Professional Documents
Culture Documents
and Cleaning,
Volume 1
K.L. Mittal,
Editor
VSP
Surface Contamination and Cleaning, Volume 1
This page intentionally left blank
SURFACE
CONTAMINATION
AND CLEANING
VOLUME 1
Editor:
K.L. Mittal
© VSP BV 2003
ISBN 90-6764-376-9
All rights reserved. No part of this publication may be reproduced, stored in a retrieval
system, or transmitted in any form or by any means, electronic, mechanical, photocopy-
ing, recording or otherwise, without the prior permission of the copyright owner.
Preface vii
Preface
which is generally the case with many symposia proceedings − rather it reflects
information which has passed peer scrutiny. The topics covered include: mapping
of surface contaminants; various techniques for cleaning surfaces; various tech-
niques for monitoring level of cleanliness; acceptable cleanliness levels, ionic
contamination; pharmaceutical cleaning validations; cleaning of glass surfaces;
decontamination of sensitive equipment; no-chemistry process cleaning; waterjet
cleaning; cleaning with solid carbon dioxide pellet blasting; cleanroom wipers;
dust removal from solar panels and spacecraft on Mars; laser cleaning of silicon
surfaces; particle removal; implications of surface contamination and cleaning;
and future of industrial cleaning and related public policy-making.
I sincerely hope that this volume addressing many aspects and recent develop-
ments in the domain of surface contamination and cleaning will be of interest to a
wide range of people working in many different industries.
Acknowledgements
It is always a pleasure to write this particular segment of a book as it offers the
opportunity to thank those who helped in many ways. First, my sincere thanks are
extended to my colleague and friend, Dr. Robert H. Lacombe, for taking care of
the organizational aspects of this symposium. The comments from the peers are a
sine qua non to maintain the highest standard of a publication, so I am most ap-
preciative of the time and efforts of the unsung heroes (reviewers) in providing
many valuable comments. I am profusely thankful to the authors for their interest,
enthusiasm and contribution without which this book would not have seen the
light of day. In closing, my thanks go to the staff of VSP (publisher) for giving
this book a body form.
K.L. Mittal
P.O. Box 1280
Hopewell Jct., NY 12533
*Molysmophobia means fear of dirt or contamination, from Mrs. Byrne’s Dictionary of Unusual,
Obscure, and Preposterous Words, University Books, Secaucus, NJ (1974).
Surface Contamination and Cleaning, Vol. 1, pp. 1–22
Ed. K.L. Mittal
© VSP 2003
Abstract—We report the development of a new, real-time non-contacting monitor for cleanliness
verification based on tunable infrared-laser methods. New analytical capabilities are required to
maximize the efficiency of cleaning operations at a variety of federal (Department of Defense
[DoD] and Department of Energy [DOE]) and industrial facilities. These methods will lead to a re-
duction in the generation of waste streams while improving the quality of subsequent processes and
the long-term reliability of manufactured, repaired or refurbished parts.
We have demonstrated the feasibility of tunable infrared-laser imaging for the detection of con-
taminant residues common to DoD and DOE components. The approach relies on the technique of
infrared reflection spectroscopy for the detection of residues.
An optical interface for the laser-imaging method was constructed, and a series of test surfaces
was prepared with known amounts of contaminants. Independent calibration of the laser reflectance
images was performed with Fourier transform infrared (FTIR) spectroscopy. The performance of
both optical techniques was evaluated as a function of several variables, including the amount of
contaminant, surface roughness of the panel, and the presence of possible interfering species (such
as water). FTIR spectra demonstrated that a water film up to 7 µm thick would not interfere with the
effectiveness of the laser-imaging instrument. The instrumental detection limit for the laser reflec-
tance imager was determined to be on the order of a 10-20 nm thick film of a general hydrocarbon
contaminant.
1. INTRODUCTION
∗
To whom all correspondence should be addressed. Phone: (925) 294-3526,
Fax: (925) 294-3410, E-mail: smsicka@sandia.gov
2 D. Ottesen et al.
2. EXPERIMENTAL
The laser-based instrument described in this report offers the capability to rapidly
survey large surface areas and to determine the location and extent of residual hy-
drocarbon contaminants following cleaning operations. In contrast, a spectro-
scopic analysis by an FTIR-based infrared reflectance instrument is able to char-
acterize a very broad range of organic constituents and many inorganic species.
However, a surface-probing FTIR instrument measures a spectrum at only a sin-
gle small area on a sample, thus requiring broad area surveys to be done by se-
quentially probing many points. Even at a rate of ~ 10 seconds per measurement
point, this can be a time-consuming process. The rate of measurement by FTIR
spectroscopy is constrained by the relatively low spectral brightness (compared to
a laser) of the incandescent illumination sources. This makes it necessary to use
relatively long integration times to achieve an acceptable signal-to-noise ratio.
The tunable-laser-based instrument overcomes these limitations by illuminating
a broad surface area with a high-brightness infrared laser. This approach allows a
4 D. Ottesen et al.
The infrared laser light is incident on the sample surface at an angle of 60°
from the surface normal, and the specularly reflected component is detected by an
InSb focal-plane array (FPA) camera with an infrared macro-lens assembly and
an array size of 256 x 256 pixels. The FPA camera is located approximately 0.3 m
from the sample surface, and the resulting field of view is 20 x 35 mm.
FTIR instruments at both Sandia and NFESC were used to characterize the
mid-infrared spectra of contaminated surfaces via optical interfaces for grazing-
angle reflectance spectroscopy. The system at NFESC uses a commercially avail-
able sampling accessory that permits a variable angle of incidence from 30 to 80°,
which is convenient for evaluating detection limits for contaminants on a variety
of surfaces. The optical interface used by the Sandia National Laboratories FTIR
instrument was constructed with a fixed 60° angle of incidence with optics exter-
nal to the spectrometer. It also differs from the NFESC system in the large solid-
angle used both to illuminate the surface and collect reflected light. This feature is
particularly useful in the examination of rougher surfaces that cause significant
scattering of the infrared beam, with a consequent degradation in both sig-
nal/noise ratio and detection limits. Both systems use infrared polarizers to en-
hance the sensitivity of the measurements by restricting the surface illumination
to p-polarization [4]. Unless otherwise noted, all reflectance spectra presented in
this paper are for p-polarized measurements.
Table 1.
Contaminant materials used for preparation of test coupon for calibration
from the mean surface level, in millionths of an inch; therefore, a Ra value of 1.5
= 0.00000015 inches (3.8 µm). Due to the nature of metal-shop finishing proc-
esses, surface roughness values vary considerably across a given surface area.
Finishing operations also result in a directional “grain” parallel to the sample
coupons’ longitudinal direction. Surface roughness measurements, therefore, ex-
hibit large variations between measurements taken along an orientation longitudi-
nal or transverse to the polishing axis. Two surface roughness levels, 600 and 220
grit, were obtained for the other metal alloys.
Prior to contaminant application, the aluminum alloy coupons were cleaned
with acetone and underwent sonication with a clean-rinsing aqueous cleaner.
They were then thoroughly rinsed in distilled water and dried in an oven at 50°C.
Once cooled, they were weighed on a microbalance with a precision of 0.01 mg.
Two or three weighings were averaged.
Both drawing agent and lubricant contaminated Al-7075 coupons were pro-
duced by two primary deposition methods – airbrushing and manual brushing.
Several other techniques were attempted, including “wire-cator” drawing, coupon
spinning, and “manual drop and spread.” These techniques were not used to pro-
duce test samples for calibration for these particular contaminants due to the supe-
rior results obtained from airbrushing and manual brushing. Three levels of draw-
ing agent were applied by airbrushing to three Al test coupons for each of the six
surface finishes, creating a suite of 18 panels. Varying concentrations of drawing
agent in water were prepared for the airbrush solutions. Similarly, four levels of
lubricant were applied to four Al test coupons for each of six surface finishes,
creating a suite of 24 panels. Manual brushing was used for all but the least con-
taminated samples, which were airbrushed. Lubricant solutions for both tech-
Tunable IR-laser mapping of surface contaminants 9
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 gently 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.
Optical constants (n and k) were derived for the contaminant C-H stretching
vibrations using the Sandia reflectance code and a dispersion model to calculate a
fit to the experimental data for one of the test coupons [7]. Reflectance-absorption
spectra for the 2800-3000 cm-1 range were calculated for 1-µm thick films of a
specific hydrocarbon contaminant on an aluminum surface at either 60 or 75° an-
gle of incidence. This function was then used as a linear variable in conjunction
with a second-order polynomial to produce a least-squares fit of the experimental
reflectance data for the test coupons. An example is shown in Figure 2 for the
longitudinal measurements of three thicknesses of drawing-agent contaminant at
Figure 2. Linear least-squares fit of experimental reflectance data for drawing-agent contaminant on
600 grit polished aluminum surfaces. Average film thickness: (Top) 0.9 µm, (Middle) 0.4 µm, (Bot-
tom) 0.1 µm.
Tunable IR-laser mapping of surface contaminants 11
Figure 5. Integrated reflection-absorption intensities of C-H stretching bands for lubricant films de-
posited on aluminum test coupons with varying degree of surface roughness for longitudinal illumi-
nation.
Figure 6. Potential interference effects of water on C-H stretching bands of hydrocarbon lubricant
film (0.7 µm) on aluminum. Three thicknesses of water film were examined (1 µm, top; 3 µm, mid-
dle; and 7 µm, bottom).
broad H-OH stretching bands centered near 3400 cm-1, however, do not obscure
the C-H stretching bands near 2900 cm-1. This is particularly important for the ef-
fective and accurate use of the tunable infrared-laser imaging instrument, since
images are acquired for only a small number of frequencies near 3000 cm-1, in
contrast to the broad-band spectral data collected by the FTIR instrument.
Figure 7. Gray-scale on-resonance (2915 cm-1, top) and off-resonance (3000 cm-1, bottom) images
for an aluminum test panel contaminated with hydrocarbon drawing agent of 0.9-µm thickness.
tor was determined by the average intensity ratio of a clean surface (the uncon-
taminated back surface of the test panel) for the two measurement frequencies.
The ratios of successive images using the PPLN-based laser system showed a
noise level of 0.44% for the entire 65,536-pixel image under our current operating
conditions. This noise level corresponds to a hydrocarbon film thickness of ap-
proximately 10-20 nm for the species examined in this report, and is the primary
factor in determining the present instrumental detection limit.
Gray-scale images at these two frequencies for the hydrocarbon drawing-agent
(thickness of 0.9 µm on aluminum) are shown in Figure 7. Structure in the images
is primarily in the form of vertical lines that represent ridges in the aluminum
substrate formed during surface polishing operations. A darker vertical band near
the center of the image manifests the presence of an absorbing hydrocarbon in the
Tunable IR-laser mapping of surface contaminants 17
Figure 8. Reflectance images and line-intensity profiles for an aluminum test panel contaminated
with a hydrocarbon drawing-agent of 0.9-µm thickness. Laser coherence noise (A) and results of
Gaussian smoothing (C) are illustrated with corresponding intensity profiles (B and D, respectively)
sampled along the horizontal lines superimposed on the images.
Examples of reflectance ratio images for several test surfaces are shown in Fig-
ures 9 and 11 in false color. A calibrated color-table (“Rainbow”) for these false-
color images is shown in Figure 10. Images for a series of 600-grit polished alu-
minum substrates contaminated with drawing agent are presented in Figure 9.
These are the same specimens whose FTIR spectra are shown in Figure 2. Aver-
age film thicknesses for the three samples are 0.9 µm (top, left), 0.4 µm (middle,
left), and 0.1 µm (bottom, left).
The images are presented in false color format with identical dynamic range to
help visualize the location of contaminants. Hydrocarbon material was manually
deposited along the orientation of the surface polishing grooves, which is oriented
vertically in these images. Heavy deposits of the hydrocarbon residue are easily
Figure 9. False-color reflectance images and thickness profiles for three aluminum test panels con-
taminated with a hydrocarbon drawing agent (thicknesses are: 0.9 µm, top-left; 0.4 µm, middle-left;
0.1 µm, bottom-left). Corresponding line thickness profiles are shown to the right of each false-color
image.
Tunable IR-laser mapping of surface contaminants 19
Figure 10. Color bar for false-color images presented in Figures 9 and 11. Film thickness was cali-
brated by weight-gain measurements during sample preparation and by comparison with FTIR re-
flectance data.
visible in the top reflectance image (red and yellow indicating the lowest reflec-
tance, hence the thickest deposit, locations), with a particularly thick vertical band
near the center. Very few areas in this image possess high reflectance values (dark
blue) characteristic of low contamination. A horizontal line across the center of
the image indicates the thickness profile, shown in Figure 9 (top, right) for this
sample. Reflectance values have been converted to thickness of the drawing-agent
hydrocarbon contaminant using the FTIR data analysis discussed above. The data
shown here indicate the thickness averaging about 0.7 µm along the profile line,
with heavier deposits up to 2 µm.
False color images of the test surfaces contaminated with lower amounts of hy-
drocarbon (Fig. 9, middle and bottom) show much less spatial variation in the dis-
tribution of hydrocarbon residue. Hydrocarbon residues are thinner and appear as
predominantly green and light blue in the false-color images while the line pro-
files show quantitatively the thickness of lubricant in these images. The average
thickness values of the three profiles presented here are consistent with the weight
change and thickness values determined by FTIR.
The potential value of the infrared-laser imaging method for cleanliness verifi-
cation is clearly demonstrated for these test panels. For these samples distribution
of the residual hydrocarbon contaminant is quite variable. In the case of the
heaviest contaminated sample, a localized cleaning to effect substantial removal
can be profitably applied to the most heavily contaminated areas.
20 D. Ottesen et al.
Figure 11. False-color reflectance images and thickness profiles for three aluminum test panels with
a hydrocarbon drawing-agent contaminant (surface polishes are: 600-grit, top-left; 220-grit, middle-
left; 80-grit, bottom-left). Corresponding line thickness profiles are shown to the right of each false-
color image.
We also acquired reflectance ratio images for test surfaces with rougher fin-
ishes for average hydrocarbon thicknesses of 0.9 µm, again using transverse illu-
mination. False-color images and corresponding thickness profiles for these two
samples are compared to the 0.9-µm thick hydrocarbon residue deposited on the
smoothest, 600-grit polished surface in Figure 11. Average thickness values from
the three profiles are in reasonable agreement for all three test panels, demonstrat-
ing that large changes in surface roughness (0.5, 2.1, and 6.1 µm) do not substan-
tially affect the measured thickness of hydrocarbon residue.
We observe a qualitative change in the false-color images in Figure 11. In-
creasingly rough test surfaces (middle and bottom) exhibit a grainier image qual-
Tunable IR-laser mapping of surface contaminants 21
ity due to the large diversity of surface orientations relative to the infrared laser il-
lumination beam. Distribution of the hydrocarbon residue on the 220-grit surface,
however, is much more even (Fig. 11, middle, left) than for the smoothest surface
(Fig. 11, top, left). The drawing-agent material shows a strong thickness gradient
toward the right-hand side of the image for the roughest, 80-grit, surface (Fig. 11,
bottom, left) that is clearly visible despite the grainy image appearance.
4. CONCLUSIONS
The work presented in this report has shown tunable infrared-laser imaging to be
an extremely attractive method for on-line detection of hydrocarbon contaminants
and determination of their spatial distribution for efficient cleaning operations.
Calibrated test panels of hydrocarbon contaminants on metallic substrates were
prepared and characterized with FTIR grazing-angle reflectance spectroscopy.
Measurements were made over a range of film thicknesses and surface roughness,
and the derived instrument sensitivity was quite robust with respect to the degree
of surface roughness and the orientation of the reflectance unit to the direction of
polishing grooves.
Tunable infrared-laser images were acquired at both absorbing and non-
absorbing frequencies for hydrocarbon contaminants on aluminum test panels.
The thickness of the contaminant layers calculated from the laser images showed
good agreement with the measured film thickness determined by spatially aver-
aged FTIR spectroscopic results. The laser images clearly reveal the heterogene-
ous distribution of the contaminant species on the component surfaces for a vari-
ety of film thicknesses and degree of surface roughness.
Primarily, the effects of laser-coherence noise determine the current detection
limits of the laser-imaging method. The noise is introduced when an image ratio
is formed from images taken at absorbing and non-absorbing wavelengths. For
typical hydrocarbon species, the detection limit appears to be on the order of 10-
20 nm for film thickness. Improvements in the system despeckling and projection
optics may substantially decrease this noise level with an attendant increase in
sensitivity.
The configuration of a future prototype imaging system instrument will be
strongly determined by system formats that employ either a pulsed or continuous-
wave laser, and staring focal-plane array (FPA) cameras or raster-scanned imag-
ers. The design of an imaging system will include a consideration of the ultimate
instrument cost. At the present time, it appears that a continuous-wave system
with a scanned imager offers the system with the lowest cost. However, the per-
formance of some newly developed inexpensive infrared microbolometer arrays
will also be evaluated as a possible component of a low-cost pulsed imager. Fu-
ture work will enlarge both the laser illumination area and image field of view in
order to develop a prototype instrument capable of rapid large-area surveys during
cleaning verification.
22 D. Ottesen et al.
Acknowledgments
We gratefully acknowledge the financial support for these investigations by the
Department of Defense through the Strategic Environmental Research and Devel-
opment Program.
REFERENCES
1. T.A. Hoffard, C.A. Kodres and D.R. Polly, Technical Memorandum, NFESC-TM-2335-SHR
(2000).∗
2. C.A. Kodres, D.R. Polly and T.A. Hoffard, Technical Report, NFESC-TR-2067-ENV (1997).*
3. C.A. Kodres, D.R. Polly and T.A. Hoffard, Metal Finishing 95, 48-53 (1997).
4. R.G. Greenler, J. Chem. Phys. 44, 310-315 (1966).
5. D.L. Allara, in: Characterization of Metal and Polymer Surfaces, L.H. Lee (Ed.), Vol. 2, pp.
193-206, Academic Press, New York (1977).
6. W.G. Golden, in: Fourier Transform Infrared Spectroscopy-Applications to Chemical Systems,
J.R. Ferraro and L.J. Basile (Eds.), Vol. 4, pp. 315-344, Academic Press, New York (1985).
7. D.K. Ottesen, J. Electrochem. Soc. 132, 2250-2257 (1985).
8. D.K. Ottesen, L.R. Thorne and R.W. Bradshaw, Sandia Report, SAND86-8789 (1986).*
9. R.W. Bradshaw, D.K. Ottesen, L.R. Thorne, A.L. Newman and L.N. Tallerico, Sandia Report,
SAND87-8241 (1987).*
10. P.E. Powers, T.J. Kulp and S.E. Bisson, Optics Letters 23, 159-169 (1998).
∗
NFESC technical reports may be ordered from the web at www.dtic.mil. Reports from Sandia
National Laboratories may be ordered by contacting Sandia National Laboratories’ Technical Li-
braries at (505) 845-8287 or the National Technical Information Service (NTIS) at www.ntis.gov.
Surface Contamination and Cleaning, Vol. 1, pp. 23–41
Ed. K.L. Mittal
© VSP 2003
MANTOSH K. CHAWLA∗
Photo Emission Tech., Inc., 3255 Grande Vista Drive, Newbury Park, CA 91320
Abstract—Defining and maintaining a “proper” level of surface cleanliness is, at best, subjective.
Often the failure of surface preparation processes is not discovered until problems, such as poor ad-
hesion, occur down stream. Surface cleanliness is critical for good surface finish or success of sub-
sequent operations that depend on surface cleanliness. To assure consistent quality of surface
cleanliness, it is important to: understand the types of contaminants that need to be monitored, most
common cleanliness monitoring methods and their strengths and limitations, factors to be consid-
ered in choosing appropriate cleanliness monitoring method(s), and cost impact of various cleanli-
ness levels.
The selection of a cleanliness monitoring method should take into account several factors, such
as the type of substrate and the types of contaminants to be monitored, etc.
In order to define “Acceptable” level of cleanliness, a total cost approach is needed. Total cost is
defined as the cost of cleaning added to the cost of non-conformance related to a particular level of
surface cleanliness. An acceptable level of cleanliness is the one that minimizes or optimizes this
“total cost”.
Keywords: Acceptable cleanliness levels; optimum cleanliness level; total cost of cleaning; cleanliness
monitoring methods.
1. INTRODUCTION
Defining and maintaining the surface preparation at “proper” levels is the key to
good surface finish. However defining a “proper” level of surface cleanliness is,
at best, subjective. For consistent results, it is important to define “how clean is
clean”. Often the inadequacy of surface preparation processes is not discovered
until problems, such as poor adhesion, occur downstream resulting in non-
conformance due to poor surface cleanliness. To assure consistent quality of sur-
face cleanliness, it is important to: understand the types of contaminants to be
monitored; most common cleanliness monitoring techniques and their strengths
and limitations; factors that affect the choice of cleanliness monitoring tech-
nique(s); select an appropriate cleanliness monitoring method; specify a desirable
∗
Phone: (805) 499-7667, Fax: (805) 499-6854, E-mail: mchawla@photoemission.com
24 M.K. Chawla
2. TYPES OF CONTAMINATION
Cleanliness monitoring methods can also be generally classified into three differ-
ent categories: 1) Indirect Methods, 2) Direct Methods, and 3) Analytical Meth-
ods. All of these methods have certain strengths and limitations, which will be
discussed later; hence, it is important to select the method that will be most ap-
propriate for a particular application. Most of these methods are appropriate for
thin film or molecular contamination.
(1) Indirect methods – Any technique that does not take a measurement on the
actual surface or area of interest would be classified as an indirect method.
See Table 1 for some of the most common indirect methods along with their
features.
(2) Direct methods – Any technique that takes a measurement directly from the
actual surface or area of interest but does not directly identify the species of
contamination present would be classified as a direct method. Some of the
most common direct methods along with their features are listed in Table 1.
(3) Analytical methods – Any technique that identifies the species of, and meas-
ures the amount of contamination would be classified as an analytical tech-
nique. Analytical techniques can be direct or indirect; however all of them
usually determine the amount of and the species of contamination. Some of
the most common analytical methods along with their features are listed in
Table 2.
Some of the most common indirect, direct and analytical methods, with a brief
discussion of their principles of operation, are presented below.
ments have been broadly accepted for material surface analysis related to wetting,
adhesion, and absorption.
4.2.5. Optically stimulated electron emission (OSEE) [4]
A probe illuminates the surface to be tested with ultraviolet light of a particular
wavelength. This illumination stimulates the emission of electrons from the metal
surface. The emitted electrons are collected and measured as current by the in-
strument. Contamination reduces the electron emission and, therefore, the current
measured. The equipment may be connected to a computerized scanning system
that can scan a flat or cylindrical surface for cleanliness. The results can be pre-
sented as a color map or 3-D map. The user can define the level of cleanliness
each color represents in the graphic presentation of the results. This feature makes
it easy to compare “before” and “after” effect of a cleaning process or side-by-
side comparison of two pieces cleaned in alternative cleaners. OSEE is simple to
operate, fast, and relatively inexpensive. In addition, it is quantitative, non-
destructive, and non-contact. This technique detects both organic and inorganic
contamination, such as oxides, and can be used on any shape of parts as long as
the geometry of the part is presented to the sensor in a consistent manner. This
system lends itself to scan small parts or large surface areas very quickly. This
test can be used in the production line as well as for on-line real time measure-
ment of surface cleanliness. The surface of interest must emit electrons for the
technique to work. Nearly all materials of engineering importance emit electrons
when exposed to UV light.
4.2.6. MESERAN surface analyzer – (measurement and evaluation of surfaces by
evaporative rate analysis) [5]
A measurement begins by depositing onto the test surface a small volume of test
solution. A thin- end-window Geiger Müller detector is positioned above the
droplet and a metered flow of gaseous nitrogen is passed between the detector and
the test surface. To sense the volatile compound, organic compounds are used in
which one or more of the carbon atoms are Carbon-14. The β-particles given off
by the C-14 molecules at the surface are counted. Specifically measurements are
made of how many molecules there are, how many are evaporating away, how
fast they are evaporating away and, how many remain retained on the surface.
Measuring molecules provides a high degree of sensitivity and the opportunity to
analyze surfaces on a molecular scale with observations and results available in
only a few minutes. The choice of volatile chemical compounds determines
whether they react with the surface material, evaporate, or are retained by the
various physical/chemical molecular forces acting at the surface.
Chemical compounds can be found which tend to both volatilize (evaporate)
and yet tend to be retained by the surface upon which they are placed. The bal-
ance of these tendencies determines just how long the volatile compound remains
on the surface, or just how much remains. In fact, it is possible to choose a com-
pound that reacts with specific properties of the surface, or a compound where the
evaporation and/or retention are affected by certain characteristics of the surface
32 M.K. Chawla
are capable of performing these analyses, some of the most common analytical
techniques are summarized below. For a more complete list of most common ana-
lytical techniques, visit www.cea.com/table/htm, website of Charles Evans & As-
sociates. For a more comprehensive list of analytical techniques visit the website
of ESCA users group in England – www.ukesca.org/tech/list/html.
4.3.1. Chemical/elemental surface analysis techniques
4.3.1.1. Auger electron spectroscopy (AES) / scanning Auger microscopy (SAM)
[7–9]
They are used to obtain elemental composition information (and some chemical
information) from the top two to five atomic layers of a material; identify the
composition of very small features and particulates on surfaces; and provide depth
composition profiles of thin films, metals, and alloys. Micro-beam AES is also
used to study grain boundaries in high temperature alloys, and to examine fracture
surfaces to determine composition and extent of damage. The Auger electrons,
named after the discoverer of the process, are produced (among other emissions)
with discrete energies, which are specific to each element, when the surface is ir-
radiated by a finely focused electron beam. Auger electrons are collected and
measured. Auger electrons have discrete kinetic energies that are characteristic of
the emitting atoms, making this technique particularly useful for identifying ele-
mental composition. The energy level of Auger electrons is specific to a species
of contamination. The escape depth of Auger electrons (1–5 nm) makes this tech-
nique very surface sensitive.
4.3.1.2. Electron spectroscopy for chemical analysis (ESCA) [7–9]
Also known as X-ray Photoelectron Spectroscopy, or XPS, is a surface analysis
technique that provides information on both elemental identity and chemical
bonding. This information can be used to identify functional groups and molecu-
lar types. This method uses special equipment to bombard the surface of interest
with X-rays under vacuum conditions, causing electrons to be ejected from the
surface. The actual elemental composition can be quantified by measuring the en-
ergy level of ejected electrons, since each element ejects electrons at a unique en-
ergy. Its application is limited to mostly research and development, but it can be
used to calibrate and evaluate other, less sophisticated measurement methods.
4.3.1.3. Secondary ion mass spectrometry (SIMS – static) [7–9]
A surface analysis technique used for identifying molecules on a surface, as well
as for depth profiling for tracking very low concentrations of contaminants or ion-
implanted species. SIMS technique includes static SIMS (SSIMS), dynamic
SIMS, and time-of flight SIMS (TOF SIMS). SSIMS can identify organic and in-
organic species. TOF SIMS is an ultra-precise and accurate technique for measur-
ing the mass of molecules in the near-surface layers of material. A pulsed primary
ion beam is used to sputter material from the surface of the sample. Secondary
ions are collected and focused into a reflection time-of-flight (TOF) mass spec-
trometer, where they are mass analyzed. Analysis involves measuring the length
34 M.K. Chawla
of time it takes the secondary ions to reach the detector. The lighter the ion, the
less time it takes to reach the detector. From the arrival time the masses of the
species can be identified. High sensitivity depth profiling is a key feature.
4.3.1.4. Secondary ion mass spectrometry (SIMS – dynamic) [7–9]
It uses a much higher intensity bombarding beam than Static SIMS, and is a par-
ticularly sensitive (less than part-per-billion level) method for depth profiling of
dopants and trace elements in semiconductors. It can also map the X-Y distribu-
tion of atomic species with sub-micrometer spatial resolution. An energetic pri-
mary ion beam is used to sputter atoms from the sample surface. Secondary ions
emitted are mass analyzed. It is inherently a profiling technique. It uses O2 or Cs
ions to bombard a surface in high vacuum. High sensitivity depth profiling is a
key feature.
4.3.1.5. Variable-angle spectroscopic ellipsometry (VASE) [7, 8]
It is a noninvasive technique that offers information about surface composition,
layer thickness, and optical properties. Its applications include examining optical
surfaces and crystals, and measuring and analyzing band gaps in semiconductors,
optical devices, thin films, and carbon coatings on computer hard disks.
4.3.1.6. Energy dispersive X-ray (EDX) and wavelength dispersive X-ray (WDX)
analyses [7, 8]
They are often combined with a scanning electron microscope or electron micro-
probe. EDX provides simultaneous multi-element analysis and elemental mapping
capabilities for a region up to a few micrometers deep. WDX analyzes trace
amounts of one element at a time and is more quantitative than EDX. An example
of EDX application is identifying silicon nitride and titanium carbide inclusions in
stainless steel.
4.3.2. Bulk analysis techniques
The following are several analytical techniques that typically are used for chemi-
cal or elemental analysis of bulk materials, but these can also be adapted for the
characterization of surfaces and thin films. Many times these techniques are used
in industry for characterizing surfaces, sometimes without full knowledge of the
strengths and limitations of these techniques. It is hoped that information about
how these techniques work, their strengths and limitations would help the reader
in determining their usefulness and limitations for their applications.
4.3.2.1. Fourier transform infrared (FTIR) spectroscopy [7, 8]
It provides information about the chemical bonding and molecular structure of or-
ganics and some inorganic solids, liquids, gases and films. This technique is espe-
cially good for identifying unknowns when reference IR spectra are available.
When an infrared beam impinges on a surface, the molecular constituents vibrate
in the infrared regime. The identities, surrounding environments, and concentra-
tions of these oscillating chemical bonds can be determined. FTIR is a powerful
analytical tool for characterizing and identifying organic molecules. The IR spec-
trum of an organic compound serves as its fingerprint and provides information
Monitoring cleanliness and defining acceptable cleanliness levels 35
about chemical bonding and molecular structure. This information can be used to
detect the 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. The 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-ray analysis
(EDX) [7, 8]
The SEM produces 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, thereby 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
of arsenic, lead, mercury and cadmium on hazardous, chemical fume hoods.
36 M.K. Chawla
There are several factors that should be considered in selecting a method for
monitoring surface cleanliness. The factors discussed here are the ones that are
most important but by no means represent a complete list of factors that should be
considered. There may be other factors that are pertinent to a particular applica-
tion that should be considered.
(1) Type of contaminant – One of the first factors that should be considered in se-
lecting a cleanliness monitoring method is the type of contaminant that need
to be monitored. Is the contaminant particulate or thin film type? If thin film
contamination, is it organic or inorganic or both? Does the technique under
consideration monitor the type of contaminants that need to be monitored?
(2) Types of substrates – What type of substrate is going to be monitored? Are
the techniques under consideration capable of monitoring this type of sub-
strates? Are the techniques likely to damage the substrate to be monitored?
(3) Level of cleanliness to be monitored – It is important that the level of contami-
nation that is expected or tolerable can be monitored by the technique under
consideration. It is recommended that samples with different levels of contami-
nation be monitored with the technique(s) under consideration. In evaluating
the technique for suitability, prepared samples should have levels of contamina-
tion spanning a range from 0% (i.e. clean surfaces) to maybe 200% of the ex-
pected level of contamination on the surface. The technique(s) should not have
any problem in distinguishing between different levels of contamination.
(4) Features of monitoring method – It is important to consider various features
of the method under consideration. For example, is the technique non-contact
and/or non-destructive? Does the technique require deposit of some medium
on the surface? For example, the contact angle measurement requires that a
droplet of water be placed on the surface of interest. How large an area can
the technique measure? Is it sensitive to surface roughness? Can the technique
check parts of different geometries? Can the technique be used on-line? Is the
technique suitable for the environment it is going to be used in? Does the
technique cause any permanent changes to the surface? All of these questions
should be considered to determine the most appropriate monitoring technique
for a particular application.
(5) Measurement speed – Is the measurement speed critical for the application
under consideration? If so, how fast can the technique make a measurement?
Is the speed sufficient to keep up with the production flow?
(6) Acquisition and operating cost – How does the acquisition cost compare
among the techniques that meet other requirements for the application? Are
there any expendable items that would have to be purchased for continued use
of the equipment? How much does that add to the operating cost? What are
the maintenance and calibration requirements and how much these require-
Monitoring cleanliness and defining acceptable cleanliness levels 37
ments will add to the operating cost? All these questions should be answered
to truly compare the total cost of any cleanliness monitoring system.
(7) Skill level required – The operator skill level can be a key factor in the use of
some techniques, particularly the analytical techniques. Some techniques may
involve interpretation of the data to determine the quality of surface cleanliness.
These factors should also be considered in the selection of a cleanliness measur-
ing technique. A high operator skill level will result in higher operating cost. In
the event of personnel turnover, higher training costs may also be incurred.
For every level of cleanliness, there is a cost to achieve that level of cleanliness.
There is corresponding level of failure/non-conformance for each cleanliness
level, hence cost of failures/non-conformance. “Total Cost” of achieving a certain
level of cleanliness is the sum of these two costs.
As the achieved level of surface cleanliness increases, the cost of cleaning also
increases. Eventually the incremental cost of cleaning rises exponentially. Hence
the cost of surface cleaning is directly proportional to the surface cleanliness level.
Intuitively, we know that the higher the cleanliness level the lower the fail-
ure/non-conformance rate, hence cost, due to surface cleanliness. The incremental
drop in costs due to lower failure/non-conformance also exhibits exponential rela-
tionship. Hence the cost of failures/non-conformance is inversely proportional to
the surface cleanliness level. If both of these costs were plotted on a graph, the
typical result would be like the one shown in Figure 1.
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 the optimum level. A small range around the optimum level of cleanliness
can be established as the “Acceptable Level” of cleanliness.
results. These parts then must be followed through the process to measure the
level of success for each part at the subsequent operation. The level of cleanliness
that results in the desired minimum level of success is the minimum level of
cleanliness that must be achieved in production. This approach has its limitations.
For example, the results depend on what level of cleanliness is being achieved in
production. If the surface is “too clean” there may not be enough variation in the
cleanliness level to identify the point where minimum success is achieved. On the
other hand, if the surface is not clean enough the desirable success may not occur.
A more proactive approach is to prepare parts with different levels of surface
cleanliness, measure and record the cleanliness level and follow through with the
subsequent operation to correlate the success level with cleanliness level. It is
recommended that the range of cleanliness should be as wide as possible to help
identify the minimum level of cleanliness. Once again it is important that a statis-
tically significant sample be used. It is also recommended that, if possible, several
cleanliness measurements should be taken from each part and the mean cleanli-
ness level be correlated to the mean success level. Figure 2 [10] graphically de-
picts the typical result of correlating the success level to surface cleanliness level.
A minimum level of cleanliness is the one that corresponds to the target minimum
level of success.
cleanliness achieved should be monitored. The cost associated with a given level
of cleanliness and the cost of non-conformance associated with that given level of
surface cleanliness should be combined to determine the “total cost”. Changes
should be made to the cleaning process to minimize the “total cost”, i.e. optimize
the “total cost”. The level of cleanliness associated with the “optimum total cost”
should be considered the optimum or “acceptable” level of cleanliness.
Surface cleanliness monitoring system must be used to monitor the process and
assure that the desired cleanliness level is being achieved on an on-going basis.
Surface cleanliness monitoring system can be very useful in assessing how the
surface cleanliness level is affected by making changes to the cleaning process or
for evaluating alternative cleaning processes for their ability to achieve the de-
sired cleanliness level.
The required level of cleaning agent concentration in the cleaning solution can
also be objectively determined and maintained by using a surface cleanliness
monitoring system. Measuring the effect of varying the concentration level of the
cleaning agent on surface cleanliness can help determine the “optimum” concen-
tration level.
In most industries, the chemical or cleaning agent replenishment schedule is
usually time-dependent. The success of this approach relies on the level of con-
tamination on each part and the number of parts processed in a given time interval
being relatively constant. In real life, the amount of contamination can vary con-
siderably from part to part. In addition, the number of parts cleaned during a
given time frame can also vary considerably. A time-dependent replenishment
schedule is not the ideal way of assuring product quality. On-going, in-process
monitoring of surface cleanliness helps in replenishment of chemicals or cleaning
agents only when needed, and not based on a pre-determined, somewhat arbitrary
schedule.
9. SUMMARY
required to operate the system. Surface cleanliness monitoring method may be di-
rect, indirect or analytical. A monitoring method can be used to optimize the
cleaning process by varying different parameters of the cleaning process while
monitoring surface cleanliness to see how it is affected by the change. It can also
help in ongoing monitoring of the cleaning process to assure that the desirable
level of cleanliness is being achieved.
REFERENCES
1. B. Kanegsberg and M. Chawla, “Non Volatile Residue”, A2C2 Magazine, 5, No. 3, 41 (2002)
and 5, No. 4, 45 (2002).
2. R.L. Gause, “A Noncontacting Scanning Non Contact Photoelectron Emission Technique for
Bonding Surface Cleanliness Inspection”, Marshall Space Flight Center, Huntsville, AL, pre-
sented at Fifth Annual NASA Workshop, Cocoa Beach, Florida (1987).
3. B. Kanegsberg and M. Chawla, “Contact Angle”, A2C2 Magazine, 4, No. 8, 41 (2001).
4. Surface Quality Monitors Brochure, Photo Emission Tech., Inc.
5. B. Kanegsberg and M. Chawla, “MESERAN”, A2C2 Magazine, 4, No. 9, 49 (2001).
6. B. Kanegsberg and M. Chawla, “Total Organic Carbon”, A2C2 Magazine, 4, No. 10, 37 (2001).
7. Charles Evans Associates Website – www.cea.com
8. R.D. Cormia, “Problem-Solving SURFACE ANALYSIS Techniques”, Surface Sciences Labo-
ratories, Mountain View, CA: Advanced Materials & Processes, 16-23 (Dec. 1992).
9. Measurement and Characterization Website – www.nrel.gov/measurements/surface/html
10. M. Chawla, “How Clean is Clean? Measuring Surface Cleanliness and Defining Acceptable
Level of Cleanliness”, in Handbook for Critical Cleaning, B. Kanegsberg and E. Kanegsberg
(Eds.), pp. 415-430, CRC Press, Boca Raton, FL (2001).
This page intentionally left blank
Surface Contamination and Cleaning, Vol. 1, pp. 43–48
Ed. K.L. Mittal
© VSP 2003
BEVERLY NEWTON∗
Dionex Corporation, 500 Mercury Drive, Sunnyvale, CA 95032, USA
Abstract—Surface ionic contamination can cause device failures. In order to find the source of the
contamination many questions must be answered first. Are the failures due to incoming materials
that are not clean? Has there been a change in the process that is introducing contamination? What is
the exact nature of the contaminant, ionic, particulate, metallic, etc? Is there a training issue that
needs to be addressed. Can the failure be tested for or is it a long term reliability problem? These are
just a few of the questions that must be answered as part of the troubleshooting process. This paper
addresses how ion chromatography can be used to troubleshoot a manufacturing or cleaning process
and to assure the quality and reliability of electronic devices. Topics covered include:
1. What is ion chromatography.
2. How does it differ from other cleanliness testing methods.
3. How can ion chromatography be used to troubleshoot a cleaning process.
4. Real life examples showing how the use of ion chromatography has improved cleaning processes.
1. INTRODUCTION
As electronic devices and assemblies become smaller and more complex, the re-
quirements for improved quality control of product cleanliness have begun to es-
calate. Surface contamination from ions such as chloride, bromide, sodium, and
organic acids has been shown to cause failures in electronic devices [1]. Ionic
residues can cause corrosion, metal migration and electrical leakage. The failures
cased by these residues may be hard or soft failures and may occur several
months after the product has been manufactured and shipped to customers. Upon
re-testing the returned product, the failures can be intermittent or “no trouble
found” making troubleshooting the device for a root cause of the failure difficult.
These residues may be on the exposed surface of an electronic device, they may
be encapsulated in flux or resin deposits, they may be trapped under surface
mounted devices or they may be encapsulated in polymer finishes (Figure 1).
∗
Phone: (408)4814272, Fax: (408)7372470, E-mail: beverly.newton@dionex.com
44 B. Newton
Figure 2. Analysis of a cassette used to transport disk drive components during manufacture.
Ionic contamination can also be found on materials that come in contact with
electrical devices during manufacture, e.g. gloves, cassettes, etc. (Figure 2). These
manufacturing consumables can transfer contamination to the manufactured
products and need to be examined for contamination in the same way that the fi-
nal product is evaluated.
Potentially corrosive ions found on printed circuit boards and electronic devices
include:
● Bromide - commonly found in solder masks
● Sulfate - comes from a variety of materials such as oils and release agents
● Chloride - commonly found in fluxes
● Organic acids such as adipic or succinic acid - found in fluxes
Typically, the higher the concentration of corrosive ions on a particular assem-
bly, the higher the risk of electrochemical failure.
There has been a growing interest in the analysis of ionic contamination on elec-
tronic components. Absolute contamination level requirements and guidelines
have not been determined; however, Contamination Studies Laboratory (CSL,
Kokomo, IN) recommends maximum levels of chloride ion in the range 1.0 µg/sq.
in for assembled boards with sensitive components such as microBGAs. The level
recommended for bare boards is less than 2.0 µg/sq. in [3].
Ion chromatography provides the unique capability of identifying the individ-
ual ions for a given contamination issue. Since the source for chloride contamina-
tion can be much different than the source for organic acid contamination it is im-
portant to know which ions the manufacturer is dealing with in order to
understand and correct the root cause of the problem. This is not possible with re-
sistivity of solvent extract (ROSE) measurements. The capability to identify and
quantify individual ions makes ion chromatography a valuable troubleshooting
tool for process contamination issues and process monitoring programs. In addi-
tion to being the most economical analytical technique for monitoring multiple
ions, ion chromatography also provides the ability to distinguish between noncor-
rosive and corrosive ions, something that ROSE testing is unable to do.
A number of studies have been published to show the use of ion chromatogra-
phy to troubleshoot reliability issues with electronic products. One of the best
sources of case study information can be found on the web site for Contamination
Studies Laboratory (CSL) at www.residues.com. CSL regularly publishes case
studies showing the hazards of ionic contamination to electronic device reliability
on their web site and in each issue of Circuits Assembly magazine.
A good explanation of how ion chromatography has been used to identify
sources of CAF (conductive anodic filament) failures can be found in a study
completed by Ready et al. [4].
Several studies [5-7] have been completed on the analysis of ionic contamina-
tion on failed disk drive components.
As mentioned earlier, manufacturing materials and packaging can be an impor-
tant source of ionic contamination. Two recent studies by Lin and Graves [8] and
Bahten and McMullen [9] provide information on the use of ion chromatography
for the analysis of ionic contamination on materials such as pink poly film (a
common packaging material) and cleaning brushes.
6. CONCLUSION
The ion chromatography, ROSE and modified ROSE test methods have been de-
veloped to allow electronics manufacturers to identify and control ionic contami-
nation before it evolves into a failed component or board. Tracking ionic con-
tamination requires systematic troubleshooting and improved cleanliness of the
product as it is manufactured. This means cleaner raw materials and processes
which are controlled by systematic analysis using standard methods such as those
documented by IPC, IDEMA, IEST (Institute of Environmental Sciences and
Technology), and SEMI (Semiconductor Equipment and Materials International).
REFERENCES
1. D. Yang, C. Lee, Y. Yang, E. Kaiser, S. Heberling and B. Newton, Precision Cleaning, 17-23
(May 1998).
