Preparation, characterization, and activity of Cu/TiO2 catalysts: II. Effect of the catalyst morphology on the hydrogenation of 1,3-cyclooctadiene and the CO-NO reaction on Cu/TiO2 catalysts

Article abstract of DOI:10.1006/jcat.1997.1476

The hydrogenation of 1,3-cyclooctadiene and the CO-NO reaction have been studied on Cu/TiO₂ catalysts, with reference to the effects of the preparation method and of the reductive pretreatments of the samples. In the hydrogenation of 1,3-cyclooctadiene at 433 K, a marked difference in activity is observed between catalysts prepared by a chemisorption-hydrolysis method (C) and those prepared by wet impregnation (I), the former showing turnover numbers about 100 times greater than the latter. This large difference in activity is ascribed to differences in the morphology of the copper species present at the surfaces of the two different kinds of catalysts: there are three-dimensional stepped small particles of Cu in the first case (C), very efficient in hydrogen dissociation, whereas there are isolated copper sites embedded in the reduced support in the second case (I), very much less active toward hydrogen dissociation. In the CO-NO reaction, C samples show a decrease in the activity on increasing the catalyst prereduction temperature, while an opposite behavior is observed on I samples. For this reaction the rate determining step is NO dissociation; the interpretation of the apparently conflicting data is that NO dissociation occurs efficiently on copper sites on mildly reduced C catalysts exposing three-dimensional copper particles and also on reduced titania sites formed in the high-temperature reductive treatments on I catalysts.

Full text of DOI:10.1006/jcat.1997.1476

JOURNAL OF CATALYSIS 165, 140–149 (1997)
ARTICLE NO. CA971476
Preparation, Characterization, and Activity of Cu/TiO Catalysts
2
II. Effect of the Catalyst Morphology on the Hydrogenation of 1,3-Cyclooctadiene
and the CO–NO Reaction on Cu/TiO Catalysts
2
F. Boccuzzi,�,1
�
A. Chiorino, M. Gargano,† and N. Ravasio†
�
Dipartimento di Chimica I. F. M. dell’ Universit a` , Via P. Giuria 7, 10125 Torino, Italy; and †Centro C.N.R. MISO, Dipartimento
di Chimica dell’ Universit a` , Via Amendola 173, 70126 Bari, Italy
Received December 5, 1995; revised May 25, 1996; accepted September 25, 1996
small two-dimensional copper clusters are exposed which
The hydrogenation of 1,3-cyclooctadiene and the CO–NO reac- are to some extent rendered positive by the interaction
tion have been studied on Cu/TiO2 catalysts, with reference to the with the support. On similarly loaded samples prepared
effects of the preparation method and of the reductive pretreat-
ments of the samples. In the hydrogenation of 1,3-cyclooctadiene
at 433 K, a marked difference in activity is observed between cata-
lysts prepared by a chemisorption–hydrolysis method (C) and those
prepared by wet impregnation (I), the former showing turnover
over, by a study of the effect of reductive treatments on the
numbers about 100 times greater than the latter. This large differ-
ence in activity is ascribed to differences in the morphology of the
copper species present at the surfaces of the two different kinds of
catalysts: there are three-dimensional stepped small particles of Cu
by a chemisorption–hydrolysis method, in addition to the
