or desorption of the adsorbates before they are 
reacted with the 2nd reactant.[4, 63] 
1.1.3c Saturating exposures 
In order to deposit uniform and conformal films across a planar surface and 
within a 3D nanostructure, sufficient precursor exposures and time must be supplied to 
fully saturate all the accessible active sites on the surface.[7, 66]  The product of a 
precursor?s partial pressure and the time it takes to fully saturate a surface (Pt) is a useful 
measure of the exposure dose a surface has had to the precursor, 
mkTSPt pi2=     Eq. 1 
Where, S is the number of molecules to saturate a surface per area, m is the molecular 
mass of the precursor, k is Boltzmann?s constant, and T is the temperature.[66]     
Determining the proper saturating dose is usually the first step in optimizing an ALD 
process and is achieved by incrementally increasing the dose of one precursor and  
 
 
9 
 
measuring the growth rate.  As 
indicated by the self-limiting nature 
of ALD, upon dose saturation, the 
growth rate will plateau designating 
the ALD regime as seen in Figure 
5.[7]  However, when depositing 
ALD within the confines of a high 
aspect ratio nanopore it becomes 
considerably more difficult to 
provide saturating exposures, not 
only due to an increase in surface 
area, but also because the reaction 
kinetics are now diffusion 
limited.[66, 67]  Therefore, Gordon et al. proposed a model where the exposure dose 
within a high aspect ratio nanopore is a 2nd order function of the aspect ratio, a, and 
require longer exposure times to completely saturate the surface.[66] 
)234191(2 2aamkTSPt ++= pi    Eq. 2 
Optimizing the exposure dose is important in that over-dosing a precursor will require 
longer purge times and waste expensive precursor, while under-dosing can lead to 
thickness gradients and lower growth rates across a surface.[68] 
 
 
Figure 5.  Al2O3 deposition rate as a function 
of H2O exposure time and purge time.  
Saturating self-limited reactions are 
demonstrated by a plateau in thte growth rate 
once the reactant doses are at or above the 
saturation level.  Self-limiting purge times 
indicate reactions are not physisorbed on the 
surface due to excess precursor or parasitic-
CVD reactions.[7] 
 
 
10 
 
1.1.3e Purge window  
Purge time is a critical component of the ALD cycle sequence.  Insufficient purge 
times to fully exhaust any unreacted precursor or subsequent by-products can result in an 
overlap of the two ALD precursors, i.e. simultaneous presence of both precursors, thus 
allowing the ALD process to transition to a CVD mode.[6]  Similar to the temperature 
window, a purge window can be created as well (Figure 6).  If a CVD mode ensues 
upstream to the reaction zone, the reactants are likely to result in gas phase reactions and 
subsequently lower growth rates.  However, if the temperature is too low, multilayer 
adsorption (physisorborption) can occur in which the excess reactants will react with the 
next precursor pulse resulting in higher growth rates on the surface.  Water is particularly 
prone to this behavior due to its polarity and natural ?sticky? behavior requiring longer 
purge times.[68]  However, when purges are too long, desorption and thermal  
decomposition of the adsorbed 
species can take place leading to 
possible film contamination or 
compositional changes in the film.  
Figure 5 demonstrates the self-
limiting behavior of the ALD process 
when purge times were incrementally 
increased by measuring film growth 
rates to determine growth was not in a 
CVD mode. 
 
Figure 6.  Schematic of the deposition rate as 
a function of purge time.  The figure displays 
the potential behaviors found in ALD. 
 
 
11 
 
1.1.3d Pressure regime 
Maintaining atomic level thickness control with wafer scale uniformity and 
conformality within high aspect ratio nanopores is dependent upon the complex reaction 
dynamics associated with chemisorbed reactions found in different flow regimes.   An 
important issue of the ALD process is the distinction between self-limiting and 
saturation.  The self-limiting behavior of ALD asserts that a reactant in the gas phase can 
only chemisorb with complementary active surface species and not with its own 
adsorbates or by-products, thus guaranteeing that the reaction will be self-limited.[4]  
However, chemisorption is a localized reaction affected by the surface energy and 
repulsion between neighboring adsorbates, therefore only allowing certain positions on 
the surface to bind with a molecule.[64]  This means that upon extended pulse times, the 
reaction doesn?t necessarily terminate because all the active sites have been consumed.  
However, reaction saturation occurs when an equilibrium between desorption of 
excessive precursor and adsorption of further species is reached for a given temperature 
and pressure.  Therefore, if all the active surface sites have been occupied by 
chemisorbed species, then increasing the partial pressure of a reactant should have no 
effect.[4]   
The process pressure within a system is directly related to the surface residence 
time of a precursor.  Li et al. demonstrated the impact of process pressure on film growth 
rates and wafer uniformity and high aspect ratio conformality (i.e., step coverage) in an 
Al2O3 ALD process when using Al(CH3)3 and O3.  As seen in Figure 7a, the across wafer 
uniformity is compromised as a function of decreased reactor pressure; however, 
uniformity was restored by increasing the process temperature.  Similarly, Figure 7b 
 
 
12 
 
illustrates a decrease in conformality within a nanopore with an aspect ratio of 13 when 
the process pressures become too low.  The authors suggested the increase in growth 
rates at higher pressures was attributed to the change in surface saturation levels, the 
enhancement of precursor diffusion, as well as the higher probability for surface reaction 
to occur at the active sites.[65]  Both of these cases demonstrate that the rate limiting step 
within a viscous flow regime (Knudsen number, Kn, < 0.01), meaning the velocity and 
surface irregularities are small enough for a gas to flow gently around obstructions in a 
laminar streamline, is the surface reaction kinetics.[69]   
 
Figure 7.  Influence of the process pressures found in a laminar viscous flow 
regime on (a) across wafer uniformity and growth rates, as well as (b) 
conformality with in a high aspect ratio nanostructure.[65] 
 
When the flow regime transitions to a molecular flow regime (Kn > 1) often 
associated with ultra-high aspect ratio nanopores where the mean free path of a precursor 
is equal to or greater than the nanopore diameter, the deposition kinetics become subject 
to the Knudsen diffusion of the precursor.[69]  The idea in Knudsen diffusion is that a 
molecule will ballistically transport its way down the length of the pore and interact with 
the pore wall before it interacts with another molecule.  Once the molecule hits the wall it 
 
 
13 
 
can either chemisorb to the surface or it can deflect off changing its course of trajectory 
and probably a losing some kinetic energy.[69]  Therefore, when in a molecular flow 
regime the rate limiting step is the Knudsen diffusion and will require not only longer 
surface residence times of a molecule to diffuse down a nanopore, but will require longer 
purge times for unreacted precursor and by-products to diffuse out.[66] 
1.3 Porous Anodic Alumina (PAA) 
Aluminum naturally has a highly resistant coating of oxide along its surface and 
the enhancement of this oxide to produce a porous anodic oxide film has garnered a lot of 
interest in self-assembled nanotechnology.  PAA membranes are hexagonally ordered, 
self-aligned cylindrical nanopores formed by the controlled electrochemical oxidation or 
anodization of aluminum metal in certain electrolytic forming solutions.[44, 70-72]  
Depending on the type of electrolytic solution used different film growth mechanisms 
occur.  Anodization in neutral pH solutions, such as neutral boric acid solution, 
ammonium borate, or a tartrate aqueous solutions, produce completely insoluble non-
porous barrier anodic alumina (BAA) membranes; whereas, slightly acidic solutions, 
such as phosphoric, oxalic, or sulfuric acids will produce a slightly soluble porous anodic 
alumina (PAA) membrane.[70, 71]  These two types of films differ in how thick one can 
be produced and their controlling parameters.  When aluminum is anodized at a constant 
voltage held below the oxide breakdown voltage in a neutral pH solution, the native 
barrier oxide film is thickened at a constant rate of 1.4 nm V-1, therefore, forming a 
protective BAA membrane ~1,000 nm thick.[70]  However, the thicknesses of a PAA 
membrane is time dependent and can grow many times thicker than the upper limit 
 
 
14 
 
placed upon BAA films. 
PAA is a balance between field-assisted chemical dissolution at the 
electrolyte/oxide interface and the oxide formation at the oxide/metal interface.  A BAA 
layer forms when the Al metal surface reacts with O2- and OH- ions at the anode allowing 
the anions to diffuse into the oxide layer and continuously react with the metal beneath.  
Field-assisted migration moves Al3+ ions from the oxide/metal interface to the 
oxide/electrolyte interface, where Al-O bonds stretch under the high electric fields.  This 
leads to the dissolution of Al3+ ions at the oxide/electrolyte interface, a process aided by 
Joule heating due to the high fields at the pore tips.  The balance of the field-assisted and 
chemical dissolution at the electrolyte/oxide interface and the continual diffusion of free 
O2- ions across the oxide layer reacting with the Al metal at the oxide/metal interface 
results in the formation of porous alumina membranes which can continually grow as 
long as conditions remain the same.[70, 71] 
The unique behavior of PAA to self-order in to a hexagonal array of parallel 
nanopores is highly dependent upon the anodization conditions.[72]  Three types of 
electrolytic acid solutions with associated voltages have been identified to lead to 
optimum self-ordering of the PAA membrane: 2.7 wt% aqueous oxalic acid at 40 V, 20 
wt% aqueous sulfuric acid at 19 V, and 10 wt% phosphoric acid at 160 V.[73, 74]  All 
three acids lead to a hexagonal arrangement of the pores, but each provides a unique 
interpore spacing, Dint, and pore diameter, Ddia (Figure 8a), which are linearly 
proportional to the anodizing potential.[70, 73]  It has been experimentally determined 
that the interpore spacings for sulfuric (~ 25 V), oxalic (~ 40 V), and phosphoric (~ 
 
 
15 
 
200V) are 66.3 nm, 105 nm, and 501 nm, respectively, while pore diameters were 
determine to be 24 nm, 31 nm, and 158 nm, respectively.[73, 74]  Therefore, as the 
anodizing voltage increases, the oxide cell increases and consequently the pore density 
will decrease.[70]  Figure 8b-c displays SEM images of a PAA template anodized in 
0.3M of aqueous oxalic acid anodized at 40 V and 8 ?C.  Figure 8b illustrates the top 
down honeycomb structure while Figure 8c exhibits the straight, parallel nanopores of the 
PAA membrane aligned orthogonally to the underlying aluminum substrate. 
 
Figure 8.  PAA formed in 0.3M of oxalic acid at 40V, (a) PAA 3D Shcematic with 
labeled metrics, (b) Top down SEM displaying honeycomb structure and (c) 
cross-sectional view demonstrating straight, parallel nanopores adhered to an 
underlying aluminum substrate.[44] 
 
 
 
16 
 
1.4 Motivation and Objectives 
As device designs head towards the nanoscale, conventional ULSI technology is 
reaching its limits to pattern and access such small device dimensions seen in technology 
today.  Therefore, a strong driving force behind nanotechnology is the combination of 
self-limiting processes with nature?s ability to self-assemble.  Atomic layer deposition is 
proving to be a powerful self-limiting technique to deposit extremely thin films with 
excellent uniformity and unprecedented conformality even over the most demanding 
nanotopographies.  The nature of PAA to self-align and self-order in order to create a 
highly dense, parallel porous membrane over large surface areas has rendered it a useful 
material for nano-templating applications and device platform applications.   
Energy storage is one such area that has benefited from the combination of ALD 
and PAA.  Figure 9 demonstrates a Ragone plot displaying the state of energy storage 
technologies found today in terms of power density and energy density.  Ideally, energy 
storage technologies should have a high energy density (capacity to store charge) as well 
as boast high power densities (rapid charge/discharge rates).  Current battery technology 
has reasonably good energy density but lacks the power density due to its dependence on 
bulk ionic movement and chemical reactions to charge and discharge.  Recent work on 
Lithium ion superbatteries have shown that coupling nanostructures and the materials 
science of the bulk material can lead to a high power density.[75]  Electrochemical 
capacitors store charge by creating a double layer of ionic charge close to the surface of 
an electrode, therefore requiring high surface area electrodes.[76]  Nanostructures with 
high surface areas have provided electrochemical capacitors with increased energy 
 
 
17 
 
densities, as well as maintaining high power density common to this technology.[77]  
Similar to electrochemical capacitors, electrostatic capacitors are dependent upon high 
surface areas in order to achieve high energy densities.  Electrostatic capacitors make use 
of a thin dielectric material between two metal electrodes in order to store charge, and are 
therefore not limited by electrochemical reactions or ionic conduction to release or carry 
a charge.[76]  However, electrostatic capacitors lack the energy density to make them a 
viable technology as compared to electrochemical capacitors and batteries.   
 
Figure 9.  Ragone plot detailing the state of energy storage technologies by 
contrasting their power density versus energy density. 
 
Recently our group has demonstrated an operational metal-insulating-metal 
(MIM) electrostatic nanocapacitors fabricated out of multilayer ALD thin films within 
the confines of an ultra-high aspect ratio PAA membrane.  Using Al2O3 as a high-k 
 
 
18 
 
dielectric (?? ~9) we have demonstrated equivalent planar capacitances of up to 100 
?F/cm2 within a 10 ?m deep nanopore.[21]  It is believed through the use of even higher-
k materials, such as HfO2 (?k ~14) and TiO2 (?? ~80), as well as the use of higher aspect 
ratio nanopores, higher capacitance values can be achieved. 
While the favorable nature of ALD to be self-limiting and saturating to provide 
exceptional coatings of thin films across planar wafers and in ultra-high aspect ratio 
nanosturctures, the reaction dynamics associated with different flow regimes, 
chemisorbed reactions, and interfering surface topographies can be quite complex.  This 
is particularly true for the type of precursors used in ALD (organometallics, halides, 
water, and ozone) which can produce deviations from the ?ideal? monolayer self-limiting 
behavior due to multilayer physisorption, steric hindrance, secondary reactions, and 
parasitic CVD reactions.  Therefore, it is important to identify the process parameters that 
constitute an ALD process window, where the benefits of ALD are realized.  Therefore, 
the objective of this work is to gain insight on the growth behavior of ALD uniformity 
across 2D planar wafers in relation to ALD conformality within an ultra-high aspect ratio 
nanopore. 
In order to understand the relationship between the 2D and 3D scale, we first 
investigate ALD?s ability to conformally coat a complex topography found in the viscous 
flow regime.  We employ the nano-roughened surface topography of a stripped 1st 
anodized PAA surface in order to observe the influence the underlying topography will 
exhibit on the resulting ALD surface profiles as deposition conformally coats the surface.  
We employ a commercial ALD reactor in order to deposit conformal TiO2 ALD films 
 
 
19 
 
using Tetrakis(dimethylamino) titanium (TDMAT) and water.  It is imperative to 
understand the progression of ALD surface profiles in order to know the benefits or 
disadvantages of ALD to modify sharp surface structures.  This is of particular 
importance to us because PAA membranes are plagued with sharp surface asperities that 
result in localized electric fields and low field breakdown in our MIM capacitor 
technology. 
Of further interest, is the investigation of deposition mechanisms when 
transitioning from a 2D planar wafer into that of a 3D ultra-high aspect ratio nanopore.  
ALD is based upon a sequence of self-limiting and saturating behavior that often relies 
upon interdependent process parameters such as reactant dose, surface residence time, 
and purge times for specific precursors used.  When a nanopore is introduced into the 
substrate, the surface area is increased, resulting in the need for more precursor molecules 
to completely saturate a surface.  However, an increase in molecules does not necessarily 
result in conformal films down the length of a nanopore.  The molecules also need time 
to diffuse down (surface residence time) and back up (purge time) the length of the 
nanopore.  Therefore, it is imperative to study the relationship between various process 
parameters in different flow regimes, as it will be of great importance to the optimization 
of the ALD process in order to create uniform and conformal films.  For the purpose of 
this work we employ a custom-built ultra-high vacuum (UHV) ALD system to deposit 
the high-k dielectric material TiO2 using the precursors TTIP and ozone.  As a means to 
characterize these structures, we propose a novel method to extract film thicknesses from 
energy dispersive spectroscopy (EDS) line scans and a cylindrical assumption. 
 
 
20 
 
1.5 Organization of the thesis 
This thesis consists of six chapters in addition to the first Introduction chapter.  The 
summary of each chapter is given below. 
Chapter 2:  This chapter reviews experimental setups that were implemented to carry 
out this research.  It primarily focuses on the design of two ALD reactors, one custom 
built and one commercial, the nano-template synthesis process of the PAA templates, and 
a variety of ex-situ characterization techniques used to characterize the ALD thin films. 
Chapter 3:  This chapter focuses on the influence a nano-roughened topography can 
induce on ALD surface profiles during continual deposition and illuminate the use of 
ALD to modify surface structures with sharp asperities. 
Chapter 4:  This chapter focuses on the growth behavior of TiO2 ALD across planar 
surfaces when using titanium tetraisopropoxide (TTIP) and ozone as the metal and 
oxidizing precursors.  This chapter will emphasize the critical role a precursor?s surface 
residence time can have on the resulting chemistry?s growth rate and uniformity. 
Chapter 5:  This chapter focuses on a methodology to extract thin film thicknesses 
when deposited in an ultra-high aspect ratio nanopore.  It proposes the use of energy 
dispersive spectroscopy (EDS) line scans to infer film thickness through a cylindrical 
assumption. 
Chapter 6:  This chapter focuses on the influence of complex interdependent ALD 
process parameters such as precursor dose, surface residence time and purge on the 
conformality of ALD thin films within 3D ultra-high aspect ratio nanopores versus the 
 
 
21 
 
uniform growth behavior across a 2D planar surface when using TTIP and ozone as 
precursors. 
Chapter 7:  This thesis is concluded with a summary of the contributions of this 
work and some thoughts regarding the future of this work. 
 
 
22 
 
Chapter 2 : Experimental Setup 
The following chapter focuses on the tools and materials that were implemented 
in order to carry out the proposed research.  Two ALD reactors were used in this study, 
the first being a custom built ultra-high vacuum (UHV) ALD reactor.  It was designed in 
2003 and built by Dr. Wei Lei and Dr. Laurent Henn-Lecordier, two prior doctoral 
students.  The tool was originally designed to investigate the use of mass-spectrometry in 
W ALD processing for in-situ sensing and real-time wafer-state metrology.[52, 78]  After 
the completion of this work, a funded collaboration with the equipment supplier MKS 
Instruments was initiated to investigate new venues for ALD process optimization.  The 
tool was upgraded with an abundance of donated equipment, including an O3mega TM 
ozone generator, a MKS Vision 1000C mass-spectrometer, and a liquid injection system 
for low vapor pressure precursors (not used in this work).  Additionally a new delivery 
system was built to accommodate low vapor pressure precursors whose mode of delivery 
was very different than the W ALD precursors, tungsten hexafluoride and silane, which 
are both gases above atmospheric pressure at room temperature.   
A commercial Beneq TFS 500 hot wall ALD reactor was purchased and installed 
in the Maryland Nanocenter?s FabLab facilities in 2008.  Designed to be the work horse 
for the Nanocenter?s ALD thin film needs, the tool was originally configured to deposit 
such films as Al2O3, HfO2, TiO2, ZnO, and TiN.  The tool allowed users to rapidly 
deposit ultra thin films over planar and nanoscale topography with relative ease. 
In order to facilitate quantitative evaluation of ALD conformality in very high 
 
 
23 
 
aspect ratio nanopores, a nanoporous template that was easily produced and compatible 
with high resolution metrology of ALD film profiles was required.  Therefore, custom 
made porous anodic alumina membranes were used as a versatile and easy platform to 
investigate the influence of key process parameters on the conformality of titanium 
dioxide (TiO2) ALD thin films in an ultra-high aspect ratio nanopore. 
Being able to easily characterize the ALD films deposited was essential to 
evaluating and rapidly optimizing the ALD process performances at the 2D wafer and 3D 
nanopore scales.  A Sopra GES5 spectroscopic ellipsometer was acquired to provide 
maps of thicknesses and optical properties across the 2D wafer on a run-to-run basis.  
Through the use of the Maryland Nanocenter?s NISPLab, a Hitachi SU-70 high 
resolution scanning electron microscope (SEM) equipped with a Bruker energy 
dispersive spectrometer (EDS) and a JEOL 2100F Field Emission transmission electron 
microscope equipped with an Oxford EDS were both readily available in order to image 
samples and determine film composition and thicknesses. A MFP-3D atomic force 
microscope was easily accessible through Dr. Seog?s lab in order to map ALD surface 
profiles across a roughened sample. 
 
 
24 
 
2.1 ALD Systems 
2.1.1 UHV-ALD Reactor 
2.1.1a Reactor design 
A custom-built ultra-high vacuum (UHV) stainless steel ALD reactor was used in 
this study (Figure 10a).  The system consists of two 8?-diameter 6-way stainless steel 
conflate (CF) chambers employed as a load-lock and ALD process chambers.  A small 50 
L/s turbomolecular pump (TMP) is used to pump down the load-lock from atmospheric 
pressure (760 Torr) to pressures below 1E-5 Torr within 5 minutes.  The load-lock is 
imperative for rapid wafer throughput by limiting exposure of the main process chamber 
to atmosphere while minimizing potential surface contamination of the ALD reactor 
environment and wall memory effects during wafer loading and unloading.  In order to 
minimize the amount of water physisorbed on the stainless steel walls of the ALD 
process chamber, heating tapes were used to maintain a temperature of 150 ?C while a 
Pfeiffer TPU261 turbomolecular pump (TMP) with a pumping speed of 260 L/s provided 
a base pressure below 1E-8 Torr with no gas load present.   
 
 
25 
 
 
Due to the digital process sequence of ALD, fast cycle times of reactant gases 
with sufficient purge times are critical to avoid parasitic CVD-like reactions.  Therefore 
the large volume of the process chamber, ~16 liters, makes fast cycling times of reactant 
gases rather difficult and would require extremely long purge times and high 
temperatures to suffice desorption of reactive species out of the chamber and off the 
walls.  Therefore, a smaller mini-reactor was embedded in the main UHV process 
chamber as seen in Figure 10b.  The volume of the mini-reactor measured 0.2 liters, 
therefore allowing rapid gas exchange during ALD exposures and shorter purge times.  
Compared to the large scale process reactor, the use of a mini-reactor allowed for shorter 
ALD mini- reactor
a)
b)
100 mm wafer
Substrate 
heater
Z-axis pneumatic
actuator
ALD chamber
0.1-1 Torr
10-5 Torr
Gas InletGas Outlet
Moveable cap
c)
Load-lock
 
Figure 10.  Custom built UHV ALD reactor ? a) ALD platform with load-lock 
and ALD chamber. b) Embedded mini ALD reactor in UHV chamber with cap 
in the down position. c) Schematic of ALD chamber. 
 
 
26 
 
residence times ~1 sec and minimized the ratio of wall to wafer surface areas to about 3 
to 1 making it a suitable environment for in-situ process sensing at the wafer scale using 
mass-spectrometry, as well as, minimizing chamber ?memory effects? often associated 
with adsorption/desorption phenomena.[52, 68, 78] 
 The mini-reactor features a pneumatically actuated stainless steel cap that is 
lowered and raised in order to easily transfer 100 mm wafers from the load-lock onto the 
substrate heater.  When the cap is in the down position, a rough vacuum seal is provided 
between the mini-reactor and the main chamber; therefore, creating a pressure differential 
respectively from ~0.1 - 1 Torr to 1E-5 ? 1E-4 Torr depending on the gas load applied 
(Figure 10c). 
 The mini-reactor consists of a custom designed 100 mm stainless steel substrate 
heater.  The base of the heater was integrated with ?? diameter tubing that operated as 
gas inlet and exhaust outlet lines.  Process gases were introduced and exhausted through 
two 1 mm wide by 100 mm long, diametrically opposing slits positioned 3mm above the 
 12
Substrate 
heater
 
Figure 11 ? Cross-section of mini-reactor 
demonstrating gas flow direction. 
Figure 12 ? View of the 100 mm 
diameter substrate heater with 
visualization of flows across the 
wafer. 
 
 
27 
 
wafer as seen in Figure 11.  Figure 12 shows a top down view of the 100 mm substrate 
heater demonstrating the cross-flow direction of gases across a wafer.  An MKS 
Instruments capacitance monometer pressure gauge was integrated in the downstream 
exhaust line in order to provide the pressure reading for the mini-reactor.  The operating 
pressure within the mini-reactor was held around 130 mTorr while the gases where 
exhausted to a 20 ft3/min mechanical roughing pump.  Under these conditions a viscous, 
laminar flow was achieved within the mini-reactor and across the wafer.   
 The temperature of the substrate heating coil is regulated by a proportional 
integral derivative (PID) control loop based on the feedback of a thermocouple attached 
on the underside of the heater.  A 100 mm 13-point SensArrayTM thermocouple wafer 
was used to calibrate the substrate temperature (Appendix I, Figure a).  At 250 ?C and 0.1 
Torr the temperature non-uniformity was measured to be less than 5% across the wafer 
determined by the standard deviation divided by the average temperature (?/?).  It was 
critical to calibrate the actual substrate temperature as a function of the PID control 
temperature as pressure and the outer UHV chamber temperature will significantly 
impact the thermal coupling between the heater and the wafer. The mini-reactor is 
considered an isothermal environment meaning the outer UHV chamber has significant 
effect on the substrate temperature which creates a significant offset from the PID 
controller setpoint temperature and the actual substrate temperature (Appendix I, Figure 
b, c).  Therefore, all temperatures hereafter refer to the estimated actual wafer 
temperature.  The cap is only conductively heated through contact with the substrate 
heater and therefore is about 20 ?C lower than the heater temperature and was subject to 
deposition as well. 
 
