New Approaches to Titania and Silica CVD by Michael L

New Approaches to Titania and Silica CVD by Michael L. Hitchman and Sergei E. Alexandrov

he companion article by Mark transmission and distribution systems exists in three crystalline forms— Allendorf1 highlights the and for other precision optics. The coat- anatase, rutile, and brookite—with rutile importance of on-line CVD ings of interest are usually oxides other having the highest refractive index of the coating for flat glass manufac- than SnO2, such as TiO2 and SiO2. This three (2.61 – 2.90). Rutile is therefore the ture. Two smaller, but fast article considers some aspects of the most interesting for many optical appli- growing, markets are those of CVD of these materials. In particular, it cations. It is also the most thermody- Tprecision optics and telecommunica- describes a novel process for the deposi- namically stable form at all tempera- tions. Both these markets are driven by tion of the rutile form of titania at low tures, but interestingly it is the anatase the rapidly developing requirement for temperatures and new approaches to form which grows preferentially at the high throughputs. Current deposition thin film formation of silica by plasma lower temperatures that are often methods rely on PVD techniques that are enhanced CVD (PECVD) without the required for film formation on thermally augmented with cumbersome tooling to need for the use of a vacuum. sensitive substrates. So, for example, move the complex geometries and there- TiO2 films produced by a range of tech- by achieve time-averaged deposition Thin Film Growth of Titania niques from low technology processes over whole surface areas. Such tech- (such as sol gel deposition and thermal niques limit throughput and usually give Titanium dioxide (TiO2) thin films are oxidation), through PVD methods (such a non-uniform deposit on parts of the used in a wide variety of applications as reactive sputtering and ion assisted coated sample, necessitating special because of their outstanding physical deposition), to a variety of CVD systems, alignment within an optic system and and chemical properties. In particular, are invariably either amorphous or crys- thereby incurring significant costs. The the high refractive index of the material talline anatase at temperatures below non-line-of-sight attribute of CVD makes and its excellent transmittance in the vis- 400°C.2,3 Reported temperatures for the the technology ideally suited for coating ible and near-IR spectral regions has led anatase to rutile transition range from complex lens and fiber geometries that to extensive applications for antireflec- 700 to 1100°C4 and while the direct CVD are used in optical telecommunication tion coatings and waveguides. Titania growth of rutile has also been described,

FIG. 1. XRD patterns for (a) SnO2 coated glass and TiO2 films on SnO2 coated glass at (b) 257°C; (c) 268°C; and (d) 300°C.

40 The Electrochemical Society Interface • Summer 2001 it is usually for deposition temperatures These applications usually require butoxide, and water led to the unex- in excess of 500°C.5 Therefore while anatase, particularly with a controlled pected discovery of a low temperature rutile is energetically the most favored degree of crystallinity. So a number of CVD rutile process.6,7 There is nothing form of titania, the fact that anatase is CVD precursors were investigated for out of the ordinary about the type of formed at lower temperatures implies their influence on crystalline quality. deposition system required, with stan- that there is a higher kinetic activation The most widely reported precursors are dard gas handling facilities and a con- barrier for rutile than for anatase. This is titanium halides (e.g. TiCl4, TiCl3, TiI4) ventional hot-walled reactor with a typ- unfortunate because many optical appli- and titanium tetra-alkoxides (e.g. ethox- ical operating pressure of about 1 Torr. cations for TiO2 require low thermal ide, iso-propoxide and the tert-butox- What is remarkable is the transition budget processing to avoid the introduc- ide). The halides have the disadvantage from amorphous material to anatase to tion of stress or damage to temperature of producing hydrogen halide gases as rutile as the substrate temperature sensitive substrates. by-products from the hydrolysis process, increases from 200°C through 250°C to so the alkoxides are