above-mentioned isolated copper species, well-dispersed,
clean, and well-crystallized three-dimensional copper par-
ticles are present after mild reductive pretreatment. More-
overall IR transparency of the samples and on the intensity
of the CO absorption bands, it was shown that by increasing
the reduction temperature a defective suboxide of titania,
in the first case (C), very efficient in hydrogen dissociation, whereas strongly absorbing in the IR, is formed, covering a large
there are isolated copper sites embedded in the reduced support fraction of the copper particles.
in the second case (I), very much less active toward hydrogen dis-
sociation. In the CO–NO reaction, C samples show a decrease in
In thispaper we report on the catalyticactivityofthe sam-
ples characterized in Part I in two reactions, namely the se-
lective hydrogenation of 1,3-cyclooctadiene to cyclooctene
and the CO–NO reaction.
the activity on increasing the catalyst prereduction temperature,
while an opposite behavior is observed on I samples. For this reac-
tion the rate determining step is NO dissociation; the interpretation
of the apparently conflicting data is that NO dissociation occurs
efficiently on copper sites on mildly reduced C catalysts exposing
three-dimensional copper particles and also on reduced titania sites
formed in the high-temperature reductive treatments on I catalysts.
The catalytic activity in hydrogenation reactions strongly
depends on both the nature and the dispersion of metallic
particles. In particular, the presence of Cu(0) was found to
be mandatory for the hydrogenation of cyclic polyenes (2).
This reaction is favored by a high specific Cu(0) area, which
in turn depends on the preparation method and on the sup-
port (3). The origins of these effects have not yet been fully
clarified. In this paper we will attempt a correlation be-
tween the catalytic activity of the samples and the chemical
properties of copper sites of different nature present at the
surface of the examined samples.
�
c 1997 Academic Press
1
. INTRODUCTION
In Part I (1), it was shown by XRD, HRTEM, and TPR
investigations and through the analysis of FTIR spectra
following CO adsorption that the surface properties of
Cu/TiO2 are strongly affected by the preparation method,
by the copper loading, and by the reductive pretreatment.
In particular it was shown that at the surface of samples
prepared by a wet impregnation method, irrespective of
the copper loading (2–8 wt.% Cu) and reductive pretreat-
ments (423 to 673 K in H2), almost isolated copper atoms or
In the case of the CO–NO reaction, copper-based cata-
lysts for NO reduction have recently become objects of
great interest, in particular the NO decomposition and re-
duction over Cu/Al2O3 (4) and Cu-exchanged zeolites (5).
Recently the catalytic reduction of NO by CO on Cu/TiO2
prepared by wet impregnation was studied by some of us
(
6). The main results were the strong dependence of the
activity on the reduction pretreatment temperature, the
highest activity being observed on samples reduced at high
temperature. Nitric oxide dissociation on defective titania
1
To whom correspondence should be addressed. Fax: 3911 6707855.
E-Mail: Boccuzzi@ch.unito.it.
140
0021-9517/97 $25.00
Copyright �c 1997 by Academic Press
All rights of reproduction in any form reserved.

