ONLINE EXCLUSIVE: Thin Film on Glass Research at Cornell University
Glass-related research began at Cornell in 1996, with the formation of the interdisciplinary research group (IRG) Thin Films on Glass as part of Cornell's Center for Materials Research (CCMR). IRG members were drawn from physics, applied and engineering physics, chemical engineering, theoretical and applied mechanics and materials science. It formed part of the CCMR until 2003, when the center changed its research direction.
An important component in the investigation of the interaction(s) between glass and thin films is the availability of glass with a reproducible, stable and flat surface. At the time, the only glass meeting this requirement was Corning Code 1737, developed to have such properties for the fabrication of flat panel displays. The drawback of this selection, a fixed and multi-component composition, was repeatedly pointed out in National Science Foundation (NSF) reviews by panels that did not always appreciate the difficulties of producing reproducible glass surfaces. In response, however, the IRG switched in 1998 to glasses in the system CaO-SiO2-Al2O3, a model system for glasses containing column II, III and IV oxides. The subset of glasses along the joint [CaO•Al2O3]x[2SiO2]1-x is pseudo-binary, rendering it a particularly good model glass. Glasses with nine different compositions were made in cooperation with Adam Ellison, Ph.D., at Corning Inc.'s experimental glass making facility.
Using these glasses, the IRG systematically studied the interaction of thin films with glass as a function of composition. Topics studied include the nucleation and growth of silicon films, the adhesion of Cu films, mass transport at glass surfaces, and the modification of surface layers by annealing. In addition, the IRG developed the first scanning tunneling microscope (STM) for insulators. The instrument uses the optical injection of carriers by UV to generate the conductivity required for STM and permits a spectroscopic analysis of the electronic states of glass surfaces.1
Currently, the IRG investigates four subjects. The first is an experimental and theoretical study of the factors controlling adhesion of metal films to glass. This topic has a long history, but models with predictive power have yet to be achieved. The second topic is impurity transport between semiconducting films and glass substrates-a subject that is relevant to flat panel displays and thin-film solar cells. The third topic is an investigation of the modification of near-surface properties when glass is annealed in air containing different moisture levels. This subject is merging with the fourth subject, the investigation of mass transport on surfaces patterned on the nano-scale by electron beam lithography and ion bombardment.
Since there is not enough space here to adequately describe all of these topics, this article will focus only on a subset of thin film on glass research to demonstrate how the interactive nature of the research carried out by the group has led to unexpected findings relevant to both the integrated circuit (IC) and glass industries.
The Effect of CleaningThe majority of thin films are applied to cleaned rather than to as made glass. Thus, the question arises: To what extent does cleaning change the surface? The topic is particularly relevant to the fabrication of active displays, where glass substrates are frequently cleaned with variations of the "RCA" cleaning process originally invented to clean silicon.2 This process is not one most glass scientists would choose. However, the semiconductor industry is averse to introduce new cleaning chemicals, traces of which might enter the processing, and had previously used the RCA process to clean other materials, such as GaAs.
Interestingly, some IRG members, before joining the group, had already found that RCA cleaning 1737 glass without the final hydrogen fluoride (HF) dip improved the electrical characteristics of polycrystalline thin film transistors to the point where they were better than controls made on oxidized silicon wafers.3,4 This amazing finding was not understood, but it was suspected that cleaning somehow changed the impurity transport between the film and glass substrate.
To gain a better understanding of this phenomenon, researchers used total external X-ray reflection, a technique that probes near surface layers by interference and can detect the presence of nanometer thick films. The technique is similar to probing for the presence of an optical coating on a glass by measuring the reflectivity as a function of wavelength. Because X-rays have a shorter wavelength than light, nanometer layers can be measured and characterized.
Figure 1 shows the measured wavelength-dependent reflectivity. A data fit indicates that the clean generated a 6 nm thick surface layer having a 22% lower electron density than the bulk.
Transistor characteristics had implied that this layer was "non-porous." This posed a problem, as leaching was expected to leave a network of interconnected pores. However, high-resolution scanning transmission electron by another IRG member showed that the surface layer was indeed free of pores. Time of flight secondary ion mass spectrometry (SIMS) analysis by an industrial collaborator showed that the leached layer was nearly pure SiO2.5 This explained why a final HF dip destroyed the electrical benefits, as HF readily dissolves such a layer.
Influence on Impurity TransportAlthough the chemical and physical nature of the layer was now known, it was still unclear how it influenced transistor performance. Indeed, it was a mystery how the exceedingly thin layers formed at much shorter cleaning times used in the electronic study could have any influence on impurity transport. Fortunately, the IRG included an expert in tracer diffusion who designed an ingenious experiment to measure the Na transport across these nanometer thick layers. His measurements showed that the layers had a Na diffusivity approximately four and a half orders of magnitude lower6 than bulk 1737 glass,7 also measured by tracer diffusion-a very unexpected finding, that did, however, explain much of the transistor data.
The explanation as to why the Na tracer diffusion was so slow came from studies on the near-surface transport of glasses annealed in dry and wet air, which showed that the presence of moisture greatly reduced the diffusivity.8 Correlation of these findings with Fourier transform infrared spectroscopy (FTIR) proved that the reduced diffusivity was due to an increase in the near surface concentration of OH-. The finding implies that the leached SiO2 formed by RCA cleaning contains large amounts of OH-. Because the layers are exquisitely thin, the presence of residual OH- could not be confirmed directly, but its occurrence is consistent with the chemistry forming the layer.