an attractive alterna- 300°C. The XRD patterns in Fig. 1 show A Serendipitous Discovery tive. Because an alkoxide contains four this variation for TiO2 deposited on tin oxygens, there is, in principle, enough oxide coated glass; this is a preferred A recent serendipitous discovery has oxygen to form TiO2 without additional substrate for photoelectrochemical led to the first low temperature CVD for- oxygen being needed. Often, though, applications. Of course, SnO2 has a mation of rutile with growth of this crys- further O2 is used to minimize carbon rutile structure itself and it could be talline form on a number of different incorporation into the deposited layers. argued that this is influencing the crys- 6,7 substrate surfaces at 300°C. This sur- An alternative oxygen source is H2O, tallization of the CVD layer. However, prising result came about during a study which has the additional feature of low- as the Raman spectra in Fig. 2 show, of CVD titania growth at low tempera- ering the reaction temperature. This rutile can also be grown on such dis- tures for photocatalytic applications. combination of an alkoxide, the tert- parate materials as steel, aluminum, sapphire and glass, as well as silicon (not shown). On other substrates such as copper and magnesia, though, the CVD layers grown at 300°C are anatase (Fig. 2). This phenomenon of low temperature CVD rutile growth appears to be unique to the combination of titanium tetra-t-butoxide and water since investigations under similar conditions with the ethoxide, iso-propoxide, n- butoxide and sec-butoxide all yield either anatase or anatase/rutile mix- tures. The reasons behind this unexpected result are still not clear. The rup- ture and formation of Ti-O bonds is known to be the rate determining step in the anatase-rutile phase transforma- tion and can be promoted by impurities.8 One could speculate that this is perhaps facilitated in the deposition process from the t-butoxide by more carbon-containing impurities being present in the layer than when other precursors are used; β-hydrogen elimi- nation with the resulting formation of stable hydrocarbons on the surface is likely to be more complex and have a higher activation energy for a t-butoxide than for less branched alkoxides. However, further studies on this point are needed. The role of the substrate is also puzzling because the materials that promote rutile growth have a range of different structures; e.g. an hcp octahe- dral structure for sapphire and possibly for native oxides on iron and aluminum, while silicon is tetrahedral. There seems little likelihood of any epi- taxial effect with those structures, although one must remember that surface reconstruction could alter the surface energies significantly. Glass clearly -1 FIG. 2. Raman spectra of LPCVD TiO2 films deposited on different substrates. Expected frequencies/cm : has no well-defined structure, but it Anatase: 194, 397, 515, 637. Rutile: 144, 235, 448, 612. may be able to catalyse the anatase to

The Electrochemical Society Interface • Summer 2001 41 rutile transition.9 Whatever the reasons duced downstream from the discharge downstream. The highest registered for the low temperature growth of rutile generation region. The results of a com- deposition rate of 300 ± 25 nm min-1 was from titanium tetra-t-butoxide and parative analysis13 of the main features achieved with a TEOS partial pressure of water vapor, two issues are clear. First, of direct and remote PECVD show that 0.2 Torr and an RF power of 400 W. It was while theoretical considerations are of the remote technique provides less inter- shown that the properties of the deposit- value and importance for understand- dependence for many process parame- ed films were comparable to those of ing and predicting CVD chemistry, ters and so there can be better control. thermally grown SiO2 films. there are still unexpected discoveries to For example, growth of SiO2 films by A similar idea based on the use of an be made. Second, low temperature remote PECVD has been shown to yield RF plasma torch has been developed for rutile growth has considerable potential layers with remarkably low concentra- the enhancement of CVD growth of sili- for the production of high refractive tions of contaminants because of this con oxide films using tetramethoxysilane index coatings on thermally sensitive better control of the plasma phase