ABSTRACT

This volume documents the proceedings of the International Symposium on Surface Contamination and Cleaning, held in Newark, New Jersey, May 23-25, 2001. Because of the importance of this topic in many technological areas, tremendous efforts have been devoted to devise novel and more efficient ways to monitor, analyse and characterize contamination

chapter |3 pages

the crystal birefringence, tempera-

chapter |4 pages

niques were prepared using pentane as the solvent. Similar methods were used in preparing calibration samples with the mold release, solder flux, and hydraulic oil samples. All contaminated coupons were gentl y heated in an oven at 50°C for several days to remove both semi-volatile and volatile components. This served to stabi-lize the contaminants, allowing for quantification by weighing. Once the weights became stable, the coupons were cooled and weighed to determine the amount of contaminant present on the surface. When not being weighed or examined, the coupons were kept in a desiccator. 3. RESULTS AND DISCUSSION Grazing-angle incidence reflectance spectroscopy acts to enhance the detection sensitivity for thin layers of residue predominantly through improved coupling of th e electric field intensity of the incident beam with the vibrating dipoles of the surface contaminant layer perpendicular to the metallic surface. Some additional enhancement of the infrared absorption spectrum will also occur due to a length-ening of the effective path length through the absorbing thin film layer [4-6]. If the optical properties of both thin film and substrate are known (or can be de-termined), the reflection-absorption spectrum can be calculated as a function of film thickness and angle of incidence. This capability is particularly useful for in-terpreting experimental data and designing optical instrumentation. Computer codes written at Sandia [7] performed these calculations for a variety of materials. 3.1. FTIR measurements FTIR reflectance data for the full drawing-agent sample set were obtained at NFESC and Sandia using angles of incidence of 75 and 60° for average film thickness ranging from 0.1 to 1 |im, and aluminum substrates with surface finish ranging from 600 to 80 grit. Since the surface finishing operation produced a highly directional roughness, measurements were made both longitudinally and transversely with respect to the polishing grooves. R values were determined at

chapter |10 pages

of the spectral response, the integrated reflection-absorption intensity, of these samples are slightly greater than the intensity of the spectral response of the same samples measured via a 60 ° angle of incidence data (Figure 3). This behavior is expected due to the increase in reflection-absorption sensitivity with increasing angle o f incidence. Here, too, the average initial slope (and hence instrument sen-sitivity) is the same for both transverse and longitudinal orientations. The pronounced non-linearity in slope for the thickest films at 75° angle-of-incidence was unexpected. A n increasingly non-linear response may be observed for thicker absorbing films, and this effect will become more pronounced as the angle of incidence is also increased. The interpretation of the data implying that measurement of a thicker film, sampled at a steeper angle, generated the observed non-linearity in the data is not substantiated by th e calculated spectra for the pre-sent measurement conditions due to the small change from 60 to 75° in the angle of incidence. Furthermore, such a non-linear effect would be most pronounced for measurements on the smoothest substrat e (Figure 4, filled circles) where the ef-fective local orientation of the surface is most constant with respect to the illumi-nation beam. Instead of observing such non-linear behavior the measurements made on the smoothest surface are by far the most linear sample series for the 75° data . We attribute the pronounced non-linearity of the 75° data for the thickest draw-ing-agent films to the morphological characteristics of the material as deposited o n the aluminum test panel surface. As described above, the drawing-agent mate-rial is highly viscous and forms a visibly heterogeneous white film at l-|im thick-ness. Variations in the deposition process produce relatively thick local areas of drawing-agent film and result in accretion of solid residue along the polishing grooves and ridges of the aluminum substrate. Under these circumstances, illumi-natio n of the surface with the FTIR beam at an angle of 75° may result in shadow-ing by contaminant material on ridge structures for all except the smoothest (600 grit polish) surface. The 12-mm diameter focal area of the infrared beam is elon-gated by a factor o f four for this angle of incidence. In contrast, reflectance meas-urements at 60° result in only a factor of 2 elongation, and minimize the shadow-ing effect of thick films except for ridges on the roughest (80 grit polish) surfaces. This interpretation is substantiated by reflectance data for the second test set (lubricant material) as shown in Figure 5. FTIR reflectance measurements have been made at 75° angle-of-incidence for a test series similar to that of the draw-ing-agent set. An analysis of the C-H stretching frequencies shows a strikingly more linear dependence of instrument response with film thickness (with the ex-ception of a single point for one of the panels with a 220 grit surface finish). We believe that this is due to the more fluid characteristic of the lubricant material, which allows the deposited film to conform much more closely to the surface to-pography of the test coupons. This behavior may also account for the stronger de-pendence of the integrated intensity slope with surface roughness, when compared to the nearly constant results for the drawing-agent contaminant examined above.

