Barrier Properties of Titanium Nitride Films Grown by Low
`Temperature Chemical Vapor Deposition from Titanium
`
`
`
`
`
`
`
`
`Tetraiodide
`
`
`
`
`ham, Dirk Manger, Gregory Peterson,
`Cheryl Faltermeier, Cindy Goldberg, Michael Jones, Allan Up
`Janice Lau, and Alain E. Ka
`oyeros
`»
`New York State Center for Advanced Thin Film Technology and Department of Physics, The University at Albany,
`
`
`
`
`
`
`
`
`
`
`
`State University of New York, Albany, New York 12222, USA
`
`
`
`
`
`
`
`
`Barry Arkles
`
`
`
`
`
`
`
`
`
`Gelest, Incorporated, Tullytown, Pennsylvania 19007, USA
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Aiit Paranipe“
`
`
`
`
`
`Texas Instruments, Incorporated, Dallas, Texas 75265, USA
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`ABSTRACT
`
`
`
`
`
`. Results are presented from a systematic study of the composition, texture, and electrical properties of titanium nitride
`(TiN) films and their performance as diffusion barrier in multilevel interconnect schemes of ultralarge scale integration
`(ULSI) computer chip device structures. The films were grown by low temperature (<450°C) inorganic chemical vapor
`deposrtion using titanium tetraiodide as source precursor and ammonia and hydrogen as co-reactants. The TiN films were
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`nitrogen-rich, with iodine concentrations below 2 atom percent, displayed resistivities in the range 100 to 150 p!) cm
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`depending on thickness, and exhibited excellent step coverage with better than 90% conformality in both nominal 0.45 am,
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`3:1 aspect ratio and 0.25 am, 4:1 aspect ratio contact structures. A comparison of the properties of chemical vapor
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`deposited (CVD) TiN with equivalent physical vapor deposited (PVD) TiN showed that reactivity with Al—0.5 a/o Cu alloys
`
`
`
`
`
`
`
`
`
`
`
`
`
`was equivalent in both cases. In particular, a 10% increase in the Al—Cu/TiN stack sheet resistance was observed for both
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`types of TiN after a 450°C, 30 min sinter. Similarly, the characteristics of CVD tungsten and reflow plug fills were iden—
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`tical on both types of TiN films. However, barrier performance for CVD TiN in aluminum and tungsten plug technologies
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`was superior to that of PVD TiN, as evidenced by lower contact diode leakage for CVD TiN in comparison with PVD TiN
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`films of equal thickness. This improved barrier performance could be attributed to a combination of factors, which
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`include the nitrogen—rich composition, higher density, and enhanced conformality of the CVD TiN phase in comparison
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`with the PVD TiN. In view of the superior step coverage and diffusion barrier characteristics, the low temperature inor—
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`ganic CVD route to TiN seems to provide an adequate replacement for conventional PVD TiN in emerging ULSI metal—
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`lization interconnect schemes.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`plicity, controllability, and ability to coat large area sub-
`Introduction
`strates with excellent uniformity at industrially viable
`Titanium nitride (TiN) is a commonly used material in
`growth rates, CVD could potentially meet performance
`current integrated circuit (IC) technologies.1 Its applica-
`demands well into the 0.18 am device technology and
`
`
`
`
`
`
`
`
`tions range from diffusion barrier and glue layer at the
`beyond.
`
`
`
`
`
`
`
`
`
`
`
`
`contact/via level to diffusion barrier and antireflection
`Early attempts at preparing CVD TiN used mostly tita-
`
`
`
`
`
`
`coating in the interconnect stack.2 Such applications are
`
`
`
`
`
`
`nium tetrachloride (TiCl4) and ammonia (NH3) to yield stoi-
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`made possible by the desirable properties of TiN, includ—
`chiometric TiN films having good step coverage.7 The
`
`
`
`
`
`
`ing its refractory nature at elevated temperature, excellent
`impurities produced by this process, mainly chlorine, were
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`mechanical, chemical, and thermal inertness, and good
`within 1 atomic percent (a/o). Unfortunately, the process-
`
`
`
`
`
`
`
`
`resistance to corrosion. These properties allow TiN to
`
`
`
`
`
`
`
`
`ing temperatures required to produce these films were in
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`withstand the repeated thermal cycles use in multilevel
`excess of 650°C, and were thus prohibitive for use above
`
`
`
`
`
`
`
`metallization of IC devices, and make its continued use in
`
`
`
`
`
`
`the contact level. Efforts to reduce deposition tempera—
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`emerging subquarter micron device technologies highly
`tures included plasma—assisted CVD (PACVD) of TiCl4 in a
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`desirable. However, the suitability of TiN for such appli-
`mixture of nitrogen and hydrogen, electron cyclotron res—
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`cations is only possible if it is deposited with good confor-
`onance (ECR) plasma CVD of TiCl4 in a nitrogen atmos—
`
`
`
`
`
`
`
`
`
`
`
`
`mality in subquarter micron features, leading to void-free
`phere,8 and atmospheric pressure CVD (APCVD) using
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`plug formation, reduced junction leakage, and low con—
`TiCl4 and isopropylamine or tert-butylamine as the co—
`
`
`
`
`
`
`
`tact/via resistance. This requirement is further complicat—
`
`
`
`
`
`
`
`
`
`
`reactant.9 These efforts led to an appreciable reduction in
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`ed by a strong push to reduce barrier thickness, as device size
`process temperatures to within the acceptable range of
`
`
`
`
`
`
`
`
`
`
`
`
`
`about 350 to 500°C. However, various reliability issues
`shrinks, to provide the cross section of aluminum or copper
`
`
`
`
`
`
`
`
`
`
`
`
`
`conductor required for optimum device performance.
