Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The influence of additives on microstructure evolution of electrochemically deposited copper films Gao, Jie 2003

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_2004-0022.pdf [ 6.86MB ]
Metadata
JSON: 831-1.0079153.json
JSON-LD: 831-1.0079153-ld.json
RDF/XML (Pretty): 831-1.0079153-rdf.xml
RDF/JSON: 831-1.0079153-rdf.json
Turtle: 831-1.0079153-turtle.txt
N-Triples: 831-1.0079153-rdf-ntriples.txt
Original Record: 831-1.0079153-source.json
Full Text
831-1.0079153-fulltext.txt
Citation
831-1.0079153.ris

Full Text

THE INFLUENCE OF ADDITIVES ON MICROSTRUCTURE EVOLUTION OF ELECTROCHEMICALLY DEPOSITED COPPER FILMS  By JIE G A O Bachelor of Science, XinJinag University, China, 1982 Masters of Science, University of Science and Technology of China, 1997  A THESIS SUBMITTED IN P A R T I A L FULFILMENT OF THE REQUIREMENTS FOR THE D E G R E E OF M A S T E R OF SCIENCE in T H E F A C U L T Y O F G R A D U A T E STUDIES (DEPARTMENT OF M E T A L S A N D M A T E R I A L S ENGINEERING) We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH C O L U M B I A November 2003 © Jie Gao, 2003  In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department  or  by  his  or  her representatives.  It  is  understood  that  copying  or  publication of this thesis for financial gain shall not be allowed without my written permission.  Department The University of British Columbia Vancouver, Canada  DE-6 (2/88)  ABSTRACT This thesis presents an investigation of the effects of additives on kinetics of deposition and microstructure evolution of Cu films, which were electrochemically deposited (ECD) on Au substrates. The rate of bulk Cu deposition was quantified with cyclic voltammetry. The self-annealing of the as-deposited Cu films was recorded at room temperature with resistivity measurements. Furthermore, the grain size evolution during self-annealing was characterized by X-ray diffraction. The challenges associated with these measurements such as reproducibility tests, the effects of substrate quality, and the aging of chemicals is discussed in detail. A systematic error of resistivity measurements is reduced to 3% by correcting the resistivities value. Considering the sensitivity in the measurement of initial absolute resistivity, it is suggested to use the normalized resistivity for presenting the results.  X R D technique is a good way for estimating the microstructure evolution during selfannealing. However, the use of resistivity techniques is recommended to investigate selfannealing kinetics. The additives, P E G and SPS play an important role in influencing the deposition rate and self-annealing rate, whereas, the effect in JGB also cannot be neglected especially its influence during underpotential deposition (UPD) stage and on the rate of self-annealing. A suitable selection of the type and the concentration in additives will be the key to control the kinetics of deposition and self-annealing in order to match the requirements of Cu interconnects.  11  T A B L E OF CONTENTS ABSTRACT ii T A B L E OF CONTENTS iii LIST OF FIGURES v LIST OF T A B L E S vii ACKNOWLEDGMENT viii 1 INTRODUCTION 1 2 LITERATURE REVIEW 5 2.1 Advanced Copper Interconnects 5 2.1.1 V R C Delay — Replacement of A l by Cu 5 2.1.2 Advanced Cu Dual Damascene Interconnects 9 2.1.3 Copper Electrochemical Deposition (ECD) 11 2.2 Organic Additive Effect on Cu Deposition Kinetics 13 2.2.1 Kinetics of Metal Deposition 13 2.2.2 The Role of Organic Additives in Plating Solution of Cu E C D 15 2.2.3 The Additive Effects on Void-Free Cu E C D 17 2.2.3.1 The Diffusion-Adsorption Theory of Additives 17 2.2.3.2 The Effect of Additive on Void-Free Cu E C D 18 2.2.4 Aging Effect of Additives 22 2.3 Microstructure Evolution in ECD Cu Films 24 2.3.1 General Observations of Microstructure Evolution in E C D Cu Film 24 2.3.1.1 The Characteristics of Self-annealing 24 2.3.1.2 Factors Affecting S elf-Annealing 27 2.3.2 Explanations 30 3 OBJECTIVES 33 4 METHODOLOGY 35 4.1 Substrate Fabrication 35 4.1.1 The Procedures of Substrate Cleaning 36 4.1.2 Electron Beam Deposition and Thermal Evaporation Deposition 37 4.2 Fabrication of Electrochemically Deposited Copper Film 38 4.2.1 Electrolyte Recipe of Copper Plating Bath Solution : 38 4.2.2 Fabrication Procedures of Copper Film....'. 40 4.3 Cyclic Voltammetric (CV) Measurement 42 4.3.1 Cyclic Voltammetric Measurement Procedures 42 4.3.2 The Basic Meaning of Cyclic Voltammetry Curve 43 4.4 Resistivity Measurement 45 4.5 X-Ray Diffraction (XRD) Techniques 49 5 THE I N F L U E N C E OF M E A S U R E M E N T CONDITIONS O N MICROSTRUCTURE E V O L U T I O N 53 5.1 Reproducibility Tests of Resistivity and X R D 53 5.2 Aging Effect of Chemicals on Self-annealing in E C D Copper Films 65 5.3 Effect of Substrate on Self-annealing Kinetics 68 6 THE EFFECT OF ADDITIVES O N MICROSTRUCTURE E V O L U T I O N OF E C D COPPER F I L M 72 6.1 The Influence of Additives on the Cu Electrochemical Deposition 72  6.1.1 The Effects of Additives on Bulk Cu E C D 6.1.2 The Effects of Additives on Cu U P D 6.1.3 The Concentration Effect of Additives on Current Densities..... 6.2 The Influence of Additives on Self-annealing Kinetics 6.2.1 The General Effect of Different Additives 6.2.2 The Effect of SPS and JGB Concentration 6.3 The Effect of Additives on Microstructure Evolution 6.3.1 The Effect of Additives on Relative Intensity of Cu (111) Peaks 6.3.2 The Effect of Additives on F W H M of Cu (111) Peaks 7 S U M M A R Y / CONCLUSIONS A N D R E C O M M E N D E T E D F U T U R E W O R K . . . 7.1 Summary 7.2 Conclusions 7.3 Recommended Future Work 8 REFERENCES  73 75 78 81 82 84 86 86 88 92 92 94 95 96  iv  LIST OF FIGURES Figure 2.1. Intrinsic gate delay (C V/I) and interconnect R C delay at minimum design g  rules of each node  6  Figure 2.2. Numbers of metal layers vs. interconnect material option  9  Figure 2.3 A schematic of the dual damascene process  10  Figure 2.4. A schematic of Cu E C D basic principle  12  Figure 2.5 Possible variation of electroplating rate along the line cross-section  18  Figure 2.6. The I-E characteristics for Cu deposition from the various electrolytes  20  Figure 2.7 Resistivity and stress in the 2 um thick Cu film versus time after the electrodeposition  25  Figure 2.8 Room temperature transition from nanocrystalline (1) to fully annealed (3) and microstructure via abnormal grain growth (2)  26  Figure 2.9 (a) R for films with a thickness of 3.15, 2.15, 1.45, 1.15, 0.90, 0.75, 0.65 and 0.55 s  um from left to right, respectively, and deposition current density is 18mA/cm .(b) R for 2  s  films deposited at 18.0, 13.6, 9.0, 4.5, 2.5, and 1 mA/cm from the left to right, respectively. 2  The films thickness was 1.15 u.m  27  Figure 2.10 Resistance transients of 1pm thick film at room temperature  29  Figure 4.1. A diagram of the P V D evaporation equipment  35  Figure 4.2 A diagram of the three-electrode electrochemical cell set up.  41  Figure 4.3 One scan of cyclic voltammetry with the standard E C D copper solution  43  Figure 4.4 (a) A J A N D E L cylindrical four-point probe;  45  Figure 4.5 A diagram of resistivity measurement with 4-point probe  47  Figure 4.6 Diagram of the diffractometer  50  Figure 4.7 (a) The raw XRD pattern of the Cu (111) peak measured after 40 hours of selfannealing in copper film using standard recipe;(b) A XRD pattern of the Cu (111) peak after the correction of background and K a  52  2  Figure 5.1 Resistivity measurements of seven groups for the self-annealing reproducible tests  55  Figure 5.2 The fraction of transformed self-annealing for the reproducibility tests  56  Figure 5.3 Corrected resistivities based on the results shown in Figure 5.1  57  -.  v  Figure 5.4 The relationship of film thickness and self-annealing rate  58  Figure 5.5 Time evolution of the X R D pattern of E C D copper film measured over 40 hours. Small stars represent the Cu (111) peaks  60  Figure 5.6 The fraction of transformed full width at half maximum (FWHM) for the reproducible test of microstructure evolution  62  Figure 5.7 A comparison for the rate of microstructure evolution between self-annealing and X R D testes  63  Figure 5.8 Aging effect of SPS and JGB on self-annealing process of E C D Cu films.... 66 Figure 5.9 The reduction current density for differently aged copper bath solutions  67  Figure 5.10 The effect of substrates with different fabrication conditions on selfannealing Figure 5.11 The roughness effect on self-annealing in Copper E C D films  69 71  Figure 6.1 The effects of different additives used in Cu bath solution on reduction current densities  74  Figure 6.2 Underpotential deposition of copper with different additives in bath solution Figure 6.3 The charge densities on UPD range with different additives  76 77  Figure 6.4 The changes in current density (measured at - 0.5 V) with the SPS and JGB concentration  80  Figure 6.5 The effect of different additives on the ECD Cu bath solution and selfannealing  82  Figure 6.6 The concentration effect of SPS on the self-annealing process  84  Figure 6.7 The concentration effect of JGB on self-annealing  85  Figure 6.8 The concentration effect of additives on Cu (111) X R D peak relative intensities in E C D Cu films  87  Figure 6.9. The concentration effect of SPS and JGB on FWHM between the initial value of FWHM (0 hour) and the final value of FWHM (after 60 hours)  89  Figure 6.10. The effect of different additives on F W H M between the initial value of F W H M (0 hour) and the final one of F W H M (after 60 hours)  90  Figure 6.11 A comparison of the kinetic characteristic of resistivity and F W H M with different concentrations of SPS  91  vi  LIST OF TABLES Table 2-1. Some Properties of Candidate Interconnect Materials [Muraka, S.P. 1997]  8  Table 4-1 Parameters Used in P V D  37  Table 4-2 Chemical Compositions of Cu Standard Plating Bath Solution:  39  Table 5-1. The Conditions and Parameters for Reproducible Tests  54  Table 5-2 The Conditions of Additives Aging Test:  65  Table 5-3 The Parameters of P V D and ECD for Three-Group Samples  68  Table 5-4 The Parameters and Results of Self-annealing Processing for Roughness Tests  71  Table 6-1 The Additives in Different E C D Cu Bath Solutions:  73  Table 6-2 The Parameters Used in The Investigation of The Concentration Effects  79  Table 6-3 The Standard Conditions of ECD Cu Films Used in Studying the Effect of Additives on Self-annealing  81  vii  ACKNOWLEDGMENTS First of all, I would like to thank my supervisors, Dr. Matthias Militzer and Dr. Dan Bizzotto for their great support and constant encouragement throughout this research, and I am truly grateful for their advice and suggestions during my graduate studies and on this thesis. I would also like to thank my thesis readers, Dr. Tom Troczynski, Dr. Keith A.R. Mitchell, and Dr. Warren Poole, for being on my thesis committee and providing valuable feedback. I would like to thank Dr. Pavel Freundlich and Jeremy Frimer for help with the procedures.  My life and this thesis are enriched by the research colleagues, and I would like to thank to the following people including: Fateh Fazeli, Ruixing Liang, Johnson Go, James Huang, Ed Guerra, Jeff Shepherd, Robin Stoodley, John Odiko Agak, Donna Dykeman, Hany Ahmed, Plamen Petkov, Mohammad Mazinani, Sujay Sarkar, Barbak Raeisenia, and Raj at Bathla. It was a great pleasure to work with them and I also learned a lot from them. M y research work would not have been successful without contributions from the Department of Metals and Materials Engineering at U B C . I would like to thank them for their support.  An especial thank to my friends: the couple of Florence Doidge and Lloyd Doidge, Cornel Lencar, Christina Kaiser, Jennie Yang, and Ben-Wei Lu for their help and care for my family and me during my study at U B C .  1  INTRODUCTION  Modem integrated circuits (IC) consist of tens and hundreds of millions of devices that are supporting a lot of complex functions. Interconnects are the passive wirings which connect these active devices. They provide the power, the ground, the inputs and outputs and the timing signals for these devices. The number of the interconnect segments in an integrated circuit is approximately 10-100 times the number of devices.  With the progress of ultra-large scale integrated (ULSI) circuit technology, the feature size of the electronic devices is becoming smaller and smaller while the complexity and the number of the devices are increasing consistently. Therefore, a high performance interconnect network is required to match the continuously miniaturized device sizes.  For the past 30 years, interconnects have been greatly' developed. In the 1970's polysilicon technology and a single layer of A l or A l alloys interconnects were almost universally used in all very-large scale integrated (VLSI) circuits and the feature size was 2-10 um. In the early 1980's, 2-3 levels of metal interconnects appeared and were accompanied by a reduction of feature size to 1.3-2 um. In the late 1980's and early 1990's, feature size decreased to 0.35-0.5pm and the 4-5 interconnect levels were common. Recently, 6-7 interconnect levels with 0.25-0.15pm feature size were developed [Liu, R. etal., 1999].  1  Due to the continuous shrinking of the feature size to the sub-micron range, the level of interconnects (i.e. the length of interconnects) is ever increasing and their resistance and capacitance become significant for the overall performance of ICs. The associated RC delay increases rapidly while the feature size decreases in the sub-micron range, and it results in a considerable signal propagation delay on long distance interconnects. Therefore, the RC delay is becoming a limitation for running of high-speed circuits with minimal spacing.  In order to fabricate the high performance interconnects with low resistivity-capacitance delay, two approaches have been done. One of them is to lower capacitance by adapting low permitivity (\ow-K) materials as interlevel dielectrics (ILD) between interconnects. The other is to lower resistances by using interconnect materials with lower resistivity.  The practice of using Cu as an advanced interconnect material becomes inevitable due to its higher conductivity compared to that of A l . The dual damascene (DD) process employed for the fabrication of Cu interconnects is a revolutionary change in process architecture for multilevel interconnects.  Electrochemical deposition (ECD) is considered as the method of choice to deposit copper into trenches and vias during the Cu dual damascene process. The E C D process is a very attractive process, not only for its low cost, its capability for high deposition rates, and low processing temperature, but also for the realization of superfilling capability during the damascene process.  2  A good Cu electroplating process contributes to improve yield and reliability of the D D processing of Cu interconnects. Two contributions have been noted: one is the striking "superfilling" behaviour of a plating bath with inhibitor additives, which leads to a voidfree, seam-free damascene deposition. Another is the room temperature self-annealing behaviour affected by the use of additives in the Cu electroplating solution.  Additives play a key role for these two contributions. The presence of organic additives in the copper bath solution influences the deposition rate, which is related to the realization of superfilling deposition. During self-annealing both the physical and electrical properties of E C D copper films are dramatically changed due to the associated microstructure evolution.  The resistivity in as-deposited Cu film is higher than the nominal value for Cu and it drops within a few hours or few days at room temperature until it reaches the nominal value. At the same time, the compressive stress is reduced and the film undergoes spontaneous abnormal grain growth. Also the kinetics of self-annealing can be influenced by organic additives, deposition current density, temperature, and film thickness.  Controlling self-annealing is important to improve the performance and reliability of Cu interconnects.  Several attempts have been made to explain the mechanisms of  microstructure evolution during self-annealing process; however, there is still a lack in a fundamental understanding of this process.  3  Therefore, it is necessary and also interesting to investigate the phenomena and to understand the mechanisms of microstructure evolution in E C D Cu films during selfannealing. It will provide important insight for the development of reliable Cu interconnects.  This thesis is comprised of six chapters: After the Introduction, Chapter 2 gives a literature review, based on the research achievements related to advanced Cu interconnects. This is followed by a summary of the objectives of this work in Chapter 3. Chapter 4 introduces the methodology used in the experiments, which includes the fabrication of substrates and E C D copper films as well as the techniques of cyclic voltammetry, resistivity measurements, and X-ray diffraction (XRD). Chapter 5 and 6 present the experimental results and their discussion. In Chapter 5 the influence of measurement conditions on microstructure evolution of E C D Cu films is evaluated, including reproducibility of tests, aging effects of chemicals, and the influence of substrates. Chapter 6 presents the effects of additives on deposition and self-annealing kinetics. Further, X R D results are provided as direct evidence of microstructure evolution on ECD Cu films during self-annealing. Chapter 7 gives the summary, conclusions, and recommended future work.  4  2 LITERATURE REVIEW 2.1 Advanced Copper Interconnects  2.1.1  V R C Delay — Replacement of A l by C u  The progress of ultra-large scale integrated circuits technology requires the propagation speed of signals through them faster with the continuing reduction of the device size. The typical feature size has been projected to be 0.1pm and 0.05pm in 2006 and 2012, respectively [SIA Roadmap, 1997]. During this period, the device density in chips will increase greatly. Interconnects, as the basic connections of signals in the device, therefore, have to increase their length and decrease their size to meet these requirements of advanced integrated circuits.  