chapter |3 pages

the residual concentration of the

chapter |3 pages

about chemical bonding and molecular structure. This information can be used to detect th e types of organic materials present on the surface. 4.3.2.2. Raman spectroscopy (RS) [7, 8] It is used to examine the energy levels of molecules that cannot be well character-ized via infrared spectroscopy. Th e two techniques, however, are complimentary. In the RS, a sample is irradiated with a strong monochromatic light source (usu-ally a laser). Most of the radiation will scatter or "reflect off' the sample at the same energy as the incoming laser radiation. However, a small amount will scat-ter from the sample at a wavelength slightly shifted from the original wavelength. It is possible to study the molecular structure or determine the chemical identity of the sample. It is quite straightforward to identify compounds by spectral library search. Due to extensive library spectral information, the unique spectral finger-print of every compound, and the ease with which such analyses can be per-formed, the RS is a very useful technique for various applications. An important application of the RS is the rapid, nondestructive characterization of diamond, diamond-like, and amorphous-carbon films. 4.3.2.3. Scanning electron microscopy (SEM) / energy dispersive X-ra y analysis (EDX) [7, 8] The SEM produce s detailed photographs that provide important information about the surface structure and morphology of almost any kind of sample. Image analy-sis is often the first and most important step in problem solving and failure analy-sis. With SEM, a focused beam of high-energy electrons is scanned over the sur-face of a material, causing a variety of signals, secondary electrons, X-rays, photons, etc. - each of which may be used to characterize the material with re-spect to specific properties . The signals are used to modulate the brightness on a CRT display, thereb y providing a high-resolution map of the selected material property. It is a surface imaging technique, but with Energy Dispersive X-ray (EDX) it can identify elements in the near-surface region. This technique is most useful for imaging particles. 4.3.2.4. X-ray fluorescence (XRF) [7, 8] Incident X-rays are used to excite surface atoms. The atoms relax through the emission of an X-ray with energy characteristic of the parent atoms and the inten-sity proportional to the amount of the element present. It is a bulk or "total mate-rials" characterization technique for rapid, simultaneous, and nondestructive analysis of elements having an atomic number higher than that of boron. Tradi-tional bulk analysis applications include identifying metals and alloys, detecting trace elements in liquids, and identifying residues and deposits. 4.3.2.5. Total-reflection X-ray fluorescence (TXRF) [7, 8] It is a special XRF technique that provides extremely sensitive measures of the elements present in a material's outer surface. Applications include searching for metal contamination in thin films on silicon wafers and detecting picogram-levels o f arsenic, lead, mercury and cadmium on hazardous, chemical fume hoods.

