Chapter 5. Microstructural Evolution in Thin Films

Chapter 5. Microstructural Evolution in Thin Films ofElectronic Materials

Academic and Research Staff

Professor Carl V. Thompson, Dr. Craig Ballentine, Dr. Paul D. Bristowe, Dr. En Ma,' Dr. John Melngailis,Dr. Robert C. O'Handley

Visiting Scientists and Research Affiliates

Dr. Sergei Bogomolov, 2 Dr. Harold J. Frost, 3 HideoWalton 6

Graduate Students

Roland Carel, Steven S. Cooperman, Anthony D. DellaE. Inglefield, Young-Chang Joo, Harold Kahn, Brett D.Ro

Miura,4 Takauki Uda, 4 Dr. A. Wagner,5 Derek T.

Ratta, Andrew D. Dubner, Jerrold A. Floro, HeatherKnowlton, Ya-Chin Liu, Hai P. Longworth, Jaesang

Technical and Support Staff

Peter C. Heron

5.1 Electromigration andMicrostructureElectric-current-induced atomic self-diffusion, orelectromigration, causes open and short circuitinterconnect failures in integrated circuits. Electro-migration occurs primarily by diffusion along grainboundaries, and interconnect failure initiates atgrain boundaries, so that interconnect reliability isstrongly dependent on the grain sizes and orien-tations in polycrystalline interconnects. Therelationships between these microstructural char-acteristics and interconnect reliability have becomemore complex as interconnect dimensions (widthsand thicknesses) have become comparable tograin sizes. We have studied the effects of micro-structure on the reliability of submicron and largerinterconnects as well as vias and contacts. Wehave done this by using a variety of techniques forcontrolling grain sizes and grain size distributionsin thin films. We are also developing models formicrostructural evolution in order to develop accu-

rate models forinterconnects.

electromigration in submicron

5.1.1 Control of Microstructures inAluminum Films

Sponsor

Joint Services Electronics ProgramContract DAAL03-89-C-0001

Project Staff

Hai P. Longworth, Professor Carl V. Thompson

In previous years, we reported the use ofprecipitate-induced abnormal grain growth toproduce grains as large as 1 mm in 1 ym thickfilms. In the past year, we completed an extensivestudy in which we have established the require-ments for precipitate-induced abnormal graingrowth in thin films and have demonstrated similar

1 Presently at the University of Michigan, Ann Arbor, Michigan.

2 Ural Polytechnical Institute, Sverdlovsk, Russia.

3 Dartmouth College, Hanover, New Hampshire.

4 Hitachi Limited, Tokyo, Japan.

5 IBM Corporation, Yorktown Heights, New York.

6 Formerly of Dartmouth College, currently with Digital Equipment Corporation, Hudson, Massachusetts.

Chapter 5. Microstructural Evolution in Thin Films

phenomena in Al-Cu-Cr, Al-Ag-Cr, Al-Mn-Cr, andAI-W alloys. We have found that to optimize theprecipitate-induced abnormal grain growthprocess, the alloy additions should initially bepresent as pure layers in the middle of the Al film.In alloys containing Cr, the Cr serves to aid in thenucleation of precipitates, leading to localization ofthe precipitates in the middle of the film, andleading to a larger number of smaller, more closelyspaced precipitates. Through controlled abnormalgrain growth, the microstructure of thin films canbe controllably varied over broad ranges.

5.1.2 Reliability of Vias

Sponsor

Semiconductor Research Corporation

Project Staff

Harold Kahn, Professor Carl V. Thompson

Over the past year, we have developed a via teststructure that allows statistical characterization ofvia reliability by stressing multiple vias in parallel.We have used this process to characterize the reli-ability of vias with W plugs and have shown thatunder normal test conditions, joule heating of theplugs contributes to accelerated open failures of Alinterconnects. We are developing simple thermalmodels to evaluate the effect of joule heating atservice conditions. We have also demonstratedthat the grain size in the Al interconnect greatlyaffects via reliabilities and also the statistics offailure. Large grains lead to bimodal failure dis-tributions. We are developing tools for statisticalanalysis of via reliability.