com- (TMOS) as a precursor.15 The reactive cold substrates. We are currently exploiting position.12 Therefore, the technique of plasma was generated between coaxial the new process for such applications. remote PECVD is a very promising one electrodes with an RF source of 13.56 for thin film technology. MHz and an applied RF power of 80-100 Thin Film Growth of Silica However, the use of low-pressure W. With this cold plasma torch, silicon PECVD, whether it is direct or remote, oxide films were deposited on various Like titania, thin films of silicon for very large quantities of optical mate- substrates with growth rates higher than oxide have interesting physical and rials is not a particularly appropriate 600 nm min-1. The carbon content in the chemical properties that make them technique because of the high cost of deposited films was rather high, though, attractive for a wide range of applica- volume vacuum equipment. Because of being in the range 1-7 atomic percent. tions in microelectronics, optics, for the this economic drawback, some relatively Some more basic work has been done protection of metals and polymers, and new CVD processes based on the use of on film formation from TEOS and hexa- particularly as a component of multi- various types of electrical discharges sus- methyldisiloxane (HMDSO) in an layers and graded-index coatings. As tained at atmospheric pressure have atmospheric pressure glow (APG) dis- mentioned earlier, for the latter applica- attracted considerable interest. charge (15 kHz, 50 W) by studying the tions, thermal budgetting is critical. As emission spectra.16 The spectra of the a result, PECVD techniques have Atmospheric Pressure PECVD APG discharges ignited in either a TEOS- attracted considerable interest as they He-O2 mixture or a HMDSO–He-O2 mix- can provide the possibility of deposi- Several attempts have been made to ture were very similar. For both precur- tion of silicon oxide films at substrate use non-thermal plasmas generated at sors with He, but without O2, emission temperatures as low as 100 to 300°C.10 atmospheric pressure for oxide CVD lines corresponding to He, CH and H 14-16 However, it has been shown by many processes. One example is the suc- were observed while with O2 additional workers that a conventional, or direct, cessful use of a plasma jet that operates at lines for OH, CO and O appeared. The PECVD approach, in which a mixture atmospheric pressure and near room intensity of the He line decreased in the of reactant gases is introduced directly temperature for plasma deposition of sil- presence of either precursor or O2. No into the plasma region, has a number of icon oxide films.14 The plasma jet was emission lines related to Si or Si-contain- disadvantages. Thus, process parame- fed oxygen and helium between two ing species were detected. In fact, the ters such as the total pressure, RF power, coaxial electrodes that were driven by a species observed to be present from the gas composition, and gas flow rate are 13.56 MHz RF generator at 40-500 W. spectra were very similar to those found all interdependent and can be difficult Tetraethoxysilane (TEOS) was mixed in conventional low pressure PECVD. to control individually. Also, complex with the effluent of the plasma jet and This led the authors to suggest that the reactions can arise as a result of the directed onto a substrate located 1.7 cm deposition mechanism involving precur- simultaneous production of many different reacting species, some of which can participate in the film formation process, leading to films with unpre- dictable composition. In particular, there is likely to be intrinsic doping of films deposited by direct PECVD. So, for example, films deposited using silane as a precursor generally contain large concentrations of contaminants,11 such as SiHx (x=1-3) and OH groups. A great deal of research of the PECVD of various materials, and especially of silicon oxide films, has been directed toward overcoming some of the drawbacks of direct PECVD processes. One approach, known as remote PECVD, is to separate the discharge generation region and the deposition zone in space and to introduce only some of the initial gases into the discharge generation region for excitation (Fig. 3).12 The other reagents, usually the less stable precursors, are intro- FIG. 3. Schematic diagram of a remote PECVD reactor with inductively coupled RF power.