chapter 38|11 pages

M.K. Chaw la An optimum level of cleanliness is the one that minimizes the total cost . Even-tually one can arrive at a cleanliness level where the savings in the failure/non-conformance costs will not be offset by incremental cost of achieving cleanliness beyond th e optimum level. A small range around the optimum level of cleanliness can b e established as the "Acceptable Level" of cleanliness. 7. DEFINING ACCEPTABLE ("OPTIMUM") LEVEL OF CLEANLINESS It is expected that the non-conformance levels will increase as the level of cleanliness decreases or vice versa. It is important to understand the relationship between the level of cleanliness and non-conformance rate in order to establish the "acceptable level of cleanliness". For example, if the failure/non-conformance rate is too high due to the surface cleanliness level, then the surface cleanliness level will have to be improved i n order to reduce the failure rate. On the other hand, no failures or a very low failure rate due to the surface cleanliness level im-plies that the surface may be "over-cleaned." It may be desirable to optimize the cleaning process by comparing the cost of failures/non-conformance with the cost of cleaning the surface. Generally, in-creasing the level of surface cleanliness will result in increased cleaning cost. An increased level of cleanliness should lower the rate of non-conformance, which, in turn, reduces the non-conformance cost. As long as the reduction in non-conformance cost more than offsets the increased cost of cleaning, it would be cost effective to increase the achieved level of surface cleanliness. When the de-crease in non-conformance cost fails to offset the increase in the cleaning cost, then an optimum or "acceptable" level of cleanliness has been achieved. To establish the optimum level of surface cleanliness, two approaches are out-lined here. One approach utilizes the success of the subsequent operation that de-pends on surface cleanliness level. The other approach is to start monitoring the cleanliness levels achieved and corresponding level of failure/non-conformance rate. Once an acceptable level of cleanliness is established using one of the two approaches, cleaning process can be monitored in production to assure ongoing product quality. 7.1. Controlled experiment This approach requires that the measure of success be defined for the subsequent operation that depends on surface cleanliness. For example, if the parts are to be bonded, then the adhesion strength of the bond will be the measure of success. If the parts are to be coated after cleaning, then the adhesion strength of the coating should be correlated to surface cleanliness. The acceptable level of surface cleanliness is the one that results in the desired level of bond/adhesion strength. One simple approach is to start monitoring and recording the cleanliness level of each part. A statistically significant sample must be monitored to assure valid

chapter |13 pages

Octane is a non-polar liquid. Its interaction with the glass substrate is purely dispersive giving rise to the octane/water interfacial tension. Water is a polar liq-uid . Its interactions with the glass substrate may be broken down into two compo-nents: dispersive and non-dispersive. The dispersive (London) interactions are those generated by instantaneous (non-permanent) dipole interactions . The non-dispersive interactions include the permanent dipole interactions (Keesom and Debye interactions), hydrogen bonding and acid-base interactions. The latter two generate a surface charge on the glass substrate. The dissociation of silanol (SiOH) groups at the glass surface may be represented by the following reaction: The surface charge generated by these acid-base interactions increases the glass/water interfacial energy. The dispersive component of the surface tension of water is almost equa l to the surface tension of octane: y = 21.8 mJ/m [5] and Yo = 21.6 mJ/m [6], where the subscript "w" stands for water and the subscript

