<%BANNER%>

Chemical Vapor Deposition and Atomic Layer Deposition of Ta-Based Diffusion Barriers Using Tert-Butylimido Tris (Diethyl...

HIDE
 Title Page
 Dedication
 Acknowledgement
 Table of Contents
 List of Tables
 List of Figures
 Abstract
 Introduction
 Background and literature...
 Experimental setup and reactor...
 Effect of NH3 addition on TaN MOCVD...
 Ultra-thin ALD TaN films using...
 Metal organic atomic layer deposition...
 Pore sealing treatments of low-k...
 References
 Biographical sketch
 

PAGE 1

CHEMICAL VAPOR DEPOSITION AND AT OMIC LAYER DEPOSITION OF TaBASED DIFFUSION BARRIERS USING TERT-BUTYLIMIDO TRIS(DIETHYLAMIDO) TANTALUM METAL ORGANIC PRECURSOR By KEECHAN KIM A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2006

PAGE 2

Copyright 2006 By KeeChan Kim

PAGE 3

This document is dedicated to the graduate students of the University of Florida.

PAGE 4

iv ACKNOWLEDGMENTS My thesis is the result of five years hard work whereby I have been accompanied and supported by many people. I am very please d to have the opportunity to express my gratitude to each of them. The first person I would like to thank is my direct supervisor, Dr. Tim Anderson. I have been under his guidance since 2001 wh en I started my Ph.D. study. During these years I have known him as an encouraging and inspiring professor. Especially, his way of supervising students to become independent thinking that allowed me to learn how to initiate, plan, execute, analyze and summarize the research myself. This experience will give me a great strength th roughout my career. I owe him much gratitude for showing me this approached research. I could not have accomplished as much without his unconditional belief and support of me. Besides being an exce llent supervisor, he took care of me in many aspects like my father w ould have done. I am really glad that I have come to know him in my life and very proud that he was my supervisor. I would like to thank the other members of my Ph.D. committee who monitored my work and provided me with valuable commen ts during my studies and this dissertation. Dr. Lisa McElwee White, Dr. Fan Ren, and Dr. Cammy Abernathy, I thank them all. My colleagues of the ALD project each gave me their full support: DoJun Kim, OhHyun Kim and Hiral Ajimera, many thanks for your support.

PAGE 5

v This research has been supported and funded by NSF-CRC grant CHE-0304810. I am also grateful for the MAIC (Major Analytical Instrument Center) in Materials Science and Engineering Department for providing an excellent characterization environments and training. Many thanks go to Eric Lamber s, Valentin Craciun, and Kerry Siebein for their cheerful assistance. I am very grateful to my wife, Kathy, for her love and patience during my graduate studies. One of the best experiences during this period was raisi ng our daughter, Trudy Kim, who provided an additional and joyful dimension to our life mission. Finally, I would like to thank God, our fath er, for all of the graces He gave me while I was studying in Gainesvi lle. The past five years was the most important times in my life, because He led me to himself and i nvaluable friends that I could have missed in other places.

PAGE 6

vi TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES...........................................................................................................viii LIST OF FIGURES...........................................................................................................ix ABSTRACT.....................................................................................................................xi ii CHAPTER 1 INTRODUCTION........................................................................................................1 Statement of Problem...................................................................................................1 2 BACKGROUND AND LI TERATURE REVIEW......................................................7 2.1 Cu Metallization.....................................................................................................7 2.2 Superiority of Ta/TaN Bi-lay er as a Cu Diffusion Barrier...................................15 2.3 TaN CVD..............................................................................................................17 2.4 TaN ALD..............................................................................................................20 2.5 Ternary Ta Nitrides...............................................................................................23 2.6 Integration of Low-k Materials on Cu Interconnects...........................................25 3 EXPERIMENTAL SETUP AND REACTOR SIMULATION USING CFD SOFTWARE...............................................................................................................29 3.1 Reactor Design and Setup.....................................................................................29 3.2 Velocity and Temperature Profile Simu lation of Reactor using CFD Software..38 4 EFFECT OF NH3 ADDITION ON TaN MOCVD USING TBTDET.......................46 4.1 Introduction...........................................................................................................46 4.2 Experimental Details............................................................................................49 4.3 Results and Discussion.........................................................................................50 4.3.1 Resistivity and Chemical Composition......................................................50 4.3.2 Microstructure and Surface Morphology Analysis....................................52 4.3.3 Deposition Characteristics and Density Analysis.......................................54 4.3.4 Nucleation Step..........................................................................................57

PAGE 7

vii 4.3.5 Diffusion Barrier Performance...................................................................58 4.4 Conclusions...........................................................................................................61 5 ULTRA-THIN ALD TaN FI LMS USING TBTDET AND NH3 FOR Cu BARRIER APPLICATIONS......................................................................................63 5.1 Introduction...........................................................................................................63 5.2 Experimental Details............................................................................................66 5.3 Results and Discussion.........................................................................................67 5.3.1 Confirmation of Self-saturation ALD Growth...........................................67 5.3.2 Process Temperature Window....................................................................70 5.3.3 Growth Characteristic of TaN ALD...........................................................72 5.3.4 Diffusion Barrier Test of Ultra-thin TaN...................................................75 5.4 Conclusions...........................................................................................................78 6 METAL ORGANIC ATOMIC LAYER DEPOSITION OF Ta-Al-N USING TBTDET, TEA AND NH3.........................................................................................80 6.1 Introduction...........................................................................................................80 6.2 Experimental Details............................................................................................82 6.3 Results and Discussion.........................................................................................83 6.3.1 Growth Rate and Chemical Composition...................................................83 6.3.2 Chemical Bonding States...........................................................................93 6.3.3 Comparison of Microstructure a nd Diffusion Barrier Performance...........98 6.4 Conclusions.........................................................................................................102 7 PORE SEALING TREATMENTS OF LOWCDO FILMS TO PREVENT Ta PRECURSOR PENETRATION DURING TaN ALD.............................................104 7.1 Introduction.........................................................................................................104 7.2 Experimental Details..........................................................................................107 7.3 Results and Discussion.......................................................................................108 7.3.1 Chemical Bonding and Cont act Angle of CDO Films.............................108 7.3.2 Density and Surface Roughness of CDO Films.......................................111 7.3.3 Penetration of Ta precursor into CDO Films...........................................113 7.4 Conclusions.........................................................................................................116 LIST OF REFERENCES.................................................................................................118 BIOGRAPHICAL SKETCH...........................................................................................125

PAGE 8

viii LIST OF TABLES Table page 2-1 Interface products between Cu and contact materials..............................................11 2-2 Evaluation factors for Cu/Cu liners/ILD..................................................................17 2-3 Resistivity and carbon conten t of TaN from MO source.........................................19 2-4 Properties of dielectrics............................................................................................27 3-1 Properties of nitrogen gas.........................................................................................42 3-2 Boundary conditions................................................................................................44 4-1 Summary of film proper ties and diffusion barrier test results of TaN deposited with and without NH3...............................................................................................62 6-1 Estimated thickness of AlN contributing to Ta-Al-NA film thickness as a function of TEA exposure time................................................................................87 7-1 Contact angle of as deposited, O2 plasma treated, SiN capped CDO films, and SiO2........................................................................................................................111

PAGE 9

ix LIST OF FIGURES Figure page 2-1 Interconnect metallization in memory and logic device............................................8 2-2 Time delays caused by interconnect a nd gate at different feature sizes.....................9 2-3 Schematic diagram of single damascene and dual damascene process....................13 2-4 Schematic diagram of Cu electroplating..................................................................14 3-1 Process flow diagram of ALD reactor system.........................................................30 3-2 CAD drawing of reactor inlet part with shower head plate attached.......................31 3-3 CAD drawing of reactor body..................................................................................32 3-4 CAD drawing of reactor downstream......................................................................33 3-5 Schematic diagram of heater assembly....................................................................34 3-6 Manifold structure of pneumatic valves...................................................................37 3-7 Control panel of ALD growth programmed with LaBView....................................38 3-8 Control volume explaining the discretiz ation of a scalar transport equation...........41 3-9 Triangular grid mesh of ALD reactor......................................................................42 3-10 Contour of velocity magnitude (m/s) and velocity vector colored by velocity magnitude around heater area (m/s).........................................................................44 3-11 Color filled contour of static temp erature (K) and contour line of static temperature (K) around heater area..........................................................................45 4-1 Chemical structure of TBTDET...............................................................................48 4-2 Resistivity, and nitrogen, carbon and oxyge n content relative to Ta in TaN films as a function of ammonia flow rate..........................................................................50 4-3 Proposed transamination reaction mechanism between TBTDET and NH3............51

PAGE 10

x 4-4 XRD patterns of TaN films deposited at 300 C as a function of ammonia flow rate........................................................................................................................... .53 4-5 Surface morphology of TaN films deposited without NH3 at 300 C and with NH3 (30 sccm)..........................................................................................................53 4-6 Growth rate (/min) vs. reciprocal growth temperature with different NH3 flow rates of 0 to 150 sccm...............................................................................................55 4-7 Cross-sectional SEM images of TaN films deposited with various NH3 flow rates at 400 C...........................................................................................................56 4-8 XRR profiles of TaN films deposited at 400 C with different NH3 flow rates.......57 4-9 AFM scans of the initial stages of deposition from TBTDET only and TBTDET with NH3 = 50 sccm at 350 .................................................................................58 4-10 XRD spectra of Cu/TaN/Si structures annealed at 500 C for one hr for films grown with different NH3 flow rates........................................................................59 4-11 SEM images of Si surface after anneali ng the structure Cu/TaN/Si at 500 C for 1 hr, followed by the removal of Cu and TaN layers, and then etch in Secco solution to reveal etch pits if Cu penetration occurred.............................................60 5-1 XRR-profiles of ALD-TaN as a func tion of TBTDET exposure time for films deposited at 300 C for 80 cycles, and 10 sec purges and NH3 exposure................68 5-2 Growth rate of ALD-TaN as a func tion of TBTDET exposure time deposited at 300 C, and 10 sec purges and NH3 exposure..........................................................69 5-3 Growth rate of ALD-TaN as a func tion of TBTDET exposure time deposited at 250 C, and 10 sec purges and NH3 exposure..........................................................69 5-4 XRR profiles of ALD-TaN as a function of growth temperature deposited with 9 sec exposure of TBTDET and 80 cycles, and 10 sec purges and NH3 exposure.....71 5-5 Thickness of ALD-TaN f ilms as a function of growth temperature deposited with 9 sec exposure of TBTDET and 80 cycles, and 10 sec purges and NH3 exposure...................................................................................................................72 5-6 XRR profiles of ALD-TaN as a functi on of cycle number deposited at 9 sec TBTDET exposure, 10 sec purges and NH3 exposure at 300 C.............................74 5-7 Thickness of ALD-TaN as a function of cycle numbers deposited at 9 sec TBTDET exposure, 10 sec purges and NH3 exposure at 300 C.............................74 5-8 AFM scans (z: 3.0 nm / div) of the surface of TaN deposited at 9 sec TBTDET exposure and 300 C using 5, 10 and 15 cycles........................................................75

PAGE 11

xi 5-9 XRD patterns of Cu/TaN/Si annealed at 500 C for 30 min....................................77 5-10 Cross-sectional TEM micrograph of Cu/TaN/Si structure. TaN grown by ALD for 15 cycles at 9 sec TBTDET exposure time and 300 C......................................77 5-11 SEM images of Si surface after annealing Cu/TaN/Si at 500 C for 30 min followed by the removal of Cu and TaN, and Secco etching...................................78 6-1 XRR profiles of ALD AlN as a functi on of TEA exposure time deposited at 300 C and 10 sec purges and NH3 exposure for 120 cycles..........................................85 6-2 Growth rate of AlN as a function of TEA exposure time depo sited at 300 C and 10 sec purges and NH3 exposure for 120 cycles......................................................86 6-3 XRR profiles of Ta-Al-N grown with Ta-Al-N sequence (A) as a function of TEA exposure time grown at 300 C and 10 sec purges and NH3 exposure for 120 cycles.................................................................................................................88 6-4 Growth rate and chemical composition of Ta-Al-N grown with Ta-Al-N sequence (A) as a function of TEA exposure time grown at 300 and 10 sec purges and NH3 exposure for 120 cycles.................................................................88 6-5 XRR profiles of Ta-Al-N grown with Al-Ta-N sequence (B) as a function of TEA exposure time grown at 300 C and 10 sec purges and NH3 exposure for 120 cycles.................................................................................................................90 6-6 Growth rate and chemical composition of Ta-Al-N grown with Al-Ta-N sequence (B) as a function of TEA exposure time grown at 300 and 10 sec purges and NH3 exposure for 120 cycles.................................................................91 6-7 XRR profiles of Ta-Al-N as a function of TEA exposure time grown with Ta-NAl-N sequence (C) at 300 with 10 sec purges and NH3 exposure for 120 cycles........................................................................................................................9 2 6-8 Growth rate and chemical compos ition of Ta-Al-N grown with Ta-N-Al-N sequence (C) as a function of TEA exposure time grown at 300 and 10 sec purges and NH3 exposure for 120 cycles.................................................................93 6-9 XPS spectra of Ta 4f core level for Ta -Al-N deposited with sequences A, B, and C and TaN at 300 and 120 cycles after 1 min sputtering....................................95 6-10 XPS spectra of N 1s core level for Ta -Al-N deposited with sequence A, B, and C and TaN at 300 and 120 cycles after 1 min sputtering....................................97 6-11 XPS spectra of Al 2p core level for Ta -Al-N deposited with sequence A, B, and C at 300 C and 120 cycles after 1 min sputtering...................................................98

PAGE 12

xii 6-12 XRD patterns of Ta-Al-N grown w ith sequence A as a function of TEA exposure time at 300 C and 120 cycles..................................................................99 6-13 Cross-sectional TEM micrographs of TaN and Ta-Al-NA deposited at 300 C with SAD patterns..................................................................................................100 6-14 XRR profiles of TaN, Ta-Al-NA and AlN deposited at 300 C for 120 cycles......101 6-15 XRD patterns of Cu/TaN and Cu/TaAl-N/Si structures annealed at 500 for time as noted...........................................................................................................102 7-1 FTIR spectrum of as deposited CDO film on Si....................................................110 7-2 XPS spectra of as received, O2 plasma treated, and SiN capped CDO films.........110 7-3 XRR-profiles of as deposited, O2 plasma treated, and SiN capped (25 and 50 nm) CDO films.......................................................................................................112 7-4 Surface roughness of (a) as received, (b) O2 plasma treated, and (c) SiN capped (25 nm) CDO films z: 10 nm div...........................................................................113 7-5 Cross sectional TEM images of (a ) TaN/as-deposited CDO/Si, (b) TaN/O2 plasma treated CDO/Si, and (c) TaN/ SiN capping layer/CDO/Si structure superimposed with line EDX Ta signal..................................................................116

PAGE 13

xiii Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy CHEMICAL VAPOR DEPOSITION AND AT OMIC LAYER DEPOSITION of Ta BASED DIFFUSION BARRIERS USING TERT-BUTYLIMIDO TRIS(DIETHYLAMIDO) TANTALUM METAL ORGANIC PRECURSOR By KeeChan Kim December 2006 Chair: Timothy J. Anderson Major Department: Chemical Engineering Ta-N and Ta-Al-N Cu diffusion barrier s were deposited by chemical vapor deposition (CVD) and atomic layer deposition (ALD) using tert -butylimido tris(diethylamido) tantalum (TBTDET)/ tri -ethyl aluminum (TEA) metal organic precursors. The effect of NH3 addition on film properties during TaN CVD from TBTDET was examined. As the NH3 flow was increased at constant TBTDET flow, the film density, nitrogen content, and grai n size increased, while resistivity and carbon content decreased as compared to films deposited with TBTDET alone. These property changes are attributed, in part, to tran samination reaction between the diethylamido ligands in TBTDET and NH3. The higher film density and nitrogen content produced TaN films that exhibited s uperior diffusion barrier performance compared to those deposited without NH3 addition. TaN was also successfully deposited by ALD with alternating exposure to TBTDET and NH3. TBTDET adsorption was shown to be self-limiting with a single monolayer growth rate of 2.6 /cycle over the process temperatur e window of 200 to 300

PAGE 14

xiv C. An incubation period exists during th e initial cycles as evidenced by a non-linear relationship between film thickness and cycle number. Ultra-thin ALD-TaN layers, as thin as 38 effectively blocked Cu di ffusion during a 30 min anneal at 500 C. Ternary Ta-Al-N films were deposite d from TBTDET and TEA to promote formation of an amorphous film and increasi ng the recrystallization temperature. The Al mole fraction was linearly dependent on th e TEA exposure time suggesting growth was self-limiting. Although Al insertion into TaN pr omoted an amorphous structure, it also lowered the overall film density. A co mparative study of the diffusion barrier performance showed that failure occurred fo r both TaN and Ta-Al-N films at the same thickness, suggesting the increased amorphous content by adding Al was offset by the lower film density. Selecting a different re actant exposure sequence produced different film properties. A higher oxygen content wa s observed along with lower growth rate when the Al Ta N sequence was employed, compared to films deposited with the Ta Al N sequence. The compatibility of ALD TaN barriers on porous lowCDO films was investigated. Two different surface treatments (i.e., O2 plasma exposure and SiN capping) were applied to the surface of a low-k f ilm before TaN deposition to prevent Ta penetration into the pores of the lowfilms. Ta precursor infiltration was detected into the dielectric film of the untreated surface, while the O2 plasma treated and SiN capped lowfilms showed no diffusion of Ta precursor due to the densification of the near surface region of treated lowfilms.

PAGE 15

1 CHAPTER 1 INTRODUCTION Statement of Problem As the feature size of integrated circu its shrinks to nano dimensions, the total circuit delay is more dominated by the interconnect RC time delay than the intrinsic device delay [Vie99]. For this reason, copper has been replacing the aluminum alloy as the interconnect metal. C opper has a lower resistivity and higher electromigration resistance, resulting in fast er device switching speeds. Copper, however, is known as a very fast diffuser in Si and SiO2, forming deep traps and copper silicide compounds [Bra03], which can cause serious problems su ch as an increase in contact resistance, change in barrier height and leaky p-n junction in device. Th erefore Cu diffusion barriers, which prevent Cu diffusion into Si or SiO2, are needed between Cu and Si or low dielectric. Various transition metals (Ta, Ti, W) and metal nitrides ha ve been investigated as candidate Cu diffusion barriers for last decades [Bec03, Cho99, Dub94, Lee98, Mus96, Par96, Raa93, Sun94, Tsa95a, Tsa95b]. Extensive research on these materials demonstrated that a bilayer of Ta and its nitride is well suited as a liner for Cu damascene. This configuration gives the highe st Cu reliability, optimized adhesion for Cu electromigration resistance, robustness of vi a interfaces and redundant current strapping for added chip re liability [Ede02]. Most Ta/TaN layers employed so far in industry have been deposited by Physical Vapor Deposition (PVD). Because of its inherent shadowing effect, this approach has a serious limitation in achieving conformal cove rage in devices with submicron features

PAGE 16

2 and high aspect ratio contacts and via hol es. Therefore, Chemical Vapor Deposition (CVD) has received recent attention owing to its superior conformality over PVD grown barriers. There are two different approaches to TaN-CVD depending on the metal source used. TaN-CVD using halide sources such as TaCl5 [Hie74] and TaBr5 [Che97] is known to require high deposition te mperature to obtain low film resistivity and halide incorporation, which is inappropriate for Back End of Line (BEOL) processing. Cl or Br incorporation in the film is an issue because the halides in the film reduce the adhesion strength and enhance the electromigration, wh ich can be major concerns for long term device reliability [She90, Yok91]. The other approach for TaN-CVD is using Metal Organic (MO) sources. This method can sign ificantly lower the de position temperature with good conformality as compared to halide source. Films deposited with MO source alone, however, tend to show high carbon content and accordingly higher film resistivity. In this research, (NEt2)3Ta=NBut [ tert -butylimido tris(diethylamido) tantalum, TBTDET] is investigated as a single source precursor as well as paired with ammonia for TaN-CVD. Previous studies on TaN-C VD using TBTDET [Tsa 95a, Tsa95b, Tsa96a] showed the results on TaN film properties deposited with TBTDET as a single source through the thermal dissociati on reaction. Since TBTDET has a N/Ta ratio of four with the Ta bonded to four N atoms, TaN films can be grown with this single precursor by CVD. The films deposited with TBTDET in a carrier gas, however, possess the large amount of carbon impurity reaching up to 30 atom %, ascribed to the diethyl amido or tbutyl imido ligands in TBTDET. Although car bon impurity helps to increase amorphous structure, it causes the film resistivity to in crease and thus making an unsuitable barrier.

PAGE 17

3 In this program, NH3 was tested as an additional nitrogen source in the deposition chemistry. This hypothesized that NH3 will undergo a transamination reaction with the carbon containing nitrogen ligands (i.e., diethyl amido and tbutyl imido ligands). This transamination reaction should lead to lowe r carbon impurity and thus lower the film resistivity compared to using TBTDET al one. Chapter 4 presents the results of experiments on the effect of adding NH3 on the TaN film properties by measuring resistivity, chemical composition, microstr ucture, surface morphology, and film density as a function of ammonia flow. Difference in deposition rate with growth temperature was identified for TaN films deposited w ith TBTDET as a single source and TBTDET with added NH3. The Cu diffusion barrier performance was tested too. Although the ITRS road map calls for CVD of barriers in the near future, atomic layer deposition (ALD) will likely emerge in the longer term as the dominant growth method in Cu diffusion barrier deposition be cause of its superior conformality and accurate thickness control compared to other deposition methods. As the barrier thickness drops below 100 standard CVD w ill approach its usability limit due to difficulties in controlling the deposition rate. The highly conformal, ultra thin barriers afforded by ALD will be essential to minimi ze the barriers impact on the resistance per unit length in Cu interconnects [Kap02]. Early reports on TaN-ALD used a halide source (TaCl5 [Hil88a, Rit99] or TaBr5 [Ale01, Ale02]) with NH3, produced polycrystalline Ta3N5 films, which made poor Cu diffusion barriers because of high grain boundary structures. High film resistivity and high Cl or Br residue deposited at low temper ature are other issues for halide-based TaN ALD. Therefore, MO source based TaN ALD is being considered as more suitable than

PAGE 18

4 halide-based TaN ALD. Although there have be en many reports of MO based TaN ALD, their main focus has been verification of self -limiting growth conditions i.e. the region of operation in which the precursor adsorption is restricted to monolayer or fractional monolayer coverage to give a gr owth rate that is constant with respect to exposure time of the reactants. The process temperature wi ndow of ALD behavior, in which the growth rate remains constant in range of depositi on temperature, has rarely been reported for MO-based TaN ALD, while the process te mperature window for halide-based TaN has been reported [Rit99]. In addition, most Cu diffusion barrier perf ormance test using ALD-TaN have been evaluated with thic k films (> 10 nm), although ITRS roadmap clearly indicates that ultra-thin barriers will be needed in the near future. In the TaN-ALD experiments reported here, TaN was deposited with the TBTDET/NH3 sequences. Unlike previous reports of TaN-ALD, XRR (X-ray Reflectivity) technique was used to accur ately measure the TaN film thickness as a function of TBTDET exposure time to locate the self-limiti ng growth zone. The process window temperature was determined for this precursor combination for the first time by measuring the growth rate as a function of growth temperature in the range of conditions that give self-limiting growth. Additionally, the growth characte ristic of the initial stage of TaN-ALD was observed using XRR and AFM, focusing on the initial complete coverage of TaN on Si. As previously men tioned, minimum thickness of ALD-TaN film that provides a sufficient barrier for Cu diffusi on has been rarely reported. The effective minimum thickness of ALD-TaN barrier was ve rified by performing Cu diffusion barrier tests on various ultra-thin (7 to 100 ) ALDTaN films and the results are discussed in depth in Chapter 5.

PAGE 19

5 Main drawback of binary nitrides as Cu di ffusion barrier materials, especially with TiN, is its tendency to form polycrystalline w ith columnar microstructure or recrystallize, leading to significant Cu gr ain boundary diffusion [Nic95]. Th e grain structure creates boundaries that cut across the film and offer fast diffusion paths for Cu. Adding a third element to a binary transition metal nitride matrix tends to disrupt the microstructure, increasing the chance of amorphous phase form ation [Ist00, Ram00], wh ich is a superior structure for Cu diffusion barriers. In chapter 5, as-deposited binary TaN films deposited by ALD contain nanocrystal or TaN with a small distribution of h-TaN phase imbedded in an amorphous matrix. To promote amorphous material, triethyl aluminum (TEA) was inserted into TaN binary film as a third element to deposit Ta-Al-N tern ary films by ALD. Chapter 6 presents a comparison of the Cu diffusion barrier perfor mance using either TaN or Ta-Al-N films deposited with the same thickness. The micr ostructure and film density change caused by Al insertion is also examined. In additi on, it is well known that different sequence exposure of precursors induces the change in film properties in ternary ALD films. Three different exposure sequences (Ta Al N, Al Ta N and Ta N Al N ) were tested and film thickness, chemical com position, and bonding states are compared. While TaN and Ta-Al-N deposition was on Si substrates in most metallization schemes incorporate chapter 4, 5 and 6, lo w-k Carbon Doped Oxide (CDO) films were employed as a substrate for ALD-TaN in Chapter 7. As mentioned previously, Cu metallization was first introduced to diminish the resistance of me tal interconnects, and then low-k dielectrics were developed to reduce the parasitic ca pacitance, further decreasing the RC time delay. One of the most important changes in sub-65 nm node and

PAGE 20

6 beyond is the introduction of porous dielectrics for the purpose of reducing the dielectric constant. Because of the porosity, however, there exists a higher probability of contamination trapping or low-k material de gradation during diffusi on barrier deposition. Chapter 7 demonstrates the results of two different surface treatments applied to the surface of CDO films before TaN barrier de position. These treatments are intended to prevent Ta precursor penetration into the porous structure of COD films. The first treatment was O2 plasma exposure, which is expected to densify the top layers by sealing the pores on the surface of CDO films. The second treatment was SiN capping, where SiN capping layer was deposited on CDO fi lms, which is intended to prevent the precursor infiltration by closing the pores near the surface of the CDO. The contact angle (CA), surface roughness, chemical bonding states, and film density of CDO films after the surface treatments were investigated using CA meter, AFM, XPS, and XRR, respectively. The efficiency in blocking Ta precursor diffusion was examined with TEMEDX after TaN-ALD on su rface treated CDO films.

PAGE 21

7 CHAPTER 2 BACKGROUND AND LITERATURE REVIEW 2.1 Cu Metallization Metallization is the key process where the various device structures fabricated on the silicon substrate are elec trically interconnected through the metal and metal alloy layers [Sil06]. Numerous regions of each circuit element such as MOSFET (MetalOxide-Semiconductor Field Effect Transisto r), Bipolar Transistor and Capacitor are properly interconnected during the metallization step to distri bute clock and other signals and to provide power/ground, to and among, the various circuits/systems functions on a chip. Metallization processing represents more than 70 % of the IC (Integrated Circuits) fabrication steps [Bra03]. Figure 2-1 (a) shows the cross sectio nal diagram of DRAM (Dynamic Random Access Memory) and logic devices, which demonstrates that the interconnect metallization occupies a large cross-sectional area for both devices [Bra03]. As shown in Figure 2-1 (b) a diagram of a CMOS structure, metals are employed in many locations of the CMOS device including gate metal, cont acts, interconnect laye rs in multi-level metallization, and plugs in via holes. Ge nerally, poly-Si, which has a high melting temperature (1414 ) and resistance to oxidation, ha s been used as a gate metal. Tungsten (W), however, has been mainly used as a contact and via hole plug metal while the aluminum (Al) alloy with Si has been us ed as the interconnect metal. Recently, Al metal and SiO2 dielectric combination for interco nnect metallization is being rapidly replaced by the Cu, low-k c onfiguration due to the decrease in the RC time delay.

PAGE 22

8 (a) (b) Figure 2-1. Interconnect metallization in memo ry and logic device (a) 3D diagram of metals employed in CMOS structure (b) (taken from [Bra03]) As complexity and packing density incr ease in ICs, the size of interconnect metallization decreases, which leads to increas e of interconnection de lay. The total circuit delay, which is a combination of the intrinsic device delay associated with the solid state device and the interconnection RC (resistance capacitance) time delay, is dominated by DRAM Logic device Interconnect Interconnect Capacitor Transistor/ Bit line Transistor

PAGE 23

9 the interconnect RC time delay at lower feature size. This trend is illustrated in figure 22, where both the interconnection delay and intr insic delay are plotted as a function of the feature size (channel length). It is clear that below ~0.5 feature, the interconnection delay will be the main limiter for ICs performance [Vie99]. Figure 2-2. Time delays caused by interconnect a nd gate at different feature sizes (taken from [Vie99]) The RC time delay can be represented w ith the simplistic model given below [Ben00]. w L R tw L Rs d Lw C d L R td L RCs 2 2 where, t is the thickness of metal, wis the width of metal, Lis the length of metal, dis the distance between metal layers, is the dielectric constant of dielectrics, sRis the sheet resistance of metal, and is the specific resistivity.

PAGE 24

10 As the model suggests, the metal interconnect with low resistivity and dielectric materials of low dielectric constant are n eeded to minimize the RC time delay leading to higher device switching speed, si nce copper possesses lower film resistivity compared to aluminum alloys (1.67 cm for Cu vs. 2.65 cm for Al) and low-k materials have lower capacitance than conventional SiO2. Therefore, industry is rapidly adopting copper as the interconnect metal fo r the intermediate and upper wiring levels on IC devices replacing Al alloys, and low di electric constant materials are being intensely researched for replacement of the conventi onal SiO2. Transition from Al a lloys to Cu is reported to increase overall microprocessor speed by 15%. Currently, the development and qualification of a 300 nm SiCOH BEOL (Back End of Line) integration for 65 nm bulk and SOI (Silicon on Insulator) semiconductor pr oduct application have been reported by IBM [Ang05]. Intel started shipping the 65 nm technology mode Pentium 4 processors fabricated with Cu and low-k metallization scheme [Int06]. Copper also shows the highe r electromigration resistan ce than aluminum, which leads to increase in devi ce lifetime [Mur95]. Aluminum suffers from serious electromigration at high current flow densit y, where electrons flowing through the metal interconnect pass on enough momentum to carry the metal atoms with them, leading to voids (openings) and hillocks (pileups) in the interconnect wiring. In addition to resistivity and electromigration improvements, copper is reported to provide higher IC production yield than aluminum-bas ed devices with similar design. Other advantages of switching to copper interconnects include a decrease in the nu mber of interconnect levels, a roughly 30 % decrease in pow er consumption for operation at a given frequency, and a

PAGE 25

11 cost savings of roughly 30% pe r interconnect level due to in tegration of dual damascene processing [Bch04]. Even with the above mentioned advantages of copper as interc onnect metallization, several problems are encountered with copper me tallization. It is well known that Cu is a very rapid diffuser in Si and SiO2. Diffusion of Cu into Si forms Cu3Si at temperature below 200 leading to formation of deep donor le vel at the interface Table 2-1 shows the compound formed at the interface be tween Cu and the various semiconductor materials [Bra03]. Cu diffusion in devices can cause an increase in contact resistance, change in barrier height, lea ky p-n junction and destruction of electrical connections to the chip. For long-term reliability of interconne ct metallizatio n, little or no transport of Cu through adjacent layers should be required. In addition, copper films show poorer adhesion to Si and SiO2 compared to aluminum. Thus, diffusion barriers, which prevent Cu diffusion and promote the a dhesion of Cu to Si and SiO2, should be employed between Si or SiO2 and Cu. Table 2-1. Interface products between Cu a nd contact materials (taken from [Bra03]) Contact material Annealing temperature Interface products Si < 200 Cu3Si (Deep donor level) Silicide 350 ~ 450 Punch-through Al 150 ~ 250 CuAl2 (Resistivity hike) SiO2 Bias thermal stress Punch-through Polymer Room temperature Cu precipitate Oxygen 100 Cu2O (Porous oxide)

PAGE 26

12 Although the benefits of employing copper as the interconnect metals are obvious, there are difficulties in patterning of copper. Wet etching is not appropriate for patterning sub-micron structures due to its isotropic na ture. Reactive ion etchi ng (RIE) of copper is also not practical because etch byproducts of Cu are not volatile. Therefore, instead of wet etching and RIE, Cu patterning is base d on the so-called Damascene process [Pan99]. Figure 2-3 (a) shows a schematic diagram of a single damascene process where grooves are first etched in a dielectric layer, barrier /metal deposited on it, and the metal layer is chemically-mechanically polished, leaving inla id metal lines in the oxide grooves. In contrast to the conventional patterning method, where a metal layer is first deposited and then unwanted metal is etched away, leavi ng the desired pattern of wires or vias, damascene patterning involves th e same number of steps, but in reverse order. In the single damascene process, the separate fo rmation of wires and vias has as much complexity as the conventional RIE patterning process. Dual-damascene, which is schematically described in figure 2-3 (b), make s it possible to form both wires and vias in the same metal deposition process step. In this process, the pattern of vias and wires is defined using two lithography and RIE steps, but the via plugs are filled in the same step with metal lines. Dual-damascene eliminates the complexity of the patterning process by reducing the number of proces sing steps. One major benefit of dual-damascene is less risk of contact failure be tween via and metal line.

PAGE 27

13 (a) (b) Figure 2-3 Schematic diagram of single dama scene (a) and dual damascene process (b) (taken from [Bra03]) As the multi-level metallization scheme gets more complicated, filling the via/plug of high aspect ratio with Cu without void or seam features becomes more critical. With this need, electroplating has been mainly employed as the deposition method of Cu in industry. Electroplating of Cu is perfor med by immersing a conductive surface in a solution containing ions of the Cu to be depos ited. The surface is electrically connected to an external power supply, and current is passed through the surface into the solution. Via patterning -.Via Photo.& Etch Via damascene -.BM/Cu dep. -.Via CMP Metal patterning -. Metal Photo.& Etch. Metal damascene -.BM/Cu dep. -.Metal CMP Damascene patterning -.Via Photo.& Etch -.Metal trench Photo.& Etch Metal dep. & anneal -.BM/seed-Cu dep. -.EP-Cu dep. & anneal CMP -.1st Cu CMP -.2nd BM CMP

PAGE 28

14 This causes reaction of the metal ions (Cu2+) with electrons (e-) to from metal (Cu): [Shy99] Cu2+ + 2e= Cu Figure.2-4 shows a schematic of Cu electroplatin g process. The wafer is typically coated with a thin conductive layer of copper (physically vapor deposited (PVD) Cu seed layer) to provide a conductive electrode for the electroplating process. This is then immersed in a solution containing cupric ions Electrical contact is made to the seed layer, and current is passed such that the reaction Cu2+ + 2e= Cu occurs at the wafer surface. The wafer, electrically connected so that Cu2+ is reduced to Cu, is referred to as the cathode. Another electrical ly active surface, known as the anode, is present in the conductive solution to complete the electrical circuit. At th e anode, an oxidation reaction occurs that balances the current flow at the cathode, thus maintaining electrical neutrality in the solution. All cupric ions removed from solution at the wafer cathode are replaced by dissolution from a solid copper anode. Figure 2-4. Schematic diagram of Cu electroplating (taken from [Bra03])

PAGE 29

15 2.2 Superiority of Ta/TaN Bi-laye r as a Cu Diffusion Barrier For most metals to be used as interconnect materials, a thin diffusion barrier should be used between the metals and dielectrics, si nce many metals form deep levels in Si and some including Cu have high diffusivities. An additional function of the diffusion barrier is to serve as an adhesion promoter for Cu and the interlayer dielectric. If the diffusion barrier materials do not adhere to the underlying layer or inte rconnect material, or if Cu does not adhere to the diffusion barrier, the result will be gross delamination and failure, which typically occurs during CMP steps used to planarize the de posit following plating. More importantly, poor adhesion causes se vere reliability problem such as electromigration. The second important quality of a Cu diffusion barrier is to have a high in-plane electrical conductivity (low resistivity), to support as high as current density as possible without producing excessive Joule hea ting, which can cause reactions that gives the device failure. This can add redundant reliability beyond the nominal electromigration limits and help save chips from open-circuit fa ilures in the event of defects or abnormal wear out [Ede02]. Other barrier requirements for Cu diffusion include prevention of Cu diffusion; low to moderate grow th temperature range (350 to 400 ) for the compatibility with low-k material; excellent st ep coverage of high as pect ratio structures and extremely thin thickness (<5 nm) [Kim05]. The nitrides and silicon nitrides of Ta Ti and W are known to be good candidates for reliable diffusion barriers with respect to interdiffu sion requirement and provide lower electrical resistiv ity compared to their pure metal counterparts. Table 2-2 shows the evaluation factors for Cu diffusion barriers fo r Cu Damascene liner application [Ede02]. In particular, two requirements of a diffusi on barrier, namely adhesion and resistivity,

PAGE 30

16 eliminate many candidate materials. Cu pois oning and barrier furt her eliminate all but TiN/Ta and TaN/Ta bilayers. TiN/Ta requires two chambers, and TiN was found to cause Cu corrosion in some CMP processes [Ede02] In addition, TiN was unsuccessful for Cu metallization due to grain boundary formation in the film leading to severe interdiffusion of Cu through the boundary. Plasma treatment and exposure to SiH4 have been chosen to enhance the barrier quality, but couldnt fully resolve the issue.[Ede02] Clearly, TaN/Ta is the best choice among the candidates listed, a nd the only one liner material to meet all criteria. Therefore, the TaN/Ta bilayer is cu rrently being employed for current Cu interconnect metallizatio n in industry. The bi-layer stru cture is employed to optimize the adhesion properties. For dual-Damascene integration in SiO2, Ta lacks adequate adhesion to SiO2, whereas TaN/SiO2 adhesion is excellent. On the other hand, Cu/TaN adhesion is relatively poor. The liner/ILD (Inter Layer Dielectric) and Cu/liner adhesion have conflicting dependency on the N content of TaNx. Significantly, the TaN/Ta liner has very low in-plane resistivity, because when Ta is deposited on a TaN surface, the lowresistivity -phase Ta is spontaneously formed with a resistivity in the range of 15 to 25 cm. Another benefit of TaN/Ta liner selec tion is better step coverage than other lower mass metals such as Ti, even for uncollimated PVD Ta and TaN. Due to the high mass of Ta, it sputters more di rectionally from the target, and its high momentum gives it more surface mobility to redistribute into the features. For further improvements in step coverage, an ionized PVD (I-PVD) process has been developed for bilayer liner [Ede02]. The TaN/Ta liner material scheme has demonstrated excellent Cu integration performance in high volume manufacturing, in cluding: high yield, high step coverage,

PAGE 31

17 mechanical, thermal and chemical stabilit y, consistently low resistivity, and high performance CMP. This bilayer liner pr ocess satisfies high-volume manufacturing criteria, yield, control, low cost and low maintenance [Ede02]. Table 2-2. Evaluation factors for Cu/C u liners/ILD (taken from [Ede02]) Attribute Cr TiN TiN/Ti Ti/TiN TiN/Ta TaN Ta TaN/Ta TiSiN WN Cu barrier X O O O O O O O O O Adhesion to ILD O O O O O O X O O O Cu on liner adh. O X O X O X O O X ? Liner on Cu adh. O ? ? O ? O O O O ? Low in-plane R O X O O ? X X O X ? Cu poisoning O O X O O O O O ? O CMP ? X X X X/O O O O O ? Single chamber O O O O X O O O O O Via, Contact R O O X ? O O O O O ?/O Contact R O O O O O O O O O ? Cu corrosion ? X X X X/O O O O O ? Thermal stability ? O X X O O O O ? O Stress, cracking X O O O O O O O O O Step coverage ? Final X X X X X X X O X ? = CVD available = Ionized-PVD availabl e ? = Not evaluated 2.3 TaN CVD As previously mentioned, TaN/Ta bilaye r is the most prominent liner for Cu metallization in industry. It is most ofte n deposited by Physical Vapor Deposition (PVD). PVD, however, has a serious limitation, i.e., poor conformality in small feature size (< 0.35 nm) and high aspect ratio features, projecte d to reach its usability limit at the 45 nm node in 2007 [Han03]. The poor conformality i nherent to all PVD processes is caused by the directionality imparted to the atoms/clus ters traveling toward the substrate [Lee93].

