Journal of Vacuum Science & Technology a Vacuum Surfaces and Films Volume 3 Issue 3 1985 [Doi...

UV/ozone cleaning of surfacesJohn R. Vig

Citation: Journal of Vacuum Science & Technology A 3, 1027 (1985); doi: 10.1116/1.573115 View online: http://dx.doi.org/10.1116/1.573115 View Table of Contents: http://scitation.aip.org/content/avs/journal/jvsta/3/3?ver=pdfcov Published by the AVS: Science & Technology of Materials, Interfaces, and Processing

Articles you may be interested in Isolation of organic field-effect transistors by surface patterning with an UV/ozone process J. Vac. Sci. Technol. B 27, 1057 (2009); 10.1116/1.3117350

Effects of UV/ozone treatment of a polymer dielectric surface on the properties of pentacene thin films fororganic transistors J. Appl. Phys. 104, 013715 (2008); 10.1063/1.2951905

The role of proximity caps during the annealing of UV-ozone oxidized GaAs J. Appl. Phys. 101, 114321 (2007); 10.1063/1.2740359

Growth and characterization of hafnium silicate films prepared by UV/ozone oxidation J. Vac. Sci. Technol. A 22, 395 (2004); 10.1116/1.1649346

UVOzone Cleaning of GaAs for MBE J. Vac. Sci. Technol. 20, 241 (1982); 10.1116/1.571365

Redistribution subject to AVS license or copyright; see http://scitation.aip.org/termsconditions. Download to IP: 128.114.34.22 On: Mon, 01 Dec 2014 15:02:16

UV /ozone cleaning of surfaces John R. Vig u. S. Army Electronics Technology and Devices Laboratory, ERADCOM, Fort Monmouth, New Jersey 07703-5302

(Received 17 September 1984; accepted 29 October 1984) The ultraviolet (UV)lozone surface cleaning method is reviewed. The UV /ozone cleaning procedure is an effective method of removing a variety of contaminants from surfaces. It is a simple-to-use dry process which is inexpensive to set up and operate. It can rapidly produce clean surfaces, in air or in a vacuum system, at ambient temperatures. By placing properly precleaned surfaces within a few millimeters of an ozone producing UV source, the process can produce clean surfaces in less than 1 min. The technique is capable of producing near-atomically clean surfaces, as evidenced by Auger electron spectroscopy, ESCA, and ISS/SIMS studies. Topics discussed include: the variables of the process,the types of surfaces which have been successfully cleaned, the contaminants which can be removed, the construction of a UV /ozone cleaning facility, the mechanism of the process, UV /ozone cleaning in vacuum systems, rate enhancement techniques, safety considerations, effects ofUV /ozone other than cleaning, and applications.

I. INTRODUCTION The ability of ultraviolet (UV) light to decompose organic molecules has been known for a long time, but it is only during the past decade that UV cleaning of surfaces has been explored.

In 1972, Bolon and Kunz! reported the ability ofUV light to depolymerize a variety of photoresist polymers. The poly-mer films were enclosed in a quartz tube, the tube was evacu-ated, then backfilled with oxygen. The samples were irra-diated with UV light from a medium pressure mercury lamp which generated ozone. The polymer films, which had been several thousand angstroms thick, were successfully depoly-merized in less than 1 h. The major products of depolymeri-zation were found to be water and carbon dioxide. Subse-quent to depolymerization, the substrates were examined by Auger electron spectroscopy (AES) and were found to be free of carbonaceous residues. Only inorganic residues such as tin and chlorine were found. When a Pyrex filter was placed between the UV light and the films, or when a nitrogen at-mosphere was used instead of oxygen, the depolymerization was hindered. Thus, Bolon and Kunz recognized that oxy-gen and wavelengths shorter than 300 nm played a role in the depolymerization.

In 1974, Sowell et al. 2 described UV cleaning of adsorbed hydrocarbons from glass and gold surfaces, in air and in a

vacuum system. A clean glass surface was obtained after 15 h of exposure to the UV radiation in air. In a vacuum system at 10-4 Torr of oxygen, clean gold surfaces were produced after about 2 h ofUV exposure. During cleaning, the partial pressure of O2 decreased, while that of CO2 and H 20 in-creased. The UV also desorbed gases from the vacuum chamber walls. In air, gold surfaces which had been con-taminated by adsorbed hydrocarbons could be cleaned by "several hours of exposure to the UV radiation." Sowell et al. also noted that by storing clean surfaces under UV radi-ation it was possible to maintain the surface cleanliness in-definitely.

Starting in 1974, Vig et al. 3-5 described a series of experi-ments aimed at determining the optimum conditions for pro-ducing clean surfaces by UV irradiation. The variables of cleaning by UV irradiation were defined, and it was shown that, under the proper conditions, UV /ozone cleaning was capable of producing clean surfaces in less than 1 min.

II. THE VARIABLES OF UV/OZONE CLEANING

A. The wavelengths emitted by the UV sources To study the variables of the UV cleaning procedure, Vig

and LeBus5 constructed the two UV cleaning boxes shown in Fig. 1. Both were made of aluminum and contained low-

AlZAK REFlEC;T IIO:R~~~~~~~5F~9 l----l_!ll:::;i~--I-SAMPlE -

U V BOX I FIG. I. Apparatus for UV lozone cleaning experiments.

1027 J. Vac. Sci. Technol. A 3 (3), May/Jun 1985

Al BOX Al STAND

TRANSFORMER U V BOX 2

1027


1028 John R. Vlg: UV I ozone cleaning of surfaces

pressure mercury discharge lamps and an aluminum stand with Alzak6 reflectors. The two lamps produced nearly equal intensities of short wavelength UV light, about 1.6 mW /cm2 for a sample 1 cm from the tube. Both boxes con-tained room air (in a clean room) throughout these experi-ments. The boxes were completely enclosed to reduce recon-tamination by air circulation.