5.1.3 Modeling of Microstructuresin Interconnects

Sponsors

Semiconductor Research CorporationNational Science Foundation

Project Staff

Derek T. Walton, Dr. Harold J. Frost, ProfessorCarl V. Thompson

Over the years, we have developed and testedmodels for crystal nucleation and growth to formcontinuous polycrystalline films. We have alsomodeled grain growth in continuous polycry-stalline films. In the past year we have modifiedthese models to investigate the evolution of themicrostructures of polycrystalline interconnects to"bamboo" structures, as illustrated in figure 1. Wefind that the approach to a fully bamboo structureis exponential in time and that the rate of transfor-mation is inversely proportional to the square ofthe strip width. When the simulation is extendedto model grain boundary pinning due to groovingat grain boundary-free surface intersections, wefind that there exists a maximum strip width-to-thickness ratio (approximately 3) beyond whichthe transformation to the bamboo structure doesnot proceed to completion. In earlier work, wehave shown that reliability of nearly-bamboo andfully-bamboo lines is greatly different. Our mod-eling work is now allowing us to identify the con-ditions under which it is reasonable to expect toobtain only fully-bamboo lines.

0.0

0.1

Yr 0.2

0.3

0.5

I . I Y _'l) 1.0

S I ) 1.5

S I I I I 12.5

Figure 1. Evolution of the grain structure within astrip. The transformation to a "bamboo" structure isdominated by the growth of bamboo sections intoregions of grain clusters. The rate of this transforma-tion depends strongly on the strip width.

40 RLE Progress Report Number 134

Chapter 5. Microstructural Evolution in Thin Films

5.1.4 Electromigration in Bicrystals

Sponsors

Joint Services Electronics ProgramContract DAAL03-89-C-0001

Semiconductor Research Corporation

Project Staff

Hai P. Longworth, Young-Chang Joo, Brett D.Knowlton, Professor Carl V. Thompson

In previous work we have shown that individualgrain boundaries greatly affect the reliability of aninterconnect line. In order to understand the reli-ability of interconnects, it is therefore important tounderstand the reliability of individual grainboundaries. To do this we have developed a newtechnique that allows us to characterize the reli-ability of populations of lines with single, identicalgrain boundaries with controlled types, locations,and orientations. The technique is schematicallyshown in figure 2. We start by making NaCIbicrystals of the desired type, on which we growepitaxial bicrystal Al films which are transferredand sintered to oxidized silicon wafers. We thenuse conventional lithographic techniques topattern multiline test structures with the same con-trolled Al grain boundary running through all thelines. We can then investigate, as an example, theeffect of the boundary orientation on the reliabilityof the line, as illustrated in figure 3.

We have shown that (1) failure times of lines con-taining single identical boundaries are lognormallydistributed, (2) the medium time to failuredepends more strongly on the boundary orien-tations than the types of grain boundaries, (3) thedeviation in the time to failure has a large compo-nent not dependent on microstructure, and (4)both interfacial diffusion and grain boundary dif-fusion appear to contribute to failure in bicrystallines and probably in bamboo and near-bamboolines as well.

NaC)

a)

Al

NaCi

b)

S Soi02

Figure 2. The fabrication process of bicrystal test pat-terns. (a) NaCI bicrystal substrate made from two NaCIsingle crystals; (b) Al bicrystal thin film epitaxiallydeposited on the NaCI substrate; (c) Al film transferredto an oxidized Si wafer; and (d) patterning.

1000

100

I i I I I i I . ,I I

J = 2.5 x 106A/cm2T = 250"C

b) 0 = 85

c) 0 = 75

Sd) e 600>

1 10 50 90 99Cumulative Failure ()

Figure 3. Failure distributions of 113[100] bicrystallines as a function of 8, the angle between the currentdirection and the grain boundary plane.

d)

Chapter 5. Microstructural Evolution in Thin Films

5.2 Microstructural Evolutionin Polycrystalline FilmsPolycrystalline thin films are used in a wide varietyof applications, especially in electronics. In theseapplications the properties and performance ofpolycrystalline films depend strongly on theaverage grain size, the distribution of grain sizes,and the distribution of grain orientations. Theseare controlled by the deposition conditions and theconditions for post-deposition processing. In thepast, we have carried out modeling and exper-iments on the effects of crystal nucleation andgrowth and grain growth on the microstructure ofthin films. This has included work on semicon-ductors as well as metals, and has explored theeffects of dopants, ion bombardment, and rapidthermal annealing, as well as other processingconditions. Our current activity is focused onmodeling of grain growth in continuous and pat-terned films and experiments on epitaxial graingrowth.