PAGE 32

18 This directionality leads to step shadowing, where parts of the substrate surface are not seen by the incoming sputter atoms. Step shadow ing results in little or no film coverage on certain sections of the substrate. These exposed substrate areas are then vulnerable to reaction with subsequently deposited Cu atom s. Therefore, CVD of barrier has received attention because of its superior conformality on high aspect ratio structures over sputterdeposited barriers. There are two different approaches to TaN-CVD process depending on the metal sources employed. The first approach us es the halide sources such as TaCl5 [Hie74] and TaBr5 [Che97]. In general, TaN-CVD from a metal halide source is known to require high deposition temperature. For example, TaN using TaCl5, N2 and H2 is grown at a temperature > 900 to obtain the low resistivity and halide content [Hie74]. This temperature, however, is inappropriate for the subsequent IC processes. Although lower growth temperature (450 ) for TaN-CVD from TaBr5, NH3 and H2 has been reported, but still not low enough for a low-k material process [Che97]. Additionally, particle formation and Cl or Br content (0.5 to 4.5 atom %) in the film are serious issues for halide-based TaN-CVD. Especially, the halide content in the film reduces the adhesion strength and causes the delamination of the f ilm, which can be a severe problem for long term device reliability [She90, Yok91]. The other approach for TaN-CVD is to employ Metal Organic (MO) sources such as Ta(NMe2)5 [pentakis(dimethylamido) ta ntalum, PDMAT] [Fix93], Ta(NEt2)5 [pentakis(diethylamido) tantalum PDEAT] [Cho98, Cho99] and (NEt2)3Ta=NBut [tertbutylimido tris(diethylamido) tantalum, TBTDET] [Tsa95a, Tsa96b]. Although the halide

PAGE 33

19 residue in TaN is not an issue for MO s ource-based TaN CVD, it tends to produce high carbon concentration, leading to high film resistivity as a result (Table 2-3). Table 2-3. Resistivity and carbon co ntent of TaN from MO source MO source for TaN CVD Resistivity ( cm) Carbon content (atom %) Growth T ( ) NH3 as an additional N source PDMAT > 106 20 200 to 400 10 % NH3 in He PDEAT 1.2 to 6.0 X 105 30 to 1 300 to 375 0 to 25 sccm TBTDET 1.0 X 105 23 500 No NH3 TaN from PDEAT had high film resistivity (up to 60000 cm) with high carbon content (~ 30 atom %) when it was deposited from PDEAT single source. The addition of NH3 as an additional nitrogen source, however, cau sed the decrease in resistivity down to 7000 cm. It lowered the carbon content from 30 to 1 atom % as well. The growth temperature should be higher than 600 to obtain reasonable range of resistivity (<1000 cm). The grain size, film crystallinity, and film density increased with the addition of NH3. TaN from PDEAT and NH3 exhibited better Cu diffusion barrier performance compared to that from PDEAT single source because of higher film density. The step coverage of the film grown with PEDAT single source was 80 % and decreased down to 56 % with NH3 addition, which is ascribed to ma ss-transfer limited conditions when NH3 was used as an additional nitrogen source [Cho98, Cho99]. TaN deposited with PDMAT (solid state) produced the insulating Ta3N5 phase film resulting in high resistivity (>106 cm). The film also had the high carbon impurity (~ 20 atom %) even though it was deposited with NH3 as an additional N source. The

PAGE 34

20 microstructure was amorphous when deposit ed in the temperature rang of 200 to 400 [Fix93]. MO-CVD of TaN using TBTDET single sour ce has also been reported [Tsa95a, Tsa96b]. A relatively low resistivity (920 cm) could be obtained at 650C. The deposition rate was almost independent of the substrate temperature (450 to 650C), indicating this temperature range was mass transfer-controlled regime. XPS results showed that the films deposited at 600 contained 10 atom % of carbon and 5 atom % of oxygen [Tsa95a]. Comparison study of di ffusion barrier performance between PVD TaN and CVD TaN from TBTDET proved that the barrier failure occurred at 650 and 600 for PVD and CVD TaN films. This demo nstrates that the PVD TaN possesses a better diffusion barrier property due to its higher film density than CVD grown TaN [Tsa96b]. 2.4 TaN ALD ALD is a self-limiting growth method char acterized by the alternate exposure of chemical species in layer-by-layer manner wh ile CVD delivers all required reactants to the reactor simultaneously. A single precursor is exposed in the reactor at any given time, so that a uniform layer of the precursor may chemisorb to the substrate surface. The most important requirement for this step is the self-limitation for the precursor molecule adsorption process. The self-limitation mean s that it limits further adsorption of precursors by passivating the adsorption site s after the saturation coverage, roughly one monolayer or less, is reached. [Kim05] In general, self-limiting condition is satis fied by the ligands bonded to the metal atoms in the precursors, such as haloge n or organic ligands [Kim05]. Once this

PAGE 35

21 chemisorbed layer forms, the reactor is evacuated or purged with inert gas. It is important to minimize the incorporation of background impur ities into the film and also to prevent the process from shifting to CVD mode, which can occur if additional reactant from the previous exposure remains in the reactor during the subsequent exposure. A second precursor is then introduced to react with th e chemisorbed layer from the previous step, forming a new layer of material. The sec ond precursor is also self-limiting, so that reaction stops after the (sub-) monolayer of ch emisorbed material from the previous step has been consumed. Another ev acuation or purge step follows the second precursor pulse to ensure the complete removal of unreacte d reactants and byproducts, which might have formed during the second precursor e xposure [Kim05]. The highly conformal, ultra uniform barriers afforded by the self-limiting feature of ALD will be essential in the future to minimize the barriers impact on the resistance per unit length in Cu interconnects [Kap02]. The first report on ALD of TaN employed TaCl5 as metal precursor, which was reacted with NH3. The film deposited at 400 was composed of polycrystalline Ta3N5 with very high resistivity over 104 cm [Hil88a, Rit99]. The formation of Ta3N5 instead of cubic TaN, was ascribed to the low reducing power of NH3. For this particular ALD process, the growth rate was only 0.12 /cycle at 200 while it increased to 0.22 to 0.24 /cycle above 300 Meanwhile, the Cl content was above 20 % at 200 but rapidly decreased to below 0.1 % above 500 The use of DMHy as an alternative precursor of NH3 also formed high resistivity Ta3N5 films [Rit99]. A dditional reducing agents such as TMA and amines helped re duce the film resistiv ity [Ale01, Ale02]. By using Zn as an additional reducing agent, cubic TaN with low resistivity of 960 cm

PAGE 36

22 was obtained, although residual Zn incorpora tion was still a problem [Rit99]. TaN ALD using TaBr5 also produced similar results with TaCl5 [Ale01, Ale02]. As an alternate approach, the metal orga nic Ta source, TBTDET, has been reported for ALD and Plasma Enhanced ALD (PE-AL D) of TaN [Par01, Par02]. For thermal ALD, NH3 was used while for PE-ALD atomic H plasma was used as reducing agent. XRD results showed that PE-ALD TaN is co mposed of cubic TaN phase, while no peaks are evident on the as-deposited thermal ALD Ta N films. While the resistivity of thermal ALD TaN from TBTDET was ~106 cm, the resistivity from PE-ALD TaN with same precursor was as low as 400 cm, when the pulse time of H plasma reached 30 seconds. Low resistivity of PE-ALD TaN was attributed to the formation of Ta-C bond, which was not observed for thermal ALD Ta N, as confirmed by XPS and XRD [Par02]. The step coverage was excellent up to asp ect ratio of 10:1. As grown thermal ALD TaN [Str04] from TBTDET contains 5 to 8 % of carbon and oxygen with the film resistivity ranging from 500 to 1000 cm for 30 nm thick films, which is very low compared to Park et al.s result (~106 cm). It also showed excellent step coverage on 100 nm trenches patterned SiO2 with an aspect ratio of 6.5 to 11 [Str04]. PDMAT MO-source has also been reporte d for thermal TaN ALD, focusing on annealing studies of ultra-thin (<10 nm) ALD TaN [Wu04]. PDMAT and NH3 were alternately exposed for TaN deposition empl oying argon as a carrier gas of PDMAT. The growth rate was 12 /min at the deposition temperature of 275 10 nm thick TaN films contain 2 atom % of carbon and 5 atom % of oxygen, showing no significant change in the composition after annealing at 750 As deposited films showed the

PAGE 37

23 amorphous structure, however, a fcc NaCl type -TaN nanocrystalline structure was obtained, when annealed at 750 for 45 minutes. 2.5 Ternary Ta Nitrides Addition of a third element to a binary refractory metal nitride matrix tends to further disrupt the microstructure, increas ing the probability of nanocrystalline or amorphous phase formation [Ist00, Ram00]. Since grain boundaries play such an important role as diffusion pathways for c opper, the ternary phase metal nitrides are expected to be good diffusion barriers [Kim03]. The tern ary phase nitride films are reported to remain amorphous up to high annealing temperature, while the binary nitrides usually crystallize easily by annealing. Carbon, silicon, boron and aluminum are the most studied third elements added to binary metal nitride films for this purpose. TaCxNy was deposited by sputtering TaC target in an Ar/N2 atmosphere as a Cu diffusion barrier [Sun01]. TaCxNy showed better thermal stability than that of respective binary phases, and its resistance to Cu diffu sion was better than TaC because of stuffing the grain boundaries with nitrogen atom s. The films had low resistivity (~300 cm), and prohibited Cu diffusion afte r 30 min annealing at 600C. MOCVD of TaCxNy was reported using a mixt ure of PDMAT and PDEAT [Hos00]. These films prevented Cu diffusion after a 30 min anneal at 500C, but had high resistivity ( 4000 cm). PECVD of TaCxNy films was reported using PDMAT and methane as a reactive gas. Film resis tivity substantially decreased to 440 ~ 2400 cm, compared to thermal CVD of TaCxNy (6300 ~ 20000 cm). PECVD TaCxNy films of 4 ~ 10 nm thickness successfully blocked the Cu diffusion after annealing at 360 for 8 hr in H2/N2 environment [Eng02].

PAGE 38

24 There have been several reports of TaSixNy deposition as a Cu diffusion barrier application. TaSixNy was deposited by reactive sputtering of a Ta Si target in N2 [Har01] and Ar/N2 [Kol91] atmosphere. This ternary nitr ide had better barrier performance and better adhesion to Cu than a Ta/TaN dual layer barrier. Various stoichiometries of TaSixNy have been reported. Ta0.43Si0.04N0.53 films exhibited a failure temperature of 825C with high resistivity (1419 cm) [Lee99]. Amorphous Ta0.36Si0.14N0.50 films had high thermal stability, preventing inte rdiffusion between ne ighboring Cu and TiSi2 layers up to 900C, with rela tively low resistivity of 625 cm [Kol91]. The effect of nitrogen content of the TaSixNy films on barrier performance was investigated [Kim97]. Films with nitrogen content greater than 40 atom % blocked Cu diffusion after an 800C anneal, while those with lower nitrogen content failed after a 700C annealing. LPCVD of TaCl5+SiH4+NH3+H2/Ar at 500C was also used to deposit TaSixNy [Bla97]. As deposited TaSixNy showed some crystalline struct ure, with a stoichiometry of Ta0.35Si0.11N0.54, and contained Cl and O impurities ranging from 3 to 5 atom %. However, the films failed to prevent Cu diffusion after a 1 minute, 600C anneal, suggesting LPCVD TaSixNy performs no better than PVD TaN. TaSixNy PE-ALD has been performed using TaCl5, SiH4 and N2/H2 plasma. The precursor exposure sequence affected the growth rate and film properties. With a TaCl5-SiH4-plasma sequence, amorphous films with lower resistivity (< 1000 cm) were deposited, while high resistivity (> 10000 cm), polycrystalline TaSixNy was deposited from a SiH4-TaCl5plasma sequence [Kim02a]. TaxAlyN ALD deposited with the a lternate exposure of TaCl5 or TaBr5, NH3 and TMA (trimethyl Aluminum) was reported [A le01]. The pulse length of TMA did not

PAGE 39

25 have much influence on Cl and C content, which was 5 ~ 6 atom %, 20 ~ 22 atom % respectively, while it increased the Al level from 10 to 12 atom % as the TMA pulse time changed from 0.2 to 0.8 second. The change in precursor sequence caused differences in film properties, such as chemical compos ition, resistivity and growth rate. The TaCl5TMA-NH3 sequence produced less electrically resi stive films with less concentration of carbon and chlorine, compared to the TMA-TaCl5-NH3 sequence. Raising the deposition temperature from 250 to 450 increases the growth rate, ascribed to the thermal decomposition of TMA at higher temperature, while it lowers the Cl content (15 to 3 atom %) and resistivity (22000 to 5000 cm). A Cu diffusion barrier test on Cu/ TaxAlyN /Si structure showed that it failed at 600 Using TaBr5 as a different Ta source for TaxAlyN produced higher resistivity with lower deposition rate than TaCl5 [Ale01]. 2.6 Integration of Low-k Materials on Cu Interconnects As the feature sizes continue to shrink, chip performance becomes more limited by BEOL due to interconnect RC de lay and power consumption (CV2f). As previously indicated, Cu metallization was first introduced to reduce resistance of metal wirings, and then low-k dielectrics were developed to di minish the parasitic capacitance. One of the most important changes in sub 65 nm node and beyond is the introduction of porous dielectrics for the purpose of reducing the dielectric constant. Table 2-4 shows the evolution of dielectric mate rial properties which was driv en by dielectric constant reduction and also required for the next in terconnect generation [F ay03]. Organic groups were first introduced into the silicon oxide-based matrix to in duce free volume (pore size less than 1 nm) and decrease the dielectric cons tant down to three. OSG (Organo Silicate

PAGE 40

26 Glass) and CDO belong to th is generation of materials. Homogenous and larger pore sizes were developed to furthe r decrease dielectric constant of these materials with k < 2.2 by templated or porogen (Polymeric Po re Generator) removal methods. The large porosity volume close to 45% and pore size s around 3 nm are the general properties of ULK (Ultra Low k) materials reported so fa r. However, these properties can lead to serious problems in Cu/low-k integration [Fay03]. Interlayer adhesion, cohesive delamina tion, material deformation and crack formation have been reported during low k dielectric integrati on during the CMP and packaging steps [Ras02]. CMP feasibility of ULK is known to be correlated to its mechanical properties. Youngs modulus (E) larg er than 4 Gpa, hardness (H) larger than 0.5 Gpa and adhesion energy (A) larger than 5 J/m2 are required to pass the conventional CMP. However, the mechanical properties of ULK are just around the limits and even lower. (E = 4 to 6 Gpa, H < 0.8 Gpa, A = 2 to 5 J/m2) [Lin01, Sch01]. An addition of metal dummies [Tsa02], adhesion promoter [K lo02] and new CMP technique with lower shear stress [Mos01] have been reported as solutions for these mechanical problems during ULK integration. Recently, UV (Ultra Violet) and EB (Electron Beam) curing have been reported to improve significantly elas tic modulus and the adhesion strength without degrading the dielectr ic constant [Ito05, Got05]. There is a high possibility of contamina tion trapping or ULK ma terial degradation during integration process, becau se of its large porosity, whic h facilitates the diffusion of chemical species from the process. Resist stripping and diffusion ba rrier depositions are the main processes of the concern. Most st ripping processes empl oy chemicals to break organic bonds of the resist, also leading to breakage of the methyl bonds in SiOC and

PAGE 41

27 carbon depletion. The resulting SiOC films b ecome denser and more hydrophilic, causing a significant dielectric consta nt rise [Rya02]. Compatible re sist stripping processes with different chemicals (H2, NH3) or plasma conditions are being optimized [Mos01]. Table 2-4. Properties of dielec trics (taken from [Fay03]) Property Ultra Low k porous SiOC Low k SiOC SiO2 Dielectric constant k 2.2 2.9 4.4 Density (g/cm3) < 1 1.2 2.3 Carbon at. % 15 to 20 20 None Sensitivity to water Hydrophobic Hydrophobic Hydrophilic Porosity (%) ~ 45 ~ 25 None Pore size (nm) 2.5 to 3.5 < 1.0 None Young Modulus E (Gpa) 4 to 6 14 > 50 Hardness (Gpa) < 0.8 1.7 > 8 Thermal conductivity (W/mK) < 0.2 ~ 1.4 Leakage current (A/cm2) < 8 E-10 ~ 8 E-10 8 E-10 Breakdown voltage (MV/cm2) > 4 ~ 3 ~ 5 As of now, PVD is the most common me thod of diffusion barrier deposition in industry, but it suffers from conformality i ssues. Therefore, i-P VD (Ionized PVD), CVD and ALD have been developed and demonstrated better conformality than conventional PVD. However, even these advanced deposit ion techniques still have some problems when it comes to ULK porous materials. For example, thin PVD barriers deposited on porous dielectrics are disconti nuous with pinholes in the film s confirmed by ellipsometry, which could be only solved by growing th ick films [Bak01, Iac02]. For CVD and ALD processes, the precursors tend to easily pene trate into porous ULK di electrics. Penetration of chemicals into ULK dielectrics can change the material structur e and modify physical

PAGE 42

28 and chemical properties of ULK materials [Fay03]. The porosity on the surface of ULK dielectrics has been sealed pr ior to barrier deposition pro cess to prevent the precursor diffusion from CVD or ALD. Numerous pore sealing trea tments such as plasma treatment [Tsa04, Leo06, Wan05, Hoy04, Abe 04, Hum05] liner deposition [Bon03, Jez04, Toi02] and UV curing [Got06, Ito06] have been employed to serve this function. The examples of these sealing treatments will be reviewed in Chapter 7.

PAGE 43

29 CHAPTER 3 EXPERIMENTAL SETUP AND REACTOR SIMULATION USING CFD SOFTWARE 3.1 Reactor Design and Setup Figure 3-1 shows the Process Flow Diagram (PFD) of the ALD reactor system used in this work. Two bubblers, usually containi ng metal organic precursor, are immersed in the constant temperature bath and heated to a defined temperature, where sufficient vapor pressure (> 0.1 Torr for ALD) can be reached to ensure the efficient delivery of the precursor for film deposition. Two metal organi c source lines are need ed for growth of ternary films with th e exception that the second metal s ource is the gas phase (e.g., SiH4, Si2H6, and WF6). The nitrogen, argon, or hydrogen carrier gas picks up the metal precursor at saturation and delivers it to reactor or bypass line, depending on the on/off schedule of pneumatic valves located upstr eam to the reactor. The piping lines placed before and after the bubbler to the reacto r are heated with heat tapes up to 10 C higher than the bubbler temperature to prevent th e condensation of precursor during delivery. The nitrogen gas not only acts as a carrier gas but also as a sweep ing (purging) gas in purge step during ALD cycles. Non-metal precu rsors, generally gas phase, are contained in a gas cylinder within a gas cabinet. All ca rrier, purge nitrogen, and reacting gases are metered to the reactor or bypass line and then to the vent system using Brooks 850 e(m) mass flow controllers paired wi th a FM304V flow manager box.

PAGE 44

30 Figure 3-1.Process flow diagram of ALD reactor system Reactor inlet head ( ) was designed as a cone-shape (Figure 3-2) so that the reactant and purge gases can be uniformly dist ributed as they enter the reactor chamber. To further increase the uniform distribution of the feed reactants over the substrate area (2 wafer maximum), the removable shower head plate is attached at the bottom side of reactor inlet using three bolts. The holes on th e shower head plate were made larger to allow adjustment of the distribution of reactan t and purge gases. The cone-type inlet fits beneath the CF flange ( ), which is the connection point to the reactor body (Figure 33), to minimize the distance between the shower head plate and the substrate. The other purpose of this design was to lower the poten tial of precursor c ondensation by wrapping heat tapes to the bottom of th e cone shape head. Two lines ( : Straight, : 90o bent) are welded to the top of the reactor inlet head to prepare for the case of CVD runs, which requires the separation of each reactant befo re they enter the reactor to prevent the particle formation in the piping lines through the reaction. Carrier N2 Reactor Vent Reactant 1 Reactant 2 Purge N2 M1 M 2 MFC Pneumatic v/v Bubbler

PAGE 45

31 Figure 3-2.CAD drawing of reactor inlet pa rt with shower head plate attached The reactor body (Figure 3-3) has a double wall (jacket) structure, where cooling water enters into the right bottom and exits out of the left top ( ) to operate the ALD/CVD runs as a cold-wall reactor and to protect the Viton O-rings at the quick access door for sample loading. The cooling water pipes are connected via 1/4 NPT female ports. Two 600-400 conflat flanges (CF ) are welded to the side of reactor. One of them is designed to accommodate the quick access door, which has the quartz glass sealed with Viton O-ring and vacuum seal s after the sample loading. The other 600-400 CF is intended for attaching the transport chamber leading to the Cu deposition chamber

PAGE 46

32 without breaking vacuum. A quick flange (QF-25 ) port is also attached at the side for thermocouple insertion to measur e the substrate temperature. Figure 3-3.CAD drawing of reactor body The location of heater support ( ) can be adjusted vertically by employing the ultra-torr fitting with Viton o-ring, which vacuum seals between the inner piping ( ) connected to the heater s upport and the outer piping ( ) in the Figure of reactor downstream body (Figure 3-4). This design allo ws the distance between the showerhead plate and the sample substrate to be adjusted for the purpose of optim izing its location for the film growth. The QF25 adaptor ( ) is connected with the inner piping through the ultra-torr fitting, which will then be linked to electrical feedthrough of QF10 size for heater power supply. Three QF25 flanges ( ) are welded to the side of reactor

PAGE 47

33 downstream body; one is a pressure transducer port, the other is a vacuum pump port, and another one is for the ro ugh vacuum pressure gauge. Figure 3-4. CAD drawing of reactor downstream Custom designed resistive heater with 2 diameter from General Electric ceramics heats the quartz susceptor on which the subs trates are placed. K-type thermocouple from the reactor side ( in Figure 3-3) fits on 1/16 hole on the susceptor and provides the Yokokawa UT-150-AN/RET temperature contro ller with the measured temperature. After comparing the set point value with th e measured temperature, the temperature controller sends out the 4 ~ 20 mA control signal to SCR power controller (1025-10-20 from Control Concepts) adjust ing the main power of 110 VA C to 24 VAC to control the substrate temperature. The resistive heater is a pyrolytic graphite -based (PG) heater coated with pyrolytic boron nitride (PBN). Operating temperature can reach over 1500 C

PAGE 48

34 and it is chemically inert to most corrosive gases and liquids. It has high mechanical and thermal shock resistance due to its low modul us and layered BN ceramics. There is no potential of particle formation due to out-gas sing of the coated BN because CVD BN is fully dense and stable. Post type electrical connection is employed rather than heater surface electrical connection to minimize the exposure of electrical connection point to the reacting environments. For further isolation from the chemical surroundings, a cap shape heater support is used. High temp erature heater hookup wires (HTMG-1CU320S/C from Omega) from the electrical feedthrough (EFT0123058 from Kurt. J. Lesker) are electrically connected to the heater via M4 thread, 0.7 mm pitch molybdenum bolts (from Kamis). Ceramic insulator washer hats (from McAllister Technical Services) are placed up and down of the heater support cap for the electrical insulation purpose. Figure 3-5. Schematic diagram of heater assembly M4 thread Washer h Molybdenum bl Heater wire Post type contact Support cap PBN/PG h

PAGE 49

35 The reactor pressure is monitored using active Pirani gauge (from BOC Edwards) paired with active gauge displa y unit. It is designed suitab le for measuring medium range and low pressure from atmosphere to 10-4 mbar. A typical pressure range for ALD growth of metal nitrides and oxides is around 1 Torr. Maintain ing reactor pressure almost constant is very important during ALD grow th, because it keeps the amount of source pulse constant. Source is pulse d into the reactor by the pressure difference between reaction chamber and source bubbler for a very short time. If the pressure is not kept constant in the chamber, the amount of sour ce at each pulse varies and thus difficult to obtain uniform growth. For this purpose, the commercial f-120 ALD reactor (from Suntola) employs the Inert Ga s Valving System, which flow s the continuous purge gas of high flow rate for every sequence to mi nimize the pressure fluctuation when the sequence changes. Unless there considerable pressure variation exists per sequence, a pulsed purge can also be used. Reactor pressu re is manually controlled by adjusting the conductance of the angular valve, attached to the mechanical pump (E2M12 from Edwards) in the system. Our reactor, volume including the cone-s hape inlet and body, was designed as 1750 cm3, which is small enough to ensure the comp lete purge of the remnant reactants and byproducts within a very short time. Consider ing the capacity of the roughing mechanical pump, 10.2 cfm (cubic feet per minute) equivalent to 4813 cm3/sec, it takes approximately 0.36 sec (1750 cm3/ 4813 cm3/sec) to sweep the whole reactor volume. Due to its small volume, the reactor pressure could quickly reach the base pressure of 10-3 Torr. It is critical in ALD that the flows should be laminar for the uniform adsorption of precursor onto the entire substrate area. Ther efore, typical Reynolds number (Re) should

PAGE 50

36 not exceed 1000, otherwise the flow pattern becomes turbulent prohibiting the uniform adsorption. The Reynolds number for a circular tube, which is the shape of our reactor, can be described as following. D Q v Dz 4 Re D: Tube diameter, ` zv: Average velocity, : Density : Viscosity, Q: Volume flow rate Insertion of typical growth conditions of ALD runs (nitrogen purge gas with 200 sccm flow rate, 1.15 kg/m3 density and 0.0175 cp viscosity) gives a Reynolds number of 2.45, satisfying the laminar flow condition. It is important to expose or pulse sequentia lly to the purge and re actant gases to the reactor or to bypass vent in the operation of ALD mode. For this purpose, the manifold, located before the reactor, c onsists of five separate lines equipped with ten Swagelok ssbnv51-c pneumatic valves. One pneumatic valve controls (on/off) th e reactant or purge flow into reactor and the other valve manage s the gas flow to the bypass line, making the sequential pulse of reactants and purge gas po ssible. These valves are assembled as close as possible to minimize the time delay between each step during the ALD. The distance between horizontal tubes was designed at 2 minimum distance for the same reason. All tubing is 1/4 diameter and connected w ith VCR fitting, suitable for the ultra high vacuum environments. Sequential exposure of each reactant and purge gas is managed by the control program encoded with National Instrument (NI) LaBView software. Figure 3-7 shows the control panel of the program. Currently, the panel shows a typical cycle of ALD growth comprising four sequences with 6 pneumatic va lves. Each column represents a sequence

PAGE 51

37 and the row corresponds to a par ticular valve. The first sequen ce is the reactant 1 pulse stage (R1), where only reactant 1 gas enters into the reactor, while the reactant 2 and purge gas are vented via the bypass line. The second one is the purge stage (P), where N2 purge gas is the only gas into the reactor a nd other gases (reactants 1 and 2) proceed to the vent. After the purge sequen ce, reactant 2 (R2) is exposed into the reactor, whereas reactant 1 and purge gas bypass to the vent. R2 sequence is followed by another purge stage, which has exactly same configuration of valve opening with the second stage (P). Figure 3-6. Manifold structure of pneumatic valves The duration time of each sequence can be controlled by entering the number of sec in the pulse time zone under the sequence tags The sequence indicato r shows the current sequence with yellow color and the total cy cle number (one cycl e corresponds to the 4 sequences) can be entered for the purpose of film thickness control. The control signals

PAGE 52

38 from this program become physically meaningf ul electric signals of +/24 VDC through the 64-channel isolated digital I/O NI PCI-6514 interface board and WRB24SX-U Switching Mode Power Supply (SMPS), connected to the actuators of solenoid valves via electric wires. Then the solenoid valves cont rol the flow of compre ssed air gas led to the actuators of pneumatic valves, which open and close the pneumatic valves (normally closed type). Figure 3-7. Control panel of ALD growth programmed with LaBView 3.2 Velocity and Temperature Profile Simu lation of Reactor using CFD Software Simulations on flow and thermal patterns in the reactor were performed using Fluent Computational Fluid Dynamics (C FD) software. Basically, equations of conservation for mass, momentum, and energy were solved with geometry and boundary conditions specific to this reactor geometr y. The equation for conservation of mass, or continuity equation, can be written as follows: 0 ) ( v t (3-1)

PAGE 53

39 This equation is the general form of the ma ss conservation equation, simply stating that the rate of increase of the density within a small volume element fixed in space is equal to the net rate of mass influx to the element divided by its volume [ ) ( v ]. Conservation of momentum can be described by the following equation. ) ( ) ( v v P g v t (3-2) This equation states that the rate of increase of moment um per unit volume [ ) ( v t ] is equal to the sum of the rate of mome ntum gain by convection per unit volume [ ) ( v v ], the rate of momentum gain by viscous transfer per unit volume [ ], the pressure force on element per unit volume [P ], and the grav itational force on element per unit volume [ g]. This momentum balance is exactly equivalent to Newtons second law of motion, which is th e statement of mass x acceleration = sum of forces. The conservation law for energy can be written as follows. ]) [ ( ) ( ) ( ) ( )) 2 1 ( ( ) 2 1 (2 2v pv g v q v U v v U t (3-3) This equation includes two major energy te rms commonly used in computational fluid dynamics, which are kinetic energy associated with observable fluid motion and internal energy associated with the ra ndom translational and internal motions of the molecules, plus the energy of interacti on between molecules. This energy conservation law states that the rate of gain of energy per unit volume [ ) 2 1 (2v U t ] equals the sum of the rate of energy input per unit volume by convection [ ) 2 1 ( (2v U v ], the rate of

PAGE 54

40 energy input per unit volume by conduction [) ( q ], the rate of work done on fluid per unit volume by gravitational force [) ( g v ], the rate of work done on fluid per unit volume by pressure forces [) ( pv ], and the rate of work done on fluid per unit volume by viscous forces []) [ ( v ]. To solve these coupled conservation PD Es, FLUENT employs a Finite Volume (FV) method to convert the governing equations to algebraic equations that can be solved numerically. The solution doma in is subdivided into a fi nite number of contiguous control volumes (CVs), and the conservati on equations are applied to each CV. This control volume technique consists of integr ating the governing equations about each CV, yielding discrete equations th at conserve each quantity on a control volume basis. Discretization of the governing equation of the steady-state conser vation of a scalar quantity is shown as a simple example, dem onstrated by the following equation written in integral form for an arbitrary control volume Vas follows: VdV S A d A d v (3-4) where A : Surface area vector : Diffusion coefficient for : Gradient of S : Source of per unit volume Equation 3-4 is applied to each control volume, or cell, in the computational domain. The two-dimensional, triangular cell shown in Fi gure 3-8 is an exampl e of such a control volume. Discretization of equa tion 3-4 on a given cell yields

PAGE 55

41 faces N f n f faces N f f fV S A A v ) ( (3-5) where faces N : Number of faces enclosing cell f : Value of convected through face f f f fA v : Mass flux through the face n) ( : Magnitude of normal to face f By default, the discrete values of the scalar at cell centers (c0 and c1 in Figure 38) are stored. However, face values f are required for the convection terms in equation 3-5 and must be interpolated from the cell center values. This is accomplished using an upwind scheme. Upwinding means that the face value f is derived from quantities in the cell upstream, or upwind, relative to the direction of the normal velocity nv Figure 3-8. Control volume expl aining the discretization of a scalar transport equation (taken from [Flu06]) The mesh was generated using the Gambit so ftware with cylindrical coordinates in a two-dimensional format (Figure 3-9). The geom etry of the mesh is almost the same as c0 c1 Af

PAGE 56

42 that of actual reactor. An uns tructured triangular grid was em ployed, which is most often used for control volume methods. Three ty pes of boundary conditions (i.e. inlet flow velocity at reactor inlet, su rface wall temperature at heater surface, and outflow type at reactor outlet) were assigned during the me sh design step. The segregator solver was used, where the governing equa tions (momentum, continuity and scalar) are solved sequentially (i.e. segregated fr om one another), rather than in a simultaneous way. The manner in which the governing equations are linearized, took an implicit form with respect to the depe ndent variables. Figure 3-9. Triangular grid mesh of ALD reactor For a given variable, the unknown value in each cell is computed using a relation that includes both existing and unknown values from neighboring cells. Cell-based option was chosen, where cell center values are cons idered for computing the gradient. Energy 4 .0 2 .0 2 .0 4 .5 5.5 3.0 0.5 Inl e t fl ow Surface wall temperat Outlet P

PAGE 57

43 equation option was selected to simulate th e thermal pattern around the resistive heater area. Nitrogen was a fluid material that enters into reactor and its density, Cp, viscosity, and thermal conductivity were extracted from the Fluent database and listed in Table 3-1. Table 3-1. Properties of nitrogen gas Density (kg/m3) Cp (J/kg-K) Thermal conductivity (W/m-K) Viscosity (kg/m-s) 1.138 1040.67 0.0242 1.163e-05 The operating pressure was fixed at 133 Pa, a typical growth pressure for ALD, and the gravitational acceleration was turned on to the minus Y axis direction. Boundary conditions used in this simulation are summ arized in the following table (Table 3-2). Linear velocity at reactor inlet was driven from the typical flow rate of N2 gas when the ALD reactor is operating in the purge step mode. Volumetric flow rate of 200 sccm, divided by inch diameter circle area gives 0.0263 m/s. Outflow boundary condition uses a typical outlet pressure of reactor, wh ich is around 1 Torr. On every iteration, the outlet velocity and pressure are updated in a manner that is consistent with fully developed flow, satisfying all governing equatio ns. Temperature at th e heater surface was set at 600 K and there is no heat flux through the side-walls of reactor. Because of the nonlinearity of the equation, it is necessary to control the change of This is typically achieved by under-rel axation, which reduces the change of produced during each iteration. In a simp le form, the new value of the variable within a cell depends upon the old value, old the computed change in , and the underrelaxation factor, as follows: old (3-6)

PAGE 58

44 The under-relaxation factor for pressu re, density, body forces, momentum and energy are set as 0.3, 1.0, 1.0, 0.7 and 0.8. In the discretization scheme the first order upwind, where is set equal to the cell center value in the upstream cell, was set for momentum, and the second order upwind, wher e quantities at cell faces are computed using a multidimensional linear reconstruction approach, was used for energy calculation. After initializing the values from the inlet fl ow velocity boundary condition, the iteration started and converged after 541 iterations. Figures 3-10 and 3-11 show the contour of velocity magnitude and static temper ature resulting from the simulation. Table 3-2. Boundary conditions Boundary location Boundary type Specific conditions Inlet flow Velocity inlet -0.0263 m/s (Y velocity) Outlet flow Outlet Pressure 1 Torr Heater temperature Wall 600 K Reactor wall Wall No heat flux Figure 3-10. Contour of velocity magnitude (m/s) (a) and velocity vector colored by velocity magnitude around heater area (m/s) (b) (a) (b)

PAGE 59

45 Figure 3-11. Color filled contour of static temperature (K) (a) and contour line of static temperature (K) around heater area (b) (a) (b)

PAGE 60

46 CHAPTER 4 EFFECT OF NH3 ADDITION ON TaN MOCVD USING TBTDET 4.1 Introduction As the feature size of integrated circuits (IC) shrinks to nanometer dimensions, the need for conformal and reliable barrier materi als increases considerably. One of the most challenging issues in Cu-based ICs is to deve lop a suitable diffusion barrier material and method to deposit it. Extensive research on tran sition metals and their nitrides (e.g., TiN, TaN, and WN) has positioned TaN as the indus try preferred barrier material. Ta and its nitride are used as the liner for Cu damascen e interconnects due to its high Cu reliability, optimized adhesion for Cu electromigrati on resistance, robust via interfaces, and redundant current strapping for added chip reliability [Ede02] Most TaN layers tested, however, have generally been deposited with reactive sputtering, which does not provide good conformal coverage of submicron featur es and high aspect ratio contact and via holes, due to its inherent sha dowing effect. Therefore, CVD of the Ta/TaN bilayer barrier combination has received more attention because of its superior conformality compared to sputter-deposited barriers. TaN-CVD using a metal halide source normally requires a high deposition temperature to obtain low resi stivity and minimal halide content [Hie74, Che97]. As previously mentioned, halide incorporation in the film reduces th e adhesion strength of the film, which reduces long term device re liability [She90, Yok91]. Therefore, halidefree, metal organic-based Ta N CVD promises to provide lower deposition temperature without halide incorporation. Although MOCVD can lower the deposition temperature,

PAGE 61

47 the temperature should be sufficiently high to obtain low resistivity as well [Tsa95a]. In addition, MOCVD of TaN tends to yield a high carbon concen tration in the film, which leads to high resistivity, especially when Ta N is deposited with a single source precursor [Cho99]. In this work, TBTDET and ammonia were employed as co-reactants for TaN CVD. Previous studies using TBTDET were performed at a relativ ely high temperature (450 to 650 C), which is not acceptable for the modern IC processes that require a low thermal budget. Therefore, a lower deposi tion temperature window (300 to 400 C) was selected to assess the feasibility of TaN grown by MOCVD at low temperat ure as a Cu diffusion barrier. Previous research on MOCVD of TaN using TBTDET was limited to supplying only TBTDET as a single source precursor without any additional nitrogen source or reducing agent. Since TBTDET contains four nitrogen atoms bonded to each Ta atom (Figure 4-1), TaN films can be grown with this single precursor by CVD. The films deposited with TBTDET in an inert carrier ga s, however, contain a significant amount of carbon, reaching up to 30 at %, which is as cribed to incomplete cleavage of the C-N bonding involving both the diethyl and t-but yl ligands. Thus the addition of NH3 during MOCVD of TaN was explored as a mechanism to reduce the carbon content of TaN films using TBTDET. It was thought that the transamination reaction of NH3 with diethylamido or tert -butylimido ligands and possibly the reaction of added H would reduce the carbon incorporat ion in the TaN films.