Since only the light which is absorbed can be effective in producing photochemical changes, the wavelengths emitted by the UV sources are important variables. The low-pressure mercury discharge tubes generate two wavelengths of inter-est, 184.9 and 253.7 nm. The 184.9 nm wavelength is impor-tant because it is absorbed by oxygen, and it thus leads to the generation of ozone. 7 The 253.7 nm radiation is not absorbed by oxygen; it therefore does not contribute to ozone genera-tion. However, it is absorbed by most hydrocarbons89 and also by ozone.7 The absorption by ozone is principally re-sponsible for the destruction of ozone in the UV box. There-fore, when both wavelengths are present, ozone is continual-ly being formed and destroyed. An intermediate product of both the formation and destruction processes is atomic oxy-gen, which is a very strong oxidizing agent.

The tube of the UV lamplO in box 1 consisted of91 cm of "hairpin-bent" fused quartz, which transmits both the 253.7 and 184.9 nm lines. The lamp emitted about 0.1 mW /cm2 of 184.9 nm radiation, measured at 1 cm from the tube.

The lamp in box 2 had two straight and parallel, 46 cm long, high-silica glass tubes. The glass was Corning UV Glass No. 9823 which transmits at 253.7 nm but not at 184.9 nm. Since this lamp generated no measurable ozone, a sepa-rate Siemens-type ozone generator ll was built into box 2. This ozone generator did not emit UV light. Ozone was pro-duced by a "silent" discharge when high-voltage ac was ap-plied across a discharge gap formed by two concentric glass tubes, each of which was wrapped in aluminum foil elec-trodes. The ozone-generating tubes were parallel to the UV tubes, approximately 6 cm away.

UV box 1 was used to expose samples, simultaneously, to 253.7 nm, 184.9 nm, and the ozone generated by the 184.9 nm radiation. UV box 2 permitted the options of exposing samples to 253.7 nm plus ozone, 253.7 nm only, or ozone only.

Vig et al. used contact angle measurements, wettability tests, and Auger electron spectroscopy (AES) to evaluate the results of cleaning experiments. Most of the experiments were conducted on polished quartz wafers, the cleanliness of which could be evaluated by the "steam test," a highly sensi-tive wettability test.5.12,15

A "black-light" long wavelength UV source, which emit-ted wavelengths above 300 nm only, was also tried. It pro-duced no noticeable cleaning even after 24 h of irradiation.

It was found early in the studies ofVig et al. that samples could be cleaned consistently by UV irradiation only if gross contamination was first removed from the surfaces. Their precleaning procedure consisted of the following steps:

(1) Scrub the samples with a swab while it is immersed in ethyl alcohol.

(2) Degrease ultrasonically in a good solvent. (3) Boil in fresh ethyl alcohol, then agitate ultrasonically.

J. Vac. Sci. Technol. A, Vol. 3, No.3, May/Jun 1985

1028

(4) Rinse in running ultrapure (18 M cm) water. (5) Spin-dry immediately after the running water rinse. Subsequent to this precleaning procedure, the steam test

and contact angle measurements invariably indicated that the surfaces were contaminated. However, after exposure to UV /ozone in box 1, the same tests always indicated clean surfaces. On numerous occasions the cleanliness of such UV /ozone cleaned surfaces has been verified in the author's laboratory, and elsewhere by AES and electron spectroscopy for chemical analysis (ESCA).I,3,4,13-15 The effectiveness of UV /ozone cleaning has also been confirmed by ion scatter-ing spectroscopy/secondary ion mass spectroscopy (ISS/ SIMS). 16

A number of quartz wafers were precleaned and exposed to the UV light in box 1 until clean surfaces were obtained. Each of the wafers was then thoroughly contaminated with human skin oil which has been a difficult contaminant to remove. The wafers were precleaned again, groups were ex-posed to each of the four UV /ozone combinations men-tioned earlier, and the time to attain a clean surface, as indi-cated by the steam test, was measured. In each UV box, the samples were placed within 5 mm of the UV source (where the temperature was about 70C).

The wafers exposed to 253.7 nm + 184.9 nm + ozone in UV box 1 became clean in 20 s. The samples exposed to 253.7 nm + ozone in UV box 2 reached the clean condition in 90 s. Samples exposed to 253.7 nm without ozone and to ozone without UV light took about 1 hand 10 h, respectively, be-fore clean surfaces were obtained.

The conclusion one can draw is that, while both UV light without ozone and ozone without UV light can produce a slow cleaning effect in air, the combination of both short wavelength UV light and ozone, such as is obtained from a quartz UV lamp, produces a clean surface orders of magni-tude faster. Although the 184.9 nm radiation is also ab-sorbed by many hydrocarbons, it was not possible from these experiments to isolate the cleaning effect of the 184.9 nm radiation. The ozone concentrations had not been measured. As is discussed below, the concentrations vary within each box with distance from the UV source.

B. Distance between sample and UV source Another variable which can greatly affect the cleaning

rate is the distance between the sample and the UV source. Because of the shapes of the UV tubes and of the Alzak reflectors above the tubes and below the samples, the lamps in both boxes were essentially plane sources. Therefore, one might have expected that the intensity ofUV light reaching a sample would be nearly independent of distance. This was not so, however, when ozone was present, because ozone has a broad absorption band7,I7,I8 centered at about 260 nm. At 253.7 nm, the absorption coefficient of ozone is 130 cm -I atm -I. The intensity I of the 253.7 nm radiation reaching a sample therefore decreases as

I = IDe - 130pd,

where p is the average ozone pressure between the sample and the UV source in atmospheres at 0 C, and d is the dis-


1029 John R. Vig: UV/ozone cleaning of surfaces

tance to the sample in centimeters. When a quartz UV tube is used, both the ozone concentration and the UV radiation intensity decrease with distance from the UV source.

Two sets of identically precleaned samples were placed in UV box 2. One set was placed within 5 mm of the UV tube, the other, at the bottom ofthe box, about 8 cm from the tube. With the ozone generator off, there was less than a 30% difference in the time it took for the two sets of samples to attain a minimal contact angle (about 60 min vs 75 min). When the experiment was repeated with the ozone generator on, the samples near the tube became clean nearly ten times faster (about 90 s vs 13 min). Similarly, in UV box 1, samples placed within 5 mm of the tube were cleaned in 20 s vs 20-30 min for samples placed near the bottom of the box, 13 cm away. Therefore, to maximize the cleaning rate, the samples should be placed as close to the UV source as possible.