This finding, in turn, triggered an investigation of Na transport in SiO2 films deposited by chemical vapor deposition (CVD), as such films are known to contain large amounts of hydrogen. Such CVD oxide films are used widely in the IC and flat panel display industry. Na transport in these films had not been measured previously, as the common assumption was that it would equal that in thermal oxide films. However, tracer measurements on CVD films deposited at the Cornell Nanofabrication Facility showed, as expected, a very low diffusivity for Na. CVD oxide films, therefore-unlike thermal oxide layers-can impede Na transport between glass substrates and thin film electronics.6 They might retard the diffusion of other impurities as well, since the effect seems to be due to the formation of a denser network.
In any case, the observation of reduced alkali diffusion in CVD films provided the explanation for why an optimum thickness exists for CVD oxide films used as "buffers" to separate poly-Si thin film transistors from glass. Display glasses containing Al2O3 are excellent getterers for Na, a potent contaminant of polycrystalline silicon transistors. When the buffer thickness exceeds about 1 mm, it begins to adversely interfere with the beneficial gettering of Na.9
Seasonal VariationsThe display glass industry has known for some time that glass drawn at various times of the year can exhibit different surface properties. Groups adjusting the near-surface refractive index of glass substrates for waveguides by diffusing dopants into it observed similar but larger seasonal variations in glass surface properties. Neither observation was understood.
The IRG investigated the subject by measuring tracer diffusion in glass annealed at various humidity levels. For example, Figure 2 shows that the Na tracer diffusion into 1737 glass diffusion-annealed in dry and wet air leads to different residual radioactivity profiles. The pronounced kink in the latter case indicates the presence of a near-surface layer with a sharply lower Na diffusion coefficient. The modified layer is several mm thick and contains OH-, fluctuations of which lead to corresponding changes in near surface diffusivity.
An interesting question, currently under investigation, is how these modified surface layers influence mass transport near the surface. Bulk glass deforms by viscous flow, but surface diffusion is expected to dominate on nanometer scale patterned surfaces.10 The nature of near-surface mass transport has become of considerable interest, as glass surfaces are patterned on an increasingly finer scale for biological applications. If mass transport in small pattern structures occurs by a mechanism other than bulk viscous flow, the thermal stability of these structures can no longer be predicted from current knowledge.
The gratings used for such studies have a depth much smaller than the thickness of the surface layer introduced by annealing glass in air, and they can be used to study the near surface mass transport in these modified surface layers. Such studies are currently under way.
Impact on ManufacturingBy clarifying the origin and properties of process-induced surface layers, the foregoing research will allow glass manufactures to eliminate non-understood seasonal changes in product properties and to manufacture glass with near surface properties optimized for a particular applications. Corresponding improvements are expected in deposited glassy layers in IC manufacturing.
For more information about Cornell University's Glass IRG, visit http://www.ccmr.cornell.edu/IRG-GLASS .
References1. Umbach, C.C. and Blakely, J.M., "Development of a Sub-Picoampere Scanning Tunneling Microscope for Oxide Surfaces," Applied Surface Science (Elsevier), Vols. 175-176, No. 15, 2001, pp. 746-752.
2. Kern, W. and Puotinen, D.A., "Cleaning Solutions Based on Hydrogen Peroxide for Use in Silicon Semiconductor Technology," RCA Review, Vol. 30, No. 2, 1970, pp. 187-206.
3. Wang, Y.Z.; Awadelkarim, O.O.; Couillard, J.G.; and Ast, D.G., "The Effect of Substrates on the Performance and Hot-Carrier Reliability of n-Channel Thin Film Transistors," Semiconductor Science and Technology, Vol. 13, No. 5, pp. 532-535.
4. Couillard, J.G.; Moore, C.B.; Fehlner, F.P.; and Ast, D.G., "Performance and Reliability of Poly-Si TFTs on Glass Substrate," Proceedings of the Fourth Symposium on Thin Film Transistor Technologies, Ed. Y. Kuo, Pennington, NJ, Electrochem. Soc., 1999, xi+ 428, pp. 355-359.
5. Couillard, J.G.; Ast, D.G.; Umbach, C.; Blakely, J.M.; Moore, C.B.; and Fehlner, F.P., "Chemical Treatment of Glass Substrates," J. Non-Cryst. Solids, Vol. 222, 1997, pp. 429-434.
6. Tian, L. and Dieckmann, R., "Bulk Diffusion Measurements to Study the Effectiveness of Barrier Layers: II. Exchange of Sodium Between Liquid Crystal Display Glass Substrates with Different Barrier Layers," J. Appl. Phys., Vol. 90, No. 8, 2001, pp. 3810-3815.
7. Tian, L. and Dieckmann, R., "Sodium Tracer Diffusion in an Alkaline-Earth Boroaluminosilicate Glass," J. Non-Cryst. Solids, Vol. 265, Nos. 1-2, 2000, pp. 36-40.
8. Tian, L.; Dieckmann, R.; Hui, C.-Y.; and Couillard, J.G., "Effect of Water Incorporation on the Diffusion of Sodium in an Alkaline-Earth Boroaluminosilicate Glass," J. Non-Cryst. Solids, Vol. 296, Nos. 1-2, 2001, pp. 123-134.
9. Couillard, J.; Ast, D.G.; Moore, C.B.; and Fehlner, F.P., "Effect of Barrier Layer Thickness on the Performance of Thin Film Transistors on Glass Substrates," Proceedings of the SPIE-The International Society for Optical Engineering (SPIE-Int. Soc. Opt. Eng.), 3014, 1997, p. 166-169.
10. Yang, F. and Li, J.C.M., "New Fick's Law for Self-Diffusion in Liquids," J. Appl. Phys., Vol. 80, No. 11, 1996, pp. 6188-6191.