 
28 
 
2.1.1b Gas delivery and reactant dose control 
 Liquid titanium tetraisopropoxide (Ti[OCH(CH3)2]4) (TTIP) was used as the 
titanium metal precursor.  A 20mL stainless steel reservoir containing 5g of liquid TTIP 
was heated using heating tapes connected to a PID temperature controller.  The source 
was maintained at 58 oC in order to achieve a vapor pressure of ~0.6 Torr (Appendix II).  
In order to achieve reproducible doses of the TTIP precursor, fast action pneumatic ALD 
valves were implemented on the inlet and outlet of a temperature-controlled 20 mL 
volume monitored by a 0-100 Torr capacitance manometer pressure gauge (Figure 13).  
TTIP doses were controlled by timing the opening of these valves.  The inlet valve to the 
volume was opened for 1 s in order to fill the volume to the vapor pressure of the TTIP 
precursor.  Upon reaching the given time setpoint, the inlet valve between the precursor 
source and the volume was closed and the gases were released into the chamber for 0.3 s.  
This time was deemed adequate to completely empty the 20 mL volume of TTIP, with 
the exception of some adsorption on the walls.  By monitoring the minimum and 
maximum pressures during each cycle, the doses were approximated from the ideal gas 
law, pV=nkT (where p is the pressure, V the volume, n the number of ?mols, k the 
Boltzmann?s constant, and T the absolute temperature).  A 0.47 ?mol dose of TTIP was 
deemed sufficient to achieve full saturation across a Si wafer within the reactor chamber.   
 
 
29 
 
 
Figure 13.  Schematic of gas delivery setup.  TTIP is introduced in volume V for 
1 sec and the released through the outlet ALD valve.  Description of TMA 
delivery line can be found in Appendix III. 
 
Ozone was obtained by flowing 200 sccm of O2 (Grade 5, Praxair) into an MKS 
Instruments O3mega ozone generator.  The flow produced a 20psig pressure in the single 
electrical discharge cell.   At 100% power, the ozone concentration at the outlet was up to 
20 wt% (300 g/m3) with the balance being O2.  Due to the high flow and high pressure 
required for the operation of the O3megaTM a different control scheme of the ozone 
exposure was necessary.  The O2/O3 mixture was flowed continuously into an HA-series 
catalytic ozone destruct from Ozone Engineering to deactivate ozone into oxygen.  
Therefore, only a fraction of the total flow was pulsed into the reactor using a Swagelok 
medium flow metering valve in series with a fast action pneumatic ALD valve (Figure 
13).  For a given needle valve conductance, the number of ?mols of ozone was 
 
 
30 
 
determined by integrating the pressure peaks measured in the mini-reactor for different 
exposure times and comparing them to integrated peaks from known doses determined 
through pressure measurements in an attached trimethylaluminum  (TMA) vessel 
(Appendix IV). 
2.1.1c Process Control 
An automated PC-based control platform created by National Instruments 
LabViewTM was used to accurately control the valve sequences during precursor 
exposures and purges, as well as monitor pressure signals in the precursor fill volumes 
and reactor.  Recipes were created in an excel spreadsheet and uploaded to the platform 
in order to define process setpoints for each precursor, the sequence of exposures and 
purges, as well as the number of total cycles the process consist of. 
2.1.2 Beneq ALD Reactor 
2.1.2a Reactor design 
A commercial TFS 500 cross flow reactor from Beneq was used for part of this study.  
The reactor features a cold wall vacuum chamber with a constant nitrogen flow of 100 
sccm creating a base pressure of ~7 Torr.  Embedded within the vacuum chamber is a 
mini hot-walled furnace reaction chamber with a base pressure of ~1 Torr (Figure 14).  
The outer vacuum chamber is pumped down through the reaction chamber creating a 
pressure differential creating a sealing flow ensuring that all reactions occur within the 
reaction chamber and over the sample.  The mini-reactor also decreases the wall to wafer 
surface ratio helping to restrict reactions to the wafer surface.  Samples are loaded 
through a large chamber door, where the cap of the mini-reactor is removed by simply 
 
 
31 
 
lifting using an improve lever due to heat (similar to removing the lid to a pot on a stove).  
The system uses a water cooling system in order to keep the outer vacuum chamber?s 
walls cold while serpentine IR heaters surround the inner vacuum chamber radiating heat 
to the mini-reactor chamber.  Therefore, the mini-reactor and subsequent substrate within 
is heated by both radiation and conduction.  The reaction chamber utilized a cross-flow 
design, introducing and exhausting process gases through two diametrically opposing 
holes.  Gases then pass through an equally spaced porous grate allowing flow to be 
equally distributed over the wafer. 
 
Figure 14.  Commercial Beneq ALD reactor ? (a) Cold wall vacuum chamber 
with serpentine IR heaters surrounding an embedded mini ALD reaction 
chamber (b) the mini ALD reaction chamber demonstrating the cross-flow 
direction of gaseous precursors. 
 
 
 
32 
 
2.1.2b Gas delivery and reactant dose control 
Liquid tetrakis(dimethylamino) titanium ( Ti[N(CH4)2]4, TDMAT) was used as the 
titanium metal precursor.  A 200 mL stainless steel bubbler containing 20g of liquid 
TDMAT was heated in the Beneq hot source HS200 compartment.  The source was 
maintained at 60 ?C in order to achieve a vapor pressure of ~1 Torr (Appendix II).  Pulses 
of TDMAT were created by pulsing 100 sccm of nitrogen through the headspace of the 
TDAMT container (not a true bubbler system) and were introduced into the reaction 
chamber by leaving a high temperature ALD valve open for a set amount of time.  The 
downside to this pseudo-bubbler approach is that the amount of precursor is limited by 
the vapor pressure of the precursor.  If higher doses are required the set temperature on 
the source needs to be raised, which could be dangerous if heading towards the 
decomposition temperature.  The decomposition temperature of TDMAT is 
approximately 90 ?C.  Water was used as the oxidizing precursor.  A 200 mL stainless 
steel bubbler containg DI water was maintained at room temperature.  Due to water?s 
high vapor pressure (Appendix II), pulses were timed using an pneumatically actuated 3-
way valve in order to determine complete purging of the lines after pulsing. 
2.2 Porous Anodic Alumina Templates 
2.2.1 Electropolished aluminum  
Aluminum foils of high purity (99.99% or better) were cut to size, rinsed with 
acetone and electropolished in order to remove any particles and micro-roughness from 
the surface that can greatly influence the quality and growth of the final pores.  Foils 
 
 
33 
 
were placed in a 1:5 perchloric acid and ethanol bath held at ~3 ?C.  In order to keep the 
bath cool during electropolishing a jacket beaker connected to a circulator pump was 
used.  A mixture of 1:1 ethylene glycol and water was circulated through the jacketed 
portion of the beaker helping to maintain the ~3 ?C necessary to keep the aluminum foil 
from burning due to the Joule heating during electropolishing.  It was important to 
maintain a constant stir rate within the bath solution in order to help dissipate 
accumulating heat as well as help keep an even perchloric acid concentration throughout 
the solution.   
 A sheet of either aluminum or stainless steel with larger surface area than the 
sample was placed in the bath to act as the cathode, i.e. negative electrode, where the cut 
aluminum sample is the anode, i.e. positive electrode.  Samples were electropolished for 
5 minutes and the aluminum surface has a mirror like finish.  Mirror like finishes were 
unattainable if perchloric solutions started at temperatures higher than 5 ?C, however 
processes starting below 5 ?C and raised to ~ 10 ?C could still be attained.  
2.2.2 Two-step anodization procedure 
A two-step anodization procedure first reported by Masuda et al was used to 
create ultra-high aspect ratio porous anodic alumina membranes.[79]  Electropolished 
aluminum samples were placed in a bath of oxalic acid (0.3 M in deionized (DI) water) 
within a similar circulator jacket setup as the electropolishing setup.  The oxalic solution 
was cooled to ~8 ?C and constantly stirred in order to encourage even etching of the 
alumina surface.  Using an aluminum sheet as the cathode and the electropolished 
aluminum sample as the anode, samples were anodized for ~7 hrs at 40 V.  After samples 
 
 
34 
 
reached an optimum pore ordering, they were stripped of the anodic oxide formed during 
the 1st anodization in an aqueous mixture of phosphoric acid (6 wt%) and chromic acid 
(1.8 wt%) at 60 ?C.  The stripped PAA surface left an ordered pre-textured aluminum 
surface at which a second anodization of the aluminum could commence.  The second 
anodization took place under the same conditions as the first, producing well ordered 
parallel nanopores that were perpendicular to the aluminum substrate underneath.  The 
pore depth was a function of the 2nd anodization timing using a pore growth rate of ~73 
nm/min.  After the desired pore depth was achieved, pores were widened in a 0.1 M 
aqueous solution of phosphoric acid at 38 ?C.  The pore widening etch rate of the PAA 
under these conditions was ~0.79 nm/min.  The amount a pore can be widened is limited 
by the interpore spacing of the PAA membrane, roughly 105 nm for a template made in 
oxalic acid.  Therefore, pores with an initial diameter of ~35 nm can only be widened to a 
maximum of ~90 nm. 
2.3 Spectroscopic Ellipsometry 
2.3.1 Operational Principles 
Film thickness is one of the most important parameters for the application of thin 
films in devices.[80]  Due to ALD?s self-limiting properties film thickness can be 
precisely controlled by the number of ALD cycles and the growth rate per cycle can be 
determined from the change in film thickness with the number of cycles.  The film 
thickness as a function of ALD cycles is determined by ex-situ spectroscopic 
ellipsometry.  Spectroscopic ellipsometry provides a fast, non-destructive and accurate 
 
 
35 
 
method to measure film thicknesses and optical properties ranging from sub-nanometers 
to a few microns.  In addition it can also be used to characterize composition, 
crystallinity, roughness, doping concentration, and other material properties associated 
with a change in optical response. 
Ellipsometry works by measuring the change in linear polarization as light 
reflects from a surface and basically consists of a light source and a detector unit.[80-82]  
A broad spectral range light source, usually ranging from near infra-red to ultra-violet, is 
passed through a set of polarizers and retarders, thus defining the polarization of the 
incoming light on a surface.  In turn, the polarization of the relected light is determined 
by the combination of another polarizer and a detector.  Ellipsometry is a specular optical 
technique, meaning the angle of incidence is equal to the angle of reflection and span the 
plane of incidence in a material.  Consequently, the linear polarized light incident upon 
the sample can be decomposed into two components: an s component which oscillates 
perpendicular and a p component which oscillates parallel to the plane of incidence.  
Ellipsometry basically measures how p- and s- components of a wave change in relation 
to one another upon reflection on the surface.  Accordingly a known polarization is 
reflected onto the sample and the ouput polarization is measured.  This change in 
polarization can be written as the complex ellipsometric parameter ?: 
??== i
s
p e
R
R )tan(?      Eq. 1 
where, Rp and Rs , are the complex Fresnel reflection coefficients determined by the 
magnitude of the s and p components before and after reflection, and correspond to the 
 
 
36 
 
product of the amplitude ratio, tan(?), and the phase shift, ?, induced by the reflection.  
? is denotes the phase difference between p- and s- wave components respectively, 
before and after reflection, where: 
? = ?1 ? ?2     Eq. 2 
Figure 15 demonstrates the relationship between the incoming linearly polarized light and 
the outgoing elliptically polarized light.   
 
Figure 15.  Ellipsometry measures the change of polarization of a linear 
polarized light when reflected on a surface.  Upon interference with the film, the 
reflected light is elliptically polarized.  Film properties, thickness, and refractive 
index are determined by the measure in phase change, ?, and amplitude change, 
?, of the reflected light.[83] 
 
 
 
 
 
37 
 
From the data obtained, the film thickness and the complex refractive index, ?, 
over the photon energy range can be determined.[80-82]  The complex refractive index is 
expressed as 
? = n + ik     Eq. 3 
where n is the index of refraction and k is the extinction coefficient, but they are often 
represented in terms of the real, ?1, and imaginary, ?2, parts of the complex dielectric 
function ? where 
21 ??? +=      Eq. 4 
22
1 kn ?=?      Eq. 5 
nk22 =?      Eq. 6 
However, because film properties are not measured directly (only ? and ? are) but 
instead have to be extracted from models, such as Drude and Cauchey.  The models are 
setup to describe the film structures on the sample and then fit the data using a regression 
analysis.  Therefore, determining a films index of refraction and subsequent thickness can 
be difficult to determine due to the fact that the index of refraction will change with film 
thickness for very thin films and as the number of unknown properties increases, the 
more convoluted the data becomes.[82]  However, a variable angle spectroscopic 
ellipsometer such as the Sopra GES5 allows for the interrogation of complex structures 
where each wavelength and angle will provide a new set of (n, k) values and 
corresponding equations. 
 
 
38 
 
2.3.2 Sopra GES5 ellipsometer 
A Sopra GES5 spectroscopic ellipsometer was used for rapid film 
characterization.  The GES5 uses a 75W high intensity Xenon arc lamp in order to obtain 
a continuous spectrum of light from 1.2 to 6.5 eV (190-1030 nm).  The use of the ultra-
violet range is particularly useful to study high band gap materials, such as TiO2.  Figure 
16 displays the configuration of a rotating polarizer and a fixed analyzer.  At location A, 
the linearly polarized light is reflected upon the surface where it is elliptically polarized 
during location B.  As the polarizer rotates, the ellipticity of the beam is varied with time.  
The light is again linearly polarized when passed through the fixed analyzer at location C, 
however this time its amplitude varies with the relative position of the fixed analyzer.  
The GES5 uses a charge coupled device (CCD) detector adapted for a grating 
spectrograph.  A goniometer is used to vary the angle of incidence ? from 7? to 90? along 
with a 200 mm automated XY stage for wafer mapping.  With the use of microspot optics 
the beam spot size can be reduced from its nominal 1-5 nm dimensions down to 365 x 
270 ?m.  
 
Figure 16.  Sopra GES5 spectroscopic ellipsometer configuration. 
 
 
39 
 
2.4 Scanning Electron Microscopy 
2.4.1 Operational principles 
With features residing in the nanoscale, a strong need exists to image such small 
features.  Optical microscopes are limited by their lack of resolution on such small 
features therefore, electron microscopes are used to reveal information about a samples 
morphology, chemical composition, and crystalline structure and orientation of materials 
making up a sample.[84]  An SEM operates by using a focused beam of high-energy 
electrons to generate a variety of signals that can be used to image the surface of a 
sample.  The most common thermionic electron gun uses a tungsten filament cathode 
because it has a high melting point and lowest vapor pressure of all the metals, thereby 
allowing it to be heated high enough to produce the emission of electrons.  Other types of 
electron emitters are lanthanum hexaboride (LaB6) cathodes, Schottky emitters of 
zirconium oxide, or a tungsten field-emission tip.[84, 85]  Through the interaction of the 
electrons and atoms at or near the surface of the sample secondary electrons, back-
scattered electrons (BSE), characteristic x-rays, light (cathodoluminescence), and 
transmitted electrons can be produced in order to reveal details about samples 1 to 5 nm 
in size.[84, 85]   
In order to produce an image, the electron beam with an energy ranging from 0.5 
keV to 40 keV, is focused by one or two condenser lenses to a spot of about 0.4 nm to 5 
nm in diameter.[86]  The beam then passes through a pair of scanning coils which 
deflects the beam on an x and y axes so that it scans in a raster fashion over a rectangular 
 
 
40 
 
area of the sample surface.   When the beam interacts with the sample, the incoming 
electrons lose energy by repeated scattering and absorption within a teardrop-shaped 
volume of the sample called the interaction volume, which can extend from 100 nm to 
around 5 ?m into the surface depending on the atomic number of the sample, the 
sample?s density, and the accelerating voltage of the beam.[84, 85]  The energy exchange 
between the electron beam and the sample results in the reflection of high-energy 
electrons by elastic scattering, emission of secondary electrons by inelastic scattering and 
the emission of x-rays, each of which can be detected by their own specialized 
detector.[84]  Signals emitted from the sample are collected by their respective detectors 
and amplified, where the intensity is then modulated on a TV monitor and rastered in 
synchronization with the electron beam in order to produce an image.[84]  Figure 17 
displays a schematic of an SEM operational setup.  In order to increase the magnification 
of the image, the raster size on the sample is decreased.[84]   
 
 
41 
 
 
Figure 17.  Schematic of an SEM setup.[87] 
2.4.2 Hitachi SU-70 SEM 
A Hitachi SU-70 SEM with a ZrO/W Schottky field emission gun was used to 
image top down and cross-sectioned samples of AAO in order to characterize AAO pore 
sizes and ALD film thicknesses.  For high resolution images an accelerating voltage of 10 
kV was used with the probe current set on high in medium mode.  Using a working 
distance of 5 mm, secondary electrons were collected using a mixture of upper and lower 
detectors.   
 
 
42 
 
2.4.3 Cross-sectional sample preparation 
In order to fabricate porous AAO samples for cross-sectioning, samples were epoxied 
to a small piece of silicon with a notch scribed in the back center pore side down.  The 
backside of the AAO template was scratched to induce etching of the aluminum substrate 
in a Cu2Cl solution.  Templates were etched with help of a sonicator for approximately 20 
min followed by sonicating the sample in isopropanol for an additional 10 min removing 
any Cu particles left on the surface.  Samples were then cleaved using the notch in the 
back side of the silicon as an initiator.  See Figure 18 for a schematic of the sample prep 
process flow.  Before imaging samples were coated with a thin film of sputtered Au/Pd in 
order to reduce sample charging during measurements.  Au/Pd films were sputtered using 
a small table top sputter in Dr. Sang Bok Lee?s lab.  The chamber was evacuated to ~60 
mTorr and samples were coated using an argon generated plasma for 15 sec. 
 
Figure 18.  Schematic of SEM sample preparation for AAO cross-sections. 
 
 
43 
 
2.5 Transmission Electron Microscopy 
2.5.1 Operational principles 
Transmission electron microscopy (TEM) is a powerful tool for investigating the 
microstructure of samples on the nanoscale.  TEM uses a high energy beam of electrons 
to transmit electrons through a sufficiently thin sample.  The maximum thickness for an 
electron to transmit through depends upon the atomic number of the material, where this 
range typically lies between 250-500 nm.[86]  However, higher electron energies help 
transmit more electrons through the sample.   
A TEM is a complex tool composed of an illumination system, specimen stage, 
and an imaging system all within an evacuated column.  The illumination system is 
comprised of an electron source usually using a tungsten filament of a LaB6 crystal, 
together with an assembly of condenser lenses located above the sample.  The specimen 
stage can either be held stationary or moved with the use of a double tilted stage.  The 
imaging system is made up of an objective lens, intermediate lenses, a projector lens, and 
a phosphor screen all positioned beneath the sample as seen in Figure 19.  After electrons 
are emitted from their source they are accelerated to their final kinetic energy by applying 
a potential difference to a cathode and an anode, which is a metal plate containing a 
central hole in which the beam is accelerated through.  The beam is then sent through a 
strong magnetic lens called the first condenser lens.  This lens forms a demagnified 
image of about 1 ?m in diameter and is then projected on to the specimen by a second 
condenser lens with a magnification typically around 2.[86]  The final illumination spot 
 
 
44 
 
on the specimen can be a small as 2 ?m, which should be sufficient to fill the viewing 
screen.[85, 86]   
 
Figure 19.  Schematic of the path an electron beam takes within the evacuated 
column of a TEM.[88] 
 
All TEM specimens are made a standard 3 mm diameter circular disk, where the 
specimen must be thin enough to allow electrons to be transmitted.  The specimen stage 
is meant to hold samples as stationary as possible, as any vibrational causing drifts would 
be magnified in the final image.     
Electrons that have been transmitted through the sample are then focused through an 
objective lens with a small focal length.  Therefore, the objective lens produces a 
magnified real image of the specimen at ~10 cm below the center of the lens.  The beam 
then passes through an objective aperture, which is comparable to a diaphragm that helps 
produce contrast in the image and limits the amount of blurring that arises from spherical 
 
 
45 
 
and chromatic aberrations.  The beam then travels through a selected area aperture, which 
is a diaphragm that is inserted in the plane that contains the first magnified real image of 
the specimen, i.e. the image plane of the objective lens.  This diaphragm is used to record 
the electron diffraction pattern associated with the sample.  Most TEMs make use of 
several intermediate lenses between the objective lens and the projector lens.  The lenses 
are used to first, change the beams focal length into smaller step allowing the overall 
magnification of the TEM to be varied over a large range, ~ 103 to 106 and second, by 
producing an electron diffraction pattern on the TEM viewing screen.  A projector lens is 
used to produce an image or a diffraction pattern across the entire TEM screen.  In order 
to determine the magnification of the image it is an algebraic product of the 
magnification factors associated with each of the imaging lenses used.  A phosphor 
screen is used to display the electron image as a visible form.[85] 
2.5.2 JEOL 2100F field emission TEM 
A JEOL 2100F microscope with a LaB6 thermionic emission electron source 
operated at 200 keV and a post-column Gatan CCD camera were used to image TiO2 
ALD based nanotubes held on a double tilted specimen holder.  Much of the focusing and 
imaging was done by Dr. Larry Lai of the NispLab staff. 
2.5.3 Nanotube sample preparation 
TiO2 ALD based nanotubes were fabricated by the dissolution of the porous AAO 
membrane the ALD films were deposited in.  Samples were placed in a 25 wt% 
phosphoric acid solution for 24 hrs, which dissolved the alumina membrane thus 
 
 
46 
 
releasing the TiO2 nanotubes in solution.  Nanotubes were filtered using DI water and a 
drop of solution with dispersed nanotubes was placed on a standard holy-carbon TEM 
grid and left to dry for several hours. 
2.6 Energy Dispersive X-ray Spectroscopy 
2.6.1 Operational principles 
Scanning electron microscopy (SEM) and scanning transmission electron 
microscopy (STEM) in conjunction with energy dispersive x-ray spectroscopy (EDS), is 
a powerful tool for obtaining chemical information from a wide variety of materials 
found in nanotechnology.  EDS is an analytical technique which utilizes x-rays that are 
emitted from the specimen when bombarded by the electron beam of an electron 
microscope in order to identify the elemental composition of the sample.[89]  When a 
high energy electron bombards an atom in the sample, an electron is ejected from an 
inner shell leaving a vacancy in its place, see Figure 20.  This vacancy is then filled by an 
electron from a higher energy state and releases energy in the form of an x-ray.  This 
emitted x-ray is characteristic of the element from which it was emitted.  Therefore, 
elements with high atomic numbers will produce more x-rays.  This type of x-ray is 
called a characteristic x-ray.[89]  In addition to characteristic x-rays, there are also 
background x-rays called ?Bremsstrahlung? radiation.  Bremsstrahlung radiation is 
produced by the deceleration of an electron as it passes through the sample.  
Bremsstrahlung x-rays are also dependent upon the atomic number of the material being 
analyzed, but are not a good source for quantitative analysis and are in large part 
 
 
47 
 
responsible for background noise during analysis.[85, 89] 
 
Figure 20.  Characteristic x-ray production. 
When an x-ray enters the detector it passes through a beryllium window, where 
they are collected by a solid state detector comprised of a single crystal of silicon with a 
layer of lithium diffused within.  When x-rays enter the detector they create electron-hole 
pairs, where the number of electron-hole pairs are equivalent to the energy of the x-
ray.[85, 86, 89]  The detector is cooled using liquid nitrogen in order to reduce 
containments, as well as limit the production of electron-hole pairs through thermal 
generation.[85]  The charge produced from the electron-hole pair is converted to a 
voltage by a field effect transistor (FET) preamplifier and then sent to a multichannel 
analyzer where the voltages are sorted.[86]  The energy for each incident x-ray, as 
determined from the voltage measurement, is sent to a computer display for further user 
analysis.[86] 
 
 
48 
 
2.6.2 Bruker EDS detector 
A Bruker EDS detector was used in combination with a Hitachi SU-70 SEM in 
order to obtain quantitative chemical analysis of thin films.  Because SEM samples   
tend to be thicker, it is easier to focus the 
electron beam within the sample allowing 
for complete absorption of the beam in 
the sample.  This causes the beam to 
spread within the sample and leads to a 
higher generation of x-rays as seen in 
Figure 21. 
2.6.3 Oxford EDS detector 
An oxford EDS detector was used 
in conjunction with a JEOL 2100F Field 
Emission TEM.  TEM requires the use of 
very thin samples in order to transmit electrons through the sample; therefore, few x-rays 
are produced than those found in SEM, as seen in Figure 21.[90] 
2.7 Atomic Force Microscopy 
2.7.1 Operation principles 
Over the past 20 years, the quest for miniaturization has led to an increased 
interest in characterizing the unique quantum and surface phenomena that matter exhibits 
at the nanoscale.  Atomic force microscopy (AFM) is a type of scanning probe 
 
Figure 21. Schematic of the diminished 
electron beam interaction within a TEM 
sample versus an equivalent SEM 
sample.[90] 
 
 
49 
 
microscopy that provides very high resolution down to the nanometer scale to provide 3-
dimensional images of a specimen?s surface topography.  AFM operates by bringing an 
extremely sharp tip in contact or in very close proximity to a sample in order to scan the 
sample?s surface.[86]  The tip is usually a couple of microns long with a radius of 
curvature of less than 100 ? in diameter and is located at the free end of a silicon or 
silicon nitride cantilever, 100 to 200 ?m long (Figure 22).[91]  The tip is brought into 
close proximity of the sample?s surface, where the forces between the tip and the sample 
lead to a deflection of the cantilever.  Typically deflections are measured by a laser 
reflected from the top surface of the cantilever and into an array of photodiodes where the 
signal is fed into an electronic interface and computer in order to control the feedback 
keeping the force constant while scanning the sample surface (Figure 23).[86]  Other 
techniques commonly used are optical interferometry, capacitive sensing, or the use of 
piezoresistive cantilevers.[91]  Depending on the sample setup many physical properties 
can be determined through the types of forces measured, such as mechanical contact 
forces, van der Waals forces, capillary forces, electrostatic forces, magnetic forces, 
chemical bonding, etc.[86, 91, 92]  In comparison to other microscopy techniques, AFM 
provides comparable or better topographical information and unobstructed views of the 
surface because images are obtained without expensive sample preparation or conductive 
coatings often necessary in SEM and TEM imaging. 
 
