Selective decomposition of hydrogen peroxide in the epoxidation effluent of the HPPO process

Article abstract of DOI:10.1016/j.cattod.2011.09.029

This work describes the selective H₂O₂ decomposition in the exit stream of the epoxidation reactor employed in the Hydrogen Peroxide-Propylene Oxide (HPPO) process. Pd/Al₂O₃ and Pt/Al₂O₃ catalysts were tested. The effects of the reaction temperature and the pH of the solution on catalyst performance were investigated. It was found that the Pt catalyst is much more active than its Pd counterpart. An increase in the temperature and the pH of the solution resulted in an increase in the H₂O₂ decomposition rate; however, a parallel increase of by-products from PO was also observed. Working with a Pt/Al₂O₃ catalyst under optimized reaction conditions (333 K, pH = 7), hydrogen peroxide can be completely decomposed at reaction times of 120 min with no by-products produced from propylene oxide.

Full text of DOI:10.1016/j.cattod.2011.09.029

Catalysis Today 187 (2012) 168–172
Contents lists available at SciVerse ScienceDirect
Catalysis Today
journal homepage: www.elsevier.com/locate/cattod
Selective decomposition of hydrogen peroxide in the epoxidation effluent of the
HPPO process
Gema Blanco-Brievaa, M. Pilar de Frutos-Escrigb, Hilario Martínb,
Jose M. Campos-Martina, Jose L.G. Fierroa,∗
a Sustainable Energy and Chemistry Group, Instituto de Catálisis y Petroleoquímica, CSIC, Marie Curie, 2, Cantoblanco, 28049 Madrid, Spain1
b Centro de Tecnología Repsol, A-5, Km. 18, 28935 Móstoles, Madrid, Spain
a r t i c l e i n f o
a b s t r a c t
Article history:
This work describes the selective H2O2 decomposition in the exit stream of the epoxidation reactor
employed in the Hydrogen Peroxide–Propylene Oxide (HPPO) process. Pd/Al2O3 and Pt/Al2O3 catalysts
were tested. The effects of the reaction temperature and the pH of the solution on catalyst performance
were investigated. It was found that the Pt catalyst is much more active than its Pd counterpart. An
increase in the temperature and the pH of the solution resulted in an increase in the H2O2 decompo-
sition rate; however, a parallel increase of by-products from PO was also observed. Working with a
Pt/Al2O3 catalyst under optimized reaction conditions (333 K, pH = 7), hydrogen peroxide can be com-
pletely decomposed at reaction times of 120 min with no by-products produced from propylene oxide.
Received 8 August 2011
Received in revised form
15 September 2011
Accepted 18 September 2011
Available online 15 October 2011
Keywords:
Epoxidation
Hydrogen peroxide
Decomposition
HPPO
© 2011 Elsevier B.V. All rights reserved.
1. Introduction
epoxidation of propene but also the sequential hydrogenation and
oxidation of alkyl-anthraquinone. Another possible solution is the
Propylene oxide (PO) is the third-largest propylene derivative
used in the chemical industry. It is commonly used for making
polyurethane, unsaturated resins, surfactants, and other commer-
cial chemicals. PO is currently produced by two different processes
involving chlorohydrin and hydroperoxide [1]. Currently, due to
environmental and economic concerns, there is a pressing need for
new PO-synthesizing processes. An interesting alternative is the
epoxidation of propene with hydrogen peroxide. This epoxidation
process produces PO with very high selectivity (95% or higher) and
theoretically excretes only H2O as a by-product [1]. However, its
commercialization has been hindered largely by the supply of H2O2.
Hydrogen peroxide is now almost exclusively produced by the
anthraquinone process in large plants in which production costs
are offset by the scale. The transportation cost for H2O2 is too high
to be used for the production of bulk chemicals like PO. One strategy
that has been proposed thus far to address the problems of H2O2
production involves the integration of the anthraquinone process
for H2O2 synthesis with the propene epoxidation process catalyzed
by titanium silicalite (TS-1) [2]. In such a combined process, a suit-
able solvent system has to be developed to facilitate not only the
direct synthesis of H2O2 from H2/O2 mixtures with supported pal-
ladium (or palladium alloy) catalysts in acidified solvents (water,
methanol) [3–9]. The propylene oxide production based on the
direct on-site synthesis of hydrogen peroxide is abbreviated as
HPPO (Hydrogen Peroxide–Propylene Oxide).