Aluminum and its alloys are the conventional materials for interconnects which have been used for the last 30 years [Muraka, S.P. 1997]. With the changes of interconnects for the increasing layers and decreasing size, they are causing some problems which influence the reliability of integrated circuits. One of them is called the R C delay or interconnects delay. R represents the total effective resistance of interconnects and C is the total effective capacitance associated with the inter-level dielectrics (ILD). R C delay increases rapidly as the feature size decreases in the sub-micron range and the gate delay decreases slowly, as shown in Figure 2.1.  5  60 50  RC delay of 1 mm wire  &  V  I  30  <0  20  2  10  Intrinsic gate delay  0 0.1  0.13  0.18  0.25  0.35  0.5  Technology Node (um) Figure 2.1. Intrinsic gate delay ( C V / I ) and interconnect R C delay at minimum design rules o f each node g  [Liu, R. et al, 1999].  The transistor operation is characterized by an intrinsic gate delay, which is used to determine how many operations can be performed i n a time unit [Muraka, S.P. 1997]. Decreasing the feature size means that the charge carriers travel shorter distance within a transistor. A s a result, the intrinsic gate delay decreases with decreasing feature size. However, R C delay becomes the main performance-limiting factor over intrinsic gate delay since it leads  to considerable  signal propagation  delay on long  distance  interconnects and limits the high- speed running o f advanced integrated circuits with minimal spacing. R C delay can be quantified by a simple formula in multilevel interconnects ( M L I ) structures [Muraka, S.P. 1997]:  RC =  P  / 2  ILD  £  (2-1)  6  Where p, I and d  M  s  ILD  , are the resistivity, length and thickness of interconnects; d  lLD  and  are the thickness and permitivity of the inter-level dielectrics.  From the above equation, it can be seen that the increase of interconnect length and the decrease of interconnect size (the thickness of interconnects) lead to an increase of the RC delay. However, a selection of low permitivity (called low K) and/or low resistivity material as new interconnect materials is beneficial for the reduction of the R C delay.  It is known that the main requirements of interconnects are good conductivity, high electromigration resistance, low corrosion, thermal stability, and economic feasibility [Muraka, S.P. 1997]. Table 2-1 summarizes relevant properties of four metals, which are considered to be the possible material of interconnects. Among these presented metals, A l is the traditional material used in interconnects but it has a relatively high resistivity. Cu, Ag, and A u are considered as the available candidates to replace A l because all of them have a lower resistivity than A l . A u has high resistance of electromigration but only shows a small improvement in resistivity. Although A g has the lowest resistivity (1.59u.Q cm), it has poor electromigration resistance [Ono, H . et. al., 1993]. The resistivity of Cu (1.67 u.Q cm) is about 40% lower than that of A l (2.66pQ cm). Further, Cu has an electromigration resistance which is one to two orders of magnitude larger than for A l [Muraka, S.P. 1997]. The key for the good electromigration resistance of Cu is that it has the highest melting point and the smallest self-diffusivity among these elements listed. Furthermore, Cu has the highest strength of all candidate metals as shown in Table 21 [Muraka, S.P. 1997].  7  Table 2-1. Some Properties of Candidate Interconnect Materials [Muraka, S.P. 1997]. Properties Resistivity (]xQ cm) Melting point (°C) Young's modulus (GPa) Yield strength (MPa) Hardness (HV) TCRxl0 (K') Thermal conductivity (W/cm°C) Corrosion in air Electromigration resistance Availability of CVD deposition and PVD etching ECD technique Dry Etching Wet Etching Self-diffusion: Q (eV)(activation energy) D (cm s") (pre-exponential factor) t a l  3  [ b J  lcJ  ldJ  2  1  0  Cu  Ag  Au  Al  1.67  1.59  2.35  2.66  1085  962  1064  660  129.8 216 51 4.3  82.5 172 25 4.1  78.5 130 20-30 4  70.6 55 15 4.5  3.98  4.25  3.15  2.38  Low High  Low Very low  Very high High  High Low  Yes Yes Yes  ?  ?  Yes Yes  Yes Yes  Yes (?) Yes  ?  ?  ?  ?  Yes  Yes  Yes  Yes Yes  2.19 0.78  2.01 1.89  1.97 1.67  1.48 1.71  [a] TCR: temperature coefficient of resistance [b] C V D : chemical vapour deposing [c] PVD: physical vapour deposition  [d] ECD: electrochemical deposition  Therefore, copper has emerged as the choice of an advanced interconnect material to replace A l . As a matter of fact, it has been reported that using Cu and /or low-A^ dielectrics indeed improve the performance of an integrated circuit and enable faster operation [Rahmat, K . et al., 1995]. It can be seen from Figure 2.2 [Bohr, M.T. et al., 1996] that the use of Cu and low-K dielectric reduces the required number of interconnect layers, which results in both a relief of RC delay and a simple and low cost process flow.  8  0.09 2007  0.13 2004  0.18 2001  0.25 1998  0 . 3 5 (nm) 1995  Technology Generation Figure 2.2. Numbers of metal layers vs. interconnect material option [Bohr, M.T. etal., 1995]. 2.1.2  Advanced C u Dual Damascene Interconnects  The use of Cu as interconnect material has been greatly explored for the last several years. Further, the effort to implement Cu interconnects as a feasible production technology has been accelerated after IBM's and Motorola's announcements in 1997 [Gwennap, L. 1997]. However, before a new material is widely adopted, many technical challenges need to be overcome.  The main challenge is the potential device contamination. Cu can easily diffuse through SiC>2 and into Si, causing device leakage and gate oxidation, and, as a result, reliability problems. Cu does not reduce SiC>2 and thus does not adhere well. Also, no volatile inorganic compound is available for Cu and thus it is difficult to etch by conventional Reactive Ion Etching (RIE) processes, which have been traditionally used to manufacture  9  A l interconnects [Howard, B.J. and Steinbruchel, C. 1991]. Fortunately, these problems have been successfully overcome by developing diffusion barriers, e.g TaN, TiN, W N [Muraka, S.P. 1997], and using the dual damascene process, respectively.  In the conventional patterning methods a metal layer is first deposited and then unwanted metal is etched away, leaving the desired pattern of trenches or vias. However, since Cu cannot be etched, dual damascene patterning technique is employed. It consists of three main steps. Firstly, etching the dielectric oxide forms the pattern of trenches or vias. Secondly, the barrier and copper layer are deposited. Lastly, the excess is removed by chemical mechanical polishing (CMP). A schematic of the dual damascene process is shown in Figure 2.3.  |  3  Figure 2.3 A schematic of the dual damascene process (1) Make trenches/vias by etching the oxide; (2) Depositing the barrier and copper into trenches/vias; (3) Removing the excess by chemical mechanical polishing.  10  Dual damascene technique has been completely adopted for the copper metallization process. The successful transition in manufacturing from Al-RIE to Cu dual-damascene has been considered as evolutionary in tooling and revolutionary in process.  2.1.3  Copper Electrochemical Deposition (ECD)  ECD technology has been used to deposit Cu into trenches / vias in the dual damascene process. It has been developed to fill the high aspect ratio of sub-micron dual damascene features. This is a most important process introduced for Cu interconnects, because E C D offers significant reliability improvements and cost reductions relative to chemical vapor deposition (CVD) and physical vapor deposition (PVD). In particular, it provides the possibility of a void-free and seam-free copper damascene process [Edelstein, D. C. et al., 1999],  The process of electrochemical deposition of copper is based on the principle of general electrochemistry as shown in Figure 2.4. During an electrochemical reaction, the process involves the transfer of electrons between one species (for reduction such as [Cu ]) and 0  another species (for oxidization such as [Cu ]). During the reduction of C u 2+  2 +  to Cu, the  number of electrons involved in the reaction is 2, so the reaction can be expressed as: [Cu ] + 2e" <-> [Cu°] 2+  (2-2)  This reaction is called a half-cell reaction, since free electrons cannot exist in the electrolyte and the electron "donor" element must have an "acceptor" counterpart. As indicated by the double arrow used in the above equation, the electrochemical reactions can move in both directions depending on the activity (or concentration) of the species involved, the temperature, and the standard electrode potentials.  11  + Anode  Cathode  Cu ( s ^ C u * (aq) + 2e-  Cu *(aq) + 2e" ^Cu(s) 2  Figure 2.4. A schematic of Cu ECD basic principle  When no external potential is applied to the cell, the cell electrode potential, E, called equilibrium potential, depends on the activities of the copper-solution interface and it can be expressed with the Nernst equation [Bard, A.J. and Faulkner, L.R. 2000]: E =E H nF  In-——{Cu }  (2-3)  Where, R is the gas constant (8.3144 J/mole- K ) ; T is the bath temperature; F is the Faraday constant (96487 C/mole); n is the number of electrons, and n = 2 for the copper reduction process. E° is the standard electrode potential and for copper E°= 0.335V/SHE 2"F  0  (standard hydrogen electrode). {Cu } and {Cu } are the activities of the copper ions in solution and the copper, respectively, the latter being unity for pure copper. For most cases the metal ion concentration may be used in place of the activity [Schneeweiss, M . A . and Kolb, D . M . , 1999].  12  Electrode potential, E, will indicate the direction of the spontaneous reaction. When E is larger than zero, oxidative dissolution of the copper from the cathode electrode occurs and the reaction will move to the left.  When E is less than zero, copper ions in the  solution are reduced which means that Cu° start to be deposited on the cathode electrode, the reaction in equation (2-2) moves to the right. When an external negative potential is applied on the cell, the "concentration" of electrons will increase. The half-cell reaction will move towards right and Cu° will be deposited on the cathode.  Copper ECD has become the standard process used for sub-micron trenches /vias filling in the dual damascene Cu interconnect technology [Gross, M . E . et al., 1998]. Cu electroplating process contributes to improved yield and reliability of the resulting damascene Cu interconnects. It has been found that a suitable small amount of additives introduced in E C D bath solution is very helpful for the superfilling behavior and also provides a means to control the self-annealing behavior of the E C D Cu interconnects.  2.2 Organic Additive Effect on Cu Deposition Kinetics  2.2.1  Kinetics of Metal Deposition  During the metal electrochemical deposition process the characteristics of deposition kinetics can be described by the Butler-Volmer equation, which provides the relationship of current, i , and applied potential, E. The Butler-Volmer equation is expressed by [Bard, A.J. and Faulkner, L.R. 2000], -aF(E  i = i [exp{0  - E ) RT  }-exp{-  (\-a)F{E-E ) eq  RT  •}]  (2-4)  13  where  (E-E *q) e  is the overpotential, which defines the deviation from equilibrium; a is the  factor describing to measure the relative sensitivity of reduction and oxidation processes to potential changes. It is associated with the deposition rate that can be defined by the slope of the I-E curve, described by Ohm's law. f is the exchange current, which is a measurement of the equilibrium reduction/oxidation kinetics of the system, i.e. at the equilibrium (net current is zero) the equilibrium potential is the potential where the forward and reverse rate constants have the same value [Bard, A . J . and Faulkner, L.R. 2000].  With no external potential applied to the working electrode surface, the working electrode will remain at equilibrium potential (E ) relative to a reference electrode. This potential is eq  defined by the Nernst equation, as shown in equation (2-3), as the  fundamental  reduction/oxidation of cupric ion in the electrolyte. When an external negative potential (E) is applied, current will flow and copper metal atoms will begin to deposit on the working electrode.  If a metal is deposited onto a foreign metal substrate (e.g. Cu onto A u instead of Cu onto Cu) a phenomenon of underpotential deposition (UPD) is observed [Kolb, D . M . 1978]. In this case, deposition starts at positive potentials for the corresponding bulk deposition up to one complete monolayer of metal deposited on the substrate. U P D is simply the consequence of an interaction between substrate and depositing metal and it occurs when the adsorbed adatoms are more strongly bound to a foreign substrate than to a substrate of their own kind.  14  UPD is called the initial stage of metal deposition onto a foreign substrate and it usually occurs via one of the following three growth modes: pure three-dimensional growth, layerby- layer growth, or an initial layer-by-layer growth followed by three-dimensional growth [Budevski, E. et al., 1996]. U P D is usually limited to a monolayer of metal, although there are certain indications that, in some cases, it can contain 2-3 monolayers [DespC, A.R. 1983] and as soon as larger amounts are deposited, bulk metal phase is formed.  Bulk deposition can be generally described as made of a nucleation and growth process, the growth of small two-dimensional or three-dimensional clusters, and finally the formation of a thicker overlayer. The first two steps depend strongly on the substrate topography. A l l processes of metal deposition are strongly determined by the defect density of the substrate surface [Schneeweiss, M . A . and Kolb, D . M . 1999].  2.2.2  The Role of Organic Additives in Plating Solution of C u E C D  Traditional copper plating bath contains copper sulfate ( C U S O 4 ) , sulphuricacid (H2SO4), chloride ions, and several essential  organic additives that control the  copper  electroplating rate and morphology [West, A.C.2000].  The additives used in the Cu plating bath are generally organic compounds, and act as inhibitors, grain refiners, and ductilizers co-deposited with copper. The names of proprietary additives in copper E C D are usually related to their functionality. In general, additives can be divided into three groups: brighteners or accelerators, levellers, and suppressors.  15  Brighteners have three possible mechanisms: grain refining, diffusion-control levelling, and random deposition [Kardos, O. 1962]. They are usually propane sulfonic acid groups or disulfides such as Bis-(sodium sulfopropyl)-disulfide (SPS), thiourea and derivatives of thiouresa, 4,5-dithiooctane-l,8-disulfonic acid. Brighteners are typically sulphuric compounds that enhances nucleation of new grains during deposition, thereby reducing the as-deposited grain size. As the grain size decreases, surface roughness decreases and smoother films are obtained [Goodenough, M . et al., 1989; Healy, J.P. et al., 1992]. Brighteners control the grain size of the plated copper, provide growth sites and participates in the charge transfer process on interface of the electrode and electrolyte.  A levelling agent is defined by its ability to produce deposits relatively thicker in substrate depressions while relatively thinner in substrate protrusions with an ultimate decrease in the depth or height of the small surface irregularities [Oniciu, L. 1990], i.e. it is accomplished through selective adsorption on readily accessible substrate surfaces. Levellers include polyamines such as Jennus Green B and derivatives of saffronic dyes. Effective levellers tend to be organic compounds with medium molecular weight containing key functional groups. They have low solubility in the plating solution and a low coefficient of diffusion.  Suppressors consist of the surfactant or wetting agent with long-chain polymers, such as Polyethylene glycol (PEG), polypropylene glycol, and co-polymers of polyglycols. Suppressing agents are characterized by their high molecular weights (>1000g/mol), low solubility, and low coefficient of diffusion. Suppressing functions to form a continuous film on the substrate surface to inhibit Cu deposition rate [Mikkola, R.D. et al., 2000].  16  According to Kelly, J.J. et al. (1998) the adsorption of a monolayer of the suppressor polymer on the substrate surface is sufficient to completely inhibit the Cu deposition. Suppressors block the active sites and inhibit the charge transfer from electrode to copper ions in the electrolyte, subsequently reduce the Cu deposition rate. In the presence of chloride ions, the degree of adsorption and inhibition of suppressors is further enhanced [Ono,H. 1993].  In summary, suppressors and levellers decrease the deposition rate, while brightener or accelerators increase it.  2.2.3  The Additive Effects on Void-Free C u E C D  2.2.3.1 TheDiffusion-Adsorption  Theory of Additives  Superfilling, as a phenomenon during a D D process, represents the event in which Cu electroplating occurs faster at the bottom of the trenches/vias than at the sidewalls and top of trenches/vias, leading to a void-free deposition [Andricacos, P.C. 1999].  Superfilling can be described by the diffusion-adsorption theory. This theory assumes that additives inhibit the electrodeposition and are consumed on the plating surface. Due to the diffusion limitation, concentration of additives decreases in the trench bottom, which leads to a rapid deposition of Cu at the bottom. [Taephaisitphongse,T. et a l , 2001; Cao, Y . et al., 2001; Georgiadou, M . et al., 2001; Kelly, J.J. et a l , 1999; Takahashi, K . M . etal, 1999].  17  This phenomenon can be understood by comparing deposition rates at different points along the featured profile, as shown in Figure 2.5, since the electrodeposition solution with additives exhibited a lower plating rate than without additives [Hu, J.C. et al., 2000]. Additives are accumulated at the top of the trench (point A) over a short distance from the diffusion boundary, where diffusion time is relative short. In contrast, the diffusion time to the bottom of the trench (point C) is too long to keep up with the consumption of additives. Since the effect of some additives is to decrease the deposition rate (such as suppressors), as a result, the addition of additives in the plating bath leads to a higher deposition rate at the bottom of the trench than that at the top, which results in superfilling.  A  Figure 2.5 Possible variation of electroplating rate along the line cross-section [ H u J . C . etal., 2000].  