`ranging from poor step coverage (30 to 70%) in some cases,
`
`
`
`
`
`
`
`
`Conventional physical vapor deposition (PVD) routes to
`
`
`
`
`
`
`
`
`
`
`to high resistivities (> 200 a!) cm) and chlorine contami—
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`TiN, with their inherent line of sight type deposition,
`nation above several atomic percent, in other cases, pre-
`
`
`
`
`
`
`
`appear to have reached their maximum useful lifetime,
`
`
`
`
`
`vented their incorporation in the IC process flow.
`
`
`
`
`
`
`
`
`
`
`even with the addition of special features such as collima-
`There are several recent reports on metallorganic CVD
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`t0rs.3 Modified PVD processes,
`including high density
`(MOCVD) of TiN from dialkylamino derivatives of titani~
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`plasma sputtering, appear to provide acceptable near term
`um of the type Ti(NR2)4, where R is a methyl or ethyl
`
`
`
`
`
`
`
`solutions, at least for the 0.25 pm device generation.“'6
`
`
`
`
`
`
`
`
`
`group.10 Additional MOCVD studies involved the use of
`
`
`
`
`
`
`
`Chemical vapor deposition (CVD), on the other hand,
`
`
`
`
`
`
`single—source
`titanium precursors
`the
`of
`type
`
`
`
`
`
`
`
`offers a low temperature alternative which is inherently
`
`
`
`
`
`
`
`[TiC12(NHR2)(NH2R)H] and [TiCl4(NR3)2], and cyclopenta—
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`capable of conformal metal growth. By combining sim—
`dienyl—based titanium compounds of the type biscyclopen-
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`tadienyl titanium diazide, szTi(N3)2 (where Cp = C5H5).11
`3 Present address: CVC Products, Inc., Rochester, New York
`
`
`
`
`
`
`These activities led to the development of robust and ver—
`14603, USA.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`J. Electrochem. 800., Vol. 144, No. 3, March 1997 © The Electrochemical Society, Inc.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Tianma Exhibit 1014
`
`
`
`1 002
`
`
`Page 1 of 7
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Page 1 of 7
`
`Tianma Exhibit 1014
`
`

`

`J. Electrochem. 800., Vol. 144, No.3, March 1997 © The Electrochemical Society, Inc.
`
`1003
`
`satile MOCVD TiN processes,12 with MOCVD from
`
`Ti(NEt2),,, for example, yielding resistivity below 300 pt!)
`
`
`
`
`
`
`
`
`
`
`cm, carbon contamination under 3 a/o, and step coverage
`as high as 70% in 0.35 pun device structures.
`
`
`
`
`
`The strategy espoused herein has focused, instead, on
`
`
`
`
`
`
`
`low temperature inorganic CVD of TiN from titanium
`
`
`
`
`
`
`
`
`tetraiodide (TiI4). Til, was selected because the dissocia—
`
`
`
`
`
`
`
`
`tion energy of the Ti—I is relatively low, with a correspond—
`
`
`
`
`
`
`
`ing heat of formation at 298 K of —92 kcal/mol. This
`
`
`
`
`
`
`
`value is well below the heat of formation for TiCl4, name—
`
`
`
`
`
`
`
`ly, — 192 kcal/mol. Accordingly, and in view of the similar
`
`
`
`
`
`
`
`
`
`chemical characteristics of the two halide chemistries, TiI4
`
`
`
`
`
`
`
`
`
`
`is expected to yield TiN films in an ammonia atmosphere
`
`
`
`
`
`
`
`
`
`
`with properties and performance similar to those from
`
`
`
`
`
`
`
`
`
`TiCl4 but at significantly lower temperature.