The epoxidation of propylene with hydrogen peroxide on tita-
nium catalysts is very effective [10–16], but 100% conversion of
hydrogen peroxide cannot be achieved industrially in a practical
reactor size. As a consequence, small amounts of hydrogen perox-
ide still remain at the exit of the epoxidation reactor. Hydrogen
peroxide cannot be introduced in the PO purification steps because
its decomposition produces oxygen, which can produce serious
safety problems. A simple way to solve this drawback is to decom-
pose the residual hydrogen peroxide at the exit of the epoxidation
reactor once unreacted propylene has been separated. The oxygen
gas generated by hydrogen peroxide decomposition will simply be
removed by flowing an inert gas.
In the existing literature, metal oxides are often employed for
hydrogen peroxide decomposition [17–22]. However, metal oxides
decompose H2O2 into hydroxyl radicals (HO•). Hydroxyl radicals
are highly volatile and are capable of reacting with a wide number
of organic compounds [19,21], including the organic components of
the epoxidation effluent. This is detrimental for the whole process
and results in a loss of the PO product. For this reason, our aim in the
present work was the selection of supported noble metal catalysts
∗
Corresponding author. Fax: +34 915854760.
E-mail address: jlgfierro@icp.csic.es (J.L.G. Fierro).
1
http://www.icp.csic.es/eqs/.
0920-5861/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.cattod.2011.09.029

G. Blanco-Brieva et al. / Catalysis Today 187 (2012) 168–172
169
Table 1
that are active in the hydrogen peroxide decomposition with a
Textural properties of the catalysts employed.
minimal formation of hydroxyl radicals [23,24]. The metal nanopar-
ticles deposited on the carrier substrate constitute the active phase
in this reaction. In these catalysts, activity may be related to the
higher adsorption of H2O2 on reduced metal sites [23,24].
This work was undertaken with the aim to study the catalytic
decomposition of hydrogen peroxide in a solution that simulates
the epoxidation reactor exit stream after propylene removal in an
HPPO process. This research focuses on the minimal formation of
by-products derived from the organic compounds present in the
stream.
Catalyst
BET surface area (m2/g) Pore Volume (ml/g) Pore diameter (nm)
Pd/Al2O3 99.1
Pt/Al2O3 97.7
0.24
0.24
10
10
To start the reaction, the basket was lowered, putting the catalysts
in contact with the reaction mixture. Hydrogen peroxide and water
concentrations were measured by iodometric and Karl–Fischer
standard titrations, respectively. The concentration of the organic
compounds was determined by GC-FID using an Agilent 6850
instrument fitted with a DB-WAX capillary column. The reaction
mixture was prepared to simulate the effluent of the epoxidation
reactor in an HPPO process after the elimination of propylene.
This mixture consists of the following: 506 ppm acetaldehyde,
12.74 wt% OP, 73.2 wt% MeOH, 0.49 wt% 1-methoxy-2-propanol,
0.78 wt% 2-methoxy-1-propanol, 0.59 wt% 1,2-propanediol and
2 wt% H2O2 [6].
2. Experimental methods
2.1. Catalysts
Alumina-supported palladium and platinum catalysts were
employed in this work. Pd/Al2O3 and Pt/Al2O3 (0.5 wt%) shaped
as cylindrical pellets (3.2 mm × 3.2 mm) were purchased from
Johnson Matthey. The effectiveness factor is comparable in both
catalysts since this only depends on the intrinsic kinetic constant
of each catalyst and both exhibit a similar shape and size, being
the operating temperature the same in both cases. These cata-
lysts are commercially distributed in the reduced (metallic) state.
For comparative purposes, cylindrical pellets (3.2 mm × 3.2 mm) of
bare ␥-alumina were employed.
3. Results and discussion
3.1. Structural characteristics of the catalysts
Palladium and platinum catalysts were tested in the selective
H2O2 decomposition at 333 K. Both catalysts were active (Fig. 1).
However, the Pt/Al2O3 catalyst was more active than its Pd coun-
terpart, reaching total conversion in 120 min. No changes in the
concentration of propylene oxide were detected, but small changes
in PO concentration can be masked by analytical errors. For this rea-
son, it may be more accurate to monitor the time dependence of
the concentration of by-products derived from PO.