2.2.3.2 The Effect of Additive on Void-Free Cu ECD The first complete report of the chemistry and process conditions used to demonstrate supperfilling in sub-micrometer features was done by Kelly, J J . et al. (1999). They think that the most effective superfilling required the use of all four additives (CT, PEG, SPS,  18  and JGB) in the plating bath, which resulted in the filling of 90 % of the trenches examined.  The ability of superfillng depends on the additive concentration and type. It was found that the increase of leveller concentration leads to a decrease of the deposition rate [Hu, J.C. et a l , 2000], which is beneficial for superfilling. However, the ability of superfilling can be limited by different additive concentrations, i.e. in some ranges of concentration, it increases the ability and in other it deceases it [Zhu, M . et a l , 1999]. The saturation suppression of the suppressor occurs at lower concentrations than that of the leveller. A n amount of suppressor (smaller or equal with 0.3 mol/litre) is enough to inhibit Cu depositions [Sun, Z. W. et a l , 2001].  Several studies revealed that superfilling deposition in sub-micrometer feature size involves competitive interaction between species that accelerate and suppress deposition [Mayer, S. et a l , 2000; Ritzdorf, T. et a l , 1999]. It was found that adding leveller in the bath only slightly affects the behaviour of brightener concentrations on the deposition rate. The increase in brightener concentration significantly affects the Cu deposition rate. However, the leveller with high concentrations acts as a suppressor and reduces the plating rate [Sun, Z.W. e t a l , 2001].  Moffat, T.P. et al. (2000) investigated the various combinations of electrolyte containing Cl-PEG-MPSA for copper deposition, as shown in Figure 2.6. M P S A (thiol - 3-mercapto1-propanesulfonate) acts as the accelerator. They found that when no M P S A was present  19  in the solution, the deposition reaction was significantly inhibited. However, the combination of C l - M P S A (no PEG) led to an acceleration of the deposition rate because accelerators are adsorbed on the copper surface and disrupt the inhibiting function of the suppressors [Mofat, T.P. et al., 2000; Reid, J. et al., 1999]. The rate of the deposition increases with an increase in the M P S A concentration even when no CI" is present in the solution [Moffat, T.P. et al. (2001].  -1  -0.9  -0.8 -0.7 -0.6 Potential, V ( S S E )  -0.5  -0.4  Figure 2.6. The I-E characteristics for Cu deposition from the various electrolytes [Moffat, T.P. etal., 2000].  A strong superfilling requires an electrolyte with a polymer suppressor, an accelerator, and a certain leveller [Cao, Y . et al., 2001], which are called three-additive system. It is found that the electrolyte only having suppressors and levellers or only with accelerators  20  and suppressors led to the formation of voids during the D D process [Reid, J. et a l , 1999; Mayer, S. et a l , 2000].  However, plating baths that contain more types of additives may increase the resistivity of copper films [Sarma, R.L. et a l , 1981] and increase the complexity of the annealing processing. So, an attempt to decrease the amounts of organic additives in the electroplating solution has been promoted.  Using only two additives (PEG and CI") in the Cu bath solution, Hayase, M . et al. (2002) obtained the copper superfilling deposition in 0.2 um trenches. Their results indicated that the inhibiting effect of Cu E C D by P E G was strongly related to CI" concentration. The increase of CI" concentration led to an increase of the inhibition effect of PEG. CI" consumed in the E C D process caused a rapid electrodepositing on trench bottoms. These gradual decreases of inhibition effect with time are in agreement with the diffusionadsorption theory. However, when enough CI" was supplied, the inhibition effect kept constant with time. When no P E G is present, the Cu deposition rate is greatly increased by addition of chloride in the plating bath. Chloride is a strongly adsorbing anion of the supporting electrolyte, which has a strong influence on the adlayer structure as well as the deposition kinetics due to its co-adsorption with the Cu atoms [Schneeweiss, M . A . and K o l b , D . M . 1999].  Hayase, M . et al. (2003) found that a stronger suppression due to P E G is observed with the addition of Br" than that using CI". It is supposed that the halide ions work as an  21  adhesive between P E G and the copper surface. They think that during electroplating, halide ions are consumed and the concentration of halide ions in sub-micrometer trenches is reduced by diffusion-limitation, which weakens the P E G adsorption and thus superfilling is realized.  A hybrid-mode additive such as two-component PEG +BTA (BTA is the benzotriazole) has been promoted recently to replace the original functions of an accelerator and a leveller in the three-additives electrolyte system [Lin, K . C .  et a l , 2002]. The hybrid-  mode additive can simultaneously act as both accelerator and suppressor, i.e. accelerates the Cu deposition rate in the D D base as an accelerator component and inhibits the deposition rate of copper at the opening of the features as a leveller component.  Lin, K . C . et al. (2002) think that the surfilling capacity within deep sub-micron damascene obtained by using an electrolyte with the hybrid-mode additive is valuable for new generation device, because it yields a higher deposition gradient within the filled patterns than any other recipe that includes only suppressing agents.  2.2.4  Aging Effect O f Additives  It was noted that no significant change of film uniformity, surface roughness, and stress in ECD Cu film was observed at different aging baths. However, the self-annealing rate of Cu films plated using aged bath began to slow down and had longer transition time before stabilizing [Koh, L.T. et a l , 2002].  22  It is assumed that the reduction of self-annealing rate is,because more impurities are produced from the increase by products of aging chemicals and is incorporated in the copper film during the E C D process with the aged bath, which aggregated in the grain boundary and slowed down the transformation due to grain boundary pinning effect [Koh, L.T. et a l , 2002]. The decrease in additive concentrations and the increase in byproducts negatively impact filling properties [Sun, Z.W. et a l , 2001].  S M S (Secondary Ion Mass Spectrometry) results indicated an increase of the level of sulfur in aged bath, which correlated with the increase of sulfur incorporation in the Cu film plated using aged bath. The in-film sulfur incorporated during ECP process could aggregate in the grain boundary and delay self-annealing rate of as-deposited Cu film [Koh, L.T. et a l , 2002]!  Similar behaviour of aging solution also was investigated by Moffat, T.P. et al. (2003). They found a significant decrease in deposition rate with electrolyte aging within 15 hours. However, the electrode response was independent of time after 24hours, i.e the aging bath did not affect the deposition kinetics any more after 24 hours. Comparing two accelerators, M P S A (thiol-3-mercapto-l-propanesulfonate) and SPS, they observed that the electrolyte with M P S A led to an increase of deposition rate with time, and finally the system stabilized after about 500mins. In sharp contrast to the M P S A aging dynamics, SPS solution showed no such time dependent behaviour. This suggests that a stable processing environment may be established by using SPS as accelerator.  23  2.3  2.3.1  Microstructure Evolution in ECD Cu Films  General Observations of Microstructure Evolution in ECD Cu Film  2.3.1.1 The Characteristics of Self-annealing Self-annealing is a phenomenon that refers to the microstructure evolution by grain growth at room temperature. Microstructure evolution occurs during a transient period of hours following deposition, and includes an increase in grain size, decrease in resistivity, hardness, compressive  stress, and changes in preferred  crystallographic texture  [Brongersma, S.H. et a l , 2002; Harper, J.M.E. et al., 1999; Lee, D . N . et al., 1996; Kuschke, W . M . et al., 1998].  The resistivity of Cu thin films was used to explore the influence of additive chemistry on impurity incorporation and its effect on microstructure evolution. It has been reported that as-deposited Cu film have about 10-30% higher initial resistivity than that in nominal Cu (~1.68pQcm) [Herper, J.M.E. et al., 1999]. Resistivity dropped 20-25% to near-nominal value within a few hours or few days at room temperature during a self-annealing process [Ritzdorf, T. et al., (1998); Gross, M.E. et al., 1998; Dubin, V . M . et al., 1998; Gignac, L. M . ; et al., 1999; Andricacos, P.C. 1999; Moffat, T.P. et a l , 2000]. At the same time, the stress decreased to near zero from an initial compressive value (~ - 45 MPa) [Ritzdorf, T. et al., 1998], or became tensile and hardness went down [Jiang, Q.T. et al., 2001, Vas'ko, V . A . et al., 2003], as shown in Figure 2.7.  24  Time (days) Figure 2.7 Resistivity and stress in the 2 um thick Cu film versus time after the electrodeposition [Vas'ko, V . A . e t a l , 2003].  It has been confirmed by the images obtained with FEB (Focussed Ion Beam), T E M (Transmission Electron Microscopy) and EBSD (Electron Backscatter Diffraction) that ECD Cu films underwent spontaneous abnormal grain growth at room temperature and the grain sizes increased from 0.05pm - 0.1pm to several micrometer during selfannealing [Herper, J.M.E. 1999; Lingk, C. and Gross, M . E . (1998,2000); Walther, D. et . a l , 2001; Lagrange, S. et a l , 2000]. The large grains nucleated and grew rapidly within a background matrix of small grains, rather than the entire grain population growing simultaneously. This is just the characteristic of abnormal grain growth [Humphreys, F.J. and Hatherly, M . 1996]. Figure 2.8 shows the T E M and EBSD results of grain growth in ECD Cu films [Freundlich, P. et a l , 2003].  25  (1). As deposited;  (2). Partially annealed; (3). Full annealed.  Figure 2.8 Room temperature transition from nanocrystalline (1) to fully annealed (3) and microstructure via abnormal grain growth (2) [Freundlich,P. et al., 2003].  The Mayadas-Shatzke Model [Mayadas, A . E . and Shatzkes, M .  1970] is used for  resistivity calculation in as-deposited copper film. This model predicts that the ratio of resistivity with grain boundaries, po, to the resistivity without grain boundary, po, can be expressed by a simplified form ^ =1 Po  +  (1.40A)(-A_) G \-C  (2-5)  R  Where X is the intrinsic electron mean free path (39nm for single crystal Cu at room temperature) [Chambers, R.G. 1950], G is the mean grain size, and C is grain boundary R  reflection coefficient estimated at 0.2 ~ 0.4 for aluminium [Herper, J.M.E. et al., 1999].  26  2.3.1.2 Factors Affecting Self-Annealing Several factors influence the rate of self-ahnealing such as the deposition current density, temperature, thickness of the Cu film, as well as additives in the plating bath.  It was observed that an increase of plating current densities, film thickness, and deposition temperature led to an increase of the self-annealing rate [Brongersma, S.H. et al., 2002; Harper, J.M.E. et al., 1999; Lee, H . et al., 2001; Lagrange, S. et al., 2000; Chang, S.C. et al., 2002; Mikkola, R.D. et al. 2000; Hsu, H . H . et al., 2001]. The effects of current and thickness are shown in Figure 2.9, in which Rs represents the sheet resistance and R is the initial resistance. 0  100  10  1000  1  1 0  T i m e (hours)  100  1000  Time (hours)  (a)  (b)  Figure 2.9 (a) R for films with a thickness of 3.15, 2.15, 1.45, 1.15, 0.90, 0.75, 0.65 and 0.55 urn from left to right, respectively, and deposition current density is 18mA/cm .(b) R for films deposited at 18.0, 13.6, 9.0, 4.5, 2.5, and 1 mA/cm from the left to right, respectively. The films s  2  s  2  thickness was 1.15 um  [Brongersma, S.H. et al., 2002].  27  Thicker films have a shorter incubation time (the time before self-annealing occurs) and a faster rate of resistivity drop [The, W.H. et a l , 2001;..Lee, H . et a l , 2001; Lagrange, S. et a l , 2000]. The, W.H. et al.(2001) think that incubation time is associated with the value of initial resistivity. Higher initial resisitity leads to a longer incubation time.  The increase of the annealing temperature does not only lead to an increase of the selfannealing rates but also in stress changes from compressive to tensile with 40% reduction in hardness during self-annealing [Mikkola, R.D. et a l , 2000]. By increase the temperature of as-deposited Cu films, defects in the plated copper are further reduced and the resisitivity of the plated copper is lowered to the nominal value [Hsu, H.H. et a l , 2001; Chang, S.C. et a l , 2002].  Stafford, G.R. et al. (1999) and Moffat, T.P. et al (2000) used the Cl-PEG-MPSA additive system for 1pm thick Cu film deposition. It was also observed that copper films deposited from electrolyte containing no additives, or no accelerator (Cl-PEG), or no suppressor (Cl-MPSA) showed no significant resistivity change with time, i.e. no selfannealing occurred during the first 72 hours after deposition. In sharp contrast, copper films electrodeposited from electrolyte containing both accelerator and suppressor (ClPEG-MPSA) had the largest resistivity change and highest self-annealing rate with the highest initial resistance, and the resistivity dropped by 23% within the first 24 hours, as shown in Figure 2.10. They also found that the samples produced with no accelerator or for an additive-free electrolyte showed void formation during damascene process.  28  Figure 2 . 1 0 Resistance transients of lum thick film at room temperature [Moffat, T.P. et a l , 2000].  It was noted that an increase in brightener concentration causes a decrease in tensile stress, which means that a compressive stress is added to the initial stress [Brongersma, S.H. et a l , 2002]. Hu, J.C. et a l , (2000) observed a lower initial resistivity of electroplated Cu films in the absence of a brightening agent, which suggests that the asdeposited Cu films have large grain sizes.  It was found that as-deposited Cu film had a strong (111) texture and the reduction of resistivity was accompanied by an increase of the Cu (111) peak of X R D during selfannealing [Stafford, G.R. et a l , 1999; K i m , J. J. et a l , 2002; Hu, J.C. et a l , 2000]. Stress-induced voiding and electromigration strongly relates to crystallographic texture in  29  metal interconnects [Vaidya, S.  et a l , 1981; Dubin, M . V . et a l , 1998]. It has been  proved that (111) texture' significantly inhibits stress-induced voiding in aluminum interconnects [Knorr, D.B. et a l , 1996] and Cu (111) texture has higher resistance to electromigration [Awaya, N . et a l , 1998; Dubin, M . V . et a l , 1998].  An increase of plating current densities leads to the increase of intensities of Cu (111) peaks [Chang, S.C. 2002(2)] and the increase of grain growth rate [Lee, H . et a l , 2001]. After increasing the annealing temperature, Cu (111) texture is strongly enhanced in all samples tested [Hsu, H . - H . et a l , 2001].  2.3.2  Explanations  Different explanations have been proposed for the mechanisms of microstructure evolution during self-annealing in E C D Cu film. Several research groups [Harper, J.M.E. et a l , 1999; Brongersma, S.H. et a l , 1999; Lingk, C. and Gross, M . E . 1998; Stafford, G.R.  et a l , 2001; K i m , J. J. et a l , 2002] assumed that the reduction in resistivity is  mainly the result of grain boundary elimination due to grain growth. The driving pressure for grain growth is the reduction in grain boundary energy per unit volume. Due to the reduced volume of the grain boundary regions, the stress changes and hardness goes down. The grain boundary pinning by fine dispersion of particles or other pinning sites results in the small as-deposited grain size. During the incubation time the grain boundaries are pinned. The diffusion of pinning species along the grain boundaries causes the rapid start of abnormal grain growth (i.e. self-annealing occurs) after an incubation time at room temperature.  30  No self-annealing occurs at room temperature when the small amounts of impurities existed in as-deposited film are pinning the grain boundaries [Kim, J. J. et a l , 2002]. The Zener's model of pinning [Zener, C. 1948] has been considered explaining the origin of fine as-deposited grain size [Humphreys, F. J.; Ardakani, M . G.1996] and they found that the particle size was almost inversely proportional to the Zener pinning pressure. The asdeposited grains are pinned at the pinning sites. If the pinning sites are particles from the ECD bath, their volume fraction is fixed by the deposited condition, but the average particle size must change at room temperature. As particles disperse with time, they allow some grains to initiate abnormal growth. Then, the average spacing between the particles increase and their ability to pin grain boundaries decreases. Eventually, certain grain boundaries are broken from the pining sites and undergo abnormal grain growth [Humphreys, F.J. and Hatherly, M . 1995] and self-annealing occurs.  High initial resistivity is associated with the defect structure in the as-deposited Cu films, and it is the result of scattering from grain boundaries, dislocations, and vacancies. [Harper, J.M.E. et a l , 1999; Lingk, C. and Gross, M.E. 1998]. Defect reduction was the main reason for the resistivity drop because the reduced amount of defects and grain boundary leads to a less restriction of the electron movement, which promotes the resistivity drop . Therefore, it was suggested that the microstructure transformation in ECD Cu films is driven by two sources. One is the defect density in as-deposited film and the other is the driving pressure from the grain boundary energy in a fine-grained deposit. These two mechanisms compete with each other in determining the resistivity drops by self-annealing [Chang, S. C. et a l , 2002].  31  Since resistivity and stress are two different phenomena, it is assumed that the resistivity change is associated with the changes of grain structure, while the stress varies with impurity concentration of Cu film and it is not governed by grain boundary volume elimination [Brongersma, S.H. et a l , 2002].  Some research groups considered that the stress is the driving force of the microstructure evolution in E C D Cu films. Since stress is relaxed through self-annealing process from moderate compressive to slight tensile, it leads to a decrease in resistivity. [Ueno, K . et a l , 1999; Vas'ko, V . A . et a l , 2003].  