`
`
`
`
`
`
`
`Additionally, the activation energy for iodine diffusion
`
`
`
`
`
`
`
`
`
`is expected to be significantly higher than chlorine, given
`
`
`
`
`
`
`
`that I is a much heavier element than C1. This expectation
`
`
`
`
`
`is based on work on the interaction of fluorine and chlo—
`
`
`
`
`
`
`rine with (111) 81,13 and which showed that the barrier for
`
`
`
`
`
`
`
`
`chlorine penetration into the Si surface is much larger
`
`
`
`
`
`
`
`
`than that for fluorine. This behavior was a consequence of
`
`
`
`
`
`
`
`
`
`
`the larger size, and hence ionicity and resulting coulomb
`
`
`
`
`
`
`
`
`
`interaction, of the chlorine atom in comparison with fluo-
`
`
`
`
`
`
`
`
`rine. This property has important implications for the
`
`
`
`
`
`
`
`
`effects of residual halide incorporation in the deposited
`
`
`
`
`
`
`
`
`TiN, with 1 a/o iodine requiring appreciably higher ther—
`
`
`
`
`
`
`
`
`mal energy to diffuse out of TiN lattice than its chlorine
`
`
`
`
`
`
`
`counterpart. This observation is also supported by the
`
`
`
`
`
`
`
`findings presented herein, including diode leakage meas—
`
`
`
`
`
`
`
`
`urements, and which indicate that 2 a/o iodine did not
`
`
`
`
`
`
`
`
`
`
`affect TiN performance as diffusion barrier/adhesion pro—
`
`
`
`
`
`
`
`mater.
`
`
`
`
`
`
`The paper is the second in a series of reports on the iden-
`
`
`
`
`
`
`
`
`
`tification and optimization of a low temperature inorgan-
`
`
`
`
`
`
`ic CVD process for TiN from Ti14.” The first report has
`focused on the development of a low—temperature, in situ,
`
`
`
`
`
`
`
`
`
`
`
`sequential CVD process for the deposition of ultrathin
`
`
`
`
`
`
`
`Ti/TiN bilayers for applications in device ULSI technolo—
`
`
`
`
`
`
`
`
`
`gies. In this article, results are presented from a systemat—
`
`
`
`
`
`
`
`
`ic study of the microstructural, microchemical, and elec-
`
`
`
`
`
`
`
`trical characteristics of TiN films, as well as their
`
`
`
`
`
`
`
`performance in 0.45 pm, 3:1 aspect ratio contact/plug
`
`
`
`
`
`
`
`
`
`device structures. The findings from this study are com-
`
`
`
`
`
`
`
`pared and contrasted with those from PVD TiN grown by
`
`
`
`
`
`
`
`
`conventional PVD techniques.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Experimental
`
`
`The CVD reactor used for inorganic CVD of TiN from
`TiI, was a custom-made, 8 in. wafer, cold wall system, and
`was described in detail elsewhere.13 Briefly, it consisted of
`a parallel plate plasma configuration with the wafer locat-
`
`
`
`
`
`
`
`
`
`ed on the bottom electrode, which was resistively heated
`
`
`
`
`
`
`
`
`
`using an externally positioned boron nitride—coated
`
`
`
`
`
`
`
`
`graphite heater. The top active electrode was formed in the
`
`
`
`
`
`
`
`
`shape of a circular mesh to allow unrestricted flow of the
`
`
`
`
`
`
`
`
`reactants through a cone-shaped shower head located
`
`
`
`
`
`above the mesh. Pumping was achieved through eight
`
`
`
`
`
`
`
`
`
`ports which were symmetrically distributed below the
`
`
`
`
`
`
`
`
`
`
`heater chuck to permit uniform gas flow distribution. A
`
`
`
`
`
`
`standard pressure based sublimator was used to store the
`
`
`
`
`
`
`
`solid TiI,1 precursor, which was delivered to the reaction
`
`
`
`
`
`
`zone with the assistance of a hydrogen carrier gas.
`
`
`
`
`
`
`
`
`Ammonia reactant flow was delivered through a sideline
`
`
`
`
`
`
`
`
`directly to the reaction zone. A soft hydrogen plasma pre-
`
`
`
`
`
`
`
`
`deposition clean was performed on all samples prior to
`
`
`
`
`
`
`
`
`deposition. However, no plasma was used during actual
`
`
`
`
`
`
`
`deposition. Several types of wafers were processed, as
`
`
`
`
`
`
`
`
`
`shown in Table I, using the process conditions summarized
`
`
`
`
`
`
`
`
`in Table II.
`
`
`
`
`
`
`
`
`
`
`
`
`
`Methods of Analysis
`
`
`
`
`
`
`
`
`
`
`The TiN film microchemical, microstructural, and elec-
`trical properties were thoroughly analyzed at the New
`
`
`York State Center for Advanced Thin Film Technology by
`
`
`
`
`
`
`
`
`
`Page 2 of 7
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Table I. Type and number of wafers used in the study.
`
`
`
` Wafer type Number of wafers Thickness CVD TiN
`
`
`
`
`
`
`
`
`
`
`
`1000 A PETEOS/Si
`900 i 50.A
`30 i 2 A
`Si
`
`
`
`
`230 : 5 A,
`Si
`0.45 pm contact
`620 : 129A
`0.45 jun contact
`230 i 5 A°
`
`
`
`
`
`0.45 pm contact
`300 i 10 A
`
`
`
`0.45 pm contact
`400 : 10°A
`
`
`
`0.45 pm contact/nI
`30 i 2 A
`
`
`
`
`
`salicide junction
`0
`
`
`
`
`
`0.45 pm contact/n"
`230 i 5 A
`
`
`
`
`
`
`.
