METHOD OF APPLYING A COATING MATERIAL TO A SUBSTRATE , COATED SUBSTRATE AND APPARATUS
This invention relates to the fabrication of thin films of material with particular, though not exclusive, application to the production of glass optical waveguides.
The technique of spin coating provides a convenient, low cost and reproducible process for the fabrication of thin films for which a solution of the material to be deposited is prepared and applied to the surface of a rotating substrate. As the substrate rotates over time, excess material is first thrown off in the initial spin-up followed by a reduction in the film thickness which is generally found to be independent of the initial fluid volume deposited.
In the 1950s, Emslie et al investigated the mathematical description of the flow of a Newtonian fluid on a rotating disk in order to better understand the production of television screens prepared by the rotational casting of a phosphor loaded slurry [1]. By equating the centripetal and viscous forces acting at any point within the coating fluid, the final film thickness was found to be a function of the material parameters via the density and viscosity of the fluid and upon the external parameters through the spin speed and spin time.
The final film thickness h, was found to obey the expression:
1 RT Equation 1. h = _ X / — ω y pt ω: angular velocity η: fluid viscosity p: density t: spin time
Using the general expression given in equation 1, one can see that as the viscosity increases, the time required to produce a layer of a given thickness increases to a point where the time becomes incompatible with the volatile solutions used. As a consequence, spin coating is often undertaken using low viscosity solutions of the
order of 1 poise with relatively high spin speeds of the order of several thousand rpm. This region of mateπal parameters is usually satisfied by photoresist materials and other solids, which are dissolved in solvents of the correct viscosity. The past four decades have seen the application of spin coating to many areas such as paints and varnishes, photoresists for the microelectronics industry [2], and sol-gel glasses [3] and polymers [4] for optical waveguides. Theoretical work has continued to understand the effect of evaporation for materials dissolved in volatile solvents [5] and the principles involved have been implemented both through dip [*] and spin casting to generate low loss sol-gel and polymer waveguides.
When considering the direct spin coating of glasses, molten phases of soft glasses such as fluorophosphates, fluoroaluminates, chalcogenides, sulphides and germanates exhibit viscosities of 1 poise or below at their liquidus temperatures but until now have not been considered for direct spin coating because of several limitations:
1. Low viscosity glasses by their very nature tend to be unstable and susceptible to crystallisation. As a consequence, the glass must be quenched quickly from its molten phase to ensure crystals do not nucleate in the film.
2. In order to exploit the conventional technique of spin coating, a substrate material would be required which could withstand the temperatures at which the glass was molten. Furthermore, it must be matched thermally for expansion coefficient to prevent cracking on cooling and, for application as a waveguide, have a lower refractive index than the molten glass material.
3. Multilayer spin coating is not possible because the preceding spin-coated layer would be remelted on application of further molten glass.
4. Chemical reaction can occur between substrate and glass if the time for which they are in contact at high temperature is too long e.g. if the glass is heated from solid to molten whilst on the substrate.
5. The viscosity of glasses show a strong temperature dependence such that temperature variations across the substrate can potentially lead to film thickness non-uniformities.
3 Spin coating is both the cheapest and quickest of the many thin film deposition techniques such as sputtering [*], evaporation ffl, liquid phase epitaxy (LPE), pulsed laser deposition (PLD) [1 ], molecular beam epitaxy (MBE), flame hydrolysis deposition (FHD) tø, plasma enhanced chemical vapour deposition (PECVD), thermal oxidation and ion exchange [n] but until now has not been applied to molten glass. Costly vacuum systems or material precursors can be neglected and, for polymers and sol-gel waveguides produced thus far, film quality and uniformity certainly match that generated using other more expensive processes. Moreover, many of these techniques are unable to offer film thickness greater than a couple of microns: a task which is easily performed in spin coating by varying the viscosity of the coating material and/or the spinning parameters. As a consequence, it offers considerable advantages for the large scale mass-production of optical devices.