Lingk, C. et al. (2000) summarizes their results of FEB, EBSD and X R D pole figure measurements in Cu E C D film and suggested that the microstructure evolution in asdeposited Cu films can be described by three stages:(l) nucleation of new grains; (2) twinning in early stage, and (3) subsequent growth. This may provide a way to understand the emergence of the new microstructure.  32  3 OBJECTIVES From the literature review, it is clear that additives are playing an important role in controlling Cu electrochemical deposition kinetics, which in turn is closely associated with the reliability of the superfilling damascene process. Three-additive system (i.e. an electrolyte with a suppressor, an accelerator and a leveller) has been widely investigated because it meets the superfilling requirements very well. Considering that the more additives used may cause an increase in the amount of impurities into the Cu film during deposition, two-additive system (suppressor and accelerator) has been tried and it also leads to the excellent supperfilling deposition. Microstructure evolution during self-annealing has been confirmed by abnormal grain growth, resistivity drop, and by the changes in stress and texture. The kinetics of self-annealing can be influenced by the deposition current density, temperature, thickness of Cu film, and additives.  Among the investigations presented in the literature review chapter, it can be seen that a lot of work has been done in this area. However, there are still a number of challenges to fully understand the inter-relationships and process conditions during the Cu deposition process, additive concentrations have to be measured and calculated accurately to meet the requirement of superfilling deposition. The superfilling is achieved by the presence of additives in the copper E C D bath. However, the function of each additive is not clear and a physical description of copper superfilling is not yet established. Furthermore, the influence of additives on microstructure evolution during the self-annealing process was only partially investigated.  33  The objective of the thesis is to study the influences of three-additive system (PEG, SPS, and JGB) on the microstructure evolution of E C D copper films during self-annealing process using a model system with deposition on A u substrates. In detail, the following tasks are to be fulfilled: 1) Investigating  the  effect  of additives on deposition kinetics using cyclic  voltammetry technique; 2) Studying the influence of additives on the rate of self-annealing using resistivity measurements at room temperature; 3) Estimating the changes in grain sizes in E C D copper films using peak broadening of X R D during self-annealing processing; 4)  Appraising the effect  of using different techniques ( X R D and resistivity  measurement) on microstructure evolution during self-annealing.  34  4 METHODOLOGY 4.1  Substrate Fabrication  Physical vapour deposition (PVD) technique was used to fabricate the substrates for the copper films deposition. P V D is a vacuum evaporation deposition technique. During the deposition process, material is vaporized from a source and then it is transported through a vacuum chamber to the substrate where it condenses. Figure 4.1 shows a diagram of the P V D evaporation equipment.  Vacuum chamber Substrate  Substrate holder  Au vapor-  Cr vapor  At/  Au target  \\A  Cr target  [  ^  | Vacuum system Power supply  Figure 4.1. A diagram of the P V D evaporation equipment. P V D equipment (AIRCO Temescal Model CV-B) was used in our substrate fabrication process. It is available with either resistance heating (thermal evaporation heating) or electron beam heating. We first deposited a Cr layer ( 3 - 4 nm) on the surface of glass (1mm) using  35  electron beam deposition. The Cr layer was used as an adhesive to improve the cohesion ability between glass substrate and the subsequent A u film. Then the A u film (lOOnm) was deposited on the top of Cr using thermal evaporation as shown in Figure 4.1.  4.1.1  The Procedures of Substrate Cleaning  The cleaning of glass substrates is very important before P V D starts, because it will influence the deposition quality of the gold film. The cleaning procedure is that the glass substrates were firstly washed using methanol and rinsed with deionized water (DI water). This procedure was repeated three times. Then the glass substrates were immersed in hot sulfate nitrate acid solution for 5 hours and rinsed with DI water completely. Subsequently, glass substrates were cleaned in the clean room where they were rinsed with DI water again and dried completely using nitrogen. Finally, the cleaned glass substrates were mounted on the samples holder carefully. The sample holder can hold ten pieces of glass substrate at once. Subsequently, the holder with the glass substrates was mounted in the P V D chamber. To get a uniformly deposited film, the sample holder should be put between the two targets (i.e. Cr and A u sources). The distance between the sample holder and the sources is set about 40cm in the chamber. After that, chamber vacuum was set on and the deposition process started once the vacuum reached the required value. The parameters used in the P V D process are listed in Table 4-1.  36  Table 4-1 Parameters Used in P V D . Parameters Deposition  A u film  Cr film  Vacuum  (Thermal deposition)  (E-Beam deposition)  Glass substrate  (Pa) Depositio n Rate -2.93 xlfj  4.1.2  4  ~ 2.6 A/s  Thickness  Deposition  -lOOnm  Rate -2.0 A/s  Thickness  Thickness  - 3 nm  1mm  Electron Beam Deposition and Thermal Evaporation Deposition  When the vacuum reaches 10" Pa the deposition process can be performed. Using 5  electron beam evaporation technique, a Cr film (3nm) was firstly deposited on the surface of glass substrates. This technique is based on the bombardment with a high-energy electron beam on the target (the material to be deposited) to heat it. The electron beam is generated from an electron gun, which uses the thermo-ionic emission of electrons produced by an incandescent filament (cathode). After they are emitted, electrons are accelerated towards an anode by a high potential (usually several keV). The crucible containing the Cr source acts as the anode.  To get the required deposited rate and reasonable beam intensity, the emission current was selected for our tests (direct current  3.5 - 4.5A). Once the desired thickness of the  Cr film was reached, the gun filament and high voltage window were turned off and the Cr film was ready.  37  After the Cr film was deposited, a A u film with approximately lOOnm in thickness was deposited on the Cr film using thermal evaporation heating techniques. In this technique, the material (here the A u source) was heated using an electrical current passing through a filament or metal plate until melting occurs. Then melting material is evaporated and condensed on the substrate to form the thin film. The evaporation sources, i.e. the material of deposition - gold pellets (99.99%) were placed in an aluminium boat coated by tungsten. After gold was deposited, the fabrication of substrates was completed. The substrates were kept in the vacuum chamber for 30 minutes to cool down.  4.2 4.2.1  Fabrication of Electrochemically Deposited Copper Film Electrolyte Recipe of Copper Plating Bath Solution  The standard electrolyte recipe of copper bath solution used for our copper films was similar to the one used by Kelly, J.J.  et al. (1999). The basic electrolyte solution  consisted of a Cupric Sulfate - Sulfuric Acid system. Four additives were used in the standard electrolyte recipe: Potassium chlorine, Polyethylene glycol, Bis-(3-sulfopropyl)disulfide disodium salt, and Janus Green B. Additives can be grouped into three categories according to their functions: leveller, brightener and suppressor as mentioned in Chapter 2. Among the additives used, Polyethylene glycol acted as suppressor; Bis- (3sulfopropyl)-disulfide disodium salt acted as brightener and Janus Green B acted as leveller. The details of the chemical composition of these additives and their concentration in the standard bath are listed in Table 4-2.  38  Table 4-2 Chemical Compositions of Cu Standard Plating Bath Solution: Composition Concentration  Supplier  CuS0 5H 0  59.9 g/1  Fisher Scientific  Sulfuric acid  H S0  176.4 g/1  Fisher Scientific  Potassium chloride  KC1  CI", 50mg/l  Fluka  Name  Formula  Cupric sulfate  4  2  2  4  Polyethylene glycol (PEG)  H(OCH CH ) OH 2  2  n  300 mg/1  Aldrich Chemical Company Inc.  Bis-(3-sulfopropyl)P M C Specialties  disulfide disodium salt  C H Na 0 S 6  1 2  2  6  4  lmg/1  RASCHIG  (SPS) Janus Green B (JGB)  Group Inc.  Aldrich Chemical C H C1N 30  31  6  lmg/1  Company, Inc  The above bath recipe was used both for investigating the behaviour of additives on copper electrochemical deposition kinetics and on self-annealing characteristics, but the additive concentrations were changed according to the experimental requirements. The following is the chemical structures of additives used:  39  PEG:  H (OCH CH ) OH 2  2  n  SPS: C H Na 0 S 6  1 2  2  JGB: C oH ,ClN 3  4.2.2  3  6  4  6  Fabrication Procedures of Copper Film  A potentiostat (ECO Chemie) with three-electrode system was used to control the potential of the C u / C u  2+  bath solution interface to accurately measure the current flowing  to the working electrode (cathode) as well as for the deposition of copper films. A threeelectrode set up is shown in Figure 4.2, which was used for the investigation of deposition and self-annealing kinetics. In this set up the reference electrode (RE) was a copper wire immersed in the plating bath, and the counter electrode (CE) was a platinum mesh as a path of the current flowing to the working electrode.  Three sets of working electrodes were used in our experiments. The first is a rotating gold disc electrode with area 0.071cm used for the investigation of the copper deposition kinetics using cyclic voltammetry (CV) and the aging effect tests. The second also is a rotating gold working electrode (PVD A u on the circle glass substrates) with an area of  40  1.3cm for roughness effect tests. The third is a non-rotating gold electrode (PVD A u on 2  the rectangular glass substrates) with an area of 2cm used to fabricate the E C D copper films, for the study of self-annealing kinetics by resistivity and X R D measurements.  Rotating W E  Figure 4.2 A diagram of the three-electrode electrochemical cell set up.  Following is the fabrication procedures of the ECD copper films. Before the deposition of the copper film, the gold substrate was cleaned in warm acid (a mixture solution of sulphuric and nitric acid with a volume ratio of 1:1) for 30 seconds. Then it was completely rinsed with DI water. The deposition current density was set at 20mA/cm and E C D time was 135 seconds, appropriate for obtaining a 1pm thick copper film. During the ECD process, a magnetic stirring bar was used. After deposition of the copper film was completed, the sample was immediately rinsed with DI water. Then it was dried with a duster (Pro Duster) softly. Subsequently, the sample was cut into two pieces with  41  lcm area each. One was used for the resistivity measurement and the other for the X R D 2  measurement. Then the self-annealing behaviour and the microstructure evolution were monitored using these two techniques on the same sample simultaneously.  4.3 Cyclic Voltammetric (CV) Measurement  4.3.1  Cyclic Voltammetric Measurement Procedures  In cyclic voltammetric measurements,  a u-Autolab potentiostat  (ECO Chemie),  interfaced with a personal computer, was used to investigate the effects of different types and concentrations of additives added to the copper bath solution on the deposition kinetics. The cell set-up with a three-electrode system was similar to that of E C D copper film fabrication. The difference is that the working electrode was a rotating gold disc with an area of 0.071cm . Rotation speed was set to 1000 rpm (see Figure 4.2). 2  A l l tests using cyclic voltammetry were taken using a scanning potential between - 0.5V and + 0.8V with scan rate of 50mV/s. The standard bath solution was the same as mentioned in Section 4.2.1. Before each cyclic voltammetry scanning starts, the gold disc electrode was cleaned by immersing it in the warm sulfuric - nitrate acid solution for 30 seconds then rinsed with DI water. Finally, the A u disc electrode was mounted on the rotating shaft for cyclic voltammetric measurement.  42  4.3.2  The Basic Meaning of Cyclic Voltammetry Curve  A typical cyclic voltammetry (CV) curve is presented in Figure 4.3. From the C V curve, the change of current density with an applied potential is presented. During one complete cyclic voltammetry scan, two reactions occur: reduction reaction and oxidation reaction. In reduction reaction (cathodic scan), a small amount of Cu is deposited on the gold disc substrate (i.e. [Cu ]+ 2e- —» [Cu ]) and during the oxidation reaction (anodic scan), 2+  0  deposited Cu is stripped from the gold disc substrate (i.e. [Cu°]-» [Cu ] +2e-)2+  .  24— Cu+K C I + P E G + S P S + J G B  15 £  u  10 -  A  < O xida^ion  o  fl  Q  r-  f Immmmmmm 1 1  i~ -0 .6 e cu  -0.4  ,"-0.2  C  N.  0 .2  Up j)  ft  {/  I  -4 U A  mm  1  ~"  0.6  0  Reduction  -10 ,  4 4 P o t e n t i a l (V)  Figure 4.3 One scan of cyclic voltammetry with the standard E C D copper solution  In Figure 4.3, the cathodic potential limit was - 0.5V. The anodic potential limit was + 0.8V. Initially, the scanning starts at the + 0.8V potential where no significant current is measured. As the potential is swept negatively, the initial stage of reduction of [Cu ] of a 2+  monolayer of [Cu] occurs on the gold surface, resulting in reductive current. In this  43  range of potential, the deposition is called underpotential deposition (UPD), as marked in Figure 4.3. U P D is generally limited to a monolayer of deposited metal. Once UPD is completed, the potential is scanned further negatively and the bulk copper deposition occurs.  At the beginning of the bulk deposition, the reductive current increases significantly with increasing negative potential until it reached a limitation potential (at - 0.5V). The potential was then scanned positively. Anodic currents (oxidation reaction) are measured as the deposited Cu layer becomes oxidized and stripped from the gold surface. Once the entire deposited copper layer is stripped from the gold surface the oxidation current drops to zero and a completed C V curve is formed as shown in Figure 4.3.  During an entire scan, a small amount of copper is alternately deposited and stripped from the A u substrate surface. The amount of Cu stripped during the anodic scanning is the same as that deposited during the cathodic scanning. The curve shown in Figure 4.3 was obtained using the standard recipe bath solution, which is used as a reference curve. The effect of additives on the deposition kinetics can be analyzed by change the bath solution and comparing the C V results with the established reference curve.  44  4.4 Resistivity Measurement  A 4-point probe connected with a lock-in amplifier (SR850) was used to measure the resistivity evolution of the copper films. Figure 4.4(a) shows a cylindrical four-point probe head (JANDEL) that was used in our experiments. A diagram of the 4-point probe working principle is presented in Figure 4.4(b).  (a)  (b)  Figure 4.4 (a) A J A N D E L cylindrical four-point probe; (b) The diagram of 4-point probe working principle.  The four-point probe consists of two current carrying probes and two voltage-measuring probes. A current source is used to supply current through the outer two probes; a voltmeter measures the voltage across the inner two probes to determine the sample resistivity. A digital voltmeter with high impedance can minimize the current flow through the portion of the circuit including the voltmeter. Thus, there is almost no potential drop across the contact resistance associated with two voltage-measuring  45  probes; only the sample resistance between the two voltage-measuring probes is recorded. Since very little contact and spreading resistance is associated with the voltage probes in the 4-point probe, the resistance of the sample between the two voltagemeasuring probes can be accurately calculated using Ohm's law, i.e. the ratio of the voltage recorded on the digital voltmeter to the value of the output current of the power supply.  -.  The cylindrical four-point probe head used is preferable for very low resistivity materials such as copper, gold, and platinum and the limit of measurement capability for sheet resistance is from hundreds of Angstroms to several microns thickness. The 4-point probe consists of four equally spaced tungsten metal tips. Each tip is supported by springs on the other end to minimize sample damage during probing.  The measured resistivity is the sheet resistance and the resistivity, p, is given by  p=  Ky  (4-1)  where t is the thickness of the film and K is a geometric factor called correction factor. U is the measured voltage drop and I is the applied current. For the different shape of sample, K can be selected from a standard table (JANDEL).  The correction factor is determined by the geometry of the sample and it is also influenced by the position of the probes on the sample and the spacing between the probes as well. Generally, the larger the sample is, the more accurate any measurement of  46  resistivity is. For small sample, the correction factor is only valid for measurements at the centre of the sample. In our experiments, the shape of the sample is a square with an area of 1cm and 1 um thickness; the probe spacing (s) is set at 1mm. Then correction factor is 2  4.2209 according to the standard table.  The range of initial resistivity in the as-deposited copper films was from 2.1 to 2.4 u.Q cm. The measurement of such small resistivities often suffers from some problems during the test such as electrical noise due to poor contact conditions, thermally induced voltages, and offset voltages produced by devices in the current source, etc. [Meade, M . L . 1983]. However, using a lock-in amplifier solves these problems.  In our experiments, resistivity measurements were carried out using a 4-point probe connected to a SR850 DSP lock-in amplifier. A diagram of the measurement system is illustrated in Figure 4.5. A l l testes are performed at room temperature.  Sample Surface contacted by a 4-Point Pr obe  B  Lock-in Amplifier  Sine out  Figure 4.5 A diagram of resistivity measurement with 4-point probe connected to a lock-in amplifier.  