 
50 
 
 
 
 
Figure 22.  SEM image of the 
AFM Si tip at the free end of a Si 
cantilever.[92] 
 Figure 23.  Schematic of an AFM 
operational setup.[92] 
 
AFM probes can operate in two modes: constant force mode and tapping mode.  
In constant force mode the force between the tip and the sample is kept constant by 
forcing the tip to remain in contact with the sample.  In this case samples can either be 
rastered underneath the tip while the tip remains stationary or the probe can be scanned 
over the sample.  However, in either mode the sample can become damaged causing a 
loss in resolution as the sample is continually rastered.  Therefore, to reduce this problem, 
AFM systems have been modified to operate in a tapping mode.   In tapping mode, the 
cantilever is lightly tapped against the sample, reducing lateral shear forces and reducing 
damage to the sample.  The cantilever is oscillated up and down near its resonance 
frequency by a small piezoelectric element mounted in the AFM tip holder.  This mode is 
particularly useful for mapping polymers and other biological samples.[86, 91]   
 
 
51 
 
2.7.2 MFP-3D AFM 
An MFP-3D stand alone atomic force microscope was used for easy 
topographical mapping across a sample.  The MFP-3D used a silicon tip in tapping mode 
at a scan rate of 1 Hz and a driving frequency of 200 kHz.  Samples were mounted on 
glass slides where 1 ?m scans were made across sample surfaces.  The AFM stage was 
encased in a vibration damping box on top of a levitated optical table in order to 
eliminate environmental vibrations that may have influenced surface measurements. 
 
 
 
52 
 
Chapter 3 :  Profile Evolution for Conformal ALD Over 
Roughened Nanotopography 
The self-limiting reactions which distinguish atomic layer deposition (ALD) 
provide ultrathin film deposition with superb conformality over the most challenging 
topography.  This chapter addresses how the shapes (i.e. surface profiles) of 
nanostructures are modified by the conformality of ALD.  As a nanostructure template, 
we employ a highly scalloped surface formed during the 1st anodization of the porous 
anodic alumina (PAA) process, followed by removal of the alumina to expose a scalloped 
Al surface.  SEM and AFM reveal evolution of surface profiles that change with ALD 
layer thickness, influenced by the way ALD conformality decorates the underlying 
topography.  The evolution of surface profiles is modeled using a simple geometric 3D 
extrusion model, which replicates the measured complex surface topography.  Excellent 
agreement is obtained between experimental data and the results from this model, 
suggesting that for this ALD system conformality is very high even on highly structured, 
sharp features of the initial template surface. Through modeling and experimentation the 
benefits of ALD to manipulate complex surface topographies are recognized and will 
play an important role in the design and nanofabrication of next generation devices with 
increasingly high aspect ratios as well as nano-scale features. 
 
 
53 
 
3.1 Introduction 
Paralleling the trend in electronic devices, high surface area 3D architectures for 
applications in energy [21, 93], catalysis [18, 61], and sensors [30, 94] are exploiting the 
benefits found in using nano-scaled structures with increasingly high aspect ratios.  In 
exploiting these nanostructures, the 3D topography is perhaps even more complex and 
demanding than in conventional applications found in ULSI based applications such as 
those demonstrated in trench and stacked capacitors in DRAM devices [9], barrier 
technology for copper metallization[95], and gate dielectric deposition on planar Si [8] 
and Fin-FET geometries [13].  For example, Banerjee et al recently demonstrated a novel 
metal-insulator-metal (MIM) nanocapacitor using a combination of self-limiting 
processes to build multilayered functional films and self-assembled nanotemplates with 
nanopores ~50 nm wide x 10 ?m deep (aspect ratio= 200) [21].  In these structures, a 
primary performance limitation was set by the geometry of the nanoporous template 
consisting of sharp asperities, leading to enhanced local electric fields and low field 
breakdown which would ultimately limit energy storage density [21].  This example 
emphasizes the need for understanding and optimizing our ability to synthesize complex 
3D nanostructures.  Complex surface topographies consisting of high aspect ratios, sharp 
asperities, and nano-roughness significantly challenge the ability to conformally coat 
complex nanostructures using traditional physical vapor deposition (PVD) and chemical 
vapor deposition (CVD) techniques [66, 94, 96].  However, atomic layer deposition 
(ALD) offers high quality films with atomic level thickness control and unprecedented 
uniformity and conformality in the most demanding 3D nanostructures [1, 2, 67]. 
 
 
54 
 
 In this chapter, we explore the evolution of surface profiles for ALD films 
deposited on a prototype of a topographically rough surface.  We utilize highly scalloped 
Al templates formed by the removal of porous anodic alumina (PAA) membranes.  PAA 
is a self assembled, electrochemical templating technique that results in a highly dense 
close-packed hexagonal array of cylindrical nanopores on the surface of aluminum [97].  
PAA naturally lends itself to many nanotechnology applications, such as a template for 
growing nanowires and nanotubes and for self assembling these structures in parallel, 
periodic arrays used to fabricate catalysts membranes [61, 94], dye sensitized solar cells 
[98], batteries [99] and sensors [30, 94].  When the PAA film is etched off, a self-
ordered, close-packed hexagonal array of nano-dimples is left imprinted in the underlying 
aluminum surface (Figure 24a) [100].  Nearest neighbor dimples meet at sharp points and 
the overall surface topography consists of periodic peaks and valleys (Figure 24b).  Such 
surfaces are ideal for testing the influence of film conformality on surface topography.  
 
Figure 24.  A removed PAA template imaged in SEM: (a) top-down view of the 
hexagonal nano-dimple array and (b) cross-section view of the scalloped Al 
surface, emphasizing the peak structure above the nano-dimples.  
  
 
 
55 
 
ALD is a self-limiting process which utilizes sequential chemisorbed surface 
reactions of precursor molecules to achieve deposition one monolayer at a time and is 
dependent upon molecule flux, adsorption/desorption probability, and surface diffusion 
[3].  Much work has focused on analytically understanding the performance of ALD 
conformality, particularly in high aspect ratio structures where process metrics, i.e., 
reactant exposures and purge times, must be modified to compensate for the higher 
surface areas and molecular flow type dynamics occurring in 3D nanostructures.  This 
includes developing simple models for determining precursor dose for conformal 
coatings [66], using Monte Carlo techniques [67], solving simultaneously Boltzmann?s 
transport and surface reaction equations [101] and correlating them with experimental 
results while trying to optimize conformality in ultra-high aspect ratio structures.  The 
prevailing message in the above reports is that ALD yields unparalleled performance 
when it comes to depositing thin films across complex 3D nanostructures, but requires 
precise process control to ensure complete surface saturation and avoid excess deposition 
characteristic of CVD-like behavior [3].  On the other hand, the surface evolution of ALD 
films across complex surface topographies, especially with sharp asperities and nano-
roughness has not been systematically explored ? experimentally or theoretically.  Such 
complex topographies are still likely to incur complete saturation reactions and avoid 
topography induced nucleation delays that could potentially affect the conformality of the 
films and the resulting topography [3, 5].  Therefore this work is aimed at investigating 
the surface evolution of ALD films in the context of highly scalloped surfaces. 
 Our test-bed utilizing a combination of a removed PAA membrane from an Al 
 
 
56 
 
template and ALD provides us with the ability to monitor surface evolution of ALD films 
as they are deposited on a highly scalloped surface.  We use two different templates with 
different interpore spacings (Dint) to reveal a universal behavior for the growth of ALD 
films on such surfaces following simple scaling laws.  Furthermore, a surface extrusion 
model is used to verify this behavior and the resulting surface area as a function of ALD 
deposition thickness.  This work demonstrates that through modeling and 
experimentation the benefits and limitations of ALD can be realized for complex 
nanostructures. Understanding this behavior will play a crucial role in the design and 
nanofabrication of next generation energy devices with increasingly high aspect ratios as 
well as nano-scale feature sizes. 
3.2 Materials and Methods 
Scalloped aluminum templates were synthesized by the removal of PAA 
membranes formed during a one-step anodization method [79].  High purity 200 ?m 
thick aluminum foils (99.99%, Alfa Aesar), were electropolished in a 1:5 perchloric acid 
to ethanol bath held at ~ 3 ?C in order to remove micro-roughness from the surface, 
which influence the quality of the pores.  The electropolished aluminum was then 
anodized at (1) 40 V and 10 ?C in an electrolytic bath of 0.3M oxalic acid to achieve an 
interpore spacing, Dint, of 110 nm or (2) at 160 V and 4?C in an electrolytic bath of 10 
wt% phosphoric acid to achieve Dint of 450 nm.  Foils were anodized for ~7 hours so 
PAA pores would self organize and align orthogonally with respect to the underlying Al 
substrate.  After anodization, the PAA film produced was removed in an aqueous solution 
of phosphoric acid (6 wt%) and chromic acid (1.8 wt%) at 60 ?C leaving a hexagonally 
 
 
57 
 
organized nano-dimple array with a highly scalloped surface imprinted in the remaining 
aluminum. 
 Titanium dioxide (TiO2) ALD thin films were thermally deposited at 150 ?C in a 
Beneq TFS-500 cross-flow ALD reactor using Tetrakis(dimethylamido)titanium 
(TDMAT) and water.  Pulses and purges of 500 msec for both precursors resulted in 
atomic thickness control with a growth rate of 0.055 nm/cycle as a sub-monolayer of film 
was consistently deposited for each cycle.  Three samples were loaded during the first run 
and deposition was carried out in 3 intervals of a fixed number of cycles, i.e. 244 or 1000 
for templates anodized in oxalic or phosphoric acid, respectively.  After each interval one 
sample was removed from the reactor and characterized via high resolution scanning 
electron microscopy (SEM, Hitachi SU-70) and atomic force microscopy (AFM, MFP-
3D Asylum Research) in tapping mode.  Film growth rates were determined using a 
spectroscopic ellipsometer (Sopra GES5) on planar Si samples. 
3.3 Results & Discussion 
The evolution of ALD surface profiles over complex nano-topography was 
investigated with particular focus on how ALD modifies surface topography.  PAA 
membranes formed in phosphoric acid with a Dint of 450 nm were removed from the 
underlying aluminum surface revealing a highly textured surface of peaks and valleys 
that rise above an array of hexagonal dimples as seen via cross-sectional SEM (Figure 
24a-b).  Peaks are seen positioned along the [121] direction of the template, whereas 
valleys are found to traverse the [111] direction. Peak structures are multifaceted in shape 
due to the convergence of multiple hexagonal dimples.  Peaks that rise notably above the 
 
 
58 
 
average peak height and non-hexagonal dimples are considered defects.  Defects are 
preferentially found along domain boundaries and are generally created during the 
anodization process when starting from a completely disordered structure, as opposed to 
a stamped template where the ordering is considered perfect [102].  
 Templates formed in phosphoric acid were coated with three consecutive layers of 
1000 cycle TiO2 ALD.  Top-down SEM images display surface profiles of TiO2 ALD 
deposited on the scalloped Al surface after 0 cycles, 1000 cycles, 2000 cycles, and 3000 
cycles (Figure 25a-d).  Images reveal that increased ALD deposition results in the 
formation of triangular patterns originating from the peak structures across the surface.  
In Figure 25b, it appears that ALD deposition is nucleating at these sharp tips, thus 
dominating ALD growth across the surface and creating areas of higher growth over 
others as seen in Figure 25c.  As deposition continues, growth fronts originating from 
these peaks start to merge and exhibit symmetric, repeating patterns of raised triangles 
decorating the surface with ?nano-petunias? as seen in Figure 25d. 
 
 
59 
 
 
Figure 25.  Progression of TiO2 ALD profiles across a scalloped Al surface 
formed using phosphoric anodizing acid imaged in SEM after (a) 0 cycles, (b) 
1000 cycles (~55 nm), (c) 2000 cycles (~110 nm), and (d) 3000 cycles (~165 nm).  
(e) SEM image of a scalloped Al surface using oxalic anodizing acid coated with 
733 cycles (~40 nm) of TiO2.  Top-down images (d and e) display the same 
formation of triangular ALD surface profiles for the same ratio of equivalent 
thickness (t) to interpore spacing distance (Dint) for each anodizing acid. 
 
A simple scaling law was devised to adequately compare the ALD surface 
profiles deposited on templates formed with different interpore spacing using either 
phosphoric or oxalic acid.  Using the film?s growth rate, deposition on one template could 
be scaled by a factor of k defined in equation (1) as 
 
intD
tk =       (1) 
where t is the measured film thickness for a set number of cycles deposited on planar 
surfaces, in order to calculate an equivalent film thickness (?) for a template with a 
 
 
60 
 
different interpore spacing ( oDint ).  Leaving ? = oDk int* .  SEM images display templates 
produced with different interpore spacings of 450 nm for phosphoric acid (Figure 25d) 
and 110 nm for oxalic acid (Figure 25e) coated with equivalent film thicknesses of 165 
nm for 3000 cycles and 40 nm for 733 cycles of TiO2 ALD, respectively.  Both samples 
display the formation of triangular patterns across the surface, revealing a universal 
behavior for templates with similar scalloped surfaces but different dimensions. 
TiO2 films were deposited for 733 cycles to a film thickness of ~40 nm on 
scalloped Al templates formed in oxalic acid.  A tilted top-down SEM image of a thin 
delaminated film of TiO2 retaining the original hexagonal pattern from the initial 
scalloped Al surface is shown in Figure 26a.  However, the film?s surface topography no 
longer displays the sharp peak structures that extruded from the initial, now-covered 
surface.  A tilted-bottom up view of the delaminated film suggests the emergence of 
growth fronts for the evolving ALD film, where they eventually merged as indicated by 
arrows in Figure 26b.  As a TiO2 film is conformally deposited over the original 
scalloped template, the ALD film topography changes shape with deposition and displays 
substantial smoothing of the original scalloped surface. This behavior underscores the 
benefits of conformal ALD to blunt or smooth sharp asperities and surface features with 
increasing film thickness.  This is particularly important in some applications, e.g., for 
local enhancement of electric fields at asperities and consequent low field breakdown of 
an MIM nanocapacitor [21]. 
 
 
61 
 
 
Figure 26.  A delaminated conformal TiO2 ALD film imaged in SEM: (a) tilted 
top-down view of merged growth fronts influenced by the hexagonal pattern of 
the removed underlying surface and (b) tilted bottom-up view of the film where 
dashed arrows in the cross-section indicate where growth fronts merged and 
laterally moved due to film broadening at the peaks (solid black arrows) and 
causing the disappearance of the nano-dimples (solid white arrows). 
 
We attribute the display of growth fronts to the direct consequences of ALD 
conformality over surface topography that includes rounded as well as sloped regions.  
While other physical or chemical deposition mechanisms convey some measure of 
conformality over topography, ALD?s self-limiting chemistry provides unprecedented 
capacity to maintain conformality even in the most challenging nanotopography.  
Sidewall deposition, i.e. on the inclined surfaces, is as rapid as on the rounded surfaces.  
This causes the rise of directional growth fronts where there is a significant change in 
incline on the surface (i.e. shift from the sloped side walls to that of the concave valley or 
dimple).  Growth fronts merge at crossing points between the inclined and curved regions 
and move laterally as deposition proceeds as indicated by the dashed arrows found in 
Figure 26b.  When the initial surface is comprised of scalloped peaks and valleys that rise 
above lower dimples, the successive ALD layers exhibit changes in the surface profile, 
 
 
62 
 
leading to the lateral growth and rounding of the peaks and reduction (or ultimately 
disappearance) of the valleys and lower dimple areas, as seen in Figure 26b.    
This behavior is evident from AFM maps for ALD TiO2 film profiles seen in 
Figure 27.  TiO2 films deposited on templates formed in phosphoric acid for 0 cycles, 
1000 cycles (55 nm), 2000 cycles (110 nm), and 3000 cycles (165 nm) were mapped 
using AFM in order to monitor the progression of surface profiles and the decrease in 
surface area with increasing ALD film thickness.  Figure 27a displays an AFM contour 
map for 2000 ALD cycles (110 nm), with the position of underlying peaks (A, B, C, and 
D) of the initial scalloped surface denoted along with the [121] and [111] directions.  The 
map illustrates the emergence of triangular surface profiles as areas of higher topography 
allow growth fronts to extend out and finally merge along the [111] direction.  AFM line 
scans taken of increased deposition across the [121] direction were layered and separated 
by the experimental growth rate for TiO2, where 0, 1, 2, and 3 indicate the number of 
intervals for 1000 cycles, with 0 being the bare template (Figure 27b).  
 
 
 
 
63 
 
 
Figure 27.  (a) AFM contour map of a 2000 cycle (~ 110 nm) TiO2 ALD film on a 
scalloped Al surface dictating the [111] and [121] directions with peaks A, B, C, 
and D (denoted by x) across the template.  (b) AFM line scans across the 121 
direction after 0 cycles, 1000 (~ 55 nm), 2000 cycles (~110 nm), and 3000 cycles 
(~165 nm) of deposited TiO2 ALD (arrows denote locations of peaks). The 
profiles are numbered 0-3 for intervals of 1000 cycles each. 
 
These results illustrate the broadening of peaks in the original scalloped surface 
initiated at positions A, B, C, and D.  As neighboring peaks positioned at A and B as well 
as C and D expand, they first cause a sharpening of the valley between them, and then 
begin to close and diminish the valley with further ALD coverage.  At the location of the 
original dimples in the scalloped surface, e.g. between B and C, much less of this effect is 
seen due to the lower plateau centered at the dimple being considerably wider to begin 
with.  Ultimately, this behavior lends itself to the formation of the triangular pattern or 
?nano-petunias? across the surface.  Although surface features of the samples were 
smaller than the AFM tips used, an increase in deposition indicates a decrease in surface 
area as growth fronts merge and the complex peak structure is rounded.  ALD?s ability to 
 
 
64 
 
conformally coat and thereby modify a topographically rough surface will prove to be of 
great importance when engineering detailed geometries sometimes found in 3D 
nanostructures.  For example, ALD growth kinetics in high aspect ratio structures become 
primarily mass-transport limited via molecule diffusion in and out of the pores rather than 
surface reaction-limited [66].  Therefore, as the ALD deposition profile evolves, the 
entrance to a pore potentially found between B and C in Figure 27 will start to close off, 
creating an ever decreasing orifice diameter for molecules to diffuse through with each 
cycle, and increasingly challenging ALD?s ability to conformally deposit within a high 
aspect ratio nanopore. 
3.4 ALD-PAA Simulation 
A simple geometric 3D extrusion model was developed by Dr. Israel Perez, a 
former graduate student, to demonstrate the influence of a complex surface topography 
and track the progression of ALD surface profiles.  The model replicated the topography 
found on a scalloped Al template after removal of the PAA membrane by using the two 
key dimensions: interpore spacing, Dint, and the scallop?s equivalent sphere radius, rs.  
The interpore spacing of the scalloped array is a well known dimension and is related to 
the anodization potential used for a specific anodization acid [103].  For the purpose of 
this simulation Dint is taken as 105 nm corresponding to oxalic acid templates.  The rs 
value is the radius of curvature of the nano-dimple imprinted in the Al surface and was 
extrapolated from AFM data of a bare scalloped Al surface.  The maximum peak height 
was measured as the difference in the z-height between the bottom of the dimple and the 
top of the peak, which for a PAA surface anodized in oxalic acid at 40 V is 
 
 
65 
 
approximately 42 nm.  Knowing this value allows the surface to be modeled by adjusting 
rs, until the maximum ideal peak height is ~42 nm, which gives a value of rs ~65 nm.  
Therefore, an ideal scalloped surface was coded in MatLab? by spacing hemispheres 
with radius rs in a hexagonal array, with inter-sphere spacing of Dint in the closed pack 
direction, and removing all intersecting data.   
As mentioned earlier, much work already exists in the literature covering the strict 
theoretical modeling of ALD in terms of thickness as a function of cycles [104], 
nucleation and growth kinetics [53], and Monte Carlo simulations in ultra-high aspect 
ratio nanopores [67].  For the purpose of this simulation, ?ideal? ALD was assumed, 
where one monolayer of film is conformally deposited per cycle and ignores nucleation 
delays potentially present during the initial cycles of TiO2 ALD on a removed PAA 
surface [105].  Instead it used experimentally determined growth rates and assumed that 
for each cycle the entire surface will be conformally coated with a specific amount of 
material given by the growth rate for the specified ALD process.  The 2D cross-section 
schematic in Figure 28 depicts the step-by-step simulation sequence for a single ALD 
cycle.  Starting with five equally spaced points on the initial surface (Figure 28a), surface 
normals were determined (Figure 28b).  New data points were placed along each surface 
normal at a distance specified by the growth rate of the ALD process being simulated 
(Figure 27b).  The new data points were no longer equally spaced along the x-axis; 
therefore, they were best fit with a new line (Figure 28c).  A new set of equally spaced 
data points were created and placed along the best fit line, ultimately, creating the new 
surface (Figure 28d). 
 
 
66 
 
 
Figure 28.  Geometric 3D extrusion model: (a) five equally spaced data points on 
the surface of a sharp peak. (b) surface normals are projected out and the data 
points for the new surface are placed along them using the prescribed growth 
rate; (c) the new data is fit with a ?best fit? line; (d) the ?best fit? line is used to 
generate more data which is equally spaced in the x-direction. 
 
The model described above used an initial surface prepared using oxalic acid and 
therefore, cannot be directly compared with experimental results on a template prepared 
using phosphoric acid at 160 V.  However, a similar scaling factor, k, used to compare 
film thicknesses from different templates can be used to compare the experimental results 
with the simulation, where Dint is 450 nm and == simo DDint 105 nm.  Top view SEM 
images (Figure 29a-d) and top view simulation results (Figure 29e-h) were compared 
illustrating the surface evolution of ALD on a scalloped Al surface mirrors what is seen 
in the extruded surface simulation. 
 
 
67 
 
 
Figure 29.  (a-d) Scalloped Al surface imaged in SEM after TiO2 ALD deposition 
of 0 cycles, 1000 cycles (~55nm), 2000 cycles (~110nm), 3000 cycles (~165nm); (e-
h) simulation of surface evolution by extruding a model of the initial surface at 
same rate as the deposition per cycle for ALD TiO2. 
    
Cross-sectional view SEM images along the [111] direction of a scalloped Al 
surface show three discrete 1000 cycle TiO2 ALD films deposited one on top of another 
(Figure 30a).  Each film set displays a distinct interface where the surface was exposed to 
air for ex-situ characterization and indicates the presence of surface contamination.  As 
deposition continued, films took on the form of the underlying film?s surface eventually 
creating a rounded knoll atop the initial peak in the Al template?s surface signifying there 
was no substantial topography-induced nucleation.  Top down and cross-sectional 
profiles along the [111] and [121] directions of the 3D model (Figure 30b) reemphasize 
that ALD deposits in a conformal manner, and that the formation of triangular islands 
seen in both SEM images and the 3D model is an artifact of the underlying surface and 
not due to nucleation and growth resonating from the tips of the sharp peaks. 
 
 
68 
 
 
Figure 30.  Cross-sectional surface profiles as a function of film thickness: (a) 
SEM image of three discontinuous 1000 cycle TiO2 ALD films along the 111 
direction of a scalloped Al surface and (b) modeled profiles of TiO2 ALD films 
along the 111 and 121 directions of a scalloped Al surface. 
 
Surface area plays an integral part in the design and nanofabrication of next 
generation energy devices such as MIM nanocapacitors.  Matlab? simulations developed 
by Dr. Perez were used to estimate the change in surface area as a function of film 
 
 
69 
 
thickness.  The surface area was estimated by summing the areas of all the quadrilaterals 
that comprise the modeled surface as defined by the X, Y, Z coordinate data.  The planar 
equivalent area is the 2D surface equivalence to the area projected onto the XY plane 
from the 3D model.  From these estimations the initial surface of a template formed in 
oxalic acid was estimated to have a 30% larger area than its planar equivalent.  Templates 
formed in phosphoric acid would have smaller surface areas due to larger interpore 
spacings, meaning there are fewer peaks for the same planar area used above.  Starting 
with the initial surface of the scalloped Al template an ALD simulation was performed 
for 1000 cycles with a growth rate of 0.1 nm/cycle.  The normalized surface area (the 
estimated surface area divided by the planar equivalent surface area) as a function of 
ALD film thickness was plotted with corresponding 3D representations of the evolving 
surface profile (Figure 31).   
 
Figure 31.  Normalized surface area as a function of ALD film thickness. 
 