5.2.1 Modeling of Grain Growth inThin Films

Sponsor

National Science Foundation

Project Staff

Jerrold A. Floro, Derek T. Walton, Roland Carel,Dr. Harold J. Frost, Professor Carl V. Thompson

We have developed a mean field model for graingrowth in thin films which accounts for the effectsof the energies of the free surfaces and thefilm/substrate interfaces of the films. Anistropiesof these energies affect the evolution of the dis-tribution of grain orientations, as well as grainsizes. With our analysis we can predict the differ-ences in the rate of grain growth and film textureevolution that result from different specific inter-face or surface energy versus orientation relation-ships. These predictions are being compared withexperimental results.

In collaboration with Dr. H.J. Frost of DartmouthCollege, we have also developed computer simu-lations of grain growth in thin films. We haveincluded in our simulations the effects of grainboundary drag due to surface grooving. This leadsto stagnation of normal grain growth at a pointwhere the grain sizes are lognormally distributedand the average grain diameter is about three timesthe film thickness. This simulation result closely

matches experimental results in a wide variety ofsystems. We can also include the effects ofsurface and interface energy anisotropy and deter-mine the conditions which lead to surface-energy-driven abnormal or secondary grain growth. Thesesimulations allow investigation of the effect of thesurface and interface energy anisotropy on theevolution of texture in thin films. Our mean fieldmodels allow us to model and predict the behaviorof large populations of grains, while our computersimulations allow us to investigate the behavior ofsmaller populations of grains as well as specificgrain configurations. We have also recently modi-fied our computer simulations to model graingrowth in patterned films, as described in Section5.1.3.

5.2.2 Epitaxial Grain Growth

Sponsor

National Science Foundation

Project Staff

Jerrold A. Floro, Roland Carel, Dr. Paul D.Bristowe, Professor Carl V. Thompson

When polycrystalline films on single crystal sub-strates undergo grain growth, interface-energy-anisotropy leads to the preferred growth ofepitaxially oriented grains. We have demonstratedthis phenomenon in metallic films deposited onsingle crystals of NaCI, mica, and MgO. We havealso demonstrated epitaxial grain growth in poly-crystalline Ag films on single crystals of Ni. Wehave chosen to study the latter experimentalsystem because calculations of the structure andenergies of Ag/Ni interfaces for a number of dif-ferent Ag/Ni relative crystallographic orientationsexist. These calculations have been shown to beconsistent with results from rotating sphere exper-iments in which very small single crystal Agspheres rotate into low energy orientations whenannealed on single crystal Ni substrates. As-deposited epitaxial films of Ag on Ni are alsoreadily attainable. We have found that the epi-taxial orientations obtained are different in each ofthese three types of experiments (as-depositedepitaxy, rotating spheres, and epitaxial graingrowth). These results suggest that the orien-tation obtained by conventional epitaxy may bemetastable once a film is continuous. In exploringthis issue, we are using embedded atom calcula-tions to calculate interface and surface energiesand using the results from these calculations in ourgrain growth models and simulations.

42 RLE Progress Report Number 134

5.3 Magnetic Properties ofHeteroepitaxial Thin Films

Sponsors

National Science FoundationU.S. Army Research Office

Project Staff

Heather E. Inglefield, Dr. Sergei Bogomolov, Dr.Craig Ballentine, Dr. Robert C. O'Handley, Pro-fessor Carl V. Thompson

In this new program we are investigating the effectof strain and misfit dislocations on the magneticproperties of very thin (<10 nm) heteroepitaxialmetallic films. In particular, we are studying Nifilms deposited on Ni-Cu alloy films. In this andrelated systems, when overlayers are very thin,they will grow pseudomorphically, adopting thelattice parameter of the substrate. However, as theoverlayer thickens, the strain energy will becomesufficient to drive deformation and the formation ofmisfit dislocations. Grids of perpendicular arraysof misfit dislocations often result. These structuralchanges are expected to have profound effects onthe magnetic properties of the films. We are pre-paring to use measurements of the magneto-opticKerr effect (MOKE) and low energy electron spinpolarization analysis (LEESPA) to analyze changesin magnetic properties during film deposition. Wewill correlate changes in magnetic properties withchanges in structure as determined using in situreflection high energy electron diffraction(RHEED), as well as ex situ x-ray diffraction anal-ysis and transmission electron microscopy.

In addition to using in situ magnetic characteriza-tion of films to probe and monitor structuralchanges, we will also investigate the effects ofmisfit dislocations on the large-scale and small-scale magnetic properties of films. As part of thisanalysis we will use scanning electron microscopyand polarization analysis (SEMPA) to image mag-netic domains with a spatial resolution smallerthan the dimensions of the misfit dislocation grid.