2.2. Catalysts characterization
Textural properties were determined from the adsorption–
desorption isotherms of nitrogen recorded at 77 K with
a
Micromeritics TriStar 3000. The specific area was calculated by
applying the BET method to the range of relative pressures (P/P0) of
the isotherms between 0.03 and 0.3 and taking a value of 0.162 nm2
for the cross-section of adsorbed nitrogen molecule at 77 K.
Powder X-ray diffraction (XRD) patterns were recorded in
the 0.5–10◦ 2ꢀ range using step mode (0.05, 5 s) with a Seifert
3000 XRD diffractometer equipped with a PW goniometer with
Bragg–Brentano ꢀ/2ꢀ geometry, an automatic slit, and a bent
graphite monochromator.
X-ray photoelectron spectra (XPS) were acquired with a VG
Escalab 200R spectrometer equipped with a hemispherical electron
analyzer and a Mg K␣ (hꢁ = 1253.6 eV) non-monochromatic X-ray
source. The samples were degassed in the pretreatment chamber at
room temperature for 1 h prior to being transferred into the instru-
ment’s ultra-high vacuum analysis chamber. The silicon, oxygen,
sulfur and carbon signals were scanned several times at a pass
energy of 20 eV in order to obtain good signal-to-noise ratios. The
binding energies (BE) were referenced to the BE of the C 1s core-
level spectrum at 284.9 eV. The invariance of the peak shapes and
widths at the beginning and end of the analyses indicated constant
charge throughout the measurements. Peaks were fitted by a non-
linear least square fitting routine using a properly weighted sum
of Lorentzian and Gaussian component curves after background
subtraction [25].
The concentration of these compounds remained almost con-
stant for the two metal catalysts (Pd/Al2O3 and Pt/Al2O3) (Fig. 1).
A blank experiment was performed using metal-free ␥-alumina
(Al2O3) pellets. The ␥-alumina alone is active in the hydrogen per-
oxide decomposition reaction (Fig. 1), but the hydrogen peroxide
conversion is clearly smaller than that observed with the metal cat-
alysts. This reaction was previously observed [24], suggesting that
the decomposition occurs on the surface with the formation of HO•
•
or HO2 radicals. The formation of radicals affects the formation of
by-products (Fig. 1). For this system, the formation of 2-methoxy-1-
propanol, 1-methoxy-2-propanol and 1,2-propanediol is evident.
These results indicate that Pt/Al2O3 is the most active catalyst
(Fig. 1). The characterization results showed no differences in the
textural properties of both samples (Table 1). XRD analysis (not
shown here) showed the diffraction peaks of ␥-alumina without
any diffraction line originating from the metal species. This con-
firms the high dispersion of platinum, most probably as very small
clusters with a size of less than 3 nm supported on the alumina
substrate.
The chemical state of the platinum or palladium was derived
from the X-ray photoelectron spectra. The Pd 3d and Pt 4d core-
levels showed the characteristic spin–orbit splitting of Pd 3d and
Pt 4d levels. The most intense (Pd 3d5/2 or Pt 4f7/2) component
of the doublet is located at lower binding energies, and the least
intense (Pd 3d3/2 or Pt 4f5/2) component is located at higher bind-
ing energies. Chemical information can be extracted from each of
these components, but attention will only be paid to the more
intense one (Pd 3d5/2 or Pt 4f7/2). Fresh samples showed the pres-
ence of one component (Table 2) corresponding to a well-dispersed
metallic species on alumina (Pd 3d5/2 336.0 eV, Pt 4f7/2 314.0 eV).
In addition, the XPS of the used samples showed the presence of
two components (Table 2): one due to the metallic species sim-
ilar to that found in the fresh samples and another located at a
higher binding energy (Pd 3d5/2 337.8 eV or Pt 4f7/2 316.9 eV) due
2.3. Hydrogen peroxide decomposition
Catalysts were tested in the decomposition of hydrogen perox-
ide. The catalytic tests were performed in a high pressure stirred
reactor (Autoclave Engineers) equipped with a falling basket. In
a typical run, the catalyst (H2O2/metal = 1200/1 molar) was put
inside an autoclave with 325 g of liquid mixture. The reactor was
purged with N2 and pressured to 0.5 MPa. When the pressure
reached a constant value, the mixture was heated up to the reac-
tion temperature by maintaining continuous stirring (1000 rpm).