 
70 
 
As the deposition proceeds the normalized surface does not change appreciably until after 
~45 nm.  After 45 nm of film growth, the normalized surface area asymptotically decays 
towards a limiting value of 1 implying that the surface has been effectively ?planarized?.  
The shoulder at 45 nm seems to coincide with the appearance of lines in the film where 
film growth fronts have started to merge along the [111] direction of the hexagonal 
scalloped cell, see Figure 31.  This is the moment at which the curvature of the saddle 
points between the peaks in the scalloped surface is completely inverted by the continued 
extrusion of the ALD deposition over the sharp peaks.  Further deposition results in the 
extension of these lines towards the center of the scallops further unifying the surface. 
3.5 Conclusion 
TiO2 ALD films were deposited on a highly scalloped Al surface to demonstrate 
the ability of ALD to provide exceedingly conformal films over a complex topography of 
a template?s initial surface.  SEM images and AFM scans reveal the progression of 
triangular surface profiles that change with ALD layer thickness, influenced by the way 
ALD conformally coats the underlying topography, and further emphasizing that ALD 
nucleation is rather insensitive to the surface topography at least for the experimental 
process chemistry investigated here.  
We demonstrated the usefulness of a simple geometric 3D extrusion model to 
simulate the evolution of ALD films? surface profiles as they conformally coat the 
complex surface topography of a modeled scalloped Al surface.  The scalloped Al surface 
was modeled and calibrated using dimensions extracted from SEM and AFM scans.  The 
surface was then extruded in a stepwise manner using experimentally obtained growth 
 
 
71 
 
rates of the particular ALD chemistry being simulated.  The model illustrates how 
conformal ALD films cause the surface topography to change with each deposited layer, 
ultimately leading to a decrease in surface area as a function of film thickness.  Results 
from this model and experimental data are in excellent agreement, leading to the 
conclusion that no nucleation related effects were observed and the resulting triangular 
pattern is an artifact of the underlying textured Al surface.   
This work illustrates the benefits of ALD to conformally coat and manipulate 
surface topographies in which sharp asperities or nano-roughness could potentially limit 
device performance.  Ultimately this work will play an important role in the design and 
nanofabrication of devices which exploit nano-processes such as PAA and ALD.  
 
 
 
72 
 
Chapter 4 :  TiO2 ALD Thin Film Growth Behavior Across 2D 
Planar Wafers Using TTIP and Ozone 
This chapter discusses the critical role of interdependent process parameters, such 
as temperature, dose, surface residence time, and purge time, have on one another as the 
process window for TiO2 ALD thin films based on the precursors titanium 
tetraisopropoxide (TTIP) and ozone is determined across planar substrates.  As with any 
deposition technique, determining a process window of parameters is critical when 
optimizing a new ALD process.  This usually incorporates a limited number of deposition 
experiments combined with ex-situ thickness and uniformity measurements.  In the case 
of ALD, an effective regime is typically indicated by the self-limiting properties of the 
ALD precursors thus leading to a constant thickness per ALD cycle.  This behavior is 
confirmed by wafer-scale measurements of across-wafer thickness and uniformity.  As 
seen in Chapter 1, the ALD process space includes many parameters that can influence 
the growth behavior of a film including dose, surface residence time, purge time, process 
pressure and temperature.  Even when limiting ourselves to the most obvious parameters, 
the task of finding an optimal process recipe can prove challenging. 
 
 
73 
 
4.1 Introduction 
Titanium dioxide (TiO2) has been extensively studied due to its versatile chemical 
and physical attributes for a wide range of applications as a thin film and a bulk material.  
Titania?s high refractive index, ranging from 2.2-2.6 depending on the deposition method 
used, makes it a common component in optical filters and coatings.[33, 106]  Due to its 
good transmission in the visible and near infra-red regions, TiO2 is a leading 
photocatalyst and serves as a critical component in emerging photovoltaic technologies, 
as well as in photonic crystal waveguides.[16, 33, 107]  TiO2 also exhibits some very 
attractive electrical properties in terms of it high dielectric constant of 40-110, making it 
a highly sought after candidate for electrostatic nanocapacitors, DRAM, and thin-film 
transistor applications.[8, 14, 48, 106, 108]  The main limitation in using TiO2 for 
electrical applications is its high leakage current caused by oxygen deficiencies in the 
material.[14, 106]  
 A variety of thin film deposition techniques have been used to deposit TiO2, 
including chemical vapor deposition (CVD)[55], sputtering[109], plasma-enhanced CVD 
(PECVD)[110, 111], and atomic layer deposition (ALD)[14, 104, 112-114].  ALD has 
attracted widespread attention on account of its excellent atomic level thickness control 
over large surface areas with unprecedented uniformity and conformality.  The ALD 
process is a gas phase deposition technique which relies on self-limiting chemisorbed 
surface reactions through the employment of two alternating reactant precursor exposures 
separated by intermediate purge steps in order to build films atomic layer by layer.  The 
digital introduction of precursors helps diminish gas-phase particles that sometimes 
 
 
74 
 
plague CVD processes in order to create thin, continuous, pin-hole free films.  However 
the optimization of ALD process recipes is often challenging as interdependent variables, 
such as dose, surface residence time, purge time, temperature or pressure, must be 
optimized for a given reactor design and chemistry.   
Precursor chemistries play an important role in ALD in that they must be volatile, 
yet thermally stable.[4]  A wide variety of precursors have been employed for the 
deposition of TiO2.  Commonly used Ti precursors are halides, alkyl amides, or 
alkoxides, such as TiCl4, tetrakis(dimethylamino) titanium (Ti[N(CH3)2]4, TDMAT), 
titanium tetraisopropoxide (Ti[OCH(CH3)2]4, TTIP), respectively.[104-106]  Each 
precursor group has their limitations; halides produce HCl as a corrosive by-product 
provoking etching of the films or the reactor and the organometallic precursors tend to 
decompose at temperatures greater than 300 oC leading to parasitic CVD-like 
reactions.[4]  The most common oxidizer used for the deposition of TiO2 is water, but 
hydrogen peroxide, molecular oxygen, and oxygen plasma have also been used.[57, 104, 
105, 112, 113, 115]  The push to find high-k dielectric materials with low leakage current 
has lead researchers to investigate ozone (O3) as an alternative to water as an oxidizing 
agent as it reduces the incorporation of impurities, such as hydroxyl ions (OH-).[116-118]  
O3 is a strong oxidizing agent and is highly volatile allowing for lower doses, lower 
deposition temperatures, and unlike water, physisorption on the surface is not a problem 
allowing for shorter purge times.[56, 117]  However, despite the interest in O3 as an ALD 
oxygen source, the surface reaction mechanisms are not well understood.  In H2O and O2 
plasma-based ALD processes, hydroxyl groups are generally considered the reactive sites 
for the chemisorption of the metal precursor.  However, during the ALD of TiO2 from 
 
 
75 
 
TTIP and O3, surface carbonates have been identified as the reactive sites for TTIP 
chemisorption.[56]  
 The self-limiting behavior of an ALD reactant depends on many factors, such as 
process temperature, reactant dose, surface residence time, purge time, process pressure, 
etc.  However, thus far, only a small set of ALD processes have been explored in order to 
understand the detailed relation between process parameters and their surface reaction 
mechanisms that happen during deposition.  It is generally known that ALD is a 
thermally activated process, where lower temperatures usually result in the use of higher 
reactant doses in order to achieve self-limiting growth; however, this leads to longer 
surface residence times.[65]  The reactant dose (Pt) is defined as the partial pressure of 
the precursor times the exposure time.  Where the partial pressure of a reactant is 
determined by how many molecules of a precursor are introduced into the system and the 
exposure time is the amount of time needed to supply a saturating dose (S) to the 
surface.[66]  Conventional methods to increase exposure doses are typically done by 
extending the precursor exposure times, therefore, allowing more time for molecules to 
react and saturate the surface.[66, 67]  However, longer pulse times can lead to low 
throughput.  As seen in work by Li [65], it was recognized that higher process pressures 
resulted in higher surface residence times and had considerable influence on film growth 
rates.  The surface residence time is how long reactants remain exposed to a surface 
before being pumped away to the exhaust and is based on the partial pressure of a 
reactant within the system and the temperature of the reaction chamber.  Therefore, this 
chapter investigates the role of key process parameters such as temperature, reactant 
dose, surface residence time and purge time on the growth behavior of TiO2 ALD thin 
 
 
76 
 
films when using TTIP and ozone as the metal and oxidizing precursors, respectively.  
We use a novel scheme to introduce reactants into the reaction chamber by breaking up 
the reactant dose and introducing it as a set of ?pulses in parallel or in series.   
4.2 Experimental Setup 
High-k dielectric titanium dioxide (TiO2) films were deposited in a custom built 
stainless steel ultra-high vacuum (UHV) cross-flow wafer scale ALD reactor using 
alternating reactant exposures of TTIP (99.99% grade, Air Liquide) and ozone (O3).  A 
0.1L mini-reactor is embedded in the UHV process chamber equipped with a 100 mm 
substrate heater and a pneumatically actuated cap.  Precursors and nitrogen gases were 
introduced and exhausted through 2 opposing slits positioned 3 mm above the wafer.  
The mini-reactor was operated at 100 mTorr and the substrate heater was maintained at 
200 ?C.  The substrates used were 100 mm Si (100) test wafers cleaned in a 1% 
hydrofluoric acid, followed by a DI rinse and dried via N2 gas. 
 Liquid titanium tetraisopropoxide (TTIP) was used as the titanium metal 
precursor.  The TTIP precursor container was heated to 58 oC in order to achieve a vapor 
pressure of ~0.6 Torr. In order to achieve reproducible doses of the TTIP precursor, fast 
action pneumatic ALD valves were implemented on the inlet and outlet of a temperature-
controlled 20 mL volume monitored by a 0-10 Torr capacitance manometer pressure 
gauge.  TTIP doses were controlled by timing the opening of these valves.  The inlet 
valve to the volume was opened for 1 s in order to fill the volume to the vapor pressure of 
the TTIP precursor.  Upon reaching a given time setpoint, the inlet valve between the 
precursor source and the volume was closed and the gases were released into the chamber 
 
 
77 
 
for an exposure time of 0.3 s.  By monitoring the minimal and maximal pressures during 
each cycle, the number of molecules entering the system were approximated from the 
ideal gas law, pV=nkT (where p is the pressure, V the volume, n the number of ?mols, k 
the Boltzmann?s constant, and T the absolute temperature). 
Ozone was obtained by flowing 200 sccm of O2 (Grade 5, Praxair) into an MKS 
Instruments O3mega ozone generator.  The flow produced a 20psig pressure in the single 
electrical discharge cell.   At 100% power, the ozone concentration at the outlet was up to 
20 wt% (300 g/m3) with the balance being O2.  Due to the high flow and high pressure 
required for the operation of the O3megaTM a different control scheme of the ozone 
exposure was necessary.  The O2/O3 mixture was flowed continuously into an HA-series 
catalytic ozone destruct from Ozone Engineering to deactivate ozone into oxygen.  
Therefore, only a fraction of the total flow was pulsed into the reactor using a Swagelok 
medium flow metering valve in series with a fast action pneumatic ALD valve.  For a 
given needle valve conductance, the number of ?mols of ozone was determined by 
integrating the pressure peaks measured in the mini-reactor for different exposure times 
and comparing them to integrated peaks from known doses determined through pressure 
measurements in an attached trimethylaluminum vessel. 
Wafer maps of film thicknesses and refractive indexes were obtained using a 
Sopra GES5 spectroscopic ellipsometer integrated with an automated xy stage and a 75W 
high intensity Xenon arc lamp to obtain a continuous spectrum of light from 2.0-6.0 eV.  
Thickness profiles were taken at 5 equidistant points across the wafer in the direction of 
the cross-flow, with the first point being closest to the gas inlet and the last point being 
 
 
78 
 
closest to the exhaust outlet.  The percentage of non-uniformity within a wafer was 
determined by the ratio of the standard deviation over the average film thickness (?/?). 
4.3 Results 
In order to investigate the influence of substrate temperature on the deposition of 
TiO2 ALD temperatures were varied from 160 ?C to 200 ?C with a viscous flow regime 
of 100 mTorr.  TTIP doses of 0.24 Torr-sec and over-saturating exposures of ozone, 1.3 
Torr-sec, were introduced into the system followed by an evacuation purge of 20 sec for 
each precursor.  Figure 32 displays the relationship between film thicknesses and the 
changing number of deposited cycles for two different temperatures, 160 ?C and 200 ?C.  
The plot demonstrates fairly good thickness control of the titania process at a processing 
temperature of 200 ?C, where the slope is the growth rate per cycle and is equivalent to 
~0.65 ?/cycle.  Films were deposited with the same conditions for 150 cycles at a 
substrate temperature of 250 ?C and were measured to have a growth rate of ~0.66 
?/cycle.  However, for the 160 ?C case, a growth rate of only ~ 0.2 ?/cycle was 
achieved.  The term ALD window usually refers to the temperature range in which the 
growth rate is constant.[4]  Lower growth rates could be attributed to the substrate 
temperature not being high enough to overcome the activation energy required for a TTIP 
molecule to chemisorb on the surface.  Therefore the optimal temperature window for 
TiO2 ALD using TTIP and ozone was deemed to be between 200 ?C and 250?C.  
However, temperatures higher than 250 ?C where not investigated due to the 
decomposition temperature of TTIP to be reportedly around 275 ?C [105, 112, 113].   
 
 
79 
 
 
Figure 32.  Film thicknesses as a function of number of ALD cycles as 
determined by ellipsometry for two different reaction temperature, 160 ?C and 
200 ?C.  Linear regression demonstrates a growth rate of ~0.65 ?/cycle for the 
200 ?C case and that the growth never really took off for the 160 ?C case. 
 
 
80 
 
 
Figure 33.  Representation of pressure within the mini-reactor during three 
different runs: (a) single ?pulse of TTIP, (b) 3 ?pulses of TTIP in parallel and 
(c) 3 ?pulses of TTIP in series. 
 
In order, to illustrate the effect of the process parameters dose and surface 
residence times of a reactant on the deposition rates of TiO2 thin films across planar 100 
mm Si wafers a set of experiments were designed using a novel scheme to introduce the 
TTIP precursor into the mini-reactor chamber.  Figure 33 displays alternating exposures 
of an over-dosed ?pulse of ozone (1.3 Torr-sec) and three different scenarios for 
introducing TTIP doses: a single ?pulse, ?pulses in parallel, and ?pulses in series.  
Figure 33a demonstrates the first scenario of a single ?pulse of TTIP being flowed into 
the mini-reactor followed by a 20 sec purge before introducing the ozone precursor and is 
 
 
81 
 
monitored by a down stream pressure gauge.  By subtracting the base pressure and 
integrating under the curve the dose of 1 ?pulse of TTIP was determined to be 0.24 Torr-
sec.  The surface residence time was determined by monitoring how long it took for the 
increased process pressure during a precurosor half-cycle to return to the base pressure of 
100 mTorr.  Therefore, surface residence time of the TTIP precursor was determined to 
be ~10 sec and the ozone surface residence time was determined to be ~ 14 sec.  A 20 sec 
purge was deemed sufficient for both precursors to completely evacuate the reactor.   
We refer to exposures introducing TTIP ?pulses as ?in parallel? if they comprise 
of a group of consecutive rapid fire single ?pulses, meaning that the ?pulses are spaced 
so closely in time that the 2nd ?pulse subsequent in the sequence occurs well before a 
residence delay is observed for the previous ?pulse, i.e. the ?in parallel? case behaves 
like a single bigger pulse.  This allows for higher TTIP doses by increasing the partial 
pressure, but also increases the TTIP surface residence time within the mini-reactor, as 
seen in Figure 33b for 3 ?pulses again followed by a 20 sec purge.    It was noticed that 
the TTIP surface residence time for doses introduced in parallel push the limits of the 20 
sec purge by barely having sufficient time to return to the base pressure before the ozone 
?pulse is introduced.  The third scenario introduces TTIP exposures as ?in series? which 
are a combination of single ?pulses separated by a 10 sec delay allowing the pressure 
between each ?pulse to head towards the base pressure, as seen in Figure 33c for 3 
?pulses and a 20 sec purge immediately after the introduction of the last ?pulse.  After 
each ?pulse the pressure within the chamber heads towards the base pressure before the 
next ?pulse is introduced, allowing for higher doses, but without increasing the surface 
 
 
82 
 
residence time of the precursor within the system.  Therefore, the TTIP surface residence 
time will be consistent with the single ?pulse case and ensures that the 20 sec purge is 
adequate before introducing the ozone ?pulse.   
One of the hallmarks of the ALD process is the self-limiting surface chemistry, 
meaning that an increased dose should still yield a consistent film thickness.  However, 
growth rates are not only a function of the dose, but are also dependent on the surface 
residence time of the precursor within the chamber.    The reported density of TiO2 ALD 
films is 3.84 gcm-3 [119] and was used to calculate an ideal monolayer of film thickness 
as 31?= ?d  ~3.3?.[120]  Amorphous TiO2 ALD films were deposited using an over-
saturating dose of ozone with one of the three dosing schemes described above.  A 
precursor?s partial pressure within a volume is proportional to the number of molecules 
with in said volume; therefore, as the number of ?pulses increase, the partial pressure 
also increases, thus creating a direct relationship between the dose and the number of 
molecules in the system.  Figure 34 demonstrates the relationship between a film?s 
growth rate as a function of the number of molecules associated with the three dosing 
schemes used to introduce TTIP into the reaction chamber.  The plot demonstrates that 
increased doses lead to higher growth rates and subsequent film densities.  However, as 
discussed earlier, doses introduced in parallel will have higher surface residence times 
than those introduced in series.  Although the same number of TTIP molecules entered 
the system for each scheme, the increased surface residence times led to higher growth 
rates and increased film densities, again exemplifying the importance of a precursor?s 
surface residence time.  It can be believed that the increase in dose leads to an over-
 
 
83 
 
saturation of precursor within the reaction chamber.  Therefore, when a delay between 
?pulses is introduced as is the case with the ?in series? scheme, the delay behaves as an 
intermediate purge step, allowing some of the excessive molecules to purge away, as well 
as, desorb any physisorbed molecules.  The plot also demonstrates that doses introduced 
in series have higher growth rates than the single ?pulse case even though they have 
similar surface residence times within the process chamber.  This goes against the idea 
that ALD is self-limiting and results in much more complex mechanisms than originally 
thought. 
 
Figure 34.  Growth rate as a function of the number of TTIP molecules when 
doses are introduced in parallel or in series.  Demonstrates that an increase in 
molecules leads to higher growth rates and subsequent film densities, however, 
TTIP doses introduced in series have lower growth rates and densities as 
opposed to their in parallel counterparts. 
 
 
84 
 
Purge time is a crucial parameter in the ALD process sequence.  Inadequate purge 
times to fully exhaust any unreacted precursors from the reactor before the introduction 
of the next precursor will result in an overlap of the two precursors, creating parasitic 
CVD-like reactions leading to higher growth rates.  However, decreasing growth rates 
can indicate desorption of adsorbed species during extended purges contradicting the 
fundamental requirement that a chemisorbed reaction is irreversible.   TiO2 films were 
deposited for 250 cycles at a substrate temperature of 200 ?C using two different TTIP 
dosing schemes: the single ?pulse (0.24 Torr-sec) and the 3 ?pulses in parallel (0.56 
Torr-sec).  Both cases used an over-saturating dose of ozone (1.3 Torr-sec) and a 
subsequent 20 sec purge.  Growth rates plotted as a function of TTIP purge time exhibit 
the self-limiting behavior typical of a TiO2 ALD deposition (Figure 35).  While a 10 sec 
purge seemed adequate to deposit self-limiting films for the single ?pulse case, it does 
not appear to be sufficient for the dose of 3 ?pulses in parallel.  This is due to the fact that 
the surface residence time of the TTIP molecules during the 3 ?pulse case exceeds the 10 
sec purge time used; therefore creating a parasitic CVD reaction when the ozone pulse is 
introduced.  However, as purge times are increased, the saturating behavior for both cases 
demonstrates that the purge time did not affect the growth rate indicating that the higher 
growth rate was not due to multi-layer physisorbed molecules or parasitic CVD-like 
reactions and the origin of the higher growth rates can be attributed to a density increase 
of surface carbonates or hydroxyl groups as a result of higher TTIP doses and surface 
residence times.   
 
 
 
85 
 
 
Figure 35.  Growth rates as a function of TTIP purge time for two different 
TTIP doses and an over-saturating dose of ozone (1.3 Torr-sec) for 250 cycles at 
a reaction temperature of 200 ?C.   
 
Across wafer uniformity is also a useful process metric to evaluate the self-
limiting behavior of ALD.  TiO2 ALD films were deposited across 100 mm wafers for 
250 cycles using an over-saturating dose of ozone and one of the three dosing schemes 
for TTIP at a process temperature of 200 ?C.  Figure 36 displays highly uniform film 
growth rates associated with the three different TTIP dosing schemes as a function of 
position across a wafer, with 0 mm being closest to the inlet and 70 mm being closest to 
the exhaust.  The single ?pulse case displays the best film uniformity of 99.6% (standard 
deviation, 1 sigma) where the uniformity only degrades by 2% as the worst case 
associated with the ?in parallel? cases.  These results support the idea that film deposition 
was not influenced by parasitic CVD-like reactions and implies that all of the active 
surface sites are reacted. 
 
 
86 
 
 
Figure 36. Growth rate profiles of TiO2 ALD films after 250 cycles deposited on 
100mm wafers. 
 
4.4 Discussion 
It is widely accepted that in H2O and O2-plasma based ALD processes, hydroxyl 
groups play an important role as the reactive sites for the chemisorption of the metal 
precursor, including TTIP.[56, 121]  Conversely, Rai et al reported that with the use of 
ozone as the oxidizing agent, the dominating reactive species for TTIP chemisorption 
were surface carbonates.[56]   When ozone is used as an oxidizer, the reaction goes 
through a combustion process producing CO2 and H2O as by-products, leaving surface 
carbonates as the reactive sites for the TTIP molecules.[56, 122, 123]  However, if purge 
times are not deemed long enough after an ozone dose, physisorbed, i.e., weakly bound 
 
 
87 
 
molecules through Van-der Wall forces, water molecules become a source of hydroxyls, 
therefore, partially re-hydroxylating the surface for a TTIP molecule.[47]  Although, the 
use of ozone greatly reduces the presences of H2O it does not eliminate it therefore 
creating two competing reaction pathways during the TiO2 ALD process.  Reactions with 
surface carbonates are deemed 1st order reactions, where reactions with surface ?OH 
groups are considered secondary reactions.  However, the growth rates associated with 
ozone based TiO2 ALD are comparable to H2O and O2-plasma assisted works, suggesting 
that the increase in growth rates during the ALD processes are possibly associated with 
the method at which TTIP doses are introduced into the system.[54, 56, 57, 124, 125] 
Ideally, the self-limiting surface adsorption and reaction process of ALD should 
garner one monolayer of film thickness per cycle, provided sufficient doses and purge 
times are supplied.  However, a perfect layer by layer growth does not always occur, and 
the growth behavior for which it does is deemed the ?ideal? case.[126]  Therefore, most 
ALD processes deposit less than a monolayer of film and are dubbed ?non-ideal?.  
Consequently, the non-ideal behavior observed during TiO2 ALD using TTIP and ozone 
is not surprising due to the large steric hindrance associated with the large size of the 
allkoxide ligands of the TTIP molecule.  Figure 37 is a schematic of the TTIP molecule 
and Table 2 is its associated bond lengths and angles.[55]  From the associated bond 
lengths and angles the radius of the TTIP molecule was calculated as 5.1 ?, which is 
much larger than the 3.3 ? associated with one monolayer of TiO2 film. 
 
 
88 
 
 
Figure 37.  Schematic of the structures of a free TTIP molecule, where the black 
ball is the Ti atom, the dark grey balls are oxygen, the light grey balls are 
carbon, and the white balls are hydrogen.  Their associated bond lengths and 
angles are reported in Table 3.[55] 
 
Table 2. Bond lengths (?) and angles (degrees) associated with the titanium 
tetraisopropoxide molecule shown in Figure 38.[55] 
 
 
 
89 
 
Through investigative studies by Rai, it was determined that H2O is a by-product 
during the TTIP ligand exchange during TiO2 ALD.[56]  This has lead to the use of TTIP 
as both the metal containing and oxidizing precursor at elevated temperatures nearing its 
decomposition temperature.[127]  Due to this innate behavior of the TTIP precursor, the 
increase in surface residence time of the TTIP molecule is known to have a significant 
effect on the extent of decomposition and formation of excessive H2O during the TTIP 
half cycle.[128, 129]  Therefore, the use of doses in parallel will increase the H2O 
production and lead to unexpected secondary reactions leading to higher growth rates.  
However, when doses are introduced in series, the H2O production is significantly 
decreased, leading to lower growth rates than the parallel case.  Nevertheless growth rates 
were still higher than those of the single ?pulse case.  This can be attributed to the 
extended purge during the TTIP half cycle due to the 10 sec delays between ?pulses, 
where excessive H2O molecules produced during the ligand exchange will have less 
opportunity to create secondary reactions.  Although all films demonstrated an increase in 
growth rate, none exceed the deposition of a full monolayer of TiO2 film growth. 
 
 
90 
 
4.5 Conclusion 
The growth behavior of TiO2 ALD thin films have been investigated as a function 
of key interdependent ALD process parameters, such as temperature, reactant dose, 
surface residence time, and purge time.  It was determined that TiO2 ALD films deposited 
using TTIP and ozone as precursors are thermally activated between 200 ?C and 250 ?C 
and display good linearity between film thicknesses and cycle numbers.  A unique dosing 
scheme to introduce precursors was developed in order to examine the relationship 
between a precursor?s partial pressure and its surface residence time.  It was determined 
that growth rates and subsequent film densities are highly dependent upon the surface 
residence time of a precursor.  It was proposed that the use of TTIP and ozone as 
precursors leads to a more complex ALD reaction mechanism, where the production of 
H2O was assumed to be a by-product during the TTIP half-cycle leading to secondary 
reaction mechanisms and higher growth rates.  Longer purge times and excellent across 
wafer uniformity ascertained that secondary reactions are the reason for higher growth 
rates and are not associated with multilayer physisorption or parasitic CVD-like 
reactions. 
 