PAGE 62

48 Figure 4-1.Chemical structure of TBTDET Encouragingly, there have been similar reports on film property improvements of MOCVD grown metal nitrides thro ugh the simple addition of NH3. For example, the resistivity and carbon impurity of TiN using TDEAT [tetra kis-diethylamino-titanium] [Sun94, Raa93, Mus96] was reduced significantly when supplying NH3 as compared to flowing just TDEAT as a singl e source. As another example, Cho et al. [Cho99] reported that as-grown TaN films deposited with PDEAT and NH3 also exhibited lower resistivity and carbon content as compared to th e films grown with PDEAT alone. Similar effects are expected upon addition of NH3 during growth of TaN from TBTDET due to the presence of dimethylam ido ligands in both PDEAT and TBTDET and the propensity for tantalum dimethyl amido complexes to undergo transamination with ammonia [Hol90]. Alt hough TBTDET also contains a tert -butylimido ligand, reaction of imido ligands with amines such as NH3 is generally facile [Whi06]. This transamination reaction with am monia is expected to lower carbon content and resistivity by removing carbon-containing ligands in TB TDET. In the study presented in this

PAGE 63

49 chapter, TaN films were de posited with and without NH3 and the various film properties (resistivity, chemical composition, microstr ucture, surface morphology, and density), as well as the deposition characteristics, are compared using 4 point probe, AES, XRD, SEM, XRR, and AFM. A test to determine the efficacy of the Cu diffusion barrier was also performed with TaN films grown w ith a single source and a range of NH3 flow rates. 4.2 Experimental Details The deposition of TaN was performed on st andard (100) silicon substrates in a customized cold-wall, low-pressure CVD reactor. TBTDET (vapor pressure 0.1 Torr @ 90 oC, supplied by Alfa Aesar, MA) was contai ned in a stainless steel bubbler and was heated to a temperature in the range 80 to 90 C. Nitrogen with a flow rate of 50 to 100 sccm was used as a carrier gas. TaN films we re deposited at a sequence of temperatures in the range 300 to 400 oC at 25 C intervals with a typical growth pressure of 1 ~ 2 Torr, and the ammonia flow rate was varied fr om 0 to 150 sccm. The film thickness was measured by cross-sections SEM. Four point probe was employed to measure the film resistivity. The microstructure was analyzed using XRD and SEM, while AFM was used to observe the surface morphology of the film s. The chemical composition of the films was probed by AES measurement and the density extracted from XRR profiles. Cu of 100 nm thickness was deposited by sputter deposition onto CVD-TaN layer of 50 nm thickness to investig ate the diffusion barrier proper ties for Cu metallization. The samples were then annealed at 500 C for an hour in inert gas (N2) atmosphere to monitor the changes in the Cu /TaN/Si structure. The barrier failure was analyzed by observing the copper silicide peak in the XRD pa ttern after annealing. The barrier failure was also tested by monitoring the formation of etch pits on the Si surface. Cu and TaN films were wet chemically etched with HNO3:H2O = 1:20 and HF:H2O2 = 1:2 solution

PAGE 64

50 respectively, after annealing. The surface of th e Si substrates was Secco etched for 15 to 30 seconds at room temperature, and the size and density of etch pits were observed by SEM. 4.3 Results and Discussion 4.3.1 Resistivity and Chemical Composition Figure 4-2 depicts the change in th e resistivity of films grown at 400 C as a function of the NH3 flow rate. As the NH3 flow rate increased the film resistivity decreased, most dramatically as small amounts of NH3 were added. The resistivity was around 26800 with no NH3 flow and it drastically decreased to 15000 at an NH3 flow of only 15 sccm. Additional NH3 resulted in a more gradual decrease in the resistivity. Figure 4-2.Resistivity, and nitrogen, carbon and oxygen content relative to Ta in TaN films as a function of ammonia flow rate The minimum resistivity obtained was ~6300 which is still more than an order of magnitude higher than that of sputtered TaN (~250 ) [Cho99] presumably

PAGE 65

51 due to high carbon levels (~ 5 at %) in CVD TaN. A decrease in the resistivity with the addition of NH3 has also been reported for Ti N-CVD using TDEAT [Sun94, Raa93, Mus96], and TaN-CVD with PD EAT [Cho99]. This trend is seen to track changes in impurity levels in the films. Figure 4-2 also shows the impurity ratios C/Ta and O/Ta as a function of NH3 flow rate. Increasing the NH3 flow reduced the carbon content of the films presumably via transamination with ammonia, which would remove the carboncontaining diethylamido ligands as the vol atile compound diethylam ine [Gor90, Dub94]. Following the proposed mechanism of transamination of Ti(NMe2)4 and NH3 [Wei96], it is suggested that NH3 initially forms a weak in termediate adduct with Ta center in (NEt2)3Ta=NBut, which is probably not stable gi ven the steric bulk of amido and imido ligands. H-atom transf ers between the coordinated NH3 and an amido ligand probably occurs via a four-transition state followed by elimination of HNEt2 (Figure 4-3). Figure 4-3. Proposed transamination r eaction mechanism between TBTDET and NH3 (Concept taken from [Wei96])

PAGE 66

52 ALD studies of TiN using TDMAT and NH3 also confirmed that transamination reaction paired with an amine elimination st ep is the primary reaction in depositing TiN films as evidenced by FTIR data and QC M in-situ measurements [Ela03]. Isotopic labeling studies of TiN CVD using 15NH3 and ND3 demonstrated the formation of Ti15N and the use of ND3 gives DNMe2 [Dub92]. A related ammonoly sis reaction involving the tert -butyl imido ligand will also be facile under the deposition conditions [Bec03]. The oxygen content is also reduced by the addition of ammonia but that cannot be a direct result of ligand removal by transamination because the ligands contain no oxygen. TaN films grown in the presence of ammonia, however, are expected to be denser via transamination reaction, which would inhi bit post-growth incorporation of oxygen upon handling the films in air. 4.3.2 Microstructure and Surface Morphology Analysis XRD patterns (Figure 4-4) of the TaN fi lms were obtained as a function of NH3 flow rate to investigate how NH3 addition affects the film crystallinity. When TaN was deposited without NH3 flow, no peaks were observed, consistent with an amorphous structure. With the addition of NH3, the peak of h-TaN began to appear and as more NH3 flow was added, and c-TaN (111) and c-TaN (200) peaks became more apparent in the pattern. CVD of TaN from PD EAT [Cho99] showed a similar trend in that the film deposited with NH3 showed a greater extent of polyc rystalline structure than the film grown with PDEAT only as a sing le source precursor. Increa sed crystallinity with higher NH3 flow is also consistent with the observ ed lower resistivity plotted in Figure 4-2.

PAGE 67

53 Figure 4-4. XRD patterns of TaN films deposited at 300 C as a function of ammonia flow rate There was a distinct difference in the surface morphology between the films grown with TBTDET alone (Figure 4-5( a)) and those grown with NH3 (Figure 4-5(b)), although the films grown with NH3 at different flow rates (30, 60, and 150 sccm) did not exhibit much variation. NH3 addition yielded rougher surface with more hillocks and larger grain size (45 to 60 nm) than the depositio n with TBTDET alone (9 to 11 nm). Figure 4-5. Surface morphology of Ta N films deposited (a) without NH3 at 300 C and (b) with NH3 (30 sccm) a b RMS: 1.7 nm RMS: 2.8 nm

PAGE 68

54 Observation of cross-sectional SEM images (Figure 4-7) also produced the same conclusion, showing films gr own in the presence of NH3 with larger grain size, while the films from TBTDET alone show ed no microstructure and a smaller grain size which could not be recognized in the fi eld-emission SEM image (50,000). 4.3.3 Deposition Characteristics and Density Analysis The deposited films were smooth, nonpor ous, pinhole-free and adhered well to the substrate when tested with a tape. The films were mirror-like with a reflected gold color. Figure 4-6 clearly illustrates that the deposit ion behavior was signi ficantly changed when NH3 was introduced to the reactor. The deposition rate from TBTDET alone as a function of growth temperature showed a steep increase with temperature characteristic of a surface reaction controlled process. Th e Arrhenius plot shown in Figure 4-6 shows an activation energy of 0.3 eV, wh ile the overall deposition with NH3 flow had the appearance of mass transfer-controlled regime with an activation en ergy of 0.05~0.07 eV. This result suggests that NH3 addition gives better control in film thickness over single source deposition since the film thickness doesn t vary significantly with the growth temperature in mass transfer controlled reaction. The effect of adding an NH3 flow at a constant flow of TBTDET was to decrease the growth rate relativ e to the case without NH3 flow at the same growth temperature. The growth rate generally trended down w ith increasing temperature, which usually implies a thermodynamic limit or desorption dom inated process. The deposition rate, however, cannot be measured by the thickness un less the density is constant, as discussed below. A similar behavior of deposition ra te with temperature was observed with TaN CVD from PDEAT [Cho99]. Their results without NH3 flow taken over the temperature range (300 to 375 C) showed surface reaction limited gr owth at low temperature (300 to

PAGE 69

55 350 C) and a mass transfer controlled patte rn when deposited with 25 sccm NH3 flow. The same sequence was also observed in TiN CVD from TDEAT [Raa93], showing a higher activation energy (Ea = 0.5 eV, 280 to 420 C) for thermal decomposition of TDEAT than that (0.09 eV) for the tr ansamination reaction of TEDAT with NH3. Figure 4-6. Growth rate (/min) vs. recipr ocal growth temperature with different NH3 flow rates of 0 to 150 sccm Figure 4-7 shows the crosssectional SEM images of TaN films grown at 4 different NH3 flow conditions at 400 C. A decrease in the deposition rate with higher NH3 flow rate was observed in most films gr own at constant temp erature. A lower deposition rate with NH3 addition was reported for TDEAT [Mus96] as measured by XRR and step profilometry. In contrast, an increase in the deposition rate with NH3 was reported for TDEAT, where film thickness was derived from weight gain measurements [Raa93], and for PDEAT, as measured by cross sectional SE M [Cho99]. Although the reason for this conflict is not obvious, one possible reas on is that the higher NH3 flow

PAGE 70

56 increased the extent of the transamination reactions, which leads to a denser film structure. Based on the assumption that the amount of Ta incorporated into films is the same during the fixed deposition time for both runs without and with NH3, the thinner films imply a higher density. Figure 4-7. Cross-sectional SEM images of TaN films deposited with various NH3 flow rates at 400 C The film density was evaluated by XRR (Fi gure 4-8) and the data were analyzed using the WinGixa program. XRR can estimate th e film density by measuring the critical angle (starting angle of total reflection of X-ray), which is proportional to film density. A very thin SiO2 layer (~ 20 ) was assumed to exist at the interface between the Si substrate and TaN film. Also, TaN with C (~15 at %), and O (~10 at %) impurities was introduced in the near upper surface region of TaN for the simulation. As expected, the NH3: 0 sccm NH3: 30 sccm NH3: 60 sccm NH3: 150 sccm

PAGE 71

57 films deposited with higher NH3 flow rate showed higher density, which would make these films better candidates for Cu diffusion barriers if they remain essentially amorphous. When the NH3 flow was greater than 60 sccm the simulated density is 10 (g/cm3), which is about 60 % of the density of bulk TaN (16.3 g/cm3). As previously mentioned, a denser film would be expected to be more efficien t in prohibiting postgrowth oxygen incorporation, which is consis tent with the lower oxygen content in the films deposited with NH3 flow and the flattening of the profile at NH3 flow rates greater than 60 sccm. Figure 4-8. XRR profiles of TaN films deposited at 400 C with different NH3 flow rates 4.3.4 Nucleation Step For deposition both with and without NH3, there did not appear to be an induction period for nucleation. TaN began to nucleate an d form films within 5 sec of the reactant exposure (Figure 4-9) as judged by the surface roughness change from 1.5 (bare Si) to 3.6 (TaN from TBTDET alone for 5 sec) and 2.7 (TaN from TBTDET and NH3 for

PAGE 72

58 5 sec). TaN deposited from TBTDET alone fo r 20 sec showed a resi stance of ~1970 ohm, which is as close as that of Ta N (~1880 ohm) from TBTDET and NH3 for 10 sec. On the other hand, TaN deposited for 5, 10, and 15 sec without NH3 and 5 sec with NH3 had the resistance in the range 480 to 780 ohm. This indicates that full c overage of TaN on Si occurred sooner for deposition with NH3 (10 sec) than for TBTDET single source deposition (20 sec), since the full coverage of TaN will increase the film resistance. It is interesting that the growth rate from TBTD ET single source was higher than that with NH3 flow at 350 C, which shows the opposite beha vior with nucleation rate. Figure 4-9. AFM scans of the initial stages of deposition from (a) TBTDET only and (b) TBTDET with NH3 = 50 sccm at 350 4.3.5 Diffusion Barrier Performance To evaluate the diffusion barrier perfor mance, Cu of 100 nm thickness was sputterdeposited onto 50 nm thick MOCV D-TaN grown with different NH3 flow. The samples were then annealed at 500 C for an hour in an inert gas atmosphere (N2) to monitor the reactions in the Cu/TaN/Si structure. Am ong those samples, only TaN deposited from TBTDET single source showed the Cu15Si4 peak (Figure 4-10), suggesting that Cu 5s 10s 15s 20s (a) 5s 10s 15s 20s (b) 3.6 3.5 2.2 3.2 2.7 2.5 3.0 5.5

PAGE 73

59 diffused through the TaN barrier to form c opper silicide, while th e other films grown with NH3 exhibited peaks associated with TaN( 111) and Cu(111) but not any related to copper silicide. The intensity of the h-TaN(100) peak is seen to increase with increased NH3 flow, which is another indication of great er crystallinity for films deposited with NH3. Figure 4-10. XRD spectra of Cu/TaN/Si structures annealed at 500 C for one hr for films grown with different NH3 flow rates. Although XRD spectra suggested that none of the TaN films deposited with NH3 failed the barrier test, the etch pit test revealed that TaN grown at NH3 flowrate of 0 and 30 sccm did not pass the barrier test (Figure. 4-11). A high de nsity of etch pits formed by the Cu diffusion into Si that was subsequently revealed by etching the bare Si surface in a Secco solution. TaN deposited with 30 sccm NH3 exhibited a lower population of etch pits, but the presence of etch pits indicated barrier failure. In contrast SEM images of etched films grown with an NH3 flow rate of 60 sccm and greater showed no etch pits.

PAGE 74

60 Revealing Cu diffusion by etching of the Si is clearly more sensitive than detecting Cu silicides by XRD [Cho99]. Desp ite higher crystallinity, whic h is expected to provide grain boundary diffusion pathways, the higher ni trogen content and density of the films deposited with NH3 appear to improve the Cu diffusion barrier performance. In addition, the lower film resistivity and impurity levels (C and O) with NH3 addition render TaN barriers superior to those from TBTDET single source deposition. Figure 4-11. SEM images of Si surface after an nealing the structure Cu/TaN/Si at 500 C for 1 hr, followed by the removal of Cu and TaN layers, and then etch in Secco solution to reveal et ch pits if Cu penetrati on occurred. The TaN films were grown at various values of NH3 flow. As evidenced above, better diffusion ba rrier performance was observed for TaN films when NH3 was added to the reaction chamber. The film density, surface roughness, crystallinity, film conductivity, and nitroge n content each increased with increasing NH3 flow, while the C and O impurity levels decreased. It is well known that high film density is a desirable property for diffusion barriers to eliminate diffusion across voids, defects or NH3: 0 sccm NH3: 30 sccm NH3: 60 sccm NH3: 150 sccm

PAGE 75

61 loosely packed grain boundaries. As an exam ple, sputter-deposited TiN exhibited better Cu diffusion barrier performance as the f ilm density increased, which led to the conclusion that film density is an important property in achievi ng a good diffusion barrier performance [Par96]. Other experiments on TiN for application as a Cu diffusion barrier reported that the diffusion failure temperat ure increased with increased film density [Rha98]. In this research, the TaN film density increased from 9.0 to 10.0 g/cm3 as NH3 flow rate was varied from 0 to 60 sccm. Higher NH3 flow rate is expected to lead to greater transamination reaction extents between NH3 and carbon containing ligands, resulting in a more dens ely packed structure. The nitrogen content in the film is al so considered an important property for diffusion barrier performance. TaN showed the highest failure temperature among Ta, Ta2N, and TaN films as examined by sheet resistance measurements, XRD, and Secco etching methods [Kal00]. Improved performa nce with higher nitrogen content was also observed by Wang et al. from BTS (Bias Ther mal Stress) measurements [Wan98]. AES measurements on TaN films showed that nitroge n content in the film, and thus the barrier performance, increased from 0.68 to 0.90 of N/Ta ratio with the addition of NH3. In addition to higher density and nitrogen content, NH3 addition significantly reduced the carbon content and film resistivity to further improve the barriers. 4.4 Conclusions TaN films were successfully de posited by CVD from TBTDET and NH3. The chemical composition, microstructure, su rface morphology, density, resistivity, and deposition characteristics were investigated as a function of ammonia flow rate and the key results are summarized in Table 4-1. En couragingly the film resistivity decreased from 26800 to 6300 with the addition of an NH3 flow to the reactor. In the

PAGE 76

62 same films, the C/Ta atom ratio dramati cally decreased from 0.95 to 0.12 and was accompanied by an increase in the N/Ta atom ratio to Ta from 0.67 to 0.87, when 150 sccm NH3 injected as compared to films deposited with TBTDET alone. Not as promising, the films became more crystalline with larger grain size and higher density along with rougher surfaces with higher NH3 flow rate. Deposition from TBTDET alone exhibited a surf ace reaction controlled regime while the deposition when NH3 was added showed a nearly temperature independent behavior in the range 300 to 400 C. The change in the film pr operties with the addition of NH3 flow is attributed, in part, to transamination reaction between NH3 and carbon containing ligands in TBTDET precursor. The efficacy of the films as Cu diffusion barriers was evaluated by both XRD and etch pitch density determination on annealed Cu/TaN/Si samples. The results showed that TaN deposited with NH3 exhibited superior barrie r quality than TaN grown with TBTDET alone, which was believed primarily a result of the higher film density and nitrogen content. Table 4-1.Summary of film properties and di ffusion barrier test results of TaN deposited with and without NH3 Density (g/cm3) N/Ta ratio C/Ta ratio Resistivity ( ) Grain size (nm) Cu15Si4 formation: after 500 C, 1 hr Without NH3 8.99 0.6670.959 26800 9-11 Cu15Si4 observed With NH3 (30-150 sccm) 9.51-10.9 .8190.886 .124.0.183 9500-6300 45-60 Cu15Si4 not viewed No etch pits for NH3> 30 sccm

PAGE 77

63 CHAPTER 5 ULTRA-THIN ALD TaN FI LMS USING TBTDET AND NH3 FOR Cu BARRIER APPLICATIONS 5.1 Introduction Al-based metal, SiO2 dielectric-based interconnect metallization is being replaced with Cu, lowdielectric-based combination as the feature size shrinks and the speed of electronic devices increases. The lower resi stivity value of Cu and lower value of for the interlayer dielectric enable electrical si gnals to move faster by reducing the RC time delay. Cu also possesses superior resistan ce to electromigration, which is a common reliability problem with Al-Si metallization. Cu, however, is a very fast diffuser in silicon, which can cause serious problems in the device, including increased contact resistance, altered barrier height, leaky pn junctions, and destruction of interconnects [Bch03]. Therefore, an appr opriate diffusion barrier is needed between Cu and its underlying layer. In addition, a diffusion barrier acts as a pa ssivating layer protecting the Cu interconnects from corrosion and oxidation and promotes adhesion of the Cu layer. Among various transition metals and their ni trides, Ta and its nitrides are known to be suitable liners because of their high reliability as a Cu barrier, good adhesion, resistance to electromigration, robust via in terfaces, and redundant current strapping for added chip reliability [Ede02]. Currently, the bi-layer liner Ta/TaN, deposited by PVD, is being used in commercial applications. PVD has been the most widely used method for diffusion barrier deposition in current semiconductor technology (>100 nm minimum feature size). PVD-based technology, however has an inherent limitation in the

PAGE 78

64 deposition of sub-100 nm and high aspect ratio structures due to the directional nature of the depositing flux and high sticking pr obability on most materials [Kim05]. Accordingly, ALD is receiving extensive atten tion due to its ability to grow films with excellent conformality, thickness uniformity over large area, and precise thickness control at the atomic level. Early reports on TaN-ALD us ed a halide source (TaCl5) sequence with NH3. The films showed very high elec trical resistivity (> 104 ) and were polycrystalline Ta3N5, which was attributed to the relatively low reducing power of NH3. The high Cl content (> 20 at %) at 200 C was also an issue with use of TaCl5, leading to possible corrosion of the Cu interconnect lines [Hil88b]. Replacement of NH3 with dimethylhydrazine, (CH3)2NNH2), yielded TaNx films similar to those obtained from NH3 but with significant carbon incorporation [Jup00]. Addition of reducing agents such as TMA (trimethyl aluminum) or amines gave minor improvements in resistivity [Ale01]. The use of Zn as an additional reducing agent produced the cubic phase of TaN with significantly lower resistivity (960 ) [Rit99]. TaBr5 [Ale02] was also employed for TaN ALD, which produced results similar to those found with the use of TaCl5. As an alternative approach, MO (Metal Or ganic) sources such as TBTDET [Tar02, Str04] and PDMAT [Wu04] were reported as a precursor for TaN-ALD. Thermal ALD TaN films deposited with TBTDET showed lower density and higher resistivity compared to PE-ALD TaN, which takes a dvantage of the reducing power of plasma generated hydrogen or nitrogen. The low resis tivity of PE-ALD TaN was attributed to the formation of the more metallic Ta-C bond [Par02]. A fcc NaCl-t ype nanocrystalline

PAGE 79

65 structure of TaN was obtained when PDMAT was used as th e Ta source in thermal ALD [Wu04]. In this study, TBTDET and ammonia were employed for thermal ALD of TaN. Previous studies on TaNALD using TBTDET and NH3 [Ale02, Par02] did not provide specific data on thickness as a function of reactant exposure time, which is essential to establish the window for achieving the self-lim iting adsorption feature of ALD. In this dissertation, XRR (X-ray Reflec tivity) was used to measure the thickness of ultra-thin (<10 nm) TaN films, which was used to identify the self-limiting growth zone. Additionally, the ALD temperature window, whic h is also needed for implementation of precise thickness control, was determined for this precursor combination. The ALD growth characteristics during th e initial stages were investigated by measuring the film thickness as a function of number of cycles. A linear dependency of ALD film thickness on number of cycles is expected, but the dependency can be nonlinear during the initial stages of growth because the adso rption/reaction characteristics can be different on the substrate surface as co mpared to finite thickness of the material being grown [Lim00, Lim01, Sat02a]. It is cr itical to understand the evolution of growth during initial stages and the mechanism of film closure for growth of ultra-thin (<10 nm) films that will be required for future diffusion barriers. Established tests for determining the quality of diffusion barriers were developed and mostly applied to films that are thick (> 10 nm) relative to future requirements. The diffusion barrier thickness, however, is anticipated to be less than 10 nm by 2007 accord ing to the ITRS roadmap [Tra04]. In this research, the minimum thickness of ALDTaN barrier was determined by applying

PAGE 80

66 standard XRD and etch pit density measurem ent to annealed ultra-thin (7 to 100 ) Cu/ALD-TaN/Si structures. 5.2 Experimental Details ALD of TaN was performed on standard (100) silicon substrates in a customized cold-wall vertical ALD reactor. Pneumatica lly actuated valves, controlled with a LaBView-based algorithm interfaced to a Digital Input/Output interface board, were sequenced to flow each reactant and purge gas to the reactor sequentially using a run-vent design. TBTDET (supplied by Alfa Aesar, MA vapor pressure 0.1 Torr at 90 C,), contained in a stainless steel bubbler, was h eated to a constant temperature with a set point of 80 C. The carrier gas was N2. TaN films were deposited at a constant substrate temperature, with a set po int in the range 200 to 400 C with a typical growth pressure of 1 Torr. The deposition cycle began with an exposure of TBTDET for 6 to 18 sec with 20 sccm of N2 carrier gas. Then nitrogen flowed in to the reactor at 200 sccm for 10 sec to ensure the complete sweeping of excess TBTDET precursor as well as volatile byproducts. After this purge step, NH3 of 20 sccm was introduced for 10 sec, followed by another N2 purge sequence. To investigate the diffusion barrier perfor mance of ultra-thin TaN films for Cu metallization, a Cu film of 100 nm thickne ss was sputter-deposited onto the ALD TaN layer of 7 to 100 thickness. The samples were then annealed at 500 C for 30 min in N2 to monitor the changes in the Cu/TaN/Si stru cture. The barrier was judged as failed if a copper silicide peak appeared in the XRD pattern of the annealed structure. The barrier failure was also tested by observing the forma tion of etch pits on the Si surface. After annealing the Cu/TaN/Si structure, the Cu and TaN films were removed by wet chemical etch sequentially with HNO3:H2O 1:20 and HF:H2O2 1:2 solutions. The surface of the Si

PAGE 81

67 substrates was then etched with a Secco so lution and the size and density of etch pits were observed by SEM. 5.3 Results and Discussion 5.3.1 Confirmation of Self-saturation ALD Growth A series of TaN films were grown at 300 C using 80 cycles with variable TBTDET exposure time to locate the saturation zone for TB TDET. The exposure time of TBTDET was changed from 6 to 18 sec with fixed NH3 exposure times of 10 sec and purge time of 10 sec for both purges. The Ta N film thickness was then measured using XRR and the spectra are shown in Figure 5-1 for several exposure times at different temperatures. The thickness of each film wa s extracted using the WinGixa software. The simulation data provides the film thickness, surface/interface roughne ss, and film density by fitting the actual logarithmic reflectivity data. The final 2 value, which determines how the simulated reflectivity matches the measured one, was set as low as 5.0 x 10-2. A very thin SiO2 layer was assumed to exist at the Si-T aN interface. It is noted that C (~15 at %), and O (~10 at %) were introduced into the near surface re gion of the TaN surface in the simulation to ensure a realistic assessment. The growth rate based on the thickness data from XRR profiles is plotted in Figure 5-2. The error bars were estimated by consid ering % error in thickness determination from XRR [Val06] and the dashed-line is draw n to aid the viewers eye. These results suggest that the effective nu mber of Ta atoms deposited pe r cycle increased for TBTDET pulse time less than 9 sec, but for longer time s the deposition rate remained constant (2.6 /cycle) through 11 sec of exposure time, indicating it reached the apparent surface saturation growth mode, which is a feature of ALD. For exposure time less than 9 sec there is apparently insufficient time to reach saturation. The growth rate then increased

PAGE 82

68 to 3.5 /cycle at 12 sec exposure time, likely the result of sufficient time for parasitic CVD reactions to contribute to growth. As indicated in Figure 5-2, the self-limiting region at 300 C was from 9 to 11 sec TBTDET exposure time. However, it was expanded to from 9 to 14 sec, when ALD operates at 250 C as shown in Figure 5-3. This indi cates that the parasitic CVD reaction caused by the partial disruption of selflimiting adsorption at high temperature (300 C) produced the narrower self-limiting zone compar ed to that of ALD at lower temperature (250 C). The deposition rate in se lf-saturation growth zone was 2.63 /cycle, which is 53.6 % of the lattice constant (4.91 ) of h-TaN phase. Th e saturation coverage of many metal-containing precursors for metal and me tal-nitride ALD reactions is a fractional monolayer (ML). One of the most-widely acc epted reasons is the steric hindrance of adsorbed metal precursors [Kim05] and it appe ars this precursor yields ML coverage per cycle. Figure 5-1. XRR-profiles of ALD-TaN as a function of TBTDET exposure time for films deposited at 300 C for 80 cycles, and 10 sec purges and NH3 exposure

PAGE 83

69 Figure 5-2. Growth rate of ALD-TaN as a function of TBTDET exposure time deposited at 300 C, and 10 sec purges and NH3 exposure Figure 5-3. Growth rate of ALD-TaN as a function of TBTDET exposure time deposited at 250 C, and 10 sec purges and NH3 exposure

PAGE 84

70 5.3.2 Process Temperature Window The thickness of each ALD-TaN film was measured as a function of growth temperature to verify the process temperature window for this precursor combination. All depositions were performed at the self-sat uration growth condition identified in the previous section. TBTDET (9 sec, 20 sccm), N2 purge (10 sec, 200 sccm) and NH3 (10 sec, 20 sccm) pulse time and flow rate were fixed with only the growth temperature varied in the range 200 to 400 C for 80 cycles. Figure 5-4 sh ows the XRR profiles of the TaN films grown as a function of temperat ure, while Figure 5-5 plots the thickness extracted from the profile for each run. The film thickness increased with increasing temperature below 200 C. Precursor chemisorption or reaction to form the adso rbing intermediate species are thermally activated processes. Therefore, adsorption of one of the surfac e species is likely kinetically limited to give incomplete adsorp tion in this low temperature region. The same trend has been reported for ALD of several metals and metal nitrides. For instance, the deposition rate decreased for substrate temperature below 150 C for W ALD using WF6 [Kla00], and a similar decrea se was reported for ALD of Ta3N5 using TaCl5 and NH3 below 300 C [Rit99]. The film thickness remained constant (200 10 ) in the approximate temperature range 200 to 300 C, suggesting this is the ALD process temperature window for TBTDET and NH3 at these conditions. This thickness is consistent with the expected value of 196 which represents ML/cycle thickness for 80 cycles of TaN (horizontal line in Figure 5-4). When ALD is operated at a temperature that provides sufficient thermal energy for chemisorption to saturate the growth rate remains constant. The

PAGE 85

71 process temperature window allows precise control on th e reproducibility of film thickness and uniformity compared to CVD sin ce the saturation condition is reproducible. The ALD temperature window for this precurs or combination at these conditions is sufficiently wide to allow good controllability and in a reasonably low temperature range to contribute minimally to the thermal budget. Above 300 C, the film thickness increased with increasing temperature, indicating that adsorption was not self-limiting. This re sult is often attributed to additional reaction that changes the adsorption process to a C VD-like process. It is not known if this involves the TBTDET or ammonia exposur e step, although the onset of thermal decomposition of TBTDET has been reported to occur above 300 C [Str04]. Figure 5-4. XRR profiles of ALD-TaN as a function of growth temperature deposited with 9 sec exposure of TBTDET and 80 cycles, and 10 sec purges and NH3 exposure

PAGE 86

72 Figure 5-5.Thickness of ALD-TaN films as a function of growth temperature deposited with 9 sec exposure of TBTDET and 80 cycles, and 10 sec purges and NH3 exposure 5.3.3 Growth Characteristic of TaN ALD The dependency of film thickness on numb er of cycles was examined by growing a series of films at ALD growth cond itions (9 sec TBTDET pulse time at 300 ) over the range of cycle number 5 to 160. It is expected that a linear depende nce exists between the ALD film thickness and the number of cycl es. The adsorption surface chemistry, however, between the initial Si substrate and the stabilized TaN during the later stage of ALD can be significantly different. Thus the substrate surface preparation, impurities in the gas phase that react with th e surface, and the selection of in itial reactant can affect the initial adsorption step. Kinetic modeling and experimental observation of the initial and stabilized stages of TaN and TiN ALD have been reported [Lim00, Lim01, Sat02a]. These studies indicate that there exists a transient region, wher e the deposition rate increases toward the constant value and th e growth rate remains constant after the transition region.

PAGE 87

73 Figure 5-6 shows the XRR profiles of AL D-TaN as a function of cycle number grown at 300 C with 9 sec TBTDET exposure time. The film thickness was obtained by fitting the measured reflectivity using Wingixa software. Figure 5-7 summarizes the thickness of a series of ALD-TaN films that were grown varying the cycle number. In examining the data above 15 cycles, the growth rate remained constant (2.5 +/0.2 /cycle) demonstrating the expected linear dependency of thickness on number of cycles. The measured thickness of films grown for less than 15 cycles, however, are slightly below the constant value of 2.5 /cycle (solid line in Figure 5-7) a nd gradually increase toward this value. It is noted that in these experiments the bare substrate was first exposed to TBTDET. The existence of this transient region is not surprising since the adsorption characteristics of TBTDET on bare Si or SiOx/Si are expected to be different than those on TaN that had been exposed to NH3 and then purged. Furthermore, the crystal structure of Si (diam ond) and TaN differ as well as th e lattice constants (Si: 5.43 TaN: 4.90 ). Thus self-limiting adsorpti on likely did not occur during the first few cycles, but once the surface was fully covered a nd presumably a few layers thick, relaxed TaN replaced Si as the growth surface to al low repeatable, self-limiting adsorption. For W ALD using WF6 and Si2H6 on SiO2 substrate, ~10 cycles of nucleation were needed before a linear relationship in thickness was achieved [Ela01]. An even longer incubation period (up to 40 cycles) has been reported for TiN ALD using TiCl4 and NH3 precursors [Bey02a].

PAGE 88

74 Figure 5-6. XRR profiles of ALD-TaN as a function of cycle number deposited at 9 sec TBTDET exposure, 10 sec purges and NH3 exposure at 300 C Figure 5-7. Thickness of AL D-TaN as a function of cycl e numbers deposited at 9 sec TBTDET exposure, 10 sec purges and NH3 exposure at 300 C

PAGE 89

75 During growth in this transient region, a surface discontinuity exists between the starting substrate surface and th e saturated surface of the re laxed TaN, leading to a nonlinear relationship between the film thickness and number of cycles. It is possible that competitive adsorption by the reactants leads to preferential growth on areas already covered by TaN. Therefore, three dimensional (3 D) growth may occur in the initial stage. Previous studies on ALD of high dielectric constant materials have shown 3D growth during the transient region [Gre02]. The possibility of this occurring in the TaN films grown in this study was examined with AFM images. Figure 5-8 shows images of the surface of ALD TaN films for several values of low cycle number (5, 10 and 15 cycles). These scans show a rough surface with high hills for the films grown with 5 and 10 cycles, indicating 3D-growth mode prevails during the nucleation st age. The surface is seen to become considerably smoother afte r 15 cycles. Note that 15 cycles was the starting point of the linear re lationship between the film th ickness and number of cycles, which was considered to be the stage of co mplete coverage of TaN on Si substrate. Figure 5-8. AFM scans (z: 3.0 nm / div) of the surface of TaN deposited at 9 sec TBTDET exposure and 300 C using 5, 10 and 15 cycles 5.3.4 Diffusion Barrier Test of Ultra-thin TaN Cu of 100 nm thickness was sputter-deposit ed on the ALD-TaN layers of 7 to 100 thickness to verify the minimum thickness th at could prevent Cu diffusion. The test 5 cycles 10 cycles 15 cycles RMS: 0.50 nm RMS: 0.57 nm RMS: 0.30 nm

PAGE 90

76 structures were then annealed at 500 oC for 30 min in N2. To evaluate the extent if any of Cu transport across the barrier, each film wa s non-destructively evaluated with XRD for evidence of Cu silicide formation. The diffr action patterns are shown in Figure 5-9. The 7 and 16 ALD-TaN films showed the Cu15Si4 (220) reflection, i ndicating that barrier failure occurred in the Cu/TaN/Si structur e, while the other films (38 and greater) showed strong peaks assigned to TaN and Cu but no reflections assigned to copper silicide. A cross-sectional TEM (Figure 5-10) samp le was prepared on the 38 /15 cycle film to confirm the thickness data extracted from XRR. The thickness from TEM (38 ) was a slightly greater than the calculated thickness from XRR (36 ), which is reasonably well matched and within the estimat ed 5 % error. This result indicates that ultra-thin ALD-TaN is a good candidate diffu sion barrier material for the 38 nm node requirement of the roadmap, which specifies a 3.5 nm thick diffusion barrier in the year 2013 [Tra04]. Similar tests with ALD-TaN repor ted that 10 to 60 thick barriers could block Cu diffusion [Bas03, Str02]. TaN peaks could be observed with 38 and 47 thick samples after the annealing. Obviously, the higher intensity of TaN peaks is expected since the thickness increased. The destructive etch pit dens ity evaluation (etching bare Si in a Secco solution after TaN/Cu removal by wet etching) was next appl ied to the annealed samples. This method, which is a more sensitive method for the de tection of Cu diffusion than XRD, also revealed that 16 ALD-TaN film was a poor barrier as evidenced by the high density of etch pits (Figure 5-11). On the other hand, TaN the 38 sample exhibited no etch pits

PAGE 91

77 on the bared Si surface (Figure 5-11), indica ting the critical thickness of ALD-TaN as a Cu diffusion barrier exists between 16 and 38 Figure 5-9. XRD patterns of Cu/TaN/Si annealed at 500 C for 30 min Figure 5-10.Cross-sectional TEM micrograph of Cu/TaN/Si structure. TaN grown by ALD for 15 cycles at 9 sec TBTDET exposure time and 300 C

PAGE 92

78 Figure 5-11. SEM images of Si surf ace after annealing Cu/TaN/Si at 500 C for 30 min followed by the removal of Cu and TaN, and Secco etching The diffusivity ( D ) of Cu in TaN films was estimated using the critical thickness (38 ) and the characteri stic diffusion length (2 t D ), which yielded a D value of 210-172/s at 500 C. This value is similar to the repor ted Cu diffusivity in single crystal TaN, ranging from 10-18 to 10-16 2/s in the temperature range 600 to 700 C [Wan02]. The estimated Cu diffusivity in ALD-TaN at 500 C is surprisingly similar in the reported diffusivity in TaN at 600 C that was obtained on singl e crystal sputter-deposited samples. The structure of ALD-TaN is amorphous with some distribution of nanocrystallites in the films, which might provide a fast Cu diffusion pathway, as compared to the single crystal structure of Ta N. The diffusivity of Cu in TaN0.62 films (~10-142/s) at 500 C [Lin00] was much higher than that in ALD-TaN films, which is consistent with the previous observation that high nitrogen content in the film is helpful in retarding Cu diffusion in TaN films. 5.4 Conclusions TaN was successfully deposited by ALD using alternating exposure of a Si wafer to TBTDET and NH3. The XRR thickness measurements as a function of exposure time confirmed that the adsorption of TBTDET wa s self-limiting as the deposition rate (2.6 10 cycle (16 ) 15 cycle (38 )

PAGE 93

79 /cycle) was independent of the TBTDET expo sure time in the relatively narrow time range 9 to 11 sec. The process temperat ure window for ALD with this precursor combination was also identified. The thic kness was constant (200 +/10 ) in the temperature range 200 to 300 C deposited with self-limiting growth conditions and 80 cycles. This temperature window for ALD mode was sufficiently wide to allow good controllability and reasonab ly low to contribute minimally to the thermal budget. A series of films was grown with variable cycle number from 5 to 100 on Si (100) in ALD growth mode. The film thickness w ith increasing cycle number shows non-linear behavior when the number of cycles is less than 15, suggesting the adsorption characteristics of TBTD ET on bare Si or SiOx/Si are different than those on TaN. After about 15 cycles the thickness varied linearly with cycle number, consistent with a growth rate of 2.5 0.2 /cycle and monolayer grow th per cycle. Cu diffusion barrier efficacy was evaluated by searching for Cu silicides using XRD and measuring the etch pit density on Secco-etched bared Si surfaces. The results show that ultra-thin ALD-TaN as thin as 38 is a good candidate for a diffusion barrier material, which satisfies the year 2013 roadmap 38 nm feature size node (3.5 nm thick diffusion barrier).