`salicide junction
`
`
`
`
`
`
`0.45 p.m contact/nI
`400 t 10 A
`
`
`
`
`salicide junction
`
`
`
`
`
`
`
`
`
`
`2
`2
`
`2
`3
`2
`1
`1
`3
`
`3
`
`3
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Table II. Summary of processing parameters.
`
`Process parameter
`Value
`
`
`
`
`
`
`Source precursor
`Wafer temperature
`
`
`Source temperature
`Reactor pressure
`Ammonia reactant flow
`
`Hydrogen carrier gas flow
`
`In situ predeposition
`
`Hydrogen Plasma clean
`
`
`
`
`'I‘iI4
`430°C
`140°C
`0.3 torr
`600 sccm
`30 sccm
`13.56 Mhz@
`0.08 W/cm2
`
`
`
`
`
`
`
`
`
`x—ray photoelectron spectroscopy (XPS), Rutherford
`
`
`
`backseattering (RBS), x-ray diffraction (XRD), four~point
`resistivity probe, and cross—sectional scanning electron
`microscopy (CS—SEM). In these studies, the results were
`
`
`
`
`standardized using a pure TiN standard deposited by col—
`
`
`
`
`
`limated sputtering at SEMATECH. Additional composi—
`
`
`
`
`
`tional and structural characterization was also carried out
`
`
`
`
`
`
`
`independently at SEMATECH.
`
`
`
`
`
`
`
`Film composition was determined using XPS and RBS.
`
`
`
`
`
`XPS was carried out on a Perkin—Elmer PHI 5500 Multi-
`
`
`
`
`
`
`
`Technique System. A magnesium x-ray source at 15 kV
`
`
`and 300 W was used. High resolution XPS scans employed
`
`
`
`
`
`
`
`a 23.50 eV pass energy to resolve shifts from particular
`
`
`
`
`
`
`
`
`
`photoelectron peaks.
`
`
`
`
`
`
`
`
`In addition to performing compositional characteriza-
`
`
`
`
`
`
`
`
`
`tion, RBS was also employed, in conjunction with CS—
`
`
`
`
`
`
`
`
`
`SEM,
`for thickness and growth rate measurements.
`
`Rutherford backseattering (RBS) spectra were taken using
`
`
`
`
`
`a 2 MeV He+ beam, and calibrated with bulk samples of
`
`
`
`
`
`
`
`
`gold and carbon, while CS—SEM investigations employed
`
`
`
`
`
`
`a Zeiss DSM 940 microscope using a 20 keV primary elec—
`
`
`
`
`
`
`tron beam. Four—point probe resistivity measurements
`
`
`
`
`
`
`
`
`
`used a Signatone four-point probe. Deposition rates were
`
`
`
`
`
`
`defined as T/t, where T is film thickness and t is run time.
`
`
`
`
`
`
`
`
`
`
`Run time was measured starting form the instant when the
`
`
`
`
`
`precursor was actually being delivered to the reactor.
`
`
`
`
`
`
`X—ray diffraction was done on a Scintag XDS 2000 x—ray
`
`
`
`
`
`
`
`
`
`
`
`diffractometer. X-rays were generated with a Cu K“ x-ray
`
`
`
`
`
`
`
`
`
`source at typical tube operating power of 1.8 kW, which
`
`
`
`
`
`
`
`corresponds to 40 mA and 45 kV. XRD spectra were col—
`
`
`
`
`
`
`
`
`lected both in normal incidence and 5° grazing angle
`
`
`
`
`
`
`
`
`geometries. The resulting TiN XRD patterns were com—
`
`
`
`
`
`
`
`
`
`pared with the sputtered TiN standard provided by
`
`
`
`
`
`
`
`
`
`
`SEMATECH and a TiN reference pattern from the stan—
`
`
`
`
`
`
`
`
`dard JCPDS powder diffraction file (PDF).
`
`
`
`
`
`
`
`A complete evaluation of the physical properties and
`
`
`
`
`
`
`
`barrier characteristics of the TiN films was carried out at
`
`
`
`
`
`
`
`
`the Semiconductor Process and Device Center of Texas
`
`
`
`
`
`Instruments. The microchemical and electrical measure—
`
`
`
`
`
`
`
`ments performed above at the Albany Center were repeat-
`
`
`
`
`
`
`
`
`
`ed as an independent checking mechanism of TiN film
`
`
`
`
`
`
`
`properties. The parameters measured and corresponding
`
`
`
`
`
`methods of analysis are summarized in Table III and dis-
`
`
`
`
`
`
`
`
`cussed in more detail in the following sections. In these
`
`
`
`
`
`
`
`
`studies, the results were standardized using a pure TiN
`
`
`
`
`
`standard deposited by collimated sputtering at Texas
`
`
`
`
`
`
`
`
`
`Instruments.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Page 2 of 7
`
`

`

`1 004
`
`J. Electrochem. 800., Vol. 144, No.3, March 1997 © The Electrochemical Society, Inc.