 
 
 
91 
 
Chapter 5 : Metrology of ALD Conformality in 3D Ultra-high 
Aspect Ratio Nanopores 
This chapter details a method to extract film thicknesses and subsequent 
conformality of deposited ALD films in ultra-high aspect ratio nanopores with the use of 
a chemical analysis tool.  TiO2 ALD films were deposited in porous AAO membranes 
with an aspect ratio of 111.  EDS line scans were taken across the cleaved interface of the 
membrane in order to extract Ti intensity profiles as a function of pore depth.  A 
relationship between the Ti intensity profiles and film thicknesses was determined 
through a cylindrical assumption and planar film thicknesses measured by ellipsometry.  
Film thickness profiles demonstrate a strong gradient down the length of the nanopore 
and are in strong agreement with direct measurements of wall thicknesses measured from 
corresponding released nanotubes imaged via TEM.    
 
 
92 
 
5.1 Introduction 
ALD?s ability to deposit conformal thin films in ultra-high aspect ratio structures 
makes a desirable technique used in the design and fabrication of nanotechnology.  
Traditional techniques used for extracting thin film thickness and conformality within 
high aspect ratio nanostructures, such as scanning electron microscopy (SEM) and 
transmission electron microscopy (TEM), usually involve extensive sample preparation, 
including sample thinning and cross-sectioning which typically takes a lengthy amount of 
time.  Nanopores present a considerable dilemma in terms of cross-sectioning due to their 
one dimensional shape and small diameters.   Ideally, the cross-section of a nanopore 
should be cut directly along the center axis in order to reveal the thin film composing the 
nanotube within.  When pores are cut accurately film thicknesses can be measured along 
the pore side walls down the length of the pore.  However, if it is not cut accurately along 
the pore axis film thickness measurements can be skewed as a result of the pores? 
cylindrical character, meaning film thicknesses can appear larger than they actually are.   
 Previous studies on conformality of thin films within high aspect ratio nanopores 
consist of cutting a bulk nanoporous film parallel to the nanopores? long axis and 
scavenging the cross-sectioned interface in order to find pores that have been cleaved 
along their specific axis.  Gordon et al. used a dual-beam SEM-FIB in order to cleave and 
image a thin film deposited inside elliptical pores etched deep into a silicon wafer.[66]  
The SEM-FIB cleaves the silicon along the natural crystallographic axis and allows for 
continuous cutting without having to remove the sample.  Elam et al. imaged a thin film 
deposited within a porous anodic aluminum oxide (AAO) by embedding the film in 
 
 
93 
 
epoxy and perpendicularly polishing the sample along the nanoporous axis.  This allows 
for top down images of the deposited in pores to be accurately measured at different 
depths within the sample.[67]  However, in order to obtain a highly resolved thickness 
profile down the length of a high aspect ratio nanopore would require several iterations of 
polishing and SEM imaging for one sample.  Employing traditional SEM cross-
sectioning techniques to characterize thin film conformality is possible, but proves to be 
time consuming and highly dependent upon accurate cleaving along a structures central 
axis.   
 Although SEM provides a direct observation of ALD films within nanopores 
allowing for qualitative assessment of ALD conformality, quantifying film thicknesses 
and subsequent conformality has proven to be difficult.  In combination with SEM 
imaging Elam et al. employed the use of electron probe micro-analysis (EPMA) in order 
to create zinc compositional profiles as a function of pore depth of a cross-sectioned ZnO 
ALD coated porous AAO membrane.  Although this work demonstrates the deposition 
behavior of various process parameters by calculating the relative concentration of Zn 
down the length of the nanopore it does not provide an actual thickness measurement. 
Perez et al. demonstrated a methodology to experimentally determine thin film 
thicknesses and subsequent conformality using template-synthesized nanotubes 
fabricated by ALD within a porous AAO template.[44]  The novelty of using AAO is that 
once thin films are deposited within the pores the template is removed by a dissolution 
process, releasing cylindrical nanotubes comprised of the ALD thin films.  Released 
nanotubes were then imaged using bright-field TEM and then analyzed using an image 
 
 
94 
 
analysis program based on a simple geometric model and the grey scale contrast of the 
image.  This technique gave direct measurements of film thickness and subsequent 
conformality of films down the length of a nanopore.  However, the shortcoming of this 
technique is that nanotubes have a tendency to break at various points along the axis 
making it difficult to locate complete nanotubes.  Also, this technique requires very clean 
sample prep in order to use the image analysis program. 
In this chapter, a method is proposed to extract thin film thicknesses and 
subsequent conformality down the length of an ultra-high aspect ratio nanopore through 
the use of cross-sectional SEM combined with energy dispersive spectroscopy (EDS) line 
scans.  EDS is a chemical microanalysis technique that detects the energy of x-rays 
emitted from the sample when the SEM electron beam hits the surface and does not 
require intensive sample preparation to perform measurements.  Film thicknesses were 
determined through a cylindrical assumption using metal intensity signals measured 
during EDS line scans and compared to film thicknesses directly measured from the 
release of ALD based nanotubes from a porous AAO membrane in TEM.  Although 
TEM analysis is a more accurate method of measuring film thicknesses and determining 
film conformality down the length of a nanopore, the rigid sample preparation and 
serendipity of finding a full nanotube means a comparable method is necessary. 
5.2 Materials and Methods 
A porous AAO template was synthesized using a two-step anodization method set 
forth by Masuda.[79]  Aluminum foils (99.99%, Alfa Aesar) were electropolished and 
anodized at 40 V and 10 ?C in an electrolytic bath of 0.3M oxalic acid.  Foils were 
 
 
95 
 
anodized for ~7 hrs during a first anodization in order to produced ordered orthogonal 
pores with respect to the substrate.  The 1st anodized membrane was removed using an 
aqueous mixture of phosphoric acid (6 wt%) and chromic acid (1.8 wt %), leaving a 
textured and ordered aluminum surface for the 2nd anodization to initiate from.  Foils 
were anodized for a 2nd time for ~2 hrs 17 min for a pore depth of ~10 ?m deep.  Pores 
were widened to ~90 nm using a pore-widening solution of phosphoric acid (0.1M) at 38 
?C.  Membranes were not removed from the underlying aluminum substrate; therefore, 
pores were only open on one end. 
 High-k dielectric TiO2 thin films were deposited by alternating reactant exposures 
of titanium tetraisopropoxide (TTIP) (99.99%, Air Liquide) as the metal precursor and 
ozone as the oxidizing agent.  The ALD equipment consists of a stainless steel ultra-high 
vacuum outer chamber with a 100mm mini-capped reactor embedded within.  Precursors 
were introduced through 3 mm slit raised above the wafer in a cross-flow direction and 
exhausted through an opposing downstream 3 mm slit to a mechanical pump.  Porous 
AAO templates were adhered flat to a 100 mm wafer by silver paint with the open ended 
pore exposed to precursor flow.  Substrate temperatures were maintained at 200 ?C and a 
base pressure of 100 mTorr. 
Liquid titanium tetraisopropoxide (TTIP) was held in a temperature controlled 
vessel heated to 58 oC in order to achieve a vapor pressure of ~0.6 Torr.  A TTIP dose of 
0.47 ?mols was fed into the reactor for an exposure time of 0.3 sec followed by a 20 sec 
evacuation of unreacted precursor and byproducts.  Ozone was produced by flowing 200 
sccm of O2 through an MKS O3mega ozone generator. An over-dosing exposure of ozone 
 
 
96 
 
(1.82 ?mols) at a concentration of 300 g/m3 was introduced into the reactor and followed 
by another 20 sec evacuation step.  A growth rate of 0.66 ?/cycle was achieved over 250 
cycles and was determined from the films deposited across the Si planar wafers the 
porous AAO templates were attached to using a Sopra GES5 ellipsometer. 
After ALD deposition, a piece of the porous AAO sample was used for TEM 
analysis.  Samples were placed in a 25 wt% phosphoric acid solution for 24 hrs, which 
dissolved the alumina template thus releasing the TiO2 nanotubes in solution.  Nanotubes 
were filtered using DI water and a drop of solution was placed on a standard holy-carbon 
TEM grid.  TEM studies were performed on a JEOL 2100F microscope with a LaB6 
source operated at 200 keV.  Images were taken using a post-column Gatan CCD camera 
and analyzed using ImageJ open source program.  Film non-conformality was determined 
by the [max-min]/[max+min].   
The rest of the coated porous AAO sample was used to make cross-sectioned 
samples for EDS analysis.  Samples were epoxied to a small piece of silicon with a notch 
scribed in the back center pore side down.  The backside of the AAO template was 
scratched to induce etching of the aluminum substrate in a Cu2Cl solution.  Templates 
were etched with help of a sonicator for approximately 20 min followed by sonicating the 
sample in isopropanol for an additional 10 min removing any Cu particles left on the 
surface.  Samples were then cleaved using the notch in the back side of the silicon as an 
initiator.  Before EDS measurements were taken samples were coated with a thin film of 
Au/Pd in order to reduce sample charging during measurements.  EDS studies were 
performed using a Bruker?s Quantax micro-analysis system attached to a Hitachi SU-70 
 
 
97 
 
high resolution scanning electron microscope operating at a 20 kV accelerating voltage 
and a magnification of 6K.  EDS line scans were performed across the cleaved cross-
section of the coated AAO membrane parallel to the nanopores, as seen in Figure 38a, at 
various stage tilts in order to determine a zero point.  Scans consisted of 100 equally 
spaced point measurements dictated by the position and length of the user drawn EDS 
line marker. 
 
 
Figure 38.  (a) SEM image displaying a cleaved TiO2 coated AAO template epoxied to a 
silicon substrate demonstrating the position of an EDS line scan position and (b) the 
corresponding averaged Ti Intensity curves versus the length of the EDS line scan for 
the tilts 5, 0, and -5 degrees. 
5.3 Results 
5.3.1 EDS Inferred Thickness 
Porous AAO membranes with dimensions of 90 nm in diameter and 10 ?m deep, 
creating an aspect ratio of ~111 were coated with 250 cycles of TiO2 ALD at a deposition 
temperature of 200 ?C.  Membranes were attached to silicon chips by firmly pressing the 
AAO template pore side down into a small dollop of epoxy making sure the epoxy is 
 
 
98 
 
thinned under the membrane in order to prevent excessive charging or drifting of the 
sample during measurements.  However, due to uneven pressing of the sample into the 
epoxy samples could be tilted at various angles due to the amount of epoxy left under the 
sample.  After the aluminum base of the AAO membrane was etched away samples were 
broken in half exposing a cleaved cross-section of the ALD coated AAO pores as seen in 
Figure 38a.  The SEM image in Figure 38a demonstrates the difficulty found in cleaving 
a porous AAO membrane as indicated by the many sheared pores across the interface 
making it nearly impossible to visually measure film thickness down the length of a pore.   
Therefore, EDS was used to measure the intensity of Ti atoms as a means to 
measure the composition of TiO2 down the length of the AAO membrane.  Oxygen could 
not be measured due to the template being comprised of Al2O3 itself, thus convoluting the 
O signal from the TiO2 ALD film.  EDS line scans approximately 11.5 ?m long were 
placed along the cleaved surface of the AAO membrane parallel to the pores as seen in 
Figure 38a.  Scans were taken starting from slightly below the pore entrance and ending 
well into the vacuum designated area.  However, depending on the tilt of the sample this 
does not necessarily guarantee the electron beam is entirely off the sample.  The electron 
beam has a plume depth and resolution of 1 ?m and for that reason samples were 
measured at various SEM stage tilts in order to determine a zero point dictating that the 
beam is no longer measuring the sample.  Samples were measured at 5, 0, and -5 degrees, 
with 0 being the original positioning of the stage.  As seen in the graph of Figure 38b, all 
three tilts display the decreasing intensity of the Ti down the length of the scan, however, 
scans made at stage tilts 0 and -5 degrees never measure 0% Ti inferring that the beam is 
 
 
99 
 
scanning at an angle back into the sample.  It is imperative to find the 0% ? 2% in order 
to determine the Ti intensity at the bottom of the pore without any overlap into vacuum.  
The electron beam moves in small increments and measures a percentage of Ti relative to 
the maximum amount of Ti measured in the sample.  Therefore, when near the bottom of 
the pore, the beam scans over part of the sample as well as the vacuum space off the 
sample.  Meaning that when the beam is at a scan location of 10.9 ?m the intensity 
profile is still indicating that Ti is being measured even though the sample is only 10 ?m 
long.  However, when samples are tilted due to sample preparation, not only are the scans 
indicating the beam is measuring more Ti atoms at the bottom of the pore, but the 100% 
intensity at the top of the pore is skewed as well due to the beam scanning over a larger 
area coated with 100% TiO2 (ie. the surface).  However, the scan made at a 5 degree 
stage tilt displays 0% Ti at the end of the scan.  Therefore, the EDS profile for the length 
of the AAO membrane can be determined by setting the 100% Ti intensity as the pore 
entrance and going a length of 10 ?m down the scan length to determine where the 
bottom of the pore is. 
In order to extract film thicknesses as a function of pore depth a relationship was 
devised between the intensity of a specified element measured by EDS and the volume of 
a nanotube.  It was assumed that ALD films deposited within a porous AAO membrane 
form cylindrical nanotubes and therefore, allowed a cylindrical assumption to be made: 
)()( xkVxI ?      Eq. 1 
 
 
 
100 
 
where, I is the EDS intensity for a specified element, k is a    constant, and V is the 
Volume of a cylinder (Eq. 2) at a given position x down the length of a nanopore (Figure 
39a). 
LxrrxV pore ))(()( 212 ?= pi     Eq. 2 
Where, rpore is the original radius of the nanopore, r1 is the radius of the nanopore after 
deposition at a position x down the length, L, of the nanopore.  As the electron beam 
scans the surface of the cleaved interface of the porous AAO it encounter hundreds of 
nanotubes due to the resolution of the electron beam.  Therefore, the model multiplies the 
number of nanotubes for which the beam scans by the area of the cylinder (Eq. 3), where 
the number of nanotubes is calculated by the width of the beam, ?, divided by the 
interpore spacing, Dint, of the AAO membrane (Eq. 4) (Figure 39b).  
))(()( 212 xrrNLxV pore ?= pi     Eq. 3 
intD
N ?=      Eq. 4 
Assuming the electron beam to be a cube, the model also takes into consideration the 
length of the nanopore the electron beam scan at any given time by multiplying the area 
of the nanotube by the length of the beam, ?, in Eq. 5. 
?pi? ))(()( 212 xrrNLxV pore ?=    Eq. 5 
By substituting Eq. 5 into Eq. 1 the intensity of an element measured by EDS can be set 
to the area of the nanotube as a function of position x being measured by a beam of 
 
 
101 
 
length, ? (Eq. 6). 
?pi? ))(()()( 212 xrrNkLxVkxI pore ?=?    Eq. 6 
In order to extract film thickness as a function of position x, the radius of the nanotube at 
position x after deposition must be determined (Eq.7).     
212
1 ])([)( opore AxIrxr ?=     Eq. 7 
Where, 
)(
)( 221
2
int
xI
rxr
k
DA pore
o
??==
pi?     Eq. 8 
It is known that the film thickness is the difference between rpore and r1(x) (Eq. 9) (Figure 
39c).  Therefore, we can determine Ao by using the planar film thickness as measured by 
ellipsometry to calculate r1(0). 
)()( 1 xrrxt pore ?=      Eq. 9 
By using Eq. 7 in conjunction with Eq. 9, thickness profiles can be extracted for ALD 
films down the length of a porous AAO membrane by using the I(x) associated with the 
position within the pore. 
 
 
102 
 
 
Figure 39. Schematic of how the EDS inferred thickness method was derived. (a) 
Demonstrates an electron beam of width ? scans several nanotubes of length (L) 
at position x.  (b) Demonstrates the electron beam of area ?2 spanning several 
nanopores within the porous AAO template with an interpore spacing of Dint.  (c) 
Demonstrates film thickness at position x is a function of the inner radius at 
position x.  
 
Pores 90 nm in diameter and 10 ?m deep were coated with 250 cycles of TiO2 
ALD resulting in a planar film thickness of 16.6 nm.  Samples were cleaved and placed in 
the SEM at a -5 degree stage tilt using an SEM probe current of 20 kV.  EDS line scans 
11.5 ?m in length were placed across the cleaved interface parallel to the pores.  TiO2 
ALD film thickness profiles within the porous AAO membrane were extracted using the 
EDS inferred thickness method and plotted versus pore depth (Figure 40).  Thickness 
 
 
103 
 
profiles demonstrate a strong gradient down the length of the pore with a conformality of 
~55%.  Conformality within the last 0.5 ?m of the pore is ~88%. 
 
Figure 40.  Graph of EDS inferred film thickness as a function of pore length.  
Curve displays a notable gradient down the length of the pore with film 
thickness measuring around ~7nm near the bottom of the poor. 
5.3.2 TEM Nanotube Metrology 
TEM images reveal HfO2 nanotubes produced by the removal of the AAO 
membrane after ALD deposition are cylindrical in shape, characteristic of the pores in 
which they were formed, as seen in Figure 41.  Nanotubes appear as long rectangular 
ribbons with dark edges running along their lengths.  Figure 41a displays a cluster of 
nanotubes illustrating the broadening structure reminiscent of a nail-head connecting the 
nanotubes to one another.  This structure is created by the deposition of ALD on the top 
surface of the AAO membrane.  Figure 41a also demonstrated the thinning of the wall 
down the length of the nanotube indicating reduced deposition deep within the pores and 
a decrease in conformality.  Figure 41b shows a single HfO2 nanotube with no nail-head 
 
 
104 
 
top implying the nanotube broke midway down the length of the nanotube.  The inset in 
Figure 41b is the selected area diffraction (SAD) pattern confirming the amorphous 
character of the HfO2 nanotubes.  To see the fabrication method of HfO2 nanotubes 
please refer to Appendix V. 
 
 
Figure 41.  TEM images of HfO2 nanotubes made by deposition of HfO2 ALD 
into porous AAO membranes followed by membrane dissolution;  a) Group of 
HfO2 nanotubes demonstrating the nail-head top of the nanotube and thinning 
of wall thickness down the length of the nanotube and b) a single HfO2 nanotube 
broken midway down the full nanotube with an inset of the selected area 
diffraction pattern dictating that the films are amorphous.[44] 
 
Bright field TEM records the intensity of electrons transmitted through a thin 
sample in order to produce an image.  Therefore, the darker edges of the nanotube can be 
attributed to a higher material density at the nanotube wall.  Due to this correlation the 
nanotube wall thicknesses can be extracted using ImageJ image analysis program.[44]  
Nanotubes were fabricated by the deposition of TiO2 using ALD within a 10 ?m deep 
nanoporous AAO membrane followed by dissolution of the membrane.  A TEM image 
 
 
105 
 
found in Figure 42 displays the bulbous bottom half of a 10 ?m long TiO2 nanotube with 
a measured wall thickness of ~ 7 nm near the bottom and ~ 9 nm about 0.5 ?m up the 
nanotube.  This constituted 87.5 % conformality within the bottom portion of the 
nanotube.  It should be noted that the TEM sample consisted of a significant amount of 
debris which is believed to be a mixture of dried phosphoric acid particulates and dust 
particles from the sample preparation. 
 
Figure 42.  TEM image of a TiO2 nanotube with wall thickness of ~7 nm near the 
bottom and ~9 nm about 0.5 ?m up the length of the nanotube. 
5.4 Discussion 
ALD is widely emerging in the world of nanotechnology due to its precise 
atomic-level thickness control and unprecedented conformality over very demanding 3D 
nanoscale topographies.  It has already been demonstrated as a useful deposition 
technique in the fabrication of energy devices[20-22], memory and thin film transistor 
applications[9, 14, 28], optoelectronics[15, 16], catalysis[17-19], coating 
nanoparticles[32-36], as well as various MEMS[23-27] and NEMS[5, 29, 49, 130] 
devices.  A large contingent of ALD process chemistries have been identified, placing a 
 
 
106 
 
great need on rapid characterization and optimization of ALD process parameters for 
nanostructure applications.[4] 
 The distinctiveness of ALD derives from its self-limiting adsorption/desorption 
reactions during each precursor half cycle.  However, the self-limiting behavior falls 
short due to depletion effects and process complexities associated with the precise 
amount of precursor and sufficient purge times to avoid parasitic CVD-like behavior.  
While the benefits of ALD are usually kept intact for planar cases, it is not always the 
case for high aspect ratio nanostructures, such as porous AAO, where deposition 
mechanisms are now diffusion limited rather than surface kinetics limited.[66]  Due to 
this behavior, the development of new metrology techniques for measuring thin films in 
complex nanostructures is critical as is emphasized by the International Technology 
Roadmap for Semiconductors.[131]  However, qualitative and quantitative metrology for 
novel nanostructures has proven challenging due to decreasing dimensions encountered 
in next generation devices. 
The use of EDS line scans as opposed to SEM and TEM imaging is meant to help 
alleviate frustrations found during sample preparation and imaging of ALD thin films 
within nanoporous structures.  Sample preparation of this method is meant to be 
relatively easy as precise cleaving of a nanopore is not necessary due to the fact that the 
electron beam used during SEM measurements penetrates the sample well past the 
interface of the cleaved surface and creating a cascade of x-rays to be measure via EDS.  
TiO2 ALD film thicknesses extracted from the EDS inferred thickness method are in 
excellent agreement with direct measurements of nanotube wall thicknesses determined 
 
 
107 
 
in TEM.  There could be discrepancies based on how well one can measure wall 
thicknesses from TEM images. 
5.4 Conclusions 
The combination of porous AAO and ALD presents an attractive platform for the 
advancement of nanotechnology; however, measuring thin films deposited in high aspect 
ratio nanostructures has proven difficult.  This work demonstrates an EDS analysis 
method that makes use of a cleaved, ALD coated porous AAO template revealing the 
parallel nanopores along the template?s cross-section.  The sample preparation of this 
method is meant to be relatively easy as compared to methods that utilize precise cross-
sectioning and sample thinning techniques and are considered quite time consuming.   
A means to extract film thicknesses and conformality of ALD thin films deposited 
in the high aspect ratio nanopore of an AAO template was demonstrated through the use 
of an EDS inferred thickness model.  The model was based on a cylindrical assumption in 
which a nanotube formed within an AAO nanopore is comprised of the ALD film used.  
Extracted film thicknesses from the EDS inferred thickness model were validated by 
comparing results to film measurements of a released nanotube imaged via TEM.  Initial 
film thicknesses within the nanopore/tube were considered to be equivalent to film 
thicknesses measured across the planar surface.  The results from the EDS inferred 
thickness method are comparable to direct TEM measurements and demonstrate the 
usefulness of this method to extract film thickness and conformality, as will prove to be 
directly relevant to the understanding and control of ALD processes and their integration 
into nanotechnology.   
 
 
108 
 
Chapter 6: ALD Conformality and Process Optimization in 3D 
Ultrahigh Aspect Ratio Nanopores 
This chapter discusses the critical role of interdependent process parameters, such 
as reactant dose, surface residence time, and purge time on the conformality of an ALD 
thin film when transitioning from a 2D planar system into that of a 3D ultra-high aspect 
ratio nanopore.  It has been demonstrated in Chapter 4 that excellent uniformity can be 
achieved with higher precursor doses and increased surface residence times, even though 
the growth rates exhibited a non-ideal saturation of the surface.  The production of H2O 
during the TTIP half-cycle is proposed as the culprit to secondary reactions that complete 
the ALD cycle and deposit more than the expected sub-monolayer of growth.  When 
transitioning from a 2D planar case in the viscous flow regime to that of a molecular flow 
regime found in an ultra-high aspect ratio nanopore, the complex reaction dynamics that 
occur in the planar system is intensified in a nanopore.  Increased doses are necessary to 
compensate for the increase in surface area, as well as longer surface residence times for 
molecules to diffuse down the pore and extended purge times to allow unreacted 
precursor and by-products diffuse out.  However, as these process parameters are deemed 
necessary to deposit conformal films within a high aspect ratio nanopore, the complex 
reaction dynamics that occur on the planar surface will have great influence on the 
deposition within the nanopore. 
 