47  In Figure 4.5, the sample surface is contacted by the metal tips of a 4-point probe, in which points 2 and 3 are the voltage probes and points 1 and 4 are the current probes. There are three sources shown in the lock-in amplifier. A is an input source used to record the potential drop of the sample with time, which is measured by the voltage probes of 4-point probe. B is another input source connected with the current probes of 4point probe. A n applied voltage is shown in B. By using a resistor, R (509.7Q), the 0  applied voltage is transferred into the applied current. The applied current is 1.78mA for all resistivity tests. "Sine-out" is the reference source and 1kHz frequency is set to the "sine-out" in our experiments. The basic principle of a lock-in amplifier is that it measures an A C voltage (or current) and gives an output in the form of a D C voltage proportional to the value of the A C signal being measured.  Once the potential drop, U (V), on the sample (across points 2 and 3) is recorded in the x  input source A , the resistivity drop of the sample with time during self-annealing can be calculated using equation (4-1), considering the current 1.78 mA, film thickness 1pm, and the correction factor 4.2209 for the square sample with 1cm area. 2  Resistivity can be used to quantify the self-annealing. For this purpose a fraction transformed, X , can be introduced such that. X=  Where p  0  ~ Po -P« P  o  P  (4-2)  is the initial resistivity value in as-deposited film, and p is the resistivity x  value of a completely annealed film.  48  4.5 X-Ray Diffraction (XRD) Techniques  X-ray diffraction has been used to study the microstructure evolution in E C D Cu films by analyzing peak broadening. The behaviour of grain growth can be predicted from the values of the full width at half maximum (FWHM) using the Scherrer equation. Scherrer equation expresses the inversely proportional relationship of F W H M , p, and the mean grain size, G, and it is given by [Scherrer, P. 1918]  Where X is the wavelength of incident X - ray beam and 6 is the Bragg angle.  X R D experiments were performed on the samples fabricated by the same conditions as that used in the samples of resistivity measurements.  X R D experiments are performed using a Rigaku Rotaflex high-intensity, rotating anode X-ray source (RU-200BH) with a theta/2theta diffractometer system. In this system, X rays are produced from an X-ray tube, which contains a tungsten cathode filament that is heated by an A C voltage ranging from 5-20 V . The anode is a water-cooled target made from pure copper. Electrons are accelerated in vacuum under potentials of 4000-6000 volts, and produce a spectrum, which includes Cu kp, Cu k i and Cu k 2 peaks. A graphite a  monochromator is used to remove Cu kp. So, only Cu diffraction spectrum and the intensity ratio of k i to k a  better interpretation of the diffraction results, the k  a 2  a 2  and Cu k  a  a 2  peaks appear in the  is about 100 to 50. In order to get  peaks should be eliminated from the  spectrum.  49  A diagram of the diffractometer is shown in Figure 4.6. A low-noise germanium detector and four slits are used in this instrumental arrangement. The slits are used to balance the requirement of resolution and intensity for required runs. To measure each peak over the 29 range with a sufficiently large ratio of signal to noise, the effective widths of the slits were selected: the divergent slit (DS) =1°, primary scatter slit (SS) =1°, soller slit = 0.45, and the receiving slit = 0.6. The size of the incident beam depends on the window size of the divergent slit (DS). It is 2mm wide and 1cm long in our tests. The scan range, 20, is set from 30 to 100 degree, scan speed was 2.0 degree/min, and step size was O.Oldegree.  Si  Detector  Figure 4.6 Diagram of the diffractometer . (S is the X-ray source, DS is the divergent slit, SS is the scatter slit, 5/ is the soller slit 1,S is the soller slit 2 and S is the receiving slit 3.) 2  3  The sample was mounted in a glass sample holder. The sample can be spun in the plane of the glass holder to increase the random orientation of grains in the mount during the 929 scanning. The sample rotates at an angle 9, while the detector rotates at an angle 29 to meet the condition of Bragg's law. The instrument is capable of very rapid scans for identification of sample. Generally, much slower scans (scan speed is less than 2.0 50  degree/min) are usually required for the analysis of complex mixed phases, high resolution work, cell refinement, or identification of trace impurities. We used the slower scans in our samples for the observation of the change of microstructure during selfannealing. X R D patterns were measured once in 5 hours during self-annealing. At the same time the resistivity was measured using the four-point probe for the sample with the same deposition conditions as that used in X R D .  The PowderX program has been used to analyze the X R D data in our experiments. The main functions of PowderX are that it can solve two problems, which often occur in the data of powder or polycrystalline samples: (1) the intensity fluctuations in the high-angle side after the Cu K a elimination; (2) high accuracy of the data required for indexing. We 2  used this program for the background subtraction, Kct2 elimination, peak search, F W H M calculation, and indexing.  Figure 4.7(a) shows an example of the raw X R D pattern from the Cu (111) peak, Figure 4.7(b) presents the same peak as shown in Figure 4.7(a), but the background and Kcc  2  have been removed using PowderX. In order to keep the initial information as much as possible, we did not used the peak-smoothing program. The values of full width at half maximum (FWHM) were calculated using PowderX program, which mainly adopts the mixed summation shape profile of the Gaussian and Lorentzian function during profile fitting. F W H M is used to estimate the microstructure evolution of the E C D copper film, which is associated with grain growth.  51  25000  25000  43.8  Figure 4 . 7 (a) The raw X R D pattern of the Cu (111) peak measured after 40 hours of selfannealing in copper film using standard recipe;(b) A X R D pattern of the Cu (111) peak after the correction of background and Kct . 2  In order to quantify the self-annealing during X R D process, we calculated the fraction transformed in F W H M , which used the similar method as resistivity did in equation (4-2). The equation for the fraction transformed, X , in F W H M is shown in equation (4-4): Y  _  (FWHM)„-  (FWHM)  (FWHM) - (FWHM) 0  x  where (FWHM) is the initial F W H M value in as-deposited film, and (FWHM)^ is the 0  F W H M value of a completely annealed film.  52  4  4  .  0  5 THE INFLUENCE OF MEASUREMENT CONDITIONS ON MICROSTRUCTURE EVOLUTION Control of microstructure in electrochemically deposited (ECD) Cu films is important for the manufacturing of reliable and high-quality copper interconnects; however, the challenge is that the rate of self-annealing in E C D Cu films can be affected not only by current density, temperature, and film thickness, but also by the additives, the state of the chemicals and the fabrication conditions for the substrates [Harper, J.M.E. et al., 1999]. Therefore, it is necessary to estimate the reliability of experimental results before conducting systematic investigations of self-annealing in E C D copper films. The following sections present the reproducibility tests in E C D copper films as well as the influences of a number of experimental conditions on the self-annealing kinetics.  5.1  Reproducibility Tests of Resistivity and XRD  Reproducibility experiments are very important for appraising the reliability of results. In order to assess our experimental results, reproducibility experiments have been completed in ECD copper samples. The conditions of the reproducibility tests are listed in Table 5-1.  53  Table 5-1. The Conditions and Parameters for Reproducible Tests Substrate Condition Au (llOnm) Cr (4nm)  Additives in Plating Solution P E G = 300mg/l KCl=105mg/l SPS = lmg/1 JGB = lmg/1  E C D Condition Current Density: 20mA/cm Cu film Thickness: 1pm Reference Electrode: Copper Wire Counter Electrode: Platinum Mesh  Aging Time of Chemicals SPS ~ 4h JGB ~ 4h  Figure 5.1 shows the results of the resistivity measurements in E C D Cu films in the reproducibility test series with the standard copper bath solution at room temperature (see Chapter 4.1). It can be seen from Figure 5.1 that all of the as-deposited E C D copper films have a higher resistivety (from 2.1 to 2.4 pQ cm) than that of bulk copper (1.68 pQcm). The difference of initial resistivity data is ~ 10%. It also was noted that all tested samples had a similar incubation time of about 6.5 hours and after that resistivity started to drop. Resistivity drops until it reaches a stable value. The average drop of resistivity is 21% within 30 hours at room temperature, which matches the resistivity drop range from 20 to 25% reported by several research groups and the resitivity drop is associated with grain growth in copper films during self-annealing [Moffat, T.P. et al., 2000; Harper, J.M.E. etal., 1999].  54  1.6 -I 1  , 10  1  100  Time, h  Figure 5.1 Resistivity measurements of seven group's for the self-annealing reproducible tests.  The fraction transformed in resistivity obtained for the data during self-annealing in the tested ECD copper films is shown in Figure 5.2. It can be seen that all samples have the same trend of microstructure evolution but there seem to be a variable in the rate of transformation during the reproducibility tests. This is considered mainly from the difference of initial resistivity in the as-deposited film. There are a 4.3 hours difference between the fastest transformation and the slowest one at 50% self-annealing. Coefficient of variation is 11.5 %. (Coefficient of Variation =100% x Standard Deviation /mean).  55  1.2  o.o  4  1  1  10  1 100  Time, h Figure 5.2 The fraction of transformed self-annealing for the reproducibility tests.  Further, it is found that there is a general trend among all the reproducibility tests, i.e. a higher initial resistivity has a higher final resistivity and the entire resistivity curves are systematically shifted to different levels as shown in Figure 5.1.  The reason for this  behavior may be associated with a systematic error which may be caused by the geometric arrangement of the resisitivity measurements.  Considering the systematic error, it may be corrected by adjusting the final resistivities to the nominal value for Cu (1.68 uf2cm). Since it is reasonable to assume that the annealed Cu films display the nominal resistivity, the new thickness of films is calculated by the  56  ratio of new final resisitivity (1.68 u.Qcm) to the old final resistivity assuming 1 um thickness according to the equation (4-1), i.e. Paid _ P no min al ^ old ^ new  t  P no min al new  P,old  x \{fum)  (5-1)  The corrected resistivities are shown in Figure 5.3. The difference of adjusted initial resistivities is now reduced to about 3%.  2.3 ~ 3 % difference  1.6 1  10  100  Time.h Figure 5.3 Corrected resistivities based on the results shown in Figure 5.1.  Further, it is noted that the sample with higher initial resistivity has a thinner film. Also considering the sample with higher initial resistivity had a slower self-annealing as shown  57  in Figure 5.2. The relationship between thickness and time at 50% resistivity transformed is shown in Figure 5.4.  18  .c  0.88  0.90  0.92  0.94  0.96  0.98  1.00  1.02  Thickness, urn Figure 5.4 The relationship of film thickness and self-annealing rate.  Figure 5.4 shows a trend that the increasing thickness of films is associated with a faster self-annealing. This is in agreement with the results reported by The, W. H . et al. (2001), who investigated the film thickness effect on self-annealing.  A 10% change in thickness as shown in Figure 5.4 is reasonable considering the fabrication process of E C D Cu films. As mentioned in Chapter 4.2.2, E C D Cu films were made by using a non-rotating A u working electrode, in which the mass transfer is mainly  58  from diffusion. Diffusion mass transfer causes a nonuniform flow during the electron transfer between the electrolyte and working electrode, which may result in a nonuniform film. Therefore, a rotating working electrode is recommended for the fabrication of E C D Cu film since a forced convection is used for the mass transfer process.  Although one sample in Figure 5.4 had opposite behaviour to others, i.e. a thicker film caused a slower self-annealing, its error range (~ 12% difference in time) can be reasonably accepted since the self-annealing kinetics can be affected by other factors except thickness, as mentioned in Chapter 2.  Therefore, the variability of resistivity reproducibility tests can primarily be rationalized by the fact that the difference of self-annealing rates among the tested samples originates from the different thickness of films due to the non-uniform deposited conditions. Considering this systematic error, the resistivity data can be corrected to match the bulk resistivity after completion of self-annealing. The measurement error is reduced from 10% to 3% by modifying the resistivity.  X-ray diffraction (XRD) is a traditional tool to study the sample microstructure. In order to confirm the microstructure evolution in E C D copper films, X R D analysis was done for the same samples which were subjected to the resistivity measurements shown in Figure 5.1. The X R D pattern of an E C D copper film with the standard copper plating solution is shown in Figure 5.5, which was observed over 40 hours. As-deposited Cu films are facecentered cubic and posses a (111) texture. The microstructure changes with time are  59  presented by the increase of the Cu (111) peak. After 40 hours the A u (111) peak height had only slightly increased by less than 5% compared to its initial value. However, the Cu (111) peak height increased by about 51 %.  45000 40000 <*  35000 Q- 30000 o  ~  25000 20000 H  O CN N  •*-> C  *:  CN CM N  o < -i—-A  15000 H  J  -»•—-A  i . y._ 25 h  - L - . A  J  5000  o o .. t„ t._40 h  J U ^ J L _ _ j ^ J L _ 30 h  JL  10000  « - CN T - CM CO, CM,  t _5h  A K__*__  i  d_ 15  20 h  h  lOh  ^Oh  0  30  50  70  90  110  26 in degree  Figure 5.5 Time evolution of the X R D pattern of E C D copper film measured over 40 hours. Small stars represent the Cu (111) peaks.  From Figure 5.5, it can be seen that the Cu (111) peak is more pronounced than others in the as-deposited Cu film, which means that there are larger number of grains with (111) orientation than others. It was also noted that the peak height of Cu (200) increased faster than that of Cu (111) during self-annealing. The difference of the initial and final peak height ratios for Cu (200) / Cu (111) is about 60 % and the ratio of F W H M of Cu  60  (200)/(lll) is increased by over 80 %. This may suggest that the rate of grain size increase in Cu (200) orientation is getting faster than that in (111) orientation during selfannealing. This is similar to the results of Vas'ko, V . A . et al. (2003). They also found that the grain sizes with (111) oriented are much larger than that in (200) oriented grains in the as-deposited Cu films. However, after deposited several days, Cu film from (111) textured in as-deposited state became more (200) than (111) textured and the increase of the ratio of Cu (200)/(l 11) in peak intensity is ~ 40%. This means that the texture in ECD Cu films changes with time during self-annealing  It was observed that during the reproducibility test process, the copper peaks in X R D test were increasing with time until they stabilized after 40 hours, and meanwhile, the resistivity drop was coming to a stable value. This suggests that a similar trend of microstructure evolution was shared during the self-annealing process by using both resistivity and X R D tests in the E C D copper film.  In order to get further information about the microstructure evolution in as-deposited copper films, we calculated the F W H M values from the X R D patterns for each tested sample using the PowderX program (see chapter 4.5.2). Figure 5.6 presents the results of fraction transformed in F W H M of Cu (111) peaks during self-annealing. It can be seen that the F W H M had the trend of microstructure evolution similar to that observed in resistivity and shown in Figure 5.2. At the beginning, there was a ~ 3 hours incubation time, which is shorter than that in resistivity measurements (~6hours), followed by the  61  microstructures evolution in the copper films where the grains in copper films grow significantly, and until they reached a stable stage after approximately 30 hours.  100  Figure 5.6 The fraction of transformed full width at half maximum (FWHM) for the reproducible test of microstructure evolution.  It was noted that the time range at 50 % transformed based on F W H M of Cu (111) peaks averages from 6.8 to 11.6 hours with a coefficient of variation of 24.1%. Comparing this to the results in resistivity tests shown in Figure 5.2, the difference in time range was about 4.8 hours, i.e. it is similar to that observed in resistivity tests. Even though the trend in measuring the microstructure evolution in E C D copper films are similar for the different measurement techniques, the apparent rate of microstructure evolution in X R D  62  test is faster than that observed in the resistivity test for each sample, i.e. times for 50% transformed are on average 5 hours shorter.  The relationship between the results of Figures 5.2 and 5.6 is illustrated in Figure 5.7. The linear line would indicate the correspondence between these two measurements. From Figure 5.7 it shows the unmatched rates of microstructure evolution in these two different test methods. The apparent rate of microstructure evolution in all of the X R D tests is faster than that in the resistivity measurements.  6  9  12  15  18  Time of F W H M , h Figure 5.7 A comparison for the rate of microstructure evolution between self-annealing and X R D testes.  63  It is worth emphasizing that the F W H M of the (111) peak is associated with the grain size, further, F W H M can be influenced by the instrumental arrangement which may result from the x-ray source image, the shape of specimen, and axial divergence of the incident beam, specimen transparency, and receiving slit. The influence of stresses in the films is another important factor to cause the broadening of the peak and affect the F W H M values. The values of F W H M shown in Figure 5.4 were directly calculated from the original X R D data, which included not only the broadening of grain size but also the broadenings due to sample stress and instrument.  