CHARACTERIZATION AND ACTIVITY OF Cu/TiO2 CATALYSTS, II
141
in close contact with copper small clusters appeared to be spectra and mass spectra could be taken at different times
the rate determining step of the reaction. Also for other or reaction temperatures in the same experiment. All the
metals, such as Pt catalysts (7), it was observed that the ac- IR spectra were recorded with the sample at room temper-
tivity when the metal is supported on n-type semiconduct- ature (RT) using a Perkin–Elmer FTIR spectrometer, at a
�
1
ing oxides is strongly dependent on the reduction pretreat- resolution of 2 cm , the number of scans (100–200) being
ment temperature. In this paper we compare the results of varied according to the transmission of the sample, with the
the NO–CO reaction on Cu/TiO2 samples prepared by wet aim of obtaining similar levels of noise. CO, from Matheson,
impregnation with those prepared by the chemisorption– was used without purification; NO, from Matheson, freshly
hydrolysis method, the samples having been submitted distilled before use, was analyzed before the reaction with
to reductive treatments at different temperatures in both the mass spectrometer.
cases.
In the label of the examined samples (e.g., 4I573), the ini-
tial number indicates the nominal copper loading (wt.% ),
the letter I or C indicates the sample preparation method
2
. EXPERIMENTAL
(
see Part I (1)), and the following three figures indicate the
2
2
.1. Catalyst Preparation
reduction temperature of the sample, before the reaction, in
Kelvin. The initial reaction temperature was RT for all the
samples and for the less reactive ones the reaction mixture
was heated to 473 K.
The sample preparation is described in Part I (1).
.2. Catalyst Activation and Procedure for
Cyclooctadiene Hydrogenation
3
. RESULTS AND DISCUSSION
The catalyst (250–300 mg) was first treated with H2 at
73 K for 20 min, followed by raising the temperature by
0 K every 20 min up to 473 K and removing each time
3
2
3
.1. Selective Hydrogenation of 1,3-Cyclooctadiene
the water formed under vacuum for 1 min. The catalyst was
then transferred under N2 into a 60-ml stainless steel (AISI
The selective hydrogenation of 1,3-cyclooctadiene to cy-
clooctene was studied in batch and pulse mode. The relative
conversion of 1,3-cyclooctadiene to cyclooctene for the re-
action carried out in pulse mode is in agreement with the
turnover frequencies obtained in batch mode. The results
obtained in batch mode at 433 K and 3 atm H2 are reported
in Table 1.
A very large difference in activity is observed between
the catalysts prepared by the two methods, the turnover
frequencies of C samples being about 100 times higher than
those ofthe I samples, which are almost completelyinactive.
The large difference in activity cannot be explained sim-
ply by the differences in Cu(0) specific surface area (see
3
16) autoclave and the organic reactant (10 ml of a solution
of 1,3-cyclooctadiene 1.6 M in n-dodecane) was added. Hy-
drogen was charged at 3 atm, the temperature was raised to
4
33 K, and stirring was begun (1000 rpm). The rate of H2 up-
take was measured by an electronic flowmeter. The initial
rate was used to evaluate turnover frequencies. Turnover
frequencies are expressed on the assumption that only cop-
per sites are active, the copper surface area being that de-
termined by N2O decomposition, as reported in Part I (1).
The same reaction was carried out in the pulse mode in
a tubular microreactor placed in the injector port of a gas
chromatograph. The catalyst (10–20 mg) was reduced with
H2 between 373 and 473 K, raising the temperature by 20 K
every 20 min. The temperature of the microreactor was ad-
justed to 433 K before the reaction was commenced. The cy-
clooctadiene (see above) was then introduced with H2 flow
rates ranging from 20 to 150 ml/min and the products were
analyzed using a nonbonded, poly(80% -biscyanopropyl–
Table 1). Thus, comparable differences in SCu(0) (64 vs
2
7
m /gCu), exhibited by Cu/Al2O3 samples prepared by the
TABLE 1
Turnover Frequencies for the Hydrogenation of 1,3
Cyclooctadiene on Different Cu/TiO2 Catalysts Com-
a
pared with Their Specific Cu(0) Surface Areas
2
0% -cyanopropylphenyl-siloxane) capillary column.
c
Turnover frequency
SCu(0)
(m /gCu)
2
.3. Catalyst Activation and Procedure
for CO–NO Reaction
Catalystb
(h
� 1
)
2
4
8
4I
8I
C
C
530
360
<5
<5
64 � 3
55 � 3
11
Pelletized catalyst was placed in a quartz infrared cell,
equipped with KBr windows, designed to treat the sample
in situ in controlled atmospheres and at different tempera-
tures. The reaction was carried out under static conditions
on the catalyst prereduced at different temperatures as
specified below. By connecting the IR cell through a needle
valve to a mass spectrometer (VG Micromass 100), FTIR
9
a
Reaction conditions: 433 K, 3 atm H2; sample pretreatment:
calcination at 623 K in O2 for 4 hours; Tred = 373–473 K.
b
Cu(wt.% ): 4C = 4.3, 8C = 8.7, 4I = 4.6, 8I = 8.5.
c
Moles of H2/moles of exposed Cu � h.