PAGE 94

80 CHAPTER 6 METAL ORGANIC ATOMIC LAYER DEPOSITION OF Ta-Al-N USING TBTDET, TEA AND NH3 6.1 Introduction Among the transition metals and their nitr ides, binary compounds such as TiN, TaN, and WN have been intens ively investigated for Cu di ffusion barriers. The drawback of binary nitrides and especially TiN, how ever, are their tendency to deposit with a columnar polycrystalline microstructure, whic h is an unfavorable configuration for a diffusion barrier [Nic95]. This grain structure creates boundaries that cut across the film and offer fast diffusion paths. To retard the formation of columnar grains, ternary materials such as (Ti,Ta,W)-Si-N and W-(B ,C)-N have been evaluated [Kim05]. The added degrees of freedom are expected to promote formation of an amorphous structure and its retention after subsequent high temperature processing. In this study, Ta-Al-N films were de posited by thermal ALD using TBTDET, TEA, and ammonia. TMA (trimethyl aluminum) ha s already been employed as an effective reducing agent in the ALD of TiN films that were deposited with an inorganic precursor (TiCl5) [Jup01]. The resulting films exhibited relatively low resistivity and low chlorine content (<4 at %). Furthermore, carbon cont amination in the films from TMA was not harmful to the diffusion barrier properties [E iz94]. Aluminum incorporation also made the film structure nanocrystalline or am orphous and thereby improved the barrier properties. Similarly, Ta-Al-N was deposited by ALD using TaCl5 or TaBr5 and NH3 as precursors and TMA as a reducing agent [Ale 01]. Metal organic AL D of Ti-Al-N using

PAGE 95

81 TDMAT and DMAH-EPP (dimethylaluminum hydr ide-ethyl-piperidine) as the titanium and aluminum precursors was also reporte d [Koo01]. The film structure remained amorphous even after high te mperature annealing at 900 C and the Ti and Al content could be changed simply by controlling the number of TDMAT-NH3 and DMAH-EPP cycles. The Ti-Al-N films showed higher Cu diffusion failure temperature than ALD TiN, as determined by XRD and observation of the etch-pit density in bare Si substrates treated with a Secco etch [K im02b]. On the other hand, another study showed that Ti-AlN did not exhibit superior results compared to ALD TiN, as judged by barrier test results using XRD, the etch-pit test, and resistiv ity measurements [Jup01]. Although there have been no reports of using TEA (triethyl aluminum) as an Al reagent in deposition of (Ti, Ta)-Al-N, TEA is expected to lower the car bon incorporation more than using TMA as evidenced by deposition studies of AlxG1-xAs and other III-V systems. In this work, the effect of Al incorporation in TaN using TEA on diffusion ba rrier performance of Ta-AlN films along with the microstruc ture and density are examined. Another focus of this work is to understand the mechanism of aluminum incorporation in Ta-Al-N films by investigating the film thickness, chemical composition, and bonding states when varying the exposur e time of TEA and the pulse sequence (i.e., Ta Al N (Sequence A), Al Ta N (Sequence B) and Ta N Al N (Sequence C)). For (Ti, Ta)-Si-N [Min00, Rei96], a change in the precursor exposure sequence induced a dramatic change in the film prope rties including Si content, resistivity, and growth rate. These changes we re attributed to strong SiH4 adsorption on the nitride surface that blocks adsorption of the metal precursor. A lower growth rate was observed when the SiH4-metal precursor-NH3 sequence was used, compared to the sequence metal

PAGE 96

82 precursorSiH4NH3 [Rei96]. For Ta-Al-N using TaCl5, the deposition rate and chemical composition also changed dependi ng on the sequence of the precursors. A higher growth rate and lower carbon and oxygen impurity levels were observed with the TaCl5-TMA-NH3 sequence rather than the TMA-TaCl5-NH3 sequence [Ale01]. The same tendency is expected to occur with the precu rsor combination used in this study since TEA reacts with ammonia and forms AlN (Sequence A). TBTDET and TEA are not known to form any compound, which should l ead to more oxygen incorporation in AlTa-N films (Sequence B) due to TEAs high affinity to oxygen. Therefore, the growth rate is expected to increase with less oxyge n incorporation when sequence A is used compared to sequence B. Sequence C is anti cipated to produce film properties similar to those produced using sequence A since TB TDET can form TaN without added NH3. 6.2 Experimental Details ALD of Ta-Al-N was performed on standa rd (100) silicon substrates in a customized cold-wall, vertical ALD reactor. Pneumatically actuated valves, controlled with a LaBView-based program interfaced to digital input/output interface board, sequentially switched reactant and purge st reams between the reactor and by-pass line. TBTDET (supplied by Alfa Aesar, MA, vapor pressure of 0.1 Torr at 90 C), contained in a stainless steel bubbler, was maintained a c onstant temperature at a set point of 80 C. The MO precursor was delivered to the reactor using a N2 carrier. TEA (supplied by Epichem, vapor pressure 0.02 Torr at 20 C,) was used as the aluminum source and N2 as the carrier gas. Ta-Al-N film s were deposited at 300 C at a typical growth pressure of 1.5 Torr. In sequence A, the standard depos ition cycle begins with an exposure of TBTDET for 9 sec using 20 sccm of N2 carrier gas. Then the MO is switched to vent and a nitrogen flow of 200 sccm is delivered to the reactor for 10 sec to ensure complete

PAGE 97

83 removal of residual TBTDET precursor as we ll as volatile byproducts. After the purge step, TEA is next introduced for 2 to 8 sec using a carrier gas flow rate of 20 sccm, followed by another purge step. Finally, NH3 at a flow 20 sccm is delivered to the reactor for 10 sec and followed by a final purge before the cycle is repeated. Sequences B and C were also operated with the same reactant flow rates and exposure times with the exception, of course, of reactant sequence. The film thickness was evaluated from the XRR profiles using the WinGixa program assuming that a very thin SiO2 layer (~ 20 ) exists at the interface between Si and TaAlN, and that C (~20 atom %), and O (~15 atom %) impurities reside in this top layer of the Ta-Al-N film. The stoichiometr y of each film was based on the chemical compositions estimated from AES measurements after 1 min sputtering. It is noted that no standard was available for calibrating the AES measurements. The thickness data from the simulation for each sample were confirmed by cross-sectional SEM measurements. The microstructure was analyzed using XRD and TEM. The film chemical composition and atomic bonding states were probed by AES and XPS measurements. To compare the Cu diffusion ba rrier properties of TaN and Ta-Al-N films, a 100 nm thick Cu film was sputtered depos ited on the ALD TaN or Ta-Al-N film. The samples were then annealed at 500 C for 30 or 45 min in N2 atmosphere and tested for Cu transport across the TaN or Ta-Al-N barrier. 6.3 Results and Discussion 6.3.1 Growth Rate and Chemical Composition The conditions that gave self-saturated growth of TaN ALD and reported in Chapter 5 were used to estimate the growth conditions for the ternary Ta-Al-N material. Specifically, the conditions TBTDET flow rate of 20 sccm with a pulse time 9 to 11 sec

PAGE 98

84 at 300 C with a fixed purge/NH3 exposure time and flow ra te (10 sec, 200 sccm/10 sec, 20 sccm) produced ALD growth characteristic s. For Ta-Al-N ALD these same pulse times and flow rates were used to deliver TBTDET, NH3, and the purge gas, while the TEA flow rate was 20 sccm and the pulse time was varied from 2 to 8 sec. As a preliminary study, the conditions that would yield self-limiting adsorption of TEA were sought by observing the growth rate with increasing exposure time of TEA. No report on the self-limiting growth of AlN when using TEA is believed to exist. Trimethyl aluminum (TMA), which is struct urally similar to TEA, was reported to be self-limiting on Si substrates at temperatur es below 650 K for doses greater than about 100 L (Langmuirs) [May91]. The same group also showed that NH3 exhibits siteselective reaction with Al-containing surface sp ecies at temperatures greater than 550 K. Their conclusion of self-limiti ng growth from TMA is based on quantitative estimates of the saturation coverage made from the inte grated XPS peak intensities, elemental sensitivity factors, and the inelastic m ean free paths for photoelectrons from Alcontaining films [May91]. A similar study on alumina substrates using TMA and NH3 at 600K showed that TMA adsorption/reaction on alumina is a self-limiting process, as judged from the lack of shif t in the C (1s) binding ener gy. The shift in the C binding energy would be consistent with continuous deposition, since the formation of Al4C3 by the incomplete pyrolysis and reaction of TM A will shift the C binding energy [Liu95]. In contrast to these reports, Ruhela et al. obser ved no saturation of the growth rate of AlN on Si was observed with increa sing TMA pulse time at growth temperature in the range 325 to 425 C [Ruh96]. This behavior was at tributed to the thermal decomposition of TMA in this temperature range.

PAGE 99

85 Figure 6-1 depicts the XRR profiles of AlN films deposited with various TEA pulse times at 300 C. The estimated growth rate per cycle is shown in Figure 6-2 along with the lattice constant of h-AlN (4.9 ). It is seen the growth rate in creased linearly with TEA pulse time and thus the use of TEA did not exhibit self-limiting behavior at 300 C. Thermal decomposition of TEA is believed to be the reason for nonself-limiting growth, as pyrolysis studies on TEA suggest that th e thermal decomposition of TEA starts around 200 C [Tis90]. Thermal decomposition of the precursor removes ligands to allow continued reaction. Apparently thermally decomposed TEA precursor allows continuous adsorption of aluminum atoms and causes CVD-like deposition. Figure 6-1. XRR profiles of ALD AlN as a function of TEA exposure time deposited at 300 C and 10 sec purges and NH3 exposure for 120 cycles

PAGE 100

86 Figure 6-2. Growth rate of AlN as a function of TEA expos ure time deposited at 300 C and 10 sec purges and NH3 exposure for 120 cycles Figure 6-3 shows the XRR profiles for film s of Ta-Al-N deposite d with sequence A and denoted by Ta-Al-NA. This set of films was grown by varying the exposure time of TEA source from 2 to 8 sec and using 120 cycles. The growth rate and chemical composition of Ta-Al-NA films as a function of TEA exposure time are plotted in Figure 6-4. As expected from the AlN result descri bed previously, the growth rate and atomic concentration of Al in Ta-Al-NA films increased with longer TEA pulse time, which is consistent with TEA not incorporating in Ta-Al-NA films in a self-limiting way. The nitrogen concentration remained almost cons tant in the range 27 to 30 at % and the Ta concentration decreased from 46 (2 sec) to 33 at % (8 sec), as the TEA exposure time increased. Considering that TaN can be deposited from a single source without NH3 and that AlN can be formed with TEA and NH3 in sequence A, it is reasonable to use the AlN thickness data from Figure 6-2 to estimate how much of the AlN thickness contributes to

PAGE 101

87 Ta-Al-NA thickness as TEA exposure time increase s from 2 to 8 sec. Following this hypothesis and as shown in Table 61, the portion of AlN in Ta-Al-NA (i.e., AlN thickness / Ta-Al-NA thickness) increased with incr easing TEA exposure time according to 28 % (2 sec), 56 % (4 sec), 63 % (6 sec), and 81 % (8 sec). On the other hand, the concentration of Al in the films as shown in Figure 6-4 increased fr om 12 at % (2 sec), 18 at % (4 sec), 20 at % (6 s ec) to 25 at % (8 sec), whic h is around 31 to 43 % of AlN thickness contribu tion to Ta-Al-NA (Table 6-1). In other words, although the Al concentration in a film is low, the contribution of AlN to the Ta-Al-NA film thickness is as high as three times of Al at % in Ta-Al-NA. This can be explained by the film density variation. The simulation of the XRR profiles of the AlN films showed that the density of AlN films deposited at 300 C is 3.1 to 3.3 g/cm3 (bulk AlN density is 3.3 g/cm3), which is much lower (~33 %) compared to the density of TaN (9.4 g/cm3). Therefore, the small atomic concentration of Al in Ta-Al-NA films contributes proportionally more to the total film thickness. Figure 6-3 also supports this co nclusion, as the critical angle of Ta-Al-NA, which is proportional to the film density, decr eased as more Al was incorporated into the film. The carbon concentration in the film was only 2.5 to 3.6 at %, which is considerably lower than 20 to 22 at % when deposited with TaCl5 or TaBr5, TMA, and NH3 [Ale01]. The oxygen concentration ranged from 10 to 11 at % after 1 min sputtering. Table 6-1. Estimated thickness of AlN contributing to Ta-Al-NA film thickness as a function of TEA exposure time TEA exposure time (sec) 2 4 6 8 AlN thickness () Figure 6-2 236 497 645 865 Ta-Al-NA thickness () 840 880 1030 1070 AlN thickness / Ta-Al-NA thickness (%) 28 56 63 81 Al concentration in Ta-Al-NA (at %) 12 18 20 25 [Al in Ta-Al-NA] / [AlN/ Ta-Al-NA] (%) 43 32 32 31

PAGE 102

88 Figure 6-3. XRR profiles of Ta -Al-N grown with Ta-Al-N se quence (A) as a function of TEA exposure time grown at 300 C and 10 sec purges and NH3 exposure for 120 cycles Figure 6-4. Growth rate and chemical co mposition of Ta-Al-N grown with Ta-Al-N sequence (A) as a function of TEA exposure time grown at 300 and 10 sec purges and NH3 exposure for 120 cycles

PAGE 103

89 Al-Ta-NB (deposited with sequence B) films were deposited with varying exposure time of TEA in the same 2 to 8 sec range. Figure 6-5 shows the XRR profiles of Al-TaNB films grown for 120 cycles as a functi on of TEA pulse time with fixed TBTDET, NH3, purge time and flow rates as previously specified. The chemical composition and the growth rate of Al-Ta-NB are shown in Figure 6-6. The growth rate of Al-Ta-NB films increased continuously from 4.8 to 7.2 /cy cle, as the TEA exposure time changed from 2 to 8 sec, which showed a smaller growth rate compared to that of Ta-Al-NA (7.0 to 8.9 /cycle). TEA might have difficulties in fo rming AlN films in sequence B, since the coverage of TBTDET over TEA will pr ohibit the reaction between TEA and NH3. In addition, NH3 will be consumed in the transamina tion reaction with TBTDET rather than forming AlN films. The carbon level in th e film was ~ 0 %, which is less carbon compared to that in Ta-Al-NA films (2.6 to 3.6 at %), probably resulting from the removal of carbon containing ligands thr ough transamination reaction. It is also reasonable that Al would react with any oxygen in the syst em since it cannot immediately form the nitride. In addition to the lack of AlN film formation, TaN deposited with TBTDET and NH3 is known to be denser than those grown with TBTDET alone at the same conditions, which is attributed to transamination reac tions [Kim06]. Note that the TBTDET pulse was followed by the NH3 pulse in sequence B, which, assuming no TaN formation before the NH3 exposure, will result in a denser and th us thinner TaN film than deposited with sequence A via transamination reaction. The N concentration dropped from 30 to 16 at %, as soon as Al was introduced for 2 sec and then remained relatively constant in the range 10 to 15 at % for longer TEA exposure times.

PAGE 104

90 The oxygen concentration ranged from 20 to 26 at %, which are almost 10 at % higher than measured in the Ta-Al-NA films. In the Ta-Al-NA sequence, neither the N or O levels significantly changed relative to the TaN result. The combined N and O levels in samples from both sequences, however, were similar and relatively independent of exposure time, suggesting that the lower nitr ogen content can be attributed to higher oxygen content in the Al-Ta-NB films. One possible reason for the added O is that the TBTDET has an oxygen impurity in the source an d oxidizes the Al adsorbed species. In the other sequence the Al firs t reacts with the am monia to N terminate the surface. Another possibility is that TBTDET exposure of TEA does not result in formation of AlN bonds and leads to Al-O bonding upon exposure to the oxygen present in the ammonia source. TEA is known to react aggressi vely with oxygen in the absence of NH3 to form Al-O bond states [Ruh96]. Figure 6-5. XRR profiles of Ta -Al-N grown with Al-Ta-N se quence (B) as a function of TEA exposure time grown at 300 C and 10 sec purges and NH3 exposure for 120 cycles

PAGE 105

91 Figure 6-6. Growth rate and chemical co mposition of Ta-Al-N grown with Al-Ta-N sequence (B) as a function of TEA exposure time grown at 300 and 10 sec purges and NH3 exposure for 120 cycles Films were also deposited with sequence Ta-N-Al-NC, which starts the deposition with exposure to TBTDET first and follows with an NH3 pulse, a TEA pulse, and another NH3 pulse. This sequence should form a ML of TaN before the TEA exposure. The XRR profiles of Ta-N-Al-NC films deposited for 120 cycles are shown in Figure 6-7 as a function of TEA pulse time. The corresponding film chemical composition and growth rate of Ta-N-Al-NC films are depicted in Figure 6-8 us ing the same scale as Figures 6-4 and 6-6. The growth rate and atomic concentration of Al in Ta-N-Al-NC films increased with longer exposure of TEA, implying that TEA did not incorporate into the films in a self-limiting way. The growth rate of Ta-N-Al-NC films was 0.3 to 1.1 /cycle less than that for Ta-Al-NA films grown under otherwise the same conditions. This is attributed to the transamination reaction occurring to a greater extent with the TBTDET and NH3

PAGE 106

92 exposure sequence, leading to denser and thus thinner film deposit ion compared to TaAl-NA films. The atomic concentration of Ta and Al in Ta-N-Al-NC films was slightly higher (1 to 3 at %) than that of Ta-Al-NA films, which is attributed to the complete reaction between MO sources and ammonia per each cycle in sequence C. The oxygen concentration ranged from 9 to 12 at %, which is smaller than that of Al-Ta-NB and similar to that of Ta-Al-NA. Figure 6-7. XRR profiles of Ta-Al-N as a f unction of TEA exposure time grown with TaN-Al-N sequence (C) at 300 with 10 sec purges and NH3 exposure for 120 cycles

PAGE 107

93 Figure 6-8. Growth rate and chemical co mposition of Ta-Al-N grown with Ta-N-Al-N sequence (C) as a function of TEA exposure time grown at 300 and 10 sec purges and NH3 exposure for 120 cycles 6.3.2 Chemical Bonding States Figure 6-9 shows XPS spectra of Ta 4f co re level of Ta-Al-N films deposited with each sequence using a Perkin Elmer PHI 5100 ESCA System with a Mg anode (h=1253.6 eV). The spectra were taken after sputter etching the surface for 1 min to remove any carbon and oxygen from exposure to the ambient (nominal sputter rate of 20 /min). Deconvolution of XPS spectra wa s performed using RBD Analysis Suite software. The XPS spectrum of ALD-TaN after sputteri ng is also shown to allow comparison of the chemical bonding states of Ta in pur e TaN. As shown in the bottom spectrum for TaN, only one chemical state with a doublet separated by 2.2 eV (Ta 4f 7/2 and Ta 4f 5/2) is evident. The Ta 4f core level binding en ergy (BE) of ALD TaN films was larger than

PAGE 108

94 that of metallic Ta (Ta 4f 7/2 21.9 0.1 eV) [Wu04], thus meta llic Ta was not detected in the film. The Ta 4f 7/2 core level BE of the ALD TaN was located at 23.20 0.05 eV due to the bond formation of Ta and N. This bi nding energy is higher than the previously reported value of 22.5 eV [Bad88] for PVD grown TaN0.07 and lower than 24.0 eV that was measured [Gor93] in CVD Ta3N5 with higher N concentration. It is in a good agreement with the stoichiometric CVD TaN reported value of 23.1 eV [Sas90]. The film grown with the Ta-Al-NA sequence yielded a spectrum with only one doublet chemical bonding state, which exhibits a Ta 4f 7/2 core level at 23.10 0.05 eV. This value is slightly shifted by 0.1 eV to a lower value as compared to that of pure ALD TaN, likely a result of the lower concentration of N in Ta-Al-NA caused by Al incorporation. It is well known that less N content in meta l nitride shifts the metal BE to the lower value as it approaches to the meta l-metal bond state. Note that N concentration decreased from 30 to 26 at % in Figure 64, when TEA was exposed for 6 sec. In contrast to films grown by sequence A, Al-Ta-NB films exhibited two different chemical states designated as P1 and P2. The Ta 4f 7/2 of P1 state was located at 22.60 0.05 eV, which is close and shif ted by +0.4 eV to that of Ta N. The P2 state has a higher BE (Ta 4f 7/2 26.05 0.05 eV) attributed to the formation of Ta-Ox bond. This P2 state has the same BE as previously report ed XPS results on TaN grown by sputtering [Wan03] and ALD [Wu04], whic h had the oxidation state of TaN. The strong electron affinity of oxygen causes more electrons to tr ansfer from Ta atoms to oxygen, increasing the BE of the Ta 4f core level. These resu lts indicate that TEA, which was not directly exposed to NH3 but instead TBTDET, reacts with the oxygen from an unknown source to form the Al-O bonds. This high oxygen conten t might have diffused and incorporated

PAGE 109

95 into TaN films and caused the N-Ta-O-Al bond sequence. Apparently, the oxidation bond state (P2) in Al-Ta-NB films generated by higher O cont ent shifted the Ta-N bond state (P1) more to lower energy (22.6 0.05 eV ), compared to that of TaN and Ta-Al-NA. In the final sequence (Ta-N-Al-NC), only one chemical bonding state for Ta 4f 7/2 core level was observed similar to that seen in the sequence A sample. It was located at 23.00 0.05 eV, which is a 0.2 eV shift to the right to the binding energy of Ta N, ascribed to less concentration of N in Ta-N-Al-NC than TaN films. Figure 6-9. XPS spectra of Ta 4f core level for Ta-Al-N depo sited with sequences A, B, and C and TaN at 300 and 120 cycles after 1 min sputtering Figure 6-10 depicts the XPS sp ectra of the N 1s core level of TaN and Ta-Al-N films grown with sequences A, B, and C. Th e N 1s core level BE of the ALD TaN was Ta 4f 7/2 of TaN: 23.2 eV

PAGE 110

96 observed at 397.60 0.05 eV, which is higher than the reported BE of N 1s in CVD Ta3N5 (396.9 eV) [Gor93] but the same as the BE of ion-beam-assisted grown stoichiometric TaN (397.5 eV) [Zha97]. Th is BE suggests that the Ta N bond in stoichiometric TaN is a strong ionic bond. Th is result is also consistent with the 4f 7/2 core level BE of Ta just prevented, whic h is very close to the value for Ta in stoichiometric TaN, and lower than that of Ta3N5. The spectrum of the Ta-Al-NA film had the N 1s core level at 397.40 0.05 eV, closer to that of the N 1s BE of Al-N ( 397.3 eV) [Nis02] than that of TaN (397.6 eV). This suggests that the N 1s core level in Ta-Al-NA is shared by Al and Ta, since it lies between the BE of Al-N and Ta-N. Both TaN and AlN films are expected to be deposited in each cycle in sequence A, since TBTDET itself can produce TaN films without a reaction with ammonia and TEA fo rms AlN films upon exposure to NH3. The Al-Ta-NB films showed the highest BE (397.70 0.05 eV) of N 1s core level. This BE is close to the reported BE valu e of N 1s in sputtered TaNO films (398 eV) [Jon22]. It is reasonable that the hi gh concentration of oxygen in Al-Ta-NB films results in TaNO bond formation. The high reactivity of Al with oxygen in the absence of NH3 contact in sequence B might be a factor fo r higher oxygen incorporation into the films. The intensity of the N bond peak decreased with the higher oxyge n concentration. The N 1s core level in Ta-N-Al-NC films was observed at 397.50 0.05 eV, located between the N 1s BE of Al-N and Ta-N peak s, indicating that the nitrogen in Ta-N-Al-NC forms bonds with both Al and Ta. In each cycl e, TaN and AlN layers are expected to form in sequence C due to the reaction of each MO source with ammonia.

PAGE 111

97 The BE of the Al 2p level in Ta-Al-NA and Ta-N-Al-NC are located at 74.50 0.05 eV (see Figure 6-11), correspondi ng to the reported BE of Al 2p in Al-N state (74.4 eV) [Nis02]. On the other hand, th e BE of Al 2p in Al-Ta-NB was 75.00 0.05 eV, which is 0.5 eV higher than that of Ta-Al-NA and closer to Al 2p in the Al-O bonding state (74.8 eV) [Nis02]. Consistent with previous results, the hi gh oxygen level in the films deposited with sequence B is likely to de velop Al-O bond rather than Al-N bond. The XPS spectra of Al-Ta-NB did not show any peaks associated with the Al-Al metallic bonding state (71.8 to 72.8 eV) [Nis02]. Figure 6-10. XPS spectra of N 1s core leve l for Ta-Al-N deposited with sequence A, B, and C and TaN at 300 and 120 cycles after 1 min sputtering N 1s of TaN: 397.6 eV

PAGE 112

98 Figure 6-11. XPS spectra of Al 2p core leve l for Ta-Al-N deposited with sequence A, B, and C at 300 C and 120 cycl es after 1 min sputtering 6.3.3 Comparison of Microstructure and Diffusion Barrier Performance It is well known that grain boundaries in polycrystalline films play an important role as a diffusion pathway for copper. Indeed the contribution of grain boundary diffusion is greater than lattice or dislocati on diffusion processes [Ele03]. In general, an amorphous film structure is obtained at low growth temperature. Low deposition temperature, however, typically induces hi gh impurity incorpor ation and high film resistivity. The other approach to obtain an amorphous structure is to add a third element (Si, Al, B, C) to the binary transition me tal nitrides (e.g., (Ti,Ta,W)-(Si,Al)-N and W(B,C)-N [Kim05]). Figure 6-12 shows the XRD patterns of Ta-Al-N films as a function of TEA exposure time in the range 0 to 6 sec deposit ed with sequence A at 300 C. The increase

PAGE 113

99 in the TEA exposure time led to lower h-TaN peak intensity. When the TEA pulse was longer than 6 sec, the h-TaN(100) peak comple tely disappeared, which indicates that Al addition to the binary TaN promoted forma tion of an amorphous structure. As grown TaN films give a low intensity h-TaN (100) pe ak in the XRD pattern. This suggests that the structure of as grown TaN is an amor phous matrix with a small amount of h-TaN crystalline phase distributed in the amorphous matrix. Cross-sectional TEM examination (Figure 6-13(a)) also shows that the struct ure of as grown TaN films contains a low density of nano-crystalline grains. A Select ed Area Diffraction Pattern (SADP) of the as grown TaN showed a ring pattern that include d weak spots, typical of a nano-crystalline material. On the other hand, the Ta-Al-N s howed a featureless mi cro-structure and no rings in the diffraction pattern (Figure 6-13(b)), indicating th e structure of Ta-Al-N films is completely amorphous. Figure 6-12. XRD patterns of Ta-Al-N grow n with sequence A as a function of TEA exposure time at 300 C and 120 cycles

PAGE 114

100 Figure 6-13. Cross-sectional TEM mi crographs of TaN (a) and Ta-Al-NA (b) deposited at 300 C with SAD patterns The insertion of Al into TaN promoted fo rmation of an amorphous structure in the films, which is a very desirable for diffusion barriers. As more Al was incorporated into the film, however, its density became much lower as noted in the XRR profiles (Figure 614). The densities of Ta-Al-N films deposited with 6 sec of TEA exposure time for each of the sequences were 6.6 g/cm3 (Sequence A), 5.6 g/cm3 (Sequence B), and 6.3 g/cm3 (Sequence C), which are considerably lower than TaN density (9.4 g/cm3) but higher than the density of AlN (3.1-3.3 g/cm3). As discussed previously, the contribution of AlN to the thickness of the total Ta-Al-N film is thus much greater, while the Al concentration in the films was less significant, suggesting th at most of Ta-Al-N films deposited with sequence A consist of an estimated 48 % of AlN contribution to the total film thickness, which would degrade the diffusion barrier pe rformance because of the low density of AlN. Si Ta-Al-N Pt Pt Si TaN (a) (b)

PAGE 115

101 Figure 6-14. XRR profiles of TaN, Ta-Al-NA and AlN deposited at 300 C for 120 cycles To compare the diffusion barrier perf ormance, 100 nm thick Cu was sputterdeposited onto both TaN/Si and Ta-Al-N/Si structures. The thickness of both TaN and Ta-Al-N films on Si was grown to 6 nm to better compare the two materials. Figure 6-15 shows the XRD profiles of Cu/Ta(Al)N/Si stru ctures after anneali ng the structure in nitrogen at 500 C for 30 or 45 min. Both films showed XRD reflections for Cu15Si4, suggesting that barrier failure occurred in the Cu/Ta(Al)N/Si structure when annealed for 45 min. On the other hand, both films anneal ed for 30 min showed strong peaks assigned to TaN and Cu but not the copper silicide peaks, which indicates both layers were successful Cu diffusion barriers at the shorter anneal time. After annealing for 30 min, c-TaN(111) p eaks appeared in the pattern and they completely vanished after the 45 min ann ealing of both films. Cu diffusion through Ta(Al)N layer caused by further annealing is be lieved to change the c-TaN structure. The

PAGE 116

102 same Cu diffusion performance for both TaN and Ta-Al-N films may be attributed to an offset between the increase in amorphous stru cture and the decrease of film density by inserting Al into TaN films. Unfortunatel y, Ta-Al-N films deposited with longer Al exposure time (> 8 sec) showed a poorer performance compared to TaN, mainly attributed to a greater decreas e in the film density. In addi tion, Al insertion significantly increased the resistivity, which could not be measured by 4 point probe. Figure 6-15. XRD patterns of Cu/TaN and Cu /Ta-Al-N/Si structures annealed at 500 for time as noted. 6.4 Conclusions The ternary Ta-Al-N material was succe ssfully deposited by flow switching the reactants TBTDET, TEA, and NH3 as precursors. Preliminar y experiments attempting to identify conditions for ALD of AlN did not locate a regime that gave the self-limiting growth feature characteristic of ALD. A dopting the conditions th at gave ALD growth behavior of TaN, a series of experiments were performed to explore the properties of the

PAGE 117

103 ternary material. The concentration of Al in Ta-Al-N films could be controlled by adjusting TEA exposure time, however, its incorporation into the films was not selflimiting. This was attributed to the th ermal decomposition of TEA at the growth temperature of 300 C. Intere stingly, changing the reacta nt exposure sequence yielded films with different properties (i.e., chem ical composition, bonding state, and growth rate). Higher oxygen content with lower growth rate was observed with the Al Ta N sequence, as compared to the films deposited with the Ta Al N sequence. High oxygen content films grown with the Al Ta N sequence resulted in the formation of Ta-O and Al-O bond states, which were not detected in the films grown with the Ta Al N sequence. Although Al integration in to TaN ensured the amorphous film structure, it also significantly lowered the overall film density. A comparison study of the barrier performance of selected films showed that barrier failure occurred for both TaN and Ta-Al-N films at the same conditions, which is Cu/Ta-(Al)-N/Si structure annealed at 500 C for 45 min, suggesting an offset between increased amorphous content and decreased film density.

PAGE 118

104 CHAPTER 7 PORE SEALING TREATMENTS OF LOWCDO FILMS TO PREVENT Ta PRECURSOR PENETRATION DURING TaN ALD 7.1 Introduction As the increasing density of devices pushe s the microelectronic device feature size down, a large portion of the tota l circuit transmission time is delayed due to the parasitic resistance (R) and capacitance (C) of interc onnects. Therefore, the traditional Al-Cu, SiO2-based interconnect metallization is being replaced with the Cu, lowdielectric combination since they enable the electric si gnal to move faster by reducing the RC time delay. Lowdielectrics seem more promising than low resistivity metal substitution because capacitance reduction not only diminishes cross-talk between parallel running lines but also reduces the power dissipati on, which is proporti onal to the capacitance [Liu02]. Therefore, there have been extens ive studies of organic and inorganic lowmaterials as a replacement for conventional SiO2 dielectrics. In organic lowmaterials, organic groups are intr oduced into a silicon oxide-based matrix to induce free volume (pore size < 1 nm) and decrease the dielectric constant (~ 3.0). Most loworganosilicate glasses (OSG) c ontain Si-R groups, leading to a hydrophobic surface and lower density by replac ing the tetrahedral Si-O bonding with low polarizability Si-R bonds. Carbon-dope d silicon oxide (SiCOH) is commonly deposited by PE-CVD and is a promising dielectric material due to its low dielectric constant with similar electrical and integration characteristics to SiO2 [Leo06]. Templated or porogen removal techniques have been used to produce more homogeneous pores

PAGE 119

105 leading to significant reduction in the di electric constant. Large porosity volume (approaching 45% porosity) and pore size (down to 3 nm) are required to obtain ultra low(ULK) materials ( < 2.2). This large porosity, how ever, often leads to mechanical and chemical interaction issues in integrating lowdielectrics into devices [Fay03]. As examples, cohesive delamination, mate rial deformation, and crack formation have been reported to occur when attempting to integrate lowdielectric materials into device processing steps such as CMP (Chemi cal Mechanical Polishing) and packaging. Addition of a metal dummy layer for capping lowfilms [Tsa02] or new CMP and packaging processes with low shear forces [Klo02] have had to be developed to overcome these issues. In addition to the mechanical issues of lowdielectrics, metal barrier deposition on porous ULK materials is a more serious ch allenge for ULK integration. PVD barriers deposited on porous dielectrics are reported to have pinholes (discontinuous), which could only be removed by growing thick film s [Bak01, Iac02]. Chemical species from CVD or ALD process can easily diffuse into ULK material given its intrinsic large porosity. Penetration of chemical s into ULK dielectrics can ch ange the material structure and permanently modify physical and chem ical properties of ULK materials [Hoy04]. Diffusion of Ti ALD precursor into CVD SiCOH films dur ing TiN ALD increased the leakage current between metal lin es and degraded the effective value [Bey02a]. Similarly, Ti was detected in HSQ (Hydr ogen Silsesquioxane) f ilms after TiN ALD, indicating a serious diffusion of Ti precursor into the lowmaterial during metal barrier deposition. The same observations were reported on MSQ (Methyl Silsesquioxane) and hybrid dielectric materials with both TiN and WCN ALD [Bey02b]. On e approach is to

PAGE 120

106 seal the surface of the lowmaterial prior to the metal barrier deposition process to prevent the precursor diffusion. Various plasma treatments have been empl oyed to seal the pores on the surface of lowfilms. Ion bombardment in the plasma chamber densifies the top surface of lowfilms, potentially leading to isolation of the bulk lowfilms from external substances. The densified top layer, howev er, typically showed a higher value compared to the bulk lowfilms. O2-based plasma treatments on Si-O-C and -SiC:H before TiN ALD enriched the dielectric surface with OH gr oups and improved the quality of subsequent TiN films [Sat02b]. O2 plasma treatments with various conditions (r.f power, duration time, and pressure) were carried out to dete rmine the optimal conditions for pore sealing. Unfortunately the O2 plasma treatment damaged the CDO films by removing carbon from the near surface region of the film, which cause d the dielectric constant to increase and film thickness to decrease [Abe04]. A high density capping layer on the lowmaterial can improve the quality of the metal diffusion layer by successfully preventi ng precursor penetra tion during diffusion barrier deposition [Bon03]. A PECVD grown SiN cap layer showed excellent resistance to Ti precursor penetration into SiO2 aerogel films, while the Ti had penetrated uncapped SiO2 aerogel films. A MC (M olecular Caulking) layer was deposited on MSQ films using the Gorham method with [2,2]-paracyclophane precursor. The molecular caulking layer effectively prevented metal precurso r penetration into MSQ during metal MOCVD process [Jez04]. Capping SSQ (Silsesquioxane) f ilms with SiN improved the resistance to stress-corrosion cracking for all leve ls of network formation [Toi02].

PAGE 121

107 In this work, two different surface treatme nts are applied to the surface of Carbon Doped Oxide (CDO) films to prevent or redu ce Ta precursor penetration into CDO films during barrier deposition. CDO films (~1 thick), which were deposited on Si by PECVD, have a dielectric constant of 2.5 a nd a pore volume close to 25 %. One approach to close the pores was O2 plasma treatment, which is expected to densify the surface layers by stripping the C content and sealing the pores on the surface of the CDO films. The other approach was to deposit a SiN ca pping layer on the CDO material. After the surface treatment, the contact angle (CA), surface roughness, chemical bonding, and film density of the CDO surface or near surface region were characteriz ed using CA meter, AFM, XPS, and XRR, respectively. An ALD Ta N film was then deposited on the treated CDO and the efficiency of each surface treat ment was evaluated by measuring the Ta penetration with TEM-EDX. 7.2 Experimental Details The test CDO wafer was provided by Stev e Johnston from the Intel Corporation. SooHwan Jang from Dr. Fan Rens group in Chemical Engineering performed the O2 plasma treatment and SiN capping layer deposition on the CDO films. An O2 plasma was vertically directed to the surface of CDO film for 18 to 42 sec at 300 mTorr and room temperature in a plasma chamber (Techinics micro-RIE Series 800-II). The O2 flow rate was fixed at 30 sccm while the plasma power was varied in the 100 to 300 W. The SiN capping layer (25 to 50 nm) was deposited on CDO films using silane (5 sccm) and NH3 (30 sccm) as precursors by PECVD (Unaxis SLR 730 PECVD). The growth temperature was set at 250 C with the pressure of 300 mTorr. TaN was grown on as-deposited and surface treated CDO films in the same growth run by ALD. TBTDET (supplied by Alfa Aesa r, MA, vapor pressure 0.1 Torr at 90 C),

PAGE 122

108 contained in a stainless steel bubbler, was h eated to a constant temperature and depending on the run was set at a temperature in the range 70 to 80 C. The precursor was delivered to the reactor with a of N2 carrier gas. The typical growth temperature ranged from 200 to 400 C with a growth pressure of 1 Torr. The deposition cycle began with an exposure to TBTDET for 9 sec with 20 sccm of N2 carrier gas. Following the first TBTDET pulse, nitrogen at 200 sccm was introduced for 10 sec to purge of reactor of the TBTDET precursor as well as volatile byp roducts. After this purge step, NH3 at 20 sccm was introduced for 10 sec, followed by another N2 purge step. The film thickness was measured by cro ss-sectional TEM and AFM was used to assess the surface morphology. The sample preparation for TEM measurement using FIB (Focused Ion Beam) was performed by S eemant Rawal from Materials Science & Engineering Department. Water contact angle data (average value of 10 points) were collected using an OCA20 tool equipped w ith a goniometer. The chemical bonding in the near surface region was probed by XPS and th e penetration of the Ta precursor was studied by TEM-EDX. The film density was extracted from XRR measurements using the WinGixa software. 7.3 Results and Discussion 7.3.1 Chemical Bonding and Co ntact Angle of CDO Films Figure 7-1 shows a typical FTIR spectrum of an as deposited CDO film with major vibrational chemical bonds identified on th e figure. The dominant stretching mode at 1040 cm-1 indicates that Si-O network is the b ackbone of this CDO fi lm. The shoulder at 1130 cm-1 is a sign of some degree of Si-O cage st ructure formation. Vibrational bands at 800 cm-1 corresponds to Si-C and Si-CH3 wagging structure. Si-C bending mode at 1270 cm-1, C-Hn at 2900 cm-1, and Si-Hn stretching mode at 2100 to 2300 cm-1 were observed.

PAGE 123

109 The FTIR spectra of both plasma sealed a nd SiN capped CDO (not shown here) did not show any significant difference with that of as-deposited CDO. This indicates that the effect of these treatments is limited to the surface region. Modification of the CDO surface after the surface treatment could be observed by XPS measurement. As-deposited CDO exhib ited two chemical bonding states at 99.8 and 103.5 eV, representing the Si-C (100.2 to 101.5 eV) and Si-O (103.4 to 103.8 eV) bonding states, respectively [Nis02]. The O2 plasma treatment, however, almost completely removed the Si-C bond peak, l eaving only the Si-O bond state at 103.4 eV. This suggests that O2 ion bombardment stripped the carbon content from the top layers of CDO, and potentially resulted in the isolation of bulk lowfilms from external substances. O2 radicals are known to react with met hyl functional groups such as Si-CH3 and Si-H bonds, leaving Si da ngling bonds behind. The reactive Si dangling bonds easily absorb oxygen from ambient environments and form Si-O bonds. The stripping of carbon from the CDO surf ace was also confirmed by contact angle measurements. The measurements revealed that the surface of the as deposited CDO film was hydrophobic with a contact angle of 82o. The contact angle, however, significantly decreased to 32o after the surface was exposed to the O2 plasma. The plasma treated surface was hydrophilic with a contac t angle close to that of SiO2 (Table 7-1). The change from hydrophobic to hydrophilic surface is attributed to enrichment of the dielectric surface with OH groups. The SiN capped surface displayed only one chemical bonding state at 101.9 eV, representing th e Si-N bonding state (101.5 to 102.3 eV) [Nis02]. No peaks corresponding to Si-O or Si-C could be observed, indicating the entire CDO surface was successfully covered with the SiN capping layer. Unlike the porous

PAGE 124

110 lowmaterials, PECVD grown SiN dose not ha ve pores in the film, which increases resistance to precursor penetr ation. In addition, the measured value of the contact angle was close to zero indicating a highly hydrophilic feature (Tab le 7-1), which should also improve the quality of a diffusion barrier. Figure 7-1. FTIR spectrum of as deposited CDO film on Si Figure 7-2. XPS spectra of as received, O2 plasma treated, and SiN capped CDO films

PAGE 125

111 Table 7-1. Contact angle of as deposited, O2 plasma treated, SiN capped CDO films, and SiO2 CDO as deposited O2 plasma treated CDO SiN capped CDO SiO2 Contact Angle 82o 32o ~ 0o 22o 7.3.2 Density and Surface Roughness of CDO Films Both as-deposited and surface treated CDO films were probed by XRR (Figure 7-3) to assess their near surface region density. The XRR data were analyzed using the WinGixa program by fitting the raw da ta to a simulated profile with 2 ~ 10-2 accuracy. The critical angle of the as -deposited CDO film was 0.33o. This simulation result corresponds to a value for the as de posited CDO density of 1.26 g/cm3, which is typical of lowfilms. For the O2 plasma treated CDO films, the critical angel was 0.38o, indicating a slightly denser st ructure than the as-deposited CDO film. In the simulation, the top surface was assumed to contain no carbon (0 at %) due to carbon depletion confirmed by XPS. The simulation outcome sugge sts that a ~10 nm top layer has a higher density of 2.80 g/cm3 compared to the bulk CDO film density of 1.26 g/cm3. The densified top layer should efficiently inhibi t precursor penetration into the CDO film. The critical angles of the 25 and 50 nm thick SiN capped CDO films were both located at 0.46o, indicating the density of the SiN film is 2.70 g/cm3, which is about twice that of the untreated CDO film. This SiN cap layer, which has almost same density as SiO2, should also increase the resistance to th e Ta precursor penetration into the CDO film. The 50 nm thick SiN capped sample showed more fringes than the 25 nm thick sample because of the additional X-ray reflecting layers in the thicker film.