Unfortunately, polymers and sol-gel glasses are often questioned for their long term environmental stability and compatibility with fibre optic systems. Furthermore, for active devices, multi-component glasses are often required incorporating active ions such as neodymium and erbium for applications as lasers and amplifiers. In particular, such glasses have found applications in optoelectronics where a range of components is required to tailor the spectroscopic or material properties of the glass for a device. For example, in the case of the 1.3 μm amplifier, the position of peak gain is particularly sensitive to the composition: one must categorically ensure that the stoichiometry of the bulk glass is reproduced in thin film form.
Traditional high energy techniques such as sputtering and evaporation can produce non-stoichiometric films due to the differences in the vapour pressures of the various species in the glass source. In contrast, pulsed laser deposition (PLD) exploits high intensity excimer irradiation of the glass target for which heating rates are so high that the bulk glass is vaporised in a shorter time than that necessary for non-congruent evaporation to occur. For the glass composition we have used for spin coating, bulk composition was indeed duplicated by PLD, but the film quality formed from the vapour plume was generally poor and typified by micron size particles. One is apparently faced with a compromise: high quality films may be produced but the spectroscopy of the glass will be sacrificed. In contrast, PLD faithfully reproduces the glass stoichiometry but the poor topography of the films is
4 unacceptable. PLD appears to be the most promising technique for deposition from
'vapour' but one can also consider deposition from the intermediate molten phase of the glass for which the stoichiometry is inherently preserved when in the liquidus phase. Hence, spin-coating provides an alternative deposition technique for those materials where stoichiometry must be maintained.
This invention provides a method of depositing a layer or film of a glass or other material onto a substrate, the method comprising one or more of the steps of heating the substrate; dipping the substrate into a reservoir of the molten glass or other material; and spinning the substrate. Preferably the substrate is an optical waveguide such as an optical fibre. Preferably the molten material is a glass. Preferably the substrate is heated to a temperature under the liquidus temperature.
The invention also provides a coated substrate (preferably a waveguide and in particular an optical fibre) coated in this manner.
The invention also provides optical apparatus using such a waveguide. Embodiments of this invention can provide a technique for the fabrication of thin films from molten material.
Embodiments of the invention will now be described by way of example only and with reference to the accompanying drawings, in which:
Figure 1 is a schematic diagram illustrating stages in a spin coating process according to embodiments of the invention;
Figure 2 is a schematic diagram of a furnace arrangement; and
Figure 3 is a schematic diagram of a spin coating section of the furnace of
Figure 2.
Figure 1 is a schematic diagram illustrating stages in a spin coating process according to embodiments of the invention.
A glass substrate is preheated to a given temperature (usually close to the glass transition temperature but applicable at any temperature below the liquidus temperature) after which, at stage 1 of Figure 1, it is dipped, face-down into a bath of molten glass. At stage 2, the substrate is spun on withdrawal from the melt which helps to remove excess glass and thins down the deposited film. On contact with the molten glass the interface between substrate and film can reach a temperature above
the fusion temperature of the glasses for which reflow of the glass at the interface can occur.
At stage 3, the coated substrate is cleaved to provide the required article such as an optical waveguide. Example stage 1 process conditions for the deposition of fluoroaluminate glass waveguides are: preheating the substrate to 430 degrees C and the molten glass to
1000 degrees C.
In an analogy to the fibre drawing process of optical fibre waveguides, imperfections at the glass interface are smoothed out as the spinning action deforms both the waveguide and substrate glasses. Multilayer structures may be fabricated by the repeated application of the spin coating process with the possibility of burying and adjusting the original film thickness layer by the reflow of material during the subsequent spin coatings. Annealing of films can be achieved during the cool down following spin coating or subsequently. These techniques have application to the production of glass optical waveguides and devices including, but not limited to, lasers, amplifiers, splitters, routers, modulators and sensors. This might include, for example, the inclusion of dopants in a glass film such as rare earth ions or transition metal ions for application as active gain media and/or photosensitive dopants such as Sn, Ce, Eu, Tb ions for gratings or channel definition. Fabrication of channel waveguides and patterned features can be achieved before or after spin coating or even between multiple layers by the application of patterning techniques such as, but not restricted to, chemical etching, ion beam milling, ion exchange, direct UN writing or micromachining. The fabrication of thin films in this manner could easily be applicable to other functions utilising spin or dip coating such as, but not confined to, coating of laser rod endfaces, protection of integrated circuits and fabrication of photonic band gap devices. Furthermore, the technique is not limited only to glasses but could also be applied to the fabrication of thin films of other materials including, but not limited to, metals and polymers. Finally, any process could perform quenching of material to form a thin film where a cooling material is subjected to a thinning force for example vertical dip coating.