The broadening effect from X R D instrument can be considered as a constant contributing to the total broadening of X R D peaks because it is from the instrumental geometry arrangement in the polycrystalline samples [Snyder, R.L. 1999] and all of our samples were tested at the same conditions of X R D setting. However, the broadening effect of the stress in the E C D copper film may affect the final behavior of microstructure evolution substantially because it changes the structure of sample by changing the distance between the two planes (d-value) in a crystal unit cell [Klug, H . P. 1974]. The effects of stress in an as-deposited E C D copper film have caused wide interest recently [Brongersma, S.H. et a l , 2002; Vas"ko, V . A . et a l , 2003]. Therefore, it may be interesting to investigate the stress effect in more detail.  Furthermore, only the F W H M of the Cu (111) peak has been analyzed since the other peaks were too small for an accurate determination of the F W H M . Therefore, the X R D analysis provides only information on the grain size of this particular texture component,  64  which may be different from the overall grain size. However, the resistivity measurement reflects the overall characteristics of grain growth during self-annealing. Therefore, this may also be a reason that an unmatched rate of self-annealing is observed when using these two techniques.  5.2  Aging Effect of Chemicals on Self-annealing in ECD Copper Films  The rate of self-annealing in E C D Cu films can be influenced by the age of the copper plating bath solution. It has been reported that the self-annealing of E C D Cu films formed in an aged plating'bath had a longer transition time before, becoming stabilized after deposition [Koh, L.T. et al., 2002; Moffat, T.P. et al., 2003]. Similar results were found in our E C D copper films, as shown in Figure 5.8, where a comparison of E C D coppers films with differently aged SPS and JGB solutions is presented. The substrates used in samples shown in Figure 5.8 are different from that shown in Figure 5.2. The conditions for aged tests have been listed in Table 5-2.  Table 5-2 The Conditions of Additives Aging Test Samples  1 2 3 4 5  Aged Chemicals  SPS SPS SPS SPS SPS  / / / / /  JGB JGB JGB JGB JGB  Aged Time  Substrates Used in Cu Films  Resistivity Tests  (Rotating A u Disc WE)  Cr Thickness  4hours  lday  2nm  2days 3 days 6days 30days  6days 15days 30days 50 days  2nm 2nm 2nm 2nm  C V Tests  Au Thickness lOOnm lOOnm lOOnm lOOnm lOOnm  65  Comparing the transition time at 50% transformed self-annealing, the freshest bath (aged 4 hours) was found to have a shorter annealing time (6.4 hours) than that aged over one week. An 18 hour transition time at 50% transformed self-annealing has been noted for the aged bath after 30 days, as shown in Figure 5.8, which is 11.4 hours longer than that observed for the fresh bath.  1.2  100 Time, h Figure 5.8 Aging effect of SPS and JGB on self-annealing process of E C D Cu films.  In order to find out how the aged bath solution influenced the kinetics of copper film deposition, cyclic voltammetry method was used with a rotating gold-disc working electrode. Results are shown in Figure 5.9.  66  It can be seen that the rate of the current density started to slow down and tended to stabilize with the increase of aging time when bath solution was older than one week, as shown in Figure 5.9. The current density of the one-week aged bath and the one aged for more than a month show no significant difference. At a given applied potential (-0.5V), the largest reduction current density appeared at -1.39 mA/cm with the fresh bath (lday). 2  There was a 6.8% difference between the largest current density and the smallest one. The difference between the baths aged for six and fifty days, respectively, is only 1.6%. It seems that the aging effect on the deposition rate tends to slow down with time and to be stable after one week.  CM  E o E  c CD  Q c  C D i_ i_  O  -0.4  -0.3  0.2  -0.1  Potential/Ref., V Figure 5.9 The reduction current density for differently aged copper bath solutions.  67  Comparing the results shown in Figures 5.8 and 5.9, it can be seen that the aging effect produces different influences on the deposition kinetics and self-annealing kinetics. Aged chemicals affect the deposited rate marginally after one week; however, they exert a significant influence on the self-annealing kinetics for shorter aging time. These trends are similar as reported by Koh, L.T. et al. (2002) and Moffat, T.P. et al. (2003). In order to get comparable results during the investigation of self-annealing kinetics, the aging time for SPS and JGB has been controlled to be 5 hours for all of our subsequent experiments.  5.3  Effect of Substrate on Self-annealing Kinetics  Substrate type and quality also influences the behavior of self-annealing observed in the ECD copper films. A u substrates with different fabrication conditions using physical vapor deposition (PVD) have been investigated. The parameters of substrates and the fabrication conditions are listed in Table 5-3. Each substrate consists of three layers: Au/Cr/glass. Table 5-3 The Parameters of P V D and ECD for Three-Group Samples Sample  P V D Conditions Cr  Au  Sub-1  20nm  Sub-2 Sub-3  E C D Conditions: Vacuum As Deposited  Current Density  80nm  Au Deposition rate -0.17 nm/s  - 3 x l 0 " Pa  20mA/cm  2  1pm  4nm  HOnm  -0.32 nm /s  - 3 x l 0 " Pa  20mA/cm  2  1pm  2nm  lOOnm  ~0.32nm/s  ~3xl0" Pa  20mA/cm  2  1pm  4  4  4  Cu film Thickness  68  The influence of different substrates is shown in Figure 5.10. It can be seen that the thickness of Cr layer affects the rate of self-annealing. It seems that a thinner Cr film caused a faster rate of self-annealing than the thicker one did.  1.05  -i  ,  1.00 4  0.75 -I 0.1  , 1  , 10  1 100  Time, h Figure 5.10 The effect of substrates with different fabrication conditions on selfannealing.  This may be explained as follows. Cr as the adhesion material was deposited between the glass and gold film. It is possible that a thinner Cr film (~2nm) may produce an uneven surface than a thicker one because there were not enough Cr atom layers to form a uniform Cr film on the glass surface during a given time, so that some Cr islands may appear after electron beam deposition, which influenced the later gold deposition, i.e. it  69  resulted in a rougher A u surface morphology. Then this will influence the following copper deposition and self-annealing kinetics. However, the thickness of A u in the range from 80 -110 nm seems not to affect the self-annealing markedly. This suggests that a much thinner Cr layer may lead to a rougher gold film, which could cause a faster selfannealing process.  In order to confirm this assumption, the surface of the A u film was roughed by running cyclic voltammetry in the gold oxide potential region. The surface of A u film was pretreated to different positive potentials in CI' containing electrolyte. The A u surface is roughed by multiple excursions into the A u oxide potential region. This way has been used in the preparation of rough electrode surfaces for use in Surface-Enhanced Raman Scattering (SERS) [Hager, G. and Brolo, A . G., 2003; Brolo, A . G . et al., 2002].  During the roughness tests, it was found that the larger the positive potential limit is, the rougher the surface of the A u film. When the potential was set to more than + 1.2V, much stronger oxidation reaction occurred and A u film was stripped off from the glass substrate. Therefore, the potential with +1.2V is the limitation for the samples used in our tests.  The roughness effect on self-annealing kinetics is evaluated. The E C D copper film samples were deposited on the roughed A u substrates and a rotating working electrode was used during the E C D process. Figure 5.11 shows the results of roughness effect on self-annealing in two samples: (1) with low roughness (at + 0.8V) and (2) having increased roughness at a positive potential of+1.2V. It can be seen that the rate of self-  70  annealing is indeed related to the roughness of the substrate. The rougher substrate produced a faster self-annealing rate in E C D copper films. The results of roughness effect on self-annealing kinetics and the parameters used in these tests are listed in Table 5-4, which summarizes the effect of the surface roughness of A u substrates on the selfannealing kinetics. Table 5-4 The Parameters and Results of Self-annealing Processing for Roughness Tests Sample No  Potential used in Roughness test (V)  Initial Resistivity as-deposited (pQ.cm)  Incubation Time of Self-annealing (h)  Half-Time during Self-anneal (h)  Resistivity drop (Initial - final)/initial (xl00%)  1 2 3  + 0.8 +1.0 + 1.2  2.2 2.5 3.0  16.7 14.2 9.9  21.6 19.8 10.4  17.6 17.9 18.1  1.05  0.75 -I 1  , _  _|  10  100  Time, h  Figure 5.11 The roughness effect on self-annealing in Copper E C D films.  71  6 THE EFFECT OF ADDITIVES ON MICROSTRUCTURE EVOLUTION OF ECD COPPER F I L M As it is known the incorporation of additives into the Cu plating solution influences both the deposition rate and the self-annealing kinetics in E C D Cu films [Moffat, T.P. 2000; Stafford, G.R. 1999]. Similar results have been observed in our E C D Cu films using different additives. In this chapter, the effects of additives on the microstructure evolution of ECD Cu films are presented. Results and discussion are divided into three sections: (1) the influence of additives on the Cu electrochemical deposition kinetics; (2) the characteristics of self-annealing observed by a drop in resistivity; and (3) the changes of full width at half maximum (FWHM) of X R D peaks associated with grain growth during self-annealing.  6.1  The Influence of Additives on the Cu Electrochemical Deposition  Cyclic voltammetry (CV) technique is good to study the electrolysis mechanisms and electrode kinetics during the E C D process and it has been used to investigate the effect of additives on Cu electrochemical deposition in our experiments. Five-sample groups were tested with different additive combination as listed in Table 6-1. During the tests, the limiting negative potential was set at - 0.5V versus the reference electrode (RE). A rotating gold disc-working electrode was used for all samples. The results of experiments are presented and discussed in the following sections, which include the effects of additives on the bulk Cu E C D , the Cu UPD, and the UPD charge density.  72  Table 6-1 The Additives in Different E C D Cu Bath Solutions: Additives in Cu ECD Bath  Names of Samples KC1  PEG  1* KC1+PEG+SPS+JGB  •  2  KC1+PEG+SPS  •  3  KC1+PEG+JGB  4  KC1+PEG  5  KC1+SPS+JGB  •  SPS  JGB  • •  * Standard Cu plating solution with C u S 0 5 H 0 = 60 g/1, H S 0 = 176.4 g/1, P E G = 300mg/l, KC1 =105mg/l, SPS - lmg/1, JGB = lmg/1. 4  2  6.1.1  2  4  The Effects of Additives on Bulk Cu ECD  Current flow measured during cyclic voltammetry scanning is proportional to the rate of the electron transfer [Bard, A.J. and Faulkner, L.R. 2000]. Figure 6.1 presents the effect of different additives on the electron transfer process during Cu E C D . From this graph, it is obvious that the different additives greatly affect the reduction current density because they result in different rates of Cu bulk deposition.  The sample with the standard bath recipe is marked with filled circles. Comparing it with the one that did not have SPS (marked with filled triangle) in the bath, it can be seen that the deposition rate is much smaller than that containing SPS. As suggested by West, A.C. et al. (2000), SPS works as a brightener when added to the Cu E C D bath and plays an important role not only in refining the grain size but also to accelerate the rate of Cu deposition.  73  Potential vs. RE, V Figure 6.1 The effects of different additives used in Cu bath solution on reduction current densities  It is interesting that JGB as a leveller in the Cu ECD bath did not considerably influence the rate of Cu deposition. Removing JGB (open circles and triangles) from the bath solution, a marginal change in deposition rate was observed when it was compared to the sample with standard solution (filled circles) and the sample without SPS (filled triangles), respectively. Considering the samples without SPS to the one without both of SPS and JGB, it can be found that they had similar behaviour of electron transfer from working electrode to the electrolyte, i.e. JGB affects the deposition kinetics only slightly.  74  PEG, known as a suppressor, resulted in a strong blocking of the Cu deposition [Kelly, J.J. et al., 1999; Mikkola, R.D. et al., 2000]. A similar result has been observed when comparing samples without P E G (marked with filled squares) to other samples with PEG. PEG suppressed the rate of the electron transfer process significantly, as shown in Figure 6.1. Therefore, it is clear from the above discussion that the Cu deposition rate was significantly influenced by the different additives. SPS and P E G play a key role in accelerating and suppressing the rate of Cu deposition, respectively. Furthermore, JGB only slightly affects the rate of bulk Cu deposition, but it plays a significant role in the underpotential deposition and self-annealing, which will be seen later.  6.1.2  The Effects of Additives on C u U P D  In order to know how the different additives in a copper bath affect the initial behaviour of Cu ECD, the characteristic of underpotential deposition (UPD) on the gold substrate was studied. It was observed that the deposition rate strongly depends on the additives used, as shown in Figure 6.2.  It can be seen from Figure 6.2 that JGB had a significant effect on the UPD. The samples without JGB (triangles and circles) have two current density peaks appeared in the UPD range compared to other cases having JGB. One peak occurs in the higher positive potential (~ 0.15V), and the other locates at the lower positive potential (~ 0.1V). However, when adding JGB into the plating bath the current peak with higher potential disappeared. This may be explained that JGB functions as the leveller added to plating bath and through  75  selective adsorption it would trend to smear out the deposition rate by filling the defect sites where U P D occurs at higher potential such as the monatomic high steps or kinks. Once JGB occupies the positions of the defects, the peaks with higher potential disappears and results in the UPD peaks moving from 0.15 to 0.1OV.  - • - KCI+PEG+SPS+JGB •C— KCI+PEG+SPS KCI+PEG+JGB  .0.14  -j 0.00  , 0.05  , 0.10  Baseline for standard bath  , 0.15  -I 0.20  Potential vs.RE, V Figure 6.2 Underpotential deposition of copper with different additives in bath solution.  Adding P E G to the plating bath significantly suppressed the U P D process resulting in a smaller current peak as compared to the one without P E G (filled squares). In the absence of PEG, the fastest Cu deposition rate was observed with the largest current density peak at ~ 0.09 V . However, removing SPS led to the smallest deposition rate. This reinforces  76  the roles of P E G as the suppressor and SPS as the accelerator. It also suggests that the functions of additives work from the beginning of deposition, i.e. during the UPD stage.  The amount of Cu deposited onto a gold surface during U P D process was obtained by integration of the curves from Figure 6.2. Firstly, a baseline was set for each curve (a baseline is shown in Figure 6.1 for the curve with standard bath solution), and then, analyze the integration over the area that is bound by the I-E curve and the baseline. The resulting charge densities are shown in Figure 6.3.  400  -i  1  -0.05  0.00  0.05  0.10  0.15  0.20  0.25  Potential vs.RE, V Figure 6.3 The charge densities on UPD range with different additives.  Comparing samples tested in a standard bath (full triangles), it was found that removing PEG from the Cu bath solution induced 16% increase of the amount of Cu deposited charge density on the surface of the gold substrate during the U P D process (filled squares); "  77  however, the removing of SPS from the bath led to a 26% decrease in charge density (filled triangles). This shows again that the baths with SPS increase the Cu deposition, while PEG in the bath acted to block the Cu deposition.  The charge density of a monolayer of Cu deposited on A u (111) electrode surface has been measured to be 355 uC/cm in sulfate containing electrolyte [Shi, Z. et al., 1995]. As shown in Figure 6.3, the value we measured without PEG (329.2 uC/cm ) in the electrolyte 2  is very close to this ideal measurement. But with PEG in plating bath, the charge density is substantially reduced, confirming the role of PEG as the suppressor strongly affecting the Cu deposition process.  From the results shown in Figures 6.2 and 6.3, it is clear that the amount of Cu deposited (charge density) on the A u substrate is a direct proof for the influence of additives on the deposition kinetics during U P D process. Absence of SPS leads to the lowest deposition rate with the smallest Cu deposited charge density; while removing P E G from the plating bath causes the highest deposition rate with the largest Cu deposited charge density.  6.1.3  The Concentration Effect of Additives on Current Densities  Up to now, we have confirmed that the different additives strongly affect the electron transfer behaviours during Cu ECD. In this section the investigation of the effect of the concentration of SPS and JGB on the Cu E C D is discussed. Table 6-2 lists the bath concentrations used in the tests. The basic solution employed in these tests is: Cu SO4 5 H 0 = 60 g/1; H S 0 = 176g/l; PEG = 300mg/l; KC1 = 105mg/l. 2  2  4  78  Table 6-2 The Parameters Used in The Investigation of The Concentration Effects. Test No. Additive Concentra tion (mg/1)  1 2 3 4 5 6 7  JGB JGB JGB JGB JGB JGB JGB  0 0.5 1 2.5 5 7.5 10  1  2  3  4  5  6  7  SPS  SPS  SPS  SPS  SPS  SPS  SPS  0  0.5  1  2.5  5  7.5  10  •  •  •  «/  </  •  The results for the experiments are summarized in Figure 6.4. In this graph, the current densities at - 0.5V are plotted and show the influence of additive concentration on the deposition kinetics. It is clear from Figure 6.4 that the concentration of SPS and JGB affected the rate of Cu deposition. For increasing concentrations, current density increased for SPS and decreased for JGB. Further, the rate of deposition increases with SPS concentration to reach a saturation level when the SPS concentration attains 2.5mg/l. In contrast, increasing the" JGB concentration leads to a slight monotonous decrease of deposition rate.  79  £ _|  0  !  