PAGE 126

112 The surface roughness after the treatments was probed by AFM (Figure 7-4). As expected, the roughness of the plasma treate d CDO film was greater than both the asdeposited and SiN capped CDO films due to the surface damage caused by the plasma treatment. The RMS value of the surface r oughness of the as-deposited CDO and 25 nm thick SiN capped CDO films was 4.6 and 4.3 respectively, while it increased to 8.9 for the O2 plasma treated sample. It is importa nt to maintain the surface roughness of lowfilm below a certain value to ensure ba rrier continuity and good sheet resistance. Figure 7-3. XRR-profiles of as deposited, O2 plasma treated, and SiN capped (25 and 50 nm) CDO films

PAGE 127

113 Figure 7-4.Surface roughness of (a) as received, (b) O2 plasma treated, and (c) SiN capped (25 nm) CDO films z: 10 nm div 7.3.3 Penetration of Ta precursor into CDO Films TaN was deposited on CDO films (as-deposited, O2 plasma treated, and SiN capped) by the alternating exposure of TBTDET and NH3. It was necessary to keep the deposition temperature less than 350 C to avoid TaN delamination during deposition because of the thermal instability of C DO. Figure 7-5 shows cross-sectional TEM images of a set of TaN films deposited at 300 C on as-deposited and surface treated RMS: 4.3 (c) RMS: 8.9 (b) RMS: 4.6 (a)

PAGE 128

114 CDO films with 120 cycles. All TaN films we re grown in the same run. No voids or cracks in TaN films were observed and their thickness was 25 nm (on as-received), 35 nm (on O2 plasma treated), and 25 nm (on SiN cappe d), producing a growth rate of 2.1 to 2.9 /cycle, as shown in the cross-sectiona l TEM images (Figure 7-5). This range of growth rate suggests that TaN was deposited on CDO in a self-limiting growth mode as evidenced by the growth rate of ALD-TaN from Chapter 5 (expected 2.45 /cycle). Cross-sectional specimens were prepared using a FE I Strata DB235 Focused Ion Beam (FIB) system. The TEM images and imbedded EDX line scans are collected using a JEOL 2010F TEM and shown in Figure 7-5 for samples with ALD TaN grown on the untreated CDO/Si st arting material, O2 plasma treated, and SiN capped (25 nm) structures. Figure 7-5 (a) demons trates that the Ta precursor penetrates a few nm into the untreated CDO film. The thin slightly dark layer between the dark TaN film and light bulk CDO film presumably resulted from infiltration of the Ta precursor into porous CDO films. The EDX line scan also supports the observation of Ta precursor penetration showing a small Ta signal intensity in the thin slightly darker region. The cross-sectional image for the TaN/CDO/Si grown on the O2 plasma treated CDO film is shown in Figure 7-5 (b) along with an EDX line scan. Consistent with the AFM result, the interface between the TaN and O2 plasma treated CDO film retains the roughness caused by the ion bombardment during the plasma treatment. The ~10 nm darker interface region is attributed to th e densification of th e near surface region produced by the plasma treatment. In contrast to the EDX scan on the TaN/as-deposited CDO sample, no Ta signal could be detect ed in the dark in terface area, strongly

PAGE 129

115 suggesting that the densified top layer prohib ited the Ta precursor from penetrating into the CDO during TaN ALD. Compared to the diffuse TaN/CDO interface in the as-deposited and plasma treated structures, the TaN/SiN interface is very ab rupt (Figure 7-5 (c)). The Ta EDX profile shows no Ta penetration into the CDO and onl y a very small amount into the SiN layer. The higher density of the SiN compared to the CDO is apparently a very efficient barrier for Ta precursor diffusion. This result sugge sts that the SiN cap thickness could possibly be significantly reduced to 3~5 nm judging fr om the Ta diffused into SiN (Figure 7-5 (c)). It is concluded that as received CDO with ~25 % porosity was vulnerable to the Ta precursor penetration unless its porous surf ace is sealed or closed before TaN ALD process. O2 plasma treatment was successful in densifying the top surface (~10 nm) of CDO and prevented the Ta precursor diffu sion, but unfortunately it produced a rough TaN/CDO interface, which would reduce barr ier continuity and sheet resistance. Although the change of was not measured in this experi ment, there is a possibility of increase in the value of after O2 plasma treatment, since O2 plasma depletes the C from the surface. SiN layer was also a very efficien t barrier for Ta precursor penetration. This hydrophilic and high density SiN layer gave an abrupt TaN/SiN interface. The SiN (3~5 nm) layer, however, would add to the total ba rrier thickness, which needs to be thinner as the device size shrinks. Considering the pros and cons of each method, the SiN capping is more likely to be adapted as a method of sealing pores of low, only if the cap layer thickness can be significantly reduced.

PAGE 130

116 Figure 7-5.Cross sectional TEM images of (a) TaN/as-deposited CDO/Si, (b) TaN/O2 plasma treated CDO/Si, and (c) TaN/ SiN capping layer/ CDO/Si structure superimposed with line EDX Ta signal 7.4 Conclusions The compatibility of ALD TaN barrier with porous lowCDO films was investigated. Two different surface treatments (i.e., O2 plasma exposure and SiN capping) were applied to the surface of CDO films to prevent Ta prec ursor penetration into this lowmaterial. A preliminary study of the untr eated CDO surface showed that growth of ALD TaN at 350 C resulted in Ta penetration to a depth of ~5 nm. Examination of the

PAGE 131

117 O2 plasma treated CDO indicated carbon deplet ion from the near surface region of the CDO leaving primarily Si-O bonds with its hydrophilic surface ch aracteristic. The O2 plasma treatment was effective in densifying the surface (2.80 g/cm3), but unfortunately also roughened the surface (AFM RMS roughness of 8.9 ) Use of a hydrophilic SiN capping layer also gave a high de nsity barrier layer (2.70 g/cm3) with an abrupt interface with TaN. The SiN capped CDO films showed no diffusion of Ta precursor. It is concluded that SiN capping layer is a be tter method to seal CDO pores since O2 plasma roughens the interface and possi bly leads to the value of to increase as a result of C depletion.

PAGE 132

118 LIST OF REFERENCES [Abe04]. T. Abell, K. Maex, Mi croelectronic Eng., 76, 16 (2004) [Ale01]. P. Alen, M. Juppo, M. Ritala, T. Sa javaara, J. Keinonen, M. Leskela, J. Electrochem. Soc., 148(10), G566 (2001) [Ale02]. P. Alen, M. Juppo, M. R itala, M. Leskela, T. Sajavaara, J. Keinonen, J. Mater. Res. 17, 107 (2002) [Ang05]. M. Angyal, J. Gill, M. Inohara, Proceedings of AMC 2005, Colorado Springs 125 (2005) [And95] J. D. Anderson, Computationa l Fluid Dynamics, New York (1995) [Adv06] http://www.advceramics.com, August (2006) [Bad88]. Y. Bada, T. A. Sa saki, I. Takano, J. Vac. Sci. Tech. A, 6, 2945 (1988) [Bak01]. M. R. Baklanov, C. Jehoul, C. M. Flannery, K. P. Mogilnikov, R. Gore, D. Gronbeck, G. Prokopowicz, C. Sullivan, Y. You, N. Pugliano, M. Gallagher, Proceedings of AMC 2001, Tokyo, 279 (2001) [Bas03]. C. Basceri, N. Maity, K. Holtzclaw, V. D. Hou, S. Rmamrajan, D. Chopra, S. W. Russell, Conference Proceedings ULSI, XVIII, Montreal, 725 (2003) [Bch03]. O. J. Bchir, S. W. J ohnston, A. C. Cuadra, T. J. Anderson, C. G. Ortiz, B. C. Brooks, D. H. Powell, L. M. White, J of Crystal Growth, 249, 262 (2003) [Bch04]. Omar J. Bchir, Chemical vapor de position of thin films for diffusion barrier applications, Doctoral Thesis, ChE, U of Florida (2004) [Bec03]. J. S. Becker, S. Suh, S. Wang, R. G. Gordon, Chem. Mater., 15, 2969 (2003) [Ben00]. B. G. Streetman, Sanjay Bane rjee, Solid State Electronic Devices 5th edition, New York (1999) [Bey02a]. G. Beyer, A. Satta, J. Schuhmacher K. Maex, W. Besling, O. Kilpela, H. Sprey, G. Tempel, Microelectronic Eng., 64, 235 (2002) [Bey02b]. G. Beyer, A. Satta, J. Schuhmacher K. Maex, W. Besling, O. Kilpela, H. Sprey, G. Tempel, Microelectron Eng. 64, 233 (2002)

PAGE 133

119 [Bir02]. R. B. Bird, W. E. Stewart, E. N. Lightfoot, Transport Phenomena 2nd edition, New York (2002) [Bla97]. E,.Blanquet, A. M. Dutron, V. Ghetta C. Bernard, R. Madar, Microelectronic Engineering 37/38, 189 (1997) [Bon03]. J. Bonitz, S.E. Schulz, T. Gessn er, Microelectronic Eng., 70, 330 (2003) [Bra03]. B. H. Brad, Tungsten nitride depos ition using MO precursor and comparison study about diffusion barrier prop erties with other barrier materials, Master Thesis, ChE, POSTECH (2003) [Che97]. X. Chen, G. Peterson, T. Stark, H. L. Frisch, A. E. Kaloyeros, Conference proceedings VLSI Multilevel Interco nnection Conference, Tokyo, 434 (1997) [Cho98]. S. L. Cho, K. B. Ki m, S. H. Min, H. K. Shin Materials Research Society Conference Proceedings, San Francisco, 514, 531 (1998) [Cho99]. S. L. Cho, K. B. Ki m, S. H. Min, H. K. Shin, S. D. Kim, Journal of the Electrochemical Society 146(10), 3724 (1999) [Dub92]. L. H. Dubois, B. R. Zegarski, G. S. Girolami, J. Electrochem. Soc., 139, 3603, (1992) [Dub94]. L. H. DuBois, Po lyhedron, 13, 1329 (1994) [Ede02]. D. Edelstein, C. Uzoh, C. Cabral Jr., P. DeHaven, A. Simon, E. Cooney III, S. Malhotra, D. Klaus, H. Rathore, B. Ag arwala, D. Nauyen, Conference proceedings ULSI, XVII, Boston, 541 (2002) [Eiz94]. M. Eizenberg, K. Littau, S. Ghangyem, A. Mak, Y. Maeda, M. Chang, A. K. Shina, Appl. Phys. Lett., 65, 2416 (1994) [Ela01]. J. W. Elam, C. E. Nelson, R. K. Gr ubbs, S. M. George, Thin Solid Films, 386, 41 (2001) [Ela03]. J.W. Elam, M. Schuisky, J. D. Fer guson, S. M. George, Thin Solid Films, 436, 145 (2003) [Ele03]. K. E. Elers, V. Saanila, W. M. Li, P. J. Soininen, J. T. Kostamo, S. Haukka, J. Juhanoja, W.F.A. Besling, Thin Solid Films, 434, 94 (2003) [Eng02]. E. R. Engbrecht, Y. M. Sun, S. Smith, K. Pfiefer, J. Bennett, J. M. White, J. G. Ekerdt, Thin Solid Films, 418, 145 (2002) [Fay03]. M. Fayolle, G. Passemard, O. Louveau, F. Fusalba, J. Cluzel, Microelectronic Eng., 70, 255 (2003) [Fix93]. R. Fix, R. G. Gordon, D. M. Hoffm an, Chemistry of Materials, 5(5), 614 (1993)

PAGE 134

120 [Flu06] Fluent Manual from www.fluent.com August (2006) [Gor90]. R. M. Fix, R. G. Gordon, D. M. Hoffman, J. Am. Chem. Soc., 112, 7833 (1990) [Gor93]. R. Fix, R. G. Gordon, D. M. Hoffman, Chem. Mater., 5, 614 (1993) [Got05]. K. Goto, S. Hashii, M. Matsumoto, N. Miura, T. Furusawa, M. Matsuura, A. Ohsaki, Proceedings of AMC2005, Colorado springs, 45 (2005) [Gre02]. M. L. Green, M. Ho, B. Busch, G. D. Wilk, T. Sorsch, J of Appl. Phys., 92, 7168 (2002) [Han03]. Hand, A., Semiconductor International, 26(5), 46 (2003) [Har01]. Hara, T., K. Sakata, Electrochemical and Solid-State Letters, 4(10), G77 (2001) [Hie74]. K. Hieber, Thin Solid Films, 24, 157 (1974) [Hil88a]. L. Hiltunen, M. Leskela, M. Makela, L. Niinista, E. Nykanen, P. Soininen, Thin Solid Films, 166, 149 (1988) [Hil88b]. L. Hiltunen, M. M. Leskela, M. Make la, L. Ninisto, E. Nykanen, P. Soinen, Thin Solid Films, 166, 154 (1988) [Hol90]. M. M. B. Holl, P. T. Wolczanski, G. D. Vanduyne, J. Am. Chem. Soc., 112, 7989 (1990) [Hos00]. Hoshino, A., S. T., S. Hiiro, H. M achida, A. Ogura, Y. Ohshita Proceeding of AMC 2000, San Diego, 403 (2000) [Hoy04]. A. Martin Hoyas, J. Schuhmach er, C.M. Whelan, J.P. Celis, K. Maex, Microelectric Engineering, 76, 32 (2004) [Hum05]. A. Humbert, L. Mage, C. Goldberg, K. Junker, L. Proenca, J. B. Lhuillier, Microelectronic Eng., 82, 399 (2004) [Iac02]. F. Iacopi, Z. Toekei, D. Shamiryan, Q. T. Le, S. Malhouitre, M. Van Hove, K. Maex, Proceedings of AMC2002, Tokyo, 29 (2002) [Int06] http://www.intel.com/technology/si licon/65nm_technology.htm, August (2006) [Ist00]. A. A. Istratov, E. R. Weber, El ectrochemical Society Conference Proceedings, Vol.2000-27, Toronto, 90 (2000) [Ito05]. F. Ito, T. Takeuchi, Y. Hayashi, Proceedings of AMC2005, Colorado springs, 22 (2005)

PAGE 135

121 [Jez04] C. Jezewski, C.J. Wiegand, D. Ye, A. Mallikarjunan, D. Liu, C. Jin, W. A. Lanford, G. Wang, J. J. Senkevich, T. Lu, J. of Electrochemical Soc., 151 (7), F157 (2004) [Jon22]. C. A. Jong, J. Y. Gan, T. S. Chin, Jpn. J. Appl. Phys., Part I 41, 5367 (2002) [Jup00]. M. Juppo, M. Ritala, M. Keskel, J. Electrochem. Soc., 147, 3377 (2000) [Jup01]. M. Juppo, P. Alen, M. Ritala, M. Le skel, Chem. Vap. Dep., 7(5), 211 (2001) [Kal00]. A. E. Kaloyeros, E. Eisenbraun, Annu. Rev. Mater. Sci., 30, 363 (2000) [Kap02]. P. Kapur, J. P. Mc Vittie, K. C. Saraswat, I EEE Transactions on Electron Devices, 49(4) 590 (2002) [Kim02a]. J. Y. Kim, H. K. Kim, Y. Kim, Y. D. Kim, W. M. Kim, J. Korean Soc., 40, 196 (2002) [Kim02b]. H. Kim, S. M. Rossnagel, 49th International American Vacuum Society conference, Denver, 46 (2002) [Kim05]. H. Kim, IBM research report, RC22737, (W0303-012) (2005) [Kim06]. K. C. Kim, T. J. Ande rson, L. M. White, to be published [Kim91]. Y. T. Kim, S. K. Min, A pplied Physics Letters 59(8), 929 (1991) [Kla00]. J. W. Klaus, S. J. Ferro, S. M. George, Thin Solid Films, 360, 145 (2000) [Klo02]. G. Kloster, T. Scherban, G. Xu, J. Blaine, B. Sun, Y. Zhou, Proceedings of IITC, IEEE, Burlingame, 242 (2002) [Kol01]. E. Kolawa, P. J. Pokela, J. S. Reid, J. S. Chen, M. A. Nicolet, Appl. Surf. Sci. 53, 373 (1991) [Koo01]. J. Koo, J. W. Lee, T. Poh, Y. Kim, Y. D. Kim, H. Jeon, J. Vac. Sci. Tech., A19, 2831 (2001) [Lee93]. Lee, C. W., Y. T. Kim, S. K. Mi n, Applied Physics Letters 62(25), 3312 (1993) [Lee98]. J. Lee, J. Kim, H. Shin, Thin Solid Films, 320, 15 (1998) [Lee99]. Lee, Y. J., B. S. Suh, M. S. Kwon, C. O. Park, Journal of Applied Physics 85(3), 1927 (1999) [Leo06]. W. Y. Leong, C. F. Tsang, H. T. Li V. Biliznetsov, L. Y. Wong, W.H. Li, Thin Solid Films, 496, 402 (2006) [Lim00]. J. Lim, H. Park, S. Ka ng, J of Appl. Phys., 88, 6327 (2000)

PAGE 136

122 [Lim01]. J. Lim, H. Park, S. Kang, J. Electrochem. Soc., 148, C403 (2001) [Lin00]. J. C. Lin, C. S. Liu, S. L. Shue, C. H. Yu, M. S. Liang, IEEE 191 (2000) [Lin01]. S. Lin, T. S. Chin, Proceed ings of IITC, IEEE, Fremont, 146 (2001) [Liu02]. P. Liu, T.C. Chang, K.C. Hsu, T. Y. Tseng, L.M. Chen, C.J. Wang, S.M. Sze, Thin Solid Films, 414, 1 (2002) [Liu95]. H. Liu, D. C. Bertolet, J. W. Rogers Jr., Surf. Sci., 340, 88 (1995) [May91]. T. M. Mayer, J. W. Rogers, T. A. Michalske, Chem. Mater., 3, 641 (1991) [Min00]. J. S. Min, J. S. Park, H. S. Pa k, S. W. Kang, J. Electrochem. Soc., 147(10), 3868 (2000) [Mos01]. K. Mosig, H. T. Li, C. F. Tsa ng, Proceedings of IITC, IEEE, Fremont, 292 (2001) [Mur95] S. P. Murarka, S. W. Hymes, Crit. Rev. Solid State Mat. Sci. 20(2) 87 (1995) [Mus96]. J. N. Musher, R. G. Gordon J. Electrochem. Soc., 143, 736 (1996) [Nic95]. M. A. Nicolet, Appl. Sur. Sci., 91, 269 (1995) [Nis02]. NIST XPS database version 2.0 (2002) [Pan99]. C. A. Panos, The Electroch emical Society Interface, 32 (1999) [Par01]. J. S. Park, M. J. Lee, C. S. Lee, S. W. Kang, Electrochem. Solid State Lett. 4, PC17 (2001) [Par02]. J. Park, H. Park, S. Kang, J. Electrochem. Soc., 149, C28 (2002) [Par96]. K.C. Park, K. B. Kim, J. Appl. Phys., 80(10), 5674 (1996) [Raa93] I. J. Raaijmakers, J. Yang, Applied Surface Science, 73, 31, (1993) [Ram00]. C. E. Ramberg, E. Blanquet, M. P ons, C. Bernard, R. Ma dar, Microelectronic Engineering 50, 357 (2000) [Ras02]. M. Rasco, K. Mosig, J. Ling, P. El enius, R. Augur, Proceedings of IITC, IEEE, Burlingame, 113 (2002) [Rei96]. J. S. Reid, E. Kolawa, C. M. Garland, M. A. Nicolet, J. Appl. Phys., 79(2), 15 (1996) [Rha98]. S. K. Rha, S. Y. Lee, W. J. Lee, Y. S. Hwang, C. O. Park, J. Vac. Sci. Tech. B (16), 2019 (1998)

PAGE 137

123 [Rit99]. M. Ritala, P. Kalsi, D. Riihela, K. Jukli, M. Leskela, J. Jokinen, Chem. Mater., 11, 1712 (1999) [Ruh96]. D. Ruhela, M. Ritala, R. Matero, M. Leskela, J. Jokinen, P. Haassalo, Chem. Vap. Dep., 2(6) 277 (1996) [Rya02]. E. T. Ryan, J. Martin, K. Junker, J. J. Lee, T. Guenther, J. Wetzel, S. Lin, D. W. Gidley, J.Sun, Proceedings of IITC, IEEE, Burlingame, 27 (2002) [Sas90]. K. Saski, A. Noya, T. Umezawa, Jpn. J. Appl. Phys., 1(29), 1043 (1990) [Sat02a]. A. Satta, J. Schuhmacher, C. M. Wh elan, W. Vandervosrst, S. H. Brongersma, G. P. Beyer, K. Maex, A. Vantomme, M. M. Vittanen, J of Appl. Phys., 92, 7641 (2002) [Sat02b]. A. Satta, M. Baklanov, O. Richard, A. Vantomme, H. Bender, T. Conard, K. Maex, W.M. Li, K. E. Elers, S. Ha ukka, Microelectronic Eng., 60, 59 (2002) [Sch01]. T. Scherban, G. Xu, C. Merrill, C. L itteken, B. Sun, Proceedings of IITC, IEEE, Fremont, 257 (2001) [She90]. A. Sherman, J. El ectrochem. Soc., 137, 1892 (1990) [Shy95]. P. M. Shyam S. W. Hymes, Critical Reviews in Solid State and Material Sciences, 20(2), 87 (1995) [Sil06] http://www.siliconfareast. com/metallization.htm Accessed: August 2006 [Str02]. O. V. straten, Y. Zhu, E. Eise nbraun, A. Kaloyeros, IEEE, 0-7803-7216-6/02, 188 (2002) [Str04]. O. V. Straten, Y. Zhu, K. Dunn, E. T. Eisenbraun, A. E. Kaloyeros, J. Mater. Res., 19, 447 (2004) [Sun01]. S. C. Sun, H. Y. Tsai, J. Wang ., Proceedings of AMC2001, Tokyo, 587, (2001) [Sun94]. S. C. Sun, M. H. Tsai, Thin Solid Films, 253, 440 (1994) [Tio02]. Y. Toivola, J. Thurn, R. F. Cook, J. of the Electrochemical Society, 149 (3), F9, (2002) [Tis90]. M. A. Tischler, IBM J. Res. Develop, 34(6), 828 (1990) [Toi02]. Y. Toivola, J. Thurn, R.F. Cook, J. of Electrochemical Soc., 149 (3), F9 (2002) [Tra04]. M. Traving, W. Stei nhgl, G. Schindler, G. Stei nlesberger, M. Engelhardt, Proceedings of AMC2004, Tokyo, 391, (2004) [Tsa02]. S. Tsai, Y. Zhu, Proceedings of IITC, IEEE, Burlingame, 102 (2002)

PAGE 138

124 [Tsa04]. C. F. Tsang, Y. J. Su, V. N. Bliznestov, Thin Solid Films, 462, 269 (2004) [Tsa95a]. M. H. Tsai, S. C. Sun, H. T. Chiu C. E. Tsai, S. H. Chuang, Appl. Phys. Lett., 67 (8), (1995) [Tsa95b]. M. H. Tsai, S. C. Sun, C. P. Lee, H. T. Chiu, C. E. Tsai, S. H. Chuang, S. C. Wu, Thin Solid Films 270(1-2), 531 (1995) [Tsa96a]. M. H. Tsai, S. C. Sun, C. E. Tsai S. H. Chuang, H. T. Chju, J. Appl. Phys., 73 (9) (1996) [Tsa96b]. M. H. Tsai, S. C. Sun, H. T. Ch iu, S. H. Chuang, Appl. Phys. Lett., 68(10), 1412 (1996) [Vie99]. N. H. Viet, K. V. Herma, H. W. Pi erre, Proceedings of the Third International Workshop on Materials Science, Hanoi, (1999) [Wan02]. H. Wang, A. Tiwari, X. Zhang, A. Kvit, J. Narayan, Appl. Phys. Lett.,81(8), 1453 (2002) [Wan03]. Z. Wang, O. Yaegashi, H. Sakaue, T. Takahagi, S. Shingubara, J. Appl. Phys., 94(7), 4697 (2003) [Wan05]. Y.H. Wang, R. Kumar, X. Zhou, J. S. Pan, J. W. chai, Thin Solid Films, 473, 132 (2005) [Wan98]. L. M. Wang, M. Chen, J. Electrochem. Soc., 125(7), 2538 (1998) [Wei96]. B. H. Weiller, J. Am. Chem. Soc., 118, 4975 (1996) [Whi06]. L. M. White, Professor of Chemistry, U of Florida (2006) [Wu04]. Y. Y. Wu, A. Kohn, M. Eize nberg, J of Appl. Phys., 95, 6167 (2004) [Yok91]. N. Yokoyama, K. Hi node, Y. Homma, J. Electro chem. Soc., 138, 190 (1991) [Zha97]. Q. Y. Zhang, X. X. Mei, D. Z. Tang, F. X. Chen, T. C. Ma, Y. M. Wang, F. N. Teng, Nucl. Instrum. Methods. Phys. Res. B, 127/128, 664 (1997)

PAGE 139

125 BIOGRAPHICAL SKETCH KeeChan Kim was born on December 27th, 1973, in Incheon, South Korea. He graduated from Pohang University of Sc ience and Technology (POSTECH) with a bachelors degree in chemical engineering in February 1999. After graduation, he worked as a process engineer at S-Oil Refinery Company in South Korea from 1999 to 2001; his main ro le was to troubleshoot and optimize the xylene process. While he was working, he got married to Kathy Noe on October 3rd in 1999 and had a beautiful daughter named Trudy Kim on December 1st, 2000. He got admission to Chemical Engineering De partment in University of Florida in August 2001 and started his Ph.D. degree under the guidance of Dr. Tim Anderson from the August 2002. His main research area was involved with thin film deposition of Tabased Cu diffusion barriers using metal-organi c Ta precursors by atomic layer deposition and chemical vapor deposition. He graduated fr om University of Florida with a doctorate degree in December 2006.


Permanent Link: http://ufdc.ufl.edu/UFE0016060/00001

Material Information

Title: Chemical Vapor Deposition and Atomic Layer Deposition of Ta-Based Diffusion Barriers Using Tert-Butylimido Tris (Diethylamido) Tantalum Metal Organic Precursor
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0016060:00001

Permanent Link: http://ufdc.ufl.edu/UFE0016060/00001

Material Information

Title: Chemical Vapor Deposition and Atomic Layer Deposition of Ta-Based Diffusion Barriers Using Tert-Butylimido Tris (Diethylamido) Tantalum Metal Organic Precursor
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0016060:00001


This item has the following downloads:


Table of Contents
    Title Page
        Page i
        Page ii
    Dedication
        Page iii
    Acknowledgement
        Page iv
        Page v
    Table of Contents
        Page vi
        Page vii
    List of Tables
        Page viii
    List of Figures
        Page ix
        Page x
        Page xi
        Page xii
    Abstract
        Page xiii
        Page xiv
    Introduction
        Page 1
        Page 2
        Page 3
        Page 4
        Page 5
        Page 6
    Background and literature review
        Page 7
        Page 8
        Page 9
        Page 10
        Page 11
        Page 12
        Page 13
        Page 14
        Page 15
        Page 16
        Page 17
        Page 18
        Page 19
        Page 20
        Page 21
        Page 22
        Page 23
        Page 24
        Page 25
        Page 26
        Page 27
        Page 28
    Experimental setup and reactor simulation using CFD software
        Page 29
        Page 30
        Page 31
        Page 32
        Page 33
        Page 34
        Page 35
        Page 36
        Page 37
        Page 38
        Page 39
        Page 40
        Page 41
        Page 42
        Page 43
        Page 44
        Page 45
    Effect of NH3 addition on TaN MOCVD using TBTDET
        Page 46
        Page 47
        Page 48
        Page 49
        Page 50
        Page 51
        Page 52
        Page 53
        Page 54
        Page 55
        Page 56
        Page 57
        Page 58
        Page 59
        Page 60
        Page 61
        Page 62
    Ultra-thin ALD TaN films using TBTDET and NH3 for Cu barrier applications
        Page 63
        Page 64
        Page 65
        Page 66
        Page 67
        Page 68
        Page 69
        Page 70
        Page 71
        Page 72
        Page 73
        Page 74
        Page 75
        Page 76
        Page 77
        Page 78
        Page 79
    Metal organic atomic layer deposition of Ta-Al-N using TBTDET, TEA and NH3
        Page 80
        Page 81
        Page 82
        Page 83
        Page 84
        Page 85
        Page 86
        Page 87
        Page 88
        Page 89
        Page 90
        Page 91
        Page 92
        Page 93
        Page 94
        Page 95
        Page 96
        Page 97
        Page 98
        Page 99
        Page 100
        Page 101
        Page 102
        Page 103
    Pore sealing treatments of low-k CDO films to prevent Ta precursor penetration during TaN ALD
        Page 104
        Page 105
        Page 106
        Page 107
        Page 108
        Page 109
        Page 110
        Page 111
        Page 112
        Page 113
        Page 114
        Page 115
        Page 116
        Page 117
    References
        Page 118
        Page 119
        Page 120
        Page 121
        Page 122
        Page 123
        Page 124
    Biographical sketch
        Page 125
Full Text












CHEMICAL VAPOR DEPOSITION AND ATOMIC LAYER DEPOSITION OF Ta-
BASED DIFFUSION BARRIERS USING TERT-BUTYLIMIDO
TRIS(DIETHYLAMIDO) TANTALUM METAL ORGANIC PRECURSOR














By

KEECHAN KIM


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

UNIVERSITY OF FLORIDA


2006



























Copyright 2006

By

KeeChan Kim



























This document is dedicated to the graduate students of the University of Florida.
















ACKNOWLEDGMENTS


My thesis is the result of five years' hard work whereby I have been accompanied

and supported by many people. I am very pleased to have the opportunity to express my

gratitude to each of them.

The first person I would like to thank is my direct supervisor, Dr. Tim Anderson. I

have been under his guidance since 2001 when I started my Ph.D. study. During these

years I have known him as an encouraging and inspiring professor. Especially, his way of

supervising students to become independent thinking that allowed me to learn how to

initiate, plan, execute, analyze and summarize the research myself. This experience will

give me a great strength throughout my career. I owe him much gratitude for showing me

this approached research. I could not have accomplished as much without his

unconditional belief and support of me. Besides being an excellent supervisor, he took

care of me in many aspects like my father would have done. I am really glad that I have

come to know him in my life and very proud that he was my supervisor.

I would like to thank the other members of my Ph.D. committee who monitored my

work and provided me with valuable comments during my studies and this dissertation.

Dr. Lisa McElwee White, Dr. Fan Ren, and Dr. Cammy Abernathy, I thank them all.

My colleagues of the ALD project each gave me their full support: DoJun Kim,

OhHyun Kim and Hiral Ajimera, many thanks for your support.









This research has been supported and funded by NSF-CRC grant CHE-0304810. I

am also grateful for the MAIC (Major Analytical Instrument Center) in Materials Science

and Engineering Department for providing an excellent characterization environments

and training. Many thanks go to Eric Lambers, Valentin Craciun, and Kerry Siebein for

their cheerful assistance.

I am very grateful to my wife, Kathy, for her love and patience during my graduate

studies. One of the best experiences during this period was raising our daughter, Trudy

Kim, who provided an additional and joyful dimension to our life mission.

Finally, I would like to thank God, our father, for all of the graces He gave me

while I was studying in Gainesville. The past five years was the most important times in

my life, because He led me to himself and invaluable friends that I could have missed in

other places.
















TABLE OF CONTENTS



A C K N O W L E D G M E N T S ................................................................................................. iv

LIST OF TABLES .......................................... viii

LIST OF FIGURES .................................................. ............................ ix

ABSTRACT ............................................ ............. ................. xiii

CHAPTER

1 IN TR O D U C T IO N ........ .. ......................................... ..........................................1.

S tatem en t o f P rob lem ................. ..................................................... ..................... 1

2 BACKGROUND AND LITERATURE REVIEW ................................................7...

2 .1 C u M etallization ......... ....... .. .. ................................... ... 7
2.2 Superiority of Ta/TaN Bi-layer as a Cu Diffusion Barrier...............................15
2.3 TaN C V D ............................................................................. ......... ...................... 17
2 .4 T aN A L D .............................................................................................................. 2 0
2 .5 T ernary T a N itrides.......................................................................... ................ 23
2.6 Integration of Low-k Materials on Cu Interconnects ......................................25

3 EXPERIMENTAL SETUP AND REACTOR SIMULATION USING CFD
S O F T W A R E .............................................................................................................. 2 9

3.1 R eactor D esign and Setup ......................................................... ......................... 29
3.2 Velocity and Temperature Profile Simulation of Reactor using CFD Software ..38

4 EFFECT OF NH3 ADDITION ON TaN MOCVD USING TBTDET.................... 46

4 .1 In tro d u ctio n ........................................................................................................... 4 6
4.2 Experim ental D details ................... .............................................................. 49
4 .3 R results and D iscu ssion .................................................................... ............... 50
4.3.1 Resistivity and Chemical Composition ................................. ................ 50
4.3.2 Microstructure and Surface Morphology Analysis ...............................52
4.3.3 Deposition Characteristics and Density Analysis..................................54
4.3.4 N ucleation Step .. .. ................. ............................................... 57









4.3.5 D iffusion B carrier Perform ance .................... .................... .....................58
4 .4 C o n c lu sio n s ........................................................................................................... 6 1

5 ULTRA-THIN ALD TaN FILMS USING TBTDET AND NH3 FOR Cu
BARRIER A PPLICA TION S.................................... ....................... ................ 63

5 .1 In tro d u ctio n ...........................................................................................................6 3
5.2 Experim ental D details ................... .............................................................. 66
5.3 R results and D iscu ssion .................................................................... ................ 67
5.3.1 Confirmation of Self-saturation ALD Growth ......................................67
5.3.2 Process Tem perature W indow ..................................................... 70
5.3.3 Growth Characteristic of TaN ALD......................................................72
5.3.4 Diffusion Barrier Test of Ultra-thin TaN .............................................75
5 .4 C o n clu sio n s........................................................................................................... 7 8

6 METAL ORGANIC ATOMIC LAYER DEPOSITION OF Ta-Al-N USING
TB TD E T TE A A N D N H 3 .................................................................... ................ 80

6 .1 In tro d u ctio n ........................................................................................................... 8 0
6.2 Experim ental D details ................... .............................................................. 82
6.3 Results and D discussion ................... ..... ............... 83
6.3.1 Growth Rate and Chemical Composition..............................................83
6.3.2 C hem ical B onding States ...................................................... ................ 93
6.3.3 Comparison of Microstructure and Diffusion Barrier Performance...........98
6 .4 C o n clu sio n s......................................................................................................... 10 2

7 PORE SEALING TREATMENTS OF LOW-K CDO FILMS TO PREVENT Ta
PRECURSOR PENETRATION DURING TaN ALD...... ............ ...................104

7 .1 In tro du ctio n ......................................................................................................... 10 4
7.2 Experim ental D details ................. ............................................................. 107
7.3 R results and D iscu ssion ................................................................ .. ............... 108
7.3.1 Chemical Bonding and Contact Angle of CDO Films.......................... 108
7.3.2 Density and Surface Roughness of CDO Films .......................................111
7.3.3 Penetration of Ta precursor into CDO Films ................ ...................113
7 .4 C o n c lu sio n s ...................................................................... ................................... 1 1 6

LIST O F REFEREN CE S ... ................................................................... ............... 118

BIOGRAPH ICAL SKETCH .................. .............................................................. 125















LIST OF TABLES


Table page

2-1 Interface products between Cu and contact materials......................................... 11

2-2 Evaluation factors for Cu/Cu liners/ILD............................................................. 17

2-3 Resistivity and carbon content of TaN from MO source .................................... 19

2-4 P properties of dielectrics ........................................... ......................... ................ 27

3-1 P properties of nitrogen gas......................................... ........................ ................ 42

3-2 B oundary conditions .. ...................................................................... ................ 44

4-1 Summary of film properties and diffusion barrier test results of TaN deposited
w ith and w without N H 3 ............................................................................. ................ 62

6-1 Estimated thickness of AIN contributing to Ta-AI-NA film thickness as a
function of TEA exposure tim e ........................................................... ................ 87

7-1 Contact angle of as deposited, 02 plasma treated, SiN capped CDO films, and
S iO 2 ........................................................................................................ .............. 1 1 1















LIST OF FIGURES


Figure page

2-1 Interconnect metallization in memory and logic device ......................................8...

2-2 Time delays caused by interconnect and gate at different feature sizes..................9...

2-3 Schematic diagram of single damascene and dual damascene process................. 13

2-4 Schem atic diagram of Cu electroplating ............................................. ................ 14

3-1 Process flow diagram of ALD reactor system ....................................................30

3-2 CAD drawing of reactor inlet part with shower head plate attached ....................31

3-3 C A D draw ing of reactor body ............................................................. ................ 32

3-4 CAD draw ing of reactor dow nstream ................................................. ................ 33

3-5 Schem atic diagram of heater assem bly ............................................... ................ 34

3-6 M anifold structure of pneum atic valves.............................................. ................ 37

3-7 Control panel of ALD growth programmed with LaBView................................38

3-8 Control volume explaining the discretization of a scalar transport equation...........41

3-9 Triangular grid m esh of ALD reactor ................................................. ................ 42

3-10 Contour of velocity magnitude (m/s) and velocity vector colored by velocity
m agnitude around heater area (m /s) .................................................... ................ 44

3-11 Color filled contour of static temperature (K) and contour line of static
tem perature (K ) around heater area..................................................... ................ 45

4-1 Chem ical structure of TB TD ET .......................................................... ................ 48

4-2 Resistivity, and nitrogen, carbon and oxygen content relative to Ta in TaN films
as a function of am m onia flow rate..................................................... ............... 50

4-3 Proposed transamination reaction mechanism between TBTDET and NH3 ............51









4-4 XRD patterns of TaN films deposited at 300 C as a function of ammonia flow
rate ......................................................................................................... ........ .. 5 3

4-5 Surface morphology of TaN films deposited without NH3 at 300 C and with
N H 3 (30 sccm ) .......................................................................... .. . .. ............... 53

4-6 Growth rate (A/min) vs. reciprocal growth temperature with different NH3 flow
rates of 0 to 150 sccm ........................................................................ 55

4-7 Cross-sectional SEM images of TaN films deposited with various NH3 flow
rates at 4 00 C ........................................................................... . .. ............... 56

4-8 XRR profiles of TaN films deposited at 400 C with different NH3 flow rates....... 57

4-9 AFM scans of the initial stages of deposition from TBTDET only and TBTDET
w ith N H 3 = 50 sccm at 350 C ............................................................ ................ 58

4-10 XRD spectra of Cu/TaN/Si structures annealed at 500 C for one hr for films
grow n w ith different N H 3 flow rates .................................................. ................ 59

4-11 SEM images of Si surface after annealing the structure Cu/TaN/Si at 500 C for
1 hr, followed by the removal of Cu and TaN layers, and then etch in Secco
solution to reveal etch pits if Cu penetration occurred........................ ................ 60

5-1 XRR-profiles of ALD-TaN as a function of TBTDET exposure time for films
deposited at 300 C for 80 cycles, and 10 sec purges and NH3 exposure .............68

5-2 Growth rate of ALD-TaN as a function of TBTDET exposure time deposited at
300 C, and 10 sec purges and NH3 exposure ..................................... ................ 69

5-3 Growth rate of ALD-TaN as a function of TBTDET exposure time deposited at
250 C, and 10 sec purges and NH3 exposure ..................................... ................ 69

5-4 XRR profiles of ALD-TaN as a function of growth temperature deposited with 9
sec exposure of TBTDET and 80 cycles, and 10 sec purges and NH3 exposure..... 71

5-5 Thickness of ALD-TaN films as a function of growth temperature deposited
with 9 sec exposure of TBTDET and 80 cycles, and 10 sec purges and NH3
e x p o su re ................................................................................................................... 7 2

5-6 XRR profiles of ALD-TaN as a function of cycle number deposited at 9 sec
TBTDET exposure, 10 sec purges and NH3 exposure at 300 C ..........................74

5-7 Thickness of ALD-TaN as a function of cycle numbers deposited at 9 sec
TBTDET exposure, 10 sec purges and NH3 exposure at 300 C ..........................74

5-8 AFM scans (z: 3.0 nm / div) of the surface of TaN deposited at 9 sec TBTDET
exposure and 300 C using 5, 10 and 15 cycles................................... ................ 75









5-9 XRD patterns of Cu/TaN/Si annealed at 500 C for 30 min ...............................77

5-10 Cross-sectional TEM micrograph of Cu/TaN/Si structure. TaN grown by ALD
for 15 cycles at 9 sec TBTDET exposure time and 300 C.................................77

5-11 SEM images of Si surface after annealing Cu/TaN/Si at 500 C for 30 min
followed by the removal of Cu and TaN, and Secco etching...............................78

6-1 XRR profiles of ALD AIN as a function of TEA exposure time deposited at 300
C and 10 sec purges and NH3 exposure for 120 cycles .....................................85

6-2 Growth rate of AIN as a function of TEA exposure time deposited at 300 C and
10 sec purges and NH3 exposure for 120 cycles ................................. ................ 86

6-3 XRR profiles of Ta-Al-N grown with Ta-Al-N sequence (A) as a function of
TEA exposure time grown at 300 C and 10 sec purges and NH3 exposure for
12 0 cy c le s ............................................................................................................ .. 8 8

6-4 Growth rate and chemical composition of Ta-Al-N grown with Ta-Al-N
sequence (A) as a function of TEA exposure time grown at 300 C and 10 sec
purges and NH 3 exposure for 120 cycles ................................................... 88

6-5 XRR profiles of Ta-Al-N grown with Al-Ta-N sequence (B) as a function of
TEA exposure time grown at 300 C and 10 sec purges and NH3 exposure for
12 0 cy c le s ............................................................................................................ .. 9 0

6-6 Growth rate and chemical composition of Ta-Al-N grown with Al-Ta-N
sequence (B) as a function of TEA exposure time grown at 300 C and 10 sec
purges and NH 3 exposure for 120 cycles ............................................ ................ 91

6-7 XRR profiles of Ta-Al-N as a function of TEA exposure time grown with Ta-N-
Al-N sequence (C) at 300 C with 10 sec purges and NH3 exposure for 120
cy cle s ...................................................................................................... ........ .. 9 2

6-8 Growth rate and chemical composition of Ta-Al-N grown with Ta-N-Al-N
sequence (C) as a function of TEA exposure time grown at 300 C and 10 sec
purges and NH3 exposure for 120 cycles ................................................... 93

6-9 XPS spectra of Ta 4f core level for Ta-Al-N deposited with sequences A, B, and
C and TaN at 300 C and 120 cycles after 1 min sputtering................................95

6-10 XPS spectra ofN Is core level for Ta-Al-N deposited with sequence A, B, and
C and TaN at 300 C and 120 cycles after 1 min sputtering................................97

6-11 XPS spectra of Al 2p core level for Ta-Al-N deposited with sequence A, B, and
C at 300 C and 120 cycles after 1 min sputtering.............................. ................ 98









6-12 XRD patterns of Ta-Al-N grown with sequence A as a function of TEA
exposure tim e at 300 C and 120 cycles ............................................. ................ 99

6-13 Cross-sectional TEM micrographs of TaN and Ta-AI-NA deposited at 300 C
w ith SA D patterns .............. ................... ................................................ 100

6-14 XRR profiles of TaN, Ta-AI-NA and AIN deposited at 300 C for 120 cycles......101

6-15 XRD patterns of Cu/TaN and Cu/Ta-Al-N/Si structures annealed at 500 C for
tim e a s n o ted ........................................................................................................... 10 2

7-1 FTIR spectrum of as deposited CDO film on Si ......................... ..................... 110

7-2 XPS spectra of as received, 02 plasma treated, and SiN capped CDO films.........110

7-3 XRR-profiles of as deposited, 02 plasma treated, and SiN capped (25 and 50
nm ) C D O fi lm s ............. .. .................... ................ ............ .... .. ....... .... ....... .. 112

7-4 Surface roughness of (a) as received, (b) 02 plasma treated, and (c) SiN capped
(25 nm) CDO films z: 10 nm div ..........................1.............................................................13

7-5 Cross sectional TEM images of (a) TaN/as-deposited CDO/Si, (b) TaN/02
plasma treated CDO/Si, and (c) TaN/SiN capping layer/CDO/Si structure
superim posed w ith line ED X Ta signal ................... .................... ..................... 116











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

CHEMICAL VAPOR DEPOSITION AND ATOMIC LAYER DEPOSITION of Ta
BASED DIFFUSION BARRIERS USING TERT-BUTYLIMIDO
TRIS(DIETHYLAMIDO) TANTALUM METAL ORGANIC PRECURSOR

By

KeeChan Kim

December 2006

Chair: Timothy J. Anderson
Major Department: Chemical Engineering

Ta-N and Ta-Al-N Cu diffusion barriers were deposited by chemical vapor

deposition (CVD) and atomic layer deposition (ALD) using tert-butylimido

tris(diethylamido) tantalum (TBTDET)/tri-ethyl aluminum (TEA) metal organic

precursors. The effect of NH3 addition on film properties during TaN CVD from

TBTDET was examined. As the NH3 flow was increased at constant TBTDET flow, the

film density, nitrogen content, and grain size increased, while resistivity and carbon

content decreased as compared to films deposited with TBTDET alone. These property

changes are attributed, in part, to transamination reaction between the diethylamido

ligands in TBTDET and NH3. The higher film density and nitrogen content produced

TaN films that exhibited superior diffusion barrier performance compared to those

deposited without NH3 addition.