2. B. Newton, Precision Cleaning, 38-39 (March 2000).
3. D. Pauls and T. Munson, Circuits Assembly, 110-112 (September 1998).
4. W.J. Ready, B.A. Smith, L.J. Turbini and S.R. Stock, Mater. Res. Soc. Symp. Proc. 515, 45-54
(1998).
5. A. Toxen, A2C2, 13-16 (September 1998).
6. P. Mee, M. Smallen and D. Vickers, IDEMA Insight, 1 (March/April 1997).
7. J. Thompson, T. Prommanuwat, A. Siriraks and S. Heberling, IDEMA Insight, 24-29 (May/June
1999).
8. S. Lin and S. Graves, Micro, 95-106 (October 1998).
9. K. Bahten and D. McMullen, Proc. Semiconductor Pure Water and Chemicals Conference, 355-
364 (March 1999).
Surface Contamination and Cleaning, Vol. 1, pp. 49–73
Ed. K.L. Mittal
© VSP 2003
DEFINITIONS
In this paper a number of abbreviations, special terms, and trademarks are em-
ployed:
(1) µCi means microCurie, a unit of radiation which corresponds to 3.7 E 4
(37,000) disintegrations per second.
(2) Carbon-14 (C-14) refers to the radioactive isotope of the element Carbon, an
isotope which emits only soft or low energy beta particles; most C-14 beta
particles are stopped by a sheet of paper.
(3) USNRC EXEMPT means the very low level of Carbon-14 that is not regu-
lated by the U S Nuclear Regulatory Commission. No license is required for
∗
To whom all correspondence should be addressed. Phone: (816) 997-3529,
Fax: (816) 997-2049, E-mail: mbenkovich@kcp.com
†
Operated for the United States Department of Energy under prime contract
DE-AC04-01AL66850. ãCopyright Honeywell LLC, 2002.
50 M.G. Benkovich and J.L. Anderson
1. INTRODUCTION
contamination on surfaces and the crosslink density (or degree of cure) in poly-
mers. In addition, the MESERAN technique can be used for quantifying chemi-
cally active sites on surfaces [1-3].
Honeywell FM&T, KCP (henceforth KCP) has been using MESERAN Ana-
lyzers for approximately 30 years to detect and quantify organic contamination on
parts and evaluate various cleaning processes for removing organic contamina-
tion. KCP has used MESERAN Analyzers extensively to evaluate the ability of
alternate solvents and processes for removing specific organic contaminants to
eliminate the use of chlorinated and fluorinated solvents [4-10]. In recent years,
KCP has been working on several projects with The MESERAN Company to im-
prove data analysis and develop new methods for using the MESERAN technol-
ogy [11-14].
Each test takes less than 3 minutes and the amount of radiochemical employed
is EXEMPT from U S Nuclear Regulatory Commission and/or ‘Agreement State’
licensing regulations due to the very low level of C-14 involved.
For the measurement of microorganic residues, the MESERAN method may be
used:
(1) Directly on a flat or concave surface and any microorganic residue thereon
which is chemically compatible with the particular radiochemical employed, or
(2) Indirectly using an extracting solvent followed by depositing and evaporating
an aliquot amount onto a “clean” reference surface. Subsequent deposition
and evaporation of the radiochemical solution permits measurement of the
amount of deposited residue by comparing the results with previously ob-
tained standards similarly deposited from volumetric dilutions.
For non-polar and/or hydrocarbon type residues, tridecane-C14 in cyclopentane
(designated BK) is employed. For more polar residues, tetrabromoethane-C14 in
cyclopentane (designated AK) is used. In order to provide a high number of de-
tected emissions for the minimal amount of radiochemical deposited, the
tridecane-C14 has a specific activity of approximately 57 µCi/µmole (one carbon
atom of tridecane is essentially pure C14 isotope) while the tetrabromoethane has
both carbon atoms labeled (approximately 114 µCi/µmole). Approximately 200
ng of radiochemical are deposited in each test with similar levels of radioactivity.
The MESERAN method assumes that the particular radiochemical employed is
chemically compatible with the residue, that the test solution droplet covers all of
the residue, and that the test solution solvent substantially dissolves the residue
within the time period of the solvent evaporation. Attention to the avoidance of
inadvertent contamination and the maintenance of reasonably constant tempera-
ture and pressure are required for optimal reproducibility from test to test.
are thus the primary factors in determining the tendency of each molecule to re-
main in solution or conversely to escape from the liquid portion of the air/liquid
(or semisolid) interface. In the MESERAN technology, which employs only a
monolayer equivalent of the radiochemical, the observed rate of evaporation is
thus a function of the residual concentration of the non-evaporated molecules of
the Carbon-14 radiochemical. Figure 3 illustrates the typical evaporation of the
radiochemical solution from a clean surface. The A-B line represents the evapora-
tion of the low boiling solvent (e.g., cyclopentane). The rationale for the initial in-
crease in counts/second is that the C-14 soft beta emissions are partially absorbed
by the solvent molecules. B represents the point at which substantially all of the
low boiling solvent has evaporated and the maximal amount of residual radiation
reaches the GM detector. The B-C line represents the evaporation of the radio-
chemical from the surface under the conditions of the test. C represents a level
where the GM detector can no longer adequately differentiate the residual radia-
tion from background.
A solution of the high-boiling-but-volatile tridecane-C14 in higher boiling hy-
drocarbons (i.e., contamination) follows a similar but slower path than does the
evaporation of the labeled tridecane itself since the non-volatile “residue” mole-
cules occupy increasing portions of the liquid (or semi-solid) interface. The rate at
which the solvent evaporates is slowed somewhat and the rate at which the radio-
chemical evaporates is slowed considerably with the observed rate of evaporation
being a function of the amount of residue on the surface. The observed rate of
evaporation of the radiochemical (the slope expressed as a positive integer) thus
is an inverse measure of the amount of non-evaporating residue. The lower the
slope, the more the residue and vice versa. Figure 4 illustrates typical evapora-
tions of the radiochemical solution with increasing amounts of residue. ABC is
repeated from Figure 3 and illustrates a typical evaporation of the radiochemical
solution with no interactions from residue (i.e., a clean substrate). A*B*C* illus-
Figure 5. Plot of raw data showing low, medium, and high levels of organic contamination.
A new method using MESERAN technique 55
In an effort to reduce the effect of the randomness of radiation, the data (ln
(counts – background)) are “smoothed” from six seconds through 115 seconds
(for 120 second length tests) and replotted. The smoothing is accomplished by
summing the ln (counts – background) at the sixth second with the previous five
seconds and the subsequent five seconds and dividing this number by 10. A divi-
sor of 10 is used instead of 11 because it is statistically sound to take the number
of items being smoothed and subtract one from it because a degree of freedom is
lost. This process is carried out through the 115th second and the subsequent data
are replotted as smoothed (ln (counts – background)) versus time. Figure 6 repre-
sents the same data as in Figure 5 except that the data in Figure 6 are logarithmi-
cally smoothed to increase the reliability of the individual points. The plotted
smoothed curve is then analyzed via linear regression to determine the slope of
the post-peak line (down to near background) which best fits the data representing
the evaporative process. The determined slope is multiplied by –10,000 to convert
it to a positive integer; this becomes the reported MESERAN slope value with
units of smoothed (ln (counts – background))/sec x (–10,0000).
Figure 6. Logarithmic plot of smoothed data showing low, medium, and high levels of organic con-
tamination.
56 M.G. Benkovich and J.L. Anderson
Figure 8. Calibration curve for DOP on aluminum panels using MESERAN total counts.
Figure 9. Calibration curve for DOP on aluminum panels using MESERAN low variance slopes.
A new method using MESERAN technique 59
3. EXPERIMENTAL
3.1. Purpose
KCP conducted a cleanliness evaluation to determine the NVOR amounts on
aluminum and stainless steel panels which were machined using KCP machining
fluids and cleaning processes chosen for production of hardware for a particular
customer. This section of the paper describes how KCP used recent advances in
MESERAN technology to determine the NVOR amounts on four stainless steel
panels and four aluminum panels (31 in2 each, excluding edges) by extracting the
panels with methylene chloride and quantifying the extracted residues in mg/ft2.
The virgin methylene chloride solvent was also evaluated so that its contribution
could be subtracted from the solvent extracts. The customer specifically requested
that the results be reported in mg/ft2 as opposed to µg/cm2, therefore that is how
the results are reported in this paper. It is common practice in the Aerospace in-
dustry (as well as other industries) to report contamination amounts on large sur-
faces in mg/ft2. The conversion factors for these units are 1 mg/ft2 is equivalent to
1.0764 µg/cm2 or 1 µg/cm2 is equivalent to 0.929 mg/ft2.
group. This evaluation was conducted on these eight small samples to allow
cleanliness verifications to be performed at KCP prior to cleaning large samples
(25.4 cm x 25.4 cm x 1.3 cm) that would be sent to the customer for cleanliness
verification.
as possible, they still will contribute some level of error to the reported results,
and at the low levels of contamination being discussed, this error may be signifi-
cant. In addition, gravimetric analysis is typically performed after rinsing a 1 ft2
(144 in2) sample area while the samples used in this evaluation were significantly
smaller (31 in2).
20 minutes per 250 µL sample when performed at 30°C. The evaporated residue
was then tested with the MESERAN Analyzer. A minimum of 5 replicates were
run for each condition tested. The data were compared to calibrations performed
using the KCP hydrocarbon blend cutting oil (mixture of 70% Pennex N 47 and
30% Hangsterfer’s Hard Cut # 511) deposited on the same stainless steel sub-
strates. Interpolations between the calibration points were performed to determine
how much residue was in the extract. This amount was then multiplied by the ap-
propriate factor to determine the quantity of residue in the entire amount of ex-
tract that was obtained from each panel. The amount of extract obtained from
each sample was measured so this calculation could be performed. Recovery of
70–80% of the 100-mL of methylene chloride solvent was typical. The NVR of
the methylene chloride was analyzed in a similar fashion to generate a baseline
value for the solvent. Ultimately, the NVR of the methylene chloride was sub-
tracted to determine the amount of residue actually extracted from the sample
panels. This result was then converted to an amount per square foot for compari-
son with the customer’s specification.
6. The weighing trays were removed from the oven and placed in a desiccator to
cool.
7. The weighing trays were placed on an analytical balance and massed to the
nearest 0.01 mg.
8. The NVR of the methylene chloride was calculated by subtracting the result
of step 3 from the result of step 7.
Two samples of virgin methylene chloride were evaluated using this procedure.
One sample was 100 mL of methylene chloride taken straight from the original
bottle, measured in a clean graduated cylinder, poured into a precleaned glass bot-
tle, and then sealed with caps containing Teflon inserts. The second methylene
chloride NVR sample was obtained using the following procedure: (1) 100 mL of
methylene chloride from the original bottle was measured in a clean graduated
cylinder, (2) the methylene chloride was poured into a precleaned Teflon squeeze
bottle, (3) the precleaned stainless steel funnel used to hold the samples that were
extracted was rinsed with 100 mL of methylene chloride, (4) the methylene chlo-
ride drained through the funnel and was captured in a precleaned glass bottle, and
(5) the glass bottle was sealed with caps containing Teflon inserts. This process
captured 89 mL of methylene chloride for analysis.
After aliquots from the extracts were analyzed using the MicroSolventEvapora-
tor and MESERAN Analyzer, the remainder of each of the extracts was poured
into clean graduated cylinders to measure the amount of solvent left for gravimet-
ric analysis. The extracts were then poured back into their respective bottles and
sent to be evaporated and measured gravimetrically.
Table 1.
Volatility of KCP machining fluids
Figure 10. Plot of various amounts of hydrocarbon blend (70% Pennex N 47 and Hangsterfer’s
Hard Cut # 511) on stainless steel disks using smoothed Ln (counts – background). (MESERAN
tests carried out using radiochemical BK).
Figure 11. MESERAN low variance slope calibration of hydrocarbon blend (70% Pennex N 47 &
30% Hangsterfer’s Hard Cut # 511) on stainless steel disks with 0.015 inch/RPM spiral grooves.
A new method using MESERAN technique 67
same area until a large volume of solvent has been evaporated). In addition, these
stainless steel disks have grooves machined in them which act as a self-centering
mechanism so that the solvent evaporates in a confined space and does not mi-
grate elsewhere.
After the methylene chloride was evaporated onto the stainless steel disks with
the MicroSolventEvaporator, the disks were tested with the MESERAN Analyzer
using radiochemical BK (best suited for detecting nonpolar residues). Table 2
shows the average low variance MESERAN slope results for the 250-µL deposi-
tions from the methylene chloride extracts.
The results were compared to those obtained previously from the calibration
using known amounts of the KCP hydrocarbon blend machining fluid (mixture of
70% Pennex N 47 and 30% Hangsterfer’s Hard Cut # 511) on these same sub-
strates, which were shown previously in Figure 11. The average low variance
slope obtained for the methylene chloride blanks was 2412. The MESERAN
slope value has units of smoothed (ln (counts – background))/sec x (–10,0000);
however, the units are generally not given when the data are shown. A linear-log
Table 2.
MESERAN low variance slope results for 250-µL depositions from methylene chloride extractions
interpolation using the values from the calibration in Figure 11 was performed to
calculate the contamination level. The slope of 2412 corresponds to a contamina-
tion level of 3 ng (or 0.000003 mg) for the 250 µL of methylene chloride evapo-
rated on the stainless steel disks. Therefore, the NVR of the 100-mL methylene
chloride blanks would be 1.2 µg (or 0.0012 mg) as calculated below:
(3 ng/250 µL) x (1000 µL/1 mL) x (100 mL) x (1 µg/1000 ng) = 1.2 µg (or 0.0012 mg)
The contamination results obtained from the MESERAN Analysis for the me-
thylene chloride extracts from the panels tested are shown in Table 3. The average
low variance slopes obtained for the samples were converted into equivalent con-
tamination amounts (in nanograms) for the 250 µL of extract evaporated by per-
forming a linear-log interpolation using the values from the calibration curve in
Figure 11. Then the amount of contamination in all of the extract collected for
each sample was calculated (shown in micrograms) using similar equations as
shown previously for the methylene chloride blanks. Next, the methylene chloride
NVR was subtracted from the contamination amount in all of the extract collected
to determine the NVR amount extracted from each sample (this result is shown as
µg/31 in2). Finally, this contamination was converted to an amount per square
foot to compare to the customer requirement of <0.1 mg/ft2. The results of all
samples tested passed the requirement of <0.1 mg/ft2 and most of them were sig-
nificantly lower.
Table 5.
Gravimetric analysis of methylene chloride extracts from 100 in2 aluminum and stainless steel pan-
els performed by the customer
5. CONCLUSIONS
and different methods were used for determining the NVR levels, the results ob-
tained were remarkably consistent.
The NVR data determined by KCP with the MicroSolventEvaporator and ME-
SERAN technique and the NVR data determined gravimetrically by the customer
follow a similar pattern. Tables 3 and 5 indicate that the results of both test meth-
ods exhibited the same cleanliness level relationship between the aluminum and
stainless steel panels for each specific contaminant evaluated. Hydrocarbon
blend-contaminated samples had more contamination after cleaning on the
stainless steel panels than on the aluminum panels. Cimtech 200-contaminated
samples had more contamination after cleaning on the aluminum panels than on
the stainless steel panels. Trimsol-contaminated samples had more contamination
after cleaning on the aluminum panels than on the stainless steel panels. Cimstar
3700-contaminated samples had more contamination after cleaning on the
stainless steel panels than on the aluminum panels.
Acknowledgements
The work of Garth Christoff, Tom Hand, and Ed Fuller was instrumental in the
performance of these evaluations.
REFERENCES
1. J.L. Anderson, “Quantitative Detection of Surface Contaminants,” Journal of the American As-
sociation of Contamination Control, II (6),9 (1963).
2. J.L. Anderson et al., “Measurement and Evaluation of Surfaces and Surface Phenomena by
Evaporative Rate Analyses,” Journal of Paint Technology, 40, No. 523, 320-327 (August 1968).
3. J.L. Anderson, “Evaporative Rate Analysis: Its First Decade”, in: Characterization of Metal and
Polymer Surfaces, L.H. Lee (Ed.), Vol 2, pp. 409-427, Academic Press, New York (1977). This
paper summarizes all known references prior to 1975.
4. L.C. Jackson, “Solubility Parameters and Evaporative Rate Analyses in Organic Residue Char-
acterization” (Topical Report), UNCLASSIFIED, Bendix-Kansas City Division: BDX-613-
1099, March 1974. (Available from NTIS)
5. L.C. Jackson, “Contaminant Removal Based on Solubility Parameter and Evaporative Rate
Analysis Technologies” (Topical Report), UNCLASSIFIED, Bendix-Kansas City Division:
BDX-613-1128, August 1974. (Available from NTIS)
6. L.C. Jackson, “How to Select a Substrate Cleaning Solvent,” Adhesives Age, 22-31 (December
1974).
7. L.C. Jackson, “Solvent Cleaning Process Efficiency,” Adhesives Age, 31-34 (July 1976).
8. L.C. Jackson, “Removal of Silicone Grease and Oil Contaminants,” Adhesives Age, 29-32
(April 1977).
9. L.C. Jackson, “Contaminant Cleaning for Critical Electrical Assembly Areas” (Final Report),
UNCLASSIFIED, Bendix-Kansas City Division: BDX-613-1695, February 1978. (Available
from NTIS)
10. M.G. Benkovich, “Solvent Substitution for Electronic Products,” International Journal of Envi-
ronmentally Conscious Manufacturing, 1, No. 1, 27-32 (1992).
11. M.G. Benkovich and J.L. Anderson, “Measurement of Organic Residues on Surfaces to a Low
Fraction of a Monolayer,” Precision Cleaning, 16-28 (May 1996). This paper includes many of
the more current references to MESERAN technology.
A new method using MESERAN technique 73
12. M.G. Benkovich and J.L. Anderson, “Quantification of MicroOrganic Residues to Low
Nanogram Levels,” Precision Cleaning ’96 Proceedings, 115-122 (1996).
13. J.L. Anderson, R.F. Russell and M.G. Benkovich, “Quantitative Measurement of Extremely
Low Levels of Non-Volatile Residues (NVR) on Surfaces and in Liquids,” Precision Cleaning
’97 Proceedings, 96-108 (1997).
14. J.L. Anderson, R.F. Russell and M.G. Benkovich, “Solvent NVR: A Problem and a Solution,”
CleanTech ’98 Proceedings, 331-340 (1998).
15. MESERAN Analyzer Literature, ERA Systems Inc., The MESERAN Company, Chattanooga,
TN.
This page intentionally left blank
Surface Contamination and Cleaning, Vol. 1, pp. 75–84
Ed. K.L. Mittal
© VSP 2003
HERBERT J. KAISER∗
STERIS Corporation, P.O. Box 147, St. Louis, MO 63166
Abstract—Cleaning validations are very difficult to perform. They can be made easier if an appro-
priate method for analyzing the samples is used. The method used should be based on the previously
established residue limits of both the active and cleaning agent. There are many choices of analyti-
cal techniques that can potentially be used. This article describes various analytical techniques that
are available for use, particularly for cleaning agent residues.
Keywords: Pharmaceutical cleaning; cleaning validation; total organic carbon; high performance
liquid chromatography; Fourier transform infrared spectroscopy; ultraviolet detection; evaporative
light scattering detection; mass spectrometry; photoelectron emission; residue; method validation.
1. INTRODUCTION
∗
Phone: (314) 290-4725, Fax: (314) 725-5687, E-mail: herb_kaiser@steris.com
76 H.J. Kaiser
The analytical chemist must determine what is being measured. Is the residue a
drug active, formulation component, or the cleaning agent? Is the residue organic
or inorganic? Is the residue water soluble or water insoluble? Is the residue par-
ticulate, microbial or is it an endotoxin? The answers to these questions will nar-
row the choice of analytical techniques available. For example, if the residue is a
drug active, HPLC or gas chromatography (GC) could be utilized. If the residue is
a formulation component or a cleaning agent, TOC determination may be the
method of choice.
An important piece of information that the analytical chemist will require is the
target residue limit. The target residue limit should always be established prior to
the selection of the analytical technique. Once the residue limit has been estab-
lished the analytical technique can be chosen and a method developed that can de-
termine residue levels below the acceptance limit [2, 3].
Once the type of residue has been established the analytical chemist can chose be-
tween either specific or non-specific techniques. A specific technique detects a
unique compound. A non-specific technique detects any compound that produces
a certain response. Examples of specific techniques include HPLC, Fourier trans-
form infrared spectroscopy (FTIR), ion chromatography (IC), atomic absorption
(AA), inductively coupled plasma atomic emission (ICP), capillary electrophore-
sis (CE) and various protein methods. Examples of non-specific techniques in-
clude TOC determination, pH determination, acid/base titrations and conductivity.
technique where, after the manufacture of a normal drug product, the product, less
the drug active, is manufactured again. A sample is then taken of this material and
analyzed for the drug active in the previous batch. This is a questionable sampling
technique because it assumes that the residue is distributed uniformly throughout
the placebo. Another problem with placebo sampling is that the residue may be so
dilute in the placebo that the analytical technique may not be able to detect it.
This is an acceptable technique only if it is used along with swab or rinse water
data. Therefore, since swabbing or rinse water analysis must be done in conjunc-
tion with placebo sampling, this technique is rarely utilized. Placebo sampling is
important when validating the cleaning of filling equipment or small pieces of
equipment where direct rinse water measurements are not practical.
5.1. Interferences
If the detection technique is non-specific a good strategy is to first identify possi-
ble sources of interference. These interferences could be either positive or nega-
tive. For example, in TOC determinations, if the person performing the sample
collection accidentally coughs onto the surface being analyzed this would cause a
positive interference. Since TOC measures a non-specific property, the residue
amount would be calculated as if all of the measured property were due to that
residue. For example, if the target residue was aspirin and the technique utilized
was TOC determination, the TOC measured would be calculated as if all of the
TOC came from the aspirin even though there may be excipients and/or deter-
gents present. This provides an upper limit value and is a worst-case situation.
The identification of possible interferences is important to both specific and
non-specific methods. In the case of specific methods there should not be any
interferences if the method was properly developed. It is important that the
method is able to detect the analyte after exposure to the cleaning environment.
Studies should be conducted to demonstrate that the analyte does not change after
exposure to alkaline or acidic cleaners. This can be simply done by exposing the
drug active to a dilution of the cleaning agent at the temperature that will be
utilized during the cleaning process. This sample can then be analyzed to
determine if the analyte can still be detected and quantitated. If this is not done
the analyte may not be detected even though it is present in the sample. If the
cleaning environment does change the nature of the analyte a new method would
be required or modifications would have to be made to the existing method. When
utilizing a non-specific method to measure a non-specific property, any
compound that displays that property and is introduced into the sample will
interfere. Possible sources of interferences could be from the environment,
sampling technique, and/or people.
78 H.J. Kaiser
advantage is that the ELSD requires special care and cleaning of the nebulizer.
The nebulizer must produce consistent particle sizes so that an accurate detection
can occur. Therefore, careful cleaning and maintenance must be performed.
Figure 2. Chromatogram showing sodium and potassium ions in a cleaning agent using conductiv-
ity detection.
TOC analyses. There are also minimal chemical interferences. TOC determination
is easy to carry out and involves a minimal amount of method development.
Typically, samples are simply diluted and analyzed. No separation procedures
need to be developed unlike HPLC. As with other instrumentations, there are
automated sampling accessories that can be used. These accessories include on-
line monitors. On-line monitoring should only be used for monitoring water qual-
ity. On-line monitoring should not be used for monitoring a cleaning validation
study. The instrumentation for TOC determination is also very cost effective.
There are two major disadvantages to TOC determination. The first is that the
sample must be water-soluble. TOC determination cannot be used as the analyti-
cal technique if the residue cannot be dissolved in water. It should be noted that
solubility in the TOC determination sense means in the part per billion range
rather than in the part per million range. If a compound is reported to be water in-
soluble it does not automatically mean that it is not soluble enough to be analyzed
by TOC determination. The second disadvantage is that samples cannot be pre-
pared in organic solvents. If the samples were prepared in organic solvents the
TOC of the solvents and not of the residue would be determined.
TOC determination is an excellent complimentary technique to HPLC. TOC
can be used to monitor water-soluble ingredients in a formulation and HPLC can
be used to monitor the water insoluble ingredients. An example of this is a formu-
lated ibuprofen. The ibuprofen itself is not soluble even at TOC determination
levels, but HPLC can be utilized to monitor the ibuprofen. The formulation com-
ponents are soluble in water and TOC determination can be used to monitor their
contribution to the residue.
82 H.J. Kaiser
6. METHOD VALIDATION
It is time now to validate the methods once the residues have been identified, lim-
its set and method(s) chosen. Worldwide regulatory agencies require that these
methods be validated [6]. In order to be validated the method(s) must be shown to
have linearity, precision, accuracy and robustness. Ideally, there would be at least
two different analysts performing analyses on two different instruments using two
Methods for pharmaceutical cleaning validations 83
Figure 3. Example of one method for swabbing a surface to recover residues for subsequent analysis.
84 H.J. Kaiser
Applying a known amount of the test residue directly onto a swab can be used to
perform swab recoveries. The swab is then extracted and the extract analyzed us-
ing the appropriate method. The extraction solvent spiked with the same amount
should be used as the control. If the swab is retaining the residue, an alternate
swab material should be sought. The recovery factor obtained should either be in-
cluded in the analytical calculations or the acceptance limit calculations but not in
both. If the recovery factor is included in both, too tight a limit will have been set.
7. SUMMARY
REFERENCES
1. “Points to Consider for Cleaning Validation”, Technical Report No. 29, Parenteral Drug Asso-
ciation (1998).
2. H.J. Kaiser and M. Minowitz, J. Validation Technol., 7(3), 226-236 (2001).
3. H.J. Kaiser, J.F. Tirey and D.A. LeBlanc, J. Validation Technol., 6(1), 424-436 (1999).
4. M.K. Chawla, Precision Cleaning, 8(6), 36, 38 (2000).
5. M. Meltzer, C. Koester and C. Steffani, “Criteria Evaluation for Cleanliness Phase 0”, Lawrence
Livermore National Laboratory, UCRL-CR-133199 (1999).
6. USP 23, United States Pharmacopeial Convention, Rockville, MD, 1982-1984 (1995).
Surface Contamination and Cleaning, Vol. 1, pp. 85–107
Ed. K.L. Mittal
© VSP 2003
Abstract—Both wet and dry cleaning processes impact the chemical composition of glass surfaces.
The alteration of glass surface chemistry is probed by wettability measurements as a function of pH.
Wettability measurements are performed using the two-liquid method. This method consists of de-
positing sessile water drops on a substrate immersed in liquid octane. The measured contact angle
allows an estimation of the number of hydroxyl groups exposed at the glass/water interface. Meas-
urements with varying pH of the water drops probe the surface charge at the glass surface and indi-
cate the isoelectric points of predominant functional groups. Variations in glass surface composition
correlate with sensitivity to organic contaminant adsorption. Three typical species of glass were
probed: a silica surface, the surface of sodalime glass, and the surface of a Corning aluminoborosili-
cate glass. The three glass species show different contamination behavior, depending on wet or dry
cleaning of the glass. The surface compositions, as probed by wettability measurements, correlate to
grazing incidence XPS data, indicating the alteration of surface chemical sites by leaching during
the wet cleaning process. It appears that sensitivity to organic contamination is controlled by the
surface chemistry and the hydration of the exposed substrate.
1. INTRODUCTION
The cleaning of glass surfaces [1] precedes a variety of coating processes, ranging
from silane-based sol-gel coatings, to adhesion promoting films, to conductive
coatings such as ITO (indium tin oxide). These cleaning processes are generally
designed to remove particles from the glass surface and to render it uniformly
wetting. For most applications, it is desirable that this wettability be associated
with the exposure of silanol groups at the glass surface. These exposed chemical
functions form part of the glass substrate and they may be used to covalently graft
silane-based coatings to the glass substrate [2].
∗
To whom all correspondence should be addressed. Phone: +33-1-64 69 74 14,
Fax: +33-1-64 69 74 54, E-mail: birchw@corning.com
86 W. Birch et al.
2. EXPERIMENTAL
Table 1.
Composition of glass substrates. Principal components of the glass composition are given in weight
percentage. For the silica substrate, this is the composition of the native oxide layer at the surface.
The bulk glass composition is given for the other two glasses. The first column recalls the process
used to make the glass
The water used in these experiments was purified using an Elga UHP unit. It
filters deionized (18 MΩ.cm resistivity) water through an activated carbon filter
and a UV lamp. This water was used to rinse samples, as well as in the prepara-
tion of acidic or basic solutions to measure surface wettability.
The dry cleaning process used was pyrolysis (PYR). The glass substrates were
heated to 500°C, according to the following cycle: a rise in temperature from
room temperature to 500°C over four hours, followed by five hours at 500°C and
a return to room temperature over five hours. The samples were placed in a glass
rack and the rack was enclosed in a glass container, closed with a folded alumi-
num foil sheet. The containers were removed from the oven after cooling to room
temperature. Provided the lid was not opened, the glass samples remained clean
for about one week. The samples were used within 5 minutes after removing the
aluminum foil lid.
The glass samples showed no visible organic contamination on their surfaces
before cleaning by CHR or PYR. Neither technique is suitable for cleaning a glass
surface bearing macroscopic contamination. At most, the samples had ambient
contamination, in the form of a monolayer or sub-monolayer coverage of organic
contaminants, expected to have a thickness of order 1 nm or less.
Following cleaning, one glass sample per cleaning batch was verified as being
wettable by water. To achieve this, a one-microliter drop of pure water was placed
on the sample. This drop was expected to spread. When the sample was tilted, its
receding edge was not expected to dewet the glass surface. The glass substrates
were used if the sample passed these criteria of glass surface wettability. The
sample tested was not used for further experiments.
Influence of cleaning on the surface of model glasses 89
For surface composition analyses by XPS and monolayer adsorption and depo-
sition experiments, PYR was replaced by UV/ozone dry-cleaning (details are pre-
sented in Sections 2.3 and 2.4). Since the cleaning of silicon wafers with pyrolysis
led to rapid subsequent contamination on exposure to air and the XPS equipment
was far from the site where the samples were cleaned, UV/ozone cleaning was
used to improve sample cleanliness during transport. The two dry cleaning tech-
niques were expected to have a similar impact on the glass surfaces composition,
which is measured by XPS to a depth of about 5 nm.
Figure 1. Schematic diagram of a sessile water drop under liquid octane. The interfacial tensions
used in equation (1) are indicated and the definition of the contact angle, θ, is illustrated.
90 W. Birch et al.
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:
where θ is the contact angle of sessile water drop under octane at the point of zero
charge. The above equations relating wettability measurements to the non-
dispersive interaction between the glass surface and water are valid if organic
contamination is absent from the glass surface.
The wettability measurements of water drops under liquid octane were per-
formed in a cube-shaped glass cell (from Hellma, in Müllheim, Germany). This
glass cell was first cleaned with a 2% solution of Hellmanex detergent (Hellma)
in pure water for 20 minutes. It was then rinsed with pure water and blow-dried
under a flow of pure nitrogen. During rinsing, the cell was verified to be uni-
formly wetting. Immediately after cleaning, octane liquid was poured into the cell
to a depth of about two centimeters. The cleaned glass substrate was then im-
mersed in octane. Following a waiting period, as given below, three two-
microliter sessile water drops were deposited on the sample. The contact angles
on both sides of the drops were measured using a Ramé-Hart contact angle go-
niometer (Mountain Lakes, NJ, USA). These contact angle values were then aver-
aged.
The first measurements were made using unpurified octane. The immersion of
the substrate in octane was noted and the sessile water drop contact angles were
plotted as a function of the elapsed time between substrate immersion in octane
and deposition of the water drop. The second set of experiments was performed
using purified octane, cleaned by passing over silica and alumina chroma-
tographic columns. Octane (purissimum grade from Fluka, Ref. 74821) was
passed through a chromatography column filled with silica (silica gel, 60, Ref.
1.09385.1000, from Merck, Darmstadt, Germany), followed by passage through a
chromatography column filled with alumina (basic alumina, 60, activity stage I,
Ref. 1067.2000, from Merck). The silica and alumina were used as received. The
chromatography columns were cleaned with Hellmanex surfactant, as described
above for the glass cell. They were rinsed with pure water and dried in a clean
oven with aluminum foil covering their ends. The first few milliliters of octane
92 W. Birch et al.
passed through the columns were used to rinse the original octane bottle and then
discarded. The purified octane was then stored in this rinsed bottle. The reasoning
behind this purification of the liquid octane is that the high active surface area of
the silica and alumina particles will adsorb polar impurities in the octane. Given
the similarity of the surface sites on silica and alumina particles to those found on
clean glass surfaces, we expect the impurities adsorbed in the columns to be those
that would have adsorbed onto a clean glass surface.
To vary the pH of the water, aqueous solutions of hydrochloric acid or sodium
hydroxide were used. These solutions were made using calculated volumes and
verified with a pH sensitive paper. In the preliminary experiments, cleaned and
dried glass substrates were immersed into octane and the wettability of water
drops at different pH values was measured. These experiments showed trends in
the wettability as a function of pH. To reduce the error bars on the measured con-
tact angle values, the cleaned substrates were first immersed for twenty minutes in
an aqueous solution of the same pH as that of the water drops to be measured.
This was found to give the same trends in wettability as a function of pH, while
decreasing the error bar on the measured contact angle values. The increased ac-
curacy of the measurement may be due to the fact that the deposited water drop
has insufficient time for it to interact with the cleaned glass surface before settling
at the measured contact angle. Further, contact angle hysteresis may prevent the
contact angle value from changing if the surface charge of the cleaned glass
changes with its exposure the water drop.
The wettability of the cleaned glass substrates was measured after their immer-
sion in the same acidic or basic solution as used in the measuring water drops. Af-
ter cleaning, each glass substrate was cut into five pieces and each piece placed in
a solution of different pH. Following the 20-30 minute soaking time, the glass
substrates were rinsed with water and blow-dried with pure nitrogen. Sessile drop
contact angles were measured for water drops deposited five minutes after immer-
sion. The deposited water drops came from the same solution as the one in which
the substrates were immersed. For each glass species, three samples were meas-
ured at each pH and three sessile water drops were measured for each sample. The
data points used for the graphs in Figures 4 and 5 represent an average of these 18
contact angle values.
achieved the UV/ozone cleaning. The UV lamp emits radiation at the 184.9 and
253.7 nm wavelengths, where the first generates ozone from oxygen while the
second combines with ozone to oxidize hydrocarbons [10, 11].
Figure 2. Water sessile drop contact angles measured after exposure of cleaned glass substrates to
unpurified octane. The legend is the same as for Table 2. The quoted contact angles are averages of
six values for three sessile drops at times of 1, 3, and 5 minutes. The contact angle values quoted for
“silica PYR” at less than one minute are for one drop at each data point.
contact angle of the sessile water drops did not change after their deposition. The
observed increase in contact angle suggests an evolution of the cleaned surface
with time over the first five minutes after drop deposition. The increasing contact
angle values also suggest a contamination of the substrate. Other possibilities,
such as a hydration or dehydration of the cleaned surface were considered. Hydra-
tion of the cleaned surface is expected to lead to a decrease in the measured con-
tact angle value. To check for dehydration, a drop of water was placed in the oc-
tane and the cleaned substrates immersed after 48 hours. The contact angle
evolution was identical to that in Figure 2, suggesting that dehydration of the
glass surface in octane is not what causes the observed increase in contact angle
with immersion time. To verify contamination by exposure to octane, the sub-
strates were blow-dried with pure nitrogen after immersion for at least five min-
utes in the liquid octane. The wettability of sessile water drops was then measured
in air. The measured contact angles are given in Table 2. They range from 32 to
46 degrees, indicating a significant level of organic contamination adsorbed on
the glass surfaces. Further, the large error bar of about 8 degrees indicates a non-
uniform surface wettability. This behavior is typically found for random contami-
nation deposited on a clean surface. The contact angle values measured after ex-
posure to octane, when compared with the complete spreading of water drops on
the freshly cleaned glass substrates, suggest that the liquid octane contaminates
the cleaned glass substrates.
The sessile drops deposited following longer immersion time of the cleaned
substrate in octane showed a higher contact angle than those on a freshly im-
mersed substrate. In fact, a water drop deposited immediately after immersion of
the cleaned glass into octane spread to form a low contact angle, generally of 5°
96 W. Birch et al.
Table 2.
Water sessile drop contact angles (in degrees) measured on glass substrates exposed to unpurified
octane for more than five minutes. The substrates were removed from the octane and blow-dried
with pure nitrogen before wettability measurements. The error bar over several sessile drops and
several substrates is nominally 8°. The superscript number indicates the number of substrates over
which the quoted wettability was measured. Each substrate was measured using between three and
eight sessile drops. In the legend, silica refers to the native oxide layer on a silicon wafer, ABS re-
fers to Corning code 1737 aluminoborosilicate glass, and SDL refers to sodalime glass. CHR and
PYR refer to chromic acid and pyrolysis cleaning, respectively
PYR 46 5 35 7
32 2
or less. The drops, once deposited, did not change their shape, indicating that the
contact angles of the sessile drops did not evolve. Young’ s equation invokes the
notion of a force equilibrium at the three-phase line, where the two fluid phases,
namely octane and water, meet the solid surface. The absence of any increase in
the contact angle of the deposited drops, when compared to the increase in contact
angle for freshly deposited drops, suggests that an equilibrium contact angle value
is not reached for the sessile water drops under octane. The notion of a force bal-
ance at the three-phase line is valid. A force balance can be considered as a mini-
mum in the energy of a system with respect to a small displacement of the system
from this minimum. This concept is valid for systems that are free to move, where
the influence of frictional forces is negligible. For contact angles, the concept of a
contact angle hysteresis is analogous to a frictional force, preventing the contact
angle from reaching a value corresponding to the overall minimum in the free en-
ergy of the system. This is the case for water drops deposited under octane, where
the observed contact angle is not reached by the drop retracting to form a contact
angle equal to that of sessile drops deposited at later times.
Following a five minute exposure to octane, the contact angles of three or more
freshly deposited sessile water drops were measured on each of the glass sub-
strates. At least three substrates were investigated for each glass species and
cleaning procedure. Table 3 gives the averaged contact angle values for each type
of glass using CHR or PYR cleaning. For the silica surface, we find a slightly
higher contact angle following PYR cleaning than for the CHR cleaned substrate.
This is compatible with a loss of about 10% of the surface hydroxyl groups during
glass pyrolysis [14]. However, for the sodalime glass, the wettability of the sur-
face following PYR cleaning remains high, even when immersed in the liquid oc-
tane that has contaminated the other glass substrates. To obtain more information
on this behavior, the advancing and receding contact angles were measured for
the substrates exposed to liquid octane for more than five minutes. The measured
values are shown in Table 4. If these contact angle values are interpreted in terms
of surface contamination of the cleaned glass substrates by immersion in the liq-
Influence of cleaning on the surface of model glasses 97
Table 3.
Water sessile drop contact angles (in degrees) measured on cleaned glass substrates under octane,
following a five minute exposure to the octane. The error bar over several sessile drops and several
substrates is nominally 4°. The legend is the same as for Table 2
Table 4.
Advancing and receding contact angles (in degrees) of water drops under unpurified octane. The
substrates were immersed in the unpurified octane for at least five minutes before measuring the
contact angle hysteresis of one water drop per substrate. The superscript number indicates the num-
ber of substrates over which the quoted wettability was measured. The legend is the same as for Ta-
ble 2
uid octane, they suggest that the pyrolysed silica is more susceptible to adsorbing
contamination than the chromic acid cleaned silica. Further, the pyrolysed so-
dalime glass does not strongly adsorb organic contaminants from the octane,
given that the water drop spreads easily. The pyrolysed aluminoborosilicate glass
shows a behavior intermediate between the silica surface and the sodalime glass,
showing a finite advancing contact angle and zero receding contact angle. The ab-
sence of a finite receding contact angle value suggests that few or no organic con-
taminants remain adsorbed on the glass surface after passage of the water drop.
On the other hand, following chromic acid cleaning, the three glass species show
similar wettability, suggesting that they have almost the same affinities for ad-
sorbing organic contaminants in liquid octane. This suggests that the sensitivity to
adsorbing contaminants depends on the composition of the glass surface and the
procedure used to clean the glass surface. Considering that PYR cleaned silica
surface is more dehydrated than CHR cleaned silica surface, we may attribute the
reduction in the affinity for adsorbing organic contaminants to the presence of a
layer of water on the glass. PYR cleaned sodalime glass shows the lowest level of
adsorbed contaminants. Both the soluble alkaline oxide film and the ambient
moisture it adsorbs may enhance the durability of this clean glass surface. It is in-
teresting to note that the surface behavior of the glass species is similar to that of
the silica following cleaning with chromic acid.
98 W. Birch et al.
Figure 3. Schematic side-view of a drop on an inclined surface, indicating the advancing contact
angle, θa, and the receding contact angle, θr.
The kinetics of water drop spreading described for water drops on glass under
purified octane remains unexplained. The increase in contact angle observed for
glass surfaces cleaned with chromic acid is not attributable to contamination,
given the high wettability of the cleaned glass following exposure to the octane.
The unusual spreading behavior of water drops on acid cleaned Corning code
1737F glass also remains unexplained. Finally, the complete wetting of pyrolyzed
glass substrates by water drops under octane suggests a high affinity of these glass
surfaces for water. Further, the presence of some components of the glass compo-
sition that are soluble in water, such as sodium oxide and calcium oxide (see Ta-
ble 1), appears to enhance the spreading of the water drops under octane.
3.3. Surface wettability as a function of pH, surface charge and SiOH group
density
While the contact angle values for sessile water drops at neutral pH provide a first
information on the surface energy of the cleaned glass substrates, further informa-
tion can be provided by measuring the surface wettability as a function of the pH.
These experiments were conducted using purified octane to avoid the substrate
contamination described in Section 3.1. The absence of contamination of the
cleaned glass substrates and the behavior under octane of sessile water drops at
neutral pH is described in Section 3.2.
The pH of water drops deposited on the cleaned glass surfaces was varied in
the range from 0 to 13, in order to probe the surface charge variations from acidic
and basic surface groups. The surface wettability was measured at pH intervals of
1 unit for acidic pH and two units for alkaline pH values. The standard deviation
of the measurements was generally 1° to 2°. About 10% of the data points showed
standard deviations ranging from 3° to 5°.
Figure 4 shows the non-dispersive interaction energy of glass substrates
cleaned with chromic acid as a function of pH, in the range from 0 to 13. The in-
terval of one pH unit for acidic solutions and two units for alkaline solutions was
chosen arbitrarily. The contact angle values, ranging from 0 to 21°, have been
converted into the interaction energy, INDglass/water, using equation (8). Figure 5
shows data from the same measurements on glass surfaces cleaned by pyrolysis.
The two graphs in Figure 5 show the same data. The upper graph was plotted to
show differences in the behavior of Corning code 1737 glass and sodalime glass,
while the lower graph shows data for all three glasses, with silica showing a sig-
nificantly lower non-dispersive interaction energy with water than the other two
glass species.
The hydroxyl (SiOH) surface groups on the clean glass may be positively or
negatively charged, depending on the solution pH in contact with the surface. The
isoelectric point, also known as the point of zero charge, or p.z.c., defines the pH
at which the surface densities of positive and negative charges are equal, associ-
ated with an equal surface density of negatively charged SiOH groups (SiO–) and
positively charged SiOH (SiOH2+) [9]. The isoelectric point of silica has been
100 W. Birch et al.
Figure 4. Non-dispersive interaction energy between sessile water drops and Chromerge cleaned
glass substrates as a function of pH. The water drops were at the pH indicated and each point repre-
sents an average of 18 contact angle measurements. The standard deviation of the data points is
typically 1-2°, with about 10% of the data points giving a standard deviation of 3-5°.