C. Contaminants Vig et al. tested the effectiveness of the UV lozone clean-

ing procedure on a variety of contaminants. Among the con-taminants were: human skin oils, contamination adsorbed during prolonged exposure to air, a cutting oil, 19 a beeswax and rosin mixture, a lapping vehicle,20 a mechanical vacuum pump oil,21 DC 704 silicone diffusion pump oil,22 DC 705 silicone diffusion pump oil,22 a silicone vacuum grease,22 an acid (solder) flux,23 and a rosin flux from a rosin core lead-tin solder. The contaminants were applied to clean, polished quartz wafers. After contamination, the wafers were pre-cleaned, then exposed to UV lozone by being placed within a few millimeters of the tube in UV box 1. After a 60 s expo-sure, the steam test and AES indicated that all traces of the contaminants had been removed.

Since AES could not differentiate between the silicon peaks due to quartz and those due to the silicon containing contaminants, the removal of silicone diffusion pump fluids was also tested on Alzak, which normally has a silicon-free oxide surface, and on gold. AES examination of the Alzak and gold surfaces following UV lozone cleaning showed no traces of silicon present.

During the course of their studies, Vig et al. learned from colleagues working on ion implantation for integrated cir-cuits that the usual wet-cleaning procedures (with hot acids) failed to remove the photoresist from silicon wafers which had been exposed to radiation in an ion-implantation accel-erator, presumably because of crosslinking of the photore-sist. Ion-implanted silicon wafers, each with approximately a 1 }lm coating of exposed Kodak Micro Resist 747,24 were placed within a few millimeters of the source in UV box 1. After an overnight (~ 10 h) exposure to UV lozone, all traces of the photoresist were removed from the wafers, as con-firmed by AES.

Films of carbon, vacuum deposited onto quartz to make the quartz surfaces conductive for study in an electron mi-croscope, were also successfully removed by exposure to UV lozone. Inorganic contaminants such as dust and salts cannot be removed by UV I ozone. Such contaminants should be removed in the precleaning procedure.

UV lozone has also been used for waste-water treatment and for destruction of highly toxic compounds.25-28 Experi-


1029

mental work in connection with these applications has shown that UV lozone can convert a wide variety of organic and some inorganic species to relatively harmless, mostly volatile products such as CO2, CO, H20, N2, etc. Com-pounds which have been successfully destroyed in water by UV lozone include: ethanol, acetic acid, glycine, glycerol, palmitic acid; organic nitrogen, phosphorous and sulfur compounds; potassium cyanide; complexed Cd, Cu, Fe, and Ni cyanides; photographic wastes, medical wastes, secon-dary effluents; plus chlorinated organics and pesticides such as pentachlorophenol, dichlorobenzene, dichlorbutane, chloroform, malathion, Baygon, Vapam, and DDT. It has also been shown29 that the combination ofUV and ozone is more effective in destroying microbial contaminants in wa-ter, such as E. coli and streptococcus faecalis, than UV or ozone alone.

D. The precleaning Although UV lozone is able to remove contaminants such

as thick photoresist coatings and carbon films, without any precleaning, the procedure cannot, in general, remove gross contamination. For example, when a clean quartz wafer was coated thoroughly with human skin oils and placed in UV box 1 (Fig. 1) without any precleaning, even prolonged expo-sure to UV lozone failed to produce a low-contact-angle sur-face. This is possibly due to the fact that human skin oils contain materials, such as inorganic salts, which cannot be removed by photosensitized oxidation.

The UV lozone removed silicones from surfaces which had been precleaned, as described earlier, and also from sur-faces which had been just wiped with a cloth to leave a thin film. However, when the removal of a thick film was at-tempted, the UV lozone removed most of the film upon pro-longed exposure; but it also left a hard cracked residue on the surface. This may be due to the fact that many chemicals respond to radiation differently, depending on whether or not oxygen is present. In the presence of oxygen many poly-mers, for instance, degrade when irradiated; whereas, in the absence of oxygen (as would be the case for the bulk of a thick film) these same polymers crosslink. In the study of the radi-ation degradation of polymers in air, the "results obtained with thin films are often markedly different from those ob-tained using thick specimen."30

For the UV lozone cleaning procedure to work reliably, the surfaces must be precleaned: first, to remove contamin-ants such as dust and salts which cannot be changed to vola-tile products by the oxidizing action ofUV lozone, and, sec-ond, to remove thick films the bulk of which could be transformed into a UV resistant film by the crosslinking ac-tion of the UV light that penetrates the surface.

E. The substrate The UV lozone cleaning process has been successfully

used on a variety of surfaces including: glass, quartz, mica, sapphire, ceramics, metals, semiconductors, and a conduc-tive polyimide cement.

Quartz and sapphire are especially easy to clean with UV I ozone, since these materials are transparent to short wave-length UV. For example, when a pile of thin quartz plates, a


1030 John R. Vig: UV I ozone cleaning of surfaces

couple of centimeters deep, was cleaned by UV /ozone, both sides of all the blanks, even the ones at the bottom of the pile, were cleaned by the process. Since sapphire is even more transparent, it too could probably be cleaned the same way. When flat quartz plates were placed one on top of the other so that there could have been little or no ozone circulation between the plates, the UV /ozone cleaning was able to clean both sides of these plates also. (There is evidence in the litera-ture31 that photocatalytic oxidation of hydrocarbons, with-out the presence of gaseous oxygen, can occur on some oxide surfaces.)

When white alumina ceramic substrates were cleaned by UV /ozone, the surfaces could be cleaned properly. How-ever, the sides facing the UV turned yellow, probably due to the production of (UV-induced) color centers. After several days at room temperature, or a few minutes at high tempera-tures, the white color returned.

Metal surfaces could be cleaned by UV /ozone without any problems as long as the UV exposure was limited to the time required to produce a clean surface. (This time should generally be about 1 min or less for surfaces which have been properly precleaned.) However, prolonged exposure of ox-ide-forming metals to UV light can produce rapid corrosion. Silver samples, for example, turned black in UV box 1 within 1 h. Experiments with sheets of Kovar, stainless steel (type 302), gold, silver, and copper showed that, upon extended UV irradiation, the Kovar, the stainless steel, and the gold appeared unchanged; the silver and copper oxidized on both sides, but the oxide layers were darker on the sides facing away from the UV source than on the sides facing the UV. When electroless gold-plated nickel parts were stored under UV /ozone for several days, a black powdery coating gradu-ally appeared on the parts. Apparently, nickel diffused to the surface through pinholes in the gold plating, and the oxi-dized nickel eventually nearly completely covered the gold. The corrosion was observed even in UV box 2, when no ozone was being generated. The rate of corrosion increased substantially when a beaker of water was placed in the UV boxes to increase the humidity. Even Kovar showed signs of corrosion under such conditions.