Overcoming the perceived limitations of conventional spin coating techniques
The process described above addresses each of the following numbered limitations which have previously been determined for a conventional spin coating process using molten glass:
1. Low viscosity glasses, by their very nature, tend to be unstable and susceptible to crystallisation. As a consequence, the glass must be quenched quickly from its molten phase to ensure crystals do not nucleate in the film.
This perceived disadvantage is overcome as follows. As the molten glass is coated onto a cooled substrate, close to the glass transition temperature, the glass is rapidly quenched through the crystallisation temperature in a time of the order of 500ms to 2s. In prototype structures examined using a scanning electron microscope (SEM), the interface is found to be free from crystals. The fast cooling rates achievable imply that glasses with small differences in glass transition temperature (Tg) and crystallisation temperature (Tx) can be produced. Hence, the technique is very appealing to the deposition of unstable glasses. In principle a wide range of viscosities can be used provided there is no substantial evaporation of the melt or compositional change.
2. In order to exploit the conventional technique of spin coating, a substrate material would be required which could withstand the temperatures at which the glass was molten. Furthermore, it must but be matched thermally for expansion coefficient to prevent cracking on cooling and for application as a waveguide, have a lower refractive index than the molten glass material.
In the present embodiments, thermally compatible materials can be used for the substrate and the film. Consequently/ the two are thermally matched and no cracking is observed on cool down. In fact, by choosing a material for the substrate of similar melting point, the deformation of the substrate is advantageous in reflowing imperfections at the interface and lowering the losses of the waveguide. By using an inverted configuration, one prevents the molten glass from instantaneously quenching
on the substrate as the interface effectively 'sees' an infinite body of molten glass. One ensures that the underlying substrate does not melt away, by limiting the contact time of the melt and substrate. In contrast, if one was to pour molten glass onto the cooled substrate, as in conventional spin coating, the first droplets of the glass would solidify on the much larger thermal mass of the substrate like hot wax spilling from a candle.
3. Multilayer spin coating is not possible because the preceding spin-coated layer would be remelted on application of further molten glass
As mentioned in the reasoning of point (2) above, multilayer processing is easily realisable using the inverted configuration of the present embodiments. On contact, the interface of the molten glass and substrate reaches a temperature generally midway between that of the bulk of the substrate and the molten glass. This temperature is sufficient to lower the viscosity of the first deposited layer to a point where it can be deformed but not remelted. In this manner, computer simulation using a package known as "POLYFLOW" has revealed that previous layers can be reflowed thereby reducing the film thickness (and hence achieving single mode operation if embodied as a waveguide) whilst reflowing inhomogeneities and improving the propagation losses of the structure. This has been confirmed by prototype experiment through the fabrication and SEM examination of an overclad planar waveguide.
4. Chemical reaction can occur between substrate and glass if the time for which they are in contact at high temperature is too long e.g. if the glass is heated from solid to molten whilst on the substrate.
In the present embodiments, chemical reaction is not significant during the contact times of perhaps a fraction of a second.
5. The viscosity of glasses show a strong temperature dependence such that temperature variations across the substrate can potentially lead to film thickness non-uniformities.
Using the inverted configuration of the present embodiments, the substrate
'sees' a semi-infinite volume of fluid glass for which temperature variations have not been observed to contribute. The strong temperature dependence of the viscosity is actually advantageous since the film thickness can be adjusted by varying the viscosity via the temperature of substrate and melt.
Specific example
A specific example will now be described in which the technique was successfully applied to the example of a neodymium doped fluoroaluminate glass. Fabrication of a low loss (<0.5dB/cm) waveguide both in single layer (waveguide and substrate) and multi-layer format (overclad layer/waveguiding layer/substrate) was achieved.