 
109 
 
6.1 Introduction 
In order to keep up with Moore?s Law, the International Semiconductor Roadmap 
has emphasized a need to access reduced feature sizes found in the device designs of 
transistor and memory technologies.[132]  However, as capacitor technology transitions 
from the early days of planar capacitors to that of trench capacitors having aspect ratios 
greater than 60:1, conformal coatings are playing an increasing crucial role.[133]  
Conventional deposition techniques, such as the physical vapor deposition (PVD) 
methods sputtering and evaporation, as well as chemical vapor deposition (CVD), are 
found to have limited abilities to coat complex structures with increasing aspect ratios 
due to their poor step coverage.[66]  The increasing need to deposit thin, continuous pin-
hole free films in complex 3D geometries is not only limited to ULSI technologies, but is 
also essential to a wide range of other technologies, such as optoelectronics[2, 15, 16], 
catalysis[18, 49, 134], energy devices[20-22, 98, 135], gas sensors[30], micro-electro-
mechanical systems (MEMS)[23, 24, 26, 136], and nano-electro-mechanical systems 
(NEMS)[33, 37, 38, 49], that all exploit the benefits found in using nano-scaled structures 
with increasingly high aspect ratios.  Therefore, researchers have turned to atomic layer 
deposition (ALD) to offer high quality films with atomic level thickness control and 
unprecedented uniformity and conformality in the most demanding 3D nanostructures. 
ALD is a gas phase deposition technique which utilizes self-limiting surface 
chemistries to control film thickness down to the atomic scale.  This control is inherent in 
the self-limiting surface chemistry where reactants are sequentially pulsed creating 
saturated chemisorbed surface reactions in order to achieve a monolayer or less of film 
 
 
110 
 
per cycle.  Much work has focused on the interplay of key ALD process parameters, 
particularly in terms of thickness control and uniformity.[14, 68, 80, 104, 113]  It was 
determined that when in a 2D planar viscous flow regime, the rate limiting step of the 
ALD deposition is the surface reaction kinetics across a planar surface; therefore, when 
precursors are underdosed, a thickness gradient occurs due to an insufficient amount of 
precursor being supplied to completely saturate the surface.[66, 68]   
In recent years, there has been an increasing interest in the combination of ALD 
deposited within the high aspect ratio nanopores found in a porous anodic alumina (PAA) 
membrane.  PAA is a self-assembled, electrochemical templating technique that results in 
the formation of a high density of cylindrical nanopores within a close-packed hexagonal 
array on the surface of an aluminum template.[97, 137]  PAA naturally lends itself to 
many applications in nanotechnology, such as a template for nanowire and nanotube 
fabrication[29, 49, 98, 99], scaffolds for catalysts[18, 49, 134], dye sensitized solar 
cells[98], batteries[93], and sensors[30].  Recently Banerjee et al, demonstrated an 
operational metal-insulator-metal (MIM) electrostatic nanocapacitor by depositing 
multilayers of ALD TiN-Al2O3-TiN films within a PAA membrane with an aspect ratio 
of ~200.[21]  Although they demonstrated an equivalent planar capacitance of ~100 
?F/cm2 using Al2O3 as the dielectric layer (?k ~9), it is believed that with the use of high-
k dielectrics, such as HfO2 (?k ~14) or TiO2 (?k ~80), as well as, higher aspect ratio 
nanopores higher equivalent capacitance values can be achieved.[21, 138] 
The combination of ALD and PAA has led to considerable attention to optimizing 
ALD conformality within ultra-high aspect ratio nanopores in a molecular flow regime.  
 
 
111 
 
Works include simple models for determining precursor dose[66], Monte Carlo 
techniques to understand the importance of precursor sticking probability[67], solving 
simultaneously Boltzmann?s transport and surface reaction equations[101, 139] and 
correlating them with experimental results while trying to optimize conformality in ultra-
high aspect ratio structures.  These models and work demonstrate that when deposition 
transitions from a 2D planar structure in a viscous flow regime to that of a 3D 
nanostructure in a molecular flow regime different reaction mechanisms dominate the 
deposition, where in a 3D nanostructure the rate limiting step is the Knudsen diffusion of 
precursor molecules[66, 140].  Therefore, deposition within a 3D nanostructure in a 
molecular flow regime requires much more control over deposition parameters than for 
planar or micro-scale features found in a viscous flow regime.  Although there has been 
much focus on both the 2D and 3D cases separately, there has been little focus on the 
actual influence the 2D planar deposition may present on the conformality within a 3D 
ultra-high aspect ratio nanopore.  
In this chapter, we investigate the impact of key process parameters such as 
reactant dose, surface residence time, and purge time on conformality profiles of ALD 
layers deposited into the nanopores of a PAA membrane while comparing them to the 2D 
planar deposition behavior.  We focus on the deposition of TiO2 as a high-k dielectric 
material using titanium tetraisopropoxide (TTIP) and ozone as the metal containing and 
oxidizing precursors, respectively.  This study will illustrate that while uniform films can 
be achieved, the growth behavior across the planar surface will have a significant impact 
on conformality profiles within an ultra-high aspect ratio nanopore due to changing 
 
 
112 
 
aspect ratios and the interplay between diffusion and surface reactions when transition 
between viscous and molecular flow regimes. 
6.2 Experimental Setup 
PAA templates were synthesized using a two-step anodization method as seen in 
Chapter 2.[97]  In short, a piece of electropolished aluminum foil (99.99%, Alfa Aesar) 
was anodized at 40 V and 103 ?C in an electrolytic bath of 0.3 M oxalic acid.  Foils were 
anodized for ~7 hrs in order to produce well ordered parallel pores with respect to the 
aluminum surface.  These 1st anodized membranes were etched off by an aqueous 
solution of phosphoric acid (6 wt%) and chromic acid (1.8 wt%), leaving a pre-textured 
and ordered aluminum surface.  Foils were anodized for a second time for 2 hrs 17 min 
defining a pore depth of ~ 10 ?m.  Pore diameters were adjusted to ~90 nm by using a 
pore widening solution of 0.1 M of phosphoric acid at 38 ?C. 
High-k dielectric TiO2 films were deposited in a custom built ultra-high vacuum 
(UHV) cross-flow wafer scale ALD reactor, using titanium tetraisopropoxide (TTIP) and 
ozone (O3) as precursors.  Precursors were introduced and exhausted through 2 opposing 
slits positioned 3 mm above the wafer.  The mini-reactor was operated at 130 mTorr and 
the substrate heater was maintained at 200 ?C.  TTIP doses were varied using the same 
dosing scheme found in Chapter 4 (Figure 33) for a single ?pulse, 3, and 6 ?pulses in 
both parallel and in series, while ozone doses were maintained in an over-saturation 
regime.  Purge times were maintained at 20 sec between both precursors? half cycles, 
except when investigating the purge time parameter during the TTIP dose on deposition 
behavior. 
 
 
113 
 
 
TiO2 films were deposited on both flat Si substrates and in PAA templates.  Si 
samples were dipped in a 3% HF solution to remove the native oxide and PAA templates 
were adhered to the Si near the inlet of the precursors.  After deposition planar maps of 
film thickness were obtained using a Sopra GES5 spectroscopic ellipsometer with an 
automated xy stage.  Thickness profiles were taken of equidistant points across the wafer 
in the direction of the flow, with the first point being closest to the gas inlet and the last 
point being closest to the exhaust outlet.  The percentage of uniformity is determined by 
the ratio of the standard deviation over the mean value (?/?) of all points taken across the 
wafer subtracted from 100%. 
High resolution top down and cross sectional SEM was performed on bare and 
ALD coated PAA samples using a Hitachi SU-70 SEM equipped with EDS. Cross 
sectional PAA samples were made by removing the underlying aluminum in a CuCl2 
solution and cleaving the remaining PAA membrane. EDS line scans were taken across 
the cleaved ALD deposited PAA templates. Film thickness as a function of pore length 
was extracted from the EDS compositional scans by setting the intensity proportional to 
the cylindrical volume of the sample being measured.  The percentage conformality was 
determined by taking the (max - min) / (max + min)*100% and subtracted from 100%. 
6.3 Results 
PAA membranes were fabricated in oxalic acid and anodized for 2 hrs and 17 
mins allowing pores to grow to a depth of 10 ?m and pore widened for 46 min and 
achieving pore diameters of ~88 nm.  Pore diameters and depths were verified using top-
 
 
114 
 
down and cross-sectional SEM images.  The pore density was calculated from the 
interpore spacing of 110 nm and determined to be 1.27 x1010 pores/cm2.  One pore repeat 
unit consists of a hexagonal unit cell and a cylindrical pore, therefore the surface area of 
one unit is calculated to be 2.77 x 10-8 cm2.  Therefore, the surface area of a PAA 
membrane increases by ~350% as compared to a planar Si wafer. 
ALD is dependent upon sufficient doses in order to completely saturate a surface 
and deposit uniform and conformal films.  Therefore, the effects of the process 
parameters, dose and surface residence time of the TTIP reactant, on film uniformity 
across a 2D planar Si wafer versus conformality within a nanoporous PAA membrane 
were investigated.  The TTIP precursor dose was varied from an ?ideal? planar saturation 
level to that of over-saturating, while maintaining the ozone dose at an over-saturation 
level.  In order to increase the dose and surface residence time of a precursor, doses were 
broken up and introduced into the reaction chamber in the same fashion as that discussed 
in Chapter 4.  TiO2 films were deposited simultaneously across planar Si wafers and in 
PAA membranes with an aspect ratio of 111 for 250 cycles.  Figure 43 displays 4 
consecutive ALD runs demonstrating highly uniform thickness profiles across the surface 
of a 100 mm planar wafer, with the direction of flow going from 0 mm to 70 mm.  The 
single ?pulse case introduced 0.55 ?mols into the system with a surface residence time of 
about 10 sec, as determine by monitoring the increase in the reactor pressure until it 
returned to the base pressure of 100 mTorr (as seen in Chapter 4, Figure 33).  The single 
?pulse scheme was deemed the ideal planar saturation level as it produced ideal growth 
rates of ~0.66 ?/cycle, as reported in the literature, and had across-wafer uniformity of 
 
 
115 
 
99.6 % (standard deviation, 1 sigma).  As doses were introduced in parallel to achieve an 
over-saturating dose (i.e. increased TTIP partial pressure, therefore, increased amount of 
molecules) and subsequent longer surface residence times, thickness profiles 
demonstrated that wafer uniformity was maintained at ~ 98%, even as growth rates 
increased.  It was discussed in Chapter 4 that the increase in growth rates are due to 
complex secondary reaction mechanisms associated with using TTIP as a precursor.  
Interestingly, as doses were introduced ?in series?, the resulting film continued to 
maintain uniformity even though its growth rate decreased from the ?in parallel? 
counterpart as a result of the 10 sec delay between ?pulses creating an intermediate purge 
step. 
 
Figure 43.  Highly uniform films of TiO2 ALD thin films 
deposited across 2D planar wafers for 250 cycles for 
three different dosing schemes.  Demonstrates increased 
film thicknesses, yet maintained film uniformities. 
 
Figure 44a illustrates that the benefits of uniform ALD seen for planar wafers in 
 
 
116 
 
the viscous flow regime (Figure 43) do not translate into good conformality down the 
length of a high aspect ratio nanopore in the molecular flow regime.  Thicknesses with 
respect to 0 ?m pore depth correspond to the planar thicknesses at pore entrances, where 
deposition in a pore for the single ?pulse case demonstrates film thicknesses deplete from 
the initial 166 ? ? 1 ? at the pore entrance ( 0 ?m) down to roughly 70 ? ? 2 ? near the 
bottom of the pore (8.5 ?m).  This decrease in film thickness constitutes a decrease in 
conformality by ~30%.  Initially this drop in film thickness was believed to be due to 
insufficient doses as a result of higher surface area and/or insufficient surface residence 
times for a molecule to diffuse down the length of the nanopore.  However, as doses were 
increased from an ?ideal? saturating dose to an over-saturating dose ?in parallel?, films 
within the pore exhibited notable gradients down the length of the pore (Figure 44a).  
Conformality as a function of the number of molecules introduced into the system for 
each dosing scheme is shown in Figure 44b for three different regions within the 
nanopore.  Conformality was determined between 0 ?m to 8.5 ?m pore depth in order to 
determine the overall trend in conformality and avoid any influence that may be imposed 
by inaccurate computation of the EDS inferred thicknesses.  The results in Figure 44b 
demonstrate that films are ~70% conformal for the single ?pulse case (black dashed line) 
as opposed to ~35% conformal associated with the 6 ?pulse case introduced in parallel 
(blue dashed line) from 0 ?m to 8.5 ?m down the length of the pore.     
 
 
117 
 
 
 
 
Figure 44. a) TiO2 thickness profiles down the length of a 10 ?m deep nanopore 
for increasing doses introduced in parallel and b) their corresponding 
conformalities as a function of the number of TTIP molecules introduced in 
parallel for three different regions within the nanopore: 0-8.5 ?m, 0-5 ?m, and 5-
8.5 ?m. 
 
As the number of TTIP molecules is increased as a function of the dose scheme 
used, steeper gradients appear to occur within the first 5 ?m of the nanopore, implying 
two deposition regions may occur within the pore.  Figure 44b displays the conformality 
as a function of the number of molecules associated with each dosing scheme as 
deposited within 1) the top half of the nanopore from 0 ?m to 5 ?m and 2) the bottom 
half the pore from 5 ?m to 8.5 ?m.  The plot highlights that the first 5 ?m of a nanopore 
has a significant decrease in conformality as opposed to the bottom half of the pore.    
The results further indicate the decrease in conformality within the first 5 ?m as the 
number of TTIP molecules and subsequent surface residence times associated with the 
dosing schemes ?in parallel? are increased, where as the bottom half of the nanopores 
appear to be in good correlation to each other.  Therefore, the overall trend in 
conformality, from 0 ?m to 8.5 ?m, echoes the conformality within the top half of the 
 
 
118 
 
nanopore. 
Figure 44 indicated that an increase in TTIP dose and surface residence time 
produced significant gradients down the length of an ultra-high aspect ratio nanopore.  
Therefore, it was believed that an increase in TTIP purge time could significantly 
improve conformality within the pores.  Purge times were varied for a TTIP dose of 3 
?pulses introduced in parallel in combination with an over-saturating ozone dose, 250 
cycles of TiO2 were deposited simultaneously across 100 mm planar Si wafers and PAA 
membranes with an aspect ratio of 111.  Figure 45 demonstrates highly uniform films 
were maintained across planar wafers as purge times were varied from 10 sec to 60 sec.  
Thicknesses associated with a 10 sec purge resulted in significantly higher growth rates 
indicative of a CVD behavior, conversely, some CVD process are still capable of 
producing uniform films.  On the other hand, the combination of uniform films and the 
saturating behavior of the film thicknesses as purge times were increased, reiterated that 
deposition is attributed to the complex reaction dynamics associated with complex 
secondary reaction mechanisms when using TTIP as a precursor.   
 
 
 
 
 
119 
 
 
 
 
Figure 45.  Highly uniform films of 
TiO2 ALD thin films deposited across 
2D planar wafers using an over-dose of 
ozone and a TTIP dose of 3 ?pulses (1.6 
?mols) in parallel followed by various 
purge lengths. 
 Figure 46.  TiO2 thickness profiles 
down the length of a 10 ?m deep PAA 
nanopore for different purge times 
after 3 ?pulses introduced in parallel.  
Profiles are associated with the planar 
depositions found in Figure 48. 
Nonetheless, when transitioning to an ultra-high aspect ratio nanopore the benefits 
of planar uniformity still do not translate into good conformality within the pore.  Figure 
46 exhibits steep thickness gradients down the length of a 10 ?m deep nanopore.  Film 
thicknesses within the nanopores are dependent on film thicknesses found across the 
planar surface.  Planar results in Figure 45 indicate that film thicknesses associated with a 
10 sec purge exceed the 440 ? film thickness limit imposed by the nanopore diameter of 
88 nm.  Therefore, films deposited with in the PAA membrane actually clog the pore and 
result in a strong gradient within the pore.  Figure 47 plots the conformality of the films 
in an ultra high aspect ratio nanopore as a function of the purge time.  Results indicate 
that an increase in purge time does not improve film conformality rather it tends to 
saturate.  This behavior alludes to a more intricate interaction between a precursor?s dose 
and surface residence time in a molecular flow regime exists. 
 
 
120 
 
 
Figure 47. Conformality determined for thickness 
profiles found in Figure 49 as a function of the TTIP 
purge time following a dose of 3 ?pulses introduced in 
parallel. 
 
Thickness profiles displayed in Figure 46 show significant gradients down the 
length of an ultra-high aspect ratio nanopore and imply that longer purge times do not in 
fact improve conformality.  Therefore, doses were introduced in series with the fore 
thought that a 10 sec delay between ?pulses would act as an intermediate purge step, 
allowing more molecules to be introduced into the system without the increase in surface 
residence time.  TiO2 films were deposited across 100 mm planar wafers simultaneously 
with a PAA membrane with aspect ratios of 111 for 250 cycles.  As discussed earlier, 
Figure 43 shows a film deposited across a 2D planar surface with 98.6% uniformity 
(standard deviation, 1 sigma).  Although film growth rates are ~ 26% higher than those of 
the single ?pulse case, the intermediate purge step seems to dispel some of the secondary 
reactions associated with using TTIP as a precursor.  Therefore, when transitioning to an 
 
 
121 
 
ultra-high aspect ratio nanopore the benefits of planar uniformity are starting to be 
translated into good conformality within the nanopore.  Figure 48 reveals that although 
gradients still exist down the length of the pore, they are not as severe as the dose 
introduced in parallel scheme and tend to behave more like a single ?pulse.   
 
Figure 48.  TiO2 thickness profiles down the length of a 
3D ultra-high aspect ratio nanopore with an aspect ratio 
of 111 for different dosing schemes.  Profiles are 
associated with the planar depositions found in Figure 43. 
 
Figure 49 presents the conformality as a function of the number of TTIP 
molecules introduced into the system for the three different dosing schemes used, as well 
as, the various purge times investigated with the 3 ?pulses ?in parallel?  dosing scheme.  
Results signify that dose introduced ?in series? improves the conformality within a 
nanopore as compare to the single ?pulses case and doses introduced ?in parallel?.  It 
further demonstrates that the purge time had no significant effect on the conformality 
 
 
122 
 
with in an ultra-high aspect ratio nanopore.  Therefore, conformality benefits from the 
introduction of intermediate purges between ?pulses, as to allow more molecules to be 
introduced into the system to compensate for increased surface areas, but also allows 
unreacted precursor, by-products to purge out of the system while allowing any 
secondary reactions to equilibrate before more precursor is introduced. 
 
 
Figure 49.  Conformality determined for thickness profiles found in Figure 
44, 46 and 48 as a function of the number of TTIP molecules introduce 
into the system for various dosing schemes and purge times. 
 
6.4 Discussion 
   Maintaining wafer scale uniformity and nanopore conformality is dependent upon 
the complex reaction dynamics associated with chemisorbed reactions found within 
 
 
123 
 
different flow regimes.  Highly uniform TiO2 films were deposited across 2D planar 
wafers even with an observed increase in growth rates.  Chapter 4 attributed the increase 
in planar growth rates to the complex reaction dynamics associated with the production 
of H2O as a by-product when using TTIP as a precursor, where higher doses and longer 
surface residence times encouraged secondary reactions.  However, when transitioning to 
an ultra-high aspect ratio nanopore these complex reaction mechanisms are magnified 
and are highly influenced by the dose and surface residence time of the precursor.  As 
doses and surface residence times were increased to compensate for the increase in 
surface area and the diffusion of precursors down the length of an ultra-high aspect ratio 
nanopore, notable gradients were observed as a function of pore depth.  Several 
explanations can be attributed to the decrease in conformality within a nanopore, such as 
the transition from viscous flow to molecular flow, the continually changing aspect ratio, 
and the complex reaction mechanisms associated with using of TTIP and ozone as 
precursors. 
One of the trademarks of the ALD process is the self-limiting surface chemistry 
which is dependent on sufficient amounts of precursor to be supplied to completely 
saturate a surface.  Therefore, when transitioning from a 2D planar wafer to that of a 3D 
nanopore, higher doses are necessary to account for the increase in surface area.  
However, when transitioning from the 2D, viscous flow to 3D, molecular flow, higher 
doses are not the only requirement to deposit conformal films.  When transitioning from 
different flow regimes, different deposition mechanisms will dominate the deposition 
kinetics.  In a viscous flow, reactions are limited by the surface reaction times, whereas, 
 
 
124 
 
in molecular flow the rate limiting step is the Knudsen diffusion of the larger precursor 
molecules (in this case TTIP).  Therefore, in order to compensate for higher surface areas 
and longer diffusion times, the dose (i.e. the partial pressure) and surface residence time 
of the TTIP precursor were increased.  During the course of deposition the aspect ratio 
changes with every cycle requiring precursors to either be put in a constant over-
saturating regime or the doses and surface residence times need to be compensated for 
with each cycle.  Also with an increase in aspect ratio, it is imperative that unreacted 
precursor and by-products are given sufficient time to purge out of the system before the 
next precursor is introduced. 
The production of water during both the TTIP and ozone half-cycles is considered 
detrimental when in the confines of an ultra-high aspect ratio nanopore.  Water is 
considered a very ?sticky? molecule due to its high polarity and therefore, can linger 
within a system requiring extensive purge times.  In the case of ozone, if purge times are 
insufficient, a parasitic CVD-like reaction can occur within the nanopore due to both the 
metal and oxidizing precursors being present in the system simultaneously. To further 
compound the situation is the transition regime that exists between systems in viscous 
flow versus that of a molecular flow.  In the viscous flow regime, higher partial pressures 
of a precursor can lead to higher particle interactions directing the flow in a laminar 
streamline across a surface (i.e. the velocity at the wall is zero), therefore, the time it 
takes to completely saturate a surface is relatively short and decreases with higher 
pressures.  However, when in the molecular flow regime, the mean free path between two 
molecules is so large that the flow is dominated by gas-wall collisions, taking a longer 
 
 
125 
 
time for a reactant to find an active surface site.  When transitioning from one flow 
regime to another, a transition regime is created, where gas molecules not only collide 
with each other, but with the wall of the system as well.  In the case of TTIP, the 
production of water can have a detrimental impact on the deposition kinetics within this 
transition regime.  After a TTIP molecule adsorbs to the surface, its by-product of water 
can possibly self-react leading to a completed ALD reaction or it could react with the 
incoming flux of TTIP molecules, thus leading to a parasitic CVD-like reaction.  When 
CVD-like reactions occur within the pores the precursor is inevitably depleted thus 
creating thickness gradients down the length of the nanopore.  Nevertheless, some TTIP 
molecules will diffuse through the transition region unreacted and will chemisorb on the 
surface within the molecular flow regime, therefore, resuming an ALD reaction.  
However, molecules in a molecular flow regime are subject to their wall interaction and 
can therefore change direction regardless of their incoming trajectory.  Therefore, the 
probability of a molecule to continue to flow in the positive direction (i.e. towards the 
bottom of the pore) rather than in the negative direction (i.e. back towards the transition 
region) is cut in half.  Therefore, the deposition within an ultra-high aspect ratio nanopore 
can be thought of as a mixture of CVD and ALD-like behaviors. 
6.5 Conclusions 
ALD is receiving credit for some of the most demanding deposition processes in 
nanotechnology due to its exceptional ability to coat ultra-high aspect ratio structures 
with very thin films.  PAA provides a versatile and easy platform to analyze ALD 
profiles in order to optimize the ALD processes and achieve conformal films.  It was 
 
 
126 
 
demonstrated that maintaining atomic level thickness control with wafer scale uniformity 
and nanopore conformality is subject to the complex reaction dynamics associated with 
chemisorbed reactions in various flow regimes.  The influence of interdependent process 
parameters such as reactant dose, surface residence time, and purge time on uniformity 
and conformality profiles of TiO2 ALD was investigated.  The study illustrates that 
uniform films were achieved across a 2D planar wafer in viscous flow, even with 
observed growth rates.  However, the complex mechanisms associated with higher 
growth rates in the viscous flow regime are magnified when in the confines of an ultra-
high aspect ratio nanopore leading to a decrease in film conformality.  Several possible 
explanations were attributed to the decrease in conformality within a nanopore, such as 
the transition from viscous to molecular flow, the continually changing aspect ratio, and 
the complex reaction mechanisms associated with using of TTIP and ozone as precursors.  
Precursor doses, surface residence times, and purge times are only a few of the 
contributing factors in understanding the complex reaction dynamics found within the 
ALD process.  Further investigations of the complex interplay of process parameters is 
essential to the optimization of the ALD process and will play an important role in the 
design and nanofabrication of next generation energy devices. 
 
