Abstract

This work investigates and summarizes the literature on basic deposition and thin film properties of tungsten thin films grown by chemical vapor deposition (CVD). Deposition of tungsten films is generally performed by the reduction of tungsten hexafluoride (WF₆) by silane (SiH₄), hydrogen gas (H₂), or silicon (Si). Each reaction has different behavior and and can produce films with different properties. Si reduction is generally not used commercially, while silane and hydrogen reduction are used fairly extensively. Silane reduction occurs quickly and at low temperatures and is generally used to seal contacts and produce ‘seed’ layers, whereas hydrogen reduction produces more pure films with better step coverage and is usually used to grow bulk films on top of a thin seed layer. Silicon reduction is usually not used commercially because it produces unfavorable conditions for further reactions. In the Silane reduction, it appears that two factors have the largest effect on film properties: substrate temperature and reactant ratio. Low substrate temperature seems to favor a -phase tungsten while high substrate temperature seems to favor b -phase tungsten. Reactant ratios with little WF6 in comparison to silane also seem to favor b -phase formation. For the hydrogen reduction reaction, the factors that seemed to have a notable effect on film properties were temperature and total pressure, with high temperature and low total pressure favoring low tensile stress. The films produced by silicon reduction did not appear to be highly dependent on reaction conditions once the self-limiting thickness was reached.

Tungsten thin films for electronic devices can be prepared in several ways. Evaporation, sputtering and CVD are the most commonly used processes in industry. CVD has several advantages over evaporation and sputtering, including better step coverage, less radiation damage to the device during processing and low coating time. The equipment necessary to obtain high quality films is also relatively economical in comparison to other techniques (Shon, 1992). Additionally, tungsten can be deposited selectively on desired areas under certain conditions (Creighton, 1994) to remove post-processing steps such as masking and etching.

Tungsten deposition is generally accomplished industrially by the reduction of tungsten hexafluoride (WF₆) by either silane (SiH₄) or hydrogen gas (H₂). WF₆ is a popular tungsten precursor due to its low cost and easy availability at high purities (Kodas, 1994). WF₆ can also be reduced by silicon, but that reaction produces poor quality films and is self limiting, which makes it industrially impractical for use as a primary reaction (Baker, 1996). Silane reduction is the highest rate and lowest temperature reaction, occurring at temperatures as low as 90 C. While these things make silane reduction attractive, step coverage is poor due to high deposition rates, it tends to lose selectivity easily, and it may produce films contaminated with silicon. Industrially, silane reduction is generally used only as a first layer to seal a contact at low temperature. The hydrogen reduction occurs at lower rates and higher temperatures, in the 300-500 C range. It has excellent step coverage and is generally used to complete the bulk of the layer after a seed layer has been applied. Hydrogen reduction is generally not suitable as a first layer because the high reactivity of WF₆ toward the silicon substrate causes excessive Si consumption, encroachment at the Si/SiO₂ interface and wormhole formations in the contact regions (Yeh et al., 1994).

As mentioned above, tungsten can be deposited both selectively or as a blanket covering an entire wafer surface. Blanket deposition is currently the technique most commonly used in industry. After a blanket deposition has been done, the surface is masked and etched for the desired final result. Selective deposition is in many ways more attractive, with good step coverage resulting from fewer process steps. This can produce quality films in less time with less toxic waste. Unfortunately, selective deposition is far more difficult to do reliably, which is why it is not used extensively in industry at this time. Tungsten hexafluoride (WF₆) reduction by silicon is a highly selective reaction, but tends to create poor selectivity conditions for subsequent H₂ reductions. Silane reduction is a moderately selective reaction while hydrogen reduction is the most selective method. Loss of selectivity (LOS) has been found to result from sub-fluorides present during reaction, hot reactive surfaces in the reactor, silicides present during reaction (Creighton, 1994) and unfavorable ratios of reactants. Work still needs to be done to make selective deposition of tungsten a viable industrial process.

The properties of the thin film produced by a given reactant combination can differ greatly from those produced by another chemistry, or even by the same chemistry under different temperature and pressure conditions. These differences can be critical in the fabrication of semiconductor devices. In order to make intelligent process control choices and tradeoffs that do not compromise device quality, it is very beneficial to have an understanding of the reaction and material properties that are likely to result from different chemistries and conditions. The reaction behavior and film properties for tungsten films produced by the reduction of WF₆ by silane, hydrogen, and silicon will be discussed.
 