42 The Electrochemical Society Interface • Summer 2001 sor molecules in the two discharges are tion region. With this technique, high required for film formation. The other the same, although for the APG dis- quality, silicon oxide films have been type uses gaseous fuel and the precur- charge O2 and the precursor are dissoci- deposited at 400-500°C with reasonable sors are introduced into the flame ated by energy transfer from metastable values of growth rate (8-10 nm s-1). either in gaseous form or as a mist. A He atoms by a Penning mechanism. Another promising approach to number of different oxide materials 17 18 Remote PECVD has been successfully atmospheric pressure PECVD coating of including silica, but also Al2O3 and 19 used for deposition of various coatings at glass substrates in the open atmosphere CeO2, have been deposited using this atmospheric pressure (AP). Recently we is flame assisted CVD, or so-called com- process. For example, amorphous silica have developed a remote AP-PECVD sys- bustion CVD (CCVD). A flame is a type coatings were deposited at a tempera- tem for deposition of silicon oxide films of plasma that is characterized by quite ture of 750 ± 50°C using a liquid fuel of based on the use of HMDSO as the sili- low values of electron energies (less than ethanol containing 0.0025 M TEOS as con-containing precursor. Argon and 0.6 eV), electron densities (107-108 cm-3) the silicon containing precursor.17 The oxygen were passed through a region and degrees of ionization (<10-10). The layers were deposited with growth rates where barrier discharge was sustained flux of heat released by the occurrence of of about 0.8 µm h-1 and were shown to with the use of high voltage supply various exothermic reactions is the be suitable for protection of metal operated at a frequency of 50 Hz. source of energy for sustaining the plas- alloys against high temperature corro- HMDSO vapor was carried to the deposi- ma. There are two main types of CCVD sion. While the deposition temperature tion zone by argon and was introduced processes. One type uses a liquid fuel is somewhat high for glass substrates, downstream from the discharge genera- containing the dissolved precursor the results of CCVD studies have,17-19 nevertheless, suggested that, despite the high energy flux directed to the substrate (a few kW cm-2), the technique could be a versatile, inexpensive, atmospheric thin film deposition method for large scale manufacturing applications such as coating glass sam- ples. The flame supplies the heat neces- sary to promote chemical reactions on the substrate surface which must be exposed in the “flame environment” only for short periods of time (typically a few seconds) in order to prevent over- heating of the surface. This then means that for a CCVD process, deposition should be characterized by a high growth rate. Thus one has the concept that the flame must act as the reaction chamber in which the highly reactive species required for film formation are formed from the precursors. Therefore the choice of suitable precursors and the provision of a favorable temperature distribution in the flame are important problems to be solved for the development of such a process. Precursor Chemistry

Precursors should be stable in an oxidative atmosphere and at elevated temperatures as well as participating in reactions resulting in the formation of active species required for the deposition process. So with CCVD, because there is a sharp temperature profile in a flame, the positioning of the injection nozzle for the introduction of the precursor is of critical importance. This is of particular concern for the case of silicon oxide CCVD based on organosilicon precursors because of the very low thermal stability of these compounds. This instability can be illustrated by the results of thermodynamic calculations for the Si-C-O-H system when HMDSO and O2 are used as the initial reactants FIG. 4. Equilibrium gas phase composition in the system Si-C-O-H. (Fig. 4). As can be seen, silicon oxide is