chapter |3 pages

The non-dispersive interaction energy between glass and water as a function of pH is expected to reflect the surface charge generated by the exposed chemical functions on the clean glas s surface. The variations in surface charge, generated by the exposed SiOH and aluminum oxide groups, is expected to give rise to fea-tures representing the surface chemistry of the clean glass. The scatter i n the data shown in Figures 4 and 5 allows only general trends to be discerned. The p.z.c.'s at pH 3 and 9 have been described in the preceding paragraphs. It is interesting to note that the chromic acid cleaned glass surfaces behave in a similar manner, showing virtually identical trends. The pyrolysis cleaned glass surfaces show dif-ferences in their behavior across the different glass compositions. These trends correlate with those observed for organic contamination of these surfaces, as de-scribed in Section 3.1, where the chromic acid cleaned glass surfaces all showed similar behavior, while the pyrolyzed glass showed significant differences in its sensitivity to contamination. In particular, the pyrolyzed silica surface shows far lower non-dispersive interaction energy with water than the pyrolyzed Corning code 1737 or sodalime glasses. This features correlates with the high degree of adsorbed contamination, described in Section 3.1, for the pyrolyzed silica surface. The datum in Figure 5 for the non-dispersive interaction energy between a py-rolyzed silica surface and water at pH 7 corresponds to a contact angle of 31°. This is significantly higher than the contact angle of water on a pyrolyzed silica surface freshly immersed into liquid octane. While the surface cleanliness was measured after cleaning, it was not measured after substrate immersion in the acidic or alkaline solutions. It is possible that the comparatively low non-dispersive interaction energy observed for pyrolyzed silica is partially an artifact caused by contamination of the cleaned silica before immersion into liquid oc-tane. Figure 4 shows similar behavior fo r the glass surfaces, suggesting that the alu-minoborosilicate and sodalime glasses show behavior similar to that of a silica surface. This phenomenon may be due to the leaching of soluble alkaline oxides from the glass surfaces during chromic acid cleaning, leaving a surface enriched in silica that behaves essentially in the same way as a chromic acid cleaned silica surface. In Figure 5, the minimum in the non-dispersive interaction energy between glass and water at pH 9 is not present for pyrolyzed sodalime glass. This mini-mum was presumed to be associated with a high sodium ion concentration in solution, neutralizing the SiO" groups at the glass surface. The presence of sodium oxide (see Table 1) in the sodalime glass composition may generate a high so-dium environment for the the silano l groups at the glass surface. The high sodium concentration in the glass may thus be equivalent to a high sodium concentration in solution, neutralizing the p.z.c.

chapter 4|3 pages

CONCLUSION While cleaned silica-based glass surfaces have similar surface compositions, their susceptibility to strongly adsorbing organic contaminant s depends strongly on the glass composition and the cleaning procedure. For the three glass species exam-ined: silica, aluminoborosilicate, and sodalime glass , the glass surfaces behave similarly after chromic acid cleaning. They show significant differences in their properties followin g a dry cleaning procedure, such as pyrolysis or UV/ozone cleaning. The cleaned silica surfaces show a high susceptibility to adsorbing or-ganic contamination following pyrolysis cleaning, while the pyrolyzed sodalime glass appears to be virtually immune to strongly adsorbing organic molecules. Py-rolyzed aluminoborosilicate glass shows an intermediate susceptibility to adsorb-ing organic contaminants. The chromic acid cleaned glass surfaces all show an in-termediate susceptibility to contamination by adsorbed organic molecules. Thus, it may be an oversimplification to consider a clean glass surface as a high energy substrate that is bound to attract ambient organic contamination. The wettability behavior of the cleaned glass surfaces showed features associ-ated with their exposed chemical functions. The non-dispersive interaction energy between glass and water as a function of pH showed evidence of charging of the surface silanol groups. The point of zero charge for these surface chemical func-tions was observed at pH 3. An estimate of the non-dispersive interaction energy between glass and water at the point of zero charge enables a reasonable estima-tion of the density of surface silanol groups on the cleaned glass. The trends ob-served for the surface charge as a function of pH correlate with the observed sus-ceptibility for adsorbing organic contamination to the cleaned glass surfaces. Charge-adsorbed surfactant monolayers indicated a negative surface charge on the cleaned glass, as expected for silica-based glass surfaces at neutral pH. The wettability of grafted self-assembled octadecylsilane monolayers indicated high quality coatings on the cleaned glass surfaces. The coating quality was identical for all three glass species following chromic acid cleaning. The UV/ozone cleaned glass surfaces showed the highest coating quality on the silica surface, followed by the aluminoborosilicate surface and the sodalime glass surface. The trends in coating quality for all chromic acid cleaned surfaces and UV/ozone cleaned surfaces correlate with those seen for susceptibility to organic contamina-tion of the cleaned glass surfaces exposed to unpurified liquid octane. REFERENCES