Silane Reduction and Properties of Deposited Films

Silane reduction, as mentioned previously, allows films to be deposited selectively at low temperatures with high growth rates. It also has some concomitant disadvantages such as difficulty achieving good step coverage and adhesion, easy loss of selectivity (LOS), and potential silicon contamination of films. In a study by Lo et al. (1973), it was found that even one percent Si in a tungsten film can break up the columnar structure of thick b -type tungsten.

One large advantage of silane reduction chemistry is that it is relatively benign to substrates such as Si, Al, TiW, and TiSi₂ in comparison to hydrogen reduction. (Schmitz, 1989) Schmitz et al. report a rate equation for the silane reduction of:

Rw = k₂*P_WF6⁰*P_SiH4¹ (1)

for low SiH₄/WF₆ ratios. This data was analyzed by Shon, and both the orders of WF₆ and Silane were confirmed using data that was shown to not be mass transfer limited at 270 C and SiH₄/WF₆ ratios below one.

Yeh et al. (1996) characterized the deposition properties of selective CVD tungsten using silane reduction of WF₆ with an inlet SiH₄/WF₆ flow rate ratio less than 0.6 over the temperature range of 280-350 C with a total pressure of 100 mTorr. Conversion was not reported. The starting material used was p-type <100> silicon wafers and n-type <100> wafers. After RCA cleaning, a 1500 A layer of oxide was grown thermally followed by deposition of 4500 A thick BPSG. Diffusion areas on the wafers were defined using standard photolithographic techniques. The wafers were dipped in 2% HF solution prior to deposition and were then loaded into the reactor and plasma etched. Selective deposition was obtained between 280-350 C where the selective surface reaction was the dominant process. The morphology of the tungsten deposited at high temperature had a b -phase columnar structure and a higher resistivity; films deposited at low temperatures had an a -phase granular structure and lower resistivity. The morphology was also found to be dependent on SiH₄/WF₆ inlet flow ratio. At flow ratios less than 0.6, the film deposited was a structure whereas at inlet flow ratios greater than two, the film deposited was b structure.

Schmitz et al. performed LPCVD in a 2.2 liter custom cold wall reactor. All experiments were performed at 300 C. The SiH₄ flow rate was fixed at 13.5 standard cubic centimeters per minute (sccm) and the SiH₄/WF₆ flow ratio was varied between 0.55 and 15. Total pressure in all reactions was 200 mTorr with argon as a carrier gas. Their work also found that film type had a strong dependence on gas composition. At inlet reactant ratios below 1.3, X-ray diffraction patterns were observed that corresponded with a -type tungsten. At inlet ratios above 1.3, a diffraction pattern appeared that corresponded to b -type tungsten. They assumed that competitive adsorption of WF₆ and SiH₄ was resulting in two parallel reaction paths. In his thesis, Shon (1992) showed that these experiments were run under conditions of severe surface depletion of WF₆. The appearance of silicon in the layer coincided roughly with the observation of the b diffraction pattern. It was found that impurities in the films, such as fluorine and oxygen, can stabilize b -type tungsten and inhibit a -type tungsten formation. From Auger analysis, it was found that oxygen was present in the b lattice but not at all in the a lattice. The a lattice did however have a silicon content of 3%, which was quite high. They also concluded that film densities could be different at differing deposition conditions, based upon the observation that at one reactant flow ratio there was an increase in the calculated total growth rate and no corresponding change in tungsten deposition rate as measured by RBS. Growth rate was calculated using thickness measurements determined by Auger spectroscopy. Due to Shon’s analysis showing severe mass transfer limitation (MTL), however, conclusions with respect to SiH₄/WF₆ ratios must consider that the near surface gas ratio is probably quite different than the inlet ratio. Because WF₆ diffuses much more slowly than silane and two silane molecules are used up for each WF₆ molecule reacted, the near-surface ratio will vary tremendously from the inlet ratio, and will vary with WF₆ conversion in the reactor.