Figure 5. Non-dispersive interaction energy between sessile water drops and pyrolysis cleaned glass
substrates as a function of pH. The upper graph shows the data from sodalime and Corning code
1737 glass, comparable to that from Figure 4. The lower graph shows the full data for the silica sur-
face, which shows an extremely different behavior, resulting in far lower interaction energy and
higher measured contact angles. Data points are averaged over 18 values with standard deviations as
described in the legend for Figure 4.
Table 5.
Water sessile drop contact angles (in degrees) measured on cleaned glass substrates under octane.
The measured water drops are aqueous hydrochloric acid at pH 3, corresponding to the point of zero
charge of the data in Figures 4 and 5. The legend is the same as for Table 2
Table 6.
Hydroxyl group density (-OH/nm2) for glass surfaces, calculated from the contact angles given in
Table 5. These contact angles were measured on the cleaned glass surfaces, for sessile water drops
under octane at the point of zero charge. The legend is the same as for Table 2
In terms of estimated hydroxyl group density on the cleaned glass surfaces, py-
rolyzed silica shows a lower hydroxyl group density than chromic acid cleaned
silica. This is compatible with a minor amount of dehydroxylation during the py-
rolysis cycle [14]. Unexpectedly, pyrolyzed sodalime glass appears to have a
higher surface SiOH concentration than silica. This may be an artifact caused by
the soluble alkali salts (see Table 1) present on the surface of the pyrolyzed so-
dalime glass. These salts may increase the wettability to water when compared to
the same glass cleaned by chromic acid, where alkaline oxides are leached during
the cleaning procedure.
Returning to the graphs in Figures 4 and 5, the data show two minima in the
non-dispersive interaction energy between glass and water. The first and more
prominent minimum, found at pH 3, may be associated with the isoelectric point
of the SiOH groups. Recalling the acid-base reaction responsible for generating
the surface charge from the SiOH groups:
SiO – +H 3O + ↔ SiOH + H 2O,
the hydrogen ion (H3O+) concentration at this pH is sufficient to drive the equa-
tion towards the right. At higher pH values, corresponding to a lower hydrogen
ion concentration, the glass surface is negatively charged from its surface SiOH
groups. At lower pH values, the surface hydroxyl groups become positively
charged, according to the following equation:
SiOH+H 3O + ↔ SiOH +2 + H 2O.
Thus, only at the isoelectric point does the glass surface bear no overall charge
from its SiOH groups. A second, shallower, minimum in the interaction energy
may be seen at pH 9. This may correspond to the sodium ions (Na+), from the so-
dium hydroxide used to increase the pH, being present in sufficient quantity to
neutralize the SiO– ions from the surface hydroxyl groups.
For all the glass samples, the measured contact angles fall sharply above pH 10
and are zero above pH 12. This is probably due to the glass surface being de-
graded by dissolving in the alkaline solution, leading to its being fully wettable by
water.
Influence of cleaning on the surface of model glasses 103
Table 7.
Surface composition of cleaned glass measured by XPS. The penetration depth is estimated at 5
nm. The percentage concentrations of selected elements are given, resulting in a total of less than
100%. The carbon signal ranged from 4-7%, indicating a layer of ambient contamination adsorbed
during transfer of the samples from cleaning to the measurement location. For the silicon wafer,
only the signal attributed to the native oxide layer is given. Since the native oxide layer is 1-2 nm
thick and the beam penetrates 5 nm into the sample, the total of the percentage compositions is
much less than 100
The presence of surfactant monolayers coating the glass surfaces was con-
firmed by measuring the wettability of non-polar tricresyl phosphate sessile drops
in air. The contact angles on the SDS monolayers were in the range 61-66 de-
grees; those on CTAB monolayers were in the range 40-41 degrees. This wettabil-
ity is compatible with a lower packing density for the CTAB surfactant
monolayer, where a larger head group prevents close packing of the aliphatic hy-
drocarbon chains [3, 4].
Further information on the surface silanol sites can be obtained by depositing a
self-assembled OTES monolayer. This monolayer was deposited on glass sub-
strates cleaned by chromic acid or UV/ozone. A monolayer of grafted octadecyl-
silane molecules generates high contact angles for water, indicating a hydropho-
bic surface of densely-packed aliphatic hydrocarbon chains. Contact angle
hysteresis probes the defect density in the self-assembled monolayers. The reced-
ing contact angle is primarily influenced by the presence of hydrophilic defects in
the hydrophobic monolayer. Measured wetting properties for water drops are
given in Table 8.
For the UV/ozone cleaned surfaces, the data show a higher quality hydrophobic
monolayer on the silica surface. A densely packed octadecylsilane monolayer
gives rise to high contact angles for water. This dense packing is a result of a high
degree of ordering of the octadecylsilane molecules and their grafting to each
other and to the glass substrate. The aluminoborosilicate glass shows a coating
quality intermediate between that on the silica and sodalime glass surface, which
shows the lowest coating quality. The chromic acid cleaned glass surfaces show
similar coating quality, slightly higher than that seen for the UV/ozone cleaned
surfaces. This improved coating quality may be, in part, due to hydration of the
chromic acid cleaned glass. The trends in coating quality for all UV/ozone
cleaned glass surfaces and all chromic acid cleaned glass surfaces correlate with
those seen for contamination of the glass surfaces, as described in Section 3.1.
Table 8.
Wettability towards water of self-assembled monolayers of OTES deposited on the cleaned glass
surfaces. The sessile drop, advancing, and receding contact angles (in degrees) are given, with the
sessile drop value above the advancing and receding values. The wettability measurements were
made in air. For the sodalime glass (SDL), the slides used were float glass. The wettability of the
deposited monolayer coatings was measured separately on the air side and the on the float side of
this glass
4. CONCLUSION
While cleaned silica-based glass surfaces have similar surface compositions, their
susceptibility to strongly adsorbing organic contaminants 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 following 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
1. W.R. Birch, in: Sol Gel Handbook, M. Aegerter (Ed.), Kluwer Academic Publishers, New York
(to be published) and references therein.
2. K.L. Mittal (Ed.), Silanes and Other Coupling Agents, Vol. 2, VSP, Utrecht (2000).
3. W.R. Birch, S. Garoff, M. Knewtson, R.M. Suter and S. Satija, Colloids Surfaces 89, 145
(1994).
4. W.R. Birch, S. Garoff, M. Knewtson, R.M. Suter and S. Satija, Langmuir 11, 48 (1995).
Influence of cleaning on the surface of model glasses 107
5. R.J. Good and M.K. Chaudhury, in: Fundamentals of Adhesion, L.-H. Lee (Ed.), pp. 137-151.
Plenum Press, New York (1991).
6. J.J. Jasper, J. Phys. Chem. Ref. Data 1, 841 (1972).
7. F.M. Fowkes, Ind. Eng. Chem. 56 (12), 40 (1964).
8. J.N. Murrell and A.D. Jenkins, Properties of Liquids and Solutions, p. 168, John Wiley and
Sons, Chichester (1994).
9. A. Carré, F. Roger and C. Varinot, J. Colloid Interface Sci. 154, 174 (1992).
10. J.R. Vig, J. Vac. Sci. Technol. A 3, 1027 (1985).
11. J.R. Vig, in: Treatise on Clean Surface Technology, Vol. 1, K.L. Mittal (Ed.), pp. 1-26, Plenum
Press, New York (1987).
12. J. Davidovits, PhD Thesis, Détermination des conditions d’obtention de films monomoléculaires
organisés : application aux silanes auto-assemblés sur silice, Université Paris 6, Chapter 5
(1998).
13. J. Peanasky, H.M. Schneider and S. Granick, Langmuir 11, 953 (1995).
14. R.K. Iler, The Chemistry of Silica, John Wiley & Sons, New York (1979).
15. A. Carré, S. Moll, J. Schultz and M.E.R. Shanahan, in: Adhesion 11, K.W. Allen (Ed.), pp. 86-
87, Elsevier Applied Science, New York (1987).
This page intentionally left blank
Surface Contamination and Cleaning, Vol. 1, pp. 109–127
Ed. K.L. Mittal
© VSP 2003
Abstract—Most of the electronic and electro-optic equipment fielded by the military is incompati-
ble with the standard aqueous based decontamination solutions, such as 5% sodium hypochlorite so-
lution, ESI has developed a nondestructive decontamination process for such sensitive equipment.
In this process, the components to be decontaminated are immersed in an ultrasonic bath filled with
an organic solvent, if they are contaminated with chemical warfare agents (CWA), or in a solution
of a surfactant in this solvent, followed by a pure solvent rinse, if they are contaminated with bio-
logical agents or radioactive particles.
In both cases, the contaminants are dissolved or suspended in the decontamination liquid in the
bath. The contaminants are removed from the decontamination liquid by circulating it through a fil-
tration train. In the case of CWA, the filtration train consists of an activated carbon filter, a particu-
late pre-filter, and a membrane filter. In the case of biological or nuclear contaminants, the circulat-
ing liquid bypasses the activated carbon beds.
A prototype decontamination system has been built and operated to demonstrate the process. In
this program, a wide range of sensitive equipment was contaminated with a CWA simulant. The
contaminated equipment was immersed and sonicated in a flowing solvent, which recirculated
around a purification loop, until the simulant could no longer be detected, and dried. The decon-
taminated equipment was then functionally tested. In all cases:
a. no traces of simulant were found on the processed pieces, and
b. the processed items were fully functional.
1. INTRODUCTION
∗
To whom all correspondence should be addressed. Phone: 781-938-7588, x 22, Fax: 781-938-
7589, E-mail: rkaiser@entropicsystems.com
110 R. Kaiser and K. Haraldsen
eliminated from all DOD activities by the National Defense Authorization Act for
Fiscal Year 1993.
Thus, alternate methods and equipment for nondestructively decontaminating
water sensitive military equipment, such as avionics, electronic, electrical and en-
vironmental systems equipment are needed. These methods need to:
a. Be effective against a wide variety of threat agents,
b. Be nontoxic to personnel,
c. Not degrade the equipment being decontaminated, and
d. Be field deployable.
Decontamination system equipment should be highly mobile and self-
sustaining. These methods should also be able to treat equipment that is be-
smirched with battlefield soils, including dirt (particulates), dried mud, oil, etc. In
a broader context, the methods and equipment should also be capable of perform-
ing maintenance cleaning operations in a depot environment. These methods and
equipment should also comply with environmental regulations.
In terms of performance, an effective decontamination method has to be able to
remove or deactivate the contaminant without affecting the part being cleaned.
Because the type of equipment that would likely be decontaminated is both geo-
metrically complex in shape and thermally sensitive, the most effective technique
will likely use a flowing or agitated liquid, at ambient or modest temperatures, as
a means of removing the range of threat agents from the equipment.
Numerous alternate potential decontamination options have been examined, but
have been determined to be of limited effectiveness [1]. Heating an article above a
modest temperature may not be an option for decontaminating thermally sensitive
items, which leads to problems in terms of effectively removing relatively non-
volatile contaminants. Decontamination by particle blasting methods, such as car-
bon dioxide snow or plastic pellets, are limited to surfaces that are in direct line of
sight with the ejection nozzle. Such methods are not effective in terms of cleaning
blind holes, crevices, and obstructed surfaces. These types of methods can be
abrasive and destructive to the equipment being decontaminated. Capture and
processing of nonvolatile contaminant-laden particles may be a problem as well.
2. APPROACH
Table 1.
Properties of fluorinated solvents of interest
VOC, kg/kg 0 0 0 1
Ozone Depletion Potential
(CFC-11 = 1) 0 0 0 0.8
Global Warming Potential 1700 320 55 5000
[100 yr ITH (1)]
Atmospheric Lifetime, yrs 17.1 4.1 0.8 110
Flashpoint, °C None None None None
Flammability Range in Air, % None None 2.4-12.4% None
Exposure Guidelines, 8 hr TWA, 200 750 200 1000
ppm
(1) ITH: Integration time horizon
Data Compiled from Published Information
nants remaining on the treated parts by visually examining the parts for residual
fluorescent marker before the parts are removed from the cleaning chamber.
The principal objectives of ESI’s development program were to:
a. Identify and/or develop safe and environmentally compatible cleaning agents
and processes that will effectively remove biological and chemical threat
agents from sensitive equipment that may be contaminated with other soils,
without damaging this equipment.
b. Identify and/or develop means by which the threat agents suspended in the
process liquids can be subsequently deactivated, and/or safely removed from
these liquids.
c. Identify and/or develop methods of monitoring the presence of threat agents
and/or other contaminants in the recovered process liquids.
d. Demonstrate the technology developed in the above objectives on a pilot
plant scale.
These objectives were all met, and the major findings of this program are dis-
cussed next.
3. MAJOR FINDINGS
3.1. Identify and/or develop safe and environmentally compatible cleaning agents
Ultrasonic solvent cleaning processes can effectively decontaminate sensitive
equipment. Methoxyperfluorobutane (3M’s HFE-7100) is the decontamination
liquid of choice because it meets the following criteria:
1. It is compatible with a wide range of sensitive equipment – the performance
of electronic and optical equipment is not affected by immersion in HFE-
7100. In particular, it does not attack components made of poly (methyl
methacrylate) or polycarbonate, as does DuPont’s Vertrel-XF.
2. The principal chemical warfare agents (CWAs) of concern are sufficiently
soluble in HFE-7100, as indicated in Table 2.
3. HFE-7100 is effective in ultrasonic cleaning baths because it has very low
surface tension, which allows it to penetrate small features of the surface, and
because it has a low heat of evaporation, which allows the ultrasonic agitation
to produce strong shear forces in order to disrupt the boundary layer and en-
train the contaminants.
4. The principal CWAs of concern are quantitatively removed from solution in
HFE-7100 by activated carbon.
5. When agent-contaminated HFE-7100 is passed through a bed of activated
carbon, the agent adsorbs onto the activated carbon, resulting in agent-free
HFE-7100 that can be recycled and reused. In comparison, there was poor ad-
sorption of chemical agents from Vertrel MCA+ in which mustard agent (HD)
114 R. Kaiser and K. Haraldsen
exhibited a higher solubility. Vertrel MCA+ thus was not considered for two
reasons – it was too aggressive a solvent to meet material compatibility re-
quirements, and could not be reclaimed for recycling by adsorption of chemi-
cal agents.
6. It is nonflammable, nontoxic, and environmentally acceptable.
Table 2.
Solubility of chemical agents in solvents of interest [2]
GB GD HD VX
Vertrel MCA+ M (RT) M (RT) 17% (RT) M (RT)
Vertrel XP-10 M (RT) M (RT) 8% (40°C) M (RT)
Vertrel-XF M (RT) M (RT) 8% (40°C) M (RT)
HFE-7100 M (RT) M (RT) 8% (40°C) M (RT)
HFE-7200 M (RT) M (RT) 8% (40°C) M (RT)
CHP M (RT) M (RT) M (RT) M (RT)
M (RT) = Miscible at room temperature
Figure 2. Circuit boards in culture medium. Circuit boards with environmental contamination were
processed in the Cadet and then cultured in trypticase soy broth to assess effectiveness of removal of
microorganisms. The circuit board in jar A is an unprocessed control. After incubation, the culture
broth is turbid, indicating multiplication of microbial contaminants (bacteria, fungi) present on the
circuit board as a result of exposure to the environment. The circuit board in jar B was decontami-
nated using the Cadet as described in the text. The lack of turbidity in the culture indicates that the
circuit board was rendered sterile as a result of processing.
3.2. Identify and/or develop means of removing the threat agents from process
liquids
The removal of CWAs and of CWA simulants dissolved in hydrofluorocarbons
by adsorption on activated carbon was demonstrated in small-scale batch tests that
were performed by ESI for CWA simulants, and under subcontract by a surety
laboratory (Battelle Memorial Institute) for CWA.
116 R. Kaiser and K. Haraldsen
Pilot scale continuous flow adsorption tests were performed in order to estab-
lish the effect of process parameters on the rate and extent of removal of CWA
simulants from a hydrofluorocarbon solution as it flowed through an activated
carbon column, and on the adsorption capacity of the activated carbon in the col-
umn. Providing sufficient residence time is critical to obtaining a reasonable col-
umn loading before breakthrough of the contaminant occurs. A CWA simulant
loading of 3 wt-% on granular activated carbon (Norit’s 1240 GAC) was achieved
with a residence time of five minutes. The length to diameter ratio of the columns
should be larger than 3 so as to minimize liquid bypassing. Activated carbon beds
and filters that come into contact with contaminated liquid can be contained in
commercially available housings that shield the system operator from any con-
tained toxic contents. These sealed containers, and their contents, can be de-
stroyed by standard methods, such as chemical deactivation or incineration.
3.3. Identify and/or develop methods of monitoring the presence of threat agents
Spectrographic fluorimetry is an extremely sensitive method of detecting fluores-
cent materials. With this method, the detection limit of fluorescent dyes dissolved
in HFE-7100 was found to be of the order of 10 parts per trillion (ppt). CWA
simulants which contained small amounts (0.05 wt-% to 5 wt-%) of a fluorescent
dye (Try-33 made by Day-Glo Corp., Dayton, OH) were used as contaminants in
the pilot decontamination studies described in the next section. The presence of
the fluorescent dye in the contaminant allowed the decontamination process to be
easily monitored, in terms of being able to both:
1. Measure and detect low levels of contaminant in the process liquid leaving
the ultrasonic bath and the activated carbon columns.
2. Detect traces of residual contaminant on the parts being processed by illumi-
nating these parts with an ultraviolet lamp.
5. The capture and removal of waste products (agent simulant, other soluble
contaminants, and particulates) from the contaminated solution in fully en-
closed, disposable activated carbon and filter cartridges, that are easy to in-
stall, remove, and would minimize operator exposure to the contents of the
cartridges.
6. The ability to immerse parts in both pure HFE-7100 and in fluorinated surfac-
tant or coupling agent solutions in HFE-7100.
Additional objectives were to obtain operating data on specific unit operations,
such as the effect of liquid circulation rate on the removal kinetics of simulant
from contaminated test parts, the effects of liquid flow rate and contaminant con-
centration on the efficiency, and capacity of activated carbon columns.
The breadboard decontamination system consisted of three modules:
1. A Poly-KleenTM Vapor Degreaser
2. An activated carbon column train
3. A circulating water chiller
The Poly-KleenTM Vapor Degreaser is a manually operated vapor degreasing
system designed to be used with low boiling fluorinated solvents, such as per-
fluorocarbons, hydrofluorocarbons and hydrofluoroethers. Figure 3 is a photo-
graph of the inside chambers of the system. Overall dimensions are 1956 mm H x
736 mm D x 904 mm H (77" L x 29" D x 36" H). The height includes the handles
on the sliding door which are 76 mm (3") high. The Poly-KleenTM Vapor De-
greaser takes full advantage of the ease of fabrication of polypropylene and its
compatibility with fluorinated liquids to allow a high performance system to be
manufactured at a lower cost than an equivalent stainless steel system.
The basic Poly-KleenTM Vapor Degreaser is a three-sump unit that requires 140
liters (38 gallons) of fluid to operate. The principal components of the system and
their effective dimensions are:
The activated carbon module is shown in Figure 4. The modules and the chiller
are interconnected as shown in the process flow diagram presented as Figure 5.
Referring to this process flow diagram, the in-line process components from BV-
6 to CV-3 were mounted on an angle iron frame, 1829 mm L x 736 mm W x 762
mm H (72" L x 29" W x 30" H).
Decontamination of sensitive equipment 119
Cleaning trials were performed with the following pieces of sensitive equipment
a. Auto-Ranging LCD Digital Multimeters, Model No. 22-179A, Radio Shack,
A Div. of Tandy Corp., Fort Worth, TX
b. Electronic Calculator, Model No. EC-441, Radio Shack, A Div. Of Tandy
Corp., Fort Worth, TX
c. Global Positioning System (GPS) receiver, Model No. GlobalNav 212, Serial
No.005263360, Lowrance Electronics, Inc., Tulsa, OK.
d. Night Vision Binoculars, Model RO 38, 4 x 48 Nighthawk, Serial No.
982331, with Model RO45, Zoom IR Illuminator, LAN Optics International,
Burlington, MA.
e. 7.65 mm semi-automatic pistol, Model PP, Carl Walther GmbH Sportswaf-
fen, Ansberg, Germany
f. Inverter Circuit Boards, 38 mm (1.5 in) square, designed by Entropic Sys-
tems, Inc.
Numerous tests were performed with digital multimeters, which were consid-
ered to be good prototypes for sensitive equipment. These items performed a
number of electrical functions, they had a liquid crystal display covered by a clear
plastic window, they contained a variety of materials that would be damaged by
120 R. Kaiser and K. Haraldsen
many solvents, and were inexpensive enough (about $12.00 each) to be consid-
ered disposable test items.
In addition, some tests were performed with other items to test the effects of
part geometry. These items included standard 25 mm x 75 mm microscope slides
(standard flat surfaces), brass pipe nipples (easily accessible interior surfaces),
and magnet assemblies (difficult to access interior surfaces). A magnet assembly
consists of a circular piece of stainless screen (typically 100 mesh) that is sand-
wiched between two cylindrical Alnico magnets of equal diameter. The magnets
were 13 mm (½") in diameter by 6.5 mm (¼" ) in height. The soil is deposited on
the screen before forming a magnet assembly. This sandwich was then subjected
to a cleaning trial. The changes in weight of the assembly and in the appearance
of the screen were measures of the effect of the cleaning trial.
The test pieces were contaminated with a variety of neat and thickened CWA
simulants and other soils. CWA simulants used in these tests were diethyl phtha-
late (DEP), tributyl citrate (TBC), and Krytox 157 (L) and (H) fluorosurfactants.
These materials are all water insoluble oils that have a low vapor pressure at am-
bient room temperature. They also all are miscible with HFE-7100. The CWA
simulants were all doped with a fluorescent dye, Try-33, that greatly facilitated
their detection on the test pieces and in the decontamination liquid.
In some of the tests, a thickener was added to the simulant to mimic the behav-
ior of thickened CWA agents. Two different types of thickeners were used: fumed
silica (Cabosil LM-130, Cabot Corp.), and an acrylic polymer (Paraloid K-125,
Rohm & Haas Corp.). Paraloid K-125 has been used to thicken military CWA.
The consistency of the simulant depends on the amount of thickener used. At 1-2
wt-% thickener loading, the simulants flow like honey, while they become semi-
solid gels at thickener loading greater than 5 wt-%. One key difference between
colloidal silica and an acrylic polymer is that colloidal silica is not soluble in any
organic solvent, but the acrylic polymer can dissolve in a polar organic solvent.
In addition to the above simulants, test pieces were also contaminated with rep-
resentative soils that could be found on fielded equipment: mineral oil, SAE 30
motor oil (NAPA) thickened with Arizona road dust (Duke Scientific Co, Palo
Alto, CA), multi-purpose lithium grease (Lubrimatic), and dried, 50 grit SiC wa-
ter based lapping compound (Clover).
The contaminant removal tests were performed in the Poly-KleenTM system ac-
cording to the following general procedure:
a. The equipment to be processed was weighed and photographed under visible
and UV light.
b. One or more tared pieces of equipment were coated with contaminant(s) or
soil(s), photographed under visible and UV light, and re-weighed.
c. The test piece(s) were placed into the transfer basket of the Poly-Kleenä sys-
tem, which was then covered with a tight fitting screen.
d. The immersion sump of the Poly-Kleenä system contained enough HFE-7100
to cover the part in the basket. Sonication for 30 minutes degassed this liquid.
Decontamination of sensitive equipment 121
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
of 1,700 ml/minute for a finite period of time. The circulation time ranged
from 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-Kleenä system. The test
pieces were removed from the basket, visually examined, photographed under
visible and UV light, reweighed, and archived.
In 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-Kleenä 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-cleaning weight
and visual appearance of these items.
122 R. Kaiser and K. Haraldsen
Table 3.
List of sensitive equipment decontamination experiments performed
Except for the runs where there was visible attack of the substrate by the simu-
lant (as in run 1 in Table 3), there was an increase of less than a 0.1 gram in the
weight of the object after contamination and cleaning and the original (i.e. before
contamination) weight of this object. In some cases, there was a weight loss of the
order of 0.1 gram (as in the calculator in run 6 and the pistol in run 9). This was
attributed to the removal of other soils that were previously present on these test
items.
If the ratio of (weight change/contaminant weight) is used as a cleaning crite-
rion, this value is less than 10%, except for run 1 (for the reasons cited above),
and for runs 13 to 17. For these last five runs, the relatively high values of this ra-
tio is attributable to the precision of the weight measurement. The weight meas-
urements were performed on a balance that had an accuracy of ± 0.02 gram,
which would account for most of the observed weight differences.
Table 4.
Sensitive equipment cleaning results
Photographs of various test objects were taken under both normal and UV il-
luminations,
a. Before the application of contaminant(s),
b. After the application of contaminant(s), and
c. After cleaning or decontamination.
The presence or absence of contaminant on the items is clearly evident by sim-
ple examination of the photographs taken under UV illumination of the night vi-
sion goggles presented in Figure 6 and the pistol in Figure 7.
Process kinetics was examined by monitoring the concentration of contaminant
in the liquid in the immersion bath as a function of time. Removing the contami-
nant from the surfaces of the object being cleaned and transferring it into the bulk
liquid took approximately 15 minutes. Once the circulation pump to the activated
carbon columns was turned on, approximately three bath turnovers were required
to purge the immersion sump of contaminant. In this test program, turnover time
was of the order of 12 to 15 minutes.
Acknowledgements
The work described in this paper was supported by the U.S. Air Force under the
auspices of SBIR PROJECT AF 97-014, “DECONTAMINATION OF AIR-
CRAFT ELECTRONIC EQUIPMENT”, Contract Number: F41624-98-C-5061.
It was facilitated by the insights and encouragement of Dr. Ngai Wong,
AFRL/HEST, the Air Force Technical Monitor.
Decontamination of sensitive equipment 127
REFERENCES
Note: Copies of Refs. 2, 3, 4, can be obtained from the sponsoring agencies or the author of this paper.
This page intentionally left blank
Surface Contamination and Cleaning, Vol. 1, pp. 129–136
Ed. K.L. Mittal
© VSP 2003
Abstract—This paper describes a new method for cleaning of parts. The method converts soils to
water-organic emulsions via pressure waves generated by ultrasonic transducers. Underwater photo-
graphs and symbolic representations show that the process consists of: (1) contact with ultrasonic-
generated pressure waves produces an emulsion of soil in water, and (2) rinsing the emulsion from
the parts.
1. INTRODUCTION
Cleaning without chemistry is not new; however, the information in this paper is
new. This paper is not about flushing with pure water to displace particles in criti-
cal cleaning situations nor is it about slowly dissolving non-volatile residue
(NVR) [1] without rinsing residues in precision cleaning situations.
This paper is about cleaning large or small quantities of oils, greases, or soaps
from metal (chiefly) parts using only water [2]. This paper describes a new
method of solving practical cleaning problems at all levels of cleaning quality [3].
This new method is used most often in continuous processes, though it can be
used easily in batch processes.
In this paper, the new method is described and compared to normal aqueous
cleaning technology. Photographic evidence of its use is shown. Limitations are
defined. Areas for future research are discussed.
2. EXPERIMENTAL
Four photographs, Figures 1a to 1d, were taken at 32°C (90°F) in un-degassed tap
water. They illustrate the experiments carried out here to describe the two-step
process. In the first process step, the parts to be cleaned are bombarded with ultra-
∗
Phone: (830)-895-3755, Fax: (612)-677-3170, E-mail: jdurkee@precisioncleaning.com
130 J.B. Durkee
sonic waves from a transducer. This produces an emulsion of water and soil on
the parts. In the second step, the emulsion (which contains the soil) is rinsed from
the parts using any combination of water jets.
The circular-shaped ultrasonic transducer produced 2000 watts over about a
30.48 cm (12 inch) length. The part was an easily recognizable form – a bullet
formed of carbon steel. The part was held on a thin copper wire about 1.27 cm
(0.5 inch) away from the circular transducer. The soil was 10W30 motor oil ap-
plied liberally with a paintbrush before the part was immersed in water. The four
photographs in Figure 1 were taken over a period of about five seconds.
The emulsion was analyzed for oil, and part cleanliness was measured.
The fundamentals of no-chemistry process cleaning 131
3. RESULTS
4. DISCUSSION
Today’s aqueous and solvent processes both involve the same three factors: sol-
vency, temperature (heat), and mechanical force. This is shown in Figure 2a.
Naturally, solvency dominates solvent cleaning. Mechanical force dominates
aqueous cleaning.
The No-chemistry technology presented in this paper involves a different para-
digm. This is shown simply in Figure 2b. There is no chemistry so there is no di-
rect concern about solvency. The process is practiced from room temperature (or
below) up to about 51.6°C (125°F). Consequently, the addition of heat is not a
significant factor.
Table 1.
Comparison of no-chemistry cleaning to cleaning via cavitation
Item Conventional cavitation No-chemistry cleaning
Water degassing Essential Not done
Optimum temperature ~ 71°C (160°F) Any temperature
Hydraulic conditions Static flow conditions Moving flow conditions
Standing waves Avoided via frequency sweep No frequency sweep
Bubbles Observed Not Observed
Cavitation YES NA
Hence, one might assume that the nature of the force is cavitation-induced implo-
sion of tiny bubbles. This assumption is incorrect. Cavitation plays no role in this
No-chemistry cleaning process. In fact, conditions are deliberately controlled to
be opposite of those which normally produce cavitation bubbles. These conditions
are given in Table 1, and will be later shown in Figure 3a to Figure 3d.
If multiple transducers above and below the parts are used, the same frequency
is used for each, vs the referenced patent [4], which requires a difference in fre-
quency. The purpose of the multiple transducers is to expose the surface which
would not ordinarily be exposed to ultrasonic waves. Flat parts, which have two
sides, are a common example.
The rinsing nozzles can be located under water with the transducers or the parts
can be raised above the water and rinsed there. Obviously, a different design of
the rinse nozzle would be used in each case.
The rinsing step is absolutely crucial, as is shown by the following experience.
A scaled-up machine was constructed. It had two steps of emulsion production via
sonication, and no intermediate step of rinsing. It was worthless as a cleaning ma-
chine, despite 2,000 watts of ultrasonic power aimed at a few non-moving parts.
After a great deal of trial and error, an intermediate rinsing zone was added. It
was this inadvertent experiment which showed that No-chemistry cleaning was a
two-step process.
Water quality is not relevant. Excellent cleaning has been obtained in deionized
water, tap water, water produced by reverse osmosis, water containing tramp in-
soluble soil, and water containing ~ 1% emulsion.
There is no practical limit to operating temperature. Issues which determine
temperature [2] are soil and substrate. Lower temperatures make emulsion forma-
tion and removal more difficult because the emulsion is more viscous. Higher
temperatures cause steel surfaces to rust. Typical values of operating temperature
are between 26.6°C (80°F) and 51.6°C (125°F).
Excellent cleaning of molybdenum-based grease from 304 stainless steel has
been done at 90.6°C (195°F) [5]. This temperature was chosen to reduce the vis-
cosity of the grease in order to increase indirectly the rate of formation and re-
moval of emulsion. 10W50 motor oil has been removed from carbon steel at
15.6°C (60°F).
5. FUTURE RESEARCH
6. SUMMARY
Cleaning of oil and grease soils without chemistry is technically feasible. Photo-
graphs show how these soils are emulsified in water. Cleaning data show how the
emulsions are rinsed to produce clean parts. This two-step cleaning process has
been implemented in continuous or batch machines.
Acknowledgments
I would like to thank Walter Johnson, Flo-Matic Corporation, Rockford, Illinois,
USA, for his sponsorship of this work and for the freedom to experiment without
regard to commercial need. I would also like to thank my colleagues Gary
Kauffman and Vasilly Pekun of Flo-Matic.
REFERENCES
Abstract—The mission of this project was to develop a practical technology for formation of ice
particles. Our previous works demonstrated the effectiveness of the use of ice-air mixture as a clean-
ing medium. However, a practical technology for fabrication of ice particles of a desired size and at
a desired temperature is still under development. The physical properties of ice (tendency to ag-
glomerate, melting, etc.) make the formation and transportation of ice particles extremely difficult.
We developed a process for controllable generation of ice particles using a rotational crusher em-
bedded into a heat exchanger. Water was supplied at the bottom of the heat exchanger and as it
moved along the rotating auger ice was formed and crushed. A refrigerant or liquid nitrogen were
used as cooling media. At the exit of the heat exchanger the air stream entrained the generated parti-
cles. In the course of operation the cooling conditions and the auger rotation were maintained con-
stant while the water flow rate varied. The rate of production and the shape and size of the generated
particles were monitored. We also investigated the temperature distribution along the heat ex-
changer and the corresponding distribution of particle sizes. As the result of this study, process phe-
nomenology was developed and a design of the system for formation of ice particles was suggested.
Keywords: Ice; particles; solidification; solid decomposition; brittle fracturing; thermal stresses.
∗
To whom all correspondence should be addressed. Phone: (973)596-3338, Fax: (973)642-4282,
E-mail: geskin@njit.edu
138 D.V. Shishkin et al.
volved the use of ultra-clean ice particles, having a uniform grain size, for clean-
ing the surface and grooves of ferrite block (Tomoji [4]). An ice blasting device
utilizing stored particles was suggested by Harima [5]. Vissisouk [6] proposed to
use ice particles near melting temperature for surface decoating. Mesher [7] de-
veloped 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 µm 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 distance 50 mm. At these conditions the rate of derusting
ranged from 290 mm2/min at the air pressure of 0.2 MPa to 1110 mm2/min at the
air pressure of 0.76 MPa. The quality of the treated surfaces complied with ISO
8501-1 Sa 2.
In the final analysis, the adoption of the ice-jet technology is determined by the
effectiveness of the generation and handling of ice particles. Regular abrasives are
stable at practically feasible operational conditions, while ice particles can exist
only at subzero temperature. Maintaining such a temperature prior, within and
outside of the nozzle is an extremely difficult task. The adhesion between the par-
ticles increases dramatically as the temperature approaches 0°C. At this condition
ice tends to pack and clog the supply lines. Thus these lines must be maintained at
a low temperature. This and other similar problems prevent adoption of ice-
waterjet (IWJ) by the industry. In order to assure the acceptance of IWJ, it is nec-
essary to develop a practical technology for formation, transportation and accel-
eration of ice particles.
stresses in the ice are below a certain level and are applied for a short period of
time. The dynamic elastic properties of ice at –5°C are: Young’s modulus (E) =
8.9-9.9 GPa; Rigidity modulus (G) = 3.4-3.8 GPa; Bulk modulus (K) = 8.3-11.3
GPa; Poisson’s ratio (χ) = 0.31-0.36 according to Hobbs [13]. For comparison,
for an aluminum alloy 1100-H14, E = 70 GPa and G = 26 GPa. For silica glass E
= 70 GPa. If the columnar ice is stressed perpendicular to the long direction of the
column, the static Young’s modulus in bars is determined by the following equa-
tion:
E = (5.69-0.64T)*104 (1)
where, temperature T is given in °C. The dynamic Young’s modulus of ice in-
creases almost linearly from 7.2 GPa at –10°C to 8.5 GPa at –180°C, and is inde-
pendent of the direction of loading. The data above show that the ice powder can
be considered as a soft blasting material and used accordingly.
One of the main issues in the use of the ice powder is sintering of the particles
and their adhesion to the surface of the enclosure. The strength of adhesion de-
pends on ice temperature. This dependence is shown in Figure 1 (a). It follows
from this figure that it is necessary to maintain ice temperature below –30°C to
prevent sintering of the particles. Sintering also depends on the duration of parti-
cles contact. The radius of the neck, which forms between two ice spheres,
Figure 1. (a) Strength of adhesion of ice particles, and (b) schematic of the sintering of ice particles
[13].
140 D.V. Shishkin et al.
brought into contact for time t at temperature T (Figure 1 (b)) is determined by the
equation
n
æ x ö A(T )
ç ÷ = m t (2)
èrø r
where x is the radius of the neck, r is the radius of sphere, A(T) is a function of
the temperature, n and m are constants. A(T), m and n depend on the mechanism
of sintering. It follows from equation (2) that it is necessary to prevent the contact
between particles in order to avoid particles sintering.
The moisture contained in the atmosphere in the course of ice transportation
enhances the adhesion of the ice to the walls as well as sintering of ice particles.
Both phenomena result in the plug formation and clogging of the conduits.
In our previous experiments several systems for ice powder formation were
tested. One of such systems as depicted in Figure 2 was selected for further ex-
periments. The system consists of the following functionally separated blocks:
– ice making block which includes the evaporator, auger, auger driver, sealing
and liquid nitrogen cooling apparatus;
– ice unloading mechanism
– nozzle block which includes parallel nozzles and focusing device.
In our experiments water entered the heat exchanger via a special port. As it
moved along the rotating nozzle it solidified and an ice plug was formed. Decom-
position of this plug led to formation of ice particles. At the outlet of the heat ex-
changer the powder entered into the nozzle block 5 and was driven to the air gun 6.
The heat exchanger and the auger of the icemaker constituted a modified com-
mercial icemaker of the Hoshizaki Co. of America, Peachtree, GA. The design of
these parts will be changed in the next generation of the device. The cooling was
carried out by the refrigerant Galden HT-55 supplied by the cooling TurboJet ap-
paratus or by liquid nitrogen stored in a tank. In both cases the supply of the cool-
ing medium was determined by the characteristics of the source, the refrigeration
system or the nitrogen tank. We replaced Hoshizaki auger driver by a more pow-
erful device in order to prevent jamming of the ice. The rotation momentum of the
auger 4 was provided via a gearbox with gear-ratio 1:100. However, the selected
driver operated at a constant angular velocity of 100 rpm. At these conditions, the
size and temperature of particles were determined by the rate of the water supply
to the port 8, which changed gradually from 0 to 200 ml/min. Water flow rate was
precisely controlled by a special valve (Figure 2 (a)). In the final analysis, we were
able to generate the desired kind of particles at given cooling conditions by the
proper selection of the water flow rate. The attempts to improve ice production by
Development of a technology for generation of ice particles 141
Figure 2. (a) Schematic of auger type IJ system where: 1 – evaporator, 2 – refrigerant coils, 3 – in-
sulation, 4 – auger, 5 – ice reloading device, 6 – air gun, 7 – air supply port, 8 – water supply port,
9 – cooling medium port, 10 – gauges, A – air flow rate valve, B – water flow rate valve, C – cool-
ing medium valve, D – data acquisition card, and (b) schematic of gauges placement inside the
evaporator with their distances measured from water inlet port (water zero level).
an increase of the water pressure or a control of the supply of the cooling medium
were unsuccessful. Solidified ice plug moved forward along the auger helical
ways. It was determined, however, that it was absolutely necessary to eliminate
any obstructions to ice flow in order to prevent jamming of the heat exchanger. It
is quite obvious that the conditions of the ice production (specific cost and energy
consumption, process stability, uniformity of the generated particles, output per
cm2 of the outlet cross-sectional area, etc.) will be dramatically improved by the
process optimization.
An unloading mechanism delivered ice particles to the abrasive port of the air
gun. The nozzle block consisted of two nozzles and a special focusing device.
Three different sizes of the nozzles were used, however, in all cases the nozzle to
focusing tube ratio was 1:2. Ice was delivered to the nozzles abrasive port through
insulated flexible plastic tubes.
The ice crystallization was monitored through a set of thermocouples and resis-
tance gauges (10) imbedded into the evaporator (Figure 2(b)). Data acquisition
card processed the signals generated by the gauges.
The large number of control variables (rate of supply of water and coolant, rate
of auger rotation, diameter and length of the heat exchanger, distribution of tem-
perature and heat flow along the refrigerator, geometry of the auger) practically
excluded empirical process optimization. At the same time, the complexity of the
142 D.V. Shishkin et al.
Figure 3. Solid ice block decomposition at the two distinct stages of ice powder formation: a) initial
stage of block formation, and b) the intermediate stage of block decomposition.
phenomena that occur in the course of the powder formation (water freezing and
simultaneous decomposition in the turbulent layer) made it impossible to con-
struct a physical model of the process. Thus the only practical approach to the
process design entailed experimental study of the distribution of water properties
within the icemaker, construction of the process phenomenology using the ac-
quired information and then the use of commercial packages (Pro/ENGINEER,
FIDAP, etc) to evaluate the design parameters of the icemaker as well as the cor-
relation between input variables and the ice properties.
Special experiments were carried out in order to examine the state of the solid
ice in the course of solidification and particles formation. In order to attain this
goal the auger was stopped during a normal operation and then removed from the
heat exchanger. The solid phase was visually examined. The state of the ice is de-
picted in Figure 3. The anatomy of the ice in the course of freezing and particles
formation was investigated using the set of pictures developed. This information
was supplemented by the ice temperature and electrical resistance monitored prior
to the auger removal. Extensive database acquired in the course of these experi-
ments was used to develop a hypothetical mechanism of ice particles formation
The suggested hypothetical process phenomenology is described below.
the ice structure is complete and further cooling brings about only reduction of ice
temperature. Thus particles size can be controlled only at the initial stages of the
process.
The following phenomenology of particles formation has been suggested. Ice
nucleation inside the evaporator obeys the rules of granular ice formation (Ice
type I). It originates as water flow reaches the evaporator wall with a temperature
of –196°C. Ice nucleation and crystal growth depend on the conditions of cooling
and the auger spiral movement. The ice plug forms and propagates along the au-
ger stem. The generated ice acquires a multilayer pattern. A supercooled ice layer
directly adjacent to the evaporator and a moderately cooled layer in the vicinity of
auger can be identified. An explosive character of water freezing near the evapo-
rator wall leads to thermal ice expansion inside the volume constrained by the
evaporator and the auger and formation of the ice plug. This results in the com-
pression of the solid and nucleation of cracks. The nucleation of cracks under a
compressive stress is generally due to dislocation pile-up at the grain boundaries
and relief of stress concentration along the grain boundaries. The phenomenon of
crack nucleation has been discussed for low to moderate loading conditions by
Sanderson [14]. His studies indicate that crack nucleation is determined by the de-
layed elastic strain. The generated cracks as well as the cracks initially developed
in the solid ice propagate within the ice body and eventually fracture the solid
block. Block decomposition occurs almost instantaneously and can be considered
as an explosive destruction. The ice temperature in this region ranges from –15°C
to –150°C (distance from water inlet level D = 10-24 mm).
The following hypothetical mechanism of ice formation and decomposition
(Figure 4) was suggested. Liquid water freezes and forms ice plug in the bottom
zone of the evaporator. Thermal expansion essentially stops ice advance, and de-
velops intensive stresses within the plug. This brings about formation and devel-
opment of cracks and eventually fracturing the solid ice plug into particles. The
density of cracks developed in the course of freezing determines the size of parti-
cles. This density, in turn, depends on the rate of water cooling.
This explains the effect of the rate of water supply on the size of particles. At a
low rate of water supply, i.e. at a high rate of freezing the fine particles are gener-
ated. The periodic mode of formation and decomposition of ice blocks determines
the periodical character of driver current oscillations. Indeed, the ice powder exits
the icemaker periodically and the periods of ice exit exactly coincide with those
of the driver operation shown in Figure 5. Before plug formation water and ice
flow freely along the auger and thus loading of the driver is minimal. As long as
the plug stops the ice motion, the torque of the driver increases dramatically, but
drops almost instantaneously when the plug is decomposed (Figure 5). After the
plug decomposition, the geometry of particles does not change and the heat ex-
change results in ice cooling only.