chapter |12 pages

Decontamination of sensitive equipment

chapter |8 pages

e. The transfer basket containing the items to be cleaned was lowered into the immersion sump , and statically (i.e. no liquid flow) sonicated for a finite pe-riod of time, usually 15 minutes. f. After static sonication, the rinse pump was turned on and the liquid in the immersion bath was circulated through the activated carbon columns at a rate of1,700 ml/minute for a finite period of time. The circulation time ranged fro m 15 minutes to 2 hours, depending on the purpose of the test. g. The rate of decontamination was monitored by following the concentration of the contaminant in the decontamination liquid (HFE-7100). h . Steps e and f were repeated until the presence of contaminant in the circulat-ing liquid could no longer be detected. i. When the immersion sump liquid was free of contaminant, the transfer basket was moved from the immersion sump to the superheat sump and dried for 30 minutes to remove liquid drag out. j . The transfer basket was removed from the Poly-Kleen™ system. The test pieces were removed from the basket, visually examined, photographed under visible and UV light, reweighed, and archived. I n order to maximize ultrasonic power density, the minimum amount of liquid needed to cover the parts being cleaned was used. Typically, the sump contained from 130 to 180 mm (5 to 7 inches) of liquid, which corresponds to a liquid vol-ume of approximately 15 liters to 30 liters (4 to 8 gallons) and a corresponding ul-trasonic power density of 26 to 18 watts/liter (100 to 70 watts/gallon). In prelimi-nary tests, it was noted that immersing and sonicating the test samples when the immersion sump was filled to the brim (about 53 liters (14 gallons)) did not result in effective cleaning. At that volume, the ultrasonic power density had dropped to a value of 8 watts/liter (30 watts/gallon). While this value would be considered marginal in a stainless steel ultrasonic bath, where the ultrasonic waves can be re-flected from the walls back into the liquid, in a polypropylene bath in which the walls absorb rather than reflect the ultrasonic waves, this power density level is too low. If parts were also contaminated with biological agents, after Step h, they would be sonicated in a fluorinated surfactant/HFE-7100 solution that would be circu-lated through microfilters to remove suspended materials. The parts would then be rinsed in fresh HFE-7100 to remove fluorocarbon surfactant residues, and then dried as described above. Table 3 lists the sensitive equipment decontamination experiments that were carried out in the Poly-Kleen™ system during the course of the program. The combination of equipment processed, contaminants used, and monitoring method(s) examined are listed in this table. The results of the various cleaning re-sults are summarized in Table 4. This table records the weights of the items listed in Table 3, before and after contamination, as well as the post-cleáning weight and visual appearance of these items.

chapter |1 pages

Development of a technology for generation of ice

particles

chapter |2 pages

volved the use of ultra-clean ic e particles, having a uniform grain size, for clean-ing the surface and grooves of ferrite block (Tomoji [4]). An ice blasting device utilizing stored particle s was suggested by Harima [5]. Vissisouk [6] proposed to use ice particles near melting temperature for surface decoating. Mesher [7] de-velope d a nozzle for enhancement of the surface cleaning by ice blasting. Shinichi [8] suggested an inexpensive cleaning of various surfaces by mixing ice particles, cold water and air. Niechial [9] proposed an ice blasting cleaning system contain-ing an ice crusher, a separator and a blasting gun. Settles [10] suggested produc-ing ice particles of a size range below 100 urn within the apparatus just prior to the nozzle. Although the potential use of ice blasting has been suggested by a number of inventors, the practical use is much more limited. Herb and Vissisouk [11] re-ported precision cleaning of zirconium alloys in the course of production of bi-metallic tubing by ice pellets. It was shown that ice blasting improved the quality of bimetal. The use of air-ice blasting for steel derusting was reported by Liu [12]. The following operational conditions were maintained: air pressure: 0.2-0.76 MPa; grain diameter: below 2.5 mm; ice temperature -50°C; traverse rate 90 mm/min; and standoff distanc e 50 mm. At these conditions the rate of derusting ranged from 290 mm/min at the air pressure of 0.2 MPa to 1110 mm /min at the