The spin coating apparatus used is shown schematically in Figure 3 and Figure 4. Two furnaces 10, 20 are mounted vertically in a mounting frame 30, one furnace generally above the other. The lower, melting furnace 20 is used to melt the doped 'film' glass at temperatures up to 1000°C while the upper, annealing furnace 10 preheats the substrate and post-anneals the deposited film at temperatures in the range of the glass transition (Tg = 430°C). The substrate is secured on a rotating spindle 40 via a boron nitride vacuum chuck 50, connected to a vacuum line 55. Gas lines 60 and extracts are fitted to the system for the introduction of inert gases and the extraction of any harmful fumes.
The molten glass (the "melt") is held in a platinum crucible and holder 100. A ball-screw driven translation stage 70 is used to transfer the melt between furnaces and a brushless servo motor 80 drives the spinner shaft. The stage considered can accelerate in excess of lg, has a maximum velocity greater than 0.2m/s and a positional accuracy of 1 micron. A high performance brushless motor can provide directly driven spinning speeds of όOOOrpm with acceleration times of less than 30ms. The possibility exists to monitor, for example using a general purpose computer 90, when the substrate has contacted the surface of the glass using the conductivity of the molten material. This can use a simple electromechanical sensing arrangement - for
example two inert probes (not shown) can be electrically connected when they contact the electrically conductive fluid glass.
The apparatus is generally sealed by an insulating cap 110 and a connecting iris
120 between the furnaces.
Characterisation of films
Neodymium doped fluoroaluminate glass films have been fabricated with a thickness of less than 20 microns in an unclad (single layer) configuration and less than 5 microns when in an overclad (double layer) configuration.
A. Single layer
Examination of the cross-section of films deposited using this technique shows excellent film uniformity. Scanning electron microscopy reveals a substantially crystal free interface, using compositional contrast imaging, which has proved difficult to produce in a fibre configuration. A low-loss neodymium doped waveguide laser has been demonstrated using the technique. Laser operation at 1050nm has been observed in a 20 micron thick neodymium doped glass waveguide that was fabricated using this technique. Using a nominal 3.5% output coupler at 1050nm, 5mm long waveguides pumped by a TLsapphire source at 800nm have exhibited an incident slope efficiency of 10% and a minimum incident threshold of 50m W. Data is consistent with propagation losses of the order of 0.5dB/cm measured by imaging the scattered light from the waveguides and suggests that losses in the waveguides were dominated by the bulk losses of the material.
B. Multilayer
Recently, the results of computer simulation of the spin coating process have been reproduced in practice, as confirmed by SEM examination. A 50 micron thick film was originally deposited by spin coating a Nd3+ doped layer of glass as per the single layer process. The film was then successfully reflowed to a thickness of 5
microns by spin coating an un-doped cladding layer of glass on top of the 50 micron film.
Again, lasing was observed in the neodymium doped guiding layer. Propagation losses have been measured to be less than 0.5dB/cm, which is consistent with the bulk losses of the material.
Multilayer structures also have the advantage of being able to use different materials (of differing material and chemical properties) for each of the layers improving the functionality of a device. For example, a stack of glass films could be fabricated with one layer doped with Nd3+ for application as a 1.3 micron optical amplifier and one layer doped with Er3+ for application as an amplifier at 1.5 microns all on the same device. The same idea can be applied to other devices including, but not restricted to, multiple wavelength sources, cross-coupling between layers and WDM devices.
Processing parameters
Several factors have been determined, by both experiment and computational fluid dynamics, which influence the final film thickness and uniformity.
These include, but are not restricted to:
1. soin speed and acceleration time. For the particular fluoride glass system we have considered, embodiments of the invention typically employ spin speeds of 1000- 6000rpm to achieve film thicknesses of the order of 20 microns. Depending on the material properties of other materials such as heat capacity, conductivity and viscosity, one could envisage employing spin speeds over a wide range from lOOrpm to 10,000rpm.