144 D.V. Shishkin et al.
Figure 4. Hypothetical schematic of ice plug formation and decomposition in an auger spiral move-
ment: (a) bottom or solidification zone of the evaporator, (b) middle or supercooling zone of the
evaporator, (c) ice plug decomposition zone, and (d) ice particles cooling zone.
Figure 5. The correlation between the driver current with time for different water flow rates. Notice
distinct frequencies of maximum current value for different water flow rates.
Development of a technology for generation of ice particles 145
Our previous work showed the possibility of using ice particles for a wide variety of
surface processing operations. However, it was found that each specific technology
required a narrow range of particle sizes and temperature. Derusting of steel, for ex-
ample, requires particles ranging from 3 to 7 mm, while the biomedical applications
require highly homogeneous ice powder in the range 0.25 mm-0.3 mm.
Figure 6. Size distribution of ice particles for two different cooling media: (a) the average diameter
of ice particles as a function of a temperature at the evaporator outlet, and (b) the average diameter
of ice granules as a function of water flow rate.
146 D.V. Shishkin et al.
Figure 7. Pictures show the high and low density ice-airjet streams (air pressure is 0.544 MPa) and
ice particles flow rate: (a) 20 g/min, and (b) 60 g/min.
A series of the experiments were carried out in order to demonstrate the potential
application of the ice-air jet for various surface-processing operations.
Figure 8. (a) Waterproof marker was partially removed from the epidermis layer of pork skin. Then
the epidermis layer was removed too. No damage to the underneath layers was detected, and (b) wa-
terproof marker was removed from the highly sensitive surface of chicken skin. No damage to the
epidermis was observed in course of the cleaning procedure. The pictures were taken with a Sony
MVC-FD71 digital camera. Note: the marker (a) was removed without damaging the skin epidermis
layer (b).
Figure 9. (a) Decontamination of a heavily contaminated machine part, and (b) removal of highly
adhesive glue layer from a plastic surface. The pictures were taken with a Sony MVC-FD71 digital
camera. Note: the highly adherent layer of grease (Fig. 9 (a)) was removed without damaging the
painted surface of the machine part (right hand side picture). The residue of glue (Fig. 9 (b)) was
removed in course of IAJ cleaning. No damage to the plastic surface was observed (right hand side
picture).
Figure 10. (a) Rusted carbon steel surface. Notice that newly formed rust layer is highly adherent,
and (b) carbon steel plate was partially derusted using IAJ (middle part of the plate). The pictures
were taken with a Sony MVC-FD71 digital camera.
150 D.V. Shishkin et al.
7. CONCLUSION
The suggested mechanism for the formation of ice powder enables us to design an
effective industrial scale device for ice jet formation. The procedure for ice parti-
cles formation was developed and a device readily available for industrial de-
ployment was constructed. An extensive use of this device is envisioned.
Acknowledgement
The study was supported by NSF grant number DDM9312980.
REFERENCES
FRED C. YOUNG∗
Cold Jet, Inc., 455 Wards Corner Road, Loveland, Ohio 45140, USA
Abstract—Blasting with solid carbon dioxide (dry ice) pellets is a technology that is gaining wide
acceptance in industry and the military for removing coatings and contaminants from surfaces. Dry
ice pellet blasting is also being used to prepare surfaces prior to applying coatings. This paper ex-
plains the principles of dry ice pellet blasting, and includes a brief history of the development of the
technology. A discussion follows to explain how dry ice pellet blasting works by combining kinetic
energy with thermal shock. The two kinds of dry ice pellet blast systems that are commercially
available, the induction (venturi) and direct acceleration types, are also discussed. The operating
principles and performance characteristics of both types are explained and compared in detail. Blast-
ing system control parameters, such as dry ice pellet flow rate, compressed air propellant flow rate,
and dry ice pellet density, are defined and their effects on cleaning performance are presented. To
conclude, the author provides a review of actual applications for dry ice pellet blasting technology
as it is currently used in the semiconductor manufacturing industry.
Dry ice is made from liquid carbon dioxide, a recycled byproduct of several
manufacturing processes. During the blasting process the dry ice sublimates to
carbon dioxide gas, just like that exhaled by humans and found naturally in our
atmosphere. Using dry ice is safe for employees and the environment. Dry Ice
blasting uses extruded dry ice pellets, roughly the size of a grain of rice. The pel-
lets are made on a dry ice extrusion machine called a “pelletizer” or “nuggetizer”.
The pellets can be produced, stored in sealed insulated containers, shipped, and
used for blasting several days after they are produced.
Dry ice blasting accelerates solid carbon dioxide pellets with compressed air in a
subsonic or supersonic blast stream to remove unwanted surface contaminants.
Upon impact with the surface the dry ice sublimates (turns from a solid to gas
without passing through a liquid phase) into carbon dioxide. The process is dry and
non-conductive, non-abrasive and non-toxic, leaving no residue on the part or
equipment being cleaned. All that remains to be collected (by vacuuming and/or
∗
Phone: 513-831-3211 ext. 367, Fax: 513-831-1209, E-mail: fyoung@coldjet.com
152 F.C. Young
filtration) is the surface contaminant being removed. Dry ice blasting leaves no
secondary waste as from sand, bead or water blasting. This allows equipment sur-
faces to be cleaned in-place during the manufacturing process. The absence of a
secondary waste stream makes dry ice blasting a perfect non-polluting technology.
With a low temperature of –79°C, dry ice (solid carbon dioxiode) has an inherent
thermal energy ready to be tapped. In addition to the kinetic energy associated
with any accelerated medium blasting, dry ice blasting uses the inherent low tem-
perature to increase shear stress in the surface coating or contaminant, enabling
the particle impact to break-up the coating. Further, the thermal gradient between
two dissimilar materials (the contaminant and the substrate) with different thermal
expansion coefficients can serve to break the bond between the two materials. The
ability of these surface mechanisms to remove the coating or contaminant varies
depending on coating or contaminant. Thermal shock is most evident when blast-
ing a thin, non-metallic coating or contaminant bonded to a metallic substrate.
Thermal shock, a key element that makes dry ice blasting an effective cleaning
method, does not cause thermal stress in the substrate being cleaned. The tem-
perature decrease caused by dry ice blasting is localized at the surface where the
contaminant is bonded to the substrate. (See references [1, 2].)
In the early 1930’s, the manufacture of solid phase carbon dioxide (dry ice) be-
came possible. During this time, the creation of dry ice was nothing more than a
laboratory experiment. As the procedure for making dry ice became readily avail-
able, applications for this innovative substance grew. Obviously, the first use was
in refrigeration. Today, dry ice is widely used in the food industry for packaging
and protecting perishable foods.
In 1945, stories exist of the U.S. Navy experimenting with dry ice as a blast
medium for various degreasing applications. In May 1963, Reginald Lindall re-
ceived a patent for a “method of removing meat from bone” using “jetted” carbon
dioxide particles. In November 1972, Edwin Rice received a patent for a “method
for the removal of unwanted portions of an article by spraying with high velocity
dry ice particles”. Similarly, in August 1977, Calvin Fong (then working for the
Lockheed Corp.) received a patent for “sandblasting with pellets of material ca-
pable of sublimation”. The work and success of these early pioneers led to the
formation of several companies in the early 1980’s that pursued the development
of dry ice blasting technology. (See reference [3].)
Cleaning with solid carbon dioxide pellet blasting 153
In recent years, dry ice pellet blasting has found its most commercially success-
ful “market niche” for in-line (in-process) cleaning of molds (rubber, plastics,
aluminum foundry, food baking). Other emerging areas of successful commercial
application are with specialized contract cleaning services, the wood and paper
industries, the semiconductor manufacturing industry, the printing industry, and
the aerospace industry.
4.1. Overview
Carbon dioxide pellet blasting uses compressed air to accelerate frozen carbon di-
oxide “dry ice” pellets to a high velocity. A compressed air supply between 415
kPa and 620 kPa pressure is required. Dry ice pellets can be made on-site or sup-
plied. The pellets are made from liquid carbon dioxide, which is a naturally occur-
ring compound that is non-toxic, non-flammable and chemically inert. Carbon di-
oxide is inexpensive and easily stored at work sites.
coating layer itself, and 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 or 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 dry 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, dry 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. Induction (venturi) and direct acceleration blast systems – the effect of
the type of system on available kinetic energy
In a two-hose induction (venturi) carbon dioxide blasting 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”
at 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-
tic 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
Cleaning with solid carbon dioxide pellet blasting 155
immediately, and continue to gain momentum as they travel the length of the
hose. At the end of the hose, the spray nozzle “gun” actually consists of a conver-
gent-divergent nozzle, which exchanges pressure differential across the nozzle for
a huge increase in air and particle velocity. Carbon dioxide particle velocities
have been measured and substantiated in excess of 215 meters per second, and up
to as high as 270 meters per second at the nozzle exit plane. This is accomplished
at less than one third of the flow volume (only 3000 L/min compared to 10,000 or
more L/min) required by the most aggressive two-hose systems. In addition to the
lighter weight and less cumbersome hand held applicator and hose of a two-hose
system, the contaminant removal energy delivered to the surface is considerably
higher than that provided by a two-hose inductive system. Even with solid carbon
dioxide particle blasting, a significant component of the contaminant removal en-
ergy is the kinetic energy per unit of area delivered to the surface. Since kinetic
energy is a function of mass and velocity of the particles, i.e., Ke=1/2 mv2, it can
be seen that a two-fold increase in particle velocity, with equal particle mass and
equal nozzle spray area, effectively increases the impact energy delivered to the
surface by a factor of four. A three-fold particle velocity increase, from 90 to 270
meters per second, increases the blast impact energy nine times that of a two-hose
system.
4.2.2. Thermal shock
Unlike other blast media, the carbon dioxide particles have a very low tempera-
ture of –78°C. This inherent low temperature imparts the dry ice blasting process
with unique thermodynamically induced surface mechanisms that affect the coat-
ing or contaminant to a greater or lesser degree, depending on coating type. Be-
cause of the temperature differential between the dry ice particles and the surface
being cleaned, a phenomenon known as thermal stress fracturing (“fracking”) or
THERMAL SHOCK can occur. As the temperature differential between the coat-
ing and the substrate increases, the thermal shear stresses in the coating increase
and couple with the impact induced stresses to increase the coating removal rate.
A good example is the fouling which occurs on rubber, plastic, and tire curing
molds. This contaminant is a chemical compound created by the interaction of
mold release products and the base polymer under high pressure and temperature.
The contaminant or “fouling” resembles a very thin glass-like material which re-
sponds very readily to the thermal shock effect of dry ice pellet blasting. In fact,
hot molds, at or near the cure temperature of 160°C, can be cleaned three to four
times faster than the same dirty molds at room temperature.
Dry Ice particle blasting is emerging as a method of choice for critical cleaning
requirements in the semiconductor industry. The areas in the semiconductor
manufacturing process where dry ice particle blasting is currently being applied
are:
(1) Silicon wafers are polished to a high degree of surface flatness in the early
stages of the diffusion process. The polishing compound is deposited randomly on
the internal surfaces of the polishing machines, then dries as a hard abrasive coat-
ing. Flecks of this abrasive contaminant can fall back onto the surface of newly
inserted wafers, causing deep scratches that cannot be polished out, resulting in
costly scrapping of silicon wafers. Dry ice particle blast has been found to be the
best method to remove the abrasive, dried-on polishing slurry from the polishing
equipment, without damaging the expensive polishing equipment components.
Cleaning with solid carbon dioxide pellet blasting 157
(2) Integrated circuits (ICs, or Chips) are produced in molds where the finished
silicon based integrated circuit is encased in a material called EMC (Epoxy Mold
Compound). The EMC eventually builds up a light deposit on the mold surface
which can cause sticking of the finished ICs in the molds, and in other forms of
IC surface defects. The traditional method used to remove EMC deposition is to
coat the molds with melamine, let it cure, then pull the melamine off the surface.
The melamine adhesively bonds to the EMC deposit. The EMC deposit is pulled
off the mold surface together with the melamine as the melamine is removed. The
major drawback is that the melamine tends to also pull the chromium plating off
the mold surfaces, rendering the molds useless. Carbon dioxide particle blast has
been demonstrated to be very effective in cleaning the IC mold surfaces without
removing the chromium plating or otherwise damaging the mold surfaces.
(3) Silicon wafers are photo-etched as part of the process to cut the wafers into
the individual rectangular or square ICs, and as part of the process to produce the
final surface transistor circuits. A compound called photoresist is deposited on
silicon wafers to mask them in areas where the photo-etching action is not de-
sired. After the photo-etching processes, the photoresist must be removed from
the wafer surface. Carbon dioxide particle blast has shown very promising results
in this area, and a great deal of developmental effort is now underway to bring
this process into widespread use in the semiconductor industry.
(4) A contamination problem arising from the etching process is the outgassing
of compounds that redeposit on the surfaces of the etching equipment (e.g., fix-
tures, insides of chambers). This polymer-like deposit can re-contaminate subse-
quent wafers being etched in this high temperature process. Also, the deposit can
build up so thick that the wafer holding fixtures become unusable. Some of the
etching process fixtures are made of very expensive and delicate materials, like
quartz. Traditional deposit removal methods include soaking in chemical baths
with toxic solutions. Carbon dioxide particle blasting has been found to be very
effective for removing the etching contaminants on quartz and some other sub-
strates.
6. CONCLUSION
Dry ice pellet blasting is a surface cleaning and preparation technology that is
gaining increasing popularity and acceptance each year, primarily as an industrial
mold cleaning process. Industry’s acceptance of dry ice blasting is based on ran-
dom trial-and-error testing by a few companies, for the purpose of verifying ac-
ceptability for their own cleaning applications, and the great willingness of the
general industry to “assume” that the mechanics and physics behind dry ice blast-
ing were well understood, documented, and as easy to apply to any given applica-
tion (like a simple cookbook recipe). In fact, there is very little scientific basis
(through investigation, testing, evaluation, and reporting) to support the assump-
tions behind the theories of how and why dry ice blasting works. Industry has ac-
158 F.C. Young
cepted dry ice blasting for mold cleaning because it has been proven to increase
profitability and improve product quality. In the harsh environment of a typical
foundry or rubber parts molding shop, dry ice blasting is a relatively benign mold
cleaning method compared to sandblasting and other crude forms of abrasive grit
blasting.
More recently, other industries, such as semiconductor manufacturing, radioac-
tive waste decontamination, and aerospace, are beginning to view dry ice blasting
as an improvement over current methods of cleaning and surface preparation, or
even as a potential “breakthrough” technology for use in developing completely
new manufacturing or processing methods. Many of these new potential applica-
tions require the cleaning or preparation of very delicate surfaces, such as thin
metal alloys, silicon wafers, composite materials, and even populated printed cir-
cuit boards. It is for these types of applications that a much better understanding
of the dry ice blast contaminant removal phenomenon is required, so that much
more precise control over the process, and predictability of the outcome of using
the process, can be achieved.
REFERENCES
1. L.C. Archibald, “Cold Jet Thermal and Surface Cleaning Characteristics” (June 1988, The Pro-
duction Engineering Research Association of Great Britain, business name PERA). PERA is lo-
cated in Melton Mobray, Leicestershire, UK. Telephone 44-664-501501.
The information in this report is restricted to Cold Jet, Inc., 455 Wards Corner Road, Loveland
45140, USA. Please contact Cold Jet, Inc. at 513-831-3211 to obtain a copy of this report.
2. K. Lay, “An Analysis of Mold Integrity After Carbon Dioxide Blast Cleaning”, published in the
proceedings of the International Tire Exhibition and Conference (ITEC), 1996, Akron, Ohio, USA.
(Available by contacting Crain Communications, Inc. 1725 Merriman Road, Akron, Ohio
44313-5251. Telephone 330-836-9180. Fax 330-836-1005)
3. J.A. Snide, “Carbon Dioxide Pellet Cleaning - A Preliminary Evaluation”, Materials & Process
Associates, Inc., October 12, 1992.
4. D.R. Linger, “Fundamentals of Dry Ice Blast Cleaning Technology”, published in the proceed-
ings of the International Tire Exhibition and Conference (ITEC), 1996, Akron, Ohio, USA.
(Available by contacting Crain Communications, Inc. 1725 Merriman Road, Akron, Ohio
44313-5251. Telephone 330-836-9180. Fax 330-836-1005)
5. C. Cundiff, “Evaluation of the Cold Jet, Inc. Carbon dioxide Blast System for Paint Stripping”,
Battelle, 505 King Avenue, Columbus, Ohio 43201, USA, October 18, 1989.
6. F. Young, “Blast Off” article published in Tire Technology International Magazine, December,
2000. Pages 54–58.
(Available by contacting UK & International Press, Abinger House, Church Street, Dorking,
Surrey RH4 1DF, UK. Telephone +44 (0) 1306 743744. Fax +44 (0) 1306 742525. E-mail
tire@ukintpress.com)
7. F. Young, “Tire Mold Maintenance with Solid Carbon Dioxide Pellet Blasting”, published in the
proceedings of the International Tire Exhibition and Conference (ITEC), 1998, Akron, Ohio, USA.
(Available by contacting Crain Communications, Inc. 1725 Merriman Road, Akron, Ohio
44313-5251. Telephone 330-836-9180. Fax 330-836-1005)
Surface Contamination and Cleaning, Vol. 1, pp. 159–172
Ed. K.L. Mittal
© VSP 2003
Abstract—A practical procedure for utilization of available information, both numerical and lin-
guistic and identification of the operational conditions of the waterjet cleaning is presented. Neural
Networks based prediction models were constructed using previously available information. The
constructed models constituted knowledge base for the procedure. Then a single parameter, the ero-
sion strength for cleaning, was determined experimentally. The fuzzy logic technique enabled us to
determine a weighted contribution of each preliminary constructed model for the process in ques-
tion. Thus, the first approximations of the operational conditions are determined. In the course of the
further operation the developed model is improved. The developed procedure will assist a practitio-
ner in the selection of a decontamination technology for an unknown surface.
1. INTRODUCTION
∗
To whom all correspondence should be addressed. Phone: (973) 596 3338,
Fax: (973) 642-42882, E-mail: geskin@njit.edu
160 K. Babets and E.S. Geskin
from cleaning agents, etc.) require alternative methods for effective material de-
contamination based on physical coating removal techniques.
Physical coating removal techniques take advantage of differences in physical
properties between the coating and the substrate to destroy the bonding and/or
abrade the coating from the underlying substrate. Physical coating removal tech-
niques use one or more of four general types of physical mechanisms [1].
● Abrasive techniques wear the coating off with scouring action.
● Impact techniques rely on particle impact to crack the coating to remove it.
● Cryogenic techniques use extreme cold conditions to make the coating more
friable and induce differential contraction to debond the coating.
● Thermal techniques use heat input to oxidize, pyrolyze, and/or vaporize the
coating.
These techniques include but are not limited to: plastic media blasting, wheat
starch blasting, sodium bicarbonate wet blasting, high pressure water blasting and
cryogenic blasting. It is clear that the water blasting constitutes the most effective
technique. Water is readily available, comparatively inexpensive, and induces no
damage to the environment. A complete separation of water and debris facilitates
material recovery. Therefore, complete pollution prevention is feasible.
Although numerous extensive studies of waterjet-based material cleaning have
been implemented [2-4], and this topic is currently of interest to many research-
ers, there is no one universal technique that will allow practitioners to bridge the
gap between the available information about the process and the current need of a
practitioner to remove a specific contaminant. This difficulty renders the process
unusable for most practical applications. Therefore, a goal of this research was to
develop a modeling tool that would assist in practical implementation of such a
technology.
available in the literature we define an area cleaning efficiency, Ea, as the ratio of
area cleaned per unit time and power delivered by the nozzle:
Area cleaned per unit time &A
Ea = = (1)
Power delivered by nozzle P
&A = I (3)
Sc
or
I
Ea = (4)
P ⋅ Sc
Combining expressions (1) and (4) results in
Ea ∝ S c − 1 (5)
The relation (5) is the basic relation, used to derive the curves, representing the
dependence of the area cleaning efficiency Ea and erosion strength for cleaning Sc
(Figure 1).
2.2. Determination of the erosion strength based on the available cleaning examples
The available experimental database reflecting material decontamination with
pure waterjet was compiled, and the area cleaning efficiencies were calculated for
deposit types given in Table 1. The data in Table 1 were used to derive the rela-
tionships between area cleaning efficiency Ea and erosion strength for cleaning Sc
162 K. Babets and E.S. Geskin
(Figure 1). The point of departure for the line 276-310 MPa in Figure 1 was the
data for Item 1 from Table 1. The calculated cleaning efficiency for removal of
Epoxy #1 deposit was assigned an Erosion Strength Sc=103, relative units. This
line, per relation (5) was plotted on a log-log chart. The location of the rest of the
data points was based on calculated area cleaning efficiencies and working water
pressure.
To derive line for 138 MPa (Figure 1) items 1 and 2 were compared. Since the
Erosion Strength for the deposit type Epoxy #1 is known (Sc=1000 relative units)
and is constant, the first point for line 138 MPa thus could be located. Similarly to
derive the line for 70-100 mPa, item #6 was located at the 138 mPa line and com-
pared with item #7.
The main result that could be inferred from the graphical relationship between
the area cleaning efficiency Ea and erosion strength Sc shows that there is a defi-
nite relationship between Ea and Sc and that Erosion Strength for cleaning of simi-
lar materials is closely spaced together, and consequently the Sc parameter can
used to characterize an unknown deposit-substrate combination. On the other
Development of a generic procedure for modeling of waterjet cleaning 163
Table 1.
Cleaning examples
Area Power Area
Item Deposit Water Nozzle Flow cleaning delivered cleaning
# type pressure diameter rate rate by nozzle efficiencing
(MPa) (mm) m3/s (m2/hour) kW-h m2/kW-h
The problem that most of the waterjet practitioners face when dealing with an un-
known surface is the lack of information about the process or, in other words, the
unavailability of a generic technique that could be used as a first approximation of
the process. This section is concerned with the development of such an approach.
The idea behind such an approach is to combine the previous knowledge about
the process in question and based on that make an informed decision as to which
waterjet parameters to apply, as a first approximation.
There are, however, situations in which the phenomena involved are very com-
plex and not well understood and for which the first principle models are not ef-
fective. Even quite often, experimental measurements are difficult and/or expen-
sive. These difficulties led us to explore the application of Soft Computing
(Artificial Intelligence) techniques as a way of obtaining models based on ex-
perimental measurements. The field of Soft Computing is comparatively new, and
it includes fuzzy logic, neural networks, expert system, cellular automata, chaotic
systems, wavelets, complexity theory, anticipatory systems, among others. But
only fuzzy logic, neural networks and genetic algorithms have reached the stage
of development where they are used for real world problems.
Fuzzy logic systems address the imprecision of the input and output variables
directly by defining them with fuzzy sets (fuzzy numbers) that are generally ex-
pressed in linguistic terms. Moreover, they allow for very complex and nonlinear
systems to be described in very simple terms, thus making them easier to under-
stand. Another important feature of fuzzy systems is their ability to accommodate
the existing expert knowledge of a process into a model by expressing it in terms
of fuzzy rules.
Neural Networks, on the other hand, model a system by using sets of input-
output data to train some generic model of the system. Neural Networks are very
good at modeling very complex nonlinear relationships with large numbers of in-
put and output variables. Models based on neural networks are also easy to opti-
mize, since although the model itself is not given in terms of an explicitly defined
function, the gradient of this function can be found numerically.
The combination of the above two techniques often results in greater flexibility
and/or clearer representation of the model than when they are used separately.
This combination is often referred to as neuro-fuzzy model of the system. Neuro-
Fuzzy Reasoning approach also allows overcoming some traditional problems in
using fuzzy logic or neural networks, such as the problem of defining a member-
ship function, extracting fuzzy rules, etc.
We are using the notion of Erosion Strength (Sc) developed in the previous sec-
tion to classify an unknown surface together with Neural Networks Fuzzy Rea-
soning technique, suggested by Takagi and Hayashi [8], for information process-
ing.
The prediction technique construction begins with the development of fuzzy
universe for erosion strength, Sc. The experimental database allows us to construct
three fuzzy sets, based on the number of experimental situations available. The
hard epoxy deposit with erosion strength of 1000 represents fuzzy Class I in Fig-
ure 2. Using the notation of the Fuzzy Logic theory we state that the hard epoxy
deposit has the degree of membership of one in the fuzzy set Class I, or, in simple
terms, this deposit is the most representative of all deposits that might be classi-
fied as belonging to Class I. Similarly, the rust deposit, with erosion strength of
400 is assigned the degree of membership of one in the fuzzy set Class II, and, fi-
nally, the oil based paint deposit is assigned the degree of membership of one in
the fuzzy set Class III. To explain the idea of “degree of membership” we refer
Development of a generic procedure for modeling of waterjet cleaning 165
the reader to Figure 2. Carefully inspecting this figure we notice that the classes
III and II, and II and I overlap to some degree. This means that some random de-
posit, with erosion strength of say 100, will belong to both class III and class II.
The extent to which this random deposit type is represented by either of the
classes is expressed in terms of the “degree of membership”. Thus, from Figure 2
we notice that a deposit with erosion strength of 100 belongs to Class III with a
degree of membership of 0.8 and at the same time belongs to Class II with a de-
gree of membership 0.2. The higher the degree of membership in a class, the more
representative this class is for a given deposit type. Clearly, if a deposit has a de-
gree of membership of 1.0 in some class, it belongs only to that particular class,
and to no other class.
Thus each of the classes in Figure 2 is represented (with a degree of member-
ship 1) by a specific deposit type available in our experimental database. In other
words, these three fuzzy sets cover the ranges of all possible values for the mate-
rials with erosion strength for cleaning from 1 to 10000 relative units. Similarly, if
we identify a deposit-substrate combination with some value of erosion strength,
Sc, and this value happens to be inside the range [1,10000], then we can identify
the degree of membership of such a deposit-substrate combination in the three
fuzzy sets (Figure 2). This procedure alone can be very useful when trying to
classify some unknown deposit-substrate combination.
166 K. Babets and E.S. Geskin
It should be emphasized that for each of the three basic deposits (hard epoxy
paint, rust, and oil based paint) an extensive experimental study was performed.
From this point on we will refer to these deposits as “base deposits” and equiva-
lently we will refer to the three fuzzy sets (classes) they represent as “base
classes”. Since we possess a required empirical knowledge for these processes,
the appropriate numerical representation of each process can now be made using
artificial neural networks. The procedure for the application of a neural network
for process modeling and optimization was described in [9]. Thus, after each net-
work has been created, properly trained and tested we have obtained reliable nu-
merical models for the three base deposits, or equivalently, three fuzzy base
classes of erosion strength (Sc). Therefore, it can now be stated that the process of
cleaning a material with erosion strength in the range [1,10000] can now be ap-
proximated as some combination of the numerical models of the three base
classes. The computational procedure is as follows. For an unknown deposit, the
practitioner makes a simple experiment that allows him/her to calculate area
cleaning efficiency. Then using the computed Ea we can determine the corre-
sponding erosion strength (Sc) for this surface from Figure 1, for the given water
pressure. Once a corresponding Sc coefficient has been found, the fuzzy member-
ship in the three classes in Figure 2 can be determined. Separate experiments were
conducted for the auto paint deposit removal with plain waterjet. The computed
area cleaning efficiency was calculated as Ea auto paint = 0.04 m2/kW-h. From Figure
1 the corresponding coefficient Sc was found to be Sc=180 relative units. And
from Figure 2, the degree of membership (µ) in the three basic classes can be cal-
culated as µ (class I) =0, µ (class II) =0.4, µ (class III) =0.59. These degrees of membership
can be interpreted as follows. The erosion strength of the auto paint deposit is ap-
proximately midway between that of the hard epoxy paint and rust deposit. Now
that the surface was identified we could use the numerical models available for
the three base deposits (hard epoxy, rust, and oil based paint) to obtain a first ap-
proximation to the process. We supply a set of input parameters (Water Pressure,
Nozzle Traverse Rate, Nozzle Diameter, and Standoff Distance) as an input into
the numerical models represented by the neural networks. The corresponding out-
put in terms of the single strip width is obtained by each of the network. The final
result is obtained by defuzzifing the output according to Equation (6).
r
*
åµ
s =1
AS
(x i ) ⋅ u S (x i )
(6)
y =
i r
, i = 1,2 ... n
åµ
s =1
AS
(x i )
In equation (6), µ is the membership value of the deposit with erosion strength
Sc in the three base classes, Us is the output of the sth neural network, and y* is the
final defuzzified output.
Development of a generic procedure for modeling of waterjet cleaning 167
Figure 3. Generic modeling approach. NN1, NN2, NN3 – Artificial Neural Network models for the
three basic classes. X1, X2, …, Xn – Process input variables.
The output of the model is the single clean width of the strip produced on the
surface by the combination of the input parameters, which then can easily be con-
verted into area cleaning rate. This procedure is sketched in Figure 3.
Table 2.
Cleaning samples
Degree of membership
Sample Deposit Area Erosion Class I Class II Class III
number cleaning strength
efficiency (relative
(m2/kW-h) units)
1 Hard Epoxy Paint 0.04 1000 1 0 0
2 Weaker Epoxy 0.06 665 0.41 0.58 0
Paint
3 Rust from Steel 0.0975 400 0 1 0
4 Weaker Rust from 0.105 360 0 0.89 0.1
Steel
5 Auto Paint 0.04 180 0 0.4 0.59
6 Oil Based Paint 0.21 30 0 0 1
The model performance was tested on each of the test deposits by providing the
model with a set of waterjet input parameters within the working space, obtaining
the corresponding output in terms of the width of a clean strip and comparing the
results with experiments.
5. DISCUSSION
The model for prediction of the results of waterjet cleaning described in the pre-
vious sections was tested on several additional test deposits. Figures 4-9 present
the results of prediction. Analyzing the results, it is clear that the model prediction
results are acceptable at both relative error of prediction (~ 20%), and at following
the trend of the process, which is also important for any cleaning study. Neverthe-
less, as a first estimation of a cleaning efficiency for a given type of deposit, these
results constitute a reasonable approximation.
However, it should be noted that at the current stage the prediction technique
was tested only in the middle of the problem space. At the outskirts of the prob-
lem space, veritable results could not be obtained. The reason for this lies in the
limitations in the development of the three numerical models that represent the
cleaning of the base deposits (hard epoxy, rust, and oil based paint). Since the
ranges of experimental parameters used for the construction of models were dif-
ferent in each case, and there was no coordinated experimental setup, but rather
the data were compiled at later stages, there are inconsistencies in choosing the
levels of process parameters in case of a test cleaning space. For example, if for a
base model development the nozzle traverse rate was in the range from 1000
mm/min to 2500 mm/min, and a test case was run at 1500~4000 mm/min, the re-
liable model performance will be at the intersection of these ranges. The way to
Development of a generic procedure for modeling of waterjet cleaning 169
cure the above problem is to cover all the parameter space (limited by equipment
capabilities) for each process variable in all the base models. Of course, this re-
sults in quite extensive experimentation, but on the good side it needs to be done
only once, when developing the base models. Also the current model does not
cover the full range of all possible erosion strengths of different materials, but by
extending the procedure with additional base models for lower or higher degrees
of erosion strength for cleaning (Sc), this limitation can be reduced or eliminated.
Figure 4. Auto paint removal, standoff distance 13.9 cm. WP – water pressure, ND – nozzle diame-
ter, standoff – stand off distance.
Figure 5. Auto paint removal, width of strip vs. traverse rate. Experimental vs. predicted.
170 K. Babets and E.S. Geskin
Figure 6. Removal of weaker rust. Experimental vs. predicted width of clean strip. Water pressure
241 MPa, nozzle diameter 0.1778 mm.
Figure 7. Experimental vs. predicted width of clean strip. Water pressure 172 MPa, nozzle diameter
0.1778 mm.
6. CONCLUDING REMARKS
The approach presented here for modeling of waterjet cleaning process allows a
user to obtain a reliable process approximation given no or limited information
about process condition. For an unknown surface a practitioner needs to deter-
mine a single coefficient, the erosion strength for cleaning (Sc), based on a simple
experiment(s). The proposed approach utilizes this coefficient and approximates
Development of a generic procedure for modeling of waterjet cleaning 171
Figure 8. Removal of weaker epoxy paint. Nozzle diameter 0.356 mm, water pressure 207 MPa.
Figure 9. Removal of weaker epoxy paint. Nozzle diameter 0.3556 mm, water pressure 276 MPa.
the cleaning results in terms of area cleaning rate. It is believed that the current
work will assist in practical implementation of waterjet cleaning technology,
where the information deficiency on process conditions is the main reason for in-
effective application of this technology.
172 K. Babets and E.S. Geskin
REFERENCES
1. United States Environmental Protection Agency, Office of Research and Development. Guide to
Cleaner Technologies, EPA/625/R-93/015,Washington D.C. (1994).
2. A.F. Conn and G. Chahine, Proc. Third American Waterjet Conference, Pittsburgh, PA, Water-
jet Technology Association, St. Louis, MO (1985).
3. F. Erdman-Jesnitzer, A.M. Hassan and H. Louis, Proc. Third International Symposium on Jet
Cutting Technology, Chicago, IL, British Hydraulic Research Association, Cranfield, UK
(1976).
4. S.T. Johnson, Proc. 7th American Water Jet Conference, Seattle, Washington, Waterjet Tech-
nology Association, St. Louis, MO (1993).
5. A. Thiruvengadam, in Erosion by Cavitation or Impingement, STP No. 408, 22-36, ASTM,
Philadelphia, PA (1966).
6. F. Heymann, in Characterization and Determination of Erosion Resistance, STP No. 474, 212-
244, ASTM, Philadelphia, PA (1969).
7. A. Conn, Proc. Fourth American Waterjet Conference, Berkeley, CA, Waterjet Technology As-
sociation, St. Louis, MO (1987).
8. H. Takagi and I. Hayashi, Intl. J. Approximate Reasoning, 5, No. 3, 191-212 (1991).
9. K. Babets, E.S. Geskin and B. Chaudhuri, Proc. 10th American Waterjet Conference, Houston,
TX, Waterjet Technology Association, St. Louis, MO (1999).
10. K. Babets, E.S. Geskin, Intl. J. Machining Sci. Technol., 4, No. 1, 81-101 (2000).
Surface Contamination and Cleaning, Vol. 1, pp. 173–191
Ed. K.L. Mittal
© VSP 2003
Abstract—The study is concerned with the development of effective technology for derusting of a
steel surface. We have investigated the surface derusting by high-speed waterjet and determined the
optimal operational conditions. This investigation involved topographical and metallographical
studies of the substrate surfaces and subsequent classification of the substrates with respect to the
degree of rust development. Then the rust was removed by a moving waterjet at various impact con-
ditions and the generated surfaces were examined. Soft computing techniques were used to select
the optimal conditions for rust removal. Due to the extremely chaotic and fuzzy nature of input in-
formation the advanced numerical procedure based on the Neural Network Driven Fuzzy Reasoning
was employed. As the result, the realistic procedure for steel derusting was found and a practical
technique for process design was suggested.
1. INTRODUCTION
ventional way of rust prevention is to coat the metal surface. But prior to the coat-
ing, a complete cleanliness of the surface must be assured to ensure that a surface
is free of any rust.
The conventional methods of rust removal involve acid cleaning or sandblast-
ing. These techniques, though proven to be effective, can be environmentally haz-
ardous. Some new derusting techniques, such as use of rust neutralizers converts
rust into a chemically neutral surface, leaving the surface ready for coating appli-
cation. However, the use of these neutralizers is limited to oil-based types only,
and the chemicals contained in these products can be detrimental to the worker’s
health.
Waterjet surface derusting constitutes rather an efficient way to clean steel sur-
faces. The following experimental study was concerned with optimization, or at
least improvement, of jet based derusting technology.
2. EXPERIMENTAL SETUP
( )
1
2
C D ⋅π ⋅ D 2 ⋅ 2⋅ P
ρw
(
Water Consumption m 3 / m 2 = )
4⋅ Rate of Area Cleaned
(2)
where ρw is the water density, CD is the discharge coefficient of the waterjet ori-
fice, whose diameter is D. In present work CD is taken to be 0.7 [3], P – waterjet
pressure.
176 K. Babets et al.
Then several nozzle passes with 25% overlapping were made at the same op-
erational conditions. The resulting derusted area was evaluated visually and pho-
tographed with Olympus photomicroscope with 12 x magnification in order to de-
termine the degree of cleaning.
3. EXPERIMENTAL RESULTS
Figure 3 shows a typical “poorly cleaned” surface. Such a surface is free from
lightly adherent mill scale, rust and other contamination, but some firmly adherent
rust remains on the surface. Thus the most representative surface samples were
classified as either belonging to one of these classes, or, not. The “fuzzy” mem-
berships in the two fuzzy classes for the remaining surfaces were determined us-
ing the artificial neural network assisted fuzzy classification method described in
the following section.
Figure 6. Chemical composition of metal surface prior to waterjet rust removal. Oxygen content is
at 900 count.
Figure 7. Chemical composition of metal surface after waterjet rust removal. Oxygen content is at
320 count.
180 K. Babets et al.
was significantly reduced after the water jet treatment. In order to evaluate the de-
gree of derusting, the chemical analysis was supplemented by the x-ray diffrac-
tion. The corresponding diffraction patterns of the metal surfaces before and after
waterjet cleaning are presented in Figs. 8 and 9, respectively. Roughly speaking
each peak in these figures corresponds to a chemical compound. The intensity of
Figure 8. X-ray diffraction analysis. Diffraction pattern of rusted metal surface prior to waterjet
treatment.
Figure 9. X-ray diffraction analysis. Diffraction pattern of metal surface after waterjet rust removal.
Experimental and numerical investigation 181
Table 1.
Results of the metal surface chemical analysis
a peak represents the relative amount of this chemical compound. The atomic
spacing values shown just above these peaks allow us to determine the type of the
chemical compound present on the surface. From Fig. 8 and Table 1 it is clear that
a rusted surface in addition to Fe contains significant amounts of Fe2O3, and
Fe3O4. After waterjet cleaning (Fig. 9) the three still remaining peaks represent
Fe, with significantly increased intensity levels. Most of the oxides are no longer
present in the figure, and intensity level of those still present is much lower than
that of Fe content. Moreover, due to low intensity levels these peaks most proba-
bly should be attributed to the noise. The wide base peak at angles 10-25 degrees
in Figs 8 and 9 is due to the presence of the holder clay used to attach the sample
in the holder. Thus, Fig. 9 constitutes a compelling proof of the efficiency of the
waterjet rust removal.
bilities, since at the chosen levels of the nozzle traverse rate and nozzle diameter
the removal of rust could still be performed. Between these two threshold values
of water pressure the clean strip width obtained in a single nozzle pass does not
vary linearly with increasing water pressure (Figure 10). The effectiveness of the
process increases with increasing water pressure (Figure 11).
Figure 10. Experimental clean strip width vs. water pressure for nozzle diameter 0.3556 mm at dif-
ferent nozzle traverse rates.
Figure 11. Effectiveness (rate of area cleaned) vs. water pressure for nozzle diameter 0.3556 mm at
different nozzle traverse rates.
Experimental and numerical investigation 183
Figure 12. Water consumption vs. water pressure for nozzle diameter 0.3556 mm at different nozzle
traverse rates.
Still the relationship between the effectiveness and water pressure can be ap-
proximated as linear at lower values of the traverse rate. At high nozzle traverse
rates and high nozzle diameters the relationship between the process effectiveness
and water pressure can no longer be considered as linear, but rather as a polyno-
mial. Figure 12 shows that for a large nozzle diameter there exists an extremum of
the water consumption as the pressure increases. This can be attributed to the fact
that the process effectiveness is not a linear function of water pressure (at least in
the considered range of process variables). At small nozzle diameters and for the
pressure range used in these experiments no extremum is seen, although it is rea-
sonable to expect that the extremum will appear at higher values of water pressure.
Figure 13. Rate of area cleaned vs. traverse rate for nozzle diameter 0.3556 mm and for different
water pressures.
Figure 14. Clean strip width vs. traverse rate for nozzle diameter 0.3556 mm and different water
pressures (69E – 69 MPa, 172E – 172 MPa, 241E – 241 MPa, 310E – 310 MPa).
consider the experimental results for water pressure 69 MPa, traverse rate 12700
mm/min and nozzle diameters 0.1778 mm (0.007 in) and 0.254 mm (0.01 in).
Here we notice that the area of the second nozzle is almost twice the area of the
first nozzle. The calculated process effectiveness is 0.39 m2/hour and 0.52
m2/hour, respectively, with specific water consumption 0.064 m3/m2 and 0.094
m3/m2. If we now take two small nozzles then the total effectiveness will be 0.78
m2/hour while the water consumption will stay at 0.064 m3/m2. Thus it appears
that it would be beneficial to use several small nozzles rather than a big one. This
result is important from practical point of view.
Experimental and numerical investigation 185
Figure 15. Water consumption vs. traverse rate for nozzle diameter 0.3556 mm and for different
water pressures.
Figure 16. Rate of area cleaned vs. nozzle diameter for water pressure 241 MPa at different nozzle
traverse rates.
Figure 17. Water consumption vs. nozzle diameter for water pressure 69 MPa and at different noz-
zle traverse rates.
186 K. Babets et al.
Figure 18. Computational approach. (I1, I2, and I3 are the input parameters (water pressure, traverse
rate, and nozzle diameter). NNmem is the neural network that decides the membership values (w1,
w2) in each class, of the above input parameters. NN1 and NN2 determine the outputs (rate of area
cleaned) y for each class.)
Figure 19. Model prediction results for water pressure 310 MPa at different nozzle traverse rates.
Experimental and numerical investigation 189
Figure 20. Model prediction results for water pressure 241 MPa at different nozzle diameters.
Figure 21. Model prediction results for nozzle diameter 0.3556 mm at different nozzle traverse
rates.
190 K. Babets et al.
Figure 22. Model prediction results for nozzle diameter 0.127 mm at different nozzle traverse rates.
5. CONCLUDINGS REMARKS
Acknowledgements
The surface examination was carried out at the Stevens Institute of Technology.
The valuable advice of Josef Karagotskiy of the Electronic Microscopy Center at
the Stevens Institute of Technology is gladly acknowledged. This work was par-
tially supported by NSF grant # DDM-9312980.
REFERENCES
1. B. Liu, B. Jia, D. Zhang, C. Wang, H. Li and H. Yao, Proc. 7th American Water Jet Conference,
Seattle, Washington, Waterjet Technology Association, St. Louis, MO (1993).
2. International Standard ISO 8501-1:1998, Third Edition (1999).
Experimental and numerical investigation 191
3. M. Hashish, Proc. 7th American Water Jet Conference, Seattle, Washington, Waterjet Technol-
ogy Association, St. Louis, MO (1993).
4. Z. Michalewicz, Genetic Algorithms + Data Structures = Evolution Programs, Third Edition,
Springer (1996).
5. H. Takagi and I. Hayashi, Intl. J. Approximate Reasoning, 5, No. 3, 191–212 (1991).
6. T. Ross, Fuzzy Logic With Engineering Applications, McGraw-Hill (1995).
7. B.D. Cullity, Elements of X-ray Diffraction, Addison-Wesley Publishing Company (1978).
8. H. Jun, Proc. 7th American Water Jet Conference, Seattle, Washington, Waterjet Technology
Association, St. Louis, MO (1993).