TaN was also successfully deposited by ALD with alternating exposure to

TBTDET and NH3. TBTDET adsorption was shown to be self-limiting with a single

monolayer growth rate of 2.6 A/cycle over the process temperature window of 200 to 300









C. An incubation period exists during the initial cycles as evidenced by a non-linear

relationship between film thickness and cycle number. Ultra-thin ALD-TaN layers, as

thin as 38 A, effectively blocked Cu diffusion during a 30 min anneal at 500 C.

Ternary Ta-Al-N films were deposited from TBTDET and TEA to promote

formation of an amorphous film and increasing the recrystallization temperature. The Al

mole fraction was linearly dependent on the TEA exposure time suggesting growth was

self-limiting. Although Al insertion into TaN promoted an amorphous structure, it also

lowered the overall film density. A comparative study of the diffusion barrier

performance showed that failure occurred for both TaN and Ta-Al-N films at the same

thickness, suggesting the increased amorphous content by adding Al was offset by the

lower film density. Selecting a different reactant exposure sequence produced different

film properties. A higher oxygen content was observed along with lower growth rate

when the Al--Ta--N sequence was employed, compared to films deposited with the

Ta--Al--N sequence.

The compatibility of ALD TaN barriers on porous low-K CDO films was

investigated. Two different surface treatments (i.e., 02 plasma exposure and SiN capping)

were applied to the surface of a low-k film before TaN deposition to prevent Ta

penetration into the pores of the low-K films. Ta precursor infiltration was detected into

the dielectric film of the untreated surface, while the 02 plasma treated and SiN capped

low-K films showed no diffusion of Ta precursor due to the densification of the near

surface region of treated low-K films.














CHAPTER 1
INTRODUCTION

Statement of Problem

As the feature size of integrated circuits shrinks to nano dimensions, the total

circuit delay is more dominated by the interconnect RC time delay than the intrinsic

device delay [Vie99]. For this reason, copper has been replacing the aluminum alloy as

the interconnect metal. Copper has a lower resistivity and higher electromigration

resistance, resulting in faster device switching speeds. Copper, however, is known as a

very fast diffuser in Si and SiO2, forming deep traps and copper silicide compounds

[Bra03], which can cause serious problems such as an increase in contact resistance,

change in barrier height and leaky p-n junction in device. Therefore Cu diffusion barriers,

which prevent Cu diffusion into Si or SiO2, are needed between Cu and Si or low K

dielectric. Various transition metals (Ta, Ti, W) and metal nitrides have been investigated

as candidate Cu diffusion barriers for last decades [Bec03, Cho99, Dub94, Lee98,

Mus96, Par96, Raa93, Sun94, Tsa95a, Tsa95b]. Extensive research on these materials

demonstrated that a bilayer of Ta and its nitride is well suited as a liner for Cu

damascene. This configuration gives the highest Cu reliability, optimized adhesion for Cu

electromigration resistance, robustness of via interfaces and redundant current strapping

for added chip reliability [Ede02].

Most Ta/TaN layers employed so far in industry have been deposited by Physical

Vapor Deposition (PVD). Because of its inherent shadowing effect, this approach has a

serious limitation in achieving conformal coverage in devices with submicron features









and high aspect ratio contacts and via holes. Therefore, Chemical Vapor Deposition

(CVD) has received recent attention owing to its superior conformality over PVD grown

barriers.

There are two different approaches to TaN-CVD depending on the metal source

used. TaN-CVD using halide sources such as TaCl5 [Hie74] and TaBr5 [Che97] is known

to require high deposition temperature to obtain low film resistivity and halide

incorporation, which is inappropriate for Back End of Line (BEOL) processing. Cl or Br

incorporation in the film is an issue because the halides in the film reduce the adhesion

strength and enhance the electromigration, which can be major concerns for long term

device reliability [She90, Yok91]. The other approach for TaN-CVD is using Metal

Organic (MO) sources. This method can significantly lower the deposition temperature

with good conformality as compared to halide source. Films deposited with MO source

alone, however, tend to show high carbon content and accordingly higher film resistivity.

In this research, (NEt2)3Ta=NBut [tert-butylimido tris(diethylamido) tantalum,

TBTDET] is investigated as a single source precursor as well as paired with ammonia for

TaN-CVD. Previous studies on TaN-CVD using TBTDET [Tsa95a, Tsa95b, Tsa96a]

showed the results on TaN film properties deposited with TBTDET as a single source

through the thermal dissociation reaction. Since TBTDET has a N/Ta ratio of four with

the Ta bonded to four N atoms, TaN films can be grown with this single precursor by

CVD. The films deposited with TBTDET in a carrier gas, however, possess the large

amount of carbon impurity reaching up to 30 atom %, ascribed to the diethyl amido or t-

butyl imido ligands in TBTDET. Although carbon impurity helps to increase amorphous

structure, it causes the film resistivity to increase and thus making an unsuitable barrier.









In this program, NH3 was tested as an additional nitrogen source in the deposition

chemistry. This hypothesized that NH3 will undergo a transamination reaction with the

carbon containing nitrogen ligands (i.e., diethyl amido and t-butyl imido ligands). This

transamination reaction should lead to lower carbon impurity and thus lower the film

resistivity compared to using TBTDET alone. Chapter 4 presents the results of

experiments on the effect of adding NH3 on the TaN film properties by measuring

resistivity, chemical composition, microstructure, surface morphology, and film density

as a function of ammonia flow. Difference in deposition rate with growth temperature

was identified for TaN films deposited with TBTDET as a single source and TBTDET

with added NH3. The Cu diffusion barrier performance was tested too.

Although the ITRS road map calls for CVD of barriers in the near future, atomic

layer deposition (ALD) will likely emerge in the longer term as the dominant growth

method in Cu diffusion barrier deposition because of its superior conformality and

accurate thickness control compared to other deposition methods. As the barrier

thickness drops below 100 A, standard CVD will approach its usability limit due to

difficulties in controlling the deposition rate. The highly conformal, ultra thin barriers

afforded by ALD will be essential to minimize the barrier's impact on the resistance per

unit length in Cu interconnects [Kap02].

Early reports on TaN-ALD used a halide source (TaCl5 [Hil88a, Rit99] or TaBr5

[AleO1, Ale02]) with NH3, produced polycrystalline Ta3N5 films, which made poor Cu

diffusion barriers because of high grain boundary structures. High film resistivity and

high Cl or Br residue deposited at low temperature are other issues for halide-based TaN

ALD. Therefore, MO source based TaN ALD is being considered as more suitable than









halide-based TaN ALD. Although there have been many reports of MO based TaN ALD,

their main focus has been verification of self-limiting growth conditions, i.e. the region of

operation in which the precursor adsorption is restricted to monolayer or fractional

monolayer coverage to give a growth rate that is constant with respect to exposure time

of the reactants. The process temperature window of ALD behavior, in which the growth

rate remains constant in range of deposition temperature, has rarely been reported for

MO-based TaN ALD, while the process temperature window for halide-based TaN has

been reported [Rit99]. In addition, most Cu diffusion barrier performance test using

ALD-TaN have been evaluated with thick films (> 10 nm), although ITRS roadmap

clearly indicates that ultra-thin barriers will be needed in the near future.

In the TaN-ALD experiments reported here, TaN was deposited with the

TBTDET/NH3 sequences. Unlike previous reports of TaN-ALD, XRR (X-ray

Reflectivity) technique was used to accurately measure the TaN film thickness as a

function of TBTDET exposure time to locate the self-limiting growth zone. The process

window temperature was determined for this precursor combination for the first time by

measuring the growth rate as a function of growth temperature in the range of conditions

that give self-limiting growth. Additionally, the growth characteristic of the initial stage

of TaN-ALD was observed using XRR and AFM, focusing on the initial complete

coverage of TaN on Si. As previously mentioned, minimum thickness of ALD-TaN film

that provides a sufficient barrier for Cu diffusion has been rarely reported. The effective

minimum thickness of ALD-TaN barrier was verified by performing Cu diffusion barrier

tests on various ultra-thin (7 to 100 A) ALD-TaN films and the results are discussed in

depth in Chapter 5.









Main drawback of binary nitrides as Cu diffusion barrier materials, especially with

TiN, is its tendency to form polycrystalline with columnar microstructure or recrystallize,

leading to significant Cu grain boundary diffusion [Nic95]. The grain structure creates

boundaries that cut across the film and offer fast diffusion paths for Cu. Adding a third

element to a binary transition metal nitride matrix tends to disrupt the microstructure,

increasing the chance of amorphous phase formation [IstOO, RamOO], which is a superior

structure for Cu diffusion barriers.

In chapter 5, as-deposited binary TaN films deposited by ALD contain nanocrystal

or TaN with a small distribution of h-TaN phase imbedded in an amorphous matrix. To

promote amorphous material, triethyl aluminum (TEA) was inserted into TaN binary film

as a third element to deposit Ta-Al-N ternary films by ALD. Chapter 6 presents a

comparison of the Cu diffusion barrier performance using either TaN or Ta-Al-N films

deposited with the same thickness. The microstructure and film density change caused by

Al insertion is also examined. In addition, it is well known that different sequence

exposure of precursors induces the change in film properties in ternary ALD films. Three

different exposure sequences (Ta-Al--N, Al-Ta-N and Ta--N-Al-N ) were

tested and film thickness, chemical composition, and bonding states are compared.

While TaN and Ta-Al-N deposition was on Si substrates in most metallization

schemes incorporate chapter 4, 5 and 6, low-k Carbon Doped Oxide (CDO) films were

employed as a substrate for ALD-TaN in Chapter 7. As mentioned previously, Cu

metallization was first introduced to diminish the resistance of metal interconnects, and

then low-k dielectrics were developed to reduce the parasitic capacitance, further

decreasing the RC time delay. One of the most important changes in sub-65 nm node and









beyond is the introduction of porous dielectrics for the purpose of reducing the dielectric

constant. Because of the porosity, however, there exists a higher probability of

contamination trapping or low-k material degradation during diffusion barrier deposition.

Chapter 7 demonstrates the results of two different surface treatments applied to the

surface of CDO films before TaN barrier deposition. These treatments are intended to

prevent Ta precursor penetration into the porous structure of COD films. The first

treatment was 02 plasma exposure, which is expected to densify the top layers by sealing

the pores on the surface of CDO films. The second treatment was SiN capping, where

SiN capping layer was deposited on CDO films, which is intended to prevent the

precursor infiltration by closing the pores near the surface of the CDO. The contact angle

(CA), surface roughness, chemical bonding states, and film density of CDO films after

the surface treatments were investigated using CA meter, AFM, XPS, and XRR,

respectively. The efficiency in blocking Ta precursor diffusion was examined with TEM-

EDX after TaN-ALD on surface treated CDO films.














CHAPTER 2
BACKGROUND AND LITERATURE REVIEW

2.1 Cu Metallization

Metallization is the key process where the various device structures fabricated on

the silicon substrate are electrically interconnected through the metal and metal alloy

layers [Sil06]. Numerous regions of each circuit element such as MOSFET (Metal-

Oxide-Semiconductor Field Effect Transistor), Bipolar Transistor and Capacitor are

properly interconnected during the metallization step to distribute clock and other signals

and to provide power/ground, to and among, the various circuits/systems functions on a

chip. Metallization processing represents more than 70 % of the IC (Integrated Circuits)

fabrication steps [Bra03].

Figure 2-1 (a) shows the cross sectional diagram of DRAM (Dynamic Random

Access Memory) and logic devices, which demonstrates that the interconnect

metallization occupies a large cross-sectional area for both devices [Bra03]. As shown in

Figure 2-1 (b) a diagram of a CMOS structure, metals are employed in many locations of

the CMOS device including gate metal, contacts, interconnect layers in multi-level

metallization, and plugs in via holes. Generally, poly-Si, which has a high melting

temperature (1414 C) and resistance to oxidation, has been used as a gate metal.

Tungsten (W), however, has been mainly used as a contact and via hole plug metal while

the aluminum (Al) alloy with Si has been used as the interconnect metal. Recently, Al

metal and Si02 dielectric combination for interconnect metallization is being rapidly

replaced by the Cu, low-k configuration due to the decrease in the RC time delay.














terconnect







Capacitor



ransistor/
Bit line

DRAM


Interconnect









} Transistor


Logic device


-111 i --

.STI


Plate Poih

Bit Line --
Gate -


(b)
Figure 2-1. Interconnect metallization in memory and logic device (a) 3D diagram of
metals employed in CMOS structure (b) (taken from [Bra03])

As complexity and packing density increase in IC's, the size of interconnect

metallization decreases, which leads to increase of interconnection delay. The total circuit

delay, which is a combination of the intrinsic device delay associated with the solid state

device and the interconnection RC (resistance capacitance) time delay, is dominated by


Int


I








the interconnect RC time delay at lower feature size. This trend is illustrated in figure 2-

2, where both the interconnection delay and intrinsic delay are plotted as a function of the

feature size (channel length). It is clear that below -0.5 pm feature, the interconnection

delay will be the main limiter for IC's performance [Vie99].


0

ci)
0



E
F--
C>
_r)
2
QV


2.5

2.0

1.5

1.0


0.5


0.0


0.0 0.5


1.0 1.5 2.0

Gate length (pmn)


Figure 2-2. Time delays caused by interconnect and gate at different feature sizes (taken
from [Vie99])
The RC time delay can be represented with the simplistic model given below

[BenOO].

pL L Lw psL2 RsL2
R = =R, CL = R RC=
tw w d td d

where, t is the thickness of metal, w is the width of metal, L is the length of

metal, d is the distance between metal layers, s is the dielectric constant of dielectrics,

R, is the sheet resistance of metal, and p is the specific resistivity.


-*- Interconnect RC time constant L=5000pm
-0- Typical gate delay
--- J^---




^"^*-


2.5


3.0





I









As the model suggests, the metal interconnect with low resistivity and dielectric

materials of low dielectric constant are needed to minimize the RC time delay leading to

higher device switching speed, since copper possesses lower film resistivity compared to

aluminum alloys (1.67 [tQ-cm for Cu vs. 2.65 [tQ-cm for Al) and low-k materials have

lower capacitance than conventional SiO2. Therefore, industry is rapidly adopting copper

as the interconnect metal for the intermediate and upper wiring levels on IC devices

replacing Al alloys, and low dielectric constant materials are being intensely researched

for replacement of the conventional SiO2. Transition from Al alloys to Cu is reported to

increase overall microprocessor speed by 15%. Currently, the development and

qualification of a 300 nm SiCOH BEOL (Back End of Line) integration for 65 nm bulk

and SOI (Silicon on Insulator) semiconductor product application have been reported by

IBM [Ang05]. Intel started shipping the 65 nm technology mode Pentium 4 processors

fabricated with Cu and low-k metallization scheme [Int06].

Copper also shows the higher electromigration resistance than aluminum, which

leads to increase in device lifetime [Mur95]. Aluminum suffers from serious

electromigration at high current flow density, where electrons flowing through the metal

interconnect pass on enough momentum to carry the metal atoms with them, leading to

voids (openings) and hillocks (pileups) in the interconnect wiring. In addition to

resistivity and electromigration improvements, copper is reported to provide higher IC

production yield than aluminum-based devices with similar design. Other advantages of

switching to copper interconnects include a decrease in the number of interconnect levels,

a roughly 30 % decrease in power consumption for operation at a given frequency, and a









cost savings of roughly 30% per interconnect level due to integration of dual damascene

processing [Bch04].

Even with the above mentioned advantages of copper as interconnect metallization,

several problems are encountered with copper metallization. It is well known that Cu is a

very rapid diffuser in Si and SiO2. Diffusion of Cu into Si forms Cu3Si at temperature

below 200 C, leading to formation of deep donor level at the interface. Table 2-1 shows

the compound formed at the interface between Cu and the various semiconductor

materials [Bra03]. Cu diffusion in devices can cause an increase in contact resistance,

change in barrier height, leaky p-n junction and destruction of electrical connections to

the chip. For long-term reliability of interconnect metallization, little or no transport of

Cu through adjacent layers should be required. In addition, copper films show poorer

adhesion to Si and SiO2 compared to aluminum. Thus, diffusion barriers, which prevent

Cu diffusion and promote the adhesion of Cu to Si and SiO2, should be employed

between Si or SiO2 and Cu.

Table 2-1. Interface products between Cu and contact materials (taken from [Bra03])
Contact material Annealing temperature Interface products

Si < 200 C Cu3Si (Deep donor level)

Silicide 350 450 C Punch-through

Al 150 ~ 250 C CuAl2 (Resistivity hike)

SiO2 Bias thermal stress Punch-through

Polymer Room temperature Cu precipitate

Oxygen 100 C Cu20 (Porous oxide)









Although the benefits of employing copper as the interconnect metals are obvious,

there are difficulties in patterning of copper. Wet etching is not appropriate for patterning

sub-micron structures due to its isotropic nature. Reactive ion etching (RIE) of copper is

also not practical because etch byproducts of Cu are not volatile. Therefore, instead of

wet etching and RIE, Cu patterning is based on the so-called Damascene process [Pan99].

Figure 2-3 (a) shows a schematic diagram of a single damascene process where grooves

are first etched in a dielectric layer, barrier/metal deposited on it, and the metal layer is

chemically-mechanically polished, leaving inlaid metal lines in the oxide grooves. In

contrast to the conventional patterning method, where a metal layer is first deposited and

then unwanted metal is etched away, leaving the desired pattern of wires or vias,

damascene patterning involves the same number of steps, but in reverse order. In the

single damascene process, the separate formation of wires and vias has as much

complexity as the conventional RIE patterning process. Dual-damascene, which is

schematically described in figure 2-3 (b), makes it possible to form both wires and vias in

the same metal deposition process step. In this process, the pattern of vias and wires is

defined using two lithography and RIE steps, but the via plugs are filled in the same step

with metal lines. Dual-damascene eliminates the complexity of the patterning process by

reducing the number of processing steps. One major benefit of dual-damascene is less

risk of contact failure between via and metal line.






13







Up'ler MeOiI
Substrate w~


Via patterning Via damascene
-.Via Photo.& Etch -.BM/Cu dep.
-.Via CMP


Dlie ITCIF. IC j'

Unrder-MetWI


Metal patterning Metal damascene
-. Metal Photo.& -.BM/Cu dep.
Etch. -.Metal CMP

(a)


CMP





V V


ElWctircoating-Cu;


Substrate


Damascene patterning
-.Via Photo.& Etch
-.Metal trench Photo.& Etch


Metal dep. & anneal
-.BM/seed-Cu dep.
-.EP-Cu dep. & anneal


(b)
Figure 2-3 Schematic diagram of single damascene (a) and dual damascene process (b)
(taken from [Bra03])

As the multi-level metallization scheme gets more complicated, filling the via/plug

of high aspect ratio with Cu without void or seam features becomes more critical. With

this need, electroplating has been mainly employed as the deposition method of Cu in

industry. Electroplating of Cu is performed by immersing a conductive surface in a

solution containing ions of the Cu to be deposited. The surface is electrically connected

to an external power supply, and current is passed through the surface into the solution.


C7CMP
stF2- BM CMP










This causes reaction of the metal ions (Cu2+) with electrons (e-) to from metal (Cu):

[Shy99]

Cu2+ + 2e- = Cu

Figure.2-4 shows a schematic of Cu electroplating process. The wafer is typically

coated with a thin conductive layer of copper (physically vapor deposited (PVD) Cu seed

layer) to provide a conductive electrode for the electroplating process. This is then

immersed in a solution containing cupric ions. Electrical contact is made to the seed

layer, and current is passed such that the reaction Cu2+ + 2e- = Cu occurs at the wafer

surface. The wafer, electrically connected so that Cu2+ is reduced to Cu, is referred to as

the cathode. Another electrically active surface, known as the anode, is present in the

conductive solution to complete the electrical circuit. At the anode, an oxidation reaction

occurs that balances the current flow at the cathode, thus maintaining electrical neutrality

in the solution. All cupric ions removed from solution at the wafer cathode are replaced

by dissolution from a solid copper anode.



Electrolytic Copper Deposition


Current Path -
e a'


Anode Cathode
(Oxidation) (Reduction)


Figure 2-4. Schematic diagram of Cu electroplating (taken from [Bra03])









2.2 Superiority of Ta/TaN Bi-layer as a Cu Diffusion Barrier

For most metals to be used as interconnect materials, a thin diffusion barrier should

be used between the metals and dielectrics, since many metals form deep levels in Si and

some including Cu have high diffusivities. An additional function of the diffusion barrier

is to serve as an adhesion promoter for Cu and the interlayer dielectric. If the diffusion

barrier materials do not adhere to the underlying layer or interconnect material, or if Cu

does not adhere to the diffusion barrier, the result will be gross delamination and failure,

which typically occurs during CMP steps used to planarize the deposit following plating.

More importantly, poor adhesion causes severe reliability problem such as

electromigration. The second important quality of a Cu diffusion barrier is to have a high

in-plane electrical conductivity (low resistivity), to support as high as current density as

possible without producing excessive Joule heating, which can cause reactions that gives

the device failure. This can add redundant reliability beyond the nominal electromigration

limits and help save chips from open-circuit failures in the event of defects or abnormal

wear out [Ede02]. Other barrier requirements for Cu diffusion include prevention of Cu

diffusion; low to moderate growth temperature range (350 to 400 C) for the

compatibility with low-k material; excellent step coverage of high aspect ratio structures

and extremely thin thickness (<5 nm) [Kim05].

The nitrides and silicon nitrides of Ta, Ti and W are known to be good candidates

for reliable diffusion barriers with respect to interdiffusion requirement and provide

lower electrical resistivity compared to their pure metal counterparts. Table 2-2 shows the

evaluation factors for Cu diffusion barriers for Cu Damascene liner application [Ede02].

In particular, two requirements of a diffusion barrier, namely adhesion and resistivity,









eliminate many candidate materials. Cu poisoning and barrier further eliminate all but

TiN/Ta and TaN/Ta bilayers. TiN/Ta requires two chambers, and TiN was found to cause

Cu corrosion in some CMP processes [Ede02]. In addition, TiN was unsuccessful for Cu

metallization due to grain boundary formation in the film leading to severe interdiffusion

of Cu through the boundary. Plasma treatment and exposure to SiH4 have been chosen to

enhance the barrier quality, but couldn't fully resolve the issue. [Ede02] Clearly, TaN/Ta

is the best choice among the candidates listed, and the only one liner material to meet all

criteria.

Therefore, the TaN/Ta bilayer is currently being employed for current Cu

interconnect metallization in industry. The bi-layer structure is employed to optimize the

adhesion properties. For dual-Damascene integration in Si02, Ta lacks adequate adhesion

to Si02, whereas TaN/Si02 adhesion is excellent. On the other hand, Cu/TaN adhesion is

relatively poor. The liner/ILD (Inter Layer Dielectric) and Cu/liner adhesion have

conflicting dependency on the N content of TaNx. Significantly, the TaN/Ta liner has

very low in-plane resistivity, because when Ta is deposited on a TaN surface, the low-

resistivity a-phase Ta is spontaneously formed with a resistivity in the range of 15 to 25

[tQ-cm. Another benefit of TaN/Ta liner selection is better step coverage than other

lower mass metals such as Ti, even for uncollimated PVD Ta and TaN. Due to the high

mass of Ta, it sputters more directionally from the target, and its high momentum gives it

more surface mobility to redistribute into the features. For further improvements in step

coverage, an ionized PVD (I-PVD) process has been developed for bilayer liner [Ede02].

The TaN/Ta liner material scheme has demonstrated excellent Cu integration

performance in high volume manufacturing, including: high yield, high step coverage,









mechanical, thermal and chemical stability, consistently low resistivity, and high

performance CMP. This bilayer liner process satisfies high-volume manufacturing

criteria, yield, control, low cost and low maintenance [Ede02].

Table 2-2. Evaluation factors for Cu/Cu liners/ILD (taken from [Ede02])
Attribute Cr TiN TiN/Ti Ti/TiN TiN/Ta TaN Ta TaN/Ta TiSiN WN
Cu barrier X 0 0 0 0 0 0 0 0 0
Adhesion to ILD 0 0 0 0 0 O X 0 0 0
Cu on liner adh. 0 X O X 0 X 0 0 X ?
Liner on Cu adh. 0 ? ? 0 ? 0 O O O ?
Low in-plane R 0 X 0 O ? X X O X ?
Cu poisoning O O X 0 0 0 0 O ? 0
CMP ? X X X X/O 0 0 0 O ?
Single chamber 0 0 0 0 X O O O 0 0
Via, Contact R O O X ? O 0 0 0 0 ?/0
Contact R 0 0 0 0 0 0 0 0 O ?
Cu corrosion ? X X X X/O 0 0 0 O ?
Thermal stability ? O X X 0 0 0 O ? 0
Stress, cracking X O O 0 0 0 0 0 0 0
Step coverage ? ** ** *
Final X X X X X X X 0 X ?


*< = CVD available *


= Ionized-PVD available ?


= Not evaluated


2.3 TaN CVD

As previously mentioned, TaN/Ta bilayer is the most prominent liner for Cu

metallization in industry. It is most often deposited by Physical Vapor Deposition (PVD).

PVD, however, has a serious limitation, i.e., poor conformality in small feature size (<

0.35 nm) and high aspect ratio features, projected to reach its usability limit at the 45 nm

node in 2007 [Han03]. The poor conformality inherent to all PVD processes is caused by

the directionality imparted to the atoms/clusters traveling toward the substrate [Lee93].









This directionality leads to step shadowing, where parts of the substrate surface are not

seen by the incoming sputter atoms. Step shadowing results in little or no film coverage

on certain sections of the substrate. These exposed substrate areas are then vulnerable to

reaction with subsequently deposited Cu atoms. Therefore, CVD of barrier has received

attention because of its superior conformality on high aspect ratio structures over sputter-

deposited barriers.

There are two different approaches to TaN-CVD process depending on the metal

sources employed. The first approach uses the halide sources such as TaCl5 [Hie74] and

TaBr5 [Che97]. In general, TaN-CVD from a metal halide source is known to require

high deposition temperature. For example, TaN using TaCl5, N2 and H2 is grown at a

temperature > 900 C to obtain the low resistivity and halide content [Hie74]. This

temperature, however, is inappropriate for the subsequent IC processes. Although lower

growth temperature (450 C) for TaN-CVD from TaBr5, NH3 and H2 has been reported,

but still not low enough for a low-k material process [Che97]. Additionally, particle

formation and Cl or Br content (0.5 to 4.5 atom %) in the film are serious issues for

halide-based TaN-CVD. Especially, the halide content in the film reduces the adhesion

strength and causes the delamination of the film, which can be a severe problem for long

term device reliability [She90, Yok91].

The other approach for TaN-CVD is to employ Metal Organic (MO) sources such

as Ta(NMe2)5 [pentakis(dimethylamido) tantalum, PDMAT] [Fix93], Ta(NEt2)5

[pentakis(diethylamido) tantalum, PDEAT] [Cho98, Cho99] and (NEt2)3Ta=NBut [tert-

butylimido tris(diethylamido) tantalum, TBTDET] [Tsa95a, Tsa96b]. Although the halide









residue in TaN is not an issue for MO source-based TaN CVD, it tends to produce high

carbon concentration, leading to high film resistivity as a result (Table 2-3).

Table 2-3. Resistivity and carbon content of TaN from MO source
MO source for Resistivity Carbon content Growth T NH3 as an additional
TaN CVD ([t-cm) (atom %) (C) N source

PDMAT > 106 20 200 to 400 10 % NH3 in He
1.2 to 6.0 X
PDEAT 10 30 to 1 300 to 375 0 to 25 sccm
105
TBTDET 1.0 X 105 23 500 No NH3



TaN from PDEAT had high film resistivity (up to 60000 [M-cm) with high carbon

content (~ 30 atom %) when it was deposited from PDEAT single source. The addition of

NH3 as an additional nitrogen source, however, caused the decrease in resistivity down to

7000 utQ-cm. It lowered the carbon content from 30 to 1 atom % as well. The growth

temperature should be higher than 600 C to obtain reasonable range of resistivity (<1000

[M-cm). The grain size, film crystallinity, and film density increased with the addition of

NH3. TaN from PDEAT and NH3 exhibited better Cu diffusion barrier performance

compared to that from PDEAT single source because of higher film density. The step

coverage of the film grown with PEDAT single source was 80 % and decreased down to

56 % with NH3 addition, which is ascribed to mass-transfer limited conditions when NH3

was used as an additional nitrogen source [Cho98, Cho99].

TaN deposited with PDMAT (solid state) produced the insulating Ta3N5 phase film

resulting in high resistivity (>106 _M-cm). The film also had the high carbon impurity (~

20 atom %) even though it was deposited with NH3 as an additional N source. The









microstructure was amorphous when deposited in the temperature rang of 200 to 400 C

[Fix93].

MO-CVD of TaN using TBTDET single source has also been reported [Tsa95a,

Tsa96b]. A relatively low resistivity (920 [p-cm) could be obtained at 6500C. The

deposition rate was almost independent of the substrate temperature (450 to 6500C),

indicating this temperature range was mass transfer-controlled regime. XPS results

showed that the films deposited at 600 C contained 10 atom % of carbon and 5 atom %

of oxygen [Tsa95a]. Comparison study of diffusion barrier performance between PVD

TaN and CVD TaN from TBTDET proved that the barrier failure occurred at 650 C and

600 C for PVD and CVD TaN films. This demonstrates that the PVD TaN possesses a

better diffusion barrier property due to its higher film density than CVD grown TaN

[Tsa96b].

2.4 TaN ALD

ALD is a self-limiting growth method characterized by the alternate exposure of

chemical species in layer-by-layer manner while CVD delivers all required reactants to

the reactor simultaneously. A single precursor is exposed in the reactor at any given time,

so that a uniform layer of the precursor may chemisorb to the substrate surface. The most

important requirement for this step is the self-limitation for the precursor molecule

adsorption process. The self-limitation means that it limits further adsorption of

precursors by passivating the adsorption sites after the saturation coverage, roughly one

monolayer or less, is reached. [Kim05]

In general, self-limiting condition is satisfied by the ligands bonded to the metal

atoms in the precursors, such as halogen or organic ligands [Kim05]. Once this









chemisorbed layer forms, the reactor is evacuated or purged with inert gas. It is important

to minimize the incorporation of background impurities into the film and also to prevent

the process from shifting to CVD mode, which can occur if additional reactant from the

previous exposure remains in the reactor during the subsequent exposure. A second

precursor is then introduced to react with the chemisorbed layer from the previous step,

forming a new layer of material. The second precursor is also self-limiting, so that

reaction stops after the (sub-) monolayer of chemisorbed material from the previous step

has been consumed. Another evacuation or purge step follows the second precursor pulse

to ensure the complete removal of unreacted reactants and byproducts, which might have

formed during the second precursor exposure [Kim05]. The highly conformal,

ultra-uniform barriers afforded by the self-limiting feature of ALD will be essential in

the future to minimize the barrier's impact on the resistance per unit length in Cu

interconnects [Kap02].

The first report on ALD of TaN employed TaCl5 as metal precursor, which was

reacted with NH3. The film deposited at 400 C was composed of polycrystalline Ta3N5

with very high resistivity over 104 _tQ-cm [Hil88a, Rit99]. The formation of Ta3N5

instead of cubic TaN, was ascribed to the low reducing power of NH3. For this particular

ALD process, the growth rate was only 0.12 A/cycle at 200 C, while it increased to 0.22

to 0.24 A/cycle above 300 C. Meanwhile, the Cl content was above 20 % at 200 C, but

rapidly decreased to below 0.1 % above 500 C. The use of DMHy as an alternative

precursor of NH3 also formed high resistivity Ta3N5 films [Rit99]. Additional reducing

agents such as TMA and amines helped reduce the film resistivity [Ale01, Ale02]. By

using Zn as an additional reducing agent, cubic TaN with low resistivity of 960 [tQ-cm









was obtained, although residual Zn incorporation was still a problem [Rit99]. TaN ALD

using TaBr5 also produced similar results with TaCl5 [AleO1, Ale02].

As an alternate approach, the metal organic Ta source, TBTDET, has been reported

for ALD and Plasma Enhanced ALD (PE-ALD) of TaN [ParO 1, Par02]. For thermal

ALD, NH3 was used while for PE-ALD atomic H plasma was used as reducing agent.

XRD results showed that PE-ALD TaN is composed of cubic TaN phase, while no peaks

are evident on the as-deposited thermal ALD TaN films. While the resistivity of thermal

ALD TaN from TBTDET was ~106 [M-cm, the resistivity from PE-ALD TaN with same

precursor was as low as 400 [M-cm, when the pulse time of H plasma reached 30

seconds. Low resistivity of PE-ALD TaN was attributed to the formation of Ta-C bond,

which was not observed for thermal ALD TaN, as confirmed by XPS and XRD [Par02].

The step coverage was excellent up to aspect ratio of 10:1. As grown thermal ALD TaN

[Str04] from TBTDET contains 5 to 8 % of carbon and oxygen with the film resistivity

ranging from 500 to 1000 [tQ-cm for 30 nm thick films, which is very low compared to

Park et al.'s result (~106 -tQ-cm). It also showed excellent step coverage on 100 nm

trenches patterned SiO2 with an aspect ratio of 6.5 to 11 [Str04].

PDMAT MO-source has also been reported for thermal TaN ALD, focusing on

annealing studies of ultra-thin (<10 nm) ALD TaN [Wu04]. PDMAT and NH3 were

alternately exposed for TaN deposition employing argon as a carrier gas of PDMAT. The

growth rate was 12 A/min at the deposition temperature of 275 C. 10 nm thick TaN

films contain 2 atom % of carbon and 5 atom % of oxygen, showing no significant

change in the composition after annealing at 750 C. As deposited films showed the









amorphous structure, however, a fcc NaCi type c-TaN nanocrystalline structure was

obtained, when annealed at 750 C for 45 minutes.

2.5 Ternary Ta Nitrides

Addition of a third element to a binary refractory metal nitride matrix tends to

further disrupt the microstructure, increasing the probability of nanocrystalline or

amorphous phase formation [Ist00, Ram00]. Since grain boundaries play such an

important role as diffusion pathways for copper, the ternary phase metal nitrides are

expected to be good diffusion barriers [Kim03]. The ternary phase nitride films are

reported to remain amorphous up to high annealing temperature, while the binary nitrides

usually crystallize easily by annealing. Carbon, silicon, boron and aluminum are the most

studied third elements added to binary metal nitride films for this purpose.

TaCxNy was deposited by sputtering TaC target in an Ar/N2 atmosphere as a Cu

diffusion barrier [Sun01]. TaCxNy showed better thermal stability than that of respective

binary phases, and its resistance to Cu diffusion was better than TaC because of stuffing

the grain boundaries with nitrogen atoms. The films had low resistivity (-300 [tQ-cm),

and prohibited Cu diffusion after 30 min annealing at 6000C.

MOCVD of TaCxNy was reported using a mixture of PDMAT and PDEAT

[Hos00]. These films prevented Cu diffusion after a 30 min anneal at 5000C, but had high

resistivity (> 4000 [tQ-cm). PECVD of TaCxNy films was reported using PDMAT and

methane as a reactive gas. Film resistivity substantially decreased to 440 ~ 2400 [tQ-cm,

compared to thermal CVD of TaCxNy (6300 ~ 20000 [tQ-cm). PECVD TaCxNy films of

4 10 nm thickness successfully blocked the Cu diffusion after annealing at 360 C for 8

hr in H2/N2 environment [Eng02].









There have been several reports of TaSixNy deposition as a Cu diffusion barrier

application. TaSixNy was deposited by reactive sputtering of a Ta-Si target in N2 [Har01]

and Ar/N2 [Kol91] atmosphere. This ternary nitride had better barrier performance and

better adhesion to Cu than a Ta/TaN dual layer barrier. Various stoichiometries of

TaSixNy have been reported. Ta0.43Si0.04No.53 films exhibited a failure temperature of

8250C with high resistivity (1419 [QD-cm) [Lee99]. Amorphous Ta0.36Si0.14No.50 films

had high thermal stability, preventing interdiffusion between neighboring Cu and TiSi2

layers up to 9000C, with relatively low resistivity of 625 [t2-cm [Kol91]. The effect of

nitrogen content of the TaSixNy films on barrier performance was investigated [Kim97].