`
`
`
`
`
`100
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Table IV. Summary of results.
`Table III. Methods of analysis.
`
`
`
`
`
`
`
`
`
`
`
`
`
`Method
`Parameter
`Parameter
`Value
`
`
`
`
`
`
`
`
`
`Four-pointprobe (900 A TiN on
`Deposition rate (A/min)
`200
`Sheet resistance
`1000 A PETEOS/Si)
`5 to 8
`Nonuniformity (%, lo)
`
`
`Step coverage (%)
`Profilometry (900 A TiN on 1000 A
`90
`PETEOS/Si) RB:
`Resistivity (11.0. cm)
`100 to 150
`TiN on 1000 A
`Weightgain (900
`Stoichiometry (TizN)
`1:1.06
`
`
`
`
`
`
`
`PETEOS/Si)
`<2 a/o iodine
`Impurities
`
`
`
`
`
`
`RBS and XPS (900 A TiN on 1000 A
`<1 a/o oxygen
`
`
`
`
`
`
`
`
`PETEOS/Si)
`1.1 times that of PVD TiN
`Density
`
`
`
`
`
`CS—SEM (230, 400 900 A TiN on
`Golden
`Color
`
`
`
`
`
`
`
`
`0.45 and 0.25 urn contact)
`Zero
`Stability (ABS/2 h, %)
`
`
`230,400,600 AllTiN on
`Similar to PVD TiN
`CVD W plug fill
`
`
`
`
`
`
`
`
`
`
`Similar to PVD TiN
`0.45 pm contact
`A1 reflow plug fill
`
`
`
`
`Similar to PVD TiN
`Used 30 to 230 A CVD TiN on Si:
`Reactivity with Al—Cu
`
`
`
`
`
`
`Barrier property
`Superior to PVD TiN
`(2') Deposited 600 A A1—0 5 3/0 Cu
`
`
`
`
`
`
`
`then spintered at 450 to 550°C for
`
`
`
`
`
`
`
`
`
`
`
`30 to 60 min. (ii) Measured pre-
`
`
`
`
`
`
`
`
`and post Al—Cu sheet resistance.
`
`
`
`
`
`
`
`
`
`
`
`
`
`RBS analysis of TiN on Si also supported the XPS
`(iii) After sintering, metal was
`
`
`
`
`
`
`
`
`
`
`stripped and o tical microscopy
`results, and indicated the presence of about 2 a/o iodine
`
`
`
`
`
`
`used to quantify pit density1n Si.
`incorporation, as displayed in Fig. 4.
`
`
`
`
`
`
`(iv) Samples on Si, plasma TEOS,
`
`
`
`
`The composition of the CVD produced TiN films was
`
`
`
`
`
`
`
`
`
`PVD TiN were similarly processed
`
`
`
`
`nitrogen-rich in comparison with the SEMATECH colli-
`and used for comparison.
`
`
`
`
`
`
`
`
`
`
`
`
`mated PVD 'IiN. Levels of iodine in the films are not
`CS—SEM (230, 400, 900 A TiN on
`
`
`
`
`
`
`
`
`
`
`
`0.45 pm contact
`
`
`
`
`
`30to400ATiNon0.45 pun
`
`
`
`
`contact/n salicide junction
`
`
`
`30 to 400 A TiN on 0.45 pm
`
`
`
`
`
`contact/n+ salicide junction
`
`
`
`
`
`
`
`
`
`
`Thickness
`
`Density
`
`Composition
`
`Step coverage
`Barrier attack
`Barrier properties
`
`
`
`
`
`
`W plug fill
`Contact resistance
`
`Diode leakage
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`(%)
`AtomicConcentration
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`8 6
`
`
`
`
`
`0
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Sputter Time (mins.)
`
`indicates a stoichiometric TiN phase
`Fig. 1. XPS depth protili
`
`
`
`
`
`with iodine concentration be ow 2 a/o. Noooxygen or other conta-
`XPS.
`ts
`minants were found within the detection limi
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`.. Q
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`454
`
`450
`
`
`
` NonnaltzedIntensity(arbitraryunits) 470
`
`
`
`
`
`
`
`
`
`466
`
`458
`462
`Binding Energy (eV)
`
`Fig. 2. XPS high resolution spectrum of TiI2pm indicates a core
`
`
`
`
`
`peak at 455.1 eV corresponding to a pure TiN phase.