9. P. Singh, J. Munoz and W. Chen, Proc. of 11th International Symposium on Jet Cutting Tech-
nology, British Hydraulic Research Group, Dordrecht, The Netherlands (1992).
10. X. Shegxiong, H. Wangping and Z. Sheng, Proc. 7th American Water Jet Conference, Seattle,
Washington, Waterjet Technology Association, St. Louis, MO (1993).
11. C. Suryanarayana and M. Norton: X-ray Diffraction, a Practical Approach, Plenum Press, New
York (1998).
This page intentionally left blank
Surface Contamination and Cleaning, Vol. 1, pp. 193–212
Ed. K.L. Mittal
© VSP 2003
Abstract—The objective of this work was to acquire knowledge needed for the development and
deployment of manufacturing processes utilizing the enormous technological potential of water ice.
Material removal by blasting with ice media such as particles, pellets and slugs was investigated.
The ice media were accelerated by entertainment in an air stream. The ice-airjet (IAJ) can replace
sand blasting and the ice-waterjet (IWJ) can replace the abrasive waterjet (AWJ). The obvious ad-
vantage of the ice media is complete pollution prevention in course of materials treatment. With this
technique it is possible to eliminate both contamination of the substrate as well as generation of con-
taminated waste streams. In addition to the obvious environmental benefits, the use of ice media will
improve a number of key operational techniques, such as cleaning, decoating, polishing, deburring,
drilling, cutting, etc. The “just-in-time” production of ice media at minimal environmental cost con-
stitutes another advantage of ice-based technologies. Our previous studies have shown that the po-
tential applications of ice abrasives range from cutting of metals to etching of photo films and preci-
sion cleaning of electronic parts. However, the rate of the cleaning and machining operations
performed was insufficient. A key objective of this research was to improve ice blasting so that it
was not only feasible, but also technologically and economically efficient.
1. INTRODUCTION
There are a number of suggested air-ice based technologies. One of the firsts of
such technologies was a car washing machine, utilizing ice particles [1]. The
stream of the charged frozen particles controlled by a set of coils was directed at
surfaces to be cleaned [2]. Szijcs [3] proposed cleaning of sensitive surfaces by
the impact of a fine grade blast material and air. The atomization of the liquid in
the air stream and subsequent freezing of the generated fine droplets form the
blast material. The freezing is achieved by the addition of a refrigerant (N2, CO2,
Freon) into the stream in the mixing chamber or by the addition of the refrigerant
∗
To whom all correspondence should be addressed. Phone: (973)596-3338, Fax: (973)642-4282,
E-mail: geskin@njit.edu
194 D.V. Shishkin et al.
into the jet after the mixing chamber. The use of ice particles, which have a uni-
form grain size, for cleaning the surface and grooves of ferrite block, was reported
by Tomoji [4]. An ice blasting device using stored particles was suggested by Ha-
rima [5]. Vissisouk [6] proposed to use ice particles near melting temperature in
order to effectively remove the coating from the substrate. Mesher [7] suggested a
nozzle for enhancement of surface cleaning by ice blasting. Shinichi [8] suggested
cleaning inexpensively various surfaces by mixing ice particles, cold water and
air. Niechial [9] proposed an ice blasting cleaning system containing an ice
crusher, a separator and a blasting gun. Settles [10] suggested producing ice parti-
cles of a size range below 100 µm within the apparatus just prior to the nozzle.
Although the use of ice blasting is suggested by a number of inventors, the
practical application is much more limited. Herb and Vissisouk [11] report the use
of ice pellets for precision cleaning of zirconium alloys in the course of produc-
tion of bimetallic tubings. It was reported that ice blasting improved the quality of
the bimetal. The use of air-ice blasting for steel derusting was reported by Liu
[12]. The following operational conditions were maintained during blasting: air
pressure: 02-0.76 MPa, grain diameter: below 2.5 mm, ice temperature –50°C,
traverse rate 90 mm/min, and standoff distance 50 mm. Under these conditions
the rate of derusting ranged from 290 mm2/min at the air pressure of 0.2 MPa to
1110 mm2/min at the air pressure of 0.76 MPa. The quality of the cleaned surface
complied with ISO 8501-1 Sa 2.
The most important problem which actually impedes adoption of the ice-jet (IJ)
technology arises from the difficulties in the generation and handling of ice abra-
sives. Regular abrasives are stable at all practical ranges of operational conditions,
while ice particles can exist only at subzero temperature. Maintaining such a tem-
perature both within the nozzle and the jet is an extremely difficult task. Ice parti-
cles tend to pack and clog the supply lines. The adhesion between the particles in-
creases dramatically as the temperature approaches 0°C. Thus prior to entrance in
the nozzle, ice particles should be maintained at a low temperature. These and
some other problems prevent adoption of IWJ. In order to assure the acceptance
of IWJ by the industry, it is necessary to develop a practical technology for for-
mation of ice-water slurry.
The experimental prototype depicted in Figure 1 was selected for further experi-
ments. The system consisted of the following functionally separated blocks:
– ice making block which includes the evaporator, auger, auger driver, sealing
and liquid nitrogen cooling apparatus;
– ice unloading mechanism
– nozzle block which includes parallel nozzles and focusing device.
Practical applications of icejet technology in surface processing 195
Figure 1. (a) Schematic of auger type IJ system where: 1 – evaporator, 2 – refrigerant coils, 3 – in-
sulation, 4 – auger, 5 – ice reloading device, 6 – air gun, 7 – air supply port, 8 – water supply port, 9
– cooling medium port, 10 – gauges, A – air flow rate valve, B – water flow rate valve, C – cooling
medium valve, D – data acquisition card, and (b) picture of the ice reloading device with nozzle
block.
In our experiments, water entered the heat exchanger via a special port. As it
moved along the rotating auger water solidified and an ice plug was formed. So-
lidified ice plug moved forward along the auger helical ways. Decomposition of
this plug formed ice powder. The heat exchanger and the auger of the icemaker
constituted a modified commercial icemaker of Hoshizaki America Inc., Peach-
tree City, GA. The design of these parts will be changed in the next generation of
the device. The cooling was carried out by the refrigerant Galden HT-55 supplied
by the TurboJet refrigeration apparatus or by liquid nitrogen stored in a tank. We
replaced Hoshizaki auger driver by a more powerful device in order to prevent
jamming of the ice. The rotation momentum of the auger 4 was provided via a
gearbox with gear-ratio 1:100. However, the selected driver operated at a constant
speed of 100 rpm. Water flow rate was precisely controlled by a special valve
(Figure 1 (a)).
At the outlet of the heat exchanger the powder was entrained by the unloading
mechanism which directed it to the nozzle block (5). The nozzle block consisted
of two air guns (6) and a special focusing device. Three different sizes of the noz-
zles were used; however, in all cases the nozzle-to-focusing tube ratio was 1:2.
An unloading mechanism delivered ice particles via flexible plastic tubes to the
abrasive port of the air gun. In the gun the air supplied into the insulated nozzle
block at the room temperature accelerated the particles. The IAJ was formed and
directed to the substrate surface 3.
196 D.V. Shishkin et al.
3. EXPERIMENTAL PROCEDURE
In the course of IJ cleaning, the air pressure was maintained at 0.544 MPa (80
psi), the nozzle diameter was 2.5 mm, and the nozzle focusing tube diameter was
5 mm. The properties of ice abrasive medium were the following: ice temperature
was in the range from –20°C to –70°C, granulometric composition of ice powder
ranged from 0.3 mm to 7 mm, and ice flow rate was 20 g/min to 150 g/min. Ice-
airjet (IAJ) was used for cleaning various sensitive surfaces covered by moder-
ately adhering deposits. The sensitive elements of the electronic boards were cov-
ered by a conductive copper paste and cleaned by the IAJ. When assembled, these
components performed normally and the normal operational modes of the devices
were demonstrated. The feasibility of using IAJ technology as a blasting medium
for cleaning highly sensitive surfaces was shown. Another experiment involved
depainting of various substrates, including mirror-like surfaces and the surfaces of
soft substrates. A complete removal of the paint and the absence of surface dam-
age were demonstrated. The generated surfaces were inspected visually.
A number of experiments involved the use of the ice abrasive in waterjet (WJ)
cutting applications. The experimental procedure was carried out with the follow-
ing parameters: the water pressure was 306.1 MPa, the diameter of the sapphire
nozzle was 0.178 mm, average standoff distance was maintained in range 7 mm–
10 mm and the traverse rate was 1.06 mm/s. Various metals and composite mate-
rials were cleaned by IWJ. The depth and cutting rate were substantially lower
than in the case with conventional abrasive media. However, the IWJ produced a
very narrow cutting kerf compared with AWJ and had a superior cutting ability
over pure WJ. The main obstacle during ice particles entrainment in the nozzle
abrasive port was their agglomeration at the port entrance and their disintegration
in the mixing chamber due to intensive melting. This technology is still under de-
velopment and requires further investigation.
4. EXPERIMENTAL RESULTS
A series of experiments were carried out in order to evaluate the potential of the
application of IAJ for surface processing. The description of these experiments is
given below.
Figure 2. (a) photograph of the electronics board of a TV set. Notice the heavy layers of dust and
dirt on the electric and electronic components of board, (b) photograph of an assembled TV set. The
contaminated board of TV set is shown (a). After cleaning TV set worked normally.
198 D.V. Shishkin et al.
small amount of the grease remaining at hidden pockets will disrupt the TV set
performance. It is clear that the jet was able to remove soil from all the difficult-
to-reach pockets.
Another experiment involved decontamination of computer boards. Various
devices (PC, electronic watches, computer games, etc) were disassembled. The
boards were covered by a mixture of lithium grease and then decontaminated by
IAJ. Clean boards were reassembled and tested. All devices worked perfectly.
Some of the devices above were used for several tests. No deviation in the com-
puter operation was noticed. The boards above were populated by a large number
of rather fragile components such as chips, connectors, etc. Any damage to any of
these components, as well as any presence of grease on the board will disable the
device. In all performed experiments the deposit was removed completely and no
damage was induced to the board components. The examples of the boards decon-
taminated in the course of these experiments are shown in Figs. 3 (a) and 3 (b).
Figure 3. (a) photograph of the board of an electronic game containing electric conduits, microchip
and electronic matrix. The board was covered by a mixture of lithium grease and copper powder.
Notice the cross contamination of electric conduits of the board, (b) photograph of the assembled
electronic game after IAJ cleaning. The electronic game performed normally.
200 D.V. Shishkin et al.
Figure 4. (a) photograph of the CD-ROM covered by Rust-Oleum gloss protective enamel. The
paint was partially removed from the CD ROM surface. No surface damage was observed in the
course of IAJ cleaning, and (b) photograph of the CD-ROM partially cleaned using IAJ technique.
Notice that layers of both paint and emulsion were removed. No surface damage was observed in
the course of IAJ processing.
Practical applications of icejet technology in surface processing 201
Figure 5. (a) photograph of the polished steel surface. The polished steel surface was contaminated
by Rust-Oleum gloss protective enamel. The paint was partially removed from the polished surface.
No surface damage was observed in the course of IAJ cleaning, so the feasibility of the precision
cleaning of polished surfaces was demonstrated. (b) photograph of the hand-painted china plate. The
plate was covered by Rust-Oleum gloss protective enamel. Part of the deposited paint was removed
by ice etching. No modification of the original surface was noticed, and thus the feasibility of IAJ
etching of sensitive surfaces was demonstrated.
202 D.V. Shishkin et al.
Figure 6. (a) photograph of an egg. The egg surface was painted by Rust-Oleum gloss protective
enamel. After this the egg was partially decontaminated by IAJ technique. No damage to the egg
surface or penetration of the ice particles through the eggshell was noticed, so the feasibility of de-
contamination of highly unstable and brittle surfaces was demonstrated. (b) photograph of the cover
of a pharmaceutical reactor contaminated by the lithium grease. Then the grease was partially re-
moved from the surface of the cover by IAJ technique. No damage to the glass in the course of IAJ
cleaning was noticed.
Practical applications of icejet technology in surface processing 203
Figure 7. (a) photograph of the LC display of a calculator containing electronic matrix and LCD
conduits. The display was contaminated by Rust-Oleum gloss protective enamel. Then all elements
of the LC display were decontaminated by IAJ technique. On assembly of the calculator the LC dis-
play performed normally. (b) photograph of a magnification lens. The lens was contaminated by
lithium grease. The grease was partially removed from the lens surface, and no damage to the lens
surface was observed.
204 D.V. Shishkin et al.
Figure 8. (a) photograph of a PVC tube contaminated by Rust-Oleum gloss protective enamel. The
tube was partially decontaminated by IAJ technique. No damage to the tube surface in the course of
IAJ cleaning was noticed, and (b) photograph of a cotton fabric. The fabric was contaminated by
Rust-Oleum gloss protective enamel. Then the paint was partially removed from fabric surface, and
thus the feasibility of the use of ice particles for decontamination of fabrics was demonstrated.
Practical applications of icejet technology in surface processing 205
Figure 9. (a) photograph of an electrical solenoid valve with connectors contaminated by Rust-
Oleum gloss protective enamel. The contacts of solenoid valve were cleaned by IAJ technique. Af-
ter cleaning the solenoid valve was connected to an electrical supply source and performed nor-
mally. This experiment demonstrated the feasibility of using IAJ technique for decontamination and
restoration of contacts of different electronic devices. (b) photograph of a DC motor. DC motor was
disassembled and all elements were covered by a mixture of lithium grease and copper powder. DC
motor was cleaned using IJ technique and the assembled DC motor performed normally.
206 D.V. Shishkin et al.
remained on the surface. It was not possible to remove it using mechanical means.
The layer was removed completely by the IAJ (Fig. 10 (a)). A metal wall was
covered by an oil paint and then was subjected to abrasive-airjet (AAJ) (the car-
rier medium was sodium bicarbonate). Then the same procedure was carried out
using IAJ. The initial state of the graffiti covered metal surface is shown in Fig.
10 (b). The graffiti removal by conventional and ice based technologies are shown
in Figs. 11 (a) and 11 (b). Another experiment involved removal of the residual
highly adhesive Weldbond glue from plastic and rubber jointed surfaces (Fig. 12
(a) and 12 (b)). Average process duration in all these experiments was around two
minutes. The heavily contaminated machine part with grease and dust was decon-
taminated by IAJ too (Fig. 13 (a) and 13 (b)). No damage to the underlying
painted surface was noticed.
5. CONCLUDING REMARKS
Although the ice-water jet constitutes an effective material removal tool, it is nec-
essary to improve conditions of the jet formation in order to assure its adoption in
practice. However, the ice-air jet is suitable for immediate application. It can be
used for decontamination of very demanding and complex surfaces as well for
such manufacturing applications as etching. Simplicity and complete absence of
environmental damage constitute the main advantages of this process. A further
development of IAJ surface cleaning technology will involve improvement of the
control of ice particle properties and enhancement of the methods for the delivery
of ice particles to the substrate. This enhancement will enable us to modify mate-
rial polishing, surface cleaning, and, perhaps, grinding.
Practical applications of icejet technology in surface processing 207
Figure 10. (a) photograph of the aluminum surface contaminated by a thick layer of tar. The bulk of
the tar was removed by WJ and knife scrubbing. The highly adherent thin layer was removed by ice
etching. No damage to the metal surface was noticed and (b) graffiti covered painted metal surface.
The oil paint is highly adherent.
208 D.V. Shishkin et al.
Figure 11. (a) Graffiti was removed with conventional AAJ. Notice discoloration occurred in the
treated region and (b) surface was decontaminated by the IAJ. No damage to the underlying paint
layer occurred.
Practical applications of icejet technology in surface processing 209
Figure 12. (a) The Weldbond glue was used to create a joint between plastic and rubber surfaces.
Notice the highly adhesive character of the glue, and (b) the glue residue was removed by IAJ clean-
ing. No surface damage was noticed.
210 D.V. Shishkin et al.
Figure 13. (a) picture of the highly contaminated machine part with grease and dust, and (b) part
was decontaminated by IAJ cleaning. No damage was seen on the underlying painted surface.
Practical applications of icejet technology in surface processing 211
Figure 14. (a) photograph of a strip of a photo film. The photo emulsion was partially removed
from the film surface. No surface damage was observed in the course of IAJ cleaning and thus the
feasibility of complete and selective emulsion removal from thin photo film was demonstrated, and
(b) photographs of cutting of aluminum strip of thickness 3.1 mm (X65). Notice the reduced width
of the kerf generated by IWJ cutting. Also note substrate surface erosion in the vicinity of IWJ gen-
erated kerf.
212 D.V. Shishkin et al.
Figure 15. Photographs of cutting of titanium sample of thickness 0.7 mm (X65). Notice the re-
duced width of kerf in the course of IWJ cutting. Also note the intensive erosion of the substrate sur-
face in the vicinity of IWJ generated kerf.
REFERENCES
TERRY MUNSON∗
Contamination Studies Laboratory (CSL), 201 East Defenbaugh, Kokomo, Indiana 46902
Abstract—This paper explores the correlation between the cleanliness levels on electronic assem-
blies and their electrical performance. It documents an experiment conducted to explore this correla-
tion. Cleanliness was measured using Ion Chromatography (IC), and electrical performance was
measured using Surface Insulation Resistance (SIR) testing under elevated humidity and tempera-
ture conditions. Furthermore, this paper discusses electronic assembly cleanliness issues, and a new
cleanliness assessment approach for determining cleanliness levels required for the typical flux
technology of today. We conclude – from the samples examined, and based on our past 10-years of
experience analyzing similar experiments – that circuit board field performance (good or poor) is
strongly correlated to the specific amount and type of invisible and visible residues between pads
and holes in all areas of active circuitry.
Keywords: Ionic contaminants; residues; electrochemical metal migration; ion chromatography; sur-
face insulation resistance; electrical performance testing; cleanliness levels.
1. INTRODUCTION
Since the 1987 Clean Air Act, when Government legislation forced the electron-
ics industry to stop using the ozone depleting chemical Freon® as a cleaner, the
industry has been required to find new chemicals and processes. The industry be-
lieved at the time that Freon® solvent cleaning techniques effectively removed all
surface contaminants. Subsequent research has shown that the old rosin based
fluxes used in the manufacturing processes actually sealed in contaminants,
whereas the new fluxes left contaminants exposed to react with humidity in the
end-user environments. These inherent process residues must be removed to
achieve product reliability.
The change from Freon® cleaning initiated many changes in the manufacturing
processes. There was no direct substitute cleaning chemistry. The new solvents
could not be used for cleaning using many of the old processing chemistries. The
change was an opportunity for many manufacturers to change their overall proc-
esses. As the changes were made to new processing materials and cleaning sol-
∗
Phone: 765-457-8095, Fax: 765-457-9033, E-mail: residuguru@aol.com
214 T. Munson
4. New flux chemistry residues on the newest metal composition result in new
interaction issues that have effects that are inadequately understood.
Figure 1. Dendrite Growth at (a) 5 seconds, (b) 10 seconds, (c) 15 seconds, (d) 20 seconds.
Correlating cleanliness to electrical performance 217
2. EXPERIMENTAL DETAILS
2.1. Materials
To perform the experiment, 30 test coupon printed circuit boards were fabricated.
The Umpire test board (shown in Figure 2) was selected because it uses a mixed
technology assembly process – surface mount technology (SMT) and plated-
through-hole technology (PTH). It can also be used to assess different component
areas, such as the pad-to-pad or lead-to-lead, and assess entrapment effects on a
variety of components (BGA, LCC, QFP, DIP, PGA, 1206 chip, 0805 chip), and
comb patterns (B24, Bellcore).
batch until they reached the final cleaning process. Cleanliness Level 1 (L1) used
a saponifier in the cleaning solution, known to be the cleanest method. Level 2
(L2) used deionized water cleaning, known to be a marginal cleaner. Level 3 (L3)
used tap water cleaning, known to be the dirtiest cleaning method (standard rinse
water process at most fabricators).
These two tests were performed on each sample: 1) Ion Chromatography and 2)
Surface Insulation Resistance with visual inspection performed per IPC 610 pro-
tocol.
ganic Acid, Sodium, Calcium, Potassium, and NH4–. Before testing, the IC system
is calibrated to NIST (National Institute of Standards and Technology) traceable
standards. The level of sensitivity is 0.01 part per million.
CSL is able to test specific areas of boards by combining IC with an effective
extraction procedure that conforms to IPC protocol TM 650 2.3.28. Each sample
was IC tested in four location areas: 1) a top-side surface mount technology
(SMT) area, 2) a bottom-side wave solder area, 3) a rework area, and 4) a peelable
solder mask area.
3. RESULTS
The experimental results are grouped into the four test area groups: 1) top-side SMT
area, 2) bottom side wave solder area, 3) rework area, and 4) solder-mask area.
Table 1.
Top-side SMT area results for all three levels of cleanliness (L1-L3)
Table 2.
Bottom side wave solder area results for all three levels of cleanliness (L1-L3)
2. L2 boards failed electrically due to the high amount of bare board contamina-
tion (chloride HASL flux) with the exception of the PGA and DIP area. The
failure was due to the insulative flux effects creating a barrier between the
moisture and the leads.
3. L3 boards failed electrically due to the high amount of bare board contamina-
tion (chloride HASL flux).
Table 3.
Rework area results for all three levels of cleanliness (L1-L3)
Table 4.
Temporary Soldermask area results for all three levels of cleanliness (L1-L3)
4. DISCUSSION
This case study shows that the low-residue assembly process used works well in
the areas of SMT and Wave Solder with clean (level 1) bare boards. However, the
secondary processing (rework and temporary soldermask) areas, even with clean
bare boards, showed high levels of electrical failure due to excess partially or un-
reacted flux.
The data presented here are only a small snapshot of the information gathered
through this testing. This assessment is not intended to replace actual product
validation or environmental testing. This case study is intended to assess the proc-
ess effects as a baseline and to determine if process changes are good or bad in
regards to the electrical effect. As an assessment tool, this will help establish the
actual level of cleanliness required in building reliable hardware.
A cleanliness assessment approach such as this will allow electronic assemblers
the opportunity to document the effects of the process and materials as a baseline
from which to make improvements. This case study and others we have per-
formed over the last ten years have shown us that the residues from fabrication do
have a large effect on electrical performance. In addition, the residues from sec-
ondary assembly processes have just as much effect on the field performance of
the product.
Although ionic and organic IC analyses of component areas on electronic as-
semblies detect a specific amount of flux and processing residues, determining
whether the levels are good or bad is based on results from an electrical SIR
evaluation of the same areas. Over the last 10 years, CSL has developed a large
cleanliness level database, and has used this data to determine general levels of
cleanliness, but we are constantly adjusting acceptability levels because of chang-
ing component packages, increasing operating frequency designs, lower voltage
designs, and changing processing materials. The result of all the changes in the
industry is electronic equipment that is more sensitive to process residues, and has
less long-term reliability, especially in harsh environments, unless detrimental
process residues are identified, measured, and reduced.
224 T. Munson
5. CONCLUSION
Acknowledgments
We would like to acknowledge Dr. K.L. Mittal for constructive feedback and
support. We would also like to acknowledge Diversified Systems Inc. of Indian-
apolis, Indiana for their manufacturing support during the experiment.
REFERENCES
1. M.G. Fontana, Corrosion Engineering, 3rd edition, McGraw-Hill Book company, New York
(1986).
2. H. Small, Ion Chromatography, 1st edition, Plenum Press, New York (1989).
3. IPC Committee, Post Solder No-Clean Handbook (IPC-SC-62A), IPC, Chicago, IL (1999).
4. IPC TM 650, Test Methods Handbook, “Ion Chromatography,” IPC, Chicago, IL (1995).
5. IPC TM 650, Test Methods Handbook, “Surface Insulation Resistance”, IPC, Chicago, IL
(1995).
Surface Contamination and Cleaning, Vol. 1, pp. 225–239
Ed. K.L. Mittal
© VSP 2003
Abstract—Cleaning critical and high reliability printed wiring assemblies (PWAs) continues to be
important to ensure high reliability performance and to prevent premature failure. The necessary
steps to qualifying both a cleaning system and an appropriate chemistry for cleaning such PWAs are
set forth. This paper addresses a centrifugal cleaning system used in conjunction with a water-based
cleaning medium to achieve optimally low levels of contaminants on PWAs. Ionograph data, ion
chromatography profiling, residual rosin determination, and outgassing data are presented demon-
strating the effectiveness of the centrifugal cleaning system and the aqueous cleaning agent for
space flight printed wiring assemblies. It is concluded that a centrifugal cleaning system coupled
with a suitable aqueous chemistry can be successfully employed to clean high reliability PWAs.
Keywords: Aqueous cleaning (AC); conformal coating; ionic contamination testing (ICT); green-
house warming potential (GWP); multilayer board (MLB); ozone depletion potential (ODP); printed
wiring assembly (PWA); printed wiring board (PWB); rosin mildly activated (RMA); semi-aqueous
cleaning (SAC); surface mount technology (SMT); volatile organic compound (VOC).
1. INTRODUCTION
During the last decade, the challenges of cleaning printed wiring assemblies
(PWAs) have grown. Today printed wiring boards have grown more complex to
meet the continuing challenges posed by the increasing uses of microdevices, fine
pitch packages, and array devices, such as ball grid arrays, microball grid arrays,
and flip chips. Multilayer boards with a large layer count and narrow trace widths
and spaces are commonplace. The ball grid arrays, microball grid arrays, and
other small devices generally have a large number of inputs/outputs (I/Os), small
standoffs, and small pitches. The small standoff and small pitch, coupled with the
complex circuitry needed to route such components, makes cleaning an ever more
critical operation. High reliability PWAs cannot tolerate contaminants since their
∗
To whom all correspondence should be addressed. Phone: (818) 354-1320, Fax: (818) 393-
5456, E-mail: john.k.bonner@jpl.nasa.gov
226 J.K. Bonner and A. Mehta
presence can potentially degrade the board, thus compromising the intended mis-
sion.
Cleaning for high performance PWAs is normally performed as a minimum at
the following stages:
(1) At the bare printed wiring board (PWB) stage prior to the application of sol-
der mask;
(2) Immediately after the PWB + components are soldered to form the PWA;
(3) Immediately prior to the application of conformal coating.
If the PWAs are properly stored, the second and third operations are sometimes
combined. A number of contaminants are potentially introduced on the PWB sur-
face. These contaminants can be classified into three broad categories:
(1) Particulates;
(2) Ionic residues;
(3) Non-ionic residues, chiefly organic in nature [1-4].
To ensure the reliability of a PWA, cleaning is mandatory to remove these con-
taminants after the soldering operation and also directly prior to the application of
a conformal coating. In addition to cleaning, some sort of cleanliness verification
method, such as ionic contamination testing (ICT), is normally employed. ICT
can be used to ascertain that a certain level of cleanliness has been achieved. In-
dustry-recognized devices, such as an Ionograph® or Omega-Meter®, have com-
monly been used for this purpose. In addition, determining the amount of residual
rosin (assuming that a rosin-based flux or paste was used) is often done. Another
useful technique is to remove some of the components and examine for flux resi-
dues both visually and by use of a microscope.
The last decade has also seen the dramatic decrease and continuing disuse of
ozone-depleting solvents. The common chlorofluorocarbon solvents, such as
Freon® TMS, have been discontinued, and many PWA assemblers have switched
to more environmentally-friendly cleaning agents, such as a wide variety of semi-
aqueous and aqueous-based materials. To enhance the performance of such mate-
rials, the proper equipment selection plays a critical role.
2. BACKGROUND
Ten years ago the Electronic Packaging and Fabrication section at the Jet Propul-
sion Laboratory (JPL) established a dedicated facility for producing very low vol-
ume but high performance surface mount technology (SMT) assemblies known as
the SMT Laboratory. This laboratory has successfully assembled SMT PWAs for
such important JPL programs as these:
● Cassini;
● ChuG Microgyro;
● Caltech-ACE;
Qualifying a cleaning system for space flight printed wiring assemblies 227
● 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 and 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
standoffs 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 IPA 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.
The new cleaning system consists of the following equipment and materials. A
brief description of its operation is also given below.
Qualifying a cleaning system for space flight printed wiring assemblies 229
5.1. Equipment
The following equipment is required:
● Centrifugal cleaning system;
● Vacuum oven;
● Refractometer.
5.1.1. Equipment description
The equipment consists of an enclosed stainless steel cylindrical process chamber
with a series of spray nozzles located vertically. A robotic arm containing a fix-
ture holds the PWA and moves it in and out of the chamber vertically. During the
cleaning cycle, the PWA is lowered into the process chamber until it is com-
pletely sealed from ambient. (See Figure 1).
5.2. Materials
The following materials are used in the centrifugal cleaning system:
● Aqueous system containing the mixture of some alkoxypropanols with one to
three alkoxy units (ether linkages). – 20% by volume (see Section 4 above for
further details);
● Corrosion inhibitor – 1% by volume;
● Defoamer – 0.1% by volume;
230 J.K. Bonner and A. Mehta
In order to investigate the new cleaning system, a comparison was made between
it and the initial cleaning system. The following objectivcs were pertinent to this
investigation.
6.1. Objectives
The two chief objectives were:
● Investigate the new centrifugal cleaner using the aqueous cleaning solution
for flight PWAs;
● Establish the optimal cleaning cycle for the new equipment.
Qualifying a cleaning system for space flight printed wiring assemblies 231
To be able to recommend the new centrifugal cleaner using the aqueous clean-
ing solution, the procedure used was to compare the cleaning data of the older
cleaning process using the semi-aqueous (SA) solution with the centrifugal
cleaner using the aqueous cleaning solution.
● Outgassing per ASTM E595, “Standard Test Method for Total Mass Loss and
Collected Volatile Condensable Materials from Outgassing in a Vacuum En-
vironment.”
Either ionic contamination testing was performed using the Ionograph® 500 or
total LVR was determined for a given sample, but not both, i.e., the tests are mu-
tually exclusive of each other. This is because in the process of conducting the
ionic contamination test, the PWA is cleaned, thus rendering it unfit for further
cleanliness testing. This is indicated in the results (Tables 1-3) using the symbol
N/A (not applicable) in one or the other column.
However, the total low volatile residue (LVR) analysis and the residual chlo-
ride analysis (Cl–) are not mutually exclusive, and both examinations can be per-
formed on the same sample. They are not mutually exclusive because first an ex-
tract is made using D.I. water to remove the very soluble anions present (Cl–, F–,
Br–), and then an extract is made using the Freon TF/IPA to remove the rosin
residue which is insoluble in water.
The results for the new centrifugal cleaning system using the aqueous chemis-
try are reported in Tables 1-3. Table 4 gives ionic contamination levels using the
older SA cleaning system.
The outgassing test method per ASTM E595 determines the volatile content of
materials when exposed to a vacuum environment. Two parameters must be
measured: Total mass loss (TML) and collected volatile condensable material
(CVCM). In addition, since polyimide printed wiring board material can absorb
moisture, an additional parameter was determined, namely, the amount of water
vapor regained (WVR). The results for the TML, CVCM and TML-WVR values
are reported in Table 5.
6.2.4. Acceptable cleanliness levels
Per JPL D-8208, “Spacecraft Design and Fabrication Requirements for Electronic
Packaging and Cabling”, the ionic contamination level as determined by the Io-
nograph® must not exceed 10 micrograms per square inch (10 µg/in2 = 1.5
µg/cm2). If it does, the entire lot of PWAs must be recleaned and one PWA per lot
retested until this ionic cleanliness level is achieved. No acceptable standard has
been agreed upon for the amount of residual rosin; however, a limit of no more
than 150 micrograms per square inch (150 µg/in2 = 22.5 µg/cm2) seems appropri-
ate. In the case of ionic profiling using ion chromatography (IC), no acceptable
standards have been agreed upon for the amount of individual ionic species, but
one would expect that the sum of the various ionic species would be less than the
limit obtained from ionic contamination testing. Based on the outgassing determi-
nation per ASTM E595, the acceptable level for the TML must be no more than
1.00%, and the CVCM must be no more than 0.10%. If the WVR is determined,
then TML-WRV is also reported.
234 J.K. Bonner and A. Mehta
Table 1.
Cleanliness data from test run #1 (new cleaning system)
Table 2.
Cleanliness data from test run #2 (new cleaning system)
Table 3.
Cleanliness data from test run #3 (new cleaning system)
The SA cleaning system (old cleaning system) parameters were: Wash time = 5.0
min. (with saponifier); rinse time = 10.0 min. (D.I. H2O); 5.0 min. (D.I. H2O/IPA
mixture); dry time = 5.0 min. The results are presented in Table 4.
Table 4.
Cleanliness data from test run #4 (old cleaning system)
Table 5.
Outgassing data from test run #3
7. SUMMARY OF RESULTS
square inch, which is the lowest level of the flight hardware determination
standard MIL-STD-1246C Level A.
● PWAs cleaned with 3.0 minutes wash and 2.0 minutes rinse had average ionic
cleanliness level of 0.93 micrograms per square inch, far below the JPL
maximum acceptable ionic cleanliness level of 10 micrograms per square
inch. This result is much lower than that obtained by the older cleaning sys-
tem. (See Tables 2 and 4). The results, however, are not optimal.
● PWAs cleaned with 6.0 minutes wash and 6.0 minutes rinse had average ionic
cleanliness level of 1.97 micrograms per square inch, still far below the JPL
maximum acceptable ionic cleanliness level of 10 micrograms per square
inch. These somewhat higher results may be due to the inadvertant contami-
nation by handling of some of the boards. Since the ionic cleanliness level is
still significantly lower than 10 micrograms per square inch, this result does
not vitiate the overall performance of the new cleaning system.
● The anion profile analysis performed with ion chromatography showed ex-
ceedingly low levels of anion species, thus indicating very low levels of re-
maining contamination.
● One PWA after cleaning had its components removed to examine for flux
residues. Both visual and 10x magnification were used to detect residues.
Nothing was noted.
● The outgassing data for the boards cleaned using the new centrifugal cleaning
system/aqueous chemistry indicates that the total mass loss (TML) is much
less than 1.00% and the collected volatile condensable material (CVCM) is
much less than 0.10%.
• The optimal cleaning cycle suggested by the data is:
• Wash solution temperature 50°C
• Rinse solution temperature 50°C
• Dry air temperature 200°C
• Wash cycle rotational speed 150 RPM
• Wash time 5 min.
• Rinse time 2.5 min.
• Dry time 2.5 min.
Note on ESD
There was some concern that during the hot air drying stage there might be an
ESD (electrostatic discharge) problem. A medical device manufacturer that pur-
chased a Speedline Technologies ACCEL MicrocelTM Centrifugal Cleaning Sys-
tem was concerned about this and performed a thorough investigation. They
found no ESD problem. In addition, at JPL a normal cleaning cycle was run and
an ESD meter was used to see if there was any ESD build-up. No ESD was de-
Qualifying a cleaning system for space flight printed wiring assemblies 239
8. CONCLUSION
The centrifugal cleaner using the new aqueous cleaning solution based on long-
chain alcohols shows a marked improvement in cleanliness of PWAs over the
previous two-stage batch semi-aqueous (SA) cleaning system using a terpene-
based SA material and water in machine #1 for cleaning and saponifier, isopropyl
alcohol (IPA), and D.I. water in machine #2 for rinsing. The centrifugal cleaner
using the new aqueous cleaning solution not only cleans at a higher degree of
cleanliness level compared to the older SA cleaning system, but also it is cost ef-
fective to use. The total cycle time is about 50% less than the older SA cleaning
system. Also, it uses single chemical (the long-chain alcohol/aqueous solution)
with very small amount of additives compared to three chemicals used by the
older SA cleaning system. The use of hazardous isopropyl alcohol is also elimi-
nated.
Acknowledgements
The research to qualify this new cleaning system was performed at the Surface
Mount Technology Laboratory at the Jet Propulsion Laboratory, California Insti-
tute of Technology, under a contract with the National Aeronautics and Space
Administration. The authors wish especially to thank Mr. Charles J. Bodie and
Mr. Amin Mottiwala for their support and encouragement.
REFERENCES
1. J.K. Bonner, in Cleaning Printed Wiring Assemblies in Today’s Environment, L. Hymes (Ed.),
pp. 65-119, Van Nostrand Reinhold, New York (1991).
2. L. Hymes (Ed.), Cleaning Printed Wiring Assemblies in Today’s Environment. Van Nostrand
Reinhold, New York (1991).
3. C.J. Tautscher, Contamination Effects on Electronic Products. Marcel Dekker, New York
(1991).
4. C.J. Tautscher, The Contamination of Printed Wiring Boards and Assemblies. Omega Scientific
Services, Bothell, WA (1976).
5. F. Cala and A.E. Winston, Handbook of Aqueous Cleaning Technology for Electronic Assem-
blies. Electrochemical Publications, Isle of Man (British Isles) (1996).
6. J.B. Durkee, The Parts Cleaning Handbook without CFCs: How to Manage The Change. Han-
ser-Gardner, Cincinnati, OH (1994).
7. M.C. McLaughlin and A.S. Zisman, The Aqueous Cleaning Handbook: A Guide to Critical-
Cleaning Procedures, Techniques and Validation. The Morris-Lee Publishing Group, Rose-
mont, NJ (1998).
This page intentionally left blank
Surface Contamination and Cleaning, Vol. 1, pp. 241–247
Ed. K.L. Mittal
© VSP 2003
Abstract—The RCA clean is widely used in the semiconductor industry for many wet-chemical
cleaning processes. The RCA clean consists of a particle removal step, the Standard Clean 1 or SC-1
and metallic impurity removal step, the Standard Clean 2 or SC-2 step. In this work we have inves-
tigated the addition of chelating agents in SC-1 solutions to prevent metallic deposition during the
SC-1 step as well as to remove metallic contamination. We also have studied the effect of surfac-
tants in such solutions on sub-micrometer particle removal. This leads to the development of a very
fast and efficient single step RCA replacement clean. The use of a single step cleaning strategy in a
single wafer mode dramatically reduces the cycle time of cleaning.
Keywords: RCA clean; silicon wafer cleaning; chelating agent; modified SC-1.
1. INTRODUCTION
SC-1 cleaning is widely used in the semiconductor industry during various wet-
chemical cleaning processes due to its outstanding particle removal efficiency.
Although SC-1, a mixture of NH4OH/H2O2/H2O, is an efficient particle removal
solution, it inherently allows some metallic impurities in solution to deposit on the
wafer surface [1]. For this reason a conventional SC-1 is typically followed by
SC-2, a mixture of HCl/H2O2/H2O, which exhibits excellent metallic impurity re-
moval efficiency [2]. This sequence of SC-1 and SC-2 is known as the RCA clean
and has been in use for over 30 years.
The most obvious advantage of adding an appropriate chelating agent to SC-1 is
to prevent the deposition of metallic impurities during the particle removal step and
thus to eliminate the need for a follow-up metallic impurity removal step. Not only
does this reduce the number of chemical cleaning steps required, saving money and
time, it also avoids the adverse effect of particle re-deposition during typical metal-
lic impurity removal steps, such as SC-2 or an HF dip. Furthermore, an appropri-
ately chelate enhanced SC-1 solution can potentially remove metallic contamination
∗
To whom all correspondence should be addressed. Phone: (408) 584-0957, Fax: (408) 584-1132,
E-mail: christopher_beaudry@amat.com
242 C. Beaudry and S. Verhaverbeke
even more efficiently than SC-2, and its ability to bind free metal ions in solution
will potentially isolate process excursions from affecting process yield.
To understand the effect of adding a chelating agent to an SC-1 solution, it is
important to study the interaction of metallic impurities in solution and the sub-
strate in that solution. In this case, the substrate of interest is silicon. In aqueous
solutions, such as SC-1, a silicon wafer surface is hydroxyl terminated (silanol
groups: -Si-OH). The interaction between the metal ions in solution and the si-
lanol surface groups can be described by the following equation:
-Si-O-H + Mx+ ó -Si-O-M(x-1)+ + H+ (1)
x+
where M is the metallic ion. From equation (1), one can see that there are two
ways to reduce metallic ions from depositing on the wafer surface. The first way
is to increase the concentration of H+, shifting the reaction to the left. Unfortu-
nately, acidifying SC-1 will degrade particle removal effectiveness of the solution
(the high pH provides electrostatic repulsive forces while lowering the pH may
result in attractive forces between particles and the substrate). The second way to
prevent or reduce metallic ion deposition is to decrease the free metal ion concen-
tration in solution. For many years, suppliers have supported such an approach by
the development and use of ultrapure materials, chemicals, and de-ionized water.
Due to the increasingly stringent requirements of wafer surface cleanliness, this
approach alone cannot reach today’s required level of surface metals. In order to
reduce metal deposition in SC-1 solutions to meet and even exceed the current
surface metal specifications, it is necessary to not only use ultrapure components,
but to also add chelating agents to bind the free metal ions present forming com-
plexes which will remain soluble in solution. Typical chelating agents can reduce
the free metal ions in solution by 6 orders of magnitude [3].
In addition to enhancing the metallic cleaning ability of SC-1 solutions, we
have also investigated the use of a surfactant in our modified SC-1 solution. Al-
though SC-1 inherently removes particles quite effectively, megasonic energy is
often applied which dramatically increases particle removal efficiency. This is in-
creasingly important as the dimensional size of semiconductor devices continues
to decrease to even smaller sizes. With this mind, the addition of surfactants to
SC-1 will become an important component to prevent particles removed from the
wafer surface from re-deposition, thus increasing the particle removal efficiency
for small particles.
In liquids, the attraction or repulsion of particles to the wafer surface is de-
pendent on the van der Waals interaction (always attractive) and the electrostatic
double layer forces (usually repulsive). The combination of these interactions will
determine the potential energy of interaction and thus the barrier to adhesion [4,
5]. The barrier to adhesion is related to the particle size, pH of the solution, and
the respective charges on the wafer surface and particle. Cleaning down to sub-
micrometer and smaller sizes becomes increasingly difficult as the barrier to ad-
hesion decreases with decreasing particle size. Thus the tendency to re-deposit on
the wafer surface increases as the particle size decreases. Surfactants may prevent
Investigation of modified SC-1 solutions for silicon wafer cleaning 243
2. EXPERIMENTAL
Figure 1. Particle Removal - Example of Si3N4 particle wafer maps before (left) and after (right)
modified SC-1 clean (particles ≥ 0.12 µm, measured on Tencor SP-1).
244 C. Beaudry and S. Verhaverbeke
were deposited on prime 300 mm wafers. The particle measurements were per-
formed on a Tencor SP-1 instrument. Surface metal measurements were obtained
with the vapor phase decomposition-ion coupled plasma mass spectroscopy
(VPD-ICPMS) technique. Time-of-flight secondary ion mass spectrometry (TOF-
SIMS) was used to assess if any residual chelating agent or surfactant remained
on the wafer surface (after the rinsing and drying).
3. RESULTS
Figure 2 shows the particle removal efficiency for an optimized modified SC-1
solution. For one lot, consisting of 13 wafers, the average particle removal effi-
ciency was 99.5% (1σ = 0.28; measured at ≥ 0.12 µm). Typical wafer maps illus-
trating the combination of full and spot Si3N4 particle deposition pattern before
and after processing are shown in Figure 1. In order to determine the effectiveness
of the surfactant studied, Valtron SP2200, we compared final particle counts with
and without surfactant present (Figure 3). In this example we see an average of 50
less particles per 300 mm wafer. The particle cleaning performance for the modi-
fied SC-1 solution was excellent and the selected surfactant reduced the average
final particle count after SC-1 cleaning.