Films with nitrogen content greater than 40 atom % blocked Cu diffusion after an 800C

anneal, while those with lower nitrogen content failed after a 7000C annealing.

LPCVD of TaCl5+SiH4+NH3+H2/Ar at 5000C was also used to deposit TaSixNy

[Bla97]. As deposited TaSixNy showed some crystalline structure, with a stoichiometry of

Ta0.35Sio.11nN.54, and contained Cl and 0 impurities ranging from 3 to 5 atom %.

However, the films failed to prevent Cu diffusion after a 1 minute, 6000C anneal,

suggesting LPCVD TaSixNy performs no better than PVD TaN. TaSixNy PE-ALD has

been performed using TaCl5, SiH4 and N2/H2 plasma. The precursor exposure sequence

affected the growth rate and film properties. With a TaCl5-SiH4-plasma sequence,

amorphous films with lower resistivity (< 1000 [if-cm) were deposited, while high

resistivity (> 10000 [if-cm), polycrystalline TaSixNy was deposited from a SiH4-TaCl5-

plasma sequence [Kim02a].

TaxAlyN ALD deposited with the alternate exposure of TaCl5 or TaBr5, NH3 and

TMA (trimethyl Aluminum) was reported [Ale01]. The pulse length of TMA did not









have much influence on Cl and C content, which was 5 6 atom %, 20 ~ 22 atom %

respectively, while it increased the Al level from 10 to 12 atom % as the TMA pulse time

changed from 0.2 to 0.8 second. The change in precursor sequence caused differences in

film properties, such as chemical composition, resistivity and growth rate. The TaCl5-

TMA-NH3 sequence produced less electrically resistive films with less concentration of

carbon and chlorine, compared to the TMA-TaC15-NH3 sequence. Raising the deposition

temperature from 250 to 450 C increases the growth rate, ascribed to the thermal

decomposition of TMA at higher temperature, while it lowers the Cl content (15 to 3

atom %) and resistivity (22000 to 5000 [tQ-cm). A Cu diffusion barrier test on Cu/

TaxAlyN /Si structure showed that it failed at 600 C. Using TaBr5 as a different Ta

source for TaxAlyN produced higher resistivity with lower deposition rate than TaCl5

[Ale01l].

2.6 Integration of Low-k Materials on Cu Interconnects

As the feature sizes continue to shrink, chip performance becomes more limited by

BEOL due to interconnect RC delay and power consumption (CV2f). As previously

indicated, Cu metallization was first introduced to reduce resistance of metal wirings, and

then low-k dielectrics were developed to diminish the parasitic capacitance. One of the

most important changes in sub 65 nm node and beyond is the introduction of porous

dielectrics for the purpose of reducing the dielectric constant. Table 2-4 shows the

evolution of dielectric material properties which was driven by dielectric constant

reduction and also required for the next interconnect generation [Fay03]. Organic groups

were first introduced into the silicon oxide-based matrix to induce free volume (pore size

less than 1 nm) and decrease the dielectric constant down to three. OSG (Organo Silicate









Glass) and CDO belong to this generation of materials. Homogenous and larger pore

sizes were developed to further decrease dielectric constant of these materials with k <

2.2 by templated or porogen (Polymeric Pore Generator) removal methods. The large

porosity volume close to 45% and pore sizes around 3 nm are the general properties of

ULK (Ultra Low k) materials reported so far. However, these properties can lead to

serious problems in Cu/low-k integration [Fay03].

Interlayer adhesion, cohesive delamination, material deformation and crack

formation have been reported during low k dielectric integration during the CMP and

packaging steps [Ras02]. CMP feasibility of ULK is known to be correlated to its

mechanical properties. Young's modulus (E) larger than 4 Gpa, hardness (H) larger than

0.5 Gpa and adhesion energy (A) larger than 5 J/m2 are required to pass the conventional

CMP. However, the mechanical properties of ULK are just around the limits and even

lower. (E = 4 to 6 Gpa, H < 0.8 Gpa, A = 2 to 5 J/m2) [LinOl, SchO 1]. An addition of

metal dummies [Tsa02], adhesion promoter [Klo02] and new CMP technique with lower

shear stress [MosO1] have been reported as solutions for these mechanical problems

during ULK integration. Recently, UV (Ultra Violet) and EB (Electron Beam) curing

have been reported to improve significantly elastic modulus and the adhesion strength

without degrading the dielectric constant [Ito05, Got05].

There is a high possibility of contamination trapping or ULK material degradation

during integration process, because of its large porosity, which facilitates the diffusion of

chemical species from the process. Resist stripping and diffusion barrier depositions are

the main processes of the concern. Most stripping processes employ chemicals to break

organic bonds of the resist, also leading to breakage of the methyl bonds in SiOC and









carbon depletion. The resulting SiOC films become denser and more hydrophilic, causing

a significant dielectric constant rise [Rya02]. Compatible resist stripping processes with

different chemicals (H2, NH3) or plasma conditions are being optimized [Mos01].

Table 2-4. Properties of dielectrics (taken from [Fay03])
Ultra Low k
Property Low k SiOC Si02
porous SiOC

Dielectric constant k 2.2 2.9 4.4

Density (g/cm3) < 1 1.2 2.3

Carbon at. % 15 to 20 20 None

Sensitivity to water Hydrophobic Hydrophobic Hydrophilic

Porosity (%) -45 -25 None

Pore size (nm) 2.5 to 3.5 < 1.0 None

Young Modulus E (Gpa) 4 to 6 14 > 50

Hardness (Gpa) < 0.8 1.7 > 8

Thermal conductivity (W/mK) < 0.2 -1.4

Leakage current (A/cm2) < 8 E-10 8 E-10 8 E-10

Breakdown voltage (MV/cm2) > 4 3 5


As of now, PVD is the most common method of diffusion barrier deposition in

industry, but it suffers from conformality issues. Therefore, i-PVD (Ionized PVD), CVD

and ALD have been developed and demonstrated better conformality than conventional

PVD. However, even these advanced deposition techniques still have some problems

when it comes to ULK porous materials. For example, thin PVD barriers deposited on

porous dielectrics are discontinuous with pinholes in the films confirmed by ellipsometry,

which could be only solved by growing thick films [Bak01, IacO2]. For CVD and ALD

processes, the precursors tend to easily penetrate into porous ULK dielectrics. Penetration

of chemicals into ULK dielectrics can change the material structure and modify physical






28


and chemical properties of ULK materials [Fay03]. The porosity on the surface of ULK

dielectrics has been sealed prior to barrier deposition process to prevent the precursor

diffusion from CVD or ALD. Numerous pore sealing treatments such as plasma

treatment [Tsa04, Leo06, Wan05, Hoy04, Abe04, Hum05] liner deposition [Bon03,

Jez04, Toi02] and UV curing [Got06, Ito06] have been employed to serve this function.

The examples of these sealing treatments will be reviewed in Chapter 7.














CHAPTER 3
EXPERIMENTAL SETUP AND REACTOR SIMULATION USING CFD SOFTWARE

3.1 Reactor Design and Setup

Figure 3-1 shows the Process Flow Diagram (PFD) of the ALD reactor system used

in this work. Two bubblers, usually containing metal organic precursor, are immersed in

the constant temperature bath and heated to a defined temperature, where sufficient vapor

pressure (> 0.1 Torr for ALD) can be reached to ensure the efficient delivery of the

precursor for film deposition. Two metal organic source lines are needed for growth of

ternary films with the exception that the second metal source is the gas phase (e.g., SiH4,

Si2H6, and WF6). The nitrogen, argon, or hydrogen carrier gas picks up the metal

precursor at saturation and delivers it to reactor or bypass line, depending on the on/off

schedule of pneumatic valves located upstream to the reactor. The piping lines placed

before and after the bubbler to the reactor are heated with heat tapes up to 10 C higher

than the bubbler temperature to prevent the condensation of precursor during delivery.

The nitrogen gas not only acts as a carrier gas but also as a sweeping (purging) gas in

purge step during ALD cycles. Non-metal precursors, generally gas phase, are contained

in a gas cylinder within a gas cabinet. All carrier, purge nitrogen, and reacting gases are

metered to the reactor or bypass line and then to the vent system using Brooks 850 e(m)

mass flow controllers paired with a FM304V flow manager box.








MFC
Pneumatic v/v
Bubbler


U


Figure 3-1.Process flow diagram of ALD reactor system

Reactor inlet head (4) was designed as a cone-shape (Figure 3-2) so that the

reactant and purge gases can be uniformly distributed as they enter the reactor chamber.

To further increase the uniform distribution of the feed reactants over the substrate area

(2" wafer maximum), the removable shower head plate is attached at the bottom side of

reactor inlet using three bolts. The holes on the shower head plate were made larger to

allow adjustment of the distribution of reactant and purge gases. The cone-type inlet fits

beneath the CF flange (1), which is the connection point to the reactor body (Figure 3-

3), to minimize the distance between the shower head plate and the substrate. The other

purpose of this design was to lower the potential of precursor condensation by wrapping

heat tapes to the bottom of the cone shape head. Two lines (): Straight, 7): 90 bent) are

welded to the top of the reactor inlet head to prepare for the case of CVD runs, which

requires the separation of each reactant before they enter the reactor to prevent the

particle formation in the piping lines through the reaction.


1..12
























I2
(1 50)

---- 3


Figure 3-2.CAD drawing of reactor inlet part with shower head plate attached

The reactor body (Figure 3-3) has a double wall (jacket) structure, where cooling

water enters into the right bottom and exits out of the left top (6)) to operate the

ALD/CVD runs as a cold-wall reactor and to protect the Viton O-rings at the quick

access door for sample loading. The cooling water pipes are connected via 1/4" NPT

female ports. Two 600-400 conflat flanges (CF (2) are welded to the side of reactor. One

of them is designed to accommodate the quick access door, which has the quartz glass

sealed with Viton O-ring and vacuum seals after the sample loading. The other 600-400

CF is intended for attaching the transport chamber leading to the Cu deposition chamber









without breaking vacuum. A quick flange (QF-25 5 ) port is also attached at the side for

thermocouple insertion to measure the substrate temperature.


'^p (KBlAW


5-


Figure 3-3.CAD drawing of reactor body

The location of heater support (1@) can be adjusted vertically by employing the

ultra-torr fitting with Viton o-ring, which vacuum seals between the inner piping (6)

connected to the heater support and the outer piping ( 5) in the Figure of reactor

downstream body (Figure 3-4). This design allows the distance between the showerhead

plate and the sample substrate to be adjusted for the purpose of optimizing its location for

the film growth. The QF25 adaptor ( 7) is connected with the inner piping through the

ultra-torr fitting, which will then be linked to electrical feedthrough of QF 10 size for

heater power supply. Three QF25 flanges ( 3) are welded to the side of reactor









downstream body; one is a pressure transducer port, the other is a vacuum pump port, and

another one is for the rough vacuum pressure gauge.




12V. I HGTAWW)




45 GTAW GTAW














2X 237_

Figure 3-4. CAD drawing of reactor downstream

Custom designed resistive heater with 2" diameter from General Electric ceramics

heats the quartz susceptor on which the substrates are placed. K-type thermocouple from

the reactor side (5 in Figure 3-3) fits on 1/16" hole on the susceptor and provides the

Yokokawa UT-150-AN/RET temperature controller with the measured temperature.

After comparing the set point value with the measured temperature, the temperature

controller sends out the 4 20 mA control signal to SCR power controller (1025-10-20

from Control Concepts) adjusting the main power of 110 VAC to 24 VAC to control the

substrate temperature. The resistive heater is a pyrolytic graphite-based (PG) heater

coated with pyrolytic boron nitride (PBN). Operating temperature can reach over 1500 C









and it is chemically inert to most corrosive gases and liquids. It has high mechanical and

thermal shock resistance due to its low modulus and layered BN ceramics. There is no

potential of particle formation due to out-gassing of the coated BN because CVD BN is

fully dense and stable. Post type electrical connection is employed rather than heater

surface electrical connection to minimize the exposure of electrical connection point to

the reacting environments. For further isolation from the chemical surroundings, a cap

shape heater support is used. High temperature heater hookup wires (HTMG-1CU-

320S/C from Omega) from the electrical feedthrough (EFT0123058 from Kurt. J. Lesker)

are electrically connected to the heater via M4 thread, 0.7 mm pitch molybdenum bolts

(from Kamis). Ceramic insulator washer hats (from McAllister Technical Services) are

placed up and down of the heater support cap for the electrical insulation purpose.




-- Post type contact
M4 thread--


Washer -


Support cap

Molybdenum -


Figure 3-5. Schematic diagram of heater assembly









The reactor pressure is monitored using active Pirani gauge (from BOC Edwards)

paired with active gauge display unit. It is designed suitable for measuring medium range

and low pressure from atmosphere to 10-4 mbar. A typical pressure range for ALD

growth of metal nitrides and oxides is around 1 Torr. Maintaining reactor pressure almost

constant is very important during ALD growth, because it keeps the amount of source

pulse constant. Source is pulsed into the reactor by the pressure difference between

reaction chamber and source bubbler for a very short time. If the pressure is not kept

constant in the chamber, the amount of source at each pulse varies and thus difficult to

obtain uniform growth. For this purpose, the commercial f-120 ALD reactor (from

Suntola) employs the 'Inert Gas Valving System', which flows the continuous purge gas

of high flow rate for every sequence to minimize the pressure fluctuation when the

sequence changes. Unless there considerable pressure variation exists per sequence, a

pulsed purge can also be used. Reactor pressure is manually controlled by adjusting the

conductance of the angular valve, attached to the mechanical pump (E2M12 from

Edwards) in the system.

Our reactor, volume including the cone-shape inlet and body, was designed as 1750

cm3, which is small enough to ensure the complete purge of the remnant reactants and

byproducts within a very short time. Considering the capacity of the roughing mechanical

pump, 10.2 cfm (cubic feet per minute) equivalent to 4813 cm3/sec, it takes

approximately 0.36 sec (1750 cm3/ 4813 cm3/sec) to sweep the whole reactor volume.

Due to its small volume, the reactor pressure could quickly reach the base pressure of 10-3

Torr. It is critical in ALD that the flows should be laminar for the uniform adsorption of

precursor onto the entire substrate area. Therefore, typical Reynolds number (Re) should









not exceed 1000, otherwise the flow pattern becomes turbulent prohibiting the uniform

adsorption. The Reynolds number for a circular tube, which is the shape of our reactor,

can be described as following.


Re= Dp 4 Qp
/P n D/u

D: Tube diameter, : Average velocity, p : Density,

/ : Viscosity, Q: Volume flow rate

Insertion of typical growth conditions of ALD runs (nitrogen purge gas with 200 sccm

flow rate, 1.15 kg/m3 density and 0.0175 cp viscosity) gives a Reynolds number of 2.45,

satisfying the laminar flow condition.

It is important to expose or pulse sequentially to the purge and reactant gases to the

reactor or to bypass vent in the operation of ALD mode. For this purpose, the manifold,

located before the reactor, consists of five separate lines equipped with ten Swagelok ss-

bnv51-c pneumatic valves. One pneumatic valve controls (on/off) the reactant or purge

flow into reactor and the other valve manages the gas flow to the bypass line, making the

sequential pulse of reactants and purge gas possible. These valves are assembled as close

as possible to minimize the time delay between each step during the ALD. The distance

between horizontal tubes was designed at 2", minimum distance for the same reason. All

tubing is 1/4" diameter and connected with VCR fitting, suitable for the ultra high

vacuum environments.

Sequential exposure of each reactant and purge gas is managed by the control

program encoded with National Instrument (NI) LaBView software. Figure 3-7 shows the

control panel of the program. Currently, the panel shows a typical cycle of ALD growth

comprising four sequences with 6 pneumatic valves. Each column represents a sequence









and the row corresponds to a particular valve. The first sequence is the reactant 1 pulse

stage (RI), where only reactant 1 gas enters into the reactor, while the reactant 2 and

purge gas are vented via the bypass line. The second one is the purge stage (P), where N2

purge gas is the only gas into the reactor and other gases reactantss 1 and 2) proceed to

the vent. After the purge sequence, reactant 2 (R2) is exposed into the reactor, whereas

reactant 1 and purge gas bypass to the vent. R2 sequence is followed by another purge

stage, which has exactly same configuration of valve opening with the second stage (P).

























Figure 3-6. Manifold structure of pneumatic valves

The duration time of each sequence can be controlled by entering the number of sec

in the pulse time zone under the sequence tags. The sequence indicator shows the current

sequence with yellow color and the total cycle number (one cycle corresponds to the 4

sequences) can be entered for the purpose of film thickness control. The control signals









from this program become physically meaningful electric signals of +/- 24 VDC through

the 64-channel isolated digital I/O NI PCI-6514 interface board and WRB24SX-U

Switching Mode Power Supply (SMPS), connected to the actuators of solenoid valves via

electric wires. Then the solenoid valves control the flow of compressed air gas led to the

actuators of pneumatic valves, which open and close the pneumatic valves (normally

closed type).


Pulse Time 11 I
Reaclanil to Reaclo I
Reacianil toVeni 9 0
Reaclanl2 Io Reacloi, j
Rea clanl2 to Veni I 0 0 9


Sequence




3.2 Velocity and Temperaturge Profile Simulation of Reactor using CFD Software








cntinuity equation, can be written as follows:e
Cueni / Wll Cycle
__ Tolal Giowlh Time I _











Figure 3-7. Control panel ofALD growth programmed with LaBView) 0 (3-1)
Simulations on flow and thermal patterns in the reactor were performed using






conditions specific to this reactor geometry. The equation for conservation of mass, or

continuity equation, can be written as follows:


atp+ V. (pv) = 0 (3-1)









This equation is the general form of the mass conservation equation, simply stating that

the rate of increase of the density within a small volume element fixed in space is equal

to the net rate of mass influx to the element divided by its volume [ V- (pv) ].

Conservation of momentum can be described by the following equation.


(p) = pg-VP -V(pvv)-V-r (3-2)


This equation states that the rate of increase of momentum per unit volume [-(pv) ] is
at

equal to the sum of the rate of momentum gain by convection per unit volume

[ V (pvv) ], the rate of momentum gain by viscous transfer per unit volume [ V 7 ],

the pressure force on element per unit volume [- VP ], and the gravitational force on

element per unit volume [ pg ]. This momentum balance is exactly equivalent to

Newton's second law of motion, which is the statement of mass x acceleration = sum of

forces.

The conservation law for energy can be written as follows.

p(O+lv2)=_(V.pv( +lv2))-(Vq)+ p(v.g)-(V.pv)-(V.[r.v]) (3-3)
at 2 2

This equation includes two major energy terms commonly used in computational fluid

dynamics, which are kinetic energy associated with observable fluid motion and internal

energy associated with the random translational and internal motions of the molecules,

plus the energy of interaction between molecules. This energy conservation law states

that the rate of gain of energy per unit volume [- p(U + -v2) ] equals the sum of the
at 2

1
rate of energy input per unit volume by convection [- (V pv(U + -v2)], the rate of
2









energy input per unit volume by conduction [- (V q) ], the rate of work done on fluid per

unit volume by gravitational force [ p(v. g) ], the rate of work done on fluid per unit

volume by pressure forces [ (V pv) ], and the rate of work done on fluid per unit

volume by viscous forces [- (V [r v]) ].

To solve these coupled conservation PDEs, FLUENT employs a Finite Volume

(FV) method to convert the governing equations to algebraic equations that can be solved

numerically. The solution domain is subdivided into a finite number of contiguous

control volumes (CVs), and the conservation equations are applied to each CV. This

control volume technique consists of integrating the governing equations about each CV,

yielding discrete equations that conserve each quantity on a control volume basis.

Discretization of the governing equation of the steady-state conservation of a scalar

quantity q! is shown as a simple example, demonstrated by the following equation written

in integral form for an arbitrary control volume V as follows:

ptv-.dA= FV 4 -dA + SdV (3-4)


where

A: Surface area vector

F,: Diffusion coefficient for q!

Vq: Gradient of q!

S,: Source of q! per unit volume

Equation 3-4 is applied to each control volume, or cell, in the computational domain. The

two-dimensional, triangular cell shown in Figure 3-8 is an example of such a control

volume. Discretization of equation 3-4 on a given cell yields










Z fvf of -Af = V(V ),,.A +SV (3-5)
N faces N faces

where

N faces: Number of faces enclosing cell

f : Value of q! convected through face f


Pf vf Af : Mass flux through the face

(VO),,: Magnitude of Vq normal to face f

By default, the discrete values of the scalar q0 at cell centers (cO and cl in Figure 3-

8) are stored. However, face values Of are required for the convection terms in equation

3-5 and must be interpolated from the cell center values. This is accomplished using an

upwind scheme. Upwinding means that the face value of is derived from quantities in

the cell upstream, or "upwind", relative to the direction of the normal velocity v,.







Af cl









Figure 3-8. Control volume explaining the discretization of a scalar transport equation
(taken from [Flu06])

The mesh was generated using the Gambit software with cylindrical coordinates in

a two-dimensional format (Figure 3-9). The geometry of the mesh is almost the same as









that of actual reactor. An unstructured triangular grid was employed, which is most often

used for control volume methods. Three types of boundary conditions (i.e. inlet flow

velocity at reactor inlet, surface wall temperature at heater surface, and outflow type at

reactor outlet) were assigned during the mesh design step. The segregator solver was

used, where the governing equations (momentum, continuity and scalar) are solved

sequentially (i.e. segregated from one another), rather than in a simultaneous way. The

manner in which the governing equations are linearized, took an implicit form with

respect to the dependent variables.


Figure 3-9. Triangular grid mesh of ALD reactor

For a given variable, the unknown value in each cell is computed using a relation

that includes both existing and unknown values from neighboring cells. Cell-based option

was chosen, where cell center values are considered for computing the gradient. Energy









equation option was selected to simulate the thermal pattern around the resistive heater

area. Nitrogen was a fluid material that enters into reactor and its density, Cp, viscosity,

and thermal conductivity were extracted from the Fluent database and listed in Table 3-1.

Table 3-1. Properties of nitrogen gas
Density (kg/m3) Cp (J/kg-K) Thermal conductivity (W/m-K) Viscosity (kg/m-s)

1.138 1040.67 0.0242 1.163e-05

The operating pressure was fixed at 133 Pa, a typical growth pressure for ALD, and

the gravitational acceleration was turned on to the minus Y axis direction. Boundary

conditions used in this simulation are summarized in the following table (Table 3-2).

Linear velocity at reactor inlet was driven from the typical flow rate of N2 gas when the

ALD reactor is operating in the purge step mode. Volumetric flow rate of 200 sccm,

divided by 12 inch diameter circle area gives 0.0263 m/s. Outflow boundary condition

uses a typical outlet pressure of reactor, which is around 1 Torr. On every iteration, the

outlet velocity and pressure are updated in a manner that is consistent with fully

developed flow, satisfying all governing equations. Temperature at the heater surface was

set at 600 K and there is no heat flux through the side-walls of reactor.

Because of the nonlinearity of the equation, it is necessary to control the change of

qb. This is typically achieved by under-relaxation, which reduces the change of q!

produced during each iteration. In a simple form, the new value of the variable q within a

cell depends upon the old value, ,bld, the computed change ino A0!, and the under-

relaxation factor, a as follows:


0! = 0,1id+aA04(36


(3-6)









The under-relaxation factor for pressure, density, body forces, momentum and

energy are set as 0.3, 1.0, 1.0, 0.7 and 0.8. In the discretization scheme the first order

upwind, where q0 is set equal to the cell center value q0 in the upstream cell, was set for

momentum, and the second order upwind, where quantities at cell faces are computed

using a multidimensional linear reconstruction approach, was used for energy calculation.

After initializing the values from the inlet flow velocity boundary condition, the iteration

started and converged after 541 iterations. Figures 3-10 and 3-11 show the contour of

velocity magnitude and static temperature resulting from the simulation.

Table 3-2. Boundary conditions
Boundary location Boundary type Specific conditions

Inlet flow Velocity inlet -0.0263 m/s (Y velocity)

Outlet flow Outlet Pressure 1 Torr

Heater temperature Wall 600 K

Reactor wall Wall No heat flux



















Figure 3-10. Contour of velocity magnitude (m/s) (a) and velocity vector colored by
velocity magnitude around heater area (m/s) (b)



























Figure 3-11. Color filled contour of static temperature (K) (a) and contour line of static
temperature (K) around heater area (b)














CHAPTER 4
EFFECT OF NH3 ADDITION ON TaN MOCVD USING TBTDET

4.1 Introduction

As the feature size of integrated circuits (IC) shrinks to nanometer dimensions, the

need for conformal and reliable barrier materials increases considerably. One of the most

challenging issues in Cu-based ICs is to develop a suitable diffusion barrier material and

method to deposit it. Extensive research on transition metals and their nitrides (e.g., TiN,

TaN, and WN) has positioned TaN as the industry preferred barrier material. Ta and its

nitride are used as the liner for Cu damascene interconnects due to its high Cu reliability,

optimized adhesion for Cu electromigration resistance, robust via interfaces, and

redundant current strapping for added chip reliability [Ede02]. Most TaN layers tested,

however, have generally been deposited with reactive sputtering, which does not provide

good conformal coverage of submicron features and high aspect ratio contact and via

holes, due to its inherent shadowing effect. Therefore, CVD of the Ta/TaN bilayer barrier

combination has received more attention because of its superior conformality compared

to sputter-deposited barriers.

TaN-CVD using a metal halide source normally requires a high deposition

temperature to obtain low resistivity and minimal halide content [Hie74, Che97]. As

previously mentioned, halide incorporation in the film reduces the adhesion strength of

the film, which reduces long term device reliability [She90, Yok91]. Therefore, halide-

free, metal organic-based TaN CVD promises to provide lower deposition temperature

without halide incorporation. Although MOCVD can lower the deposition temperature,









the temperature should be sufficiently high to obtain low resistivity as well [Tsa95a]. In

addition, MOCVD of TaN tends to yield a high carbon concentration in the film, which

leads to high resistivity, especially when TaN is deposited with a single source precursor

[Cho99].

In this work, TBTDET and ammonia were employed as co-reactants for TaN CVD.

Previous studies using TBTDET were performed at a relatively high temperature (450 to

650 C), which is not acceptable for the modern IC processes that require a low thermal

budget. Therefore, a lower deposition temperature window (300 to 400 C) was selected

to assess the feasibility of TaN grown by MOCVD at low temperature as a Cu diffusion

barrier.

Previous research on MOCVD of TaN using TBTDET was limited to supplying

only TBTDET as a single source precursor without any additional nitrogen source or

reducing agent. Since TBTDET contains four nitrogen atoms bonded to each Ta atom

(Figure 4-1), TaN films can be grown with this single precursor by CVD. The films

deposited with TBTDET in an inert carrier gas, however, contain a significant amount of

carbon, reaching up to 30 at %, which is ascribed to incomplete cleavage of the C-N

bonding involving both the diethyl and t-butyl ligands. Thus the addition of NH3 during

MOCVD of TaN was explored as a mechanism to reduce the carbon content of TaN films

using TBTDET. It was thought that the transamination reaction of NHs with

diethylamido or tert-butylimido ligands and possibly the reaction of added H would

reduce the carbon incorporation in the TaN films.












N

-\ I
N-Ta = N

I
N



Figure 4-1.Chemical structure of TBTDET
Encouragingly, there have been similar reports on film property improvements of

MOCVD grown metal nitrides through the simple addition of NH3. For example, the

resistivity and carbon impurity of TiN using TDEAT [tetrakis-diethylamino-titanium]

[Sun94, Raa93, Mus96] was reduced significantly when supplying NH3 as compared to

flowing just TDEAT as a single source. As another example, Cho et al. [Cho99] reported

that as-grown TaN films deposited with PDEAT and NH3 also exhibited lower resistivity

and carbon content as compared to the films grown with PDEAT alone.

Similar effects are expected upon addition of NH3 during growth of TaN from

TBTDET due to the presence of dimethylamido ligands in both PDEAT and TBTDET

and the propensity for tantalum dimethylamido complexes to undergo transamination

with ammonia [Hol90]. Although TBTDET also contains a tert-butylimido ligand,

reaction of imido ligands with amines such as NH3 is generally facile [Whi06]. This

transamination reaction with ammonia is expected to lower carbon content and resistivity

by removing carbon-containing ligands in TBTDET. In the study presented in this









chapter, TaN films were deposited with and without NH3 and the various film properties

(resistivity, chemical composition, microstructure, surface morphology, and density), as

well as the deposition characteristics, are compared using 4 point probe, AES, XRD,

SEM, XRR, and AFM. A test to determine the efficacy of the Cu diffusion barrier was

also performed with TaN films grown with a single source and a range of NH3 flow rates.

4.2 Experimental Details

The deposition of TaN was performed on standard (100) silicon substrates in a

customized cold-wall, low-pressure CVD reactor. TBTDET (vapor pressure 0.1 Torr @

90 C, supplied by Alfa Aesar, MA) was contained in a stainless steel bubbler and was

heated to a temperature in the range 80 to 90 C. Nitrogen with a flow rate of 50 to 100

sccm was used as a carrier gas. TaN films were deposited at a sequence of temperatures

in the range 300 to 400 C at 25 C intervals with a typical growth pressure of 1 ~ 2 Torr,

and the ammonia flow rate was varied from 0 to 150 sccm. The film thickness was

measured by cross-sections SEM. Four point probe was employed to measure the film

resistivity. The microstructure was analyzed using XRD and SEM, while AFM was used

to observe the surface morphology of the films. The chemical composition of the films

was probed by AES measurement and the density extracted from XRR profiles.

Cu of 100 nm thickness was deposited by sputter deposition onto CVD-TaN layer

of 50 nm thickness to investigate the diffusion barrier properties for Cu metallization.

The samples were then annealed at 500 C for an hour in inert gas (N2) atmosphere to

monitor the changes in the Cu/TaN/Si structure. The barrier failure was analyzed by

observing the copper silicide peak in the XRD pattern after annealing. The barrier failure

was also tested by monitoring the formation of etch pits on the Si surface. Cu and TaN

films were wet chemically etched with HNO3:H20 = 1:20 and HF:H202 = 1:2 solution









respectively, after annealing. The surface of the Si substrates was Secco etched for 15 to

30 seconds at room temperature, and the size and density of etch pits were observed by

SEM.

4.3 Results and Discussion

4.3.1 Resistivity and Chemical Composition

Figure 4-2 depicts the change in the resistivity of films grown at 400 C as a

function of the NH3 flow rate. As the NH3 flow rate increased the film resistivity

decreased, most dramatically as small amounts of NH3 were added. The resistivity was

around 26800 [tQcmwith no NH3 flow and it drastically decreased to 15000 [tQcmat an

NH3 flow of only 15 sccm. Additional NH3 resulted in a more gradual decrease in the

resistivity.

Resistivity (pIjcm) Ratio
1
-\ N/Ta
25k -
0.8

20k
,0.6
15k -

0.4
10k .--
10k ------------ Resistivity

S"T 0.2
5k C/Ta
O/Ta_
Ok -NH3 flow
0 30 (sccm) 60 150
(sccm)
Figure 4-2.Resistivity, and nitrogen, carbon and oxygen content relative to Ta in TaN
films as a function of ammonia flow rate

The minimum resistivity obtained was -6300 [tQcm which is still more than an

order of magnitude higher than that of sputtered TaN (-250 [tQcm) [Cho99] presumably









due to high carbon levels (~ 5 at %) in CVD TaN. A decrease in the resistivity with the

addition of NH3 has also been reported for TiN-CVD using TDEAT [Sun94, Raa93,

Mus96], and TaN-CVD with PDEAT [Cho99]. This trend is seen to track changes in

impurity levels in the films. Figure 4-2 also shows the impurity ratios C/Ta and O/Ta as a

function of NH3 flow rate. Increasing the NH3 flow reduced the carbon content of the

films presumably via transamination with ammonia, which would remove the carbon-

containing diethylamido ligands as the volatile compound diethylamine [Gor90, Dub94].

Following the proposed mechanism of transamination of Ti(NMe2)4 and NH3

[Wei96], it is suggested that NH3 initially forms a weak intermediate adduct with Ta

center in (NEt2)3Ta=NBut, which is probably not stable given the steric bulk of amido

and imido ligands. H-atom transfers between the coordinated NH3 and an amido ligand

probably occurs via a four-transition state followed by elimination of HNEt2 (Figure 4-3).

NH3
NH3
Et2N NBut I
Ta f EtzN T NBut

Et2N NEt2 Et2N NEt2





H NH2
H2N T NBut E u

Ta^ _Et1 ,T NBu'

S NEt Et2N NEt2

Et2NH
Figure 4-3. Proposed transamination reaction mechanism between TBTDET and NH3
(Concept taken from [Wei96])









ALD studies of TiN using TDMAT and NH3 also confirmed that transamination

reaction paired with an amine elimination step is the primary reaction in depositing TiN

films as evidenced by FTIR data and QCM in-situ measurements [Ela03]. Isotopic

labeling studies of TiN CVD using 15NH3 and ND3 demonstrated the formation of Ti 15N

and the use of ND3 gives DNMe2 [Dub92]. A related ammonolysis reaction involving the

tert-butyl imido ligand will also be facile under the deposition conditions [Bec03]. The

oxygen content is also reduced by the addition of ammonia but that cannot be a direct

result of ligand removal by transamination because the ligands contain no oxygen. TaN

films grown in the presence of ammonia, however, are expected to be denser via

transamination reaction, which would inhibit post-growth incorporation of oxygen upon

handling the films in air.

4.3.2 Microstructure and Surface Morphology Analysis

XRD patterns (Figure 4-4) of the TaN films were obtained as a function of NH3

flow rate to investigate how NH3 addition affects the film crystallinity. When TaN was

deposited without NH3 flow, no peaks were observed, consistent with an amorphous

structure. With the addition of NH3, the peak of h-TaN began to appear and as more NH3

flow was added, and c-TaN (111) and c-TaN (200) peaks became more apparent in the

pattern. CVD of TaN from PDEAT [Cho99] showed a similar trend in that the film

deposited with NH3 showed a greater extent of polycrystalline structure than the film

grown with PDEAT only as a single source precursor. Increased crystallinity with higher

NH3 flow is also consistent with the observed lower resistivity plotted in Figure 4-2.































30 32 34 36 38 40 42 44
Figure 4-4. XRD patterns of TaN films deposited at 300 C as a function of ammonia
flow rate

There was a distinct difference in the surface morphology between the films grown

with TBTDET alone (Figure 4-5(a)) and those grown with NH3 (Figure 4-5(b)), although

the films grown with NH3 at different flow rates (30, 60, and 150 sccm) did not exhibit

much variation. NH3 addition yielded rougher surface with more hillocks and larger grain

size (45 to 60 nm) than the deposition with TBTDET alone (9 to 11 nm).

1.00 1.00



0.75 0.75



0.50 0.50



0.25 0.25



0 0
0 0.25 0.50 0.75 1.00 0 0.25 0.50 0.75 1.00
JiM JUM
Figure 4-5. Surface morphology of TaN films deposited (a) without NH3 at 300 C and
(b) with NH3 (30 sccm)









Observation of cross-sectional SEM images (Figure 4-7) also produced the same

conclusion, showing films grown in the presence of NH3 with larger grain size, while the

films from TBTDET alone showed no microstructure and a smaller grain size which

could not be recognized in the field-emission SEM image (50,000x).

4.3.3 Deposition Characteristics and Density Analysis

The deposited films were smooth, nonporous, pinhole-free and adhered well to the

substrate when tested with a tape. The films were mirror-like with a reflected gold color.

Figure 4-6 clearly illustrates that the deposition behavior was significantly changed when

NH3 was introduced to the reactor. The deposition rate from TBTDET alone as a

function of growth temperature showed a steep increase with temperature characteristic

of a surface reaction controlled process. The Arrhenius plot shown in Figure 4-6 shows

an activation energy of 0.3 + eV, while the overall deposition with NH3 flow had the

appearance of mass transfer-controlled regime with an activation energy of 0.05-0.07 eV.

This result suggests that NH3 addition gives better control in film thickness over single

source deposition since the film thickness doesn't vary significantly with the growth

temperature in mass transfer controlled reaction.

The effect of adding an NH3 flow at a constant flow of TBTDET was to decrease

the growth rate relative to the case without NH3 flow at the same growth temperature.

The growth rate generally trended down with increasing temperature, which usually

implies a thermodynamic limit or desorption dominated process. The deposition rate,

however, cannot be measured by the thickness unless the density is constant, as discussed

below. A similar behavior of deposition rate with temperature was observed with TaN

CVD from PDEAT [Cho99]. Their results without NH3 flow taken over the temperature

range (300 to 375 C) showed surface reaction limited growth at low temperature (300 to









350 C) and a mass transfer controlled pattern when deposited with 25 sccm NH3 flow.

The same sequence was also observed in TiN CVD from TDEAT [Raa93], showing a

higher activation energy (Ea = 0.5 eV, 280 to 420 C) for thermal decomposition of

TDEAT than that (0.09 eV) for the transamination reaction of TEDAT with NH3.

Ln (Dep. Rate)

4.4 0 sccm <


4.2 -
30 sccm
60 sccm
4.0 -

100 seem \
3.8 '150 sccm


3.6 -

3.4 1000/T (1/K)
3 .4 '-----------'------'----- '
1.40 1.50 1.60 1.70 1.80
Figure 4-6. Growth rate (A/min) vs. reciprocal growth temperature with different NH3
flow rates of 0 to 150 sccm

Figure 4-7 shows the cross-sectional SEM images of TaN films grown at 4

different NH3 flow conditions at 400 C. A decrease in the deposition rate with higher

NH3 flow rate was observed in most films grown at constant temperature. A lower

deposition rate with NH3 addition was reported for TDEAT [Mus96] as measured by

XRR and step profilometry. In contrast, an increase in the deposition rate with NH3 was

reported for TDEAT, where film thickness was derived from weight gain measurements

[Raa93], and for PDEAT, as measured by cross sectional SEM [Cho99]. Although the

reason for this conflict is not obvious, one possible reason is that the higher NH3 flow









increased the extent of the transamination reactions, which leads to a denser film

structure. Based on the assumption that the amount of Ta incorporated into films is the

same during the fixed deposition time for both runs without and with NH3, the thinner

films imply a higher density.


Figure 4-7. Cross-sectional SEM images of TaN films deposited with various NH3 flow
rates at 400 C

The film density was evaluated by XRR (Figure 4-8) and the data were analyzed

using the WinGixa program. XRR can estimate the film density by measuring the critical

angle (starting angle of total reflection of X-ray), which is proportional to film density. A

very thin SiO2 layer (~ 20 A) was assumed to exist at the interface between the Si

substrate and TaN film. Also, TaN with C (-15 at %), and 0 (-10 at %) impurities was

introduced in the near upper surface region of TaN for the simulation. As expected, the


NH3: 30 sccm










films deposited with higher NH3 flow rate showed higher density, which would make

these films better candidates for Cu diffusion barriers if they remain essentially

amorphous. When the NH3 flow was greater than 60 sccm, the simulated density is 10

(g/cm3), which is about 60 % of the density of bulk TaN (16.3 g/cm3). As previously

mentioned, a denser film would be expected to be more efficient in prohibiting post-

growth oxygen incorporation, which is consistent with the lower oxygen content in the

films deposited with NH3 flow and the flattening of the profile at NH3 flow rates greater

than 60 sccm.

Log (I)
a.u.
NH3 flowrate (sccm) 0 30 60 100 150
Density (glcm3) 8.99 9.51 9.99 10.0 10.1




100 sccm



150 sccm



0 seem



20
0.5 1 1.5 2 2.5 3
Figure 4-8. XRR profiles of TaN films deposited at 400 C with different NH3 flow rates

4.3.4 Nucleation Step

For deposition both with and without NH3, there did not appear to be an induction

period for nucleation. TaN began to nucleate and form films within 5 sec of the reactant

exposure (Figure 4-9) as judged by the surface roughness change from 1.5 A (bare Si) to

3.6 A (TaN from TBTDET alone for 5 sec) and 2.7 A (TaN from TBTDET and NH3 for









5 sec). TaN deposited from TBTDET alone for 20 sec showed a resistance of -1970 ohm,

which is as close as that of TaN (-1880 ohm) from TBTDET and NH3 for 10 sec. On the

other hand, TaN deposited for 5, 10, and 15 sec without NH3 and 5 sec with NH3 had the

resistance in the range 480 to 780 ohm. This indicates that full coverage of TaN on Si

occurred sooner for deposition with NH3 (10 sec) than for TBTDET single source

deposition (20 sec), since the full coverage of TaN will increase the film resistance. It is

interesting that the growth rate from TBTDET single source was higher than that with

NH3 flow at 350 C, which shows the opposite behavior with nucleation rate.