170
G. Blanco-Brieva et al. / Catalysis Today 187 (2012) 168–172
3,2
2,8
2,4
2,0
1,6
1,2
0,8
0,4
0,0
2,7
(b)
Pt/Al2O3 333 K
Pd/Al2O3 333 K
Al2O3 333 K
(a)
Pt/Al2O3 333 K
Pd/Al2O3 333 K
Al2O3 333 K
2,4
2,1
1,8
1,5
1,2
0,9
0,6
0,3
0,0
1-methoxy-2-propanol
1,2-propanediol
2-methoxy-1-propanol
0
20
40
60
80
100 120
0
20
40
60
80
100 120
Time (min)
Time (min)
Fig. 1. Hydrogen peroxide (a) and by-products (b) derived from PO concentration profiles during the H2O2 decomposition at 333 K using Pd and Pt catalysts and alumina
support.
to the presence of an oxidized species (Pd2+ or Pt2+). The presence
of these oxidized species points to a mechanism of H2O2 decompo-
sition that involves redox processes where oxidized and reduced
metal species participate.
Quantitative XPS data (Table 2) revealed that the surface
Pd(Pt)/Al ratio is higher for Pd than for Pt in both the fresh and
used catalysts. These results indicate that metal exposure is higher
in the Pd catalyst than in its Pt counterpart, meaning it has higher
dispersion. However, the Pt catalyst is clearly more active in hydro-
gen peroxide decomposition than its Pd counterpart. Thus, Pt has
higher intrinsic activity than palladium for this reaction.
reaction time and complete conversion after 30 min. In addition,
the increase in the temperature from 313 K to 353 K is accompa-
nied by a parallel increase in the formation of byproducts from
PO, yielding 1-methoxy-2-propanol, 2-methoxy-1-propanol, and
1,2-propanediol (Fig. 2). The side reactions participating in the
formation of such products are a serious drawback for an indus-
trial application because the decomposition of unreacted hydrogen
peroxide yields a loss of the desired PO product. Therefore, sep-
aration/purification units are imperative and could result in an
increase in the operation process costs and could result in an
increase in the operation process costs in order to reduce the loss
of PO selectivity.
Surface analysis of spent catalysts via XPS analysis demon-
strated the presence of two platinum species, Pt0 and Pt2+ (Table 2).
The proportion of oxidized species increased with the reaction
temperature. This effect can be attributed to the higher oxida-
tion capability of hydrogen peroxide at increased temperatures.
Quantitative XPS data, i.e., Pt(Pd)/Al surface atomic ratios, reveal
valuable information (Table 2). The Pt/Al surface atomic ratio of
the Pt catalyst does not change after use in the reaction at dif-
ferent temperatures and is very close to the value determined
on the fresh catalyst. These results indicate that the Pt cata-
lyst is a robust system and is particularly well suited for the
target reaction under the reaction conditions selected in this
work.
3.2. Catalytic activity
The catalysts were tested for the decomposition of hydrogen
peroxide present in a liquid mixture that simulates the exit stream
of an industrial HPPO reactor. The effects of reaction temperature
and the pH of the solution on catalyst performance were analyzed
in some detail.
3.2.1. Influence of temperature
The influence of reaction temperature was studied by check-
ing the performance of the Pt/Al2O3 catalyst at temperatures in
the range of 313–353 K. It was observed that an increase in reac-
tion temperature is paralleled by an increase in the decomposition
rate of hydrogen peroxide (Fig. 2). The H2O2 concentration drops
very quickly, with a conversion of greater than 90% after a 15 min
3.2.2. Effect of pH
Hydrogen peroxide decomposes in aqueous solutions at basic
pH. We have tested the effect of the solution pH on the catalyzed
decomposition of hydrogen peroxide present in a propylene-free
mixture that simulates the exit stream of an industrial HPPO reac-
tor. The hydrogen peroxide decomposition slightly increases in the
basic solution (Fig. 3), but no significant changes are observed in the
formation of PO by-products. As a consequence, the increased for-
mation of PO by-products is clearly more affected by an increase in
the reaction temperature than an increase in the pH of the solution
over this short reaction time. Nonetheless, working with a basic
solution is not recommended for an industrial process because
corrosion-resistant equipment is required, and a neutralization
unit should be incorporated. Thus, the use of a basic solution does
not result in enough of an increase in the reaction rate to justify the
Table 2
XPS analysis data for fresh and used catalysts.