Figure 4 illustrates the effectiveness of the studied chelating agent for reducing
metallic deposition. In particular, it is interesting to look at the level for Al, Fe,
and Zn. These are some of the metals that readily deposit from conventional SC-1
Figure 2. Particle Removal - Average Si3N4 particle removal after modified SC-1 clean (particles
≥ 0.12 µm, measured on Tencor SP-1).
Investigation of modified SC-1 solutions for silicon wafer cleaning 245
Figure 3. Particle Removal - Final particle counts after modified SC-1 clean with and without sur-
factant (particles = 0.1-0.14 µm, measured on Tencor SP-1).
Figure 4. Metal Deposition - Surface trace metals levels after modified SC-1 clean as determined by
VPD-ICPMS (1 sigma error bars are generally within data points).
246 C. Beaudry and S. Verhaverbeke
solutions [1]. The average surface metals levels after the modified SC-1 clean was
equal to or below today’s VPD-ICPMS detection limits. For reference, typical
levels for a conventional SC-1 last clean are: Al ~ 1x1011, Fe ~ 2x1010, and Zn ~
1x1011 (atoms/cm2). The chelating agent under investigation is efficiently binding
the free metal ions in solution reducing their deposition onto the wafer surface
and thus can eliminate the need for an additional metal removal step. Although
the modified SC-1 solution did not deposit metals from the solution we also char-
acterized the metal removal efficiency of this solution (Figures 5 and 6). The con-
centration of chelating agent and exposure time were varied while the
NH4OH:H2O2:H2O volume ratio and temperature were held constant at 1:2:80 and
80°C. Figure 5 shows the results for a 30 s processing time. The final concentra-
tion of all metals investigated was typically greater than 1E+10 atoms/cm2. Fe
removal was found to be a function of chelating agent concentration while other
Figure 5. Metal Removal - Surface trace metals levels after a 30 s modified SC-1 clean using differ-
ent chelating agent concentrations as determined by VPD-ICPMS.
Figure 6. Metal Removal - Surface trace metals levels after a 10 minute modified SC-1 clean using
different chelating agent concentrations as determined by VPD-ICPMS.
Investigation of modified SC-1 solutions for silicon wafer cleaning 247
metals did not exhibit any dependence. Figure 6 shows the final concentrations of
metal after 10 minutes of exposure. All metals were reduced by 2-3 orders of
magnitude to close to or below 1E+10 atoms/cm2. Exposure time is obviously an
important consideration for metal removal. Methods to increase the metal removal
rate are now under investigation.
One of the concerns with the use of chelating agents and/or surfactants in SC-1
last clean is the potential for organic contamination remaining on the surface of
the wafer. However, in a typical spin cleaning equipment, the rinse process can be
optimized to eliminate such concerns. The use of heated DI water rinse and high
spin rates during rinsing can effectively remove all traces of the chelating agent
and surfactant. TOF-SIMS measurements were carried out to confirm the absence
of both organic additives. No trace of additive-specific residues was observed on
the processed wafers. Thus, organic contamination through use of appropriately
selected chelating agents and/or surfactants can be eliminated through process op-
timization. The absence of any heavy metal signature in the TOF-SIMS data also
confirms our VPD-ICPMS results.
4. SUMMARY
In this paper we have shown that the addition of an appropriately selected chelat-
ing agent to SC-1 solutions can eliminate the need for an additional metal removal
step, potentially saving time and money. In addition, the use of a surfactant can
enhance particle removal efficiencies for very small particle sizes (<0.14 µm).
This modified SC-1 solution, containing both additives, was shown to have excel-
lent particle removal efficiency, to reduce metal deposition on the wafer surface
to below VPD-ICPMS detection limits, and to remove surface metal contamina-
tion. Methods to improve metal removal efficiency are currently underway. Fur-
thermore, rinsing can be optimized to eliminate all traces of the chelating agent
and surfactant residues.
REFERENCES
MICHAEL T. ANDREAS∗
Micron Technology, Inc., Mail Stop 306, 8000 S. Federal Way, Boise, ID 83707-0006, USA
Abstract—An inexpensive qualification technique is described for wafer cleaning tools used after
chemical-mechanical polishing (CMP). Pipette deposition of slurry onto a monitor wafer can pro-
vide the particle challenge needed to qualify the performance of a post-CMP wafer cleaning tool. In
addition to gauging the performance of these tools, this pipette method is faster and less expensive
than many common particle deposition techniques, including immersion, polishing and aerosol
deposition.
Keywords: Brush cleaning; chemical-mechanical polishing; CMP; particle deposition; PVA; slurry;
wafer cleaning.
1. INTRODUCTION
Surface preparation and cleaning is one of the most critical steps in semiconduc-
tor manufacturing [1]. For all wafer cleaning tools, routine qualification is neces-
sary to ensure that no particle contamination is introduced by the wafer cleaning
equipment [2]. The broad category of wafer cleaning tools includes the poly(vinyl
alcohol) (PVA) brush scrubbing tool [3]. The brush scrubber has been increas-
ingly utilized [4] in semiconductor fabrication as a preferred technique for particle
removal after CMP. Because of the high particle removal performance required of
post-CMP cleaning tools, it is critical to monitor and maintain the performance of
such tools [5]. The most direct measure of tool performance is inline inspection of
actual product wafers [6]. While inline defect analysis is invaluable, it may re-
quire a time lag of hours or even days between wafer cleaning and discovery of
high wafer defectivity. In a high-volume manufacturing environment, this delay
can lead to hundreds of product wafers with possible contamination. For this rea-
son, inline inspection of product wafers is supplemented by regular tool qualifica-
tion using less expensive particle monitor (PMON) wafers. This PMON qualifica-
tion should provide an accurate measure of the tool performance with the quickest
possible turnaround time. For post-CMP cleaning tool qualification, it is neces-
∗
Phone: (208) 368-5067, Fax: (208) 368-2548, E-mail: andreas@micron.com
250 M.T. Andreas
sary to use prepared PMON wafers with contamination analogous to that found
on polished product wafers. There are several methods for preparing these con-
taminated monitor wafers. One way is to use polished monitor wafers [7]. In this
method, test wafers can be selected which represent the surface chemistry of
product wafers without the expense of underlying circuitry. The test wafer surface
can be homogeneous or heterogeneous, depending on the process being qualified.
These test wafers can be polished under the same CMP conditions as product wa-
fers. This method requires the same resources as inline product inspection, there-
fore providing an accurate measure of post-CMP tool performance but not neces-
sarily decreasing the turnaround time. Another common method is the slurry dip
[8, 9], where monitor wafers are dipped in a wet process tank of diluted slurry.
Again, monitor wafers can be selected to represent the surface chemistry of prod-
uct wafers. The diluted slurry can be selected to simulate the CMP chemistry.
This method has the advantage of providing contaminated monitor wafers without
the time or expense of using a CMP tool. Also, it is possible to deposit dry parti-
cles using an aerosol deposition technique [10]. While it is claimed that this
method is more controllable and repeatable than aqueous slurry immersion, these
dry particles may not represent polishing residue as accurately as a CMP slurry.
Here we describe an extremely simple contamination technique – direct pipette
deposition of a small volume of undiluted slurry onto a monitor wafer.
2. EXPERIMENTAL
Bare silicon and blanket oxide wafers were used for all tests. Blanket oxide wa-
fers were prepared by plasma enhanced chemical vapor deposition (PECVD) us-
ing tetraethoxysilane (TEOS). These TEOS derived films were deposited to 350
nm thickness on 200 mm diameter silicon substrates. All wafers were cleaned us-
ing OnTrak DSS-200 Series II brush cleaning tools. These tools were run using a
dilute (<1%) basic cleaning solution. All wafers were inspected with a Tencor
SurfScan 6420 laser scattering wafer inspection tool [11, 12]. Bare silicon moni-
tor wafers were inspected for all light-scattering point defects (LPDs) >0.16 µm.
Blanket oxide wafers were inspected for LPDs > 0.18 µm.
3. RESULTS
Table 1.
Initial slurry drop test results
tion (as measured by laser scatterometry) for a given drop size. Maps showing the
LPD distribution after 2 second scrubbing are shown in Figures 2 and 3 for wafers
contaminated with alumina and colloidal silica, respectively. In all cases, the 80
second brush cleaning time was sufficient to attain particle levels below the pre-
defined production limit of 100 LPDs. Silicon monitor wafers run before and after
each group of slurry drop test wafers did not show significant slurry carryover.
Due to the simplicity of this procedure, slurry drop deposition was investigated
further as a method for routine tool qualification.
Figure 2. Wafer map showing LPD distribution after alumina slurry drop and 2 second scrub.
Figure 3. Wafer map showing LPD distribution after colloidal silica slurry drop and 2 second scrub.
254 M.T. Andreas
Table 2.
Experimental scrubber recipes
Recipe Brush height, Chemical flow, Brush rotation, Brush time, Rinse time,
mm L/min 1/min sec/brush sec
C 3.5 0.5 139 40 9
E1 1.0 0.5 139 40 9
E2 3.5 0.5 40 40 9
E3 3.5 0.5 139 10 9
E4 3.5 0.5 139 20 9
E5 3.5 0.0 139 40 9
E6 3.5 0.5 139 40 5
E7 3.5 0.5 139 40 13
Table 3.
Experimental scrubber qualification method details
Table 4.
Experimental details, dLPD results and method dynamic range (MDR) for each series
Table 5.
The full factorial of scrubber qualification tests using a known good brush installation
Figure 4. Several wafer maps from the SQM optimization tests. These silicon wafers were scrubbed
using recipe A1. The lower left note in each frame indicates drop size (1d = 1 drop, 3d = 3 drops)
and slurry composition (S = silica, A = alumina).
Performance qualification of post-CMP cleaning equipment 257
Figure 5. Wafer map for a CVD oxide wafer contaminated with one drop of alumina slurry and
scrubbed using recipe A1.
Table 6.
Recipe details for more sub-optimal scrubber recipes used to test the sensitivity of optimized SQMs.
Recipe C is the control scrubber recipe.
experimental scrubber recipes A1 and A2. The improved SQM using 3 drops of
alumina slurry was selected for comparison to the previous SQM using 1 drop of
colloidal silica slurry. Based on results with experimental recipes A1 and A2,
three more sub-optimal scrubber recipes were created. These recipes (T1, T2 and
T3) were designed to perform somewhere between recipes C (control) and A1.
These recipes are described in Table 6. Bare silicon wafers were used for all fur-
ther tests. Wafer run order was randomized to average out any brush carryover ef-
fects. Process details and results are shown in Table 7. Wafer maps showing sen-
sitivity to radial defect pattern formation are shown in Figure 6. The dLPD results
comparing optimized alumina and silica slurry drop methods are plotted in Figure
7. The alumina slurry drop SQM shows better sensitivity to inefficient scrubber
258 M.T. Andreas
operation, especially using recipe T1. All wafers using silica and alumina SQMs
met contol levels for dLPDs when using the control scrubber recipe. One of the
silica drop trials showed anomalously high residue. This may have been carryover
from the previously scrubbed wafer, which brought alumina slurry contamination.
Further tests confirmed that carryover from 3 drops of alumina appeared when us-
ing recipe T1. All experiments confirmed the improved alumina drop SQM as
more sensitive to conditions which may cause radial defect patterns. After imple-
menting this improved SQM, no further radial defect patterns were discovered on
product wafers.
Table 7.
Process details and inspection results for improved slurry drop SQM comparison. All tests were run
using bare silicon wafers. The high dLPD result for 1dS wafer 19 (5312 adders) may be due to
carryover from 3dA wafer 18
Figure 6. Representative wafer maps from the 3 drops alumina (3dA) qualification method on sili-
con wafers, showing sensitivity to radial defect patterns when using recipes T1 and T3.
Performance qualification of post-CMP cleaning equipment 259
Figure 7. Plot of dLPD versus scrubber recipe for optimized alumina and silica slurry drop SQM trials.
4. CONCLUSION
Manual pipette deposition of CMP slurry onto a monitor wafer is a quick and ef-
fective way to provide qualitative particle challenges to wafer cleaning equip-
ment. This method is much faster and cheaper than other common particle deposi-
tion techniques, including polishing, aerosol deposition and immersion in dilute
slurry. We have shown that this method provides enough particle loading to de-
termine whether or not a post-CMP cleaning tool will perform within acceptable
particle removal limits. This, in turn, translates to lower defects on product wafers
and improved yields at a lower quality control cost.
REFERENCES
1. W. Kern (Ed.), Handbook of Semiconductor Wafer Cleaning Technology, pp. 416-419, Noyes
Publications, Park Ridge, NJ (1993).
2. F.W. Kern, Jr. and G.W. Gale, in: Handbook of Semiconductor Manufacturing Technology,
Y. Nishi and R. Doering (Eds.), pp. 87-104, Marcel Dekker, New York (2000).
3. W. Krusell, J.M. de Larios and J. Zhang, Solid State Technol., 38, No. 6, 109-114 (1995).
4. R. DeJule, Semiconductor Intl., 56-64 (Nov. 1998).
5. J.M. de Larios, J. Zhang, E. Zhao, T. Gockel and M. Ravkin, MICRO, 15, No. 5, 61-73 (1997).
6. C. Dennison, MICRO 16, No. 2, 31-42 (1998).
7. D.W. Cooper, R.C. Linke and M.T. Andreas, MICRO 17, No. 7, 55-64 (1999).
8. A.A. Busnaina, N. Moumen, M. Guarrera and J. Piboontum, in: Semiconductor Fabtech – 9th
Edition, M.J. Osborne (Ed.), pp. 279-282, ICG Publishing, London (1999).
9. S. Ramachandran, A.A. Busnaina, R. Small, C. Shang and Z. Chen, in: Semiconductor Fabtech
– 13th Edition, G. Oliver (Ed.), pp. 271-277, ICG Publishing, London (2001).
260 M.T. Andreas
10. Y.H. Liu, S.H. Yoo, S.K. Chae, J.J. Sun, K. Christenson, J. Butterbaugh, J.F. Weygand and
N. Narayanswami, Semiconductor Intl., 145-152 (June 2000).
11. R.S. Howland, Semiconductor Intl., 164-170 (Aug. 1994).
12. J.J. Shen and L.M. Cook, MICRO 15, No. 3, 53-66 (1997).
13. N. Moumen, M. Guarrera, J. Piboontum and A.A. Busnaina, in: Proceedings, 10th Annual
IEEE/SEMI Advanced Semiconductor Manufacturing Conference, pp. 250-254 (1999).
Surface Contamination and Cleaning, Vol. 1, pp. 261–266
Ed. K.L. Mittal
© VSP 2003
MOSHE OLIM∗
Seagate Technology, 7801 Computer Avenue, South Bloomington, Minnesota 55435, USA
1. INTRODUCTION
The typical processed surface is dry as the process starts. If a chemical reagent is
dispensed onto a dry surface it is likely that different parts of the surface will be
subjected to the reagent for different time intervals thus resulting in a nonuniform
processing result. Therefore, prior to dispensing chemical reagents on the surface,
it is imperative that the surface be covered with an inert liquid. This liquid is typi-
cally deionized water which covers the whole surface and fills the trenches. The
time required to fill a trench depends strongly on the characteristics of the surface
processed, the surface tension of the water, and the width and depth of the trench.
If the trench is hydrophilic, the capillary action of the water/air interface will en-
sure that the trench fills with water. The geometry of the process is shown in Fig.
1. The pressure of the gas trapped in the feature is increased due to the capillary
force, and this increase in pressure enhances the diffusion of the gas into the liq-
uid. The process continues until the gas trapped in the cavity is completely con-
sumed by this diffusion process. The upper limit on the time required to fill the
trench can be estimated (for details see [1]) as follows:
∗
Phone: 952-402-5888, E-mail: moshe.olim@seagate.com
262 M. Olim
Figure 1. Geometry of the trench filling process. PL is the pressure in the liquid, ∆P is the capillary
pressure difference across the interface, w and h are the trench width and depth, respectively, and θ
is the contact angle.
Figure 2. Time required to complete the trench filling process as a function of trench width for con-
tact angle values of 30 and 60 degrees. The surface tension of the liquid is 70 mN/m, and trench as-
pect ratio h/w = 10.
k æ 1 + Ph ö 2
t fill = ç ÷h (1)
DRT è ∆P ø
where k, D, R, T, and h are, respectively, a proportionality constant, vapor/air dif-
fusion coefficient, universal gas constant, temperature, and trench depth, and Ph
and ∆P are the atmospheric pressure and capillary pressure difference, respec-
Spatial and temporal scales in wet processing of deep submicrometer features 263
tively, across the interface. The results of the calculation are shown in Fig. 2. It
can be seen that the time required to fill a 0.25 µm wide trench is well below one
second even for a trench whose aspect ratio is 10.
With the trench full of DI water, a chemical reagent is dispensed onto the sub-
strate. In the interest of uniformity of processing along the full depth of the
trench, it is important that the spatial concentration of the reagent along the trench
depth be kept as uniform as possible. For analysis purposes, one may assume that
the trench is full of water and the top of the trench is covered with liquid reagent.
The reagent may penetrate the trench through either (a) convective or (b) diffusive
mixing. In order for convective mixing to take place, the flow characteristics must
allow for vortices to exist. The possibility of vortex existence may be ruled out by
comparing the relevant geometric parameters to the smallest vortex diameter pre-
dicted by Kholmogorov scales (see [2]) using the following equation:
−3/4
η æ ul ö
≈ç ÷ (2)
l èvø
where η is the smallest length scale that can sustain turbulence (i.e. the smallest
vortex diameter possible), l is a characteristic length of the system (in this case it
is the width of a trench), and u and v are the flow velocity and kinematic viscos-
ity, respectively. With relatively large trench width and velocity values of l = 0.5
µm and u = 1 m/s one obtains η/l ≈ 1, and with more realistic values of l = 0.25
µm and u = 0.01 m/s one obtains η/l ≈ 100. Since the smallest possible vortex di-
ameter is noticeably larger than the trench width, it is clear that turbulent mixing
cannot take place in the trench. This implies that the reagent is transferred into the
trench by diffusion only. The same argument applies to transfer of reagents out of
the trench when DI water is dispensed onto the surface in order to stop the reac-
tion.
Since it is desired that the results of the chemical process be uniform along the
depth of the trench, it is clear that the exposure time of any point on the trench
wall to the reagent should be as close as possible to that at any other point along
the trench wall. The uniformity of the process may be estimated by comparing the
time it takes for the reagent concentration at the bottom of the trench to equalize
with that at the top of the trench. For practical purposes, let us consider the con-
centrations equalized when the concentration at the bottom reaches 95% of the
concentration at the top. The time required for the reagent concentration at the
trench bottom to reach a given concentration level can be estimated (see [1]) us-
ing the following one-dimensional diffusion equation:
264 M. Olim
Figure 3. Normalized chemical concentration at the trench bottom vs time. u is the velocity of the
liquid at the top of the trench, and D is the diffusion coefficient of the chemical in the liquid.
∂C ∂ 2C
+ 2 =0 (3)
∂t ∂z
where D = 1.e-9 m2/s is the diffusion coefficient and C is the reagent concentra-
tion subject to C(t,z=h) = 1, Cz(t,z=0) = 0, C(t=0,z<h) = 0. The results shown in Fig. 3
clearly show that the concentration at the trench bottom reaches 95% of the
concentration at the trench top in approximately 25 ms.
The transport of the reagent out of the trench is described by the same equation
with different initial and boundary conditions C(t,z=h) = 0, Cz(t,z=0) = 0, C(t=0,z<h) = 1
and calculations show that within 100 ms the maximum concentration is reduced
by seven orders of magnitude.
3. DRAINAGE
Typically, the de-ionized water used to rinse the substrate must be removed (in
the liquid state) from the substrate and out of the features fabricated in the sub-
strate. Water that is not removed in the liquid state would evaporate leaving origi-
nally dissolved contaminants to coagulate on the substrate and in the features thus
adversely affecting the yield of the manufacturing process.
Spatial and temporal scales in wet processing of deep submicrometer features 265
Substrate rotation, since it increases the body force on the water in the features,
is often utilized as a means of water removal enhancement. However, as the typi-
cal size of the features decreases, the importance of body forces compared to sur-
face tension (manifested in the Bond number) also decreases thus reducing the ef-
ficiency of rotation as a mechanism for water removal. It has been shown [4] that
the amount of water removed from a trench for a given contact angle does not
change significantly below Bond number of 0.1 which may, therefore, be defined
as the critical Bond number. The parameter that determines the amount of water
removed when the Bond number is below the critical value is the contact angle θ,
and the maximum amount of water removed from the trench of width L does not
exceed πL2/4 per unit length of a long trench. Since the depth of the trench is sig-
nificantly larger than its width, very little water is removed from the trench in the
liquid phase and most of the water must evaporate.
4. EVAPORATION
This step is typically achieved by purging the process chamber using dry nitrogen
at ambient temperature and pressure. The equations describing the time interval
required to dry by evaporation a feature of a given depth are developed in [3] and
they allow for any temporal variation of vapor concentration in the ambient.
The rate of evaporation through a stagnant gas in one dimension is expressed in
Equation (17.2-15a) in [5]. A slightly modified version of this equation is pre-
sented below:
PD æ P − Pv ,top ö
N = ln ç ÷ (4)
RTz çè P − Pv ,int ÷ø
where P and T, respectively, are the pressure and temperature in the system, D is
the diffusivity of the vapor in the ambient gas, R is the universal gas constant, z is
the distance between the top of the feature and the liquid/gas interface, and Pv is
the partial pressure of the vapor either at the top of the feature (top) or at the liq-
uid/gas interface (int). This may be converted into an equation showing the rate of
recession of the air/liquid interface which, in turn, may be integrated (for details
see [3]) to yield
%
∆s æ 1− P v ,top ö
%z 2 = ò ln ç ÷÷ ds (5)
ç %
è 1− Psat ø
0
where
1/2
æ ρ L RT ö Pv ,top
%z = z ç ; %
P % = Psat ,
÷ v ,top = ; Psat (6)
è 2 M L PD τ ø P P
266 M. Olim
5. SUMMARY
REFERENCES
Abstract—As the state of technology advances in data storage and integrated circuits, the need to
remove smaller particles becomes more critical in order to maintain economical yields and avoid
product failures due to contamination. Present textile materials are approaching the limits of their
ability to achieve particle removal. New materials are being developed and investigated for provid-
ing the high level of cleaning efficiency required. These new materials fall into the general class
known as microdeniers. They are being shown to have the properties required to overcome the
shortcomings of traditional textiles.
Keywords: Contamination control; wipers; microdenier; cleanroom wipers; particle removal; fiber
construction.
1. INTRODUCTION
The critical dimensions in data storage and integrated circuit technologies con-
tinue to get smaller and smaller, putting these high-tech products at risk from sub-
micrometer particles. As flying heights approach 25 nm, contamination of disk
media and read-write heads during their manufacture becomes even more of a
concern. A half-micrometer (500 nm) particle is more than ten times the read-
write gap, perhaps leading to a read-write error or a crash. Similarly, integrated
circuits continue to have ever-decreasing “line width” ground rules, requiring the
control of particles of ever decreasing size limits. Particles hundredths of a mi-
crometer in size can be “killers” [1].
The relatively large fiber diameters of standard wipers make them less efficient
in picking up these small particles. Microfibers have been shown to be more effi-
cient in this application based on their geometry and the physics of particle re-
moval [2].
∗
To whom all correspondence should be addressed. Phone: 201-327-9100 X330, Fax: 201-327-
5945, E-mail: jskoufis@texwipe.com
†
Now retired.
268 J. Skoufis and D.W. Cooper
During plant and cleanroom construction, and then throughout operations, rigor-
ous cleaning is needed, using such consumable materials as wipers, swabs, clean-
ing compounds, and cleaning tapes. Such materials need to be selected carefully
both with respect to contaminating potential and ability to clean. Cleanliness is
measured by examining materials, extracting particulate matter in agitated liquid
or in air, and by extracting chemical constituents such as ionics, hydrocarbons,
non-volatile residue (NVR or residue after evaporation) etc., with appropriate sol-
vents [3].
Cleanroom wipers (and swab heads) selected for cleanrooms of different levels
of cleanliness generally follow these guidelines:
A. Cleanest, Fed-Std-209E Class 1 (= Class M1.5) to Class 10 (= Class M2.5)
rooms:
laundered sealed-edge polyester knit (lowest contaminants) or Nylon knit
(somewhat higher non-volatile residues). A portion of such a wiper is shown
in Figure 1.
B. Class 10 to Class 100 (= Class M3.5) rooms:
laundered polyester knit or hydroentangled polyester (somewhat higher parti-
cles and fibers due to unsealed edges).
C. Class 100 to Class 10,000 (= Class M5.5) rooms:
hydroentangled polyester/cellulose blends.
Figure 1. Laundered sealed edge wiper representing the highest level of cleanliness for critical
cleanroom applications.
Microdenier fabrics for cleanroom wipers 269
D. Classes >10,000:
These consist of various composites, including polypropylene and natural fi-
bers such as cotton and polyurethane foam.
The progression of wiper offerings is logical:
1. sealed-edge products are cleaner than unsealed;
2. laundered products are cleaner than unlaundered;
3. continuous filaments give less fiber contamination than do cut (“staple”) fi-
bers;
4. natural fibers contain contaminants harder to control than those in synthetic
fibers, and natural fibers are only available as staple fibers.
The efficacy of a wiper includes absorbency and ability to pick up and retain
particulate matter. This will depend on the fiber or foam base material and on
construction details. Inevitably, some contaminants will reach the work surface
where non-contaminating materials are needed to remove them.
3. MICRODENIER FABRICS
The denier of a yarn or fiber is the linear density expressed as the number of
grams in 9000 m of the yarn or fiber. Microdenier fibers (filaments) are defined as
1 denier or less, where 1 denier for polyester corresponds to a circular cylinder
with a diameter D of about 11 µm [4].
For fabrics of the same fiber composition of denier d:
– the total length of fibers will be proportional to 1/d
– the radius of a fiber will be proportional to d
– the pores created by bundles of such fibers will have dimensions proportional
to d .
– the total surface area of fibers will be proportional to 1/ d
– the cross-sectional area of a fiber will be proportional to d.
The techniques for creating such fine fibers include:
1. extruding a two-component mixture, then dissolving one component
2. extruding a two-component mixture, then fracturing the fiber with high-
pressure water, mechanical action, or chemical stress
3. extruding a single component, then elongating and thinning the fibers with
high-temperature gas jets.
The first two types, known as “islands in the sea” and “pie,” respectively, are
the ones normally used for microdenier and ultra-microdenier (below 0.1 denier)
materials used for cleanroom applications. They are generally made into woven or
knitted goods. The third type is used for “nonwovens” only, because of the diffi-
culties associated with weaving and knitting these fibers.
270 J. Skoufis and D.W. Cooper
Figure 2. Various manufacturing techniques for forming microdenier fibers. Other technologies can
be used but generally will fall into one of the four shown. The first is a single component while the
others are multicomponent fibers.
Mochizuki et al. [5] of Unitika Ltd. of Japan have described the formation of
ultra-fine fibrous materials by splitting bi-component fibers after they have been
formatted as a spunbonded fabric (molten continuous fibers are laid down and ad-
here where the fibers cross). They noted that three methods were conventionally
used (Figure 2) to form microdenier fibers from bi-component fibers: dissolution
of one of the components, separation of the components by swelling or shrinking,
and separation by mechanical distortion, the last being the approach they used. A
“sunflower” pattern (Figure 3) of six polyester fibers surrounding a polyethylene
core was split apart by flexing, taking a 3-denier fiber and making six 0.25-denier
fibers and a 1.5-denier core segment. The thin fibers contributed softness and
flexibility; the core allowed convenient thermal bonding. The density of the fabric
decreased, giving greater porosity. Triboelectrification of the fibers from dissimi-
lar materials can produce charging that enhances the effectiveness of dry wiping
in picking up dust particles [6]. Of course, this is not recommended for contact
with semiconductors.
Teijin Ltd. of Japan markets its Micro-StarTM material, made of bi-component
nylon/polyester fibers, that are arranged into 16 filaments that have wedge-shaped
cross sections (Figure 4). Such filaments are typically 0.16 denier, but can be made
smaller. The materials are useful for absorbing both oil and water. Tests done on
cleaning tapes made from these materials generally showed lower levels of ionic
Microdenier fabrics for cleanroom wipers 271
Figure 3. Sunflower pattern showing six polyester fibers surrounding a polyethylene core. Splitting
results in making 3 denier fiber into six 0.25 denier fibers with a 1.5 denier core. The left figure
shows an oblique view and the right figure shows a cross section of the bicomponent filaments.
Courtesy of Unitika Ltd.
contaminants than from tapes made with conventional denier fibers (1.5 denier).
This is not an inherent quality but the result of a more rigorous cleaning step.
Kuraray Ltd. of Japan produces its SOLIV (R)TM fibers by splitting polyester
fibers longitudinally, forming fibers that have roughly rectangular cross sections,
claimed to facilitate wiping through a scraping mechanism (Figure 5) [7].
Toray Industries, Inc. of Japan markets its microdenier cleaning cloth as Toray-
seeTM and LuminexTM. The polyester fibers are produced by splitting thicker fibers,
creating material with a 2-micrometer diameter (0.06 denier) and having sharp
edges rather than being round. Figure 6 shows a comparison against materials
made with larger, traditional microdenier fibers (ca. 5 micrometers in diameter).
The current technology allows the formation of woven and knit goods from fi-
bers having deniers as low as 0.06. It also allows the formation of nonwoven
items, typically synthetic leathers, having deniers as low as 0.0001. While these
“nanofiber” materials have not yet been utilized into cleanroom wipers, the tech-
nology is nearly available to allow weaving and knitting these fibers into cleaning
materials. These fine fibers will probably be much more fragile in an unbonded
state and their usefulness as wiping cloths need to be determined.
272
J. Skoufis and D.W. Cooper
Figure 5. Longitudinally split fibers showing how the rectangular shapes facilitate wiping through a scraping mechanism. The upper left shows the
construction of a single filament while the upper right shows the same filaments split and knit into a wiper.
Courtesy of Kuraray Ltd.
Microdenier fabrics for cleanroom wipers 273
Figure 6. Ultra microdenier polyester cleaning cloth showing split wedge shaped fibers approxi-
mately 2 micrometers in diameter (0.06 denier) is shown on the left. Comparison is made with two
traditional larger microfiber cloths on the right.
Courtesy of Toray Ltd.
The force needed to break a fiber (in tension) is generally proportional to its
cross-sectional area, a, so that the tensile strength of a textile fiber is often given
as gpd, grams per denier, which should be roughly independent of fiber diameter.
For polyester this is about 3-10 gpd and is commonly called the fiber “tenacity”.
The yarns of interest to us are in the range of 50 to 150 denier (composed of many
filaments), meaning a single yarn could suspend 150 to 1500 g without breaking,
depending on the tenacity of the particular polyester material [8].
The strength of an individual filament determines whether it breaks when
snagged; the strength is proportional to (denier)/(# filaments) = d/n. Bending
strength is usually also proportional to cross-sectional area, so the same propor-
tionality can be expected. To prevent pilling (formation of lint balls), most
polyester in commercial use is low tenacity. Higher tenacity fibers are needed for
wipers required to have abrasion resistance.
The geometry of a twisted yarn is complicated. The densest plausible packing
would be that of a unit cell that is a hexagon of contiguous cylinders surrounding
a central cylinder, which would give less than 10% open cross-sectional area. A
more plausible approximation is a square unit cell, with sides 4r in length, cir-
cumscribing four cylinders, of radius r, shown in Figure 7. This gives a porosity
274 J. Skoufis and D.W. Cooper
Figure 7. Calculation of cross-sectional porosity, e, or open space, of a twisted yarn from the dens-
est possible packing of filaments.
e = 1-(π/4) = 0.215 = 21.5%. The spaces around these cylinders are complicated,
but they are roughly cylindrical pores with radii that are about half that of the
filament cylinders, r/2. r/2 will be taken here to be the characteristic dimension of
these inter-filament pores in the yarn. Note that this model is quite approximate,
as this geometry would in fact let liquid enter only from the ends of the yarn, and
not from the periphery, which clearly is not the case in practice.
The packing density is a determining factor in many of the properties of the
wiper: the absorbency, ability to pick up particles, and feel are all affected. Once
the filament size is determined, the science of fabric construction comes into play
in order to provide optimum properties.
inference, the boundary layer will be proportional to the square root of the fila-
ment cross section and therefore to (d/n) . Some boundary-layer reduction ad-
vantage is expected for microdenier fabrics based on this analysis.
The height to which a non-volatile liquid can rise in a fabric by absorption is
inversely proportional to the pore size. Microdenier fabrics, because of their
smaller pore sizes compared to standard denier fabrics, should raise liquids
higher. The speed the liquid will travel horizontally through the fabric is propor-
tional to the pore size. Microdenier fabrics provide less speed of absorption. This
may not be critical for many applications but in cases where it is, fabric construc-
tion can overcome this disadvantage.
The amount of liquid a wiper can hold is roughly proportional to its thickness.
For the same knit, the thickness can be expected to be roughly proportional to the
square root of the yarn denier, so wipers with larger deniers should offer greater
absorption capacity. However, the weight of the wiper will be the product of the
denier and the length of yarn used to knit it. To keep the basis weight the same,
the length of yarn will have to be inversely proportional to the denier. This would
mean fewer or shorter loops. In the limit of a very large denier, one would have
almost a very loose weave, rather than a knit, which will adversely affect absorp-
tion capacity. Larger denier yarns would also feel stiffer and more difficult to get
into tight spaces.
It is not completely clear what the implications of denier are in absorption ca-
pacity as so much is dependent on basis weight and construction. Experience has
shown that the ability to hold liquid is of less interest in critical cleaning than the
ability to pick up small particles and reduce boundary layers.
For different fabrics with the same basis weight (g/m2), there will be the same
volume of polyester contained in a total length of filaments L'. The volume of the
filaments would be πr2L'. The surface area would be 2πrL'. The surface area per
unit volume (thus per unit weight) would be 2/πrL', inversely proportional to the
filament radius, thus the surface area would be proportional to (n/d) . The
greater the surface area, the higher the liquid absorption, but cleaning the fabric
may be more difficult.
Figure 8. If one circle represents a particle and the other a fiber, the force, Fv, required to lift the
particle can be calculated. The smaller the fiber, the smaller the particle it can push up from the sur-
face.
The contact (idealized) between the cylindrical fiber and the spherical particle
occurs at their surfaces, along a line from one center to the other. The vertical
component of the force is then
Fv = F sinθ,
where F is the force at the contact and θ is the angle with respect to the surface or
Fv = F (r - R) / (r + R),
with the smaller object being pushed toward the surface and the larger object be-
ing pushed away. The smaller the denier, the smaller the particle it can push up
from the surface.
Besides the size advantage microdeniers offer in picking up very small parti-
cles, microdeniers can be made in a variety of shapes, even within the same
unsplit bicomponent filament. Common shapes include: shovel, wedge, star,
multi-lobe, and interior hollows. The wide variety of shapes provides opportuni-
ties for investigation into their suitability for particle removal. Specific shapes
may be optimal only for specific particle types.
Microdenier fabrics for cleanroom wipers 277
7. FUTURE ACTIVITIES
While microdenier materials have been in wide use throughout Asia in clean-
rooms, they have been commercialized slowly in the United States. One reason is
the lack of US producers; another is their higher cost compared to standard denier
products.
Interest in microdenier materials is gradually increasing in the U.S. More so-
phisticated microelectronic products demand better contamination control. The
price of microdeniers is decreasing due to competition as more microdenier mate-
rial manufacturers are entering the market.
Several manufacturers who intend to revolutionize fiber production and signifi-
cantly reduce manufacturing costs have designed new equipment in the U.S. Sev-
eral companies and universities have purchased pilot-scale and production equip-
ment. Initial trials look promising. The equipment is able to change polymer types
quickly, and configurations not previously possible are now feasible. These ma-
chines can produce nanofibers of various shapes and various polymers by simple
economical screen changes, rather than the time-consuming changes in expensive
spinnerets previously needed.
8. CONCLUSIONS
The new microdenier fibers being used for textiles should be able to produce
highly absorbent and very clean and effective wiping materials for cleanroom use.
The technical developments of the industry are driving the need for fibers and
fabrics which are more effective in removing ever smaller particles. Asia has led
in development and use of microdenier fibers, but in the U.S. manufacturers are
catching up, using economical and flexible designs and methods.
REFERENCES
Abstract—Optical reflectometry was used to study the attachment and subsequent detachment of
silica particles (diameter 25 nm) from the surfaces of titanium dioxide wafers under well-defined
hydrodynamic conditions. The rate of detachment and maximum detached amount was studied as a
function of both pH and added linear polyphosphate solutions. The latter have the general formula
[PnO3n+1](n+2)- where n is the number of phosphorous atoms in the molecule. The maximum detached
amount increased with increasing pH. The maximum detached amount also increased with n.
Atomic force microscopy was used to measure the interaction between silica spheres (diameter
7 µm) and titanium dioxide wafers under the same solution conditions. The detachment force
needed to separate the surfaces decreased with increasing pH as well as with n in direct agreement
with the reflectometry data. It was shown that, in addition to repulsive electrical double layer forces,
adsorbed polyphosphates provided a short-ranged steric layer that reduced the lateral interaction be-
tween the surfaces. The use of these two complementary techniques has given valuable insight into
the processes responsible for fine particle detachment and has particular application to surface
cleaning.
Keywords: Fine particle detachment; reflectometry; atomic force microscopy; particle adhesion.
1. INTRODUCTION
nanosized silica particles onto titanium dioxide surfaces has been studied as a
function of pH and addition of solutions of linear polyphosphates. Previous in-situ
infra-red studies [2] have shown that linear polyphosphates selectively adsorb
onto titanium dioxide forming strong chemical bonds. Streaming potential meas-
urements [3] have shown that for a fixed polyphosphate concentration, the tita-
nium dioxide becomes more negative with increasing n, which was attributed to
an increased charge density with n. In addition, direct force measurements [3-5]
have shown that adsorbed polyphosphate introduces a steric layer which leads to a
short-ranged repulsive interaction.
Reflectometry combined with a stagnation point flow cell was used to measure
silica particle attachment and detachment. The stagnation point flow is ideally
suited for the study of colloidal particle attachment and detachment processes in
which their transport is governed only by diffusion [6].
2. EXPERIMENTAL
2.1. Materials
2.1.1. Titanium dioxide wafers
A titanium dioxide layer was deposited on the surface of silicon wafers by sput-
tering (prepared at Philips Research, The Netherlands) [3]. Ellipsometry meas-
urements showed the titanium dioxide layer to be 40 nm thick and XPS analysis
showed the composition of the deposited layer to be pure titanium dioxide. X-ray
diffraction showed the deposited TiO2 to be amorphous. Imaging by AFM showed
an rms roughness of 0.3 nm over an area of 1 µm2 with a maximum peak height
of 2 nm. The wafers were cleaned by detergent washing followed by rinsing with
high-purity water, ethanol, heptane and copious amounts of more high-purity wa-
ter. Finally, the wafers were blown dry in a stream of nitrogen and plasma cleaned
(Harrick Plasma Cleaner/Steriliser PDC-32) for 1 minute immediately prior to
use. The isoelectric point of the titanium dioxide covered wafers was determined
to be pH 4.2 using streaming potential measurements [3].
2.1.2. Silica
Suspensions of silica particles used for reflectometry experiments were prepared
from LudoxTM AS40 (DuPont). The particles were dialyzed for 2 days in Milli Q
water and then suspended in solutions of KNO3 (10-3 M) with a particle concentra-
tion of 100 mg/l. The mean particle radius was R = 12 ± 2 nm, determined by
transmission electron microscopy (CESMA, Adelaide University). The silica
spheres used for AFM measurements were obtained from Allied-Signal, (Chi-
cago, Illinois). XPS analysis (CSIRO Division of Molecular Science) showed the
composition of the sample to be pure silica. The typical diameter was found to be
7 µm. AFM imaging of the spheres over an area of 500 nm2 showed the rms
roughness to be 0.8 nm.
Fine particle detachment studied by reflectometry and atomic force microscopy 281
2.2. Methods
2.2.1. Reflectometry
The experimental setup for the combined reflectometry and stagnation point fluid
cell is shown schematically in Figure 1. A detailed description of the experimental
procedure and the theory behind the technique is given elsewhere [7, 8]. Here a
Figure 1. Schematic diagram of the combined reflectometry and stagnation point flow cell.
282 A. Feiler and J. Ralston
brief description of the essential features is given. The reflected light from a plane
polarized He/Ne laser off an adsorbing surface is measured. The adsorbing sur-
face in this work was titanium dioxide covered silicon wafers. The reflected light
is split into its parallel (p) and perpendicular (s) components and the measured
signal, S, is the ratio of these intensities:
IP
S= (1)
IS
This ratio depends on the refractive index profile close to the surface of the
substrate. Material adsorbed at the interface, in this case silica nanoparticles, will
change the refractive index profile and hence result in a change in S. Quantitative
measurements of the attached particle amount (Γ) can be obtained from the
change in signal via:
1 ∆S
Γ= (2)
As S 0
where S0 is the intensity ratio prior to adsorption and ∆S is the change in intensity
ratio. The sensitivity factor, As, takes into account the explicit refractive index
contributions from the surface, the adsorbed material and the aqueous medium.
2.2.1.1. Stagnant point flow
In the stagnant point fluid cell, the collector surface is positioned at a critical dis-
tance from the inlet tube such that a stagnant point flow is generated at the point
where the fluid impinges the surface. Under these conditions the hydrodynamics
can be very well defined [9]. The particles arrive at the surface under the influ-
ence of surface forces and Brownian diffusion only. Solutions were gravity fed
into the reflectometry cell from high-density polyethylene (HDPE) containers
(250 ml) mounted on adjustable laboratory jacks. The height of the liquid above
the inlet port to the fluid cell determined the flow rate. The flow rate was main-
tained at 1.5 cm3/min with a height of the liquid 13 cm above the inlet port. A
valve was used to switch between solutions entering the cell. The cell volume was
30 cm3. A vacuum pump was used to suck excess solution from the cell.
2.2.2. AFM
A sphere attached to the end of an AFM cantilever comprises a colloid probe. A
silica sphere was attached to a cantilever using a heat softening resin (Epikote
1004, Shell) using a micromanipulation arm attached to a metallurgical micro-
scope (Olympus BH2). The cantilevers were silicon nitride, tipless, 200 µm long,
wide legged from Nanoprobe (Park Scientific, USA). Spring constants were de-
termined to be 0.1 ± 0.05 N/m by measuring the resonance frequency of the canti-
levers with added known masses [10]. Prior to measurement the colloid probe was
rinsed with ethanol, dried in a stream of nitrogen and plasma cleaned (Harrick
Plasma Cleaner/Steriliser PDC-32) for 1 minute. A Nanoscope III controller and
Fine particle detachment studied by reflectometry and atomic force microscopy 283
Figure 2. Typical reflectometry raw data showing change in signal with time upon introduction of a
particle suspension at arrow (a) followed by the introduction of a displacing solution at arrow (b).
284 A. Feiler and J. Ralston
equilibrium with the solution in the reservoirs. The baseline signal, S0 was moni-
tored during this time to ensure that the signal was stable prior to measurement.
The background electrolyte solution was permitted to flow into the cell for a fur-
ther 100 seconds after the measurement began. After this time (point (a) in Figure
2), the valve was switched to allow silica particles into the cell (the electrolyte
concentration remained unchanged). The reflectometer signal increases due to the
attachment of the silica particles. Initially the signal increases linearly with time.