1  !  2  !  3  !  4  !  5  !  6  !  7  !  8  (  9  f  10  1  11  Concentration, mg/l Figure 6.4 The changes in current density (measured at - 0.5 V) with the SPS and JGB concentration  These results indicate that JGB as the leveller in the bath only slightly suppressed the Cu bulk deposition rate. However, the brightener concentrations (here SPS) significantly affect the Cu deposition rate. But this ability as an accelerator or anti-suppressor seems to be limited when the concentration reaches values over 2.5mg/l.  In summary, it can be seen that the additives PEG, SPS, and JGB influence the kinetics of copper electrochemical deposition significantly. Their effects are visible from the initial stage of Cu deposition (UPD) then also during Cu bulk deposition. JGB plays a special role by selective adsorption during the U P D process. P E G and JGB decrease the rate of Cu  80  deposition and SPS increases the rate of Cu deposition. The concentration effect of additives suggests that it might be necessary to choose suitable additive concentration for control purposes.  6.2  The Influence of Additives on Self-annealing Kinetics  We have confirmed that the type of additives and their concentrations strongly influence the deposition kinetics. It also is interesting to know how sensitive the rate of selfannealing is to these types and concentrations of additives. In this section, the concentration effect of SPS and JGB is discussed and the effect of different additives on the self-annealing is presented. Experimental parameters used in these tests are listed in Table 6-3.  Table 6-3 The Standard Conditions of E C D Cu Films Used in Studying the Effect of Additives on Self-annealing. Substrate Condition Au = 110 nm Cr = 4 n m  Additives in Standard Solution P E G = 300mg/l  E C D Condition (Working electrode: Au)  _  KC1= 105mg/l  Current Density = 20mA/cm Cu film thickness = 1 pm  SPS = lmg/1  Reference electrode: copper wire  JGB = lmg/1  Counter electrode: platinum mesh  Aged Time of Chemicals SPS ~ 5hrs. JGB ~ 5hrs.  81  6.2.1  The General Effect of Different Additives  In order to understand the contribution of each additive added into the copper sulfatesulfuric acid electrolyte solution during the self-annealing process, the resistivity of asdeposited Cu films was tested and the results are summarized in Figure 6.5.  3.4 3.2 3.0  a. No KC! b. No S P S & JGB c. No SPS d. No PEG e. No JGB f. OnlyKCI g. No Additives  1.8  10  100  Time, h Figure 6.5 The effect of different additives on the E C D Cu bath solution and selfannealing.  It was found that i f one of the additives was removed from the plating solution no selfannealing process occurred during the first 70 to 100 hours except in the case without JGB. Not having JGB still has a slowing rate of self-annealing. This is not in agreement  82  with the results reported by Kelly, J.J. et al. (1999). They found that self-annealing occurred only when all additives appear in the Cu bath.  It was also noted that different additives added to the bath solution produced different initial resistivities in as-deposited Cu films and the sample with the additive-free solution had the lowest initial resistivity, as shown in Figure 6.5. Similar results have been reported by Moffat, T.P. et al. (2000) and Stafford, G.R. et al. (1999). They found that the plating solution with no additives, or no accelerator, or no suppressor has only minor changes of the initial resistivity and minimal perturbations of the microstructure, however, the combination of accelerator and suppressor obtained a large changes in initial resistivity. This is a direct indication of additive incorporation and grain refinement. Since the initial resistivity is related to the initial grain size in as-deposited Cu film, it is possible that the different functions of additives and their incorporation influence the microstructure of ECD Cu film.  From the results shown in Figure 6.5, it can be seen that the influence of additives on self-annealing kinetics is different from that on the deposition kinetics as described in Figure 6.1. It may be understood that the deposition kinetics depends on the interaction between the electrolyte with additives and the working electrode (Au substrate), which is reflected by the rate of electron transfer during deposition process. However, the selfannealing kinetics is associated with the microstructure evolution of E C D Cu films, which is affected by the initial gain size and incubation time.  83  6.2.2  The Effect of SPS and J G B Concentration  The effects of additive concentration by studying SPS and JGB on self-annealing were tested and the results are presented in Figures 6.6 and 6.7. It can be seen from Figure 6.6 that SPS concentration affects the self-annealing significantly. Without SPS in the Cu bath solution, the as-deposited E C D Cu film did not show a resistivity drop within 100 hours, i.e. no self-annealing occurred. The sample with standard bath solution (filled circle) had a maximum rate of self-annealing and there is a 20.9% resistivity drop within 32 hours with a completed self-annealing process. The one with ten times higher concentration of SPS, however, resulted in a noticeable decrease of the self-annealing rate with a smaller drop in resistivity 19.6% within 45 hours.  1.05  1.00  4  0.75  -I  1  :  ^  n  :  10  1  100  Time, h  Figure 6.6 The concentration effect of SPS on the self-annealing process.  84  It was noted during the experiments that the surface of the as-deposited copper film with SPS had a shinier smoother surface, but a rougher surface is observed in the absence of SPS. It suggests that having SPS in the Cu bath solution produces a very small-grained deposit. This is in agreement with the definition of SPS, i.e. SPS as a brightener or accelerator used to produce fine grains promotes self-annealing [Oniciu, L. and Muresanm, L. 1990].  1.05 -,  ,  1.00 A  0.75 -| 1  , 10  1  100  Time, h Figure 6.7 The concentration effect of JGB on self-annealing.  In Figure 6.7, it can be seen that without JGB in the Cu plating solution, self-annealing still occurred in the as-deposited E C D Cu film, which is accompanied by a 12% resistivity drop over the first 100 hours with an unfinished self-annealing. This result is in  85  sharp contrast to that without SPS as shown in Figure 6.6. However, the behaviour of ten times concentration of JGB is similar to that of SPS, i.e. it does not increase the selfannealing rate. Therefore, the concentration change of JGB does not affect the existence of self-annealing but it changes the kinetics of self-annealing.  In summary, the maximum self-annealing rate occurred in the sample with the standard bath (SPS =lmg/l, JGB = lmg/1) and it has a 20.9% resistivity drop within 32 hours. Increasing the concentration of SPS or JGB ten fold, however, does not promote the rate of self-annealing. No JGB delays the self-annealing rate, however no self-annealing occurred when SPS is absent.  6.3 The Effect of Additives on Microstructure Evolution  6.3.1  The Effect of Additives on Relative Intensity of Cu (111) Peaks  As illustrated in Figure 5.5, a strong orientation of Cu (111) is shown in the as-deposited Cu film and Cu peak heights on the X R D pattern grew over time during the selfannealing. For the standard plating bath, A u (111) peak height shows a less than 5 % change compared the 51 % changes in Cu (111) peak height within first 40 hours after deposition. Since the A u (111) peak height is more stable with time than the peaks for the Cu film, it was selected as a reference height used for the calculation of the relative intensity of Cu peaks with the different concentrations and types of additives. The calculated relative peak intensities with different concentrations of additives are shown in Figure 6.8.  86  200  Time, h Figure 6.8 The concentration effect of additives on Cu (111) X R D peak relative intensities in E C D Cu films  It was noted that the changes of relative peak intensity are greatly dependent on the concentration and the type of additives. The growth rate of X R D peak height for the sample with the standard bath showed the largest increase of peak height within first 30 hours, however, without SPS in the plating solution it kept a much stable peak height (less than 6% change) within 60 hours, i.e. no self-annealing occurred. Without JGB a slow increase in peak intensity was still observed. However, increasing the concentration of SPS or JGB ten fold did not improve the growth rate of peak height, in reverse, it  87  seems that the increase of concentrations of SPS and JGB delays the growth of peak height, as shown in Figure 6.8.  Therefore, it can be seen that there is a similar trend of microstructure evolution between the results shown in Figure 6.8 and the one shown in Section 6.2 during the selfannealing. It means that a drop in resistivity is associated with an increase of X R D peak height in X R D for E C D Cu film. The type and the concentration of additives affect the behavior of microstructure evolution.  6.3.2  The Effect of Additives on F W H M of C u (111) Peaks  In order to estimate the evolution of grain growth during self-annealing, the values of F W H M of Cu (111) peaks was calculated from the X R D pattern of the E C D Cu films. A comparison of F W H M kinetic characteristic with different additives is presented in Figures 6.9 and 6.10.  From these graphs, it can be seen that there is a decrease of F W H M within 60 hours. The difference among the F W H M reductions depends on the different additives added to the bath. The sample with the standard plating solution has the largest reduction (65%) of F W H M . However, without SPS, or without both SPS and JGB, or without P E G there were almost no changes in their F W H M values (less than 5 % change). But without JGB still caused a 25 % decrease of F W H M . When increasing the concentration of SPS or JGB ten times did not promote the decrease of F W H M . This is a similar finding to the  88  resistivity measurements shown in Section 6.2, i.e. both of them present the same characteristics of self-annealing kinetics, which is affected by the additives. According to Sherrer Equation (4-3) in Chapter 4, the narrowing of F W H M during the self-annealing is associated with the microstructure evolution or the grain growth with time. At the beginning, just after the Cu film was deposited (at 0 hour), the standard sample had the largest F W H M , which is related to the smallest average grain size. Over time, F W H M was reduced, which means grain sizes were increasing. After the first 60 hours, F W H M had decreases by 65% in the standard sample and after that the value of F W H M stabilized. This means that the grain size had reached a limiting value well above the mean free path 39nm (see the equation (2-5) in Chapter 2), and the self-annealing was completed. Therefore, the F W H M changes indeed reflect the microstructure evolution and it can be used to estimate the rate of grain growth.  0.35  JGB=1mg/l  s  p  s  = 1°mg/l  SPS = 0  JGB = 10mg/l  JGB = 0  Concentration of Additives in Bath  Figure 6.9. The concentration effect of SPS and JGB on FWHM between the initial value of FWHM (0 hour) and the final value of FWHM (after 60 hours).  89  0.35  0 Hour I ~1 60 Hours  0.30 A  a a u O) 0J  X  0.25 H  0.20  0.15  LL  0.10  0.05 A  0.00 S P S = 1mg/l J G B = 1mg/l  JGB = 0  SPS = 0  PEG = 0  SPS = JGB = 0  Concentration of Additives in Bath  Figure 6.10. The effect of different additives on F W H M between the initial value of F W H M (0 hour) and the final one of F W H M (after 60 hours)  In order to fully understand the relationship between the resistivity drop and the decrease in F W H M , a comparison of the kinetic characteristics of resistivity and F W H M with different concentration of SPS is presented in Figure 6.11.  It was noted that the results shown in Figure 6.11 are similar to that presented in Chapter 5.1, in which X R D shows faster apparent self-annealing rates than resistivity tests. In Figure 6.11 it shows that this trend is still followed even the SPS concentration is increased. It seems to suggest a general feature of using X R D and resistivity to present the self-annealing kinetics in the same sample, i.e. it is independent of the measurement conditions.  90  1.25  Time, h Figure 6.11 A comparison of the kinetic characteristic of resistivity and F W H M with different concentrations of SPS .  91  7 S U M M A R Y / CONCLUSIONS AND R E C O M M E N D E T E D FUTURE W O R K 7.1 Summary Three kinds of additives were used in copper electrochemical plating solution (PEG, SPS, and JGB). The effects of additives have been investigated at room temperature both on the deposition kinetics and microstructure evolution during self-annealing. The reproducibility results indicate that there was a 10% systematic error of initial absolute resistivity value, which is related to the non-uniform thickness in as-deposited Cu films. The error is reduced to 3% in the corrected resistivities. Thicker film leads to a faster self-annealing. Considering the sensitivity level in the measurement of initial absolute resistivity, it is suggested to use the normalized resistivity for representing the results.  Using two different measurement techniques (resistivity and XRD), a similar trend of microstructure evolution in as-deposited Cu films was observed. However, there is several hours' difference in the microstructure evolution on the same E C D Cu film between the two measurement systems during self-annealing. There is a shorter incubation time in X R D than that of resistivity before self-annealing starts. This suggests that the way to describe self-annealing is different when using X R D and resistivity measurements. It may be explained that resistivity measurement directly reflects the microstructure evolution in the ECD Cu film during the self-annealing, while the F W H M results of X R D indicates not only the information of grain growth but also the broaden  92  effects of experimental instruments  and stress in the samples during the measurement  process.  The age of the bath chemicals influenced the self-annealing kinetics significantly, however, it did not considerably affect the deposition kinetics. Also the fabrication condition of the substrate is an important factor in affecting self-annealing kinetics. The thickness of the Cr layer is believed to activate the surface morphology of the A u substrate. A thinner Cr layer creates a rougher substrate, which promotes a rougher A u surface and, in turn, a faster rate of self-annealing.  The influence of additives on Cu deposition kinetics is enormous. The functions of SPS and PEG as accelerator and suppressor, respectively, are active starting with the first monolayer of Cu deposition. JGB affects the UPD significantly by the selective absorption on the A u surface. Additives in the Cu bath solution firstly influence the initial stage of Cu ECD (i.e. the UPD), and subsequently they influence the Cu bulk deposition. SPS increases the rate of deposition and P E G and JGB decrease the rate of deposition.  Additives do not only affect the deposition kinetics but also the characteristics of selfannealing in as-deposited E C D Cu films. The concentration of additives influences both the rate of deposition and self-annealing. A n optimum self-annealing rate appears to be associated with the standard bath solution. No self-annealing was found within the investigated time range when one of the additives was absent in the bath solution with the exception of JGB. Without JGB in plating bath, there is still a slow self-annealing. It was  93  confirmed that the microstructure evolution of the E C D copper film was associated not only with a resistivity drop but also with narrowing of F W H M of X R D peaks during selfannealing. The X R D results share the same trend as the resisitivity data during selfannealing. The value of F W H M in X R D is suggested for estimating the change of grain size during self-annealing.  7.2 Conclusions The systematic error in reproducibility tests of resistivity originates from the nonuniformed thickness in as-deposited films. It suggests that the control of thickness during the film fabrication process results in a general challenge of resisitivity measurements. A recommended method is to use the rotating working electrode instead of a stationary working electrode with a stirred solution. The former will produce a uniformed surface of film with better film quality since the mass transfer with a rotating working electrode is forced convection.  The control of substrate conditions and aging chemicals also are important to guarantee the reliability of the research results and also the fabrication of the high quality of ECD Cu film. A substrate with ~3nm Cr layer with a bath solution aged over one week is recommended to be used in investigating the self-annealing rate.  X R D technique is good for estimating the microstructure evolution during self-annealing. X R D is a simple technique but its interpretation is not straightforward. F W H M depends on a number of factors and it is then difficult to easily calculate the overall microstructure  94  evolution. Thus, EBSD would be a preferred method to monitor microstructure evolution during self-annealing.  PEG and SPS play an important role in influencing the deposition rate and self-annealing rate, whereas, the effect in JGB also cannot be neglected especially its influence during UPD stage and on the rate of self-annealing. A suitable selection of the type and the concentration in additives will be beneficial to control the rate of self-annealing and to match the requirements of Cu interconnects. Furthermore, a detailed understanding of the mechanism and the kinetics of self-annealing using the quantitative analysis are necessary for further research.  7.3 Recommended Future Work Explaining and modeling the microstructure mechanisms of self-annealing in E C D Cu films based on experimental investigation by: (1)  Fabricating E C D Cu films using the rotating working electrode;  (2)  Analyzing the microstructure evolution in E C D copper films using EBSD techniques during self-annealing;  (3)  Quenching the as-deposited Cu films for better analysis during EBSD process;  (4)  Developing a relationship to describe the effect of additives on the kinetics of Cu deposition and self-annealing;  (5) Quantitative analysis of the changes in grain sizes by building a model.  