chapter 1|11 pages

(b)) is determined by the

chapter |2 pages

coating layer itself, an d at the interface between the coating and the substrate, causes instant fracturing and separation of coating material from the surface. In general, if a coating or contaminant is CHEMICALLY bonded to a surface, dry ice particle blasting will NOT effectively remove the coating. If the bond is PHYSICAL o r MECHANICAL in nature, such as a coating of rubber residue which is "anchored" into the porous surface of an aluminum casting, then there is a good chance that dr y ice blasting will work. Contaminants which are etched, or stained into the surfaces of metals, ceramics, plastics, or other materials typically cannot be removed with dry ice blasting. If the surface of the substrate is extremely porous or rough, providing strong mechanical "anchoring" for the contaminant or coating, dr y ice blasting may not be able to remove all of the coating, or the rate of removal may be too slow to allow dry ice blasting to be cost effective. The classic example of a contaminant that does NOT respond to dry ice blast-ing is RUST. Rust is both chemically and strongly mechanically bonded to steel substrate. Advanced stages of rust must be "chiseled" away with abrasive sand blasting. Only the thin film of powderized "flash" rust on a fresh steel surface can be effectively removed with dry ice blasting. 4.2.1.1. Inductio n (venturi) and direct acceleration blast systems - the effect of the typ e of system on available kinetic energy In a two-hose induction (venturi) carbon dioxide blastin g system, the medium particles are moved from the hopper to the "gun" chamber by suction, where they drop to a very low velocity before being induced into the outflow of the nozzle by a large flow volume of compressed air. Some more advanced two-hose systems employ a small positive pressure to the pellet delivery hose. In any type of two-hose system, since the blast medium particles have only a short distance in which to gain momentum and accelerate to the nozzle exit (usually only 200 to 300 mm), the final particle average velocity is limited to between 60 and 120 meters per second. So, in general, two-hose systems, although not so costly, are limited in their ability to deliver contaminant removal kinetic energy to the surface to be cleaned. When more blasting energy is required, these systems must be "boosted" a t the expense of much more air volume required, and higher blast pressure is re-quired as well, with much more nozzle back thrust, and very much more blast noise generated at the nozzle exit plane. The other type of solid carbon dioxide medium blasting system is like the "pressurized pot" abrasive blasting system common in the sand blasting and Plas-ti c Media Blasting industries. These systems use a single delivery hose from the hopper to the "nozzle" applicator in which both the medium particles and the compressed air travel. These systems are more complex and a little more costly than the inductive two-hose systems, but the advantages gained greatly outweigh the extra initial expense. In a single-hose solid carbon dioxide particle blasting system, sometimes referred to as a "direct acceleration " system, the medium is introduced from the hopper into a single, pre-pressurized blast hose through a sealed airlock feeder. The particles begin their acceleration and velocity increase

chapter |3 pages

F.C. Young

chapter |13 pages

Development of a generic procedure for modeling of

water jet cleaning

chapter |1 pages

REFERENCES

chapter |6 pages

the process

chapter |7 pages

• MISER; • SEAWINDS; • Pathfinder. and many others. Since the assemblies produced in this laboratory always fall in the high performance, high reliability category, cleaning is mandatory, not op-tional . With the demise of the ozone-depleting solvents that were the mainstay of the electronics industry for twenty years, it was necessary to turn to alternative chemistries an d cleaning systems to ensure cleanliness and high reliability of the surface mount assemblies (SMAs). The initial cleaning system chosen for the SMT Laboratory was a two-stage batch semi-aqueous (SA) cleaning system. Although this system worked satisfac-torily for a number of years, the decision was reached recently to replace it. Part of the reason was the increasing complexity of the SMT PWAs. Equipment to en-sure that the cleaning solution would successfully penetrate under the small standoff s and tight spacings found under the newer components now being in-creasing employed was considered mandatory. Another factor in the decision was that the initial equipment manufacturer sold off this portion of the business and no longer supported the equipment. It proved increasingly more difficult to maintain it in good working condition. In addition, isopropyl alcohol (IPA), used in the original equipment, came under increasing scrutiny by the South Coast Air Qual-ity Management District (SCAQMD). Because IP A is a volatile organic com-pound (VOC), its emission into the atmosphere is tightly controlled. The decision was made to investigate a new cleaning system and a chemistry that would sup-port JPL's need for clean PWAs to meet the newer challenges. 3. PERTINENT PROCESS INFORMATION The following JPL process information is pertinent to the discussion: • Rosin-based fluxes and pastes are used to produce all electronic hardware. Using the terminology of Mil-F-14256, the classification of these products is rosin mildly activated (RMA). • The solder paste is applied using a semi-automated screen printer ensuring that the paste is deposited in a uniform and consistent manner. Only stainless steel stencils are used in conjunction with a stainless steel squeegee. All boards are visually inspected for proper paste deposition after the stencil operation. • A laser-based solder paste height and width measurement system is used with a resolution of 0.0001 inch (2.5 jxm). This system provides real time informa-tion on the uniformity of solder paste deposition. All boards are subjected to this measurement prior to the reflow operation. • A batch reflow operation is used to create the solder joints of the SMT PWAs. The SMT PWAs are thermally profiled using aM.O.L.E.® - a thermocouple