The rate of attachment decreases markedly close to saturation coverage. At point
(b), the valve was switched to introduce a new solution, containing either a parti-
cle-free electrolyte solution at high pH or a solution of linear polyphosphate. The
decrease in the signal at point (b) is due to the detachment of the silica particles.
Initially the decrease in signal is very rapid but the rate of detachment slows down
as the maximum detached amount is reached. The signal reaches a “plateau de-
tached amount” before complete detachment of the particles has been obtained.
The attachment of silica particles to a titanium dioxide wafer at pH 4, followed
by their detachment upon introduction of electrolyte solutions at higher pH’s, is
shown in Figure 3. The rate of attachment and the maximum attached amount,
Γmax, was the same for each experiment. The linear attachment regime is indica-
tive of a rate limited only by the mass transport of the particles from solution to
Figure 3. Amount of silica particles attached at the titanium dioxide surface as a function of time af-
ter the introduction of a particle suspension in 10-3 M KNO3 at pH 4 followed (at t = 1000 s) by in-
troduction of particle-free solutions of 10-3 M KNO3 at higher pH’s.
Fine particle detachment studied by reflectometry and atomic force microscopy 285
the surface [16, 17]. A similar observation was made in other reflectometry stud-
ies of nanometre-size particles [8, 18, 19]. No detachment was measured when
electrolyte solutions below pH 6 were introduced into the cell. Upon switching to
electrolyte solutions at pH 6 and higher, detachment of silica particles was de-
tected. Both the rate of detachment and the maximum detached amount increased
with increasing pH. At pH 6, only a small quantity (10%) of silica particles were
detached and the detachment rate was slow compared to initial rate of attachment
at pH 4. At pH 9, half of the pre-attached particles were detached and the initial
rate of detachment was faster than the initial rate of attachment.
In Figure 4 the detachment force measured by AFM between a silica sphere
and titanium dioxide substrate as a function of pH under the same solution condi-
tions as in Figure 3 is presented. A pH dependent adhesion force is evident. Pre-
vious studies have shown [3, 20] that the interaction force between silica and tita-
nium dioxide surfaces is well described by the DLVO theory of colloidal stability.
The adhesion results presented here may be rationalised in terms of the combined
effects of an attractive van der Waals force and pH-dependent electrical double
layer interactions. The isoelectric point (iep) of the silica particles and the tita-
nium dioxide wafers have been measured to be at pH ~ 2 and pH ~ 4.5 respec-
tively [3, 20]. At pH values below the iep of the titanium dioxide, in addition to
Figure 4. Force-distance curves for the retraction of a titanium dioxide wafer from a silica sphere (R
= 3.5 µm) as a function of pH in 10-3 M KNO3. The curves correspond to, from top to bottom, data
at pH 9, 8, 7, 6, and 5.6.
286 A. Feiler and J. Ralston
Figure 5. Amount of silica particles attached at the titanium dioxide surface as a function of time af-
ter the introduction of a particle suspension in 10-3 M KNO3 at pH 4 followed (at t = 1000 s) by in-
troduction of solutions of linear polyphosphates (10-5 M) at pH 4 in 10-3 M KNO3. The subscript in
Pn refers to the number of phosphorous atoms in the molecule.
the attractive van der Waals force, there will be an attractive electrical double
layer interaction between the oppositely charged surfaces. At pH values above the
iep of titanium dioxide both surfaces will be negatively charged and the electrical
double layer interaction will be repulsive. A tensile force is needed to separate the
surfaces from intimate contact at pH 5.6. This is indicated by the negative value
of the normalised force at the point at which the surfaces jump out of contact. At
this pH, the electrical double layer interaction will be weakly repulsive and the
adhesion is due to the attractive van der Waals forces. At higher pH values the
electrical double layer interactions become increasingly repulsive and the surfaces
are seen to separate from contact even in the presence of a positive applied force.
The separation force curves correlate well with the detachment data seen in Figure
3 and explain why no detachment of silica particles was detected below pH 6 and
also why the detached amount increased with pH.
The attachment of silica particles at pH 4 onto titanium dioxide substrates fol-
lowed by their detachment upon switching to solutions of linear polyphosphates
(10-5 M) is presented in Figure 5. As discussed in the Introduction, adsorption of
polyphosphate onto titanium dioxide modifies the surface rendering it negatively
charged. Note that whereas changes in pH affected both the silica and titanium
Fine particle detachment studied by reflectometry and atomic force microscopy 287
Figure 6. Force-distance curves for the retraction of a titanium dioxide wafer from a silica sphere (R
= 3.5 µm) at pH 4 in 10-3 M KNO3 in the presence of linear polyphosphate solutions (10-5 M) from
top downwards P<10>, P3, P2 and P1.
Figure 7. Amount of detached particles measured by reflectometry against the normalised detach-
ment force measured by AFM under similar solution conditions. The symbols refer to the data
measured as a function of pH (♦) and in the presence of polyphosphate (•).
measurements [3] and can be directly related to chemical modification of the tita-
nium dioxide surface due to adsorbed polyphosphate. The increase in the repul-
sive interaction with n is due to an increased negative surface potential on the ti-
tanium dioxide and also the presence of a steric layer, whose thickness, δ,
increases with n (δ ≈ 0.4-0.6 nm for n = 1-3 and δ ≈ 1.6 nm for n = <10>) [3].
A good correlation between the detached amounts of silica particles measured
by reflectometry and the normalised detachment force measured by AFM is seen
in Figure 7. The detached amount of particles is plotted as a percentage of the to-
tal attached amount of particles prior to switching to the displacing solution. The
detached amount of particles increases as the normalised detachment force be-
comes more positive (more repulsive). The detachment of the silica particles is
clearly sensitive to the variation in the electrical double layer interactions brought
about by changes in pH as well as due to adsorbed polyphosphate. The fact that
detachment of silica particles is detected at all in the presence of P1 despite the
force curves showing an adhesional interaction is evidence that the steric layer
due to adsorbed polyphosphate is important in the detachment process. The slight
discrepancy from a linear trend between the detached amount and detachment
force seen for the P<10> data point can be understood in terms of the polydisperse
Fine particle detachment studied by reflectometry and atomic force microscopy 289
Figure 8. The effect of the sequence of addition of silica particles and P3 solution. Curve I is for the
introduction of silica particles followed by the introduction of P3 (10-5 M) at t = 1500 s. Curve II is
for the introduction of silica particles after pretreatment by flowing P3 solution into the cell for 1000
seconds and rinsing with electrolyte solution for 100 seconds. Curve III is for the introduction of a
suspension containing a mixture of silica particles and P3 (10-5 M). All experiments were conducted
in 10-3 M KNO3 at pH 4.
nature of the P<10> sample which contains a range of linear polyphosphates from n
= 1-19. In the reflectometry measurements, the diffusion rate of the polyphos-
phate species becomes a critical factor. The diffusion rate of the polyphosphate
species will decrease with n. Thus, although the larger n species will impart a
more repulsive force on the silica-titanium dioxide system, the smaller n species
will diffuse to the surface more quickly. The consequence is a competition be-
tween the rate of polyphosphate adsorption and the modification of the resultant
polyphosphate adsorption.
Finally, it is of interest from an application viewpoint as to the most efficient
use of a dispersing agent such as polyphosphate in preventing particle attachment.
In Figure 8 the order of addition of silica particles and P3 solution (10-5 M) to the
cell is investigated. Curve I shows the attachment of particles in the absence of
polyphosphate followed by their detachment upon introduction of a P3 solution
(10-5 M). This sequence of addition is identical to that shown in previous figures.
Curve II shows the attachment of silica particles on a titanium dioxide wafer that
290 A. Feiler and J. Ralston
4. CONCLUSIONS
The attachment of nanosized silica particles onto titanium dioxide surfaces and
their subsequent detachment due to changes in solution pH or in the presence of
linear polyphosphates was studied using reflectometry. The rate of detachment
and maximum detached amount increased with both pH and n. It was seen that in-
creasing the pH led to repulsive electrical double layer interactions, which were
responsible for the detachment of the particles. In addition to electrical double
layer interactions, adsorption of linear polyphosphate onto titanium dioxide pro-
vided a steric component, which facilitated the particle detachment. A good corre-
lation was seen between the amount of detached particles and the AFM measured
detachment force. Furthermore it was shown that a solution containing excess lin-
ear polyphosphate could prevent particle attachment all together.
REFERENCES
1. K. L. Mittal (Ed.), Particles on Surfaces 5 & 6: Detection, Adhesion and Removal, VSP, Utrecht
(1999).
2. A. P. Michelmore, W. Gong, P. Jenkins and J. Ralston, Phys. Chem. Chem. Phys., 2, 2985
(2000).
3. A. Feiler, P. Jenkins and J. Ralston, Phys. Chem. Chem. Phys., 2, 5678 (2000).
4. A. Feiler, I. Larson, P. Jenkins and P. Attard, Langmuir, 16, 10269 (2000).
5. Y. K. Leong, P. J. Scales, T. W. Healy, D. Boger and R. J. Buscall, Chem. Soc. Faraday Trans.,
89, 2473 (1993).
Fine particle detachment studied by reflectometry and atomic force microscopy 291
6. Z. Adamczyk, B. Siwek, M. Zembala and P. Belouschek, Adv. Colloid Interface Sci., 48, 151
(1994).
7. J. C. Dijt, PhD thesis, Wageningen University, The Netherlands (1993).
8. M. R. Bohmer, J. Colloid Interface Sci., 197, 251 (1998).
9. T. Dabros and T. G. M. van de Ven, Colloid Polym. Sci., 261, 694 (1983).
10. J. P. Cleveland, S. Manne, D. Bocek and P. K. Hansma, Rev. Sci. Instrum., 64, 403 (1993).
11. W. A. Ducker, T. J. Senden and R. M. Pashley, Nature, 353, 239 (1991).
12. I. Larson, C. J. Drummond, D. Y. C. Chan and F. Grieser, J. Am. Chem. Soc., 115, 11885
(1993).
13. P. G. Hartley, I. Larson and P. J. Scales, Langmuir, 13, 2207 (1997).
14. M. Jaschke and H.-J. Butt, Rev. Sci. Instrum., 66, 1258 (1995).
15. J. N. Israelachvili, Intermolecular & Surface Forces, 2nd ed., Academic Press, London (1992).
16. Z. Adamczyk, L. Szyk and P. Warszynski, J. Colloid Interface Sci., 209, 350 (1999).
17. N. Kallay, M. Tomic, B. Biskup, I. Kunjasic and E. Matijevic, Colloids Surfaces, 28, 185
(1987).
18. M. R. Bohmer, E. A. van der Zeeuw and G. J. M. Koper, J. Colloid Interface Sci., 197, 242
(1998).
19. R. A. Hayes, M. R. Bohmer and L. G. Fokkink, J. Langmuir, 15, 2865 (1999).
20. I. Larson, C. J. Drummond, D. Y. C. Chan and F. Grieser, J. Phys. Chem., 99, 2114 (1995).
This page intentionally left blank
Surface Contamination and Cleaning, Vol. 1, pp. 293–310
Ed. K.L. Mittal
© VSP 2003
Abstract—In Lunar or Martian habitat systems it is impossible to avoid contact with dust. Martian
dust storms, containing submicrometer to 50 µm particles, are an environmental threat to solar cells,
spacecraft, and spacesuits. Because of the high electrostatic charge of the dust and its strong adhe-
sion properties, its deposition onto life support equipment could damage or degrade equipment, re-
ducing the mission duration and endangering personnel. The inhalation of electrostatically charged
airborne dust is also a health hazard to astronauts inside the habitat. Ways to minimize or eliminate
the potential hazards caused by charged particles on space life support equipment are therefore
needed. Specifically, the following topics are discussed in this paper: (1) tribocharging of insulating
materials, (2) the design of a sensor to measure particle size and electrostatic charge distributions of
Mars dust on a single particle basis and in real-time, (3) an experimental plan to minimize deposi-
tion of charged particles on solar cells and life support equipment, and (4) a novel method for re-
moving deposited dust particles.
1. INTRODUCTION
∗
To whom all correspondence should be addressed. Phone: 501-569-8007, Fax: 501-569-8020,
E-mail: mkmazumder1@ualr.edu
294 S. Trigwell et al.
The settling of this dust, especially during a Martian dust storm, can have a
significant effect on the efficiency of solar panels, due to the settled dust imped-
ing the sunlight from the cells. Results from the Materials Adherence Experiment
(MAE) on the Mars Pathfinder mission measured an obscuration of the solar ar-
rays due to dust deposition at a rate of about 0.28% per day [3] with an estimate
that settling dust may cause degradation in performance of a solar panel of be-
tween 22% and 89% over the course of two years [4].
Particles may also settle on the solar arrays by a process known as saltation, in
which particles are lifted from the surface by the wind. These have a size range of
1–200 µm in diameter and an average trajectory of 10 to 20 cm off the surface [5].
Due to the low barometric pressure in the atmosphere, of about 10 mbar, saltation
occurs at wind velocities greater than 15 m/s, which has been recorded at Viking
lander sites [5]. Dust accumulation can also occur due to settling from the atmos-
phere. However, the real deposition rate will also depend on the geographical lo-
cation and from season to season.
The removal of dust settled on an array by natural wind forces on Mars has been
ruled unlikely due to the low atmospheric pressure, which will necessitate high wind
velocities of the order of 35 m/s [6]. The measurements of wind velocities at Viking
sites showed that maximum peak wind velocity was only 25 m/s, with winds over 15
m/s occurring only 1% of the time [6]. Therefore, it was concluded that for long-
duration missions, prevention of deposition or periodic removal of accumulated dust
must be performed to maintain the efficiency of the solar power arrays.
The problem with the design of any mechanism that has to work on a Mars
spacecraft is the hostile environment in which it is expected to perform. The at-
mosphere of Mars is quite different from that of Earth in that it is composed pri-
marily of carbon dioxide (95.3%) with minor amounts of other gases (nitrogen –
2.7%, argon – 1.6%, oxygen – 0.13%, and trace amounts of water and neon) [2].
Although the water content of the atmosphere is about 1/1000 that of Earth, it can
condense out forming clouds and even ground frost in the winter. The most sig-
nificant factor is the temperature on the surface. The average recorded tempera-
ture on Mars is –63°C, with a maximum and minimum of approximately 20°C
and –140°C, respectively. However, the temperature variation depends on the lo-
cation. Temperatures of –133°C are observed at the winter poles, while tempera-
tures as high as 27°C are observed on the dayside during the summer [2].
In this study, the goal was to develop an understanding of the principles of par-
ticle charging and to perform theoretical and experimental studies on the adhesion
and removal of charged particles. Specifically, the following are presented: (1) a
study of the effects of tribocharging of insulating materials and how it can play a
role in dust accumulation on solar panels, (2) design of a sensor to measure parti-
cle size and charge distributions of Mars dust, (3) development a self-cleaning
panel with electrodynamic screens to repel charged dust from settling on solar
panels, and (4) development of an electrostatic wiper type brush for removing de-
posited particles utilizing minimum mechanical parts.
Dust removal from solar panels and spacecraft on Mars 295
2. TRIBOCHARGING OF PARTICLES
There exist two primary mechanisms of charge transfer for contact or tribocharg-
ing between two dissimilar materials. The first is electron transfer in that a linear
relationship is observed between charge transferred during contact of two dissimi-
lar materials. The second is ion transfer in that real surfaces of metals or insula-
tors are covered by adsorbed layers, which are frequently ionic in nature, and that
charge transfer is by positive or negative ion transfer between the materials. A
third theory is postulated that involves material transfer that carries an associated
charge.
Figure 1. (a) The insulator is uncharged with states filled below the neutral level. (b) On contact,
empty states below the Fermi level are filled and the insulator charge is now proportional to φ - φI.
of polymers and insulators regarding the electron energy levels and the role of
impurities is needed.
2.3. Experimental
In order to better understand the factors involved in how the surface properties of
materials influence the charge that can be transferred to a material, the surface
work function and the surface chemical composition of various metallic, ceramic,
polymeric, and mineral materials were analyzed. The metal and polymer materials
used were typical of those that are or may be used in a Mars mission. Pulverized
quartz (SiO2), alumina (Al2O3), and pyrite (FeS2) were used as examples of min-
erals as spectroscopic analysis of Mars dust had shown it to be composed of sili-
cates and iron and magnesium rich sulfates [15]. The samples were analyzed in
the as-received condition with no prior cleaning except for the polymeric materi-
als that were scraped with a clean scalpel blade to expose a fresh surface. The
samples were then analyzed by X-ray photoelectron spectroscopy (XPS) to de-
termine the surface chemistry, and by ultra-violet photoelectron spectroscopy
(UPS) in air to measure the surface work function.
The XPS data were obtained on a PHI Quantum 2000 ESCA Spectrometer us-
ing a focussed monochromatic Al Kα (hν = 1486.7 eV) x-ray source. The x-ray
beam used was a 100 W, 100 µm diameter beam and was rastered over a 1.5 mm
by 0.2 mm area. The survey scans were collected using a pass energy of 117.4 eV
producing a Full Width at Half-Maximum (FWHM) of less than 1.6 eV for the
Ag 3d 5/2 peak. The high energy resolution data were collected using a pass en-
ergy of 23.5 eV, producing a FWHM of less than 0.75 eV for the Ag 3d 5/2 peak.
The collected data were referenced to an energy scale with binding energies for
Cu 2p 3/2 at 932.67 +/– 0.05 eV, and Au at 84.0 +/– 0.05 eV. On some insulating
samples, positive charging of the surface was observed due to the loss of elec-
trons, causing the peaks to shift during data acquisition For these cases, low en-
ergy electrons were used to flood the specimen to neutralize the surface.
The UPS data were obtained on a Riken Keiki AC-2 UV photoelectron spec-
trometer. The samples and detector were placed in open air. The UV source was a
deuterium (D2) lamp with a spot diameter of 2 mm by 2 mm. For the samples with
a high efficiency of photoemission (the metals and graphite) the light source
power was 49.9 nW, and for those samples that have a low efficiency of photo-
emission (the polymers, coal, pyrite, and vitrinite) the light source power was in-
creased to 600.2 nW. The resolution of the instrument for precision measurements
is given as 0.02 eV. The samples were analyzed at a temperature of 22°C, a rela-
tive humidity of 40%, and a pressure of 1 MPa.
2.4. Results
The XPS data are presented in Table 1. The relative atomic concentrations of the
observed elements as reported in Table 1 were obtained by integrating the area
under each peak of interest and normalizing with sensitivity factors supplied by
Dust removal from solar panels and spacecraft on Mars 299
Table 1.
Surface element concentrations (atomic %) of selected materials as measured by XPS
C O N Na Si Cu Fe Cr Ag Al Mg F Ca Zn Cl S
Copper 70 21 – – 1.4 7.4 – – – – – – – – – –
316L steel 53 35 0.8 – 0.9 – 8.7 1.7 – – – – – – – –
Electro. 316L 69 24 2.1 1.8 1.3 – 0.8 0.7 – 1.1 – 0.2 – 0.3 0.3 –
Silver 40 16 – – – – – – 40 – – – – – 2.6 –
Aluminum 23 53 – – – – – – – 24 1.2 – – – – –
PTFE 30 0.2 – – – – – – – – – 70 – – – –
Nylon 66 79 13 7.8 – 0.2 – – – – – – – – – – –
Polystyrene 75 21 2.4 1.2 – – – – – – – – – – – 0.7
Glass 15 55 – 9.5 17 – – – – – 2.9 – 0.3 – – –
SiO2 9.9 64 – 0.6 26 – – – – – – – 0.7 – – –
Pyrite 51 27 0.4 0.6 – – 5.4 – – – – – – – 1.4 14
Graphite 98 2.3 – – – – – – – – – – – – – –
Polycarbonate 84 16 – 0.1 0.2 – – – – – – – – 0.1 0.1 –
Acrylic 73 27 – – – – – – – – – – – – – –
the instrument manufacturer. The data showed that the metal specimens had par-
ticularly high levels of surface contamination (carbon and oxygen) compared to
the ceramic and polymeric specimens. For example, copper showed 70 atomic %
of carbon on the as-received surface, compared to the quartz (SiO2) specimen
with only 9.9 atomic % of carbon. For PTFE, the XPS data showed the surface
composition to be 30 atomic % carbon and 70 atomic % fluorine, which is very
close to the CF2 stoichiometric composition of PTFE. This supports the fact that
polymers do not pick up surface contamination in air as readily as metals. The
graphite specimen shows only a minimal oxygen concentration (2.3 atomic %).
The measured UPS data are shown in Table 2. The value of the work function
for each material is compared with the value reported in the literature.
It was observed in Table 2 that the measured work function for each material
was higher than the reported values. A closer examination showed that the ratio of
the measured-to-reported work function value for copper (1.17:1) was higher than
for the other metals; silver (1.08:1), and aluminum (1.05:1); and these metals
showed less carbon surface contamination by XPS than the copper. In contrast,
the measured-to-reported work function value for PTFE was approximately unity,
and the PTFE showed no discernible carbon contamination by XPS. Clearly, a
correlation between the amount of surface contamination and increase in work
function can be observed.
300 S. Trigwell et al.
Table 2.
Measured work functions by UPS in air compared to work function values reported in the literature.
Measurement errors were +/– 0.002 eV
(a)
(b)
(c)
Figure 3. High resolution XPS scans of peaks for pyrite (FeS2). a) Carbon C1s peak, b) Sulfur S2p
peak, and c) Iron Fe 2p peak.
302 S. Trigwell et al.
The XPS survey spectrum shows the peaks associated with the surface compo-
sition. In addition to the peaks for carbon and oxygen, minor amounts of sodium,
nitrogen, and chlorine were also detected. In the high resolution scans, the peak
for each element was curve fitted with the component peaks as reported in nu-
merous reference literature and data bases [16-18]. What can be observed from
the high resolution peaks is that the carbon present on the mineral is predomi-
nantly C-C or C-H bonding indicating carbonaceous contamination, with minor
amounts of carbon-oxygen species. A small amount of organic carbon-sulfur was
also detected. However, the sulfur peak shows the sulfur to be mainly in the sul-
fate form with very little, if any, iron sulfide present. This is confirmed in the iron
peaks where the iron is present predominantly as iron sulfate, and in this case no
iron sulfide is detected. The data show that for the case of pyrite what is actually
chemically present on the surface is very different from the bulk composition,
which can affect the value of the work function. This is important in understand-
ing how materials charge relative to each other in the triboelectric series. How-
ever, these data were taken in an Earth environment, and so it is of great interest
to determine what the surface composition of minerals would be in a Martian en-
vironment.
3. DESIGN OF SENSOR
There are a number of instruments that can be used to characterize the aerody-
namic size distribution of particles. Instruments such as a Faraday cup are avail-
able to estimate the net average electrostatic charge on particles samples. How-
ever, the choice of instruments for real-time simultaneous measurements of both
aerodynamic diameter and electrostatic charge distributions of particles on a sin-
gle particle basis is limited. The Electrical Single Particle Aerodynamic Relaxa-
tion Time (E-SPART) analyzer is used extensively for simultaneous characteriza-
tion of particle size distribution (PSD) and electrostatic charge distribution [19].
The analyzer can be used in the diameter range from 0.5 to 50 µm and charges in
the range from 0 to their saturation charge limit.
The E-SPART analyzer, as shown in Figure 4, uses an AC electric drive to os-
cillate the particles in air. The resultant oscillatory motion of the particle lags be-
hind the external AC field. The phase lag (φ) relates to the aerodynamic diameter
of the particle, and the amplitude of the particle trajectory determines the particle
charge and polarity, as shown in Figure 5(a). Figure 5(b) shows a still from a
video image taken of charged particle tracks in the chamber. The particle tracks
are analyzed by Laser Doppler Velocimetry. The details of the operation of the
instrument are available [20-23].
Figures 6(a) and (b) show typical particle size and charge distribution, respec-
tively, for a positive copier toner with a mean particle size of 8 µm. Although the
data show the overall particle count, individual particle data can be extracted.
Dust removal from solar panels and spacecraft on Mars 303
(a)
(b)
Figure 5. (a) Principle of operation of E-SPART, and (b) video image of particle tracks. The E-
SPART analyzer uses an AC electric drive to oscillate the charged particles.
304 S. Trigwell et al.
(a)
(b)
Figure 6. (a) Particle size distribution, and (b) charge distribution for a positive toner.
Dust removal from solar panels and spacecraft on Mars 305
Dust settling out of the atmosphere onto any horizontal surface is a potential prob-
lem in the obscuration of solar arrays. A method is required to periodically re-
move the dust, or prevent the dust from settling in the first place. Ideally, a
method that requires no moving parts and is robust in operation is most desirable,
as shown in Figure 7.
The static charges on the particles provide an opportunity to prevent dust deposi-
tion by using an AC voltage driven electrode screen. This type of screen creates a
repelling force to the charged particles regardless of their polarity. The device con-
sists of an electrode screen that contains a number of parallel electrodes placed
equidistant from each other, embedded in a insulating coating as shown in Figure 7.
Figure 8. Property change versus temperature for martensitic transformation in NiTi alloy.
lower transformation temperatures (< –50°C) are available. Data from the Path-
finder mission showed a high level of consistency in the surface temperature
range at the Ares Vallis landing site [25], as shown in Figure 9. From these data, a
NiTi alloy with a transformation temperature of approximately –50°C (223°K)
would suffice.
308 S. Trigwell et al.
(a)
(b)
(c)
Figure 10. (a) NiTi spring in tension. (b) Upon heating through the transformation temperature, and
(c) after heating.
Dust removal from solar panels and spacecraft on Mars 309
6. CONCLUSIONS
The tribocharging of Mars dust can contribute to strong adhesion of dust particles to
solar panels, spacecraft, and spacesuits. Both the polarity of the charged dust parti-
cles as well as the amount of charge depend upon the surface composition of the
particles and the contacting materials. It has been shown that in an ambient Earth
environment, surface contamination and oxidation produce significant changes to
the surface composition and hence the work function of a variety of materials.
An instrument has been developed that can simultaneously measure both size
and charge distributions on an individual particle basis. A smaller, more robust
version of which is proposed to measure size and charge distributions of Mars
dust in situ.
An electrodynamic screen shows promise for preventing deposition of charged
dust particles, and an electrostatic brush for dust removal has been developed us-
ing NiTi shape memory alloy. The electrodynamic screen and cleaning device
proposed have no mechanical parts so the probability of failure is minimized.
will be investigated against stainless steel and PTFE to determine the charging
characteristics in a Martian type environment. Simultaneously, an Ultra-violet
Photoelectron Spectrometer is being developed to allow work function measure-
ments to be taken as a function of relative humidity within the chamber.
A compact E-SPART analyzer that is robust enough for space flight is also be-
ing developed, as well as the development of the electrodynamic screen is being
continued.
REFERENCES
1. G.A. Landis, Paper presented at the Intersociety Energy Conversion Engineering Conference,
Honolulu, HI, July 27–August 1, 1997.
2. Mars News, www.marsnews.com/planetology, November 1999.
3. G.A. Landis and P.P. Jenkins, Proc. 26th IEEE Photovoltaic Specialists Conference, 865-869
(1997).
4. G.A. Landis, Acta Astronautica, 38, 885-891 (1996).
5. R. Greeley, N. Lancaster, S. Lee and P. Thomas, Mars, pp. 835-933, University of Arizona
Press, Tuscon, AZ (1992).
6. J.R. Gaier, M.E. Perez-Davis and M. Marabito, Paper presented at the 16th AIAA/NASA/
ASTM/IES Space Simulation Conference, Albuquerque, NM, November 5–8, 1990.
7. J. Cross, Electrostatics: Principles, Problems and Applications, Adam Hilger, Bristol, England
(1987).
8. Y. Murata, Jap. J. App. Phys., 18, 1-8 (1979).
9. J. Lowell and A.C. Rose-Innes, Adv. Phys., 29, 947-1023 (1980).
10. C.B. Duke and T.J. Fabish, Phys. Rev. Lett., 37, 1075-1078 (1976).
11. D.A. Hays, Proc. International Conf. on Modern Electrostatics, Ruinian Li (Ed.), Beijing, China,
pp. 327-330, International Academic Publishers, New York (1988).
12. H. Bauser, Dechema Monographs, 72, 11-28 (1974).
13. W.R. Harper, Contact and Frictional Electrification, Laplacian Press, Morgan Hill, CA (1998).
14. D.M. Taylor and P.E. Secker, Industrial Electrostatics: Fundamentals and Measurements, John
Wiley & Sons, New York (1994).
15. C.D. Cooper and J.F. Mustard, Paper #6164 presented at The Fifth International Mars Science
Conference, Pasadena, CA (1999).
16. Ph. De Donato, C. Mustin, R. Benoit and R. Erre, Appl. Surface Sci., 68, 81-93 (1993).
17. C.D. Wagner, W.M. Riggs, L.E. Davis and J.F. Moulder, Handbook of X-Ray Photoelectron
Spectroscopy, Perkin-Elmer Corp., Eden Prairie, MN (1983).
18. NIST XPS Database, http://srdata.nist.gov/xps/Bind_e_spec_query.asp
19. M.K. Mazumder and R.E. Ware, US Patent #4633714 (1987).
20. P.A. Baron, M.K. Mazumder and Y.S. Cheng, in: Aerosol Measurements: Principles, P. Baron
and K. Willeke (Eds.), Chap. 17, Van Nostrand Reinhold, New York (1992).
21. M.K. Mazumder, S. Banerjee, R.E. Ware, C. Mu, N. Kay and C.C. Huang, IEEE Trans. Ind.
Applications, 30, 365-369 (1994).
22. M.K. Mazumder, S. Banerjee and C. Mu, in: Dispersion and Aggregation, B.M. Moudgil and P.
Somasundaran (Eds.), Engineering Foundation, New York (1994).
23. M.K. Mazumder, R.E. Ware, J.D. Wilson, R.G. Renniniger, F.C. Hiller, P.C. McLeod, R.W.
Raible and M.K. Testerman, J. Aerosol Sci., 10, 561-569 (1979).
24. C.R. Wayman, MRS Bull., 49-56, April 1993.
25. Mars Pathfinder website: http://www.mars.jpl.nasa.gov (1997).
26. S. Trigwell, R.D. Hayden, K.F. Nelson and G. Selvaduray, Surface Interface Anal., 26, 483-489
(1998).
Surface Contamination and Cleaning, Vol. 1, pp. 311–334
Ed. K.L. Mittal
VSP 2003
Abstract—We report on experiments on the underlying physical mechanisms in the Dry- (DLC)
and Steam Laser Cleaning (SLC) processes. Using a frequency doubled, Q-switched Nd:YAG laser
(FWHM=8 ns) we removed polystyrene (PS) particles with diameters in the range of 110 nm to
2000 nm from industrial silicon wafers by the DLC process. The experiments have been carried out
both in ambient conditions as well as in high vacuum (10−6 mbar) and the cleaned areas have been
characterized by atomic force microscopy for damage inspection. In DLC we have determined the
cleaning laser fluence thresholds for a large interval of particle sizes. Additionally we could show that
particle removal was due to a combination of at least three effects: substrate thermal expansion, local
substrate ablation as a consequence of field enhancement at the particle, and explosive evaporation
of moisture adsorbed from the air. Which effect dominates the process depends on the boundary
conditions. For our laser parameters no damage-free DLC was possible, i.e. whenever a particle was
removed by DLC we damaged the substrate by local field enhancement. In our SLC experiments
we determined the amount of superheating of a liquid layer adjacent to surfaces with controlled
roughness. On silicon wafers the water layer could be superheated to 250 ◦ C prior to the onset of laser
induced bubble nucleation. The heat transfer from the silicon substrate into the liquid was found to be
limited by a thermal boundary resistance which can be characterized by a heat transfer coefficient of
3 · 107 W/(m2 K). Based on the knowledge about the particle removal mechanisms and the determined
cleaning efficiency we discuss the advantages and disadvantages of DLC and SLC as possible future
industrial surface cleaning procedures.
1. INTRODUCTION
The removal of particle contamination from surfaces is one of the crucial prerequi-
sites for a further increase in the integration density of ICs and for the progress in
nanotechnology. At all stages of the production of ICs, e.g., from the bare Si wafer
to the patterned chip, particles even smaller than 100 nm in size can cause a damage
to the produced structure and hence be responsible for the failure of the final device.
In the late 1980s, the experts in the field of cleaning technology predicted that
traditional cleaning methods such as ultrasonics and wet techniques would reach
their limit of capability [1, 2]. In addition, these traditional techniques were and
still are harmful to the environment as they consume large quantities of aggressive
chemicals and water. Although the traditional methods have been continuously
improved [3], still particle contamination causes considerable production losses
[4], and with further shrinking of line widths [5] there is a definite need to replace
traditional methods by new cleaning technologies.
One of these new approaches is called laser cleaning. In Dry Laser Cleaning
(DLC) [6– 8] the surface to be cleaned is irradiated by a short laser pulse. In Steam
Laser Cleaning (SLC) [6, 7, 9, 10] prior to the application of the laser pulse a liquid,
e.g. a water-alcohol mixture, is condensed onto the surface.
After the first attempts of implementing laser cleaning in prototype cleaning
tools [11], this strategy was not pursued any further as there were too many open
questions related to the underlying physics. In the following years several research
laboratories around the world [12– 20] started to investigate the physical processes
involved both in SLC and DLC, starting from the simple scenarios suggested by
the authors of the first publications on the subject. These scenarios also formed the
basis for certain models to describe laser cleaning and to interpret the experimental
results obtained [6, 7, 10, 18, 19, 21– 33]. However, recent experiments [31, 33–
41] show that both the DLC and SLC scenarios that have been taken as common
sense so far do not incorporate all the important cleaning mechanisms and hence
are oversimplified.
In this article we will first summarize our knowledge on the cleaning processes
involved in laser cleaning and their interplay, and then present the results of
systematic measurements of cleaning efficiencies in both DLC and SLC for particle
sizes from 110 nm up to 2000 nm. The interpretation of these results will clearly
point out the importance of the cleaning mechanisms neglected in the original SLC
and DLC scenarios. Against this background we will discuss briefly the state of
theoretical modeling of laser cleaning. Based on the previous sections we will finish
the article with a statement of the prospects and problems of laser cleaning as an
industrial cleaning process from today’s state of knowledge.
2. EXPERIMENTAL ASPECTS
Figure 1. Typical sample as used in the laser cleaning experiments imaged in a scanning electron
microscope. The displayed area is 4.8 µm × 4.8 µm and the particle size is 110 nm.
removal efficiencies for various, well-defined sizes. Their spherical shape addition-
ally facilitates a comparison with theoretical models, as adhesion forces of particles
are mostly calculated for the geometry of a sphere on a flat substrate. Some ex-
periments were also performed using irregularly shaped Al2 O3 particles (Summit
Chemicals Europe GmbH, Düsseldorf, Germany) as contaminants.
As substrate we used industrial silicon (100) wafers (Wacker Siltronic, Burghau-
sen, Germany) that were cleaned in isopropyl alcohol (IPA) in an ultrasonic bath
before applying the contaminants.
The particles were deposited on the silicon substrate by a spin coating process,
described in detail in [17, 41]. We were able to prepare samples with more than
95% of isolated spheres at particle densities above 1000 per cm2 . A typical example
can be seen in Fig. 1 where 110 nm sized PS particles were deposited onto a
Si wafer. Care was taken to prevent particle agglomeration, which is important
for quantitative experiments, as agglomerates exhibit a different cleaning behaviour
compared to single particles [15, 35].
The determination of laser fluences for the nanosecond pulses is described in detail
in [17]. Briefly, the laser fluence was determined relative to the well-known melting
threshold fluence of Si, making use of the higher reflectivity of the molten with
respect to the solid phase. This was done by time-resolved monitoring of the
reflected light of a HeNe laser (λ=633nm). Simultaneously the laser’s scattered
light was detected in order to probe the cleaning efficiency.
The techniques described above are not suitable for determining laser fluences
in bulk liquids. Hence for the bubble nucleation experiments we chose another,
intrinsic laser fluence calibration method. During the experiment we found a distinct
change in the reflectivity of the water/silicon system. As computer simulations of
the temperature dependency of this reflectivity show, this change is primarily caused
by the temperature change in the water. Based on this, it is possible to derive [42]
that the integral
9µs
Rp (t)
W := − 1 dt ∼ F (1)
4µs R0,p
containing the reflectivity change in p-polarization (R0,p initial reflectivity, Rp (t)
reflectivity after bubble nucleation) is directly proportional to the applied laser
fluence F . The upper boundary of the integral is given by the end of the detection
time, the lower boundary of 4 µs was chosen as the time where no gas bubbles
were present on the surface any more. Hence, the applied local laser fluence right
in the area where bubble nucleation is probed can be determined from the measured
reflectivity changes of the system.
Laser cleaning of silicon wafers: Prospects and problems 315
Figure 3. Optical determination of laser induced bubble nucleation (for details see text).
Already the authors of the first publications on laser cleaning [6, 7, 9, 10] suggested
physical mechanisms to describe the particle removal process. In DLC the removal
process was ascribed to the thermal expansion of the substrate due to the heating
with the laser pulse. SLC was explained as a consequence of the explosive
evaporation of the applied liquid. However, recent research [40, 44, 45] in the field
of laser cleaning has shown that several other mechanisms are of importance for
the process. In the following sections we will describe the cleaning mechanisms
experimentally verified so far.
From this it is clear that a measurement of the dynamics of the substrate expansion
upon illumination with a laser pulse provides insight into this cleaning mechanism.
Detailed discussions of our experiments can be found in [41, 43], therefore we will
only highlight the results here.
Using an interferometric setup as described in Section 2 we measured the
expansion of a silicon substrate after illumination with a Nd:YAG laser pulse
(FWHM = 10-20 ns, λ=532 nm) with ns time resolution. Typical values of this
displacement were a few nm. By numerical derivation of the averaged displacement
as a function of time of 600 individual experiments we obtained the deceleration of
the substrate surface. This deceleration of the surface, rather than its acceleration as
discussed by most authors in the description of DLC, is the relevant parameter for
the particle removal [45]. For typical laser cleaning fluences of a few 100 mJ/cm2
the decelerations are in the order of about 107 m/s2 , just the order of magnitude that
is thought to be necessary for removal of such small particles [7].
The experimental values are in good agreement with a simple theoretical descrip-
tion of the thermal expansion of the substrate. Taking into account only the 1D
heat equation2 and denoting by R the reflectivity, by α the linear thermal expansion
coefficient, by CP the specific heat of the substrate, by ρ its density, by f (t) the
time dependent normalized intensity of the laser, and by F its fluence the surface
displacement is given by
t
α
d(t) = (1 − R)F f (t ) dt (2)
ρ CP −∞
ns2 F
amax = −1.71 · 106 g 2 τ2
(3)
mJ/cm
2 Thermoelastic effects [46, 47] are not taken into account, as the area where the displacement is
detected by the interferometer was at least ten times smaller than the illuminated spot and was located
in its center. Thus lateral thermal gradients are neglected at this spot.
318 M. Mosbacher et al.
Figure 4. Intensity (E 2 ) of the electromagnetic field around a PS particle (diameter 1700 nm,
refractive index 1.59) in vacuum when illuminated by a laser at λ=800 nm. The laser enters from
the left, and the light is polarized in the image plane. At the light averted side (i.e., not directly
illuminated by the laser) of the particle the intensity is enhanced by a factor of about 30.
Figure 5. Surface damage caused by local melting and ablation of the substrate. The melting was
induced by the enhancement of laser intensity in the near field of the particles at the surface. Shown
are AFM images in top view (left) and a 3D illustration (right) of the same site.
320 M. Mosbacher et al.
Although laser cleaning has been known for more than ten years now, still many
questions related to the underlying physical processes are not answered yet. The
authors of the first publications on the subject [6– 10] suggested simple physical
scenarios that were accepted for the interpretation of experimental data thereafter
and were taken as a basis for theoretical modeling: DLC was explained solely by
the thermal expansion of the substrate and the adhering particle, SLC by explosive
evaporation of the liquid condensed onto the surface [6, 7, 10, 18, 19, 21– 33].
However, it is still not clear whether these scenarios accurately reflect the actual
particle removal process for all laser parameters and cleaning environments studied
so far, even in the experiments that are quoted to justify the models. This is
because the models suffer from major drawbacks, such as the neglect of the cleaning
mechanisms field enhancement induced local substrate ablation.
In addition to the thermal expansion of the substrate the early publications on
laser cleaning [6, 7, 9, 10] suggested a second mechanism for the removal of dirt
particles: the explosive evaporation of a liquid such as water or alcohol. Here the
liquid layer is heated directly [10] (laser energy absorbed in the liquid) or indirectly
[6, 7] (laser energy absorbed in the substrate and transferred to the liquid via heat
conduction) by the laser pulse. Due to the short time scale of this heating the liquid
is superheated to a certain extent, becomes intrinsically unstable and subsequently
evaporates explosively. The momentum transfer of this explosion onto the particle
is then thought to lift the particles off the surface. In the following, we will discuss
this cleaning mechanism in three variations: the evaporation of liquid adsorbed at
the particle-surface area as it is the case for DLC in ambient air, the evaporation of
a liquid film as it is found in SLC, and the laser induced nucleation of bubbles in a
bulk liquid that provides considerable information on the underlying physics.
Laser cleaning of silicon wafers: Prospects and problems 321
Typically, these studies were carried out on rather rough surfaces such as metal
films. Here quite moderate superheatings of about 20 K were found to be
sufficient for the nucleation of laser induced bubbles, a value much lower than the
theoretically predicted one of 200 K for water [57]. This has been attributed to the
surface roughness of these substrates. But to our knowledge there are no systematic
studies that confirm this roughness-dependency of the threshold. Yet this is the
crucial point in the transfer of the obtained results from the water/metal film system
to the water/silicon wafer system under consideration in laser cleaning studies. With
respect to this several open questions arise: 1) Supposing rough metal films provide
nucleation sites for gas bubbles and cause only moderate superheatings of the liquid,
what will in contrast be the degree of superheating on smooth substrates like silicon
wafers? 2) Is it justified to determine the temperature of the superheated liquid
by measuring [58] or computing [28] the temperature of the silicon substrate and
simply assuming a perfect heat transfer from the substrate into the liquid?
3.3.3.1. Bubble nucleation thresholds. We, therefore, studied [40, 42] the nu-
cleation of bubbles at a superheated liquid/solid interface under controlled surface
roughness. When the incident laser fluence reached a well defined threshold, scat-
tered light was observed. This indicates a sharp nucleation threshold. The thresh-
old decreases with increasing starting temperature T0 of the water as less energy
is needed to reach the nucleation temperature. Figure 6 shows these temperature
dependent thresholds for the examined systems. The lines are linear extrapolations
to vanishing fluence and yield the nucleation temperature by their intersection with
Figure 6. Comparison of the bubble nucleation thresholds and extrapolated superheating limits of
water on substrates of different roughnesses.
Laser cleaning of silicon wafers: Prospects and problems 323
Figure 7. Computed maximum temperatures in water on a silicon surface as function of the heat
transfer coefficient (HTC) ζH2 O for different starting temperatures T0 . As laser fluence we used the
experimentally determined threshold fluence for bubble nucleation.
different for each starting temperature. It can be seen from these computations that
for all starting temperatures of the water the graphs of the calculated maximum
temperatures as a function of the assumed heat transfer coefficient intersect at
one single point. This point is given by a value of ζH2 O = 3 · 107 W/(m2 K)
and a maximum water temperature of 250 ◦ C – just as determined experimentally.
Therefore, this ζ -value represents the heat transfer coefficient in the studied system.
Interestingly, the computations exclude values of ζ < 3 · 106 W/(m2 K), as in this
case the equilibrium boiling temperature of 100 ◦ C is not reached for all starting
temperatures.