95  REFERENCES Andricacos, P.C.; Searson, P. C ; Reidsema-Simpson, C ; Allongue, P.; Stickney, J.L. and Oleszek, G . M . (Ed.) (1999). Electrochemical Processing in ULSI Fabrication and Semiconductor/Metal Deposition II. In: Proceedings of the Electrochemical Society International Symposium P V 99-9, 3-6 May 1999, Seattle, Washington, [In: Proc. - Electrochem. Soc, 1999; 99-9], pp.400 Bard, A.J. and Faulkner, L.R. (2001). Electrochemical Methods Fundamentals and Applications. John Wiley, New York. pp. 833 Bohr, M.T.; Ahmed, S. S.; Ahmed, S. U . ; Bost, M . ; Ghani, T.; Greason, J.; Hainsey, R.; Jan, C.; Packan, P.; Sivakumar, S.;Thompson, S.;Tsai, J.; Yang, S.(1996). A high performance 0.25 pm logic technology optimized for 1.8 V operation. Technical Digest - International Electron Devices Meeting, pp.847-850. Brolo, A. G.; Germain, P.; Hager, G. (2002), Investigation of the Adsorption ofLCysteine on a Polycrystalline Silver Electrode by Surface-Enhanced Raman Scattering (SERS) and Surface-Enhanced Second Harmonic Generation (SESHG). Journal of Physical Chemistry B 106(23), pp. 5982-5987. Brongersma, E.;Richard, S.Ff.;Vervoort, L.; Bender, H. ;Vandervorst, W.;Lagrange, S.; Beher, G. and Maex, K . (1999). Two-step room temperature grain growth in electroplated copper. J. Appl. Phys. 86: pp.3642-3645 Brongersma, S. H.; Richard, E.; Vervoort, I.; Maex, K . (1999). Stress in electrochemically deposited copper. AIP Conference Proceedings, 491 (Stress Induced Phenomena in Metallization), pp. 249-254. Brongersma, S.H.; Kerr, E.;Vervoort, I.; Saerens, A. and Maex, K.(2002). Grain growth, stress, and impurities in electroplated copper. J. Mater. Res. 17(3): pp.582 - 589 Budevski, E.; Staikov, G. and Lorenz, W.J. (1996). Electrochemical phase formation and growth: an introduction to the initial stages of metal deposition, Weinheim New York : V C H pp.410. Cao, Y.; Searson, P.C. and West, A.C. (2001). Direct Numerical Simulation of Nucleation and Three-Dimensional, Diffusion-Controlled Growth. J. Electrochem. Soc. 148(5): C376-C382. Cao, Y.; Taephaisitphongse, T.; Chalupa, R. and West, A . C . (2001). Three-Additive Model of'Superfilling of Copper. J. Electrochem. Soc. 148(7): C466-C472 Chambers, R.G. (1950). The Conductivity of Thin Wires in a Magnetic Field. Proc. R. Soc. A 202: pp.378-394 96  Chang, S.C.; Shieh, J. M.;Dai, B.T. and Feng, M.S.( 2002). Reduction of Resistivity of Electroplated Copper by Rapid Thermal Annealing. Electrochem Solid-State Lett. 5(6): C67-C70 Chang, S.C.; Shieh, J.M.; Dai, B.T. ; Feng, M.S. and L i , Y . H . (2002). The Effect of Plating Current Densities on Self-Annealing Behaviors of Electroplated Copper Films. J. Electrochem. Soc. 149(9): G535-G538 DespC, A . R. (1983). Kinetics and mechanisms of electrode processes In Comprehensive Treatise of Electrochemistry, Vol.7, Bockris, J. O ' M (ed.). New York: Plenum Press, pp. 451 Dubin, V . M . ; Morales, G.; Ryu, C. and Wong, S.S. (1998). Microstructure and Mechanical Properties of Electroplated Cu Films for Damascene ULSI Metallization. In: Thin-Films—Stresses and Mechanical Properties VII. Cammarata, R. C ; Nastasi, M . A . ; Busso, E. P. and Oliver, W. C. (Ed.). Proc. Mater. Res. Soc. Symp. 505: pp.137-142 Edelstein, D.C.; Andricacos, P.C.; Agarwala, B.'; Carnell, C. ; Chung, D . C o o n e y HI, E.; Cote, W. ; Locke, P. ; Megivern, C. ; Wachnick, R. and Walton, E. (1999). Copper Interconnect Technology in Semiconductor Manufacturing. In: Electrochemical Processing in ULSI Fabrication and Semiconductor/Metal Deposition II). Proceedings of the Electrochemical Society 3-6 May 1999, Seattle, Washington. Andricacos, P.C.; Searson, P. C ; Reidsema-Simpson, C. ; Allongue, P.; Stickney, J.L. and Oleszek, G . M . (Ed.). P V 99-9:pp.l-8 Freundlich, P.; Millitzer M . ; and Bizzotto, D. (2003). Poster for MRS Spring Meeting 2003, San Francisco. Georgiadou, M . ; Veyret, D. ; Sani, R.L. and Alkire, R. C. (2001). Simulation of Shape Evolution during Electrodeposition of Copper in the Presence of Additive. J. Electrochem. Soc. 148(1): C54-C58 Gignac, L. M . ; Rodbell, K . P.; Cabral, C , Jr.; Andricacos, P. C ; Rice, P. M . ; Beyers, R. B.; Locke, P. S.; Klepeis, S. J. (1999), Characterization of plated Cu thin film microstructures. Materials Research Society Symposium Proceedings 562(Polycrystalline Metal and Magnetic Thin Films), pp.209-214 Goodenough, M . and Whitlaw, K . J. (1989), Studies of copper deposition for high aspect ratio printed circuit boards. Transactions of the Institute of Metal Finishing 67(3), pp.57-62. Gordon, J. G.; Melroy, O. R.; Toney, M . F. (1995), Structure of metal-electrolyte interfaces: copper on gold(lll), water on silver(lll). Electrochimica Acta 40(1), pp.3-8.  97  Gross, M.E.; Lingk, C.; Siegrist, T.; Coleman, E.; Brown, W.L.; Ueno, K.; Tsuchiya, Y.; Itoh, N . ; Ritzdorf, T.; Turner, J.; Gibbons, K . ; Klawuhn, E. ; Biberger, M . ; Lai, W.Y.C.; Miner, J.F. ; Wu, G. and Zhang, F. (1998). In: Advanced Interconnects and Contact Materials and Processes for Future Integrated Circuits. Murarka, S. P.; Eizenberg, M . ; Fraser, D. B. ; Madar, R. and Tung, R. (Ed). Proceedings of Materials Research Society Symposium 514: pp.293-299 Gross, M . E.; Drese, R.; Lingk, C ; Brown, W. L.; Evans-Lutterodt, K . ; Barr, D.; Golovin, D.; Ritzdorf, T.; Turner, J.; Graham, L. (1999), Electroplated damascene copper: process influences on recrystallization and texture. Materials Research Society Symposium Proceedings 564(Advanced Interconnects and Contacts), pp.379-386. Gross, M.E.; Lingk, C ; Brown, W.L. and Dress, R. (1999). Implications of damascene topography for electroplated copper interconnects. Solid State Technol. 42(8): pp.47-54. Gwennap, L. 1997. Microprocessor Report, Vol.11, pp.14 Hager, G. and Brolo, A . G. (2003), Adsorption/desorption behavior of cysteine and cystine in neutral and basic media: electrochemical evidence for differing thiol and disulfide adsorption to a Au(lll) single crystal electrode. Journal of Electroanalytical Chemistry, Vol. 550-551, pp.291-301. Harper, J.M.E.; Cabral, Jr., C ; Andricacos, P.C.; Gignac, L.; Noyan, L C ; Rodbell, K.P. and Hu, C.K.(1999). Mechanisms for microstructure evolution in electroplated copper thin films near room temperature. J. Appl. Phys. 86: pp.2516-2525 Hayasse, M . ; Taketani, M . ; Aizawa, K. ; Hatsuzawa, T. and Hayabusa, K . (2002). Copper Bottom-up Deposition by Breakdown of PEG-Cl Inhibition. Electrochem. SolidState Lett. 5(10): C98-C101. Hayasse, M . ; Taketani, M . ; Zawa, T. H. and Huabusa, K. (2003). Preferential Copper Electrodeposition at Submicrometer Trenches by Consumption of Halide Ion. Electrochem. Solid-State Lett. 6(6): C92-C95. Healy, J.P.; Pletcher, D. and Goodenough, M . (1992). The chemistry of the additives in an acid copper electroplating bath : Part III. The mechanism of brightening by 4,5-dithia-octane-l, 8-disulphonic acid. J. Electroanal. Chem. 338: pp. 179-187 Howard, B.J. and Steinbruchel, C. (1991). Reactive ion etching of copper in SiCU-based plasmas. Appl. Phys. Lett. 59(8): pp.914-916. Hsu, H . H , Lin, K . H . ; Lin, S.J. and Yeh, J.W. (2001). Electroless Copper Deposition for Ultralarge-Scale Integration. J. Electrochem. Soc. 148(1): C47-C53  98  Hu, J.C.; Chang, T.C.; Wu, C.W. ; Chen, L.J. Hsiung, C.S.; Hsieh, W.Y.; Lur, W. and Yew, T.R.(2000). Effects of a new combination of additives in electroplating solution on the properties of Cu films in ULSI applications. J. Vac. Sci. Technol. A. 18(4): pp.1207-1210. Humphreys, F. J.; Ardakani, M . G. (1996).Gram boundary migration and Zenerpinning in particle-containing copper crystals. Acta Materialia 44(7): pp. 2717-2727'. Humphreys, F.J. and Hatherly, M . (1995). Recrystallization and Related Annealing Phenomena., Oxford, U K . , Pergamon, pp.497 Jiang Q.T. and Thomas, M.E.(2001). Recrystallization effects in Cu electrodeposits used in fine line damascene structures. J. Vac. Sci. Technol. B. 19(3): pp.762-766 Kardos, O.; Foulke, D.G. and Tobias, C.W. (Ed).(1962). Advances in Electrochemistry and Electrochemical Engineering, Vol. 2, Interscience Publishers, New York Kelly J.J., and West, A . C . (1998). Copper Deposition in the Presence of Polyethylene Glycol. I. Quartz Crystal Microbalance Study. J. Electrochem. Soc. 145(10): pp.3472-3476 Kelly, J.J.; Tian, C. and West, A. C. (1999). Leveling and Microstructural Effects of Additives for Copper Electrodeposition. J. Electrochem. Soc. 146(7): pp.25402545 Kim, J. J.; Kim, S.K. and Bae, J.U. (2002). Investigation of copper deposition in the presence of benzotriazole. Thin Solid Films. 415: pp.101-107 Klug, H. P. and Alexander., L. E. (1974). X-Ray Diffraction Procedures for polycrystalline and amorphous materials. New York. J. Wiley & Sons. pp. 716 Knorr, D.B.; Tracy, D.P. and Rodbell, K.P.(1991). Correlation of texture with electromigration behavior in Al metallization. Appl. Phys. Lett. 59(25): 32413243. Knorr, D.B. and Rodbell, K.P. (1996). The role of texture in the electromigration behavior of pure aluminum lines J. Appl. Phys. 79: pp.2409-2417. Koh, L.T.; You, G.Z.; L i , C.Y. and Foo, P.D.(2002). Investigation of the effects of byproduct components in Cu plating for advanced interconnect metallization. Microelectronics Journal. 33(3): pp. 229-234. Kolb, D . M . (1978). Advances in Electrochemistry and Electrochemical Engineering, Vol.11, Interscience Publishers, New York, pp.125  99  Kuschke, W . M . ; Kretschmann, A.; Keller, R.M.; Vinci, R.P.; Kaufmann, C. and Arzt, E. (1998). Textures of thin copperfilms.J. Mater. Res. 13: pp.2962-2968 Lagrange, S.; Brongersma, S.H. ; Judelewicz, M . ; Suerens, A.; Vervoort, I. ; Richard, E.R.; Palamans, R. and Maex, K.(2000).Self-annealing characterization of electroplated copper films. Microelectronic Engineering. 50: pp.449-457 Lee, D.N. (1996). Texture and Related Phenomena of Copper Electrodeposits. In Mater. Res. Soc. Symp. Proc. Advanced Metallization for Future ULSI. Tu, K. N.;Mayer, J. W.; Poate, J. M . and Chen, L. J. (Ed.), 427, pp.167-178 Lee, H.; Lopatin, S. D. and Wong, S.S. (2000). Correlation of stress and texture evolution during self- and thermal annealing of electroplated Cu films. In: Proceedings of the IEEE Int. Interconnect Technology Conf. San Francisco, C A , 5-7 June, pp.114-117 Lee, H.; Lopatin, S. D.; Marshall, A . F. and Wong, S. S. (2001). Evidence of dislocation loops as a driving force for self annealing in electroplated Cu films. In: Procedings of the IEEE Int. Interconnect Technology Conf., Burlingame, C A , 4-6 June, pp. 236-240 Lin, K.C.; Shieh, J.M. ; Chang, S.C.; Dai, B.T.; Chen, C F . and Feng, M.S.( 2002). Electroplating copper in sub-100 nm gaps by additives with low consumption and diffusion ability. J. Vac. Sci. Technol. B. 20(3): pp.940-945 Lin, K.C.; Shieh, J.M. ; Dai, B.T.; Chen, C.F.; Feng, M.S. and L i , Y . H . (2002;. Leveling effects of copper electrolytes with hybrid-mode additives. J. Vac. Sci. Technol. B. 20(6): pp.2233-2237 Lingk, C. and Gross, M . E . (1998). Recrystallization kinetics of electroplated Cu in damascene trenches at room temperature. J. Appl. Phys. 84: pp.5547-5553. Lingk, C ; Gross, M . E . and Brown, W.L. (2000). Texture development of blanket electroplated copper films. J. Appl. Phys. 87: pp.2232-2236 Liu, R.; Pai,C.S.; Martinez, E. (1999). Interconnect technology trend for microelectronics. Solid-state Electronics. 43(6): pp. 1003-1009 Mayadas, A.E. and Shatzkes, M . (1970). Electrical-Resistivity Model for Poly crystalline Films: the Case of Arbitrary Reflection at External Surfaces. Phys. Rev. B.l(4): pp.1382-1389.  100  Mayer, S.; Contolini, R.; Jackson, R.; Reid, J. ; Martin, J. ; Morrissey, D.; Schetty, R. (1999). Integration of Copper PVD and Electroplating processes for Damascene feature electrofilling. In Interconnect and Contact Metallization for ULSI. Proceedings - Electrochemical Society. Arita, Y.; Mathad, G.S. and Rathore, H.R. (Eds). Honolulu/Hawaii, Fall 1999. P V 99-31: pp.174-189 Meade, M . L . (198D. Lock-in Amplifiers: principles and applications: London, U K : Peregrinus. Pp. 232. Mikkola, R. D.; Jiang, Q.-T.; Carpenter, B. (2000), Copper electroplating for advanced interconnect technology. Plating and Surface Finishing 87(3), pp.81-85. Moffat, T P . ; Boenvich, J.E.; Huber, W.H. ; Stanisheysky, A. ; Kelly, D.R. ; Stafford, G.R. and Josell, D. ( 2000). Superconformal Electrodeposition of Copper in 50090 nm Features. J. Electrochem. Soc. 147(12): pp.4524-4535 Moffat, T P . ; Wheeler, D.; Huber, W.H. and Josell, D. (2001). Superconformal Electrodeposition of Copper. Electrochem. Solid-State Lett. 4(4): C26-C29 Moffat, T P . ; Wheeler, D.; Witt, C. and Josell, D. (2002). Superconformal Electrodeposition Using Derivitized Substrates. Electrochem. Solid-State Lett. 5(12): C110-C112 Moffat, T P . ; Baker, B. ; Wheeler, D. and Josell, D. ( 2003). Accelerator Aging Effects During Copper Electrodeposition. Electrochem. Solid-State Lett. 6(4): C59-C62 Muraka, S.P. (1997). Multilevel interconnections for ULSI and GSI era. Materials Science and Engineering, R19(3-4): pp.87-151 Oniciu, L. and Muresan, L. (1991), Some fundamental aspects of leveling and brightening in metal electrodeposition. Journal of Applied Electrochemistry 21(7), pp.565-74. Ono, H.; Iijima, T.; Ninomiya, N . ; Nishiyama, A . ; Ushiku, Y . and Iwai, H . (1993). Topology of Silver Films Annealed in Air, Japan Society of Appl. Phys, 40 Spring Meeting, Ext. Abstract, A p r , pp. 814  th  Porter, D.A. and Easterling, K . E . (1981). Phase Transformations in Metals and Alloys. New York: Van Nostrand Reinhold. pp. 446 Rahmat, K ; Nakagawa, O. S.; Oh, S-Y.; Moll, J.; Lynch, W. T. (1995), A scaling scheme for interconnect in deep-submicron processes. Technical Digest International Electron Devices Meeting pp. 245-248.  101  Reid, J.; Bhaskaran, V.; Contolini, R.; Patton, E.; Jackson, R.; Broadbent, E.; Walsh, T.; Mayer, S.; Schetty, R.; Martin, J.; Toben, M . ; Menard, S. (1999), Optimization of damascene feature fill for copper electroplating process. IEEE International Interconnect Technology Conference, Proceedings, San Francisco, May 24-26, pp. 284-286 Ritzdorf, T.; Graham, L.; Jin, S.; Mu, C. and Fraser, D. (1998). Self annealing of electrochemically deposited copperfilmsin advanced interconnect applications. In: Proceedings of the IEEE International Interconnect Technology Conference, San Francisco, C A , June 1-3, 1998. pp.166-168. Ritzdorf, T.; Chen, L.; Fulton, D.; Dundas, C. (1999), Comparative investigation of plating conditions on self-annealing of electrochemically deposited copper films. IEEE International Interconnect Technology Conference, Proceedings, San Francisco, May 24-26, pp.287-289. Sarma, R.; Lekshmana; Nageswar, S. (198U, Electrodeposition of copper in the presence of 2-mercaptoethanol. Surface Technology 12(4), pp.377-82. Schneeweiss, M.A.and Kolb, D . M . (1999). The Initial Stages of Copper Deposition on Bare and Chemically Modified Gold Electrodes. Phs. Stat. Sol. (a) 173(1): pp.5171 Seah, C.H.; Mridha, S. and Chan, L.H. (1999).-Annealing of copper electrodeposits. J. Vac. Sci. Technol. A . 17(4): pp. 1963-1967. Semiconductor Industrial Association (SIA) Roadmap. (1997). Scherrer, P., (1918), Estimation of the size and internal structure of colloidal particles by means ofRdntgen rays. Nachr. Ges. Wiss. Gdttingen, pp.96-100. Shi, Zhichao; Wu, Shijie; Lipkowski, Jacek. (1995), Coadsorption of metal atoms and anions: Cu UPD in the presence of S042-, CI- and Br-. Electrochimica Acta 40(l),pp.9-15. Snyder, R.L.(1999). X-ray Characterization of Materials in: X-Ray Diffraction, Edited by Lifshin, E., Weinheim ; Chichester : bWiley-VCH, pp.105. Stafford, G.R.; Vandin, M.D. ; Moffat, T.P.; Armstrong, N . ; Jovic, V . D . and Kelly., D.R. (1999), 77ze influence of additives on the room-temperature recrystallization of electrodeposited Copper. In: Electrochemical Technology Applications in Electronics III. Proceedings of Electrochemical Society. Madore, C ; Osaka, T.; Romankiw, L. T. and Yamazaki, Y . (Eds.), Honolulu, Hawaii, Fall 1999 P V 9934: pp.340-350  102  Stafford, G.R.; Moffat, T.P. ; Jovic, V.D.; Kelly, D.R. ; Bonevich, J. ; Josell, D.; Vaudin, M . ; Armstrong, N . ; Huber, W.and Stanishevsky, A . ;(200V. Cu electrodeposition for on-chip interconnections. 2000 International Conference of American Institute of Physics: Characterization and Metrology for ULSI Technology. Gaithersburg, Maryland, 26-29 June 2000. D.G. Seiler, A . C . Diebold, T.J. Shaffher, R. McDonald, W. Murray Bullis, P.J. Smith, E . M . Secula (Eds.) 550: pp.402-406 Sun, Z.W. and Dixit, G. (2001). Void-free copper deposition. Solid State Technology. 44(11): pp. 97-100 Taephaisitphongse, T.; Cao, Y . and West, A . C. ( 2001). Electrochemical and Fill Studies of a Multicomponent Additive Package for Copper Deposition. J. Electrochem. Soc. 148(7): C492-C497 Takahashi, K . M . and Gross, M . (1999). Transport Phenomena That Control Electroplated Copper Filling of Submicron Vias and Trenches. J. Electrochem. Soc. 146(12): pp. 4499-4503 The, W.H.; Koh, L.T.; Chen, S.M.; Xie, J.; L i , C.Y. and Fob, P.D.(2001). Study of microstructure and resistivity evolution for electroplated copper films at nearroom temperature. Microelectronics Journal. 32(7): pp.579-585 Ueno, K ; Ritzdorf, T. and Grace, S. (1999) Seed layer dependence of room -temperature recrystallization in electroplated copper films, Journal of Applied Physics, 86 (9), pp.4930-4935. Vas'ko,V.A.; Tabakovic, I. and Riemer, S.C. (2003). Structure and Room-Temperature Recrystallization of Electrodeposited Copper. Electrochem. Solid-State Lett. 6(7): C100-C102 Walther, D.; Gross, M.E.; Evans-Lutterodt, K.; Brown, W.L.; Merchant, M.Oh.S. and Naresh, P. (2001). Texture, Microstructure, and room temperature Recrystallization in electroplated Copper for advanced Interconnects. In: Mater. Res. Soc. Symp. Proc. April 23 - 27, San Francisco, Ca. 612: D10.1. West, A.C. (2000). Theory of Filling of High-Aspect Ratio Trenches and Vias in Presence of Additives. J. Electrochem. Soc. 147(1): pp.227-232 West, A . C ; Mayer, S. and Reid, J. (2001). A Superfilling Model that Predicts Bump Formation. Electrochem. Solid-State Lett. 4(7): C50-C53 Zener, C.(1948). Private communication to Smith C S . (1948) ,Trans. Metall. Soc. A.I.M.E. 175, pp.15  103  Zhu, M . ; Papapanayiotou, D. ; Lee, Y . and Ting, C. H . (1999). Recent advancements in gap filling Cu electroplating technology. In Electrochemical Technology Applications in Electronics III. Electrochemical Society Proceedings. Madore, C ; Osaka, T.; Romankiw, L. T. and Yamazaki, Y . (Ed). Honolulu, Hawaii, Fall 1999 PV 99-34: 38-43  104  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0079153/manifest

Comment

Related Items