chapter 6|7 pages

2.5. Test runs

chapter |12 pages

Microdenier fabrics for cleanroom wipers

chapter |1 pages

A METHOD FOR REMOVING DEPOSITED DUST

chapter |2 pages

the dynamics of the substrate expansion

chapter |4 pages

electromagnetic field at the particl e has to be computed numerically. An example of such a computation using a program based on [49] is given in Fig. 4. But not only doe s the Mie theory describe an enhancement of the laser intensity in the particles' near field, it also predicts that for certain values of the size parameter nd/X (d denoting the particle diameter, À the laser wavelength) the enhancement should be particularly efficient, resulting in a resonant intensity enhancement, the so-called "Mie-resonances". 3.2.2. Near-field induced substrate damage When inspecting contaminated samples by scanning electron microscopy (SEM) or atomic force microscopy (AFM ) after DLC using ns laser pulses, the consequences of the field enhancement process became obvious: all over the cleaned areas w e found substrate damages localized exactly at the former particle positions [35, 37-39]. These damages manifested as melting pools or even holes in the surface, typical examples can be seen in Fig. 5. The consequences for the laser cleaning process are obvious. The intensity enhancement reduces the maximum laser fluence that can be applied in the process. Usually in laser cleaning studies [19, 31 ] the laser fluence corresponding to the melting threshold of a bare surface is taken as the damage threshold fluence. Our experiments show clearly that this is an inadequate definition. Instead one must take into account the enhanced laser fluence underneath the particles, as it will be discussed in Section 4. Fro m the obtained AFM images we were able to analyse in detail the surface profile at the damaged sites. Here we found that for high field enhancement factors the silicon substrate was not only molten , but that some material was even ablated (see Sec. 4). The momentum transfer to the particles during the ablation process significantly contributes to the cleanin g process and hence local substrate ablation