( 15s 0 20s














Figure 4-9. AFM scans of the initial stages of deposition from (a) TBTDET only and (b)
TBTDET with NH3 = 50 sccm at 350 0C

4.3.5 Diffusion Barrier Performance


To evaluate the diffusion barrier performance, Cu of 100 nm thickness was sputter-

deposited onto 50 nm thick MOCVD-TaN grown with different NH3 flow. The samples

were then annealed at 500 C for an hour in an inert gas atmosphere (N2) to monitor the

reactions in the Cu/TaN/Si structure. Among those samples, only TaN deposited from

TBTDET single source showed the Cul5Si4 peak (Figure 4-10), suggesting that Cu









diffused through the TaN barrier to form copper silicide, while the other films grown

with NH3 exhibited peaks associated with TaN(1 11) and Cu(1 11) but not any related to

copper silicide. The intensity of the h-TaN(100) peak is seen to increase with increased

NH3 flow, which is another indication of greater crystallinity for films deposited with

NH3.

a.u Cu15Si4(220)ll h-TaN(100) Cu(1I11)

150 sccm










30 sccm
30 sccm






20 25 30 35 40 45 50
Figure 4-10. XRD spectra of Cu/TaN/Si structures annealed at 500 C for one hr for films
grown with different NH3 flow rates.

Although XRD spectra suggested that none of the TaN films deposited with NH3

failed the barrier test, the etch pit test revealed that TaN grown at NH3 flowrate of 0 and

30 sccm did not pass the barrier test (Figure. 4-11). A high density of etch pits formed by

the Cu diffusion into Si that was subsequently revealed by etching the bare Si surface in a

Secco solution. TaN deposited with 30 sccm NH3 exhibited a lower population of etch

pits, but the presence of etch pits indicated barrier failure. In contrast SEM images of

etched films grown with an NH3 flow rate of 60 sccm and greater showed no etch pits.










Revealing Cu diffusion by etching of the Si is clearly more sensitive than detecting Cu

silicides by XRD [Cho99]. Despite higher crystallinity, which is expected to provide

grain boundary diffusion pathways, the higher nitrogen content and density of the films

deposited with NH3 appear to improve the Cu diffusion barrier performance. In addition,

the lower film resistivity and impurity levels (C and 0) with NH3 addition render TaN

barriers superior to those from TBTDET single source deposition.


Figure 4-11. SEM images of Si surface after annealing the structure Cu/TaN/Si at 500 C
for 1 hr, followed by the removal of Cu and TaN layers, and then etch in
Secco solution to reveal etch pits if Cu penetration occurred. The TaN films
were grown at various values of NH3 flow.

As evidenced above, better diffusion barrier performance was observed for TaN

films when NH3 was added to the reaction chamber. The film density, surface roughness,

crystallinity, film conductivity, and nitrogen content each increased with increasing NH3

flow, while the C and 0 impurity levels decreased. It is well known that high film density

is a desirable property for diffusion barriers to eliminate diffusion across voids, defects or


NH3:,O sccm










MAI ,I I 1..",v 11- 1 1
NH3: 60 sccm










MAI: "A.- in -., "'! I I


NH3: 30 sccm











NH3:150 sccm









loosely packed grain boundaries. As an example, sputter-deposited TiN exhibited better

Cu diffusion barrier performance as the film density increased, which led to the

conclusion that film density is an important property in achieving a good diffusion barrier

performance [Par96]. Other experiments on TiN for application as a Cu diffusion barrier

reported that the diffusion failure temperature increased with increased film density

[Rha98]. In this research, the TaN film density increased from 9.0 to 10.0 g/cm3 as NH3

flow rate was varied from 0 to 60 sccm. Higher NH3 flow rate is expected to lead to

greater transamination reaction extents between NH3 and carbon containing ligands,

resulting in a more densely packed structure.

The nitrogen content in the film is also considered an important property for

diffusion barrier performance. TaN showed the highest failure temperature among Ta,

Ta2N, and TaN films as examined by sheet resistance measurements, XRD, and Secco

etching methods [Kal00]. Improved performance with higher nitrogen content was also

observed by Wang et al. from BTS (Bias Thermal Stress) measurements [Wan98]. AES

measurements on TaN films showed that nitrogen content in the film, and thus the barrier

performance, increased from 0.68 to 0.90 of N/Ta ratio with the addition of NH3. In

addition to higher density and nitrogen content, NH3 addition significantly reduced the

carbon content and film resistivity to further improve the barriers.

4.4 Conclusions

TaN films were successfully deposited by CVD from TBTDET and NH3. The

chemical composition, microstructure, surface morphology, density, resistivity, and

deposition characteristics were investigated as a function of ammonia flow rate and the

key results are summarized in Table 4-1. Encouragingly the film resistivity decreased

from 26800 aQcnto 6300 aQcmwith the addition of an NH3 flow to the reactor. In the









same films, the C/Ta atom ratio dramatically decreased from 0.95 to 0.12 and was

accompanied by an increase in the N/Ta atom ratio to Ta from 0.67 to 0.87, when 150

sccm NH3 injected as compared to films deposited with TBTDET alone.

Not as promising, the films became more crystalline with larger grain size and

higher density along with rougher surfaces with higher NH3 flow rate. Deposition from

TBTDET alone exhibited a surface reaction controlled regime, while the deposition when

NH3 was added showed a nearly temperature independent behavior in the range 300 to

400 C. The change in the film properties with the addition of NH3 flow is attributed, in

part, to transamination reaction between NH3 and carbon containing ligands in TBTDET

precursor. The efficacy of the films as Cu diffusion barriers was evaluated by both XRD

and etch pitch density determination on annealed Cu/TaN/Si samples. The results

showed that TaN deposited with NH3 exhibited superior barrier quality than TaN grown

with TBTDET alone, which was believed primarily a result of the higher film density and

nitrogen content.

Table 4-1. Summary of film properties and diffusion barrier test results of TaN deposited
with and without NH3
Density N/Ta C/Ta Resistivity Grain size Cu15Si4 formation:
(g/cm3) ratio ratio ([tacm) (nm) after 500 C, 1 hr
Without 8.99 0.667 0.959 26800 9-11 Cu15Si4 observed
NH3
With
Cui5Si4 not viewed
3 9.51-10.9 *88 .1- 9500-6300 45-60 No etch pits for
(30-150 0.886 .0.183 3 3
sccmNH3> 30 sccm
sccm)














CHAPTER 5
ULTRA-THIN ALD TaN FILMS USING TBTDET AND NH3 FOR Cu BARRIER
APPLICATIONS

5.1 Introduction

Al-based metal, Si02 dielectric-based interconnect metallization is being replaced

with Cu, low-K dielectric-based combination as the feature size shrinks and the speed of

electronic devices increases. The lower resistivity value of Cu and lower value of K for

the interlayer dielectric enable electrical signals to move faster by reducing the RC time

delay. Cu also possesses superior resistance to electromigration, which is a common

reliability problem with Al-Si metallization. Cu, however, is a very fast diffuser in

silicon, which can cause serious problems in the device, including increased contact

resistance, altered barrier height, leaky p-n junctions, and destruction of interconnects

[Bch03]. Therefore, an appropriate diffusion barrier is needed between Cu and its

underlying layer. In addition, a diffusion barrier acts as a passivating layer protecting the

Cu interconnects from corrosion and oxidation and promotes adhesion of the Cu layer.

Among various transition metals and their nitrides, Ta and its nitrides are known to

be suitable liners because of their high reliability as a Cu barrier, good adhesion,

resistance to electromigration, robust via interfaces, and redundant current strapping for

added chip reliability [Ede02]. Currently, the bi-layer liner Ta/TaN, deposited by PVD, is

being used in commercial applications. PVD has been the most widely used method for

diffusion barrier deposition in current semiconductor technology (>100 nm minimum

feature size). PVD-based technology, however, has an inherent limitation in the









deposition of sub-100 nm and high aspect ratio structures due to the directional nature of

the depositing flux and high sticking probability on most materials [Kim05].

Accordingly, ALD is receiving extensive attention due to its ability to grow films with

excellent conformality, thickness uniformity over large area, and precise thickness

control at the atomic level.

Early reports on TaN-ALD used a halide source (TaCl5) sequence with NH3. The

films showed very high electrical resistivity (> 104 [Ctf2)c and were polycrystalline Ta3N5,

which was attributed to the relatively low reducing power of NH3. The high Cl content (>

20 at %) at 200 C was also an issue with use of TaCl5, leading to possible corrosion of

the Cu interconnect lines [Hil88b]. Replacement of NH3 with dimethylhydrazine,

(CH3)2NNH2), yielded TaNx films similar to those obtained from NH3 but with

significant carbon incorporation [Jup00]. Addition of reducing agents such as TMA

(trimethyl aluminum) or amines gave minor improvements in resistivity [Ale01l]. The use

of Zn as an additional reducing agent produced the cubic phase of TaN with significantly

lower resistivity (960 [tcmn) [Rit99]. TaBr5 [Ale02] was also employed for TaN ALD,

which produced results similar to those found with the use of TaCl5.

As an alternative approach, MO (Metal Organic) sources such as TBTDET [Tar02,

Str04] and PDMAT [Wu04] were reported as a precursor for TaN-ALD. Thermal ALD

TaN films deposited with TBTDET showed lower density and higher resistivity

compared to PE-ALD TaN, which takes advantage of the reducing power of plasma

generated hydrogen or nitrogen. The low resistivity of PE-ALD TaN was attributed to the

formation of the more metallic Ta-C bond [Par02]. A fcc NaCl-type nanocrystalline









structure of TaN was obtained when PDMAT was used as the Ta source in thermal ALD

[Wu04].

In this study, TBTDET and ammonia were employed for thermal ALD of TaN.

Previous studies on TaN-ALD using TBTDET and NH3 [Ale02, Par02] did not provide

specific data on thickness as a function of reactant exposure time, which is essential to

establish the window for achieving the self-limiting adsorption feature of ALD. In this

dissertation, XRR (X-ray Reflectivity) was used to measure the thickness of ultra-thin

(<10 nm) TaN films, which was used to identify the self-limiting growth zone.

Additionally, the ALD temperature window, which is also needed for implementation of

precise thickness control, was determined for this precursor combination.

The ALD growth characteristics during the initial stages were investigated by

measuring the film thickness as a function of number of cycles. A linear dependency of

ALD film thickness on number of cycles is expected, but the dependency can be non-

linear during the initial stages of growth because the adsorption/reaction characteristics

can be different on the substrate surface as compared to finite thickness of the material

being grown [LimOO, LimO1, Sat02a]. It is critical to understand the evolution of growth

during initial stages and the mechanism of film closure for growth of ultra-thin (<10 nm)

films that will be required for future diffusion barriers. Established tests for determining

the quality of diffusion barriers were developed and mostly applied to films that are thick

(> 10 nm) relative to future requirements. The diffusion barrier thickness, however, is

anticipated to be less than 10 nm by 2007 according to the ITRS roadmap [Tra04]. In this

research, the minimum thickness of ALD-TaN barrier was determined by applying









standard XRD and etch pit density measurement to annealed ultra-thin (7 to 100 A)

Cu/ALD-TaN/Si structures.

5.2 Experimental Details

ALD of TaN was performed on standard (100) silicon substrates in a customized

cold-wall vertical ALD reactor. Pneumatically actuated valves, controlled with a

LaBView-based algorithm interfaced to a Digital Input/Output interface board, were

sequenced to flow each reactant and purge gas to the reactor sequentially using a run-vent

design. TBTDET (supplied by Alfa Aesar, MA vapor pressure 0.1 Torr at 90 C,),

contained in a stainless steel bubbler, was heated to a constant temperature with a set

point of 80 C. The carrier gas was N2. TaN films were deposited at a constant substrate

temperature, with a set point in the range 200 to 400 C with a typical growth pressure of

1 Torr. The deposition cycle began with an exposure of TBTDET for 6 to 18 sec with 20

sccm of N2 carrier gas. Then nitrogen flowed into the reactor at 200 sccm for 10 sec to

ensure the complete sweeping of excess TBTDET precursor as well as volatile

byproducts. After this purge step, NH3 of 20 sccm was introduced for 10 sec, followed by

another N2 purge sequence.

To investigate the diffusion barrier performance of ultra-thin TaN films for Cu

metallization, a Cu film of 100 nm thickness was sputter-deposited onto the ALD TaN

layer of 7 to 100 A thickness. The samples were then annealed at 500 C for 30 min in N2

to monitor the changes in the Cu/TaN/Si structure. The barrier was judged as failed if a

copper silicide peak appeared in the XRD pattern of the annealed structure. The barrier

failure was also tested by observing the formation of etch pits on the Si surface. After

annealing the Cu/TaN/Si structure, the Cu and TaN films were removed by wet chemical

etch sequentially with HNO3:H20 1:20 and HF:H202 1:2 solutions. The surface of the Si









substrates was then etched with a Secco solution and the size and density of etch pits

were observed by SEM.

5.3 Results and Discussion

5.3.1 Confirmation of Self-saturation ALD Growth

A series of TaN films were grown at 300 C using 80 cycles with variable

TBTDET exposure time to locate the saturation zone for TBTDET. The exposure time of

TBTDET was changed from 6 to 18 sec with fixed NH3 exposure times of 10 sec and

purge time of 10 sec for both purges. The TaN film thickness was then measured using

XRR and the spectra are shown in Figure 5-1 for several exposure times at different

temperatures. The thickness of each film was extracted using the WinGixa software. The

simulation data provides the film thickness, surface/interface roughness, and film density

by fitting the actual logarithmic reflectivity data. The final x2 value, which determines

how the simulated reflectivity matches the measured one, was set as low as 5.0 x 10-2. A

very thin SiO2 layer was assumed to exist at the Si-TaN interface. It is noted that C (-15

at %), and 0 (-10 at %) were introduced into the near surface region of the TaN surface

in the simulation to ensure a realistic assessment.

The growth rate based on the thickness data from XRR profiles is plotted in Figure

5-2. The error bars were estimated by considering +5 % error in thickness determination

from XRR [Val06] and the dashed-line is drawn to aid the viewer's eye. These results

suggest that the effective number of Ta atoms deposited per cycle increased for TBTDET

pulse time less than 9 sec, but for longer times the deposition rate remained constant (2.6

A/cycle) through 11 sec of exposure time, indicating it reached the apparent surface

saturation growth mode, which is a feature of ALD. For exposure time less than 9 sec

there is apparently insufficient time to reach saturation. The growth rate then increased









to 3.5 A/cycle at 12 sec exposure time, likely the result of sufficient time for parasitic

CVD reactions to contribute to growth.

As indicated in Figure 5-2, the self-limiting region at 300 C was from 9 to 11 sec

TBTDET exposure time. However, it was expanded to from 9 to 14 sec, when ALD

operates at 250 C as shown in Figure 5-3. This indicates that the parasitic CVD reaction

caused by the partial disruption of self-limiting adsorption at high temperature (300 C)

produced the narrower self-limiting zone compared to that of ALD at lower temperature

(250 C). The deposition rate in self-saturation growth zone was 2.63 A/cycle, which is

53.6 % of the lattice constant (4.91 A) of h-TaN phase. The saturation coverage of many

metal-containing precursors for metal and metal-nitride ALD reactions is a fractional

monolayer (ML). One of the most-widely accepted reasons is the steric hindrance of

adsorbed metal precursors [Kim05] and it appears this precursor yields 12 ML coverage

per cycle.

a.u.


-- v-\ ^^^^^ V 12s (276 A)



10s (214 A)

-9s (207 A)



6s (118 A)

20
0.5 1.5 2.5 3.5
Figure 5-1. XRR-profiles of ALD-TaN as a function of TBTDET exposure time for
films deposited at 300 C for 80 cycles, and 10 sec purges and NH3 exposure




































Figure 5-2. Growth rate of ALD-TaN as a function of TBTDET exposure time deposited
at 300 C, and 10 sec purges and NH3 exposure


4 6 8 10 12 14 16 18
Figure 5-3. Growth rate of ALD-TaN as a function of TBTDET exposure time deposited
at 250 C, and 10 sec purges and NH3 exposure


Growth rate (Alcycle)



/
/
/
/

*Half Monolayer (2.45 A) -








TBTDET Exposure time (sec)
s<


Growth Rate (A/cycle)
$
$
/
/
/
/
/
/
I

/

Half Monolayer (2.45 A)y o
', ..4.. ..,. ... .,..,. .. rT ".*"'."........... ..'". "... >. ......*......*... ... *.. *....... ... *........
r
s

'I
T
s
s

STBTDET Exposure Time









5.3.2 Process Temperature Window


The thickness of each ALD-TaN film was measured as a function of growth

temperature to verify the process temperature window for this precursor combination. All

depositions were performed at the self-saturation growth condition identified in the

previous section. TBTDET (9 sec, 20 sccm), N2 purge (10 sec, 200 sccm) and NH3 (10

sec, 20 sccm) pulse time and flow rate were fixed with only the growth temperature

varied in the range 200 to 400 C for 80 cycles. Figure 5-4 shows the XRR profiles of the

TaN films grown as a function of temperature, while Figure 5-5 plots the thickness

extracted from the profile for each run.

The film thickness increased with increasing temperature below 200 C. Precursor

chemisorption or reaction to form the adsorbing intermediate species are thermally

activated processes. Therefore, adsorption of one of the surface species is likely

kinetically limited to give incomplete adsorption in this low temperature region. The

same trend has been reported for ALD of several metals and metal nitrides. For instance,

the deposition rate decreased for substrate temperature below 150 C for W ALD using

WF6 [KlaOO], and a similar decrease was reported for ALD of Ta3N5 using TaCl5 and

NH3 below 300 C [Rit99].

The film thickness remained constant (200 + 10 A) in the approximate temperature

range 200 to 300 C, suggesting this is the ALD process temperature window for

TBTDET and NH3 at these conditions. This thickness is consistent with the expected

value of 196 A, which represents 12 ML/cycle thickness for 80 cycles of TaN (horizontal

line in Figure 5-4). When ALD is operated at a temperature that provides sufficient

thermal energy for chemisorption to saturate, the growth rate remains constant. The









process temperature window allows precise control on the reproducibility of film

thickness and uniformity compared to CVD since the saturation condition is reproducible.

The ALD temperature window for this precursor combination at these conditions is

sufficiently wide to allow good controllability and in a reasonably low temperature range

to contribute minimally to the thermal budget.

Above 300 C, the film thickness increased with increasing temperature, indicating

that adsorption was not self-limiting. This result is often attributed to additional reaction

that changes the adsorption process to a CVD-like process. It is not known if this

involves the TBTDET or ammonia exposure step, although the onset of thermal

decomposition of TBTDET has been reported to occur above 300 C [Str04].












250 C (191 A)







20
0.5 1.5 2.5 3.5
Figure 5-4. XRR profiles of ALD-TaN as a function of growth temperature deposited
with 9 sec exposure of TBTDET and 80 cycles, and 10 sec purges and NH3
exposure










Thickness (A)

300 ,
/
270 /
/
240 /

210

Half Monolayer GR thickness
180 (2.45 A X 80 cycles = 196.4 A)

150 Temperature (oC)
100 200 300 400
Figure 5-5.Thickness of ALD-TaN films as a function of growth temperature deposited
with 9 sec exposure of TBTDET and 80 cycles, and 10 sec purges and NH3
exposure

5.3.3 Growth Characteristic of TaN ALD

The dependency of film thickness on number of cycles was examined by growing a

series of films at ALD growth conditions (9 sec TBTDET pulse time at 300 C) over the

range of cycle number 5 to 160. It is expected that a linear dependence exists between the

ALD film thickness and the number of cycles. The adsorption surface chemistry,

however, between the initial Si substrate and the stabilized TaN during the later stage of

ALD can be significantly different. Thus the substrate surface preparation, impurities in

the gas phase that react with the surface, and the selection of initial reactant can affect the

initial adsorption step. Kinetic modeling and experimental observation of the initial and

stabilized stages of TaN and TiN ALD have been reported [LimOO, Lim01, Sat02a].

These studies indicate that there exists a transient region, where the deposition rate

increases toward the constant value and the growth rate remains constant after the

transition region.









Figure 5-6 shows the XRR profiles of ALD-TaN as a function of cycle number

grown at 300 C with 9 sec TBTDET exposure time. The film thickness was obtained by

fitting the measured reflectivity using Wingixa software. Figure 5-7 summarizes the

thickness of a series of ALD-TaN films that were grown varying the cycle number. In

examining the data above 15 cycles, the growth rate remained constant (2.5 +/- 0.2

A/cycle) demonstrating the expected linear dependency of thickness on number of cycles.

The measured thickness of films grown for less than 15 cycles, however, are slightly

below the constant value of 2.5 A/cycle (solid line in Figure 5-7) and gradually increase

toward this value. It is noted that in these experiments the bare substrate was first

exposed to TBTDET. The existence of this transient region is not surprising since the

adsorption characteristics of TBTDET on bare Si or SiOx/Si are expected to be different

than those on TaN that had been exposed to NH3 and then purged. Furthermore, the

crystal structure of Si (diamond) and TaN differ as well as the lattice constants (Si: 5.43

A, TaN: 4.90 A). Thus self-limiting adsorption likely did not occur during the first few

cycles, but once the surface was fully covered and presumably a few layers thick, relaxed

TaN replaced Si as the growth surface to allow repeatable, self-limiting adsorption. For

W ALD using WF6 and Si2H6 on SiO2 substrate, -10 cycles of nucleation were needed

before a linear relationship in thickness was achieved [Ela01 ]. An even longer incubation

period (up to 40 cycles) has been reported for TiN ALD using TiCl4 and NH3 precursors

[Bey02a].






74







S30 141.2 A
30 -99.7 A





0.0 0.5 1.0 1.5 2.0 2.5 3.0 20




Figure 5-6. XRR profiles of ALD-TaN as a function of cycle number deposited at 9 sec
TBTDET exposure, 10 sec purges and NH3 exposure at 300 C

450
Thickness (A)
400 .0

3508
300 2-.- -5 A


150 0 o
10 38.0 A












0.0 0. 0 5 10 15 20 25 30 35


0 's # of cycles
0 20 40 60 80 100 120 140 160 180
Figure 5-7. Thickness of ALD-TaN as a function of cycle numbers deposited at 9 sec
TBTDET exposure, 10 sec purges and NH3 exposure at 300 C
TBTDET exposure, 10 sec purges and NH3 exposure at 300 C









During growth in this transient region, a surface discontinuity exists between the

starting substrate surface and the saturated surface of the relaxed TaN, leading to a non-

linear relationship between the film thickness and number of cycles. It is possible that

competitive adsorption by the reactants leads to preferential growth on areas already

covered by TaN. Therefore, three dimensional (3D) growth may occur in the initial stage.

Previous studies on ALD of high dielectric constant materials have shown 3D growth

during the transient region [Gre02]. The possibility of this occurring in the TaN films

grown in this study was examined with AFM images. Figure 5-8 shows images of the

surface of ALD TaN films for several values of low cycle number (5, 10 and 15 cycles).

These scans show a rough surface with high hills for the films grown with 5 and 10

cycles, indicating 3D-growth mode prevails during the nucleation stage. The surface is

seen to become considerably smoother after 15 cycles. Note that 15 cycles was the

starting point of the linear relationship between the film thickness and number of cycles,

which was considered to be the stage of complete coverage of TaN on Si substrate.
" cycles 10 cycles 15 cycles








'50
Figure 5-8. AFM scans (z: 3.0 nm / div) of the surface of TaN deposited at 9 sec
TBTDET exposure and 300 C using 5, 10 and 15 cycles

5.3.4 Diffusion Barrier Test of Ultra-thin TaN


Cu of 100 nm thickness was sputter-deposited on the ALD-TaN layers of 7 to 100

A thickness to verify the minimum thickness that could prevent Cu diffusion. The test









structures were then annealed at 500 C for 30 min in N2. To evaluate the extent if any of

Cu transport across the barrier, each film was non-destructively evaluated with XRD for

evidence of Cu silicide formation. The diffraction patterns are shown in Figure 5-9. The

7 and 16 A ALD-TaN films showed the Cui5Si4 (220) reflection, indicating that barrier

failure occurred in the Cu/TaN/Si structure, while the other films (38 A and greater)

showed strong peaks assigned to TaN and Cu but no reflections assigned to copper

silicide.

A cross-sectional TEM (Figure 5-10) sample was prepared on the 38 A/ 15 cycle

film to confirm the thickness data extracted from XRR. The thickness from TEM (38 A)

was a slightly greater than the calculated thickness from XRR (36 A), which is

reasonably well matched and within the estimated 5 % error. This result indicates that

ultra-thin ALD-TaN is a good candidate diffusion barrier material for the 38 nm node

requirement of the roadmap, which specifies a 3.5 nm thick diffusion barrier in the year

2013 [Tra04]. Similar tests with ALD-TaN reported that 10 to 60 A thick barriers could

block Cu diffusion [Bas03, Str02].

TaN peaks could be observed with 38 and 47 A thick samples after the annealing.

Obviously, the higher intensity of TaN peaks is expected since the thickness increased.

The destructive etch pit density evaluation (etching bare Si in a Secco solution after

TaN/Cu removal by wet etching) was next applied to the annealed samples. This method,

which is a more sensitive method for the detection of Cu diffusion than XRD, also

revealed that 16 A ALD-TaN film was a poor barrier as evidenced by the high density of

etch pits (Figure 5-11). On the other hand, TaN the 38 A sample exhibited no etch pits








on the bared Si surface (Figure 5-11), indicating the critical thickness of ALD-TaN as a

Cu diffusion barrier exists between 16 and 38 A.

a.u. h-TaN
(100) Cu h T,,
I c-TaN (111) h-TaN
(11)(10 20 cy ,cI,



15 cycle
(38 A)

CuisSi4 (220) Cu15Si4 10 cycle
(16 A)

5 cycle
b (7 A)

20 30 40 50 e 60
Figure 5-9. XRD patterns of Cu/TaN/Si annealed at 500 C for 30 min


Figure 5-10.Cross-sectional TEM micrograph of Cu/TaN/Si structure. TaN grown by
ALD for 15 cycles at 9 sec TBTDET exposure time and 300 C























Figure 5-11. SEM images of Si surface after annealing Cu/TaN/Si at 500 C for 30 min
followed by the removal of Cu and TaN, and Secco etching

The diffusivity (D) of Cu in TaN films was estimated using the critical thickness

(38 A) and the characteristic diffusion length (2 D t ), which yielded a D value of

2 x 10- 17cr/s at 500 'C. This value is similar to the reported Cu diffusivity in single crystal

TaN, ranging from 10-18 to 10-16 Crm/s in the temperature range 600 to 700 C [Wan02].

The estimated Cu diffusivity in ALD-TaN at 500 C is surprisingly similar in the reported

diffusivity in TaN at 600 C that was obtained on single crystal sputter-deposited

samples. The structure of ALD-TaN is amorphous with some distribution of nano-

crystallites in the films, which might provide a fast Cu diffusion pathway, as compared to

the single crystal structure of TaN. The diffusivity of Cu in TaNO.62 films (-10-14Crm/s) at

500 C [LinOO] was much higher than that in ALD-TaN films, which is consistent with

the previous observation that high nitrogen content in the film is helpful in retarding Cu

diffusion in TaN films.

5.4 Conclusions

TaN was successfully deposited by ALD using alternating exposure of a Si wafer to

TBTDET and NH3. The XRR thickness measurements as a function of exposure time

confirmed that the adsorption of TBTDET was self-limiting as the deposition rate (2.6


10 cycle (16 A)









4
1"nkv .... .......


15 cycle (38 A)










Vd$v XV.W11 Ij W1 I /









A/cycle) was independent of the TBTDET exposure time in the relatively narrow time

range 9 to 11 sec. The process temperature window for ALD with this precursor

combination was also identified. The thickness was constant (200 +/- 10 A) in the

temperature range 200 to 300 C deposited with self-limiting growth conditions and 80

cycles. This temperature window for ALD mode was sufficiently wide to allow good

controllability and reasonably low to contribute minimally to the thermal budget.

A series of films was grown with variable cycle number from 5 to 100 on Si (100)

in ALD growth mode. The film thickness with increasing cycle number shows non-linear

behavior when the number of cycles is less than 15, suggesting the adsorption

characteristics of TBTDET on bare Si or SiOx/Si are different than those on TaN. After

about 15 cycles the thickness varied linearly with cycle number, consistent with a growth

rate of 2.5 0.2 A/cycle and 12 monolayer growth per cycle. Cu diffusion barrier efficacy

was evaluated by searching for Cu silicides using XRD and measuring the etch pit

density on Secco-etched bared Si surfaces. The results show that ultra-thin ALD-TaN as

thin as 38 A is a good candidate for a diffusion barrier material, which satisfies the year

2013 roadmap 38 nm feature size node (3.5 nm thick diffusion barrier).














CHAPTER 6
METAL ORGANIC ATOMIC LAYER DEPOSITION OF Ta-Al-N USING TBTDET,
TEA AND NH3

6.1 Introduction

Among the transition metals and their nitrides, binary compounds such as TiN,

TaN, and WN have been intensively investigated for Cu diffusion barriers. The drawback

of binary nitrides and especially TiN, however, are their tendency to deposit with a

columnar polycrystalline microstructure, which is an unfavorable configuration for a

diffusion barrier [Nic95]. This grain structure creates boundaries that cut across the film

and offer fast diffusion paths. To retard the formation of columnar grains, ternary

materials such as (Ti,Ta,W)-Si-N and W-(B,C)-N have been evaluated [Kim05]. The

added degrees of freedom are expected to promote formation of an amorphous structure

and its retention after subsequent high temperature processing.

In this study, Ta-Al-N films were deposited by thermal ALD using TBTDET, TEA,

and ammonia. TMA (trimethyl aluminum) has already been employed as an effective

reducing agent in the ALD of TiN films that were deposited with an inorganic precursor

(TiCl5) [Jup01]. The resulting films exhibited relatively low resistivity and low chlorine

content (<4 at %). Furthermore, carbon contamination in the films from TMA was not

harmful to the diffusion barrier properties [Eiz94]. Aluminum incorporation also made

the film structure nanocrystalline or amorphous and thereby improved the barrier

properties. Similarly, Ta-Al-N was deposited by ALD using TaCl5 or TaBr5 and NH3 as

precursors and TMA as a reducing agent [Ale01]. Metal organic ALD of Ti-Al-N using









TDMAT and DMAH-EPP (dimethylaluminum hydride-ethyl-piperidine) as the titanium

and aluminum precursors was also reported [Koo01]. The film structure remained

amorphous even after high temperature annealing at 900 C and the Ti and Al content

could be changed simply by controlling the number of TDMAT-NH3 and DMAH-EPP

cycles. The Ti-Al-N films showed higher Cu diffusion failure temperature than ALD

TiN, as determined by XRD and observation of the etch-pit density in bare Si substrates

treated with a Secco etch [Kim02b]. On the other hand, another study showed that Ti-Al-

N did not exhibit superior results compared to ALD TiN, as judged by barrier test results

using XRD, the etch-pit test, and resistivity measurements [Jup01]. Although there have

been no reports of using TEA (triethyl aluminum) as an Al reagent in deposition of (Ti,

Ta)-Al-N, TEA is expected to lower the carbon incorporation more than using TMA as

evidenced by deposition studies of AlxGi-xAs and other III-V systems. In this work, the

effect of Al incorporation in TaN using TEA on diffusion barrier performance of Ta-Al-

N films along with the microstructure and density are examined.

Another focus of this work is to understand the mechanism of aluminum

incorporation in Ta-Al-N films by investigating the film thickness, chemical composition,

and bonding states when varying the exposure time of TEA and the pulse sequence (i.e.,

Ta-Al-N (Sequence A), Al-Ta-N (Sequence B) and Ta-N-Al-N (Sequence

C)). For (Ti, Ta)-Si-N [MinOO, Rei96], a change in the precursor exposure sequence

induced a dramatic change in the film properties including Si content, resistivity, and

growth rate. These changes were attributed to strong SiH4 adsorption on the nitride

surface that blocks adsorption of the metal precursor. A lower growth rate was observed

when the SiH4-metal precursor-NH3 sequence was used, compared to the sequence metal









precursor- SiH4- NH3 [Rei96]. For Ta-Al-N using TaCl5, the deposition rate and

chemical composition also changed depending on the sequence of the precursors. A

higher growth rate and lower carbon and oxygen impurity levels were observed with the

TaCl5-TMA-NH3 sequence rather than the TMA-TaC15-NH3 sequence [Ale01]. The same

tendency is expected to occur with the precursor combination used in this study since

TEA reacts with ammonia and forms AIN (Sequence A). TBTDET and TEA are not

known to form any compound, which should lead to more oxygen incorporation in Al-

Ta-N films (Sequence B) due to TEA's high affinity to oxygen. Therefore, the growth

rate is expected to increase with less oxygen incorporation when sequence A is used

compared to sequence B. Sequence C is anticipated to produce film properties similar to

those produced using sequence A since TBTDET can form TaN without added NH3.

6.2 Experimental Details

ALD of Ta-Al-N was performed on standard (100) silicon substrates in a

customized cold-wall, vertical ALD reactor. Pneumatically actuated valves, controlled

with a LaBView-based program interfaced to digital input/output interface board,

sequentially switched reactant and purge streams between the reactor and by-pass line.

TBTDET (supplied by Alfa Aesar, MA, vapor pressure of 0.1 Torr at 90 C), contained in

a stainless steel bubbler, was maintained a constant temperature at a set point of 80 C.

The MO precursor was delivered to the reactor using a N2 carrier. TEA (supplied by

Epichem, vapor pressure 0.02 Torr at 20 C,) was used as the aluminum source and N2 as

the carrier gas. Ta-Al-N films were deposited at 300 C at a typical growth pressure of

1.5 Torr. In sequence A, the standard deposition cycle begins with an exposure of

TBTDET for 9 sec using 20 sccm of N2 carrier gas. Then the MO is switched to vent and

a nitrogen flow of 200 sccm is delivered to the reactor for 10 sec to ensure complete









removal of residual TBTDET precursor as well as volatile byproducts. After the purge

step, TEA is next introduced for 2 to 8 sec using a carrier gas flow rate of 20 sccm,

followed by another purge step. Finally, NH3 at a flow 20 sccm is delivered to the reactor

for 10 sec and followed by a final purge before the cycle is repeated. Sequences B and C

were also operated with the same reactant flow rates and exposure times with the

exception, of course, of reactant sequence.

The film thickness was evaluated from the XRR profiles using the WinGixa

program assuming that a very thin Si02 layer (~ 20 A) exists at the interface between Si

and TaAIN, and that C (-20 atom %), and 0 (-15 atom %) impurities reside in this top

layer of the Ta-Al-N film. The stoichiometry of each film was based on the chemical

compositions estimated from AES measurements after 1 min sputtering. It is noted that

no standard was available for calibrating the AES measurements. The thickness data

from the simulation for each sample were confirmed by cross-sectional SEM

measurements. The microstructure was analyzed using XRD and TEM. The film

chemical composition and atomic bonding states were probed by AES and XPS

measurements. To compare the Cu diffusion barrier properties of TaN and Ta-Al-N films,

a 100 nm thick Cu film was sputtered deposited on the ALD TaN or Ta-Al-N film. The

samples were then annealed at 500 C for 30 or 45 min in N2 atmosphere and tested for

Cu transport across the TaN or Ta-Al-N barrier.

6.3 Results and Discussion

6.3.1 Growth Rate and Chemical Composition

The conditions that gave self-saturated growth of TaN ALD and reported in

Chapter 5 were used to estimate the growth conditions for the ternary Ta-Al-N material.

Specifically, the conditions TBTDET flow rate of 20 sccm with a pulse time 9 to 11 sec









at 300 C with a fixed purge/NH3 exposure time and flow rate (10 sec, 200 sccm/10 sec,

20 sccm) produced ALD growth characteristics. For Ta-Al-N ALD these same pulse

times and flow rates were used to deliver TBTDET, NH3, and the purge gas, while the

TEA flow rate was 20 sccm and the pulse time was varied from 2 to 8 sec.

As a preliminary study, the conditions that would yield self-limiting adsorption of

TEA were sought by observing the growth rate with increasing exposure time of TEA.

No report on the self-limiting growth of AIN when using TEA is believed to exist.

Trimethyl aluminum (TMA), which is structurally similar to TEA, was reported to be

self-limiting on Si substrates at temperatures below 650 K for doses greater than about

100 L (Langmuirs) [May91]. The same group also showed that NH3 exhibits site-

selective reaction with Al-containing surface species at temperatures greater than 550 K.

Their conclusion of self-limiting growth from TMA is based on quantitative estimates of

the saturation coverage made from the integrated XPS peak intensities, elemental

sensitivity factors, and the inelastic mean free paths for photoelectrons from Al-

containing films [May91]. A similar study on alumina substrates using TMA and NH3 at

600K showed that TMA adsorption/reaction on alumina is a self-limiting process, as

judged from the lack of shift in the C (Is) binding energy. The shift in the C binding

energy would be consistent with continuous deposition, since the formation of A14C3 by

the incomplete pyrolysis and reaction of TMA will shift the C binding energy [Liu95]. In

contrast to these reports, Ruhela et al. observed no saturation of the growth rate of AIN

on Si was observed with increasing TMA pulse time at growth temperature in the range

325 to 425 C [Ruh96]. This behavior was attributed to the thermal decomposition of

TMA in this temperature range.









Figure 6-1 depicts the XRR profiles of AIN films deposited with various TEA pulse

times at 300 C. The estimated growth rate per cycle is shown in Figure 6-2 along with

the lattice constant of h-AIN (4.9 A). It is seen the growth rate increased linearly with

TEA pulse time and thus the use of TEA did not exhibit self-limiting behavior at 300 C.

Thermal decomposition of TEA is believed to be the reason for non-self-limiting growth,

as pyrolysis studies on TEA suggest that the thermal decomposition of TEA starts around

200 C [Tis90]. Thermal decomposition of the precursor removes ligands to allow

continued reaction. Apparently, thermally decomposed TEA precursor allows continuous

adsorption of aluminum atoms and causes CVD-like deposition.

a.u.
















20
0.3 0.8 1.3 1.8 2.3 2.8 3.3 3.8
Figure 6-1. XRR profiles of ALD AIN as a function of TEA exposure time deposited at

300 C and 10 sec purges and NH3 exposure for 120 cycles
3 00 'C and 10 sec purges and NH3 exposure for 120 cycles









8 Growth rate (A/cycle)

7

6
h-AIN lattice constant (4.9 A)
5 ----------------------- --------------------

4

3

2
TEA pulse time (s)
I I I I I I I

1 2 3 4 5 6 7 8 9

Figure 6-2. Growth rate of AIN as a function of TEA exposure time deposited at 300 C
and 10 sec purges and NH3 exposure for 120 cycles

Figure 6-3 shows the XRR profiles for films of Ta-Al-N deposited with sequence A

and denoted by Ta-AI-NA. This set of films was grown by varying the exposure time of

TEA source from 2 to 8 sec and using 120 cycles. The growth rate and chemical

composition of Ta-AI-NA films as a function of TEA exposure time are plotted in Figure

6-4. As expected from the AIN result described previously, the growth rate and atomic

concentration of Al in Ta-AI-NA films increased with longer TEA pulse time, which is

consistent with TEA not incorporating in Ta-AI-NA films in a self-limiting way. The

nitrogen concentration remained almost constant in the range 27 to 30 at % and the Ta

concentration decreased from 46 (2 sec) to 33 at % (8 sec), as the TEA exposure time

increased. Considering that TaN can be deposited from a single source without NH3 and

that AIN can be formed with TEA and NH3 in sequence A, it is reasonable to use the A1N

thickness data from Figure 6-2 to estimate how much of the AIN thickness contributes to