Lee et al. (1993) investigated the effects of SF₆ and CF₄ plasma pretreatments on the adhesion and junction leakage characteristics of selectively deposited CVD W films on p-type (100) Si wafers. A commercial cold wall LPCVD reactor was used and wafers were heated using an IR lamp. Contact resistances were measured using a test pattern of 360 contact holes. Junction leakage current was measured using n⁺/p diodes reverse biased at 5V. With only an HF dip as wafer pretreatment, peeling was not a problem for as-deposited tungsten films. However, peeling did start to occur during annealing at 450 C in N₂. Contact resistances in peeled areas were found to be very high, on the order of > 1 KW . It was found that contact resistance for Al/n⁺ contacts was about 100 W lower than W/n⁺ contacts. It was found that junction leakage currents in W/n⁺ junctions pretreated with SF₆ plasma were considerably higher than those from the Al/n⁺ control contacts. They postulate that junction leakages and high contact resistance are mainly caused by lateral and vertical Si consumptions due to the high etch rate of SF₆ plasma. Etching done with CF₄ plasma at a much lower rate produced contact resistances and leakage currents comparable to the Al/n⁺ control. Adhesion problems even with plasma etching pretreatment presented themselves as films became thick, on the order of 500 nm. This is probably due to the asymmetric grain growth induced stress mentioned above. They found that improvement in junction and leakage characteristics could be obtained by following a three step process of Silane reduction at 300 C for 40 seconds, followed by annealing at 450 C for 120 seconds, followed by silane reduction at 300 C until the desired thickness was obtained.

Tsutsumi et al. (1990) deposited tungsten films in a cold wall, lamp-heated, single-wafer reactor using WF₆, SiH₄, and H₂. Deposition temperature and pressure were 300 C and 0.1 Torr, respectively. The inlet flow ratio of SiH₄/WF₆ was varied between 0.4 and 1.0. Inlet gas flow rates of WF₆ and H₂ were held constant at 6 and 1000 sccm. Film thickness and morphologies were studied using SEM. X-ray diffraction measurements were performed to examine crystal orientation and composition depth profiling was done with Auger electron spectroscopy. They examined the growth behaviors and film properties of WSi₂, TiSi₂ and n⁺ Si as a function of SiH₄/WF₆ ratio. They found that as the ratio increases, the growth rates of the films increase rapidly at critical values. At these values, the morphologies of the films also change from films with more symmetric grains to films with columnar grains. They also found that the critical ratios appeared to depend on substrate material, with the critical ratio being 0.8 for Si. Under the critical values, the orientation of the films is a -phase while over the critical ratio the film becomes b -phase. Auger electron spectroscopic analysis showed a large amount of silicon content in the films at high ratios compared to low ratios. They speculated, as have many others, that impurity atoms may correlate to the formation of unstable b -tungsten at the expense of a -tungsten.
 

Hydrogen Reduction and Properties of Deposited Films

Hydrogen reduction of WF₆ can achieve relatively high rates depending on process conditions, but is slower than silane reduction under the same conditions. Tungsten films produced by hydrogen reduction need to be grown on top of a ‘seed’ layer of hydrogen dissociating material such as Pt, Pd, W, or Ni. In the literature, there seems to be general agreement that the deposition rate for this reaction depends on WF₆⁰ and H₂^1/2.

Hydrogen reduction is the most selective of the common industrial tungsten deposition reactions and achieves excellent step coverage, well-crystallized films and little potential for contamination if deposited on top of an existing layer. The hydrogen reduction used alone, however, tends to allow WF₆ to attack bare substrates, such as silicon, if a seed layer hasn’t first been deposited. This can produce junction leakage problems associated with encroachment and tunneling (Lee et al., 1993). Hydrogen reduction requires higher temperatures than silane reduction for comparable deposition rates.

Metz et al. (1984) deposited thin tungsten films selectively onto silicon surfaces by LPCVD using silicon reduction followed by hydrogen reduction of WF₆. The silicon and hydrogen reductions were both performed at 300 C. The reactor used was a conventional quartz hot wall reduced pressure reactor. A deposition time of 10 minutes was used. The substrates were thermally oxidized silicon wafers, coated with 5000 A of polysilicon done in a previous LPCVD step. SEM micrographs were used to examine selectivity. Selective films were found to be quite smooth and very adherent to both polysilicon and single crystal silicon surfaces. X-ray diffraction analysis was done to prove that the films were very well crystallized in spite of the low deposition temperature. The (110) peak was the most intense, followed by the (211) and (200) peaks. A non-temperature-dependent residual resistivity of 12 microohm-centimeters (mW -cm) was found. It is speculated that this is due to the oxygen impurities that were located by Auger electron spectroscopy and could be reduced by process improvements. Auger electron spectroscopic analysis indicated no residual fluorine contamination and 0.9% residual oxygen contamination. Secondary ion mass spectrometry was also done to confirm the fluorine results. Hall effect was examined and indicated that the material was primarily a hole conductor, with a relatively low Hall mobility that is consistent with the residual resistivity found. Difficulties with interpretation of the Hall effect were discussed, because Hall effect is a difficult property to quantify in metals and especially tungsten.