5.4 Interface Reactions inMultilayer Thin Films

Sponsors

Hitachi LimitedIBM Corporation

Chapter 5. Microstructural Evolution in Thin Films

Project Staff

Dr. En Ma, Hideo Miura, Takauki Uda, ProfessorCarl V. Thompson

When two materials are brought into contact andheated, interdiffusion will lead to compound for-mation. This phenomena is important throughouttechnology, but is especially important in elec-tronics in understanding unwanted reactionsbetween films and substrates, as well as desiredreactions that form silicides and diffusion barriers.Despite the importance of interface reactions, theinitial processes that control which compoundphase will form during a reaction (phase selection)is not well understood. We have investigatedphase selection during interface reactions by usingmultilayer films in which we have alternatelydeposited pure layers with layer thicknesses as lowas a few nanometers. By using multilayer films,we can focus on the earliest stages of interfacereactions because these early reactions consumeall or most of the volume of the overall films. Thisallows the use of tools such as transmission elec-tron microscopy, x-ray analysis, and differentialscanning calorimetry (DSC) to study formation ofcompound layers that are only a few nanometersthick. DSC is especially useful in studying thekinetics as well as the thermodynamics of theseprocesses. We have used these techniques tostudy reactions in Ni/Si, Ti/Si, V/Si, Co/Si, andNi/Al multilayer films. We have shown that inter-diffusion always precedes nucleation of a com-pound and have argued that interdiffusion is infact a necessary precursor to compound formation.Phase selection is therefore constrained by therates of self-diffusion and interdiffusion. Thisinsight has allowed us to explain our frequentobservations that the initial phase to form is oftenmetastable. We are also now able to make pred-ictions about phase selection sequences in inter-face reactions.

5.5 Focused Ion Beam InducedChemical Vapor Deposition

Sponsor

U.S. Army Research Office

Project Staff

Anthony D. Della Ratta, Jaesang Ro, Andrew D.Dubner, Dr. A. Wagner, Dr. John Melngailis, Pro-fessor Carl V. Thompson

A focused ion beam (FIB) can be used to causelocal deposition of a film to write lines for x-raymask repair or circuit restructuring. Ions are accel-erated to the substrate where each ion can causetens of decomposition events in a chemical vapor

Chapter 5. Microstructural Evolution in Thin Films

deposition process. We have carried out exper-iments focused on understanding the energytransfer mechanism which leads to FIB-inducedchemical vapor deposition. The process is sub-strate mediated and scales with the ion energy andion mass in such a way as to strongly suggest thatdecomposition is caused by collisions betweenknock-on substrate atoms and adsorbedmonolayers of unreacted molecules. It alsoappears that removal of the reactant product canbe rate limiting. It is found that film purities areimproved when films are deposited at lower ratesor higher temperatures. These insights will help inthe development of processes for deposition ofhigh purity lines with controlled widths and aspectratios.

5.6 Publications

Dubner, A.D., A. Wagner, J. Melngailis, and C.V.Thompson. "The Role of the Ion/Solid Inter-action in Ion Beam Induced Deposition ofGold." J. Appl. Phys. 70: 665 (1991).

Evans, P.V., D.A. Smith, and C.V. Thompson."Absence of Electrical Activity at High AngleGrain Boundaries in Zone- Melt- RecrystallisedSilicon-on-Insulator Films." Appl. Phys. Lett.60: 439 (1992).

Floro, J.A., and C.V. Thompson. "Epitaxial GrainGrowth and Orientation Metastability in Heter-oepitaxial Thin Films." Materials ResearchSociety Symposium Proceedings 187: 274(1991).

Frost, H.J., C.V. Thompson, and D.T. Walton."Simulation of Thin Film Grain Structures: II.Abnormal Grain Growth." Acta Metall. Mat.Forthcoming.

Frost, H.J., C.V. Thompson, and D.T. Walton."Abnormal Grain Growth in Thin Films Due toAnisotropy of Free-Surface Energies." Pro-ceedings of the First International Conferenceon Grain Growth in Polycrystalline Materials,Rome, 1991. Forthcoming.

Jiran, E., and C.V. Thompson. "Capillary Instabili-ties in Thin Continuous Films." Thin SolidFilms. Forthcoming.