The corrosion may possibly be explained by the fact that, as is known in the science of air-pollution control, in the presence of short wavelength UV light, impurities in air, such as oxides of nitrogen and sulfur, combine with water vapor to form a corrosive mist of nitric and sulfuric acids. The use of controlled atmospheres in the UV box should, therefore, minimize the corrosion problem.

Since UV /ozone dissociates organic molecules, it may be useful for cleaning some organic materials just as etching and electropolishing are sometimes useful for cleaning met-als. The process has been used successfully to clean quartz crystal resonators which have been bonded with silver-filled polyimide cement.32 Teflon (TFE) tape exposed to UV / ozone in UV box 1 for 10 days experienced a weight loss of 2.5%.33 Also, the contact angles measured on clean quartz plates increased after a piece of Teflon was placed next to the plates in a UV box.34 Similarly, Viton shavings taken from an 0 ring experienced a weight loss of 3.7% after 24 h in UV box 1. At the end of the 24 h, the Viton surfaces had become


1030

sticky.33 Semiconductor surfaces have been successfully UV /

ozone-cleaned without adversely affecting the functioning of the devices. For example, after a 4K static RAM integrated circuit was exposed to UV /ozone for 120 min in a commer-cial UV /ozone cleaner, the device continued to function without any change in performance. (This IC has been made using n-channel silicon gate technology, with 1 to 1.5 f-lm junction depths.35)

F. Rate enhancement techniques UV /ozone cleaning "rate enhancement" techniques have

been investigated by Zafonte and Chiu.36 Experiments on gas phase enhancement techniques included a comparison of the cleaning rates in dry air, dry oxygen, moist air, and moist oxygen. The moist air and moist oxygen consisted of gases that had been bubbled through water. Oxygen that had been bubbled through hydrogen peroxide was also tried. Experi-ments on liquid-enhancement techniques consisted of drop-wise addition of either distilled water or of hydrogen perox-ide solutions of various concentrations to the sample surfaces. Most of the sample surfaces consisted of various types of photoresist on silicon wafers.

The gas phase "enhancement" techniques resulted in neg-ligible to slight increases in the rates of photoresist removal (3-20 A/min without enhancement vs 3-30 A/min with en-hancement). The water and hydrogen peroxide liquid phase enhancement techniques both resulted in significant rate en-hancements (100-200 A/min) for non-ion implanted resists. The heavily ion implanted resists (1015_1016 atoms/cm2) were not significantly affected by UV /ozone, whether "en-hanced" or not.

III. THE MECHANISM OF UV/OZONE CLEANING The available evidence indicates that UV /ozone cleaning

is primarily the result of photosensitized oxidation pro-cesses, as is represented schematically in Fig. 2. The con-taminant molecules are excited and/or dissociated by the absorption of short wavelength UV light. Simultaneously, atomic oxygen and ozone l718 are produced when O2 is disso-ciated by the absorption of UV with a wavelength less than 245.4 nm. Atomic oxygen is also produced l718 when ozone is dissociated by the absorption of the UV and longer wave-lengths of radiation. The excited contaminant molecules, and the free radicals produced by the dissociation of con-

1 IONS CONTAMINANT FREE RADICALS MOLECULES + hv 1 EXCITED MOLECULES NEUTRAL MOLECULES VOLATILE MOLECULES (C02,H20,N2,etc.1

FIG. 2. Simplified schematic representation ofUV /ozone cleaning process.



taminant molecules, react with atomic oxygen to form simpler volatile molecules, such as CO2, H 20, N 2, etc.

The energy required to dissociate an O2 molecule into two ground state 0 atoms corresponds to 245.4 nm. However, at and just below 245.4 nm the absorption of O2 is very weak.7,17,18 The absorption coefficient increases rapidly with decreasing wavelengths. A convenient wavelength for pro-ducing 0 3 is the 184.9 nm emitted by low-pressure Hg dis-charge lamps in fused quartz envelopes. Similarly, since most hydrocarbons have a strong absorption band between 200 and 300 nm, the 253.7 nm wavelength emitted by the same lamps is useful for exciting or dissociating contaminant molecules. The energy required to dissociate ozone corre-sponds to 1140 nm. The absorption by ozone reaches a maxi-mum near the 253.7 nm wavelength. The actual photo-chemical processes occurring during UV lozone cleaning are more complex than shown in Fig. 2. For example, the rate of production of ozone by 184.9 nm photons is promoted by the presence of other molecules, such as N2 and CO2,

As described earlier, the combination of short wavelength UV light and ozone produced clean surfaces about 200 to 2000 times faster than UV light alone or ozone alone. Simi-larly, Prengle et al. 25 ,28 had found in their studies of waste-water treatment that UV enhances the reaction with ozone 102-fold to 104-fold, and the products of the reactions are materials such as CO2, H20, and N 2. Increasing the tem-perature was found to increase the reaction rates. Mattox37 has also found that mild heating increases the UV lozone cleaning rates. Bolon and Kunz,l on the other hand, had found that the rate ofUV lozone depolymerization of photo-resists did not change significantly between 100 and 300 DC. The rate of destruction of microorganisms was similarly in-sensitive to a temperature increase from room temperature to 40 Dc. 29

IV. UV/OZONE CLEANING IN VACUUM SYSTEMS Sowell et ae reported that, when 10-4 Torr pressure of

oxygen was present in a vacuum system, short wavelength UV desorbed gases from the walls of the system. During UV irradiation, the partial pressure of oxygen decreased, while that of CO2 and H 20 increased.