The Electrochemical Society Interface • Summer 2001 43 the thermodynamically stable phase at methoxysilane (MTMOS), H3Si(OCH3)3, of silicon oxide films. It is thus clear temperatures below 2000 K. At higher and trimethylmethoxysilane (TMMOS), why TEOS and HMDSO are the most temperatures silicon monoxide and (CH3)3 SiOCH3); group 4: ethoxysilanes widely used precursors for PECVD of sil- atomic silicon prevail over other silicon such as tetraethoxysilane (TEOS), icon oxide films. containing species. The concentration Si(OC2H5)4, ethyltriethoxysilane of methane—one of the most stable (ETEOS), C2H5Si (OC2H5)3, dimeth- Conclusions hydrocarbons—sharply decreases with yldiethoxysilane (DMDEOS), (CH3)2 increasing temperature above about 500 Si(OC2H5)2, and trimethylethoxysilane CVD is currently not extensively K. One can also note the increase in (TMEOS), (CH3)3 SiOC2H5). used for coating optical components. As concentrations of the elements in atom- A comparative study of the use of the market demands grow, though, and ic form and of radicals (OH, HO2) with plasmas for oxygen/organosilicon pre- costs need to be kept down one can temperatures above 3000 K. Most signif- cursors representative of groups 2-4 expect to see an increasing use made of icantly, the equilibrium concentrations above has been made for remote CVD in order to take advantage of the of silicon-hydrogen and silicon-oxygen- PECVD.23 The mechanisms of deposi- non-line-of-sight possibilities it offers hydrogen species are negligible at tem- tion are not known and seem to be for increasing deposition throughput peratures higher than 1000 K. extremely complex. In order to compare and also for coating complex substrate For CCVD, molecular chemistry is the various precursors, the dependen- geometries. In addition, one must never obviously a major factor in choosing cies of conversion ratios per substrate forget the C in CVD and the potential the precursor. This can be equally true area on the so-called Remote Composite that has for developing novel processes. for other PECVD processes because of Parameter (RCP) [W cm-1] has been The further feature of deposition at low the complex reactions that occur either studied. This parameter reflects the temperatures by the introduction of in the plasma or downstream from the combination of the microwave power plasma sources and the economic discharge region. As a result, one of the absorbed in the plasma and the distance attraction of not having to use vacuum main research directions for the CVD of of the substrate from the center of the systems by exploiting atmospheric pres- silicon oxide films has been to find the plasma source. It can be interpreted as sure discharges strongly suggests that best silicon containing precursors to the critical flow rate of excited oxygen CVD will make a significant contribu- provide higher growth rates and species coming from the discharge gen- tion to the burgeoning areas of telecom- improvements in film quality. Silicon eration region for complete reaction munications and precision optics. containing precursors for PECVD of sili- with the monomer.23 A monomer mol- con oxide can be divided into two ecule that needs more oxygen atoms for References major groups: silicon containing inor- a high conversion ratio should show a ganic compounds (e.g. silicon halides, higher RCP. It was shown that there was 1. M.D. Allendorf, Interface, Vol. 10, No. 2, p. chlorosilanes, and silane)10-12 and sili- an increase of RCP with decreasing 34 (2001). con containing organic compounds (e.g. number of alkoxy groups: 2. N. Rausch and E. P. Burte, J. Electrochem. organosilicon monomers such as hexa- Soc., 140, 145 (1993). 20 3. P. Vaija, J. Lahouidak, and J. Durand, Euro. methyldisiloxane (HMDSO), tetra- RCPcritTMOS < RCPcritMTMOS < Ceram. Soc., 11, 551 (1993). 21 ethylorthosilicate (TEOS), and tetra- RCPcritTMMOS 4. G. Skinner, H. L. Johnston, and C. Beckett, methylsilane (TMS).22 Titanium and Its Compounds, p. 21, Johnston Enterprises, Columbus, Ohio, The main advantage of silane as an RCPcritTEOS < RCPcritETEOS < (1940). inorganic precursor with weak Si-H RCPcritDMDEOS < RCPcritTMEOS bonds is that there is the possibility of 5. M. Ritala, M. Leskela, E. Nykanen, P. Soininen, and L. Niinisto, Thin Solid Films, depositing high quality silicon oxide Other authors24,25 have found as well 225, 288 (1995). layers at quite low temperatures and that precursors with a high number of 6. M. L. Hitchman and J. F. Zhao, J. Phys. IV with reasonable growth rates. However, alkoxy groups require less excited oxy- France, 9, 357 (1999). for an atmospheric pressure PECVD sys- gen species for deposition of good qual- 7. M. L. Hitchman and J. F. Zhao, International Patent Application, tem, the design has to take into account ity silicon oxide films. It has also been the high flammability and explosive PCT/GB00/00432 (1999). shown that while SiO-CH3 bonds are 8. R. D. Shannon and J. A. Pask, J. Am. Ceram. nature of silane. Therefore, for CVD of easily broken, even