Wulu et al. (1989) studied film stress in films deposited by H₂ reduction of WF₆. They used device-quality (100) silicon wafers coated with 40nm of thermal oxide and 100 nm of phosphorous doped LTO. Wafers were analyzed for curvature before tungsten deposition and those which showed excessive curvature were not used in the analysis. Tungsten was deposited using the hydrogen reduction reaction at 305 C in a hot wall LPCVD reactor with a high H₂ flow. Film thickness was determined by weight, assuming a 100% dense film, which is probably not the best approximation. They found that tensile stress is a function of film thickness and rises rapidly for films between 15 and 40 nm thick. For films greater than 50 nm thick, tensile stress decreased. They observed that this behavior is also seen in Cr and Ag thin films where the behavior is usually explained by the coalescence of separate grains or densification of the metal film network during the initial stages of growth. They also found that stresses on the order of 1 GPa in the film did produce a noticeable increase in leakage current in shallow junctions.

Leusink (1994) studied the development of growth stressed in W films deposited by hydrogen reduction using in situ wafer curvature measurements. Reactor details can be found below in the silicon reduction section. He found that stress in tungsten films deposited by hydrogen on a seed layer of tungsten initially starts to increase rapidly, followed by a decline and progression into compressive stress, after which point it increases again. Surface roughness was found to increase from approximately 10 nm to 100 nm as films increased from 100 nm to 1000 nm in thickness. It was concluded that surface roughness has a power law dependence on film thickness with an exponent of approximately 0.58. Resistivity was found to decrease with film thickness. The results are attributed to a reduction in grain boundary density with increasing thickness and a concomitant reduction in grain boundary scattering of electrons.

Tensile and compressive stress mechanisms were found to depend in opposite ways on growth rate and deposition temperature. Tensile stress decreases with increasing temperature and increases with increasing growth rate. The initial tensile strength increase was ascribed to formation and relaxation of grain boundaries during formation of a continuous film, whereas ‘coherency’ stresses associated with the enlargement of grains that are constrained by other grains results during the intermediate stages of growth. As grain boundaries relax, this stress mechanism dissipates and tensile stresses may be encountered again. The grain size, morphology, impurity content, and resistivity of the films appeared insensitive to process conditions. Unfortunately, the conditions favoring low tensile stress, high temperature and low total pressure, conflict with conditions that produce good step coverage, low temperature and high total pressure.
 

Silicon Reduction and Properties of Deposited Films

Silicon reduction is a reaction that is not often used in industry due to its self-limiting nature. While it can be used to deposit an initial seed layer of tungsten for further H₂ or SiH₄ reduction, it provides poor conditions for good selectively for subsequent reductions due to the production of SiF₄. Leusink (1994) proposes a rate law for mass transfer limited WF₆ reaction with Si as follows:

s_av.t_f= (Es/6(1-n_s))t²_s(1/R_tf-1/R_o), (3)

where

s_av = thickness averaged biaxial state of stress in a film t_f
Es = Young’s modulus
n_s = Poisson’s ratio
t_s = bending of a thick substrate
1/R_tf-1/R_o = change in substrate curvature due to film.

Reflectivity measurements were done using a tungsten halogen lamp as a light source and dispersing the specularly reflected light by a spectrograph grating of 300 grooves/mm. The intensities of eight wavelengths were measured using a diode array detector. Reflectivity measurements can be used to evaluate the surface roughness of a film. Sheet resistance was measured using a four-point probe. Crystal structure was examined by X-ray diffraction and film morphology was analyzed by high resolution scanning electron microscopy. Impurity concentration and distribution was analyzed with SIMS and Auger electron spectroscopy and electron probe micro-analysis (EPMA).