One must exercise caution in using a mercury UV source in a vacuum system because, should the lamp envelope break or leak, mercury would enter and ruin the usefulness of the system. Mercury has a high vapor pressure; its complete re-moval from a vacuum chamber is a difficult task. Other types ofUV sources, such as xenon or deuterium lamps, may be safer to use in vacuum systems. The UV light can also be radiated into systems through sapphire or quartz windows, or through deep-UV fiber-optic bundles. A small partial pressure of oxygen should be present during UV cleaning.

Caution must also be exercised when using UV lozone in a cryopumped vacuum system, since cryopumped ozone is po-tentially explosive,39 particularly during regeneration of the cryopump. A convenient method of dealing with this poten-tial hazard is to use two kinds of UV sources, one an ozone-generating source, the other an "ozone killer" source40 (see Sec. V.).


1031

V. SAFETY CONSIDERATIONS In the construction of a UV cleaning facility, one should

be aware of the safety hazards associated with short wave-length UV light. Exposure to intense short wavelength UV light can cause serious skin and eye injury within a short time. For the UV boxes used in the Vig et al. experiments, switches are attached to the doors in such a manner that when the doors are opened, the UV lamps are shut off auto-matically. If the application demands that the UV lamps be used without being completely enclosed (for example, as might be the case if an UV cleaning facility is incorporated into a thermocompression bonder), then proper clothing and eye protection should be worn to prevent skin burns and eye injury.

Another safety hazard is ozone, which is highly toxic. In setting up a UV cleaning facility, one must ensure that the ozone levels to which people are exposed do not exceed 0.1 ppm, the OSHA standard.38 Ozone is also a potential hazard in a cryopumped vacuum system because cryopumped ozone can become explosive under certain conditions. 39

One method of minimizing the hazards associated with ozone is to use two types of short wavelength ultraviolet sources for UV lozone cleaning.40 One, an ozone generating UV lamp, e.g., a low-pressure mercury light in a fused quartz envelope. The other, a UV lamp that does not generate ozone but one which emits one or more wavelengths that are strongly absorbed by ozone, e.g., a low-pressure mercury light in a high silica glass tube, which emits primarily at 253.7 nm. Such a nonozone generating UV source can be used as an "ozone killer." For example, in one cryopumped vacuum system it was found that after the UV lozone clean-ing, in up to 20 Torr of oxygen, was completed and the ozone generating UV lamp was turned off, 10 min of "ozone killer" UV light reduced the concentration of ozone to less than 0.01 ppm, a level that is safe for immediate cryopumping.41 With the ozone killer lamp, ozone concentrations were re-duced by at least a factor of 100 within 10 min. Without the ozone killer lamp, the half life of ozone is 3 days at 20 DC.42

VI. UV/OZONE CLEANING FACILITY CONSTRUCTION

The material chosen for the construction of a UV lozone cleaning facility should be one which is not corroded by ex-tended exposure to UV lozone. One material that can be used is polished aluminum with a relatively thick anodized oxide layer, such as Alzak. 6 Such materials are resistant to corrosion, have a high thermal conductivity which helps to prevent heat buildup, and are also good reflectors of short wavelength UV. Most other metals, including silver, are poor reflectors in this range.

Vig et al. initially used ordinary, shop variety aluminum sheet for UV box construction. After a while, however, it was noticed that a thin coating of a white powder (probably aluminum oxide particles) appeared at the bottom of the boxes. Even in a UV box made of standard Alzak, after a couple of years' usage white spots started appearing on the Alzak. To avoid the possibility of particles being generated inside the UV lozone cleaning facility, the facility should be


1032 John R. Vig: UV/ozone cleaning of surfaces

inspected periodically for signs of corrosion. The use of "class M" Alzak may also help, since this material has a much thicker oxide coating. It is made for "exterior marine service," instead of the "mild interior service" specified for standard Alzak. Commercially available UV lozone clean-ers are constructed of stainless steel. To date, no corrosion problems have been reported with such cleaners.

Organic materials should not be present in the UV clean-ing box. For example, the plastic insulation usually found on the leads of UV lamps should be replaced with inorganic insulation, such as glass or ceramic. The box should be en-closed so as to minimize recontamination by circulating air, and to prevent accidental UV exposure.

The most widely available sources of short wavelength UV light are mercury arc lamps. Low-pressure mercury lamps in pure, fused quartz envelopes operate near room temperature, emit approximately 90% at 253.7 nm, and gen-erate sufficient ozone for effective surface cleaning. Approx-imately 5% of the output of these lamps is at 184.9 nm. Medium and high-pressure UV lamps 7 generally have a much higher output in the short wavelength UV range. These lamps also emit a variety of additional wavelengths below 253.7 nm, which may enhance their cleaning action. However, they operate at high temperatures (the envelopes are near red hot), have a shorter lifetime, a higher cost, and present a greater safety hazard. The mercury tubes can be fabricated in a variety of shapes to fit different applications. In addition to mercury arc lamps, microwave-powered mer-cury vapor UV lamps are also available.

Other good sources of short wavelength UV, such as xe-non lamps and deuterium lamps, are also available. These lamps must also be in an envelope transparent to short wave-length UV, such as quartz. In setting up a UV cleaning facili-ty, one should choose a UV source which will generate enough UV lozone to allow for rapid photosensitized oxida-tion of contaminants; however, too high an output at the ozone generating wavelengths can be counterproductive be-cause a high concentration of ozone will absorb most of the UV light before it reaches the samples. The samples should be placed as close to the UV source as possible to maximize the intensity reaching the samples. In the UV cleaning box I of Vig et al., the parts to be cleaned are placed on an Alzak stand, the height of which can be adjusted to bring the parts close to the UV lamp. The parts to be cleaned can also be placed directly onto the tube if the box is built so that the tube is on the bottom of the box.43

VII. APPLICATIONS The UV lozone cleaning procedure has found numerous

applications during the past several years. A major use is substrate cleaning prior to thin-film deposition as is widely used in the quartz crystal industry during the manufacture of quartz crystal resonators. There is probably no other de-vice of which the performance is so critically dependent on surface cleanliness. For example, the aging requirement for a 5 MHz resonator is that the frequency change no more than two parts on 1010 per week, whereas adsorption or desorp-tion of a monolayer of contamination from such a device changes the frequency by about one part in 106. The surface


1032

cleanliness must therefore be such that the rate of transfer of contamination within the (hermetically sealed) resonator en-closure is less than 10-4 mono layers per week! In the auth-or's quartz resonator fabrication laboratory, UV lozone is used at several points during the fabrication sequence, such as for cleaning and storing metal tools, masks, resonator parts, and storage containers.