without oxygen, the Soc., 48, 391 (1965). silicon oxide films based on the use of a scission of Si-CH3 bonds requires oxy- 9. J. H Jean and S. H. Lin, J. Mater. Res., 14, thermal plasma or with combustion gen radicals; this conclusion is consis- 2922 (1999). CVD, for non-electronic applications, tent with the comparative bonding 10. A. C. Adams in "Plasma Deposited Thin Films," F. Jansen and J. Mort, eds., CRS, New such as the coating of glass substrates, energies. Because of this, TMS is not silicon halides and chlorosilanes are York (1986). widely used for PECVD of SiO2 films 11. Y. B. Park, J. K. Kang, and S-W. Rhee, Thin probably the most promising precur- and, in particular, it is not a promising Solid Films, 280, 43 (1998). sors, especially from an economic point precursor for remote PECVD. Thus, the 12. G. Lucovsky, D. V. Tsu, R. A. Rudder, and R. of view. results of the RCP studies demonstrate J. Markunas, in Thin Films Processes II, p. Several groups of the most generally that alkoxy groups have improved reac- 565, J. L. Vossen and W. Kern, eds., Academic Press, London. used organosilicon precursors can be tivity compared to alkyl groups and that identified. Group 1: alkyl silanes such as 13. S. E. Alexandrov, J. Phy. IV, C5, 567 (1995). monomers with ethoxy groups form sil- 14. S. E. Babayan, J. Y. Jeong, V. J. Tu, J. Park, G. tetramethylsilane, Si(CH3)4; group 2: icon oxide films more readily than S. Selwyn, and R. F. Hicks, Plasm. Sourc. Sci. disiloxanes such as hexamethyl- those with methoxy groups. The RCP Technol., 7, 286 (1998). disiloxane, (CH3)3SiOSi(CH3)3 and studies also showed that the transfer of 15. K. Inomata, H. Ha, K. A. Chaudhary, and H. tetramethyldisiloxane (TMDSO), (CH3)2 excited oxygen species from the dis- Koinuma, Appl. Phys. Lett., 64, 46 (1994). 16. Y. Sawada, S. Ogawa, and M. Kogoma, J. HSiOSiH(CH3)2); group 3: methoxysi- charge generation region to the sub- lanes such as tetramethoxysilane Phys. D: Appl. Phys., 28, 1661 (1995). strate surface can be identified as the 17. J. M. Hampikian and W. B. Carter, Mat. Sci. (TMOS), Si(OCH3)4, methyltri- rate limiting step in the remote PECVD Eng., A267, 7 (1999).

44 The Electrochemical Society Interface • Summer 2001 18. M. R. Hendrick, J. M. Hampikian, and W. B. Carter, J. Electrochem.Soc., 145, 3986 About the Authors (1998). Sergei E Alexandrov has been a professor at St. Petersburg State Technical 19. W. B. Carter, G. W. Book, T. A. Polley, D. University in the Department of Electronic Materials Technology since 1995. His W. Stollberg, and J. M. Hampikian, Thin research activities involve the development of a range of CVD processes including Solid Films, 347, 25 (1999). atomic layer depositions, plasma enhanced CVD systems, in situ diagnostics of CVD 20. D. Korzec, D. Theirich, K. Traub, F. reactions, and low temperature methods for deposition of various nitride, oxide, and Werner, and J. Engemann, Surf. Coat. oxynitride dielectric layers. He may be reached at [email protected]. Technol., 74-75, 67 (1995). 21. C. S. Pai and C. D. Chang, J. Appl. Phys., Michael L Hitchman graduated with Honours in Chemistry at Queen Mary College, 68, 793 (1990). London in 1962. He studied for a further five years at the University of Oxford, first 22. A. C. Adams, in Reduced Temperature Processing for VLSI, R. Reif and G. R. as a DPhil student then as an ICI research fellow. This was when he learned about Srinivasan, Editors, PV 86-5, p. 111, The homogeneous kinetics, heterogeneous kinetics, mass transport, rotating disc electrodes, Electrochemical Society Proceedings and the wonders of electrochemistry. He spent six years at the RCA laboratory in Series, Penninngton, NJ (1986). Zurich where, after discovering the limitations of electrochemistry for electrochromic 23. Ch. Bayer, E. Bapin, and Ph. R. von Rohr, displays, he turned his attention to another system involving homogeneous and Surf. Coat. Technol., 118-119, 874 (1999). heterogeneous kinetics and mass transport, namely CVD; and found that rotating discs 24. Y. Inoue and O. Takai, Plasma Sources Sci. proved as useful and powerful for CVD as for electrochemistry. In 1979, he become a Technol., 5, 339 (1996). lecturer at the University of Salford and in 1984 he was appointed to 25. R. Delsol, P. Raynaud, Y. Segui, M. the Young Chair of Chemical Technology at the University of Strathclyde. He may be Latreche, L. Agres, and L. Mage, Thin Solid reached at [email protected]. Films, 289, 170 (1996).

The Electrochemical Society Interface • Summer 2001 45