chapter 5|2 pages

7. Cleaning thresholds and process efficiency A systematic determination of cleaning thresholds in both DLC and SLC should provide key information for the application of laser cleaning, as it allows to predict the minimum particle size that ca n be removed and to judge which of the two processes DLC or SLC is more efficient. On the basis of our measurements this comparison ca n be done for the first time and for a large size interval of particles. Perhaps the most striking differenc e in the two laser cleaning methods is the dependence of the cleaning threshold fluence on particl e size. Whereas in SLC this threshold appeared to be universal, i.e. size- and material-independent for the investigated particles, in DLC we found in agreement with other authors a size dependent threshold, with smaller particles being harder to remove than larger ones. Fro m this it is obvious that SLC is a more efficient method for small particles, i.e. for particles smaller than about 400 nm in diameter (for particles larger than 400 nm see below ) which is the most interesting size regarding the cleaning of bare silicon wafers in the semiconductor industry. In addition, SLC is superior to DLC in the minimum particle size that could be cleaned from silicon wafers. Recalling that the current minimum line width in ICs is 13 0 nm, which means that particles of about 60-70 nm in size have to be removed, this is a key information on the quality of a cleaning method. The lower size limit of particles that could be remove d by DLC was found to be 110 nm, compared to 60 nm and an efficiency above 90% in SLC. Summarizing the above, SL C is superior to DLC due to three crucial characteristics: its universal cleaning threshold, its lower threshold fluences for the relevant particle sizes, and its capability of removing sub 100 nm-particles. 5.2. Consequences of cleaning mechanisms involved Although in DLC no particles smaller than 100 nm could be removed, at a first glance it seems to be the more appropriate method for larger particles as its cleaning thresholds are distinctly lower than th e universal SLC threshold. However, for a judgement of the perspectives of SLC and DLC it is not sufficient to solely determine and compare cleaning efficiency and laser cleaning threshold fluence. On the contrary, as our studies above show very clearly, this comparison must be put into perspective by taking a closer look at the cleaning mechanisms involved. The most important physical process not taken into account in traditional investigations and only recently [34, 35, 38-40] studied is the local substrate ablation due to the enhancement of the laser intensity in the near field of the particles. The first, and most obvious, consequence of field enhancement is a locally increased laser fluence underneath the particle, and hence a decrease in the incident laser fluence necessary for particle removal. At a first sight this looks like a positive effect, but obviously a locally enhanced laser intensity drastically lowers the threshold for surface damage, and indeed we did observe surface damage caused either by melting (small particles) or local substrate ablation (large particles)

chapter |1 pages

the cleaning mechanisms are not independent of each other, e.g. adsorbed moisture may influence the field enhancement pattern. Compared to DLC the state of modeling the SLC process is at a rather initial stage . Two groups [28, 65] have suggested models to describe it. Yet these models rely on far-reaching assumptions in the description of the processes of laser induced bubble nucleation and growth as well as on the assumption of the temperature of the superheated water layer as growth medium. As our experiments on the last aspect show, it is impossible to transfer the results gained on rough metal films [50, 55, 56] to the water film/silicon system. Furthermore, it is not clear neither qualitatively no r quantitatively how the explosive evaporation differs between bulk water (as in our investigations) and water films (as in SLC) or even small water menisc i as they can be found in ambient environment DLC. Therefore, a good deal of future research on the dynamics of laser induced bubble nucleation and th e explosive evaporation in all these systems is necessary to accurately describe SLC. 6. SUMMARY In thi s paper we have described our state of knowledge on the cleaning mechanisms responsible for particle removal in laser cleaning. Beside s the well-known thermal expansion of the substrate and the explosive evaporation of a water film we identified local substrate ablation a s another cleaning mechanism. Additionally we have shown the significant impact of the explosive evaporation of atmospheric moisture adsorbed at the particles for DLC. Local substrate ablation caused by field enhancement in the particles' near field not only causes particle remova l in DLC, but inevitably also causes substrate damage. Furthermore a damage-free DLC process was not possible with the laser parameters we used in our experiments. Steam laser cleaning, o n the contrary, proved to be superior to the DLC process due to its higher efficiency, universal cleaning threshold and its capability to remove much smaller particles. These findings argue for the application of SLC in wafer cleaning and underline the need for further research on the physics of both DLC and SLC as only this knowledge will ensure a successful implementation of the technique in future industrial applications. Acknowledgements We thank Prof. B. Luk'yanchuk (DSI, Singapore) and Dr. Nikita Arnold (Johannes-Kepler-University, Linz, Austria) for useful discussions. The authors would also like to thank Dr. Bernd-Uwe Runge, Christof Bartels, Johannes Graf, Florian Lang, and Michael Olapinski (all of University of Konstanz) for constructive discussions of the findings of our experiments. Financial support by the EU TMR project "Laser Cleaning" (No. ERBFMRXCT98 0188) and the Konstanz Center for Modern

Registered in England & Wales No. 3099067
5 Howick Place | London | SW1P 1WG© 2024 Informa UK Limited