The process is also being applied in a hermetic sealing method which relies on the adhesion between clean surfaces in an ultrahigh vacuum. 14,43-46 It has been shown that metal surfaces will weld together under near-zero forces if the sur-faces are atomically clean. A gold gasket between gold met-allized (UV lozone cleaned) aluminum oxide sealing surfaces is currently providing excellent hermetic seals in the produc-tion of ceramic flatpack -enclosed quartz resonators. The fea-sibility of achieving good hermetic seals by pressing a clean aluminum gasket between two clean, unmetallized alumi-num oxide ceramic surfaces has also been shown.44-46

The same adhesion phenomenon between clean (UV I ozone cleaned) gold surfaces has been applied to the con-struction of a novel surface contaminant detector.47.48 The rate of decrease in the coefficient of adhesion between freshly cleaned gold contacts is used as a measure of the gaseous condensable contaminant level in the atmosphere.

The process has also been applied to improve the reliabil-ity of wire bonds, especially at reduced temperatures. It has been shown,49.5o for example, that the thermocompression bonding process is highly temperature dependent when or-ganic contaminants are present pn the bonding surfaces. The temperature dependence can be eliminated by UV lozone cleaning of the surfaces just prior to bonding. In a study of the effects of cleaning methods on gold ball bond shear strength, UV lozone cleaning was found to be the most effec-tive method of cleaning contaminants from gold surfaces.51 UV lozone is also being used for cleaning alumina substrate surfaces during the processing of thin-film hybrid circuits.52

When the nonuniform appearance of thermal/flash pro-tective electro-optic goggles was traced to organic contamin-ants on the electro-optic wafers, a number of cleaning meth-ods were tested. UV lozone proved to be the most effective method for removing these contaminants, and thus it was chosen for use in the production of the goggles. 53

Other applications which have been described are photo-resist removal, 1.5.13 the cleaning of: vacuum chamber walls,2 photomasks,54 silicon wafers (for enhancing photoresist ad-hesion), 54 lenses, 53 mirrors, 54 solar panels, 54 sapphire (before the deposition of HgCdTe),54 cold-rolled steel (for paint and zinc coating adhesion improvement),54 surface acoustic wave and other fine linewidth devices,54.55 inertial guidance subcomponents (glass, chromium-oxide surfaced gas bear-ings, and beryllium),55 and gallium arsenide wafers.57 Since short wavelength UV can generate radicals and ions, a side benefit ofUV I ozone cleaning of insulator surfaces can be the neutralization of static charges. 58

UV lozone cleaning of silicon substrates in silicon molecu-lar beam epitaxy (MBE) has been found to be effective in producing near-defect-free MBE films. 59 By using UV I ozone cleaning, the above 1200 C temperatures required for removing surface carbon in the conventional method can be


1033 John R. Vlg: UV/ozone cleaning of surfaces

lowered to below 1000 0c. The slip lines resulting from ther-mal stresses and thermal pits that are often produced by the high temperature treatment are minimized in the lower tem-perature processing. Impurity redistribution in the substrate is also reduced.

VIII. EFFECTS OTHER THAN CLEANING Short wavelength UV, ozone, and the combination of the

two, can have effects other than surface cleaning. Among the more significant of these effects are the following:

A. Oxidation Ozone's oxidation power is second only to that of fluorine.

Ozone can oxidize most inorganic compounds to their final oxidative state.42 For most substrates, UV lozone cleaning for the minimum time necessary to obtain a clean surface will not cause a significant amount of oxidation. However, extended storage under UV lozone may be detrimental for some oxidizable surfaces. In some cases, the enhanced oxide formation may be beneficial. For example, whereas the "na-tive" oxide on GaAs is only about 30 A thick, UV lozone produces an oxide layer that is 100 to 300 A thick,60 i.e., UV I ozone can produce a clean, enhanced oxide "passivated" surface. 10 min of UV lozone cleaning increased the oxide thickness on silicon substrates from 0.9 to 1.2 nm.59 Similar-ly, the native UV lozone produced oxide layer at the inter-face of HgCdTe-Si02 has been found to enhance the inter-face properties.61 Solar radiation and atmospheric ozone have been found to markedly enhance the sulfidation of cop-per.62 Extended exposure to UV lozone has been found to significantly increase the oxide layer thickness on aluminum surfaces.63 Whereas the oxide thickness on air exposed alu-minum surfaces is normally limited to about 50 A, after 10 min ofUV lozone exposure, the oxide thickness was found to be 90 A, and after 60 min, 200 A.

B. UV-enhanced outgassing Short wavelength UV has been found to enhance the out-

gassing of glasses. 64 The UV light produced evolution of sig-nificant quantities of hydrogen, and also water, carbon diox-ide, and carbon monoxide. The hydrogen evolution was proportional to the amount of radiation incident onto the samples. For UV -opaque glasses, the evolution occurred from the side exposed to the UV; for high transmission sam-ples, the gas evolved from both sides.

C. Other surface/lnterface effects Energetic radiation such as UV and y-radiation has been

reported to produce dehydration and the formation of free radicals on silica surfaces.65 However, dehydrated (or silox-inated) silica surfaces are hydrophobic,66.67 whereas UV I ozone cleaned silica (quartz) surfaces exhibit a very low (less than 4) contact angle, thus indicating that the UV lozone cleaning does not dehydrate the surfaces, nor does it modify surface silanol groups the way high-temperature vacuum baking does.68 Short wavelength UV has also been found to produce a bleaching effect in Si-Si3 interfaces with thin ox-ides,69 and has also been found to produce yellowing (color


1033

centers) during the cleaning of aluminum oxide ceramics. The yellowing can be readily bleached by heating the sam-ple.