 
 
127 
 
Chapter 7: Conclusions and Future Work 
7.1 Conclusions 
Nature?s ability to self-assemble and self-limit is leading the way for next 
generation energy devices.  Atomic Layer Deposition is receiving responsibility for some 
of the most demanding deposition processes in the nano-community, by demonstrating its 
unique ability to process ultrathin films with unprecedented uniformity and conformality 
across 2D planar and complex 3D nanostructure topographies.  PAA has proven itself to 
be a versatile and easy platform for device design as well as nanostructure templating.  
Our recent development of an operational MIM electrostatic nanocapictor requires the 
use of high-k dielectric thin films deposited into the confines of an ultra-high aspect ratio 
PAA membrane.  Currently our work is limited by the use of Al2O3 as our high-k 
dielectric material.  It is believed that through the use of higher-k materials, such as TiO2, 
higher capacitance values can be achieved.  As no two ALD chemistries or reactors are 
the same, the deposition of TiO2 ALD required an in depth study of the growth behavior 
when transitioning from a 2D planar system to that of a 3D ultra-high aspect ratio 
nanopore in order to optimize the process conditions to achieve conformal ALD films.   
Using a commercial Beneq ALD reactor, I reemphasized the conformal behavior 
of ALD to coat the complex surface topography of a highly scalloped Al surface 
produced after the removal of the 1st anodized PAA membrane.  I demonstrated that the 
evolution of surface profiles change with the ALD layer thickness and are influenced by 
the way ALD conformally decorates the underlying topography.  Excellent agreement 
 
 
128 
 
was seen between the experimental data and the results of a simple geometric 3D 
extrusion model developed by Dr. Israel Perez confirming that ALD produces highly 
conformal films even on highly structured, sharp features of a complex nanotopography.  
The model demonstrates how the surface effectively becomes planarized with increased 
ALD film thickness, where ALD growth fronts from higher surface features dominate the 
surface profiles and eventually merge.  This work illustrates the benefits of ALD to 
conformally coat and modify surface topographies in which sharp asperities or nano-
roughness could potentially limit device performance. 
I iterated that ALD is a self-limiting and self-saturating gas phase deposition 
technique that is highly dependent on the precursor chemistries used and the interplay of 
critical process parameters, such as temperature, reactant dose, surface residence time, 
and purge time.  Using a custom built UHV ALD system; I deposited high-k TiO2 thin 
films using TTIP and ozone while employing a unique dosing scheme in order to evaluate 
the relationship between film metrics, growth rate and uniformity, and the process 
parameters: precursor?s partial pressure, surface residence time, and purge time.  Results 
demonstrated increased growth rates as a function of increased reactant surface residence 
times within the process chamber.  It was proposed that the complex reaction 
mechanisms associated with using TTIP as a precursor led to secondary reaction 
mechanisms and higher growth rates due to the production of H2O as a by-product during 
the TTIP half cycle of the ALD process.  Longer purge times confirmed that the 
increased growth was not due to physisorbed precursor molecules or parasitic CVD 
reactions.  Although the films exhibited higher growth rates, they still revealed 
 
 
129 
 
exceptionally uniform films across the wafer, further emphasizing that the ALD reaction 
was self-limiting and self-saturating. 
I proposed a method to extract film thicknesses and subsequent conformality of 
deposited ALD films within ultra-high aspect ratio nanopores through the use of EDS 
line scans.  A relationship between the deposited ALD film thickness and intensity 
profiles from EDS line scans along a cleaved PAA template was determined through a 
cylindrical assumption and the growth rate determined from planar film thicknesses.  
Film thickness profiles were compared to direct measurements from TEM images of 
released ALD based nanotubes.  EDS inferred profiles and TEM images demonstrated 
that although there was a strong depletion of film thickness down the length of the 
nanotube, films could still be accurately measured. 
I demonstrated the crucial interplay of process parameters, such as reactant dose, 
surface residence time, and purge time on the deposition of thin films in ultra-high aspect 
ratio nanopores as compared to deposition across a planar wafer.  I deposited high-k TiO2 
thin films using TTIP and ozone as precursors, while employing a unique dosing scheme 
to evaluate the relationship between film metrics, uniformity and conformality, and the 
process parameters.  Results indicate that uniform films across a 2D planar wafer do not 
translate into conformal films within a nanopore when less than optimized processes are 
used.  Several explanations were attributed to the decrease in film conformality and the 
subsequent rise of CVD-like behaviors within an ultra-high aspect ratio nanopore.  
Possible explanations include the transition from viscous to molecular flow, continually 
changing aspect ratios, and the complex reactions associated with using TTIP and ozone.  
 
 
130 
 
7.2 Future Work 
Growth rates, uniformity and conformality are to a large extent the golden 
standards to evaluate the performance of an ALD process.  We have shown that our 
cross-flow ALD system is a versatile platform to study the impact of reactant exposures, 
surface residence times, and purge times on film growth behaviors across planar 2D 
wafers as well as complex 3D nanostructures.  TiO2 ALD films have a strong potential to 
be the next generation high-k dielectric material.[141]  Throughout this work a number of 
tools and methodologies have been developed to examine the extrinsic (growth rate, 
thickness uniformity and conformality) properties of TiO2 ALD films using TTIP and 
ozone as the self-limiting precursors.  Therefore, a logical continuation of this work is to 
further the investigation of the complex reaction mechanisms associated with the use of 
the precursors TTIP and ozone through in-situ chemical analysis using mass 
spectrometry.  It is also important to characterize the intrinsic (electrical, optical, 
compositional) properties of the TiO2 films not only across planar 2D wafers but in 
complex 3D nanostructures is a key component to the next generation of MIM 
electrostatic nanocapacitors. 
 TiO2 ALD exhibits its highest dielectric constant in its rutile phase at ?k ~80.[141]  
However, as deposited TiO2 films are amorphous and require high temperature annealing 
that could potentially lead to grain growth in the high-k dielectric causing an increase in 
the leakage current, as well as potentially damage the metal electrode within an MIM 
electrostatic nanocapacitors.[141, 142]  It has been noticed that the deposition of TiO2 on 
the conductive metal Ru leads to the deposition of rutile TiO2.[125]  Ru and RuO2 are 
 
 
131 
 
both conducting materials that can both be used as the bottom electrode of an MIM 
nanocapacitor.[60, 141]  Recently, our UHV ALD reactor has been outfitted for the 
deposition of Ru and RuO ALD.  Preliminary results achieved by Keith Gregorczyk 
illustrate the complex deposition mechanisms of depositing Ru within a PAA membrane.  
Ru ALD films were deposited at 250 ?C using Ru(Cyprus)TM and molecular oxygen (O2) 
on four different pretreated substrates: Si, SiO2, Al2O3, and TiO2.  Planar results found 
that after 200 cycles of deposition, Ru was deposited on the Si, SiO2, and TiO2 films with 
a growth rate of ~0.5 ?/cycle with a very short nucleation time.  However, films were 
extremely rough as seen by SEM images of the surface.  Films deposited on Al2O3 coated 
wafers demonstrated a lengthy nucleation period of greater than 200 cycles, thus 
potentially leading to problems in using Ru as a bottom electrode in our MIM 
nanocapacitor work.  We are currently investigating the use of a TiO2 ALD nucleation 
layer under the Ru in order to push forward with the high-k rutile TiO2 creating an 
insulator-metal-insulator-metal (IMIM) nanocapacitor.  However, as with depositing any 
new chemistry the process metrics have to be optimized in order to find the optimal 
conditions for both extrinsic (growth rate, uniformity, conformality) and intrinsic 
(electrical, optical, compositional) properties for Ru ALD deposited across planar 2D 
wafers, as well as in ultra-high aspect ratio nanopores. 
  Future work in the commercial Beneq reactor encompasses pushing the limits of 
ALD conformality within ultra-high aspect ratio nanopores.  The Beneq tool has been 
touted for it unprecedented conformality within high aspect ratio structures, however, this 
process metric has only been seen for ideal ALD chemistries Al2O3 and ZnO using high 
vapor pressure sources.  Conformality studies of ZnO ALD deposited within PAA 
 
 
132 
 
membranes using diethylzinc (DEZ) and water has highlighted the complex 
adsorption/desorption mechanisms that can often be associated with using water as an 
oxidizing precursor.  Results demonstrated that longer purge times during the ALD 
process allowed for the manipulation of film deposition down the length of a nanopore 
due to water?s ability to physisorb multilayers.  It was found that longer purge times 
assist in desorption of these multilayers, thus effecting the film?s resulting film thickness 
and subsequent conformality within an ultra-high aspect ratio nanopore.  This work could 
prove to be useful in the fabrication and design of new energy devices, such as solar cells.   
Additional process optimization on the Beneq tool is being continued by Alex 
Kozen in order to further the conformality studies of ALD chemistries whose precursors 
require the use of a bubbler system due to their low vapor pressures.  However, due to the 
virtual valving system the Beneq system employs it makes changing the precursor dose 
extremely difficult.  It is has been discussed in this paper that the conformality within a 
high aspect ratio nanopore is dependent upon higher doses as well as higher surface 
residence times.  Therefore, machinery to enforce a quasistatic flow is being introduced.  
ALD conformality is vital to next generation nanotechnology and requires the process 
development and optimization of many ALD processes. 
 
 
 
133 
 
Appendices 
Appendix I.  Substrate Temperature Calibration 
Figure II. Calibration of ALD substrate heater as measured by a 13 point thermocouple 
wafer. a) Diagram of 100 mm 13 point SensArrayTM thermocouple wafer. b) Wafer 
temperature as a function of controller temperature at 100 mTorr.  c) Wafer temperature 
as a function of pressure at 175 ?C setpoint temperature. 
a) 
 
 
 
 
134 
 
b) 
 
c) 
 
 
 
 
135 
 
Appendix II.  Reactant Vapor Pressure 
Figure III.  Reactant vapor pressure vs temperture for titanium tetraisopropoxide (TTIP). 
 
Vapor pressure laws for liquid precursors (Source: Air Liquide, France) 
TTIP  log P [Torr] = 9.465 - 3222/T [K] 
        TDMAT       log P [Torr] = 8.6 ? 2850/T [K] 
        Water       log P [Torr] = 8.07 ? 1730/T [K] 
 
 
136 
 
Appendix III. Trimethylaluminum (TMA) Gas Line Delivery System 
Reproducible precursor doses of TMA were achieved by implementing a pressure end-
point control system where each precursor vapor is introduced through a needle valve 
into a temperature-controlled volume.  Upon reaching a given pressure setpoint, the inlet 
valve between the precursor and the volume was closed and the gases were released into 
the chamber for 0.3 s.  The minimum and maximum pressures were monitored for each 
cycle in order to determine the number of ?mols introduced into the system.  After each 
pulse, any gas remaining between the fill valve and the needle valve, where dumped into 
the exhaust vacuum through a bypass valve. 
Figure IV.  Record of maximum and minimum pressures in TMA delivery volume over 
100 cycles.  Calculated doses (open diamonds) are shown on the right axis. 
 
 
 
 
137 
 
Appendix IV.  Calibration of Ozone Dose 
Figure V. Calibration of Ozone does is only controlled by a needle valve and exposure 
time.  Compared the pressure increase in the mini-reactor (PRGA) during known TMA 
doses and O3 pulses to establish exact O3 does in mols (done in conditions where no 
chemical reactions occurred). Calculated integrated peak for PRGA under different known 
TMA doses to establish regression. 
 
 
 
 
138 
 
Appendix V. HfO2 ALD Nanotube Fabrication[44] 
A porous AAO template was synthesized by a two-step anodization method.[79]. 
In brief, a piece of electropolished aluminum foil (99.99%, Alfa Aesar) was anodized at 
40V and 10?C in an electrolytic bath of 0.3M oxalic acid. In the first step of the 
anodization process, the foil was anodized for long enough (~ 7 hours) so that the pores 
were ordered and growing orthogonally with respect to the substrate. The AAO film that 
was created was then etched off by an aqueous mixture of phosphoric acid (6 wt %) and 
chromic acid (1.8 wt %), leaving a pre-textured and ordered aluminum surface. In the 
second step of the anodization process, the foil was anodized for a set amount of time (~1 
hour) defining the depth of the pores to be ~4um. Pore diameters were adjusted to ~60 
nm using a pore-widening solution of phosphoric acid (0.1 M) at 38 ?C.  Membranes 
were not detached from the substrate before ALD processing and therefore its nanopores 
only had one open end. 
High-K dielectric HfO2 thin films were deposited by alternating reactant 
exposures of tetrakis(ethylmethylamino)hafnium (TEMAH), that is, Hf[N(CH3)(C2H5)]4 
[99.99% grade, Sigma Aldrich], as the organometallic precursor and DI water as an 
oxidant.  The ALD equipment consists of a stainless steel tube (25? in length and 3? in 
diameter) in a tube furnace.  Precursors and nitrogen gases were introduced through one 
end of the reactor, while un-reacted precursors and reaction byproducts were exhausted at 
the opposite end of the tube to a rotary vane vacuum pump.  Substrates were introduced 
into the system through a removable flange at the exhaust end of the reactor.  The wafer 
temperature was measured via a thermocouple to be 200 ?C.  During each of the self-
 
 
139 
 
limited half reactions corresponding to the alternating exposures of TEMAH and water, 
reactants were dosed to achieve full surface saturation resulting in excellent thickness 
control at the atomic level as a sub-monolayer of oxide was consistently deposited for 
each cycle. 
A reservoir containing 5 g of liquid TEMAH was placed in an incubation oven set 
to 105 ?C to achieve a vapor pressure of approximately 2 Torr.  Prior to each exposure, 
the TEMAH gas was fed into a 20 mL isothermal volume and TEMAH doses were 
controlled by timing the opening of a pneumatically actuated valve.  A 0.85 ?mol dose of 
TEMAH was determined to be sufficient to achieve full saturation on a Si substrate 
located in next to the AAO template.  In the case of water, the vessel was kept at room 
temperature.  Accurate water dosage was achieved by filling up a 20 mL isothermal 
volume up to a targeted pressure monitored by a capacitance gauge and then discharging 
the gas into the reactor for a fixed amount of time. This pressure end-point control 
approach was required to minimize the effect of the cooling of the water source over the 
process duration due to the forced vaporization of the precursor and the resulting 
decrease of its vapor pressure.  Full saturation was achieved by filling up the volume at 7 
Torr.  We estimate the partial pressures of both the water and TEMAH doses to be ~7 
mTorr from the pressure spike recorded by a downstream pressure sensor (average 
change in pressure over peak width). 
Throughout the deposition process, 30 sccm of nitrogen was continuously flowed 
into the reactor maintaining a reactor pressure of 96 mTorr.  A growth rate of 1 ? cycle-1 
was achieved over 100 cycle process.  Prior to each pulse, a 10 s N2 purge was initiated to 
ensure that un-reacted precursor and reaction byproduct, mainly (CH3)(C2H5)NH, were 
 
 
140 
 
adequately purge out of the reactor. 
After ALD deposition, samples were place in 0.1M NaOH solution for 1 hour 
which dissolved away the alumina membrane freeing the HfO2 film in solution.  After 
filtration with DI water, a drop of the solution was placed on a carbon coated Cu TEM 
grid.  
 
 
141 
 
 
References 
[1]  M. Leskela and M. Ritala, "Atomic layer deposition (ALD): from precursors to 
thin film structures," Thin Solid Films, vol. 409, pp. 138-146, 2002. 
[2]  L. Niinisto, J. Paivasaari, J. Niinisto, M. Putkonen, and M. Nieminen, "Advanced 
electronic and optoelectronic materials by Atomic Layer Deposition: An overview 
with special emphasis on recent progress in processing of high-k dielectrics and 
other oxide materials," Physica Status Solidi a-Applied Research, vol. 201, pp. 
1443-1452, 2004. 
[3]  H. Kim, H. B. R. Lee, and W. J. Maeng, "Applications of atomic layer deposition 
to nanofabrication and emerging nanodevices," Thin Solid Films, vol. 517, pp. 
2563-2580, 2009. 
[4]  R. L. Puurunen, "Surface chemistry of atomic layer deposition: A case study for 
the trimethylaluminum/water process," Journal of Applied Physics, vol. 97, 2005. 
[5]  M. Leskela, M. Kemell, K. Kukli, V. Pore, E. Santala, M. Ritala, and J. Lu, 
"Exploitation of atomic layer deposition for nanostructured materials," Materials 
Science & Engineering C-Biomimetic and Supramolecular Systems, vol. 27, pp. 
1504-1508, 2007. 
[6]  O. Sneh, M. A. Cameron, and S. M. George, "Adsorption and desorption kinetics 
of H2O on a fully hydroxylated SiO2 surface," Surface Science, vol. 364, pp. 61-
78, 1996. 
[7]  O. Sneh, R. B. Clark-Phelps, A. R. Londergan, J. Winkler, and T. E. Seidel, 
"Thin film atomic layer deposition equipment for semiconductor processing," 
Thin Solid Films, vol. 402, pp. 248-261, 2002. 
[8]  G. D. Wilk, R. M. Wallace, and J. M. Anthony, "High-kappa gate dielectrics: 
Current status and materials properties considerations," Journal of Applied 
Physics, vol. 89, pp. 5243-5275, 2001. 
[9]  E. Gerritsen, N. Emonet, C. Caillat, N. Jourdan, M. Piazza, D. Fraboulet, B. 
Boeck, A. Berthelot, S. Smith, and P. Mazoyer, "Evolution of materials 
technology for stacked-capacitors in 65 nm embedded-DRAM," Solid-State 
Electronics, vol. 49, pp. 1767-1775, 2005. 
[10]  M. Leskela and M. Ritala, "Atomic layer deposition chemistry: Recent 
developments and future challenges," Angewandte Chemie-International Edition, 
vol. 42, pp. 5548-5554, 2003. 
[11]  M. Schumacher, P. K. Baumann, and T. Seidel, "AVD and ALD as two 
complementary technology solutions for next generation dielectric and conductive 
thin-film processing," Chemical Vapor Deposition, vol. 12, pp. 99-108, 2006. 
[12]  M. Knez, K. Niesch, and L. Niinisto, "Synthesis and surface engineering of 
complex nanostructures by atomic layer deposition," Advanced Materials, vol. 19, 
pp. 3425-3438, 2007. 
[13]  C. R. Manoj and V. R. Rao, "Impact of high-k gate dielectrics on the device and 
circuit performance of nanoscale FinFETs," Ieee Electron Device Letters, vol. 28, 
 
 
142 
 
pp. 295-297, 2007. 
[14]  S. K. Kim, G. J. Choi, J. H. Kim, and C. S. Hwang, "Growth behavior of Al-
doped TiO2 thin films by atomic layer deposition," Chemistry of Materials, vol. 
20, pp. 3723-3727, 2008. 
[15]  D. M. King, X. H. Du, A. S. Cavanagh, and A. Weimer, "Quantum confinement 
in amorphous TiO2 films studied via atomic layer deposition," Nanotechnology, 
vol. 19, p. 6, Nov 2008. 
[16]  J. S. King, D. Heineman, E. Graugnard, and C. J. Summers, "Atomic layer 
deposition in porous structures: 3D photonic crystals," Applied Surface Science, 
vol. 244, pp. 511-516, 2005. 
[17]  M. Kemell, V. Pore, M. Ritala, M. Leskela, and M. Linden, "Atomic layer 
deposition in nanometer-level replication of cellulosic substances and preparation 
of photocatalytic TiO2/cellulose composites," Journal of the American Chemical 
Society, vol. 127, pp. 14178-14179, 2005. 
[18]  M. J. Pellin, P. C. Stair, G. Xiong, J. W. Elam, J. Birrell, L. Curtiss, S. M. 
George, C. Y. Han, L. Iton, H. Kung, M. Kung, and H. H. Wang, "Mesoporous 
catalytic membranes: Synthetic control of pore size and wall composition," 
Catalysis Letters, vol. 102, pp. 127-130, 2005. 
[19]  E. Santala, M. Kemell, M. Leskela, and M. Ritala, "The preparation of reusable 
magnetic and photocatalytic composite nanofibers by electrospinning and atomic 
layer deposition," Nanotechnology, vol. 20, 2009. 
[20] P. Banerjee, Perez, I., Robertson, E., Lee, S.B., Rubloff, G.W., "Lateral two 
terminal nanotube devices and method for their formation,"  U.S., 2007. 
[21]  P. Banerjee, Perez, I., Henn-Lecordier, L., Lee, S.B., Rubloff, G.W., 
"Nanotubular metal-insulator-metal capacitor arrays for energy storage," Nature 
Nanotechnology, vol. 4, pp. 292-296, 2009. 
[22]  P. Banerjee, Perez, I., Henn-Lecordier, L., Lee, S.B., Rubloff, G.W., "ALD based 
metal-insulator-metal (MIM) nanocapacitors for energy storage," ECS 
Transactions, 2009. 
[23]  C. Nistorica, J. F. Liu, I. Gory, G. D. Skidmore, F. M. Mantiziba, B. E. Gnade, 
and J. Kim, "Tribological and wear studies of coatings fabricated by atomic layer 
deposition and by successive ionic layer adsorption and reaction for 
microelectromechanical devices," Journal of Vacuum Science & Technology A, 
vol. 23, pp. 836-840, 2005. 
[24]  C. F. Herrmann, F. W. DelRio, D. C. Miller, S. M. George, V. M. Bright, J. L. 
Ebel, R. E. Strawser, R. Cortez, and K. D. Leedy, "Alternative dielectric films for 
rf MEMS capacitive switches deposited using atomic layer deposited Al2O3/ZnO 
alloys," Sensors and Actuators a-Physical, vol. 135, pp. 262-272, 2007. 
[25]  T. M. Mayer, J. W. Elam, S. M. George, P. G. Kotula, and R. S. Goeke, "Atomic-
layer deposition of wear-resistant coatings for microelectromechanical devices," 
Applied Physics Letters, vol. 82, pp. 2883-2885, 2003. 
[26]  N. D. Hoivik, J. W. Elam, R. J. Linderman, V. M. Bright, S. M. George, and Y. 
C. Lee, "Atomic layer deposited protective coatings for micro-electromechanical 
systems," Sensors and Actuators a-Physical, vol. 103, pp. 100-108, 2003. 
[27]  J. F. Liu, C. Nistorica, I. Gory, G. Skidmore, F. M. Mantiziba, and B. E. Gnade, 
"Layer-by-layer deposition of zirconium oxide films from aqueous solutions for 
 
 
143 
 
friction reduction in silicon-based micro electromechanical system devices," Thin 
Solid Films, vol. 492, pp. 6-12, 2005. 
[28]  A. Javey, H. Kim, M. Brink, Q. Wang, A. Ural, J. Guo, P. McIntyre, P. McEuen, 
M. Lundstrom, and H. J. Dai, "High-kappa dielectrics for advanced carbon-
nanotube transistors and logic gates," Nature Materials, vol. 1, pp. 241-246, 2002. 
[29]  J. S. Lee, B. Min, K. Cho, S. Kim, J. Park, Y. T. Lee, N. S. Kim, M. S. Lee, S. O. 
Park, and J. T. Moon, "Al2O3 nanotubes and nanorods fabricated by coating and 
filling of carbon nanotubes with atomic-layer deposition," Journal of Crystal 
Growth, vol. 254, pp. 443-448, 2003. 
[30]  R. Artzi-Gerlitz, K. D. Benkstein, D. L. Lahr, J. L. Hertz, C. B. Montgomery, J. 
E. Bonevich, S. Semancik, and M. J. Tarlov, "Fabrication and gas sensing 
performance of parallel assemblies of metal oxide nanotubes supported by porous 
aluminum oxide membranes," Sensors and Actuators B-Chemical, vol. 136, pp. 
257-264, 2009. 
[31]  M. A. Cameron, I. P. Gartland, J. A. Smith, S. F. Diaz, and S. M. George, 
"Atomic layer deposition of SiO2 and TiO2 in alumina tubular membranes: Pore 
reduction and effect of surface species on gas transport," Langmuir, vol. 16, pp. 
7435-7444, 2000. 
[32]  J. D. Ferguson, A. W. Weimer, and S. M. George, "Atomic layer deposition of 
ultrathin and conformal Al2O3 films on BN particles," Thin Solid Films, vol. 371, 
pp. 95-104, 2000. 
[33]  D. M. King, X. H. Liang, Y. Zhou, C. S. Carney, L. F. Hakim, P. Li, and A. W. 
Weimer, "Atomic layer deposition of TiO2 films on particles in a fluidized bed 
reactor," Powder Technology, vol. 183, pp. 356-363, 2008. 
[34]  J. D. Ferguson, A. W. Weimer, and S. M. George, "Atomic layer deposition of 
SiO2 films on BN particles using sequential surface reactions," Chemistry of 
Materials, vol. 12, pp. 3472-3480, 2000. 
[35]  J. D. Ferguson, A. W. Weimer, and S. M. George, "Atomic layer deposition of 
Al2O3 films on polyethylene particles," Chemistry of Materials, vol. 16, pp. 
5602-5609, 2004. 
[36]  J. D. Ferguson, A. R. Yoder, A. W. Weimer, and S. M. George, "TiO2 atomic 
layer deposition on ZrO2 particles using alternating exposures of TiCl4 and 
H2O," Applied Surface Science, vol. 226, pp. 393-404, 2004. 
[37]  A. Johansson, T. Torndahl, L. M. Ottosson, M. Boman, and J. O. Carlsson, 
"Copper nanoparticles deposited inside the pores of anodized aluminium oxide 
using atomic layer deposition," Materials Science & Engineering C-Biomimetic 
and Supramolecular Systems, vol. 23, pp. 823-826, 2003. 
[38]  A. Johansson, J. Lu, J. O. Carlsson, and M. Boman, "Deposition of palladium 
nanoparticles on the pore walls of anodic alumina using sequential electroless 
deposition," Journal of Applied Physics, vol. 96, pp. 5189-5194, 2004. 
[39]  J. Biener, T. F. Baumann, Y. M. Wang, E. J. Nelson, S. O. Kucheyev, A. V. 
Hamza, M. Kemell, M. Ritala, and M. Leskela, "Ruthenium/aerogel 
nanocomposites via atomic layer deposition," Nanotechnology, vol. 18, 2007. 
[40]  R. G. Gordon, J. Becker, D. Hausmann, and S. Suh, "Vapor deposition of metal 
oxides and silicates: Possible gate insulators for future microelectronics," 
Chemistry of Materials, vol. 13, pp. 2463-2464, 2001. 
 