2. the temperature n melt and substrate.. In principle, any range of substrate temperatures below the liquidus can be considered. However, it is advantageous to choose the temperatures such that appreciable deformation of the substrate does not occur and no crystallisation at the interface is observed. Furthermore, one can adjust the temperature of the melt to lower temperatures close to the fusion
temperature where the melt of glass may be contacted and quenched on the surface of the substrate.
3. the material viscosity/temperature dependence. For the fluoride glass system used, the viscosity is a strong function of temperature with a viscosity of less than 1 poise above its liquidus temperature of 900°C.
4. contact-time of melt and substrate. A contact time sufficient for full and complete wetting of the substrate by the molten film is required. For the example of the fluoroaluminate glass system, typical contact times of the order of several milliseconds to several seconds have been employed. For other materials, this is determined by the film thickness required and the particular material properties, such as, but not exclusive to, heat capacity and thermal conductivity.
5. dip depth of substrate into melt. The influence of dip time, contact time and dip speed are all inter-related for the reasons given in point 4. For example, if one increases the dip depth, then for the same dip speed, the contact time will be increased.
6. vertical speed at which the substrate is dipped into the melt. See point 5.
7. melt/substrate separation during spin coating. The distance above the melt during spin coating is determined by the requirement that an appreciable quenching rate is required for crystal free solidification. However, for the fluoride glass system, we have demonstrated that deposited films can be thinned further by holding the as deposited film several millimetres above the melt whilst continuing to spin the substrate. This is because the heat generated by the melt continues to heat the glass film, thus reducing its viscosity and allowing further film thinning to occur.
8. melt volume, thickness and density. The melt can represent a considerable thickness of glass in comparison to the substrate when one is requiring a large thermal mass to contact the substrate to prevent rapid cooling of the molten glass on contact. In contrast, a thin layer can be used when one wishes the melt to quickly quench on the surface of the substrate.
9. substrate volume, thickness and density. Substrate volume is important in determining how quickly the molten glass will cool on the substrate. Small substrates exhibit a low thermal mass so that it heats up quickly which can be advantageous if flow of the substrate is required for thinning down the film. In
contrast, a thick substrate might be employed in an example where patterned features have been made in the substrate or first spin coated film. In this case, one wants minimal deformation of the film or substrate so a large, cool thermal mass is used to contact the melt so that the melt/substrate interface cools quickly. 10. angling of the substrate with respect to the spinning axis to prevent the formation of a trapped air bubble. Spin coated films have been observed with a central uncoated region caused by a trapped air bubble forming at the melt/substrate contact point during initial contact. By angling the surface of the substrate one provides an escape route for the trapped gas thus leading to an improvement in film uniformity. In addition, surface tension can be used advantageously in order to
'bow' the surface of the melt and prevent the formation of an air bubble on contact. This may require the use of small melt volumes in order for such surface tension features to be observed.
11. Substrate circumference cross-section. For example, chamfering of the substrate edge allows the molten glass to be spun off the substrate more easily leading to larger, more uniform areas of film.
12. Subsequent overcoating and reflowing of the original film by repeated application of the coating technique with materials of identical or different properties for example, for the case of glasses: chalcogenide glasses, phosphates or fluorides.
PT ΓRT TΓ ATTON REFERENCES
1. A.G. Emslie et al, Journal of Applied Physics 29 (5) (195S) 2. D. Meyerhofer Journal of Applied Physics 49 (7) ( 1978) pp. 3993-3997
3. X.M.Du et al, Journal of the American Ceramic Society 78 (8) (1995) pp. 2254- 2256
4. F.L.Givens et al, Journal of the Electrochemical Society 126 (2) (1979) pp. 269-272
5. C.J. Lawrence Physics of Fluids 31 (10) (1988) pp. 2785-2795 6. D.W. Hewak etal, Applied Optics 27 (21) (1988}pp.4562-4564
7. G.R.J. Robertson et al, Applied Optics 30 (3) pp276-278
8. E. Doran et al. Journal of Applied Physics (1997) 81(2) pp679-684
9. C.N.Afonso et al, Applied Surface Science 96-9^(1996) pp.760-763 10J.R.Bonar et al, Electronics Letters 30 (3) pp. 1 l.R.NRamaswamy et al, Journal of Lightwave Technology 6(6) pp.984-1002