D. Etching Short wavelength (193 nm) UV laser irradiation ofbiologi-

cal and polymeric materials has been shown to be able to etch the materials with great precision, via "ablative photo-decomposition," and without significant heating of the sam-ples. Linewidths 5 pm wide have been etChed onto a plastic film to demonstrate the capability of this technique. 70 Oxy-gen does not appear to have the same significance in this process as it does in UV lozone cleaning. The etch depth versus fluence in vacuum and in air were found to be the same. 71 The UV lozone has been found to etch Teflon33.34 and Vi ton, 33 and wil1likely etch other organic materials as well.72.73

IX. SUMMARY AND CONCLUSIONS The UV lozone-cleaning procedure has been shown to be

a highly effective method of removing a variety of contam-inants from surfaces. It is a simple-to-use dry process which is inexpensive to set up and operate. It can produce clean surfaces at room temperature, either in a room atmosphere or in a controlled atmosphere.

The variables of the UV cleaning procedure are: the con-taminants initially present, the precleaning procedure, the wavelengths emitted by the UV source, the atmosphere between the source and sample, the distance between the source and sample, and the time of exposure. For surfaces which are properly precleaned and placed within a few milli-meters of an ozone-producing UV source, the process can produce a clean surface in less than 1 min. The combination of short wavelength UV light plus ozone produces a clean surface substantially faster than either short wavelength UV light without ozone or ozone without UV light. Clean sur-faces will remain clean indefinitely during storage under UV lozone, but prolonged exposure of oxide-forming metals to UV lozone in room air can produce rapid corrosion.

The cleaning mechanism seems to be a photosensitized oxidation process in which the contaminant molecules are excited and/or dissociated by the absorption of short wave-length UV light. Simultaneously, atomic oxygen is generated when molecular oxygen is dissociated and when ozone is dissociated by the absorption of both short and long wave-lengths of radiation. The products of the excitation of con-taminant molecules react with atomic oxygen to form simpler molecules, such as CO2 and H20, which desorb from the surfaces.

'D. A. Bolon and C. O. Kunz, Polym. Eng. Sci. 12,109 (1972); also D. A. Bolon, U. S. Patent No.3 890 176, June 17, 1975.

2R. R. Sowell, R. E. Cuthrell, D. M. Mattox, and R. D. Bland, J. Vac. Sci. Techno!. 11,474 (1974).

'J. R. Vig, C. F. Cook, Jr., K.Schwidtal, J. W. LeBus, and E. Hafner, in Proceedings of the 28th Annual Symposium Frequency Control (US Army Electrons Command, Ft. Monmouth, NJ, 1974), ADA 011113, pp 96-108. Article reprinted as ECOM Technical Report 4251, AD 785513, 1984.

4J. R. Vig, J. W. LeBus, and R. L. Filler, in Proceedings of the 29th Annual



Symposium on Frequency Control, pp. 220-229, 1975. Copies available from the National Technical Information Service, Springfield, V A, ADA017466.

5J. R. Vig and J. W. LeBus, UV /Ozone Cleaning afSurfaces (IEEE Trans. Parts, Hybrids and Packag., 1976), Vol. PHP-12, pp. 365-370.

6Alzak is an aluminum reflector material with a corrosion resistant oxide coating. The Alzak process is licensed to several manufacturers by the Aluminum Co. of America, Pittsburgh, PA 15219.

7J. G. Clavert and J. N. Pitts, Jr., Photochemistry (Wiley, New York, 1966), pp.205-209,687-705.

"V. S. Fikhtengolt's, R. V. Zolotareva, and Yu A. L'vov, Ultraviolet Spec-trum of Elastomers and Rubber Chemicals (Plenum, New York, 1966).

L.Lang, Absorption Spectra in the Ultraviolet and Visible Region (Aca-demic, New York, 1965).

IOModel No. R-52 Mineralight Lamp, Ultraviolet Products, Inc., San Ga-briel, CA 91778.

"See, e.g., Encyclopaedic Dictionary of Physics (Pergamon, New York, 1962), Vol. 5, p. 275.

12M. E. Schrader, in Surface Contamination: Its Genesis, Detection and Con-trol, edited by K. L. Mittal (Plenum, New York, 1979), Vol. 2, pp. 541-555.

"P. H. Hollaway and D. W. Bushmire, Proceedings of the 12th Annual Reliability Physics Symposium (Institute of Electrical and Electronic En-gineers, Piscataway, NJ, 1974), pp. 180-186.

I4R. D. Peters, Proceedings of the 30th Annual Symposium on Frequency Control, pp. 224-231, 1976. Copies available from the National Technical Information Service, Springfield, V A, ADA046089.

15e. E. Bryson and L. 1. Sharpen, in Surface Contamination: Its Genesis, Detection and Control, edited by K. L. Mittal (Plenum, New York, 1979), Vol.2, pp. 687-696.

16W. L. Baun, Surf. Sci. 6, 39-45 (1980). 17J. R. McNesby and H. Okabe, "Oxygen and Ozone," in Advances in Pho-

tochemistry, edited by W. A. Noyes, lr., G. S. Hammond, and J. N. Pitts (Interscience, New York, 1964). Vol. 3, pp. 166--174.

'"D. H. Volman, in Advances in Photochemistry, edited by W. A. Noyes. G. S. Hammond, and 1. N. Pitts (Interscience, New York, 1963), Vol. I, pp. 43-82.

lOp. R. Hoffman Co., Carlisle, PA. 2John Crane Lapping Vehicle 3M, Crane Packing Co., Morton Grove, IL

60053. "Welch Duo-Seal, Sargent-Welch Scientific Co., Skokie, IL 60076. "Dow Corning Corp., Midland, MI 48640. "Dutch Boy No. 205, National Lead Co., New York, NY 10006 2'Eastman Kodak Co., Rochester, NY 14650. "H. W. Prengle, e. E. Mauk, R. W. Legan, and e. G. Hewes, Hydrocarbon

Process. 54, 82 (1975). 26H. W. Prengle Jr., e. E. Mauk, and J. E. Payne, Forum on Ozone Disin-

fection, 1976; International Ozone Institute, Warren Bldg., Suite 206, 14805 Detroit Ave., Lakewood, OH 44107

27H. W. Prengle, Jr., and e. E. Mauk, Workshop on Ozone/Chlorine Diox-ide Oxidation Products of Organic Materials, EPA/International Ozone Institute Warren Bldg., Suite 206, 14805 Detroit Ave., Lakewood, OH 44107, Nov. 1976.