 
144 
 
[41]  A. W. Ott, J. W. Klaus, J. M. Johnson, S. M. George, K. C. McCarley, and J. D. 
Way, "Modification of porous alumina membranes using Al2O3 atomic layer 
controlled deposition," Chemistry of Materials, vol. 9, pp. 707-714, 1997. 
[42]  R. L. Puurunen, W. Vandervorst, W. F. A. Besling, O. Richard, H. Bender, T. 
Conard, C. Zhao, A. Delabie, M. Caymax, S. De Gendt, M. Heyns, M. M. 
Viitanen, M. de Ridder, H. H. Brongersma, Y. Tamminga, T. Dao, T. de Win, M. 
Verheijen, M. Kaiser, and M. Tuominen, "Island growth in the atomic layer 
deposition of zirconium oxide and aluminum oxide on hydrogen-terminated 
silicon: Growth mode modeling and transmission electron microscopy," Journal 
of Applied Physics, vol. 96, pp. 4878-4889, 2004. 
[43]  D. R. G. Mitchell, G. Triani, D. J. Attard, K. S. Finnie, P. J. Evans, C. J. Barbe, 
and J. R. Bartlett, "Atomic layer deposition of TiO2 and Al2O3 thin films and 
nanolaminates," Smart Materials & Structures, vol. 15, pp. S57-S64, 2006. 
[44]  I. Perez, E. Robertson, P. Banerjee, L. Henn-Lecordier, S. J. Son, S. B. Lee, and 
G. W. Rubloff, "TEM-based metrology for HfO2 layers and nanotubes formed in 
anodic aluminum oxide nanopore structures," Small, vol. 4, pp. 1223-1232, 2008. 
[45]  L. Nyns, A. Delabie, M. Caymax, M. M. Heyns, S. Van Elshocht, C. Vinckier, 
and S. De Gendt, "HfO2 Atomic Layer Deposition Using HfCl4/H2O: The First 
Reaction Cycle," Journal of the Electrochemical Society, vol. 155, pp. G269-
G273, 2008. 
[46]  W. J. Maeng and H. Kim, "Electrical properties of atomic layer deposition HfO2 
and HfOxNy on si substrates with various crystal orientations," Journal of the 
Electrochemical Society, vol. 155, pp. H267-H271, 2008. 
[47]  X. Y. Liu, S. Ramanathan, A. Longdergan, A. Srivastava, E. Lee, T. E. Seidel, J. 
T. Barton, D. Pang, and R. G. Gordon, "AID of hafnium oxide thin film from 
tetrakis(ethylmethylamino)hafnium and ozone," Journal of the Electrochemical 
Society, vol. 152, pp. G213-G219, 2005. 
[48]  W. J. Maeng and H. Kim, "Thermal and plasma-enhanced ALD of Ta and Ti 
oxide thin films from alkylamide precursors," Electrochemical and Solid State 
Letters, vol. 9, pp. G191-G194, 2006. 
[49]  M. Kemell, E. Harkonen, V. Pore, M. Ritala, and M. Leskela, "Ta2O5- and 
TiO2-based nanostructures made by atomic layer deposition," Nanotechnology, 
vol. 21. 
[50]  J. W. Elam, J. A. Libera, M. J. Pellin, A. V. Zinovev, J. P. Greene, and J. A. 
Nolen, "Atomic layer deposition of W on nanoporous carbon aerogels," Applied 
Physics Letters, vol. 89, 2006. 
[51]  R. K. Grubbs, C. E. Nelson, N. J. Steinmetz, and S. M. George, "Nucleation and 
growth during the atomic layer deposition of W on Al2O3 and Al2O3 on W," 
Thin Solid Films, vol. 467, pp. 16-27, 2004. 
[52]  W. Lei, L. Henn-Lecordier, M. Anderle, G. W. Rubloff, M. Barozzi, and M. 
Bersani, "Real-time observation and optimization of tungsten atomic layer 
deposition process cycle," Journal of Vacuum Science & Technology B, vol. 24, 
pp. 780-789, 2006. 
[53]  J. W. Elam, C. E. Nelson, R. K. Grubbs, and S. M. George, "Nucleation and 
growth during tungsten atomic layer deposition on SiO2 surfaces," Thin Solid 
Films, vol. 386, pp. 41-52, 2001. 
 
 
145 
 
[54]  A. Rahtu and M. Ritala, "Reaction mechanism studies on titanium isopropoxide-
water atomic layer deposition process," Chemical Vapor Deposition, vol. 8, pp. 
21-28, 2002. 
[55]  A. Palma and A. Alavi, "An ab initio study of titanium tetra-iso-propoxide 
(TTIP) adsorption mechanism on a Si(100) surface," Computational Materials 
Science, vol. 33, pp. 244-249, 2005. 
[56]  V. R. Rai and S. Agarwal, "Surface reaction mechanisms during ozone-based 
atomic layer deposition of titanium dioxide," Journal of Physical Chemistry C, 
vol. 112, pp. 9552-9554, 2008. 
[57]  V. R. Rai and S. Agarwal, "Surface Reaction Mechanisms during Plasma-
Assisted Atomic Layer Deposition of Titanium Dioxide," Journal of Physical 
Chemistry C, vol. 113, pp. 12962-12965, Jul 2009. 
[58]  C. Ducso, N. Q. Khanh, Z. Horvath, I. Barsony, M. Utriainen, S. Lehto, M. 
Nieminen, and L. Niinisto, "Deposition of tin oxide into porous silicon by atomic 
layer epitaxy," Journal of the Electrochemical Society, vol. 143, pp. 683-687, 
1996. 
[59]  S. K. Kim, C. S. Hwang, S. H. K. Park, and S. J. Yun, "Comparison between 
ZnO films grown by atomic layer deposition using H2O or O-3 as oxidant," Thin 
Solid Films, vol. 478, pp. 103-108, 2005. 
[60]  J. H. Kim, D. S. Kil, S. J. Yeom, J. S. Roh, N. J. Kwak, and J. W. Kim, 
"Modified atomic layer deposition of RuO2 thin films for capacitor electrodes," 
Applied Physics Letters, vol. 91, 2007. 
[61]  J. W. Elam, A. Zinovev, C. Y. Han, H. H. Wang, U. Welp, J. N. Hryn, and M. J. 
Pellin, "Atomic layer deposition of palladium films on Al2O3 surfaces," Thin 
Solid Films, vol. 515, pp. 1664-1673, 2006. 
[62]  A. C. Jones, H. C. Aspinall, P. R. Chalker, R. J. Potter, T. D. Manning, Y. F. 
Loo, R. O'Kane, J. M. Gaskell, and L. M. Smith, "MOCVD and ALD of high-
kappa dielectric oxides using alkoxide precursors," Chemical Vapor Deposition, 
vol. 12, pp. 83-98, 2006. 
[63]  S. M. George, "Atomic Layer Deposition: An Overview," Chemical Reviews, 
vol. 110, pp. 111-131, Jan. 
[64]  M. Ylilammi, "Monolayer thickness in atomic layer deposition," Thin Solid 
Films, vol. 279, pp. 124-130, 1996. 
[65]  M. Y. Li, Y. Y. Chang, H. C. Wu, C. S. Huang, J. C. Chen, J. L. Lue, and S. M. 
Chang, "Effect of process pressure on atomic layer deposition of Al2O3," Journal 
of the Electrochemical Society, vol. 154, pp. H967-H972, 2007. 
[66]  R. G. Gordon, D. Hausmann, E. Kim, and J. Shepard, "A kinetic model for step 
coverage by atomic layer deposition in narrow holes or trenches," Chemical 
Vapor Deposition, vol. 9, pp. 73-78, 2003. 
[67]  J. W. Elam, D. Routkevitch, P. P. Mardilovich, and S. M. George, "Conformal 
coating on ultrahigh-aspect-ratio nanopores of anodic alumina by atomic layer 
deposition," Chemistry of Materials, vol. 15, pp. 3507-3517, 2003. 
[68]  L. Henn-Lecordier, M. Anderle, E. Robertson, and G. W. Rubloff, "Exploring the 
limits of atomic layer deposition: the process window for Al2O3 ALD from TMA 
and H2O," Chemical Vapor Deposition, submitted. 
[69] J. F. O'Hanlon, A User's Guide to Vacuum Technology, 2nd ed. New York: John 
 
 
146 
 
Wiley & Sons, 1989. 
[70]  J. W. Diggle, T. C. Downie, and C. W. Goulding, "ANODIC OXIDE FILMS ON 
ALUMINUM," Chemical Reviews, vol. 69, pp. 365-&, 1969. 
[71]  G. E. Thompson, "Porous anodic alumina: Fabrication, characterization and 
applications," Thin Solid Films, vol. 297, pp. 192-201, 1997. 
[72]  H. Masuda and K. Fukuda, "Ordered Metal Nanohole Arrays Made By a 2-step 
Replication of Honeycomb Structures of Anodic Alumina," Science, vol. 268, pp. 
1466-1468, 1995. 
[73]  W. Lee, K. Nielsch, and U. Gosele, "Self-ordering behavior of nanoporous 
anodic aluminum oxide (AAO) in malonic acid anodization," Nanotechnology, 
vol. 18, Nov 2007. 
[74]  A. P. Li, F. Muller, A. Birner, K. Nielsch, and U. Gosele, "Hexagonal pore arrays 
with a 50-420 nm interpore distance formed by self-organization in anodic 
alumina," Journal of Applied Physics, vol. 84, pp. 6023-6026, 1998. 
[75]  K. S. Kang, Y. S. Meng, J. Breger, C. P. Grey, and G. Ceder, "Electrodes with 
high power and high capacity for rechargeable lithium batteries," Science, vol. 
311, pp. 977-980, 2006. 
[76]  C. Liu, F. Li, L. P. Ma, and H. M. Cheng, "Advanced Materials for Energy 
Storage," Advanced Materials, vol. 22, pp. E28-+. 
[77]  R. Liu, S. Il Cho, and S. B. Lee, "Poly( 3,4-ethylenedioxythiophene) nanotubes 
as electrode materials for a high-powered supercapacitor," Nanotechnology, vol. 
19, 2008. 
[78]  L. Henn-Lecordier, W. Lei, M. Anderle, and G. W. Rubloff, "Real-time sensing 
and metrology for atomic layer deposition processes and manufacturing," Journal 
of Vacuum Science & Technology B, vol. 25, pp. 130-139, 2007. 
[79]  H. Masuda and M. Satoh, "Fabrication of gold nanodot array using anodic porous 
alumina as an evaporation mask," Japanese Journal of Applied Physics Part 2-
Letters, vol. 35, pp. L126-L129, Jan 1996. 
[80]  E. Langereis, S. B. S. Heil, H. C. M. Knoops, W. Keuning, M. C. M. van de 
Sanden, and W. M. M. Kessels, "In situ spectroscopic ellipsometry as a versatile 
tool for studying atomic layer deposition," Journal of Physics D-Applied Physics, 
vol. 42, 2009. 
[81] J.A. Woollam Co., Inc., http://www.jawoollam.com/tutorial_1.html, March 20, 
2010 
[82] H. G. Tompkins and W. A. McGahan, Spectroscopic ellipsometry and 
reflectometry, a user's guide: Wiley-Interscience, 1999. 
[83] Trinity College School of Physics, 
http://www.tcd.ie/Physics/Surfaces/ellipsometry2.php, March 20, 2010 
[84] W. Vanderlinde, "SEM Short Course: Scanning Electron Microscopy,"  College 
Park: Laboratory for Physical Sciences, 2008. 
[85] R. Egerton, Physical Principles of Electron Microscopy: An Introduction to TEM, 
SEM, AEM. New York: Springer, 2007. 
[86] P. Flewitt and R. Wild, Physical Methods For Materials Characterization. 
Philadelphia: IOP Publishing, 2003. 
[87] Purdu University, http://www.purdue.edu/REM/rs/sem.htm, March 30, 2010 
[88] University of Nebraska-Lincoln, http://www.unl.edu/CMRAcfem/temoptic.htm, 
 
 
147 
 
March 30, 2010 
[89] W. Vanderlinde, "SEM Short Course: Energy Dispersive X-ray Analysis," 
Laboratory for Physical Sciences, 2008. 
[90] Oxford Instruments PLC, http://www.x-raymicroanalysis.com/x-ray-
microanalysis-explained/pages/tutorial3/interaction.htm, 2010 
[91] Galloway, "Atomic Force Microscopy: A Guide to Understanding and Using the 
AFM,"  2004. 
[92] Wikipedia, http://en.wikipedia.org/wiki/Atomic_force_microscopy, March 31, 
2010 
[93]  R. Liu and S. B. Lee, "MnO2/Poly(3,4-ethylenedioxythiophene) coaxial 
nanowires by one-step coelectrodeposition for electrochemical energy storage," 
Journal of the American Chemical Society, vol. 130, pp. 2942-2943, 2008. 
[94]  J. W. Elam, G. Xiong, C. Y. Han, H. H. Wang, J. P. Birrell, U. Welp, J. N. Hryn, 
M. J. Pellin, T. F. Baumann, J. F. Poco, and J. H. Satcher, "Atomic layer 
deposition for the conformal coating of nanoporous materials," Journal of 
Nanomaterials, 2006. 
[95]  R. Rosenberg, D. C. Edelstein, C. K. Hu, and K. P. Rodbell, "Copper 
metallization for high performance silicon technology," Annual Review of 
Materials Science, vol. 30, pp. 229-262, 2000. 
[96]  S. M. Gates, "Surface chemistry in the chemical vapor deposition of electronic 
materials," Chemical Reviews, vol. 96, pp. 1519-1532, 1996. 
[97]  H. Masuda, H. Yamada, M. Satoh, H. Asoh, M. Nakao, and T. Tamamura, 
"Highly ordered nanochannel-array architecture in anodic alumina," Applied 
Physics Letters, vol. 71, pp. 2770-2772, 1997. 
[98]  A. B. F. Martinson, J. W. Elam, J. T. Hupp, and M. J. Pellin, "ZnO nanotube 
based dye-sensitized solar cells ZnO nanotube based dye-sensitized solar cells," 
Nano Letters, vol. 7, pp. 2183-2187, 2007. 
[99]  Y. K. Zhou, C. M. Shen, and H. L. Li, "Synthesis of high-ordered LiCoO2 
nanowire arrays by AAO template," Solid State Ionics, vol. 146, pp. 81-86, 2002. 
[100]  H. Masuda, F. Hasegwa, and S. Ono, "Self-ordering of cell arrangement of 
anodic porous alumina formed in sulfuric acid solution," Journal of the 
Electrochemical Society, vol. 144, pp. L127-L130, 1997. 
[101]  M. K. Gobbert, V. Prasad, and T. S. Cale, "Predictive modeling of atomic layer 
deposition on the feature scale," Thin Solid Films, vol. 410, pp. 129-141, 2002. 
[102]  H. Asoh, K. Nishio, M. Nakao, T. Tamamura, and H. Masuda, "Conditions for 
fabrication of ideally ordered anodic porous alumina using pretextured Al," 
Journal of the Electrochemical Society, vol. 148, pp. B152-B156, 2001. 
[103]  W. Lee, R. Ji, U. Gosele, and K. Nielsch, "Fast fabrication of long-range ordered 
porous alumina membranes by hard anodization," Nature Materials, vol. 5, pp. 
741-747, 2006. 
[104]  Q. Xie, J. Musschoot, D. Deduytsche, R. L. Van Meirhaeghe, C. Detavernier, S. 
Van den Berghe, Y. L. Jiang, G. P. Ru, B. Z. Li, and X. P. Qu, "Growth kinetics 
and crystallization behavior of TiO2 films prepared by plasma enhanced atomic 
layer deposition," Journal of the Electrochemical Society, vol. 155, pp. H688-
H692, 2008. 
[105]  Q. Xie, Y. L. Jiang, C. Detavernier, D. Deduytsche, R. L. Van Meirhaeghe, G. P. 
 
 
148 
 
Ru, B. Z. Li, and X. P. Qu, "Atomic layer deposition of TiO2 from tetrakis-
dimethyl-amido titanium or Ti isopropoxide precursors and H2O," Journal of 
Applied Physics, vol. 102, 2007. 
[106]  A. Niskanen, K. Arstila, M. Leskela, and M. Ritala, "Radical enhanced atomic 
layer deposition of titanium dioxide," Chemical Vapor Deposition, vol. 13, pp. 
152-157, 2007. 
[107]  L. K. Tan, H. Gao, Y. Zong, and W. Knoll, "Atomic Layer Deposition of TiO2 to 
Bond Free-Standing Nanoporous Alumina Templates to Gold-Coated Substrates 
as Planar Optical Waveguide Sensors," Journal of Physical Chemistry C, vol. 
112, pp. 17576-17580, 2008. 
[108]  I. Jogi, K. Kukli, J. Aarik, A. Aidla, and J. Lu, "Precursor-dependent structural 
and electrical characteristics of atomic layer deposited films: Case study on 
titanium oxide," Materials Science in Semiconductor Processing, vol. 9, pp. 1084-
1089, 2006. 
[109]  S. Duenas, H. Castan, H. Garcia, E. San Andres, M. Toledano-Luque, I. Martil, 
G. Gonzalez-Diaz, K. Kukli, T. Uustare, and J. Aarik, "A comparative study of 
the electrical properties of TiO2 films grown by high-pressure reactive sputtering 
and atomic layer deposition," Semiconductor Science and Technology, vol. 20, pp. 
1044-1051, 2005. 
[110]  G. A. Battiston, R. Gerbasi, A. Gregori, M. Porchia, S. Cattarin, and G. A. Rizzi, 
"PECVD of amorphous TiO2 thin films: effect of growth temperature and plasma 
gas composition," Thin Solid Films, vol. 371, pp. 126-131, 2000. 
[111]  N. G. Kubala, P. C. Rowlette, and C. A. Wolden, "Self-Limiting Deposition of 
Anatase TiO2 at Low Temperature by Pulsed PECVD," Electrochemical and 
Solid State Letters, vol. 12, pp. H259-H262, 2009. 
[112]  M. Ritala, M. Leskela, L. Niinisto, and P. Haussalo, "TITANIUM 
ISOPROPOXIDE AS A PRECURSOR IN ATOMIC LAYER EPITAXY OF 
TITANIUM-DIOXIDE THIN-FILMS," Chemistry of Materials, vol. 5, pp. 1174-
1181, 1993. 
[113]  J. Aarik, A. Aidla, T. Uustare, M. Ritala, and M. Leskela, "Titanium 
isopropoxide as a precursor for atomic layer deposition: characterization of 
titanium dioxide growth process," Applied Surface Science, vol. 161, pp. 385-395, 
2000. 
[114]  K. Bange, C. R. Ottermann, O. Anderson, U. Jeschkowski, M. Laube, and R. 
Feile, "INVESTIGATIONS OF TIO2 FILMS DEPOSITED BY DIFFERENT 
TECHNIQUES," Thin Solid Films, vol. 197, pp. 279-285, 1991. 
[115]  G. T. Lim and D. H. Kim, "Characteristics of TiOx films prepared by chemical 
vapor deposition using tetrakis-dimethyl-amido-titanium and water," Thin Solid 
Films, vol. 498, pp. 254-258, 2006. 
[116]  J. H. Kwon, M. Dai, M. D. Halls, and Y. J. Chabal, "Detection of a formate 
surface intermediate in the atomic layer deposition of high-k dielectrics using 
ozone," Chemistry of Materials, vol. 20, pp. 3248-3250, 2008. 
[117]  R. Katamreddy, R. Inman, G. Jursich, A. Soulet, and C. Takoudis, "Atomic layer 
deposition of HfO2, Al2O3, and HfAlOx using O-3 and metal(diethylamino) 
precursors," Journal of Materials Research, vol. 22, pp. 3455-3464, 2007. 
[118]  S. K. Kim, S. Y. Lee, M. Seo, G. J. Choi, and C. S. Hwang, "Impact of O-3 
 
 
149 
 
feeding time on TiO2 films grown by atomic layer deposition for memory 
capacitor applications," Journal of Applied Physics, vol. 102, 2007. 
[119]  C. R. Ottermann and K. Bange, "Correlation between the density of TiO2 films 
and their properties," Thin Solid Films, vol. 286, pp. 32-34, 1996. 
[120]  A. W. Ott, J. W. Klaus, J. M. Johnson, and S. M. George, "Al3O3 thin film 
growth on Si(100) using binary reaction sequence chemistry," Thin Solid Films, 
vol. 292, pp. 135-144, 1997. 
[121]  R. Matero, A. Rahtu, M. Ritala, M. Leskela, and T. Sajavaara, "Effect of water 
dose on the atomic layer deposition rate of oxide thin films," Thin Solid Films, 
vol. 368, pp. 1-7, 2000. 
[122]  M. Rose, J. Niinisto, I. Endler, J. W. Bartha, P. Kucher, and M. Ritala, "In Situ 
Reaction Mechanism Studies on Ozone-Based Atomic Layer Deposition of 
Al2O3 and HfO2," Acs Applied Materials & Interfaces, vol. 2, pp. 347-350. 
[123]  D. N. Goldstein, J. A. McCormick, and S. M. George, "Al2O3 Atomic Layer 
Deposition with Trimethylaluminum and Ozone Studied by in Situ Transmission 
FTIR Spectroscopy and Quadrupole Mass Spectrometry," Journal of Physical 
Chemistry C, vol. 112, pp. 19530-19539, 2008. 
[124] E. Langereis, W. Keuning, S. Rushworth, A.Kingsley, F. Roozeboom, M. C. M. 
v. d. Sanden, and W. M. M. Kessels, "Remote plasma ALD of TiO2 and SrTiO3: 
Ti precursor evaluation and substrate temperature dependence," in ALD AVS 
Monterey, CA, 2009. 
[125]  S. K. Kim, W. D. Kim, K. M. Kim, C. S. Hwang, and J. Jeong, "High dielectric 
constant TiO2 thin films on a Ru electrode grown at 250 degrees C by atomic-
layer deposition," Applied Physics Letters, vol. 85, pp. 4112-4114, 2004. 
[126]  H. S. Park, J. S. Min, J. W. Lim, and S. W. Kang, "Theoretical evaluation of film 
growth rate during atomic layer epitaxy," Applied Surface Science, vol. 158, pp. 
81-91, 2000. 
[127]  A. Rahtu, M. Ritala, and M. Leskela, "Atomic layer deposition of zirconium 
titanium oxide from titanium isopropoxide and zirconium chloride," Chemistry of 
Materials, vol. 13, pp. 1528-1532, 2001. 
[128]  Y. Gao, "In-situ IR and spectroscopic ellipsometric analysis of growth process 
and structural properties of Ti1-xNbxO2 thin films by metal-organic chemical 
vapor deposition," Thin Solid Films, vol. 346, pp. 73-81, 1999. 
[129]  G. A. Battiston, R. Gerbasi, M. Porchia, and A. Marigo, "INFLUENCE OF 
SUBSTRATE ON STRUCTURAL-PROPERTIES OF TIO2 THIN-FILMS 
OBTAINED VIA MOCVD," Thin Solid Films, vol. 239, pp. 186-191, 1994. 
[130]  M. Kemell, M. Ritala, M. Leskela, R. Groenen, and S. Lindfors, "Coating of 
Highly Porous Fiber Matrices by Atomic Layer Deposition," Chemical Vapor 
Deposition, vol. 14, pp. 347-352, 2008. 
[131] "International Technology Roadmap for Semiconductors," International 
Sematech, Texas,  2005. 
[132]  J. Robertson, "High dielectric constant oxides," European Physical Journal-
Applied Physics, vol. 28, pp. 265-291, 2004. 
[133]  F. Wiest, V. Capodieci, O. Blank, M. Gutsche, J. Schulze, I. Eisele, J. Matusche, 
and U. I. Schmidt, "Conformal aluminum oxide coating of high aspect ratio 
structures using metalorganic chemical vapor deposition," Thin Solid Films, vol. 
 
 
150 
 
496, pp. 240-246, 2006. 
[134]  P. C. Stair, C. Marshall, G. Xiong, H. Feng, M. J. Pellin, J. W. Elam, L. Curtiss, 
L. Iton, H. Kung, M. Kung, and H. H. Wang, "Novel, uniform nanostructured 
catalytic membranes," Topics in Catalysis, vol. 39, pp. 181-186, 2006. 
[135]  M. Berginski, J. Huepkes, W. Reetz, B. Rech, and M. Wuttig, "Recent 
development on surface-textured ZnO : Al films prepared by sputtering for thin-
film solar cell application," Thin Solid Films, vol. 516, pp. 5836-5841, 2008. 
[136]  E. R. Parker, B. J. Thibeault, M. F. Aimi, M. P. Rao, and N. C. MacDonald, 
"Inductively coupled plasma etching of bulk titanium for MEMS applications," 
Journal of the Electrochemical Society, vol. 152, pp. C675-C683, 2005. 
[137]  G. Xiong, J. W. Elam, H. Feng, C. Y. Han, H. H. Wang, L. E. Iton, L. A. Curtiss, 
M. J. Pellin, M. Kung, H. Kung, and P. C. Stair, "Effect of atomic layer 
deposition coatings on the surface structure of anodic aluminum oxide 
membranes," Journal of Physical Chemistry B, vol. 109, pp. 14059-14063, Jul 
2005. 
[138]  A. C. Jones, H. C. Aspinall, P. R. Chalker, R. J. Potter, K. Kukli, A. Rahtu, M. 
Ritala, and M. Leskela, "Some recent developments in the MOCVD and ALD of 
high-kappa dielectric oxides," Journal of Materials Chemistry, vol. 14, pp. 3101-
3112, 2004. 
[139]  M. K. Gobbert, V. Prasad, and T. S. Cale, "Modeling and simulation of atomic 
layer deposition at the feature scale," Journal of Vacuum Science & Technology 
B, vol. 20, pp. 1031-1043, 2002. 
[140]  S. O. Kucheyev, J. Biener, T. F. Baumann, Y. M. Wang, A. V. Hamza, Z. Li, D. 
K. Lee, and R. G. Gordon, "Mechanisms of atomic layer deposition on substrates 
with ultkahigh aspect ratios," Langmuir, vol. 24, pp. 943-948, 2008. 
[141]  S. K. Kim, K. M. Kim, O. S. Kwon, S. W. Lee, C. B. Jeon, W. Y. Park, C. S. 
Hwang, and J. Jeong, "Structurally and electrically uniform deposition of high-k 
TiO2 thin films on a Ru electrode in three-dimensional contact holes using atomic 
layer deposition," Electrochemical and Solid State Letters, vol. 8, pp. F59-F62, 
2005. 
[142]  S. W. Jeong, H. J. Lee, K. S. Kim, M. T. You, Y. Roh, T. Noguchi, W. Xianyu, 
and J. Jung, "Effects of annealing temperature on the characteristics of ALD-
deposited HfO2 in MIM capacitors," Thin Solid Films, vol. 515, pp. 526-530, 
2006.