Kahn, H., and C.V. Thompson. "A Statistical Char-acterization of Electromigration-Induced OpenFailures in 2-Level Metal Structures." MaterialsResearch Society Symposium Proceedings 225:15 (1991).

Kahn, H., and C.V. Thompson. "The Effect ofApplied Mechanical Stress on the Electro-migration Failure Times of Aluminum Intercon-nects." App. Phys. Lett. 56: 1308 (1991).

Longworth, H.P., and C.V. Thompson. "AbnormalGrain Growth in Aluminum Alloy Thin Films."J. Appl. Phys. 69: 3929 (1991).

Longworth, H.P., and C.V. Thompson. "An Exper-imental Study of Electromigration in BicrystalAl Lines." Appl. Phys. Lett. Forthcoming.

Ma, E., and C.V. Thompson. "Nucleation andGrowth During Reactions in Multilayer AI/NiFilms: The Early Stages of AI3Ni Formation." J.App. Phys. 69: 2211 (1991).

Ma, E., L.A. Clevenger, C.V. Thompson, and K.N.Tu. "Kinetic and Thermodynamic Aspects ofPhase Evolution." Materials Research SocietySymposium Proceedings 187: 83 (1991).

Ma, E., C.V. Thompson, and L.A. Clevenger. "ACalorimetric Study of the Kinetics of AI3 NiNucleation and Growth During Reactions inAI/Ni Thin Films." Materials Research SocietySymposium Proceedings 157: 455 (1991).

Miura, H., E. Ma, and C.V. Thompson. "InitialSequence and Kinetics of Silicide Formation inthe Cobalt/Amorphous-Silicon Multilayer ThinFilms." J. Appl. Phys. 70: 4287 (1991).

Thompson, C.V., J.Y. Tsao, and D.J. Srolovitz,editors. "Evolution of Thin Film and SurfaceMicrostructure." Materials Research SocietySymposium Proceedings 202 (1991).

Thompson, C.V. "On the Role of Diffusion inPhase Selection During Reactions atInterfaces." J. Mater. Res. 7: 367 (1992).

Thompson, C.V., L.A. Clevenger, R. DeAvillez, E.Ma, and H. Miura. "Kinetics and Thermo-dynamics of Amorphous Silicide Formation inMetal/Amorphous Multilayer Thin Films."(Invited Paper) Materials Research SocietySymposium Proceedings 187: 61 (1991).

Thompson, C.V. "Experimental and TheoreticalAspects of Grain Growth in Thin Films."(Invited Paper) Proceedings of the First Inter-national Conference on Grain Growth in Poly-crystalline Materials, Rome, 1991.Forthcoming.

Walton, D.T., H.J. Frost, and C.V. Thompson."The Development of Bamboo and Near-Bamboo Microstructures in Thin Film Strips."Appl. Phys. Lett. Forthcoming.

44 RLE Progress Report Number 134

Walton, D.T., H.J. Frost, and C.V. Thompson."Computer Simulation of Grain Growth inThin-Film Interconnect Lines." MaterialsResearch Society Symposium Proceedings 225:219 (1991).

Walton, D.T., H.J. Frost, and C.V. Thompson."Modeling of Grain Growth in Thin FilmStrips." Proceedings of the First InternationalConference on Grain Growth in PolycrystallineMaterials, Rome, 1991. Forthcoming.

5.7 Theses

Cooperman, S. Electromigration in Submicron-Wide Interconnects. S.M thesis. Dept. ofMater. Sci. and Eng., MIT, 1992.

Liu, Y. Coplanar-Spherical-Particle Coarsening inDiscontinuous Ultrathin Polycrystalline Gold

Chapter 5. Microstructural Evolution in Thin Films

Films on Amorphous Silicon-Nitride Substrates.Ph.D. diss. Dept. of Mater. Sci. and Eng., MIT,1991.

Longworth, H. Electromigration in Aluminum andAluminum Alloy Thin Films with ControlledMicrostructures. Ph.D. diss. Dept. of Mater.Sci. and Eng., MIT, 1992.

Ro, J. Microstructural Evolution During FocusedIon Beam Induced Deposition. Ph.D. diss.Co-supervised with Dr. J. Melngailis. Dept. ofMater. Sci. and Eng., MIT, 1991.

From left: Senior Research Scientist Dr. John Melngailis with Research Engineer Sergey Etchin of MIT'sMicrosystems Technology Laboratories, and Research Assistants Christian R. Musil and Henri J. Lezec

46 RLE Progress Report Number 134