28H. W. Prengle, Jr., in Proceedings of the lnternational Ozone Institute Ozone Symposium, Warren Bldg., Suite 206, 14805 Detroit Ave., Lake-wood, OH 44107,1978.

20J. D. Zeff, R. R. Barton, B. Smiley and E. Alhadeff, US Army Medical Research and Development Command, Final Report, Contract No. DADA 17073-C-3138, Sept. 1974. Copies available from NTIS, AD A004205.

30H. V. Boenig, Structure and Properties of Polymers (Wiley, New York, 1973), p. 246.

3IV. N. Filimonov, in Elementary Photoprecesses in Molecules, edited by B. S. Neporent (Consultants Bureau, New York, 1968), pp. 248-259.

"R. L. Filler, J. M. Frank, R. D. Peters, and J. R. Vig, Proceedings of the 32nd Annual Symposium on Frequency Control, pp. 290-298, 1978. Copies available from Electronics Industries Assoc., 2001 Eye St., NW, Washington, D. e. 20006.

33J. W. LeBus and J. R. Vig (unpublished). 34J. Kusters (personal communication). 35E. Lasky (personal communication).


1034

36L. Zafonte and R. Chiu, Technical Report on UV-Ozone Resist Strip Feasibility Study, UVP, Inc., 5100 Walnut Grove Avenue, San Gabriel, CA 91778, September, 1983; to be presented at the SPIE Santa Clara Conference on Microlithography in March 1984.

"D. M. Mattox, Thin Solid Films 53,81 (1978). '""Occupational Safety and Health Standards", Vol. I, General Industry

Standards and Interpretations, Oct. 1972, Pt. 1910, 1000, Table Z-I, Air Contaminants, P. 642,4 as per change 10, June 26, 1975.

39e. W. Chen and R. G.Struss, Cryogenics 9,131 (1969). 4J.R. Vig and 1. W. LeBus, U. S. Patent NO.4 028135, June 7,1977. 4ID. A. Ehlers, Report No. PT81-004, General Electric Neutron Devices

Dept., St. Petersburg, FL, 1981. 42Matheson Gas Data Book, 6th ed. (Matheson Gas Products Co., East

Rutherford, NJ, 1980), pp. 574-577. 43R. D. Peters (personal communication). 44J. R. Vig and E. Hafner, Technical Report ECOM-4134, US Army Elec-

tronics R&D Command, Fort Monmouth, NJ, July 1973. Copies avail-able from NTIS, AD 763215.

"E. Hafner and 1. R. Vig, U. S. Patent No.3 914 836, Oct. 28, 1975. 46p. D. Wilcox, G. S. Snow, E. Hafner, and 1. R. Vig, Proceedings of the

29th Annual Symposium on Frequency Control, pp. 202-210,1975. See Ref. No.4 above for availability information.

47R. E. Cuthrell and D. W. Tipping, Rev. Sci. Instrum. 47, 555 (1976). 48R. E. Cuthrell, in Surface Contamination: Its Genesis, Detection and Con-

trol, edited by K. L. Mittal (Plenum, New York, 1979), Vol. 2, pp. 831-841.

49J. L. Jellison, IEEE Trans. Parts Hybrids Packag. PHpn, 206 (1975). 51. L. lellison, in Surface Contamination: Its Genesis, Detection and Con-


"J. A. Weiner, et al., Proceedings of the Electronic Components Confer-ence,pp.208-220,1983.

"R. Tramposch, Circuits Manufacturing 23, 30 (1983). 53J. A. Wagner, in Surface Contamination: Its Genesis, Detection and Con-


54E. Lasky (personal communcation). "H. I. Smith (unpublished and personal communications). 561. R. Stemniski and R. L. King, Jr., in Adhesivesfor Industry (Technology

Conferences, EI Segundo, CA, in cooperation with the So. Calif. Section, Soc. of Plastics Engineers, 1980), pp. 212-228.

571. A. McClintock, R. A. Wilson, and N. E. Byer, J. Vac. Sci. Technol. 20, 241 (1982).

'"D. H. Baird, Final Technical Report No. TR 76-807.1, Dec. 1976. Copies available from NTIS, AD A037463.

59M. Tabe, Appl. Phys. Lett. 45,1073 (1984). 6OJ. A. McClintock (personal communication). 61B. K. Janousek, R. e. Carscallen, and P. A. Bertrand, J. Vac. Sci. Tech-

nol. (to be published). 6'T. E. Graedel et al., Science 224, 599 (1984). 6'G. V. Clatterbaugh et al., Proceedings of the 34th Electronic Components

Conference, pp. 21-30,1984. MV. O. Altemose, in Vacuum Physics and Technology, Vol. 14 of Methods

of Experimental Physics, edited by G. L. Weissler and R. W. Carlson (Academic, New York, 1979), Chap. 7, pp. 329-333.

"'M. M. Tagieva and V. F. Kiseler, Russ. J. Phys. Chern. 35, 680 (1961). MM. L. Hair, Proceedings of the 27th Annual Symposium on Frequency

Control, AD771042, pp. 73-78, 1973. 67M. L. White, Proceedings of the 27th Annual Symposium on Frequency

Control, AD77I042, pp. 79-88,1973; also in Clean Surfaces: Their Prep-aration and Characterization for Interfacial Studies, edited by G. Gold-finger (Marcel Dekker, New York, 1970), pp. 361-373.

bRR. N. Lamb and D. N. Furlong, J. Chern. Soc. Faraday Trans. I 78,61 (1982).

bOp. J. Caplan, E. H. Poindexter, and S. R. Morrison, J. Appl. Phys. 53, 541 (1982).

7R. Srinivasan, Science News 123, 396 (1983). 71R. Srinivasan and B. Braren, 1. Polym. Sci.: Polym. Chern. edition 22,

2601 (1984). 72G. S. Alberts, U. S. Patent No.3 767490, Oct. 23,1973. 73A. N. Wright, U. S. Patent No.3 664 899, May 23,1972.


Journal of Vacuum Science & Technology a Vacuum Surfaces and Films Volume 3 Issue 3 1985 [Doi...

Documents

Transcript of Journal of Vacuum Science & Technology a Vacuum Surfaces and Films Volume 3 Issue 3 1985 [Doi...