Catalyst
Binding Energy (eV) of Pd
3d5/2 and Pt 4d5/2 Levels
Surface Atomic
Ratio (Pt or Pd)/Al
Pd/Al2O3 fresh
Pd/Al2O3 used
336.0
0.0083
337.9 (34)
336.0 (66)
314.0
316.8 (42)
314.2 (58)
316.9 (53)
314.2 (47)
316.9 (42)
314.2 (58)
0.0120
0.0041
Pt/Al2O3 fresh
Pt/Al2O3 used at
333 K
Pt/Al2O3 used at
353 K
0.0042
0.0042
0.0040
Pt/Al2O3 used at
353 K and pH = 10

G. Blanco-Brieva et al. / Catalysis Today 187 (2012) 168–172
171
3.2
2.8
2.4
2.0
1.6
1.2
0.8
0.4
0.0
2.7
Pt/Al2O3 333 K
Pt/Al2O3 353 K
Pt/Al2O3 333 K
Pt/Al2O3 353 K
2.4
2.1
1.8
1.5
1.2
0.9
0.6
0.3
0.0
1-methoxy-2-propanol
1,2-propanediol
2-methoxy-1-propanol
0
30
60
90
120
0
20
40
60
80 100 120
Time / min
Time (min)
Fig. 2. Hydrogen peroxide and by-products derived from PO concentration profiles during the H2O2 decomposition at 333 K and 353 K using a Pt catalyst.
3.2
2.8
2.4
2.0
1.6
1.2
0.8
0.4
0.0
2.7
2.4
2.1
1.8
1.5
1.2
0.9
0.6
0.3
0.0
Pt/Al2O3 353 K
Without catalyst 353 K pH=10
Pt/Al2O3 353 K pH=10
Pt/Al2O3 353 K
Without catalyst 353 K pH=10
Pt/Al2O3353 K pH=10
1-methoxy-2-propanol
Monopropylene glycol
2-methoxy-1-propanol
0
20
40
60
80 100 120
0
20
40
60
80 100 120
Time (min)
Time / min
Fig. 3. Hydrogen peroxide and by-products derived from PO concentration profiles during the H2O2 decomposition at 353 K and varying pH using Pt.
complication in the industrial process needed for its implementa-
tion.
that Pt atoms display a higher intrinsic activity in the reaction of
hydrogen peroxide decomposition than their Pd counterpart.
Accurate measurement of the binding energies of core-electrons
in the used catalysts revealed the presence of two Pt species,
suggesting that the decomposition reaction occurs via a redox
mechanism where oxidized and reduced metal species are
involved. Platinum dispersion does not seem to be modified after
use in the reaction under several reaction conditions, and Pt cat-
alyst remains similar to that of the fresh catalyst. These findings
indicate that Pt/Al2O3 is a robust and stable catalyst for the target
reaction.
Increases in the reaction temperature and/or pH resulted in a
higher H2O2 decomposition rate, but they also increased the for-
mation of by-products from PO. The loss of PO selectivity and
by-product formation are not desirable from an industrial point
of view because separation/purification units have to be added to
4. Conclusions
The selective decomposition of hydrogen peroxide present in
the exit stream of the epoxidation reactor in an HPPO process can
be performed with an alumina-supported platinum (or palladium)
catalyst without significant formation of by-products derived from
propylene oxide. A Pt/Al2O3 catalyst working at the optimum
temperature (333 K) completely decomposes the hydrogen per-
oxide present in the exit stream at reaction times no longer than
120 min.
Quantitative XPS data revealed that the Pd catalyst has a higher
number of exposed atoms on the surface than its Pt counterpart,
even though the Pt catalyst is much more active. This indicates

172
G. Blanco-Brieva et al. / Catalysis Today 187 (2012) 168–172
the plant hardware, resulting in an increase in operation process
costs and process economy.
[8] J.M. Campos-Martin, G. Blanco-Brieva, J.L.G. Fierro, Angew. Chem. Int. Ed. 45
(2006) 6962.