4. EFFICIENCY MEASUREMENTS
4.1. Steam laser cleaning
For the SLC we used [17] the same experimental setup as in the DLC experiments,
but supplemented it with a steam providing unit [11]: a controlled flow of filtered,
pressurized air was directed through a reservoir of a water/IPA mixture (90% water)
heated to 330 K. Then the steam/air mixture was directed to the sample via a nozzle
at a distance of 1.5 cm from the area to be cleaned. The IPA’s role was to improve
the wetting of the steam condensed onto the silicon wafer leading to the formation
of a liquid film. We estimated the film thickness using ellipsometric measurements
to be about 200-400 nm.
The cleaning process was found [35] to be statistical in a way that the number Nr
of the remaining particles after n cleaning steps is given by Nr /N0 = (1 − p)n as
function of the single shot cleaning efficiency p and the original particle number N0 .
Therefore, in Fig. 8 we plotted only the cleaning efficiencies after 20 cleaning steps
(steam and laser pulse) for the sake of clarity. This figure shows the dependence
of the cleaning efficiency on the applied laser fluence for particles of different sizes
(60 nm-800 nm), materials (PS, SiO2 , Al2 O3 ) and geometries (spherical PS, SiO2
and arbitrarily shaped Al2 O3 ).
Some important results can be obtained from this graph. A predominant feature
is the existence of a universal cleaning threshold for all particles investigated
at a laser fluence of about 110 mJ/cm2 . For slightly higher laser fluences we
Figure 8. Cleaning efficiency for various colloidal particles in SLC with λ=532 nm, FWHM=7 ns.
The cleaning threshold is found to be the same for all particles.
326 M. Mosbacher et al.
observed a very steep increase in the cleaning efficiency, and values above 90%
after 20 cleaning steps (laser pulse plus steam jet) are reached well below the
melting threshold of a bare silicon substrate, even for particles as small as 60 nm.
The threshold cleaning fluence of 110 mJ/cm2 found in SLC is larger than the
threshold fluence of 80 mJ/cm2 determined for bubble nucleation in bulk water
(see. Sec. 3.3). This needs further experimental investigation, however there are
two possible explanations. First, the bubble nucleation process may depend on the
film thickness, and hence there would be differences in the process dynamics when
comparing bulk liquid to liquid films. Second, it is not clear so far whether the mere
nucleation of bubbles is sufficient for an efficient cleaning or whether more laser
fluence is needed to create larger or faster growing bubbles.
Figure 9. Thresholds in the applied laser fluence for particle removal in DLC in ambient air. Particles
smaller in diameter than 110 nm could not be removed.
Laser cleaning of silicon wafers: Prospects and problems 327
of the structure of the silicon wafer, i.e. the silicon substrate and the native oxide
layer of specified thickness, as induced by melting has to be strictly avoided. From
experiments [63] this melting threshold is known to be about 280 mJ/cm2 for our
laser parameters, which therefore represents the upper limit of applicable laser
fluence. The second threshold, also indicated in Figure 9, is the cleaning threshold
of the SLC process as discussed in Section 4.1.
In order to obtain information on the dependence of the cleaning threshold on
the particle size we investigated many different particle diameters in the range of
110-2000 nm. It turned out that only by the investigation of this large variety of
particle sizes some main characteristics of the dependence of the cleaning threshold
on the particle diameter could be revealed. At first sight the shape of the curve
roughly follows a 1/rk -trend, r denoting the particle radius and 1< k <2. This
monotonic behavior was predicted by the first DLC models, and in fact already the
first publications on DLC reported that smaller particles were harder to remove than
larger ones due to the nature of the adhesion forces [6– 8].
Taking a closer look, however, one discovers an oscillating behavior of the thresh-
old fluences which we attribute to optical resonances as discussed in Section 3.2.3.
This is illustrated in the graph by the line connecting the data points. However, it
should be pointed out that this line is just a guide to the eye and does not describe
the field enhancement efficiency as function of the particle diameter. The number
of particle sizes used in our experiments is not sufficient to resolve this dependency.
In DLC using sub-ns pulses the removal of a particle is always accompanied by
the formation of a hole [38], i.e. the hole formation threshold is identical with
cleaning threshold. From this, one can conclude that here the dominant cleaning
mechanism is local substrate ablation. Against the background of field enhancement
as the origin of surface damage, it is a natural consequence also for nanosecond laser
pulses to determine not only the cleaning threshold fluences in DLC, but also the
local melting/ablation thresholds. The latter ones, instead of the melting threshold
of the bare silicon surface, represent the true upper limit for the applicable laser
fluence and are by their nature particle dependent. For its determination we made
use of the Gaussian spatial beam profile of our cleaning laser. Due to this profile a
spatial variation of the position in the cleaned area corresponds to a variation in the
locally applied laser fluence. In a post process analysis we investigated the cleaned
areas of our samples with an atomic force microscope (AFM, Digital Instruments).
By imaging damage sites such as displayed in Figure 5 at different locations in
the cleaned areas and especially at their borders, corresponding to the cleaning
threshold fluence, we determined the damage threshold for each particle size. For
all particles investigated the cleaning threshold was identical with the damage
threshold. Damage-free DLC was not possible by applying the laser parameters
we used!
The AFM images of damage sites contain even more information on the particle
removal mechanism as they reveal quantitative topographical information. All the
investigated damage sites showed the same features: a “trench” surrounding a
328 M. Mosbacher et al.
Figure 10. Trench depth at the damage sites as a function of the particle size when the particles were
removed in DLC applying the threshold cleaning fluence. For particles larger than 250 nm in diameter
the depth increases strongly with the particle size, for smaller particles it remains almost constant.
central “hillock”. Generally speaking, at high laser fluences the hillock was lower
and the trench deeper, whereas for low laser fluences a hillock was detectable but
the trench almost disappeared. In Figure 10 we plotted for the investigated particles
the mean trench depths for damages that occurred at the cleaning/damage threshold
in a double logarithmic graph.
This plot clearly shows two regimes: for particles smaller than about 250 nm in
diameter the depth remains almost constant at about 1 nm. For larger particles we
found a strong increase in the trench depth and the volume of the hillock was smaller
than that of the trench, i.e., ablation of substrate material had taken place.
From this observation we conclude that even for DLC using ns pulses local
ablation of the substrate plays a role as a cleaning mechanism for “large” particles
where the field enhancement is high and thus provides fluences high enough for
ablation. For smaller particles field enhancement probably causes local melting, but
no ablation at the threshold cleaning fluence.
Figure 11. Comparison of the cleaning thresholds in ambient air and high vacuum. In vacuum the
thresholds are distinctly higher.
The results and the comparison with the ambient values are presented in Figure 11.
In vacuum the laser fluences necessary for particle removal are higher for particles
smaller than about 800 nm in diameter. For larger particles no difference in the
threshold was detected between ambient conditions and HV.
We attribute this to the predominance of different cleaning mechanisms for differ-
ent particle sizes. Cleaning in ambient conditions is facilitated by explosive evapo-
ration of adsorbed moisture from the air. When exposed to vacuum for several hours
the amount of moisture at the particle-surface contact is significantly decreased, the
contribution of explosive evaporation to the cleaning forces decreases, and conse-
quently the cleaning threshold increases. This argument should also be valid for
particles larger than 800 nm in diameter. However, as discussed above, for large
particles local ablation induced by field enhancement appears to strongly contribute
to the cleaning process. And, of course, this is the case as well in ambient air as in
vacuum and hence no large difference in the cleaning thresholds for the two differ-
ent environments is detected.
Against the backdrop of the results presented above we now discuss the prospects
as well as the problems of using laser cleaning as an industrial cleaning technique.
In this context we would like to highlight three main aspects: the systematic
determination of cleaning thresholds, the role of different cleaning mechanisms,
and the consequences for the applicability of theoretical models proposed so far.
330 M. Mosbacher et al.
whenever a particle was removed in DLC. This means that damage-free DLC was
impossible with our laser parameters. In fact, the local substrate ablation combined
with a momentum transfer to the particles was found to be the dominating cleaning
mechanism with large size parameters and sub-nanosecond laser cleaning [40].
A second consequence of field enhancement also argues against a technological
application of DLC. The DLC results obtained both in ambient conditions and in
HV confirm a general trend already obtained by other authors [6, 7, 19, 31, 33, 64]
and predicted by their models: cleaning thresholds for smaller particles tend to
be higher than for larger ones. Yet this is not a strict rule. In contrast to the
above cited experiments we used a large variety of particle diameters ranging from
110 nm to 4100 nm and could show that the dependence of the cleaning threshold
as a function of the particle diameter was non-monotonous as a consequence of the
optical resonances in the near field of the particles. This non-monotonous behaviour
in DLC makes it difficult to apply the correct cleaning fluence for the removal of a
specific particle size, unlike the universal threshold in SLC.
It should be pointed out that field enhancement is an intrinsic physical effect that
cannot be avoided in laser cleaning. However, it can be minimized by choosing
suitable wavelengths to reduce the size parameter or by carrying it out in an
environment with a refractive index close to that of the particles.
5.3. Models
Models which describe the laser cleaning process accurately would be an important
tool in the real world application, as they could predict the optimum cleaning
conditions for various substrates, particles, lasers and cleaning environments.
However, such models must inevitably incorporate all the knowledge gained
regarding the physical processes behind both SLC and DLC.
So far in DLC the particle removal has always been solely ascribed to the
thermal expansion of the substrate. Field enhancement was taken into account
only as increased laser fluence [19, 34], but not as the origin of an additional
cleaning mechanism via local ablation. The third cleaning mechanism, evaporation
of adsorbed ambient moisture, is not incorporated into any model published so
far. The latter is even more important, as most of the experiments carried out
to compare with laser cleaning models were performed in ambient conditions
[22, 24, 26– 28, 30, 64]. Only in [19] the experiments were conducted in vacuum,
and Vereecke et al. [31] varied the relative humidity and reported a higher cleaning
efficiency at increased humidity levels. In addition to this, only recent models
[44, 45] consider the elastic properties of the substrate/particle system which might
have a great influence on the cleaning process as well.
Future models should incorporate at least these three cleaning mechanisms and
treat DLC as an interplay of all of them. Depending on the process parameters
(laser wavelength, pulse duration, optical constants of the materials, etc.) their
overall contribution to the cleaning will vary. It should also be pointed out that
332 M. Mosbacher et al.
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 nor
quantitatively how the explosive evaporation differs between bulk water (as in our
investigations) and water films (as in SLC) or even small water menisci 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 the explosive evaporation
in all these systems is necessary to accurately describe SLC.
6. SUMMARY
In this paper we have described our state of knowledge on the cleaning mechanisms
responsible for particle removal in laser cleaning. Besides the well-known thermal
expansion of the substrate and the explosive evaporation of a water film we
identified local substrate ablation as 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 removal 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, on 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
Laser cleaning of silicon wafers: Prospects and problems 333
REFERENCES
1. S.A. Hoenig, in: Particles on Surfaces 1: Detection, Adhesion and Removal, K.L. Mittal (Ed.), p.
3, Plenum, New York (1988).
2. T. Hattori, Solid State Technol., 8 (July 1990).
3. R. Kohli, in: Particles on Surfaces 7: Detection, Adhesion and Removal, K.L. Mittal (Ed.), VSP,
Utrecht, in press.
4. R. DeJule, Semiconductor Intl., 65-68 (August 1998).
5. International Technology Roadmap for Semiconductors, Sematech International (2000).
6. W. Zapka, W. Ziemlich and A.C. Tam, Appl. Phys. Lett. 58, 2217 (1991).
7. A.C. Tam, W.P. Leung, W. Zapka and W. Ziemlich, J. Appl. Phys. 71, 3515 (1992).
8. A.C. Engelsberg, Mater. Res. Soc. Symp. Proc. 315, 255 (1993).
9. I.V. Beklemyshev, V.V. Makarov, I.I. Makhonin, Y.N. Petrov, A.M. Prokhorov and V.I. Pustovoi,
JETP Lett. 46, 347-350 (1987).
10. K. Imen, J. Lee and S.D. Allen, Appl. Phys. Lett. 58, 203-205 (1991).
11. H.K. Park, C.P. Grigoropoulos, W.P. Leung and A.C. Tam, IEEE Trans. CPMT A, 17, 631 (1994).
12. J.B. Heroux, S. Boughaba, I. Ressejac, E. Sacher and M. Meunier, J. Appl. Phys. 79, 2857 (1996).
13. S. Boughaba, X. Wu, E. Sacher and M. Meunier, J. Adhesion, 61, 293-307 (1997).
14. A.C. Tam, H.K. Park and C.P. Grigoropoulos, Appl. Surf. Sci. 127-129, 721-725 (1998).
15. M. She, D. Kim and C.P. Grigoropoulos, J. Appl. Phys. 86, 6519 (1999).
16. M. Mosbacher, N. Chaoui, J. Siegel, V. Dobler, J. Solis, J. Boneberg, C.N. Afonso and P. Leiderer,
Appl. Phys. A 69[Supp.], 331 (1999).
17. M. Mosbacher, V. Dobler, J. Boneberg and P. Leiderer, Appl. Phys. A 70, 669 (2000).
18. J.M. Lee, K.G. Watkins and W.M. Steen, Appl. Phys. A 71, 671-674 (2000).
19. Y.F. Lu, Y.W. Zheng and W.D. Song, J. Appl. Phys. 87, 1584 (2000).
20. D. Bäuerle, Laser Processing and Chemistry, 2nd Ed., Springer (1996).
21. J.D. Kelley and F.E. Hovis, Microelectronic Engg. 20, 159 (1993).
22. Y.F. Lu, W.D. Song, M.H. Hong, B.S. Teo, T.C. Chong and T.S. Low, J. Appl. Phys. 80, 499
(1996).
23. Y.F. Lu, W.D. Song, B.W. Ang, M.H. Hong, D.S.H. Chan and T.S. Low, Appl. Phys. A 65, 9
(1997).
24. Y.F. Lu, W.D. Song, K.D. Ye, M.H. Hong, D.M. Liu, D.S.H. Chan and T.S. Low, Appl. Surf. Sci.
120, 317 (1997).
25. Y.F. Lu, W.D. Song, Y. Zhang and T.S. Low, Proc. SPIE 3550, 7 (1998).
26. Y.F. Lu, W.D. Song and T.S. Low, Mater. Chem. Phys. 54, 181 (1998).
27. Y.F. Lu, W.D. Song, M.H. Hong, D.S.H. Chan and T.S. Low, Proc. SPIE 3097, 352 (1997).
28. Y.F. Lu, Y. Zhang, Y.H. Wan and W.D. Song, Appl. Surf. Sci. 138-139, 140 (1999).
29. Y.F. Lu, Y.W. Zheng and W.D. Song, Appl. Phys. A 68, 569 (1999).
30. W.D. Song, Y.F. Lu, K.D. Ye, C.K. Tee, M.H. Hong, D.M. Liu and T.S. Low, Proc. SPIE 3184,
158 (1997).
31. G. Vereecke, E. Röhr and M.M. Heyns, J. Appl. Phys. 85, 3837 (1999).
32. X. Wu, E. Sacher and M. Meunier, J. Appl. Phys. 87, 3618 (2000).
33. T. Fourrier, G. Schrems, T. Mühlberger, J. Heitz, N. Arnold, D. Bäuerle, M. Mosbacher,
J. Boneberg and P. Leiderer, Appl. Phys. A 72, 1-6 (2001).
34. B.S. Lukyanchuk, Y.W. Zheng and Y.F. Lu, Proc. SPIE 4065, 576-587 (2000).
334 M. Mosbacher et al.
35. P. Leiderer, J. Boneberg, V. Dobler, M. Mosbacher, H.-J. Münzer, N. Chaoui, J. Siegel, J. Solis,
C.N. Afonso, T. Fourrier, G. Schrems and D. Bäuerle, Proc. SPIE 4065, 249-259 (2000).
36. P. Leiderer, J. Boneberg, M. Mosbacher, A. Schilling and O. Yavas, Proc. SPIE 3274, 68 (1998).
37. M. Mosbacher, presentation at the 5th International Conference on Laser Ablation COLA 99, July
1999, Göttingen, Germany.
38. M. Mosbacher, H.-J. Münzer, J. Zimmermann, J. Solis, J. Boneberg and P. Leiderer, Appl. Phys.
A 72, 41 (2001).
39. H.-J. Münzer, M. Mosbacher, M. Bertsch, J. Zimmermann, P. Leiderer and J. Boneberg, J.
Microscopy 202, 129 (2001).
40. M. Mosbacher, M. Bertsch, H.-J. Münzer, V. Dobler, B.-U. Runge, D. Bäuerle, J. Boneberg and
P. Leiderer, Proc. SPIE (in print).
41. M. Mosbacher, H.-J. Münzer, M. Bertsch, V. Dobler, N. Chaoui, J. Siegel, R. Oltra, D. Bäuerle,
J. Boneberg and P. Leiderer, in Particles on Surfaces 7: Detection, Adhesion and Removal, K.L.
Mittal (Ed.), VSP, Utrecht, in press.
42. V. Dobler, PhD thesis, University of Konstanz, Germany (2002).
43. V. Dobler, R. Oltra, J.P. Boquillon, M. Mosbacher, J. Boneberg and P. Leiderer, Appl. Phys. A 69,
335 (1999).
44. B.S. Lukyanchuk, Y.W. Zheng and Y.F. Lu, Proc. SPIE 4423, 115 (2001).
45. N. Arnold, G. Schrems, T. Mühlberger, M. Bertsch, M. Mosbacher, P. Leiderer and D. Bäuerle,
Proc. SPIE (in print).
46. S.V. Buntsents, S.G. Dmitriev and O.G. Shagimuratov, Phys. Solid Stat., 38, 552-557 (1996).
47. M. Vicanek, A. Rosch, F. Piron and G. Simon, Appl. Phys. A, 59 407-412 (1994).
48. G. Mie, Ann. Physik 4, 377-445 (1908).
49. P.W. Barber and S.C. Hill, Light Scattering by Particles: Computational Methods, World
Scientific Publishing, Singapore (1989).
50. O. Yavas, P. Leiderer, H.K. Park, C.P. Grigoropoulos, C.C. Poon, W.P. Leung, N. Do and A.C.
Tam, Phys. Rev. Lett. 70, 1830 (1993).
51. H.K. Park, C.P. Grigoropoulos, C.C. Poon, A.C. Tam, O. Yavas and P. Leiderer, Proc. SPIE 2498,
32-40 (1994).
52. O. Yavas, P. Leiderer, H.K. Park, C.P. Grigoropoulos, C.C. Poon, W.P. Leung, N. Do and A.C.
Tam, Appl. Phys. A 58, 407-415 (1994).
53. H.K. Park and C.P. Grigoropoulos, C.C. Poon and A.C. Tam, Appl. Phys. Lett. 68, 596-598
(1996).
54. O. Yavas, A. Schilling, J. Bischof, J. Boneberg and P. Leiderer, Laser Phys. 7, 343-348 (1997).
55. H.K. Park, D. Kim, C.P. Grigoropoulos, J. Appl. Phys. 80, 4072-4081 (1996).
56. O. Yavas, A. Schilling, J. Bischof, J. Boneberg and P. Leiderer, Appl. Phys. A 64, 331-339 (1997).
57. C.T. Avesidian, J. Phys. Chem. Ref. Data 14, 695-729 (1985).
58. A.C. Tam, W.P. Leung and W. Zapka, in: Particles on Surfaces: Detection, Adhesion and
Removal, K.L. Mittal (Ed.), pp. 405-418, Marcel Dekker, New York (1995).
59. E.T. Swartz and R.O. Pohl, Rev. Mod. Phys. 61, 605-668 (1989).
60. G. Cerbe and H.J. Hoffmann, Einführung in die Thermodynamik, Munich (1999).
61. G. Vereecke, E. Röhr and M.M. Heyns, Appl. Surf. Sci. 157, 67 (2000).
62. A.A. Kolomenskii, H.A. Schuessler, V.G. Mikhalevich and A.A. Maznev, J. Appl. Phys. 84, 2404-
2410 (1998).
63. J. Boneberg, Metallische Dünnfilmschmelzen von Halbleiterschichten nach ns-Laser-Annealing,
PhD thesis, University of Konstanz, Germany (1996).
64. D.R. Halfpenny and D.M. Kane, J. Appl. Phys. 86, 6641 (1999).
65. X. Wu, E. Sacher and M. Meunier, J. Appl. Phys. 87, 3618 (2000).
Surface Contamination and Cleaning, Vol. 1, pp. 335–343
Ed. K.L. Mittal
© VSP 2003
Abstract—A new photonic cleaning process that minimizes exposure of the substrate is introduced.
The concepts of the Resonant Laser Detachment (RLD) are described. The RLD process uses a laser
light source with intensity modulation to remove sub-micrometer contaminant particles from a sub-
strate. Unlike other laser removal methods, which eject the particle from the surface with a single
high-intensity laser pulse, RLD uses a continuous series of low-intensity laser pulses. The timing
and shape of these laser pulses are tuned to exploit the kinematic properties of the particle-surface
system and laser-material interactions. Theoretical analysis suggests a correlation between system
resonant frequency and the particle separation mechanism. This technique results in an efficient par-
ticle removal mechanism that minimizes stress and heat loading to the underlying substrate.
1. INTRODUCTION
The trends in the manufacturing of Integrated Circuits (IC’s) and Flat Panel Dis-
plays (FPD’s) and in other related manufacturing industries and the requirements
for environmentally safe cleaning technologies are driving the need to investigate
advanced dry photonic based cleaning methods.
Contamination remaining on the substrate after a processing step may create
device defects, rendering final end products useless. It is recognized that sub-
micrometer particles can cause major defects that can result in defective compo-
nents and lower production yield. With the trend towards reduced feature size in
IC’s, contamination will have an even greater impact on manufacturing yield. The
need is to increase yield, thus lowering production cost and increasing manufac-
turing competitiveness.
This investigation focuses on the use of a light based cleaning technology for
the purpose of developing a commercially viable critical cleaning process.
∗
To whom all correspondence should be addressed. Phone: 585-292-5610, Fax: 585-427-8422,
E-mail: phammond@lightforcetech.com
336 K. Kearney and P. Hammond
Strong adhesion forces exist between particles and surfaces due to van der Waals,
electrostatic, and capillary attraction mechanisms. Bowling [1] has discussed
these forces in detail, while a concise summary can be found in Tam et al. [2].
van der Waals forces are comprised of London-dispersion force, dipole-dipole
and dipole-induced dipole interactions. Capillary forces arise when atmospheric
moisture condenses in the gap between a particle and a substrate, and are a func-
tion of particle radius and liquid surface tension. Coulomb electrostatic forces
originate from the electrostatic double-layer formed between the particle and the
substrate.
Tam et al. [2] note that as particle size decreases, the relative adhesion force
increases dramatically. Particle mass (m) decreases as the cube of the diameter,
while adhesion forces (F) decrease directly with diameter; hence, the acceleration
(a=F/m) required to detach a particle from a surface scales inversely with the
square of the diameter. As will be shown later, adhesion forces on a micrometer-
size particle greatly exceed gravitational forces.
As noted by Bowling [1], the contact area of a particle in contact with a surface
is quite small, resulting in tremendous pressures at the particle-substrate contact
point. For a typical 1-µm particle, force per unit area is estimated to be 10.9 N/m2
which is enough to deform the particle and increase the particle to surface contact
area. Since the force of adhesion depends on the contact area, this effect will fur-
ther strengthen the binding of the particle to the surface.
æ h öd
FvdW = ç ÷ 2 , (1)
è 6π ø z
where h is the material dependent Lifshitz-van der Waals constant.
Notice that for a given separation z, the van der Waals forces scale with the
particle diameter d, while the mass of the particle scales as (d/2)3. Hence, the
force per unit mass (acceleration) scales with particle size as 1/d2: smaller parti-
cles require greater acceleration than larger particles to remove them from the
substrate. From Eq. (1), we see that the van der Waals forces on a sub-micrometer
particle can greatly exceed the gravitational force on the same particle resting on
the surface of the substrate.
As an example, consider a 1 µm diameter particle of silicon resting on a silicon
substrate. The gravitational force on the particle is approximately 1.5x10–15 N,
while the van der Waals forces at a distance of 0.4 nm from the substrate (the meas-
ured equilibrium separation distance) are 7x10–8 N, a nearly 50x107 times greater.
Particle removal using resonant laser detachment 337
Figure 1. The particle contact area increases with the time the particle remains on the surface due to
deformation forces, thus increasing the total particle adhesion force.
Figure 2. A particle resting on a substrate surface is trapped in a particle-surface potential well. The
potential energy of a particle at a given separation distance from a surface is shown.
with the orbital electrons of the particle and surface molecules. The two forces are
estimated to be equal at a distance of 0.4 nm from the surface [1]. This distance,
indicated in Figure 2, represents the minimum of the potential well in which the
contaminant particle is trapped. Although the exact shape of the potential well is
difficult to calculate precisely, it will have the general functional form shown in
Figure 2. A particle bound to a surface may thus be considered an oscillator. As
discussed below, the RLD technique exploits this kinematic behavior by exciting
the resonant frequencies of the oscillator.
Figure 3. Laser light sources utilize photothermal energy for a direct energy transfer to the particle
and surface. The net effect is a rapid thermal expansion of the particle and substrate surface, dis-
lodging the particle from the surface.
4. RADIATION FORCES
(axial) is directed along the laser beam axis. The second component (transverse)
arises due to the transverse light intensity gradient (e.g., a gaussian beam profile).
The combined effects of refraction and the transverse intensity gradient are such
that an unequal angular distribution of light (and hence momentum) exits at the
spherical surface of the particle, producing a net effective force towards the center
of the beam. This effect has been well demonstrated experimentally, and can be
used to form an optical “trap” for small particles [14]. In practice, the presence of
a viscous medium is required to stabilize oscillations about the center of the po-
tential well formed by the intensity profile. For typical beam diameters and light
intensities, the transverse force can be made comparable to the axial force, and
both can exceed the particle’s weight by many orders of magnitude.
The radiometric force has been exhaustively verified, and is commonly used in
aerosol studies [13]. The radiometric force is caused by uneven heating of the par-
ticle surface, and the interaction of the heated particle with the ambient back-
ground gas. Under optimum gas pressure conditions, the radiometric force can be
several orders of magnitude greater than the radiation force.
4.2. Photophoresis
Photophoresis has been investigated by Periasamy [16] as a mechanism to inhibit
particle attachment to a substrate. The motivation in their research was to prevent
particles from initial contamination. Our proposal is to use the mechanism of pho-
tophoresis to transport particles, which have been laser-detached from the sub-
strate, away from the surface and, ultimately, to a particle collection device such
as an electrostatic filter or vacuum line.
Figure 4. A number (n) of light sources (L) of various pulse frequencies and incident angles (Q) can
be applied to achieve the desired resonant response.
forces and heat loads are kept very small. Figure 4 demonstrates the application of
a single light source at a given incident angle to the substrate surface. Several in-
cident light sources may be required to achieve an effective particle detachment.
Defining the particle movement to be along a path orthogonal to the surface,
we can assume that the particle is bound in a potential well with an equilibrium
distance of approximately 0.4 nm above the surface. By applying a small, pertur-
bative force to the particle it is possible to cause the particle to oscillate about this
equilibrium position. For very small displacements, we can represent the potential
near the minimum as a parabola – i.e., a harmonic oscillator potential. This ap-
proximation will break down as the oscillations increase in amplitude. However,
it is sufficient to obtain an order-of-magnitude estimate of the resonant vibrational
frequency of the particle in the potential well.
For a 1 µm silicon particle resting on a silicon substrate with a contact diameter
of 0.03 µm we can estimate the resonant frequency by calculating the total adhe-
sion force (van der Waals and deformation forces, F) and applying Hooke’s law
to calculate the effective spring constant (k) for the particle-surface system:
F = −kx (3)
Again, assuming the distance x at which the particle is initially separated from the
substrate to be 0.4 nm gives a resonant frequency
1 1 k (4)
f = ω= ≈ 94 MHz
2π 2π m
The actual frequency scales with the effective particle separation distance as
1 / x . For a separation distance of 5 nm the frequency decreases to approxi-
mately 1.5 MHz. For a 0.1 µm particle the initial resonant frequency can exceed
1 GHz. In any case, these order of magnitude frequencies are within the limits of
342 K. Kearney and P. Hammond
Figure 5. The calculated resonant frequencies for 0.5 µm, 1 µm, 5 µm silicon particles on a silicon
substrate as a function of particle-to-surface separation distance is shown. The curves show the de-
pendence of resonant frequency on initial separation distance, particle size, and strongly suggest that
the applied driving force must be adaptive in relation to separation distance.
6. CONCLUSION
particle surface roughness the initial resonant frequency can exceed 1 GHz for a
0.01 µm silicon particle resting on a silicon substrate. The resonant frequency of a
particle on a surface decreases as the particle-to-surface distance increases requir-
ing an adaptive, chirped light source(s) for complete particle detachment. Particle
removal is accomplished by collecting particles, after resonant detachment from
the surface, using electrostatic gettering and vacuum suction to prevent reattach-
ment.
REFERENCES
CAROLE LEBLANC∗
Toxics Use Reduction Institute, University of Massachusetts Lowell, One University Avenue, Lowell,
MA 01854-2866
Abstract—In this paper, the author presents some of her findings in the pursuit of safer and greener
chemical solvents for hard-surface cleaning, as well as some of the new directions that the science
of cleaning may take in the next five to ten years. Specifically, innovative methods of research and
development into cleaning alternatives are explained, including molecular modeling, data mining,
and the use of ionic liquids. A discussion on chemical risk assessment ensues, in light of the scien-
tific concepts of hormesis and endocrine disruption. This is followed by a comparative analysis of
the European approach to policy-making, known as the Precautionary Principle and recent events
pertaining to cleaning issues in the U.S. Finally, conclusions are drawn based on a hypothetical case
of ‘over-cleaning’.
Keywords: Data mining; designer molecules; endocrine disruption; hormesis; ionic liquids; molecu-
lar modeling; precautionary principle.
1. INTRODUCTION
If the physical properties of liquids were any indication, successful solvent re-
placement with aqueous cleaners would never be possible. The components and
the behavior of the water molecule are nothing like those of a typical organic sol-
vent used for cleaning, as evidenced by the high polarity and dipole moment of
water. To illustrate, some important physical properties of the chlorinated solvent
trichloroethylene (TCE) and water are compared in Table 1.
Chemists often try to come as close as possible to the physical characteristics of the
original solvent in formulating new cleaners for the same application. This is entirely
understandable. Devising test protocols in which the solvent behaves in a certain
known fashion and expecting a water-based cleaner to function in the same manner,
however, is neither very realistic nor even a fair analysis of its potential performance.
∗
Phone: 978-934-3249, Fax: 978-934-3050, E-mail: Carole_LeBlanc@uml.edu
346 C. LeBlanc
Table 1.
Physical properties of a chlorinated solvent (TCE) and water
For this reason, the author described the significance of choosing the right
piece of mechanical equipment in process conversions involving aqueous clean-
ing in her thesis, “The Search for Safer and Greener Chemical Solvents in Sur-
face Cleaning: A Proposed Tool to Support Environmental Decision-Making”.a
Nevertheless, the author is familiar with at least two situations in which an over-
dependence on chemical properties led investigators to a much narrower field of
replacement candidates than the computer program, The Aqueous Way to Go,b the
tool developed during her doctoral research, would have recommended. In both
cases, an organic solvent was replaced with yet another organic solvent.
As a result, only incremental improvements, if any, were made to the health
and safety of workers and to the protection of the environment upon implement-
ing the alternative cleaners. The replacement cleaners shared a variety of traits
with the original solvents and the inherent dangers in using any organic and/or
chlorinated compound remained the same.
a
Completed for Erasmus University’s (Rotterdam, the Netherlands) international program in
cleaner production, cleaner products, industrial ecology and sustainability, in conjunction with the
Toxics Use Reduction Institute at the University of Massachusetts Lowell, leading to a Ph.D. in Sus-
tainable Development and Management (www.eur.nl/fsw/gsem/phd), the first awarded to an Ameri-
can woman.
b
A web-based, interactive matrix, i.e., a tool designed to enhance decision making in solvent sub-
stitution (www.angelfire.com/band2/greencleaners/doctoralthesis.html).
The future of industrial cleaning and related public policy-making 347
application was limited to the company’s Axarel® line of products. The Aqueous
Way to Go further merges the function of a computer program with Hansen-like,
actual performance criteria.
While application-specific testing is still required, the results of pertinent clean-
ing tests from the Toxics Use Reduction Institute’s Surface Solutions Laboratory
(SSL) are stored in The Aqueous Way to Go program. Additional performance in-
formation from other databases is also inserted into the program and serves to (1)
further decrease the time required to identify greener chemical cleaners and (2)
further increase the proficiency of the final selection. Like the molecular model-
ing used to accelerate chemical formulating described in the previous section, a
mechanism is needed, preferably computer-based for speed and accuracy, to (1)
sort through a plethora of data that may, or may not, be relevant and (2) determine
what chemical interactions, if any, reveal important trends for cleaning. Table 2
contains some of the newer tools available to conduct this kind of research.
Recently, algorithmic programming has been applied to advance the cause of
solvent substitution. In March of 2000, three simulation programs with different
algorithms were reviewed for designing greener solvents by Cabezas, Harten and
Green [3]. The three simulations were: (1) the U.S. EPA’s Program for Assisting
the Replacement of Industrial Solvents (Paris II), (2) the Technical University of
Denmark’s software, Computer Added Molecular Design (CAMD) and (3) Mo-
lecular Knowledge Systems’ chemical design software, Synapse. The Paris II al-
gorithm (www.tds.cc) uses chemicals from the Design Institute for Physical Prop-
erty Research (DIPPR) database and “looks for potential replacement solvents
whose properties are as close to the required parameters as possible”. The CAMD
solvent-design algorithm (www.capec.kt.dtu.dk) operates in a five-stage process
using valence (i.e., molecular charge) rules. The Synapse algorithm
(www.molknow.com) “generates candidate chemical structures, which are then
screened as potential solvent replacements in a four-step methodology”.
Unlike these programs that focus on theoretical scenarios with data that are
primarily intended for the scientific community, The Aqueous Way to Go concen-
trates on actual performance data of existing cleaners for the end-user community,
in addition to applications development. Data mining, or knowledge discovery in
databases, offers the best approach for manipulating this kind of information to
arrive at meaningful insights from observed tendencies (for example, the per-
formance of certain surfactants) in would-be relational databases.
Nevertheless, it would still be possible to use any, or a combination of, the re-
maining computer tools described in Table 2 for research and development into
greener cleaners. The web site http://surfactants.net/huibers/greenchem.html lists
a number of computer programs developed for property prediction, solvent re-
placement studies, and reaction design as well as additional solvent substitution
resources on the World Wide Web, in particular, the U.S. EPA’s Envio$en$e’s
(http://es.epa.gov) links to data systems.
350 C. LeBlanc
Table 2.
Examples of math-based/computer-enhanced research tools
Figure 5. The combination of an ionic liquid and supercritical CO2 to separate an organic com-
pound from solution.
Unlike water-soluble compounds that can be extracted with water, or the re-
moval of chemicals with high vapor pressures by distillation, ionic liquids require
very high temperatures to effect separation of compounds. This post-separation of
chemical products from ionic liquids may be difficult to achieve since the heat
needed may cause the products to degrade. Furthermore, the energy needed to
drive these reactions may be too expensive. If these problems can be solved, ionic
liquids may become safer, greener solvents since they do not possess any measur-
352 C. LeBlanc
able vapor pressure and so, unlike chlorinated/organic solvents, do not evaporate
to be inhaled by workers or to be emitted into the atmosphere and cause air pollu-
tion. (The dermatological consequences of exposure to ionic liquids as well as
their impact on water pollution are currently unknown.)
To address this separation issue, Brennecke and Beckman performed experi-
ments using a combination of an ionic liquid and supercritical carbon dioxide at
room temperature [5]. Their experiment, first reported in 1999, is diagramed in
Figure 5.
Both the carbon dioxide and the ionic liquid are recoverable for reuse. The
same system used for the separation of naphthalene could theoretically be used
for the removal of organic surface contaminants. While liquid-liquid (as opposed
to liquid-CO2) extractions would still be possible, they would invariably return the
system to the use of organic solvents, depending on the coefficient of partition, or
to the use of water, which would be almost entirely ineffectual for the separation
of most hydrocarbons. The use of various polymers, surfactants or solubilizers
may enhance the extraction/cleaning process.
To date, no toxicological or environmental fate studies have been published on
ionic liquids. This is urgently needed before much more additional application
work is done.
3. DISCUSSION
Figure 7. Identification of hormetic zone of zinc affecting cell reproduction. Source: H. Rubin,
Proc. Natl. Acad. Sci. (USA), 72, 1676 (1975).
by the onset of cancer. All three stages of the disease – initiation, promotion and
proliferation – have been linked to hormetic behavior [8].
Whether or not a chemical is said to exhibit a dose-dependent beneficial or
deleterious reaction depends upon the conditions defined at the time of the expo-
sure. For example, many chemicals used in the treatment of Acquired Immune
Deficiency Syndrome (AIDS) are considered toxic under almost every other non-
diseased circumstance; AIDS patients themselves need to be monitored closely
for toxicity levels during treatment. Problems arise when conditions are not de-
fined prior to a chemical’s release into the general environment, turning the bio-
sphere, if not the patient, into a laboratory. This is descriptive of the use of most
of mankind’s synthesized chemicals, including the detergents and solvents used
for cleaning.
The point of this discussion on hormetic behavior is that exposure may be more
harmful at lower, as opposed to higher, concentrations for the same chemical, tox-
icity notwithstanding. In fact, hormesis contains the root word hormones, which
are very powerful, biologically-active compounds that function effectively at low
concentrations. This refutes the principle learned by most chemists trained before
1990 that “dilution is the solution to the problem” and demonstrates the impor-
tance of identifying potential chemical hazards before they enter the biosphere, to
avoid the difficulty of separating minute amounts of powerful toxins (for exam-
ple, dioxin) from various waste streams.
Table 3.
Reasons for scientific uncertainty [20]
Table 4.
Overview of global policies affecting suspect endocrine disrupting chemicals [1]
In fact, consultants to the U.S. space agency have recommended the use of bu-
tane (lighter fluid) in some part-cleaning operations since this policy change. It
would appear that as older scientists retire and/or are replaced by younger, inex-
perienced researchers/contractors, there is a lack of a common understanding of
the past lessons learned from the use of these solvents. This may cause American
society to repeat some of the same mistakes made earlier. In other words, the U.S.
public may be facing a retreat to increased exposure to hydrocarbon products, and
their associated health hazards, used for cleaning prior to the discovery of the de-
struction of the ozone layer by chlorofluorocarbons (CFCs). These developments
are especially troubling in light of the United States’ active opposition to the
Kyoto Protocol to decrease global-warming (i.e., carbon-based) emissions. The
The future of industrial cleaning and related public policy-making 359
ponentially. Table 5 contains the water quality guidelines for chip manufacture
and blood dialysis as a means to compare each system’s level of desired contami-
nation control.
The goal of this treatise is not to expose the semiconductor industry’s over-
reliance on energy and water resources; other researchers such as Ted Smith,
founder and director of the Silicon Valley Toxics Coalition, are far more familiar
with the industry and have documented this dependency. The purpose here is to
reflect on the unsustainable nature of the current technology (indeed, most com-
puters are considered obsolete within eighteen months of manufacture), more spe-
cifically, the unhealthy, unnatural conditions to which workers are exposed in
cleanrooms.
Table 5.
Water quality guidelines for semiconductors (A) and hemodialysis (B)
A B
Contaminant Maximum Contaminant mg/l*
concentration*
Suspended solids: residue / 0.1 ppm / 500 counts Aluminum (Al) 0.01
particulates per liter Arsenic (Ar) 0.005
Dissolved solids: organic 0.020 ppm Barium (Ba) 0.1
(TOC)
Cadmium (Cd) 0.001
Ionic: resistivity / dissolved 18.3 megohm-cm /
silica (SiO2) 3 ppb Calcium (Ca) 2.0
Aluminum (Al) cations 0.2 ppb Choramines 0.1
Ammonium (NH4) cations 0.3 ppb Chlorine (Cl) 0.5
Chromium (Cr) cations 0.02 ppb Chromium (Cr) 0.014
Copper (Cu) cations 0.002 ppb Copper (Cu) 0.1
Iron (Fe) cations 0.02 ppb Fluoride (F) 0.2
Manganese (Mn) cations 0.05 ppb Lead (Pb) 0.005
Potassium (K) cations 0.1 ppb Magnesium (Mg) 4.0
Sodium (Na) cations 0.05 ppb Mercury (Hg) 0.0002
Zinc (Zn) cations 0.02 ppb Nitrate (AS N) 2.0
Bromide (Br) anions 0.1 ppb Potassium (K) 8.0
Chloride (Cl) anions 0.05 ppb Selenium (Se) 0.09
Nitrite (NO2) anions 0.05 ppb Silver (Ag) 0.005
Nitrate (NO3) anions 0.1 ppb Sodium (Na) 70.0
Phosphate (PO4) anions 0.2 ppb Sulfate (SO4) 100.0
Sulfate (SO4) anions 0.05 ppb Zinc (Zn) 0.1
Bacteria 0 counts per 100 ml
Source: P. Cartwright, Proc. of the Precision Cleaning meeting held in Rosemont, IL, May 1995.
*Measurement system chosen by industry.
The future of industrial cleaning and related public policy-making 361
c
Staphylococci are gram positive bacteria that are typically arranged in clumps or grape-like clus-
ters. They can be distinguished from streptococci in that only the staphylococi are catalase-positive
(catalase is an enzyme that liberates oxygen from hydrogen peroxide).
d
Enterococci are gram-positive bacteria that are widely distributed in nature and are part of the
normal flora of the gastrointestinal and genital tracts.
362 C. LeBlanc
of antibiotics, appears to be the root cause of these infectious mutations. The dis-
ease involves prions. Prions are tiny biological bits that may or may not be alive
and so are impossible to ‘kill’ (in the conventional sense of the term) to prevent
the infection’s spread. Some of these prions can be viewed three-dimensionally
with RosMol/Chime at www.mad-cow.org/prion_structure_folder/viewers.html.
aware of the developments in the cleaning industry so that other studies, like
those conducted at Tufts medical school, can be undertaken.
In the meantime, unless and until the chemical industry provides complete
chemical disclosure on a global basis, institutions such as the Toxics Use Reduc-
tion Institute’s research facility at the University of Massachusetts Lowell should
assist in formulating green chemical cleaners, in addition to providing education
and training programs and state-of-the-art laboratory testing of existing products.
This could be accomplished through partnerships with commercial enterprises
and/or other research facilities. Remaining pertinent issues, some having nothing
to do with cleaning performance, could then be addressed. These include studies
on the chemical additives of fragrances (over 80% of the odorants now used are
synthetic in origin) and dyes or colorants (often added for worker safety in prod-
uct recognition). Tighter quality control on percentages of ingredients could also
be maintained, since currently the concentrations of a cleaner’s components re-
ported on its Material Safety Data Sheet can vary by as much as 400%. Most im-
portantly, chronic, hormetic and synergistic chemical-exposure tests need to be
developed and implemented before cleaners are marketed.
The development of the computer program, The Aqueous Way to Go revealed
many of these trends and potential hazards in cleaning processes and chemicals.
The successful technical diffusion of this tool will require an educational compo-
nent, the topic of an upcoming paper.
REFERENCES