`
`
`
`
`10
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`W O
`
`:
`
`
`
`a8
`
`402
`
`398
`Binding Energy (eV)
`
`394
`
`390
`
`A N
`
`:
`
`Q
`
`
`
`NormalizedIntensity(arbitraryunits)
`
`
`
`
`
`.3 The XPS N Is binding energy at 397.3 eV indicates a pure
`TiNiphase.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`The 900 A CVD TiN films on plasma tetraethylorthosil-
`icate (PETEOS) were analyzed for sheet resistance, sheet
`thickness, density and composi—
`resistance uniformity,
`tion/stoichiometry. Sheet resistance was measured using a
`
`
`
`
`
`
`
`
`four-point probe, thickness was calculated using a pro-
`
`
`
`
`
`
`
`filometer, density was determined from weight gain
`
`
`
`
`
`measurements, and composition/stoichiometry was
`
`
`
`
`
`
`determined from RBS analysis. For the RBS analysis, a
`
`
`
`
`
`
`
`PVD (conventional sputtering) TiN film was used as
`
`
`
`
`
`
`control for comparison.
`
`
`
`The 30 to 230 A CVD TiN films on Si were used for bar-
`
`
`
`
`
`
`
`
`rier studies. A 6000 A A1— 0.5 a/o Cu film was sputter
`
`
`
`
`
`
`
`deposited on the CVD TiN and the stack was subjected to
`
`
`sintering at 450 to 550°C for 30 to 60 min. Sheet resistance
`
`
`
`
`
`
`
`
`
`
`
`
`
`measurements prior to and after sintering were used to
`
`
`
`
`
`
`
`
`
`
`quantify the reaction between Al—Cu and CVD TiN. After
`
`
`
`
`
`
`
`
`
`
`sintering was complete, the metal was stripped and the
`
`
`
`
`
`
`
`
`
`
`
`density of pits in the silicon was measured on an optical
`
`
`
`
`
`
`
`
`microscope to quantify the quality of the CVD TiN barri-
`
`
`
`
`
`
`
`
`er. 6000 A A1—0. 5 a/o Cu films deposited directly on Si,
`
`
`
`
`
`
`
`
`plasma TEOS, and PVD TiN were also subjected to simi-
`
`
`
`
`
`
`
`
`
`
`lar sinter cycles for comparison. The patterned contact
`
`
`
`
`
`
`
`
`
`wafers were used for step coverage measurements and to
`
`
`
`
`
`
`
`
`
`
`evaluate nucleation, barrier attack, and plug—fill during
`
`
`
`
`
`
`
`
`
`CVD tungsten deposition. Finally, the contacts on the n"
`
`
`
`
`
`
`
`salicided junctions were used to measure contact resist—
`
`
`
`
`
`
`
`
`ance and contact—induced diode leakage.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Results
`
`
`
`
`
`
`
`The TiN films thus produced were metallic, mirror—like,
`
`
`
`
`and gold colored. Their physical properties are summa—
`rized in Table IV and discussed in more detail below. In
`particular, deposition nonuniformity is slightly high, but is
`
`
`
`
`
`
`
`an equipment not a process issue, given that a custom-
`
`
`
`
`
`
`
`made, nonoptimum, reactor was used in the study.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Film composition—The composition of the films was
`
`
`
`
`
`
`
`
`
`examined by XPS and RBS. XPS depth profiling, as
`
`
`
`
`
`
`
`shown in Fig. 1, indicated a nitrogen-rich TiN,“ phase
`with iodine concentration below 2 a/o. No oxygen or other
`
`
`
`
`
`
`contaminants were found within the detection limits of XPS
`
`
`
`
`
`
`
`
`(~1a/o). Additionally, XPS high resolution spectra of ele-
`
`
`
`
`
`
`
`
`mental core levels yielded a Ti 2p3,2 core peak at 455.1 eV,
`
`
`
`
`
`
`
`
`as displayed in Fig. 2, corresponding to a pure TiN phase.
`
`
`
`
`
`
`
`
`Figure 3 shows the N Is binding energy at 397.3 eV, also
`
`
`
`
`
`
`
`indicating a pure TiN phase.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Page 3 of 7
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Page 3 of 7
`
`

`

`J. Electrochem. 800., Vol. 144, No.3, March 1997 © The Electrochemical Society, Inc.
`
`1 005
`
`30
`
`toQ
`
`
`
`NormalizedYield
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`10
`
`O
`
`100
`
`200
`
`300
`Channel
`
`500
`
`600
`
`Fig. 4. R35 analysis of TiN on Si sup
`rts the XPS results, and
`indicates the incorporation of about 2 f0Iodine.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`expected to pose any reliability problems, such as Si
`
`wormhole formation and metal corrosion. This assessment
`is attributed to the higher activation energy for iodine dif-
`fusion as compared with chlorine, given that I is much
`
`
`
`
`
`
`
`
`heavier than Cl. Accordingly,
`it is expected that 1 a/o
`
`
`
`
`
`
`iodine might require appreciably higher thermal energy to
`
`
`
`
`
`
`
`
`
`diffuse out of the TiN lattice than its chlorine counterpart,
`
`
`
`
`
`
`
`
`
`thus posing less of a reliability problem. This assumption
`
`
`
`
`
`
`
`
`
`is supported by the findings presented herein and which
`
`
`
`
`
`
`
`demonstrate that TiN barrier integrity and device perfor-
`
`
`
`
`
`
`
`
`
`mance are not compromised by 2 a/o residual iodine levels.