[9] J. Zhao, J. Zhou, J. Su, H. Guo, X. Wang, W. Gong, AlChE J. 53 (2007) 3204.
[10] M.C. Capel-Sanchez, G. Blanco-Brieva, J.M. Campos-Martin, M.P. De Frutos, W.
Wen, J.A. Rodriguez, J.L.G. Fierro, Langmuir 25 (2009) 7148.
[11] M.C. Capel-Sanchez, J.M. Campos-Martin, J.L.G. Fierro, J. Catal. 217 (2003) 195.
[12] M.C. Capel-Sanchez, J.M. Campos-Martin, J.L.G. Fierro, Appl. Catal. A: Gen. 246
(2003) 69.
[13] M.C. Capel-Sanchez, J.M. Campos-Martin, J.L.G. Fierro, J. Catal. 234 (2005) 488.
[14] M.C. Capel-Sanchez, J.M. Campos-Martin, J.L.G. Fierro, M.P. De Frutos, A. Padilla
Polo, Chem. Commun. (2000) 855.
Acknowledgements
The authors acknowledge financial support from Repsol (Spain)
and the MICINN (Spain) through the PSE-310200-2006-2, FIT-
320100-2006-88 and ENE2007-07345-C03-01/ALT projects. GBB
gratefully acknowledges a fellowship granted by Repsol.
[15] B. Notari, in: W.O.H.D.D. Eley, B. Gates (Eds.), Adv. Catal., vol. 41, Academic
Press, 1996, p. 253.
[16] P. Ratnasamy, D. Srinivas, H. Knözinger, in: B.C. Gates, H. Knözinger (Eds.), Adv.
Catal., vol. 48, Academic Press, 2004, p. 1.
References
[17] M.A. Hasan, M.I. Zaki, L. Pasupulety, K. Kumari, Appl. Catal. A: Gen. 181 (1999)
171.
[1] T.A. Nijhuis, M. Makkee, J.A. Moulijn, B.M. Weckhuysen, Ind. Eng. Chem. Res. 45
(2006) 3447.
[2] C. Wang, B. Wang, X. Meng, Z. Mi, Catal. Today 74 (2002) 15.
[3] G. Blanco-Brieva, J.M. Campos-Martin, M.P. De Frutos, J.L.G. Fierro, Catal. Today
158 (2010) 97.
[18] A. Hiroki, J.A. LaVerne, J. Phys. Chem. B 109 (2005) 3364.
[19] H.-H. Huang, M.-C. Lu, J.-N. Chen, Water Res. 35 (2001) 2291.
[20] W.P. Kwan, B.M. Voelker, Environ. Sci. Technol. 37 (2003) 1150.
[21] C.M. Miller, R.L. Valentine, J. Hazard. Mater. 41 (1995) 105.
[22] H. Zhou, Y.F. Shen, J.Y. Wang, X. Chen, C.-L. O‘Young, S.L. Suib, J. Catal. 176 (1998)
321.
[23] V.R. Choudhary, C. Samanta, P. Jana, Appl. Catal. A: Gen 332 (2007) 70.
[24] M.A. Hasnat, M.M. Rahman, S.M. Borhanuddin, A. Siddiqui, N.M. Bahadur, M.R.
Karim, Catal. Commun. 12 (2010) 286.
[4] G. Blanco-Brieva, E. Cano-Serrano, J.M. Campos-Martin, J.L.G. Fierro, Chem.
Commun. 10 (2004) 1184.
[5] G. Blanco-Brieva, M.C. Capel-Sanchez, M.P. de Frutos, A. Padilla-Polo, J.M.
Campos-Martin, J.L.G. Fierro, Ind. Eng. Chem. Res. 47 (2008) 8011.
[6] G. Blanco-Brieva, M.C. Capel-Sanchez, M.P. De Frutos, A. Padilla-Polo, J.M.
Campos-Martin, J.L.G. Fierro, Ind. Eng. Chem. Res. 47 (2008) 8011.
[7] G. Blanco-Brieva, M.P. De Frutos Escrig, J.M. Campos-Martin, J.L.G. Fierro, Green
Chem. 12 (2010) 1163.
[25] C.D. Wagner, L.E. Davis, M.V. Zeller, J.A. Taylor, R.H. Raymond, L.H. Gale, Surf.
Interface Anal. 3 (1981) 211.