Results of the in situ wafer curvature analysis showed three distinct regimes of stress development. At depositions done at low temperatures, force per unit width increased with time and became constant. At high temperatures with low WF₆ flow rates, the force per unit width increased with time as long as WF₆ was admitted. At high temperatures with high WF₆ flow rates, the force per unit width first increased and then continuously decreased as a function of time. Results showed that the self-limiting film thickness produced by silicon reduction decreased as temperature decreased but growth rate remained relatively constant in the low temperature regime. In the high-temperature low-flow regime, nearly 100% WSi₂ was produced. In the high temperature, high WF₆ flow rate regime, mostly a tungsten is produced with some b and some WSi₂. He found that initially tensile stress developed, followed by the development of compressive stress. He stipulated that the compressive stress found in films under these conditions were due to homogeneous nucleation and growth of WSi₂. Overall, Leusink found that thickness averaged intrinsic stress in self-limiting W films did not depend significantly on temperature, total pressure, WF₆ inlet pressure or self-limiting film thickness. Since the stress became constant when the self-limiting thickness was reached, the development of stress must occur during film growth. Self-limiting W films analyzed by SEM consisted of tapered columnar grains with domed tops with the column diameter estimated at 30 nm.
 

Summary

The literature shows that each reaction seems to have different factors that contribute to differences in deposited film properties. It is difficult to make this judgement absolutely, however, as the experiments performed using the different chemistries were not equivalent and thus are difficult to compare. There do appear to be some general trends that can be summarized. Silane reduction, it appears that two factors have the largest effect on film properties: substrate temperature and reactant ratio. Low substrate temperature seems to favor a -phase tungsten while high substrate temperature seems to favor b -phase tungsten. Reactant ratios with little WF6 in comparison to silane also seem to favor b -phase formation. For the hydrogen reduction reaction, the factors that seemed to have a notable effect on film properties were temperature and total pressure, with high temperature and low total pressure favoring low tensile stress. The films produced by silicon reduction did not appear to be highly dependent on reaction conditions once the self-limiting thickness was reached.
 

Bibliography

Shon, Y. (1992) "Transient Modeling of Selective Chemical Vapor Deposition of Tungsten During a Via Filling Process," M.S. Thesis, Colorado State University.

Creighton, J.R. (1994) "The Surface Chemistry and Kinetics of Tungsten Chemical Vapor Deposition and Selectivity Loss, " Thin Solid Films, Vol. 241.

Kodas, T. and Hampdon-Smith, M.J. (1994) "The Chemistry of Metal CVD," VCH.

Baker, Kevin, "Heat transfer at low pressure and kinetic study of tungsten chemical vapor deposition utilizing a batch reactor and in situ measurement", MS Thesis, Colorado State University, 1996.

Yeh, W.K.; Chen, M.C.; Wang, P.J.; Liu, L.M.; Lin, M.S., "Deposition Properties of Selective Tungsten Chemical Vapor Deposition." Materials Chemistry and Physics, Vol. 45, 1996, 284-287.

Lo, J.S.; Haskell, R.W.; Byrne, J.G. and Sosin, A., 4^th International Conference on CVD, The Electrochemical Society, Princeton NJ, 74, 1973.

Schmitz, J.E. J.; Buiting M.J. and Ellwanger, R.C., "Study of the SiH4-WF6 Chemistry Used in Selective W Deposition", Tungsten and Other Refractory Metals for VLSI Applications IV, R.A. Blewer and C.M. McConica ed. Materials Research Society, Pittsvurgh, PA, 27-34, 1989.

Shon, Y. (1992) "Transient Modeling of Selective Chemical Vapor Deposition of Tungsten During a Via Filling Process," M.S. Thesis, Colorado State University.

Lee, C.; Im, Y.J.; Lee, J.G., "Adhesion and leakage current characteristics of selective CVD tungsten films on the silicon substrate," Journal De Physique IV, Vol. 3, 1993, 433-440.

Tsutsumi, T.; Kotani, H.; Katayama, T.; Miyatake, H.; Okamoto,T. and Nagao, S., Tungsten and other Refractory Metals for VLSI/ULSI Applications V, S.S. Wong and S. Furukawa ed. Materials Research Society, Pittsburgh, PA, 31, 1990.

Metz, W.A.; Mahan, J.E.; Malhotra, V.; and Martin, T.L., "Electrical Properties of Selectively Deposited Tungsten Thin Films," Applied Physics Letters, Vol. 44, No. 12,

15 June 1984, 1139-1141.

Wulu, H.C.; Gardener, D.; Saraswat,K., "CVD W Film Stress and Calculation of the Effect of Stress on p-n Junction Edge Leakage Current", Tungsten and Other Refractory Metals for VLSI Applications IV, R.A. Blewer and C.M. McConica ed. Materials Research Society, Pittsvurgh, PA, 27-34, 1989.

Leusink, Gert, "Growth and properties of CVD-W films for microelectronic applications: A study on growth-related stress" , PhD dissertation, Delft University, Netherlands, 1994.