`
`
`
`
`
`
`
`It is also in agreement with similar work reported by
`
`
`
`
`
`
`
`
`Yokoyama et al. on the mobility of chlorine in CVD TiN
`
`
`
`
`
`
`
`produced from the reaction of 'I‘iCl4 and NH3.15
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Film texture—Figure 5 shows a typical XRD spectrum
`
`
`
`
`
`
`
`
`
`
`of a CVD TiN film. The spectrum exhibited reflections cor—
`
`
`
`
`
`
`responding to 20 = 42.340 (111) and 61.98 (220), with the
`(111) orientation exhibiting the highest diffraction inten-
`
`
`
`
`
`
`sity. Interestingly, the XRD spectrum displayed in Fig. 6
`
`
`
`
`
`
`
`
`from SEMATECH’S sputter—deposited TiN film showed a
`
`
`
`
`
`
`
`
`
`different XRD texture, with the (200) orientation showing
`
`
`
`
`
`
`the strongest diffraction peak. Interestingly, Yokoyama et al.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`also cfiibserved textural variations between CVD and PVD
`TiN.1
`
`
`
`
`
`
`
`The textural variations observed might have significant
`
`
`
`
`
`
`
`implications for the barrier characteristics and associated
`
`
`
`
`
`
`
`differences in the peformance of CVD and PVD grown
`films. Additionally, it has been shown that the texture of
`
`
`
`
`
`
`the TiN film can significantly affect that of the overlaying
`
`
`
`
`
`
`aluminum film. Primarily, predominantly (200) textures
`
`
`
`
`
`
`
`
`TiN leads to the formation of similarly textured alu-
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Intensity(ArbitraryUnits)
`
`30
`
`‘ 0
`
`0
`
`60
`
`70
`
`Angle 29
`
`(degrees)
`
`Fig. 5. Typical XRD spectrum of a CVD TiN film. The diffraction
`k locations and intensifies are in good agreement with those
`fmfn a standard TiN powder sample, indicating that the TiN filmIs
`
`
`polycrystalline.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Page 4 of 7
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Units)
`Intensitu(Arbitraru
`
`
`3o
`
`40
`
`60
`50
`Angle 29 (degrees)
`
`70
`
`Fig. 6. XRD spectrum of a sputter-deposited TiN film.
`
`
`
`
`
`
`
`
`
`
`
`minum, while (111) oriented TiN produces an identical
`aluminum orientation. This effect is important since pre-
`dominantly (111) oriented aluminum exhibits enhanced
`electromigration resistance and increased mean time to
`
`
`
`
`
`
`
`failure (MTTF) under electromigration stress conditions in
`
`
`
`
`
`
`
`comparison with its (200) counterpart.17 Additional stud—
`
`
`
`
`
`ies are underway to study these effects and will be report—
`
`
`
`
`
`
`ed in a subsequent publication.
`
`
`
`
`
`
`
`
`
`
`
`
`Aggressive structure fill and resistivity—Figure 7 dis—
`
`
`
`
`
`
`
`
`
`
`plays typical CS-SEM micrographs of TiN step coverage
`
`
`
`in aggressive device trench structures, with nominal fea—
`ture size of 0.45 pm, 3:1 aspect ratio. As can be seen, step
`
`
`
`
`
`coverage is excellent, with 90% conformality observed in
`
`
`
`
`
`
`
`the structures examined. Interestingly, highly conformal
`
`
`
`
`
`
`
`TiN coverage was achieved across a wide process window,
`
`
`
`
`
`
`
`
`
`
`
`
`as demonstrated in Fig. 8 which shows 90% TiN step cov—
`
`
`
`
`
`
`
`erage in 0.25 pm, 4:1 aspect ratio structures. These sam—
`
`
`
`
`
`ples were processed at a substrate temperature of 425°C,
`
`
`
`
`
`
`
`
`source temperature of 140°C, reactor pressure of 0.5 Torr,
`
`
`
`
`
`
`
`
`
`
`and hydrogen carrier gas and ammonia reactant flows of,
`
`
`
`
`
`
`
`
`
`respectively, 30 and 400 sccm. The values observed are sig—
`
`
`
`
`
`
`
`
`nificantly higher than those that can be obtained using
`
`
`
`
`
`
`
`
`collimated PVD TiN. Film resistivity ranged from 100 I10
`
`
`
`
`
`
`
`
`cm for 1000 A films up to 150 an em for 100 A films.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Barrier properties—The density of the film is high, 1.1
`
`
`
`
`
`
`
`
`times that of the collimated PVD TiN as tested by Texas
`
`
`
`
`
`
`
`
`
`
`
`
`
`Instruments, and the films have a strong golden color.
`Films are stable in air and showed no evidence of oxida—
`
`
`
`
`
`
`
`
`tion, even after prolonged exposure to air. Plug filling with
`
`
`
`
`
`
`
`
`
`
`CVD tungsten and Al reflow are similar for vias lined with
`
`
`
`
`
`
`
`
`CVD TiN or PVD TiN. Reactivity with Al—0.5% Cu during
`
`
`
`
`
`
`
`
`
`
`a 450°C, 30 min forming gas sinter was similar to the reac-
`
`
`
`
`
`
`
`
`
`tivity of PVD TiN. The sheet resistance increase upon sin—
`
`
`
`
`
`
`
`
`
`
`tering is marginally lower for CVD TiN compared to PVD
`
`
`
`
`
`
`
`
`
`TiN, as observed in Fig. 9. The various splits indicate that
`
`
`
`
`
`
`
`
`
`
`
`the sheet resistance increase is due to interaction between
`
`
`
`
`
`
`
`
`
`the Al—Cu and TiN rather than between Al—Cu and Si.
`
`
`
`
`
`
`
`
`
`After a 550°C, 60 min sinter, splits 1, 2, and 4 show pitting
`
`
`
`
`
`
`
`
`
`
`of the Si surface indicative of barrier failure. However the
`
`
`
`
`
`
`
`
`density of pits is maximum for the case of no barrier, and
`
`
`
`
`
`
`
`
`
`is minimum for the CVD TiN barrier.
`
`
`
`
`
`
`
`
`
`
`
`Superiority of the CVD TiN barrier can also be gauged
`
`
`
`
`
`
`
`
`
`from the contact—induced diode leakage for a diode with
`
`
`
`
`
`
`
`
`
`
`
`250,000 contacts (Fig 10). The diode leakage characteris—
`
`
`
`
`
`
`tics for a 230 A CVD TiN film are equivalent to those for a
`
`
`
`
`
`
`
`
`
`500 A PVD TiN film. Increasing the CVD TiN thickness to
`
`
`
`
`
`
`
`
`400 A decreases the leakage further. There is no apprecia—
`
`
`
`
`
`
`
`ble difference1n diode leakage for block diodes with few
`
`
`
`
`
`
`
`
`
`
`
`
`contacts except for the case of 30 A CVD TiN which
`
`
`
`
`
`
`
`
`
`
`appears to be too thin, or possibly not even continuous, to
`
`
`
`
`
`
`
`
`
`yield a reliable measurement (Fig. 11). Additionally, the
`
`
`
`
`
`
`
`
`
`wider dispersion'in the diode leakage current for the 400 A
`
`
`
`
`
`
`
`
`
`
`TiN film is attributed to a larger than typical thickness
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Page 4 of 7
`
`

`

`1006
`
`J. Electrochem. 800., Vol. 144, No.3, March 1997 © The Electrochemical Society, Inc.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`— 5
`
`08 no
`
`Fig. 7. Typical CS-SEM micrograph of TiN step coverage in 0.45 pm, 3:1 aspect ratio contacts.
`
`
`
`
`
`
`
`
`
`
`
`
`variation across the specific wafer used in those measure-
`ments. Accordingly, the worst leakage value corresponded to
`a spot on the wafer where the TiN was thinner than 400 A.
`
`
`
`
`
`
`
`
`Discussion and Conclusions
`
`
`
`
`
`
`
`The results discussed above demonstrate that the low
`
`
`
`
`
`
`
`
`
`
`
`
`temperature inorganic CVD route is a viable approach for
`
`the deposition of TiN for applications as barrier layer and
`adhesion promoter in emerging subquarter micron device
`
`
`
`
`
`
`technologies. The TiN films thus produced were gold col-
`
`
`
`
`
`
`
`ored, stoichiometric, and exhibited metallic conductivity,
`
`
`
`
`
`
`
`
`
`with resistivities in the range of 100 to 150 p!) cm, depend—
`
`
`
`
`
`
`ing on film thickness. Microchemical analyses showed that
`
`
`
`
`
`
`
`
`the films were free from oxygen or carbon contamination,
`
`
`
`
`
`within the detection limits of XPS, with iodine concentra-
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`tions below 2 3/0. Such levels of iodine concentrations are
`not expected to cause reliability problems in device oper-
`ation. This expectation is based on the assumption that
`iodine is significantly heavier than chlorine and will thus
`
`
`
`
`
`
`
`
`
`require higher activation energy for diffusion out of the
`
`
`
`
`
`
`
`
`TiN matrix. The diode leakage data presented herein
`
`
`
`
`
`
`
`
`seems to also support this assessment.
`
`
`
`
`
`
`
`
`A systematic evaluation was also carried out of the bar—
`
`
`
`
`
`
`
`
`rier characteristics of CVD TiN using PVD TiN as base line
`
`
`
`
`
`
`
`material. The data showed that reactivity with Al—0.5 a/o
`
`
`
`
`
`Cu alloys was equivalent in both cases, with the CVD TiN
`
`
`
`
`
`
`
`
`
`film exhibiting a marginally better performance as com—
`
`
`
`
`
`
`
`
`
`
`pared to its PVD counterpart. Similarly, the characteris—
`
`
`
`
`
`
`
`
`tics of CVD tungsten and reflow plug fills were identical