1
42
BOCCUZZI ET AL.
two methods, were found by some of us to give much copper sites. Recent measurements on different Cu planes
�
1
lower differences in turnover (130 vs 9 h ) (3). Among the (10) show that the Cu (111) plane presents the lowest ac-
Cu/TiO2 catalysts, the C samples are so much more active tivation barrier for H2 dissociation with respect to other
that some other factor must be taken into account.
examined planes, such as the (100) and the (110) planes,
The diene molecules are probably adsorbed on the sam- accounting in this way for the specific turnover frequency
ples prepared with both preparations both on the support differences observed.
and on the metallic particles. Therefore the difference in
3
.2. Mass Spectral Data on CO–NO Reaction
the activity is probably related to a different capability of
the copper sites generated in the two preparations to dis-
sociate H2 and/or to coordinate in close contact both the
olefin and the hydrogen. In Part I (1) evidence was provided
that the main difference between the C and I samples con-
sists in the presence in C samples of well-formed small Cu
crystallites, exposing a significant fraction of step and edge
sites whereas I samples expose only almost-isolated copper
sites embedded in an amorphous layer of reduced titania.
According to the model often used in describing the ad-
sorption of hydrogen on metal surfaces, when the molecule
approaches the surface it passes through a physisorption
well and then crosses a barrier caused by Pauli repulsion
between the H2 1�g orbital and the metal orbitals to enter
a region of chemical interaction (8). If the barrier for dis-
sociation is low enough, the molecule will end up as two
chemisorbed atoms. For noble metals, Cu and low-index Ni
and Fe surfaces the H2 chemisorption is energetically acti-
vated, whereas on transition metals such as Pd and W and
high-indexNi, Pt, and Cu planesthe dissociative chemisorp-
tion is almost spontaneous (9). The energetics of this inter-
action of H2 on metal surfaces is generally discussed by a
one-dimensional representation of the H/H2–metal poten-
tial as a function of the particle distance; instead it turns
out that there is a distribution of barriers with different ac-
tivation energy requirements depending on the chemisorp-
tion surface site. There is much experimental evidence that
hydrogen dissociation occurs preferentially on stepped sur-
faces of different metals (10–12).
Mass spectra of the gas phase composition and FTIR
spectra of the adsorbed species were taken simultaneously
at different stages of the reaction on differently prepared
and pretrated samples. Since the reaction of CO and NO
might be expected to yield CO2, N2, and N2O, m/z = 28 (CO
and N2), 30 (NO), and 44 (CO2 and N2O) were in partic-
ular examined: as an indication of the advancement of the
reaction the (m/z 44)/(m/z 30) ratio (N2O + CO2)/NO is re-
ported in Table 2 for the two differently prepared samples.
Since this quotient is the ratio between two reaction prod-
ucts and a reactant, its value rises with the catalyst activity.
It appears from the reported data that on the C sample
with low copper loading (2% ), by increasing the reduction
temperature, a decrease in the activity is observed, this ac-
tivity being restored by reoxidation followed by a second
mild reduction (2C423(II)). On the I samples the opposite
behavior is observed, since the activity increases by increas-
ing the reduction temperature; moreover, by increasing the
copper loading (4 and 8% ), an absolute increase in the ac-
tivity is observed and on these samples the activity remains
quite high also after more severe reductive treatments up
to 673 K.
The data reported in Table 2are absolute data. Asalready
shown in Part I (1) the exposed copper sites on the two cata-
lysts strongly differ as regards their nature and their abun-
dance depending on the preparation method and the reduc-
tive pretreatment. The opposite trends in catalytic activity
observed by increasing the reduction temperature can pos-
sibly be rationalized by looking at the activity and reactivity
Therefore, samples C, exposing small metal particles with
a large fraction of step sites, are expected to be very active in
hydrogen dissociation. On the contrary, I samples, exposing
isolated copper atoms embedded in reduced titania, are
probably inactive in hydrogen activation because of their
partial “positivization,” as shown by the close similarity of
the CO IR absorption band frequency on these sites and
TABLE 2
a
Ratio [(N O + CO )/NO] Determined after CO–NO Interac-
2
2
tions (Initial Ratios CO–NO 1 : 1) on Two Differently Prepared
and Pretreated Cu/TiO2 Samples
+
10
that of the absorption band of CO adsorbed on Cu (d )
sites.
Ratio (N2O + CO2)/NO
As for the higher specific turnover frequency of the 4C
sample with respect to the 8C one (Table 1), we recall here
that there are indications coming from the analysis of CO
absorption spectra reported in Part I (1) that the 4C sample
exposes small metal particles with mainly (111) microfacets
and high fraction of step and borderline sites; the 8C sam-
ple, together with step, borderline, and (111) sites, exposes
also (100) microfacets. Therefore, on the latter sample the
Sample
15 min RT
15 h RT
15 min 373 K
15 min 473 K
2
2
2
C423
C523
C673
0.24
0.12
0.06
0.3
—
0.1
—
—
—
4.0
—
� ꢀ
—
3.34
� ꢀ
1.25
5.0
0.53
5.5
0.33
2.0
—
2C423(II)
—
2
2
2
I523
I623
I673
0.08
0.25
0.77
0.27
11.1
a
(
111) surface sites are a lower fraction of the total exposed
Mass spectrometric ratio: (m/z 44)/(m/z 30).

CHARACTERIZATION AND ACTIVITY OF Cu/TiO2 CATALYSTS, II
143
of the copper sites and of the reduced titania exposed at the
surface at the end of the different pretreatments. In order to
obtain a deeper understanding of the mass spectral results
on the NO–CO reaction presented in Table 2, we present
and discuss below the FTIR spectra of the adsorbed phase
during the same experiments.
3
.3. FTIR Data of NO Interaction on CO
Precovered Samples
The full range IR transmission spectra and the ab-
sorption bands of the adsorbed species in the 2400–
�
1
2
000 cm range produced by interaction first of CO and
then of NO on sample 2C423 are shown in Figs. 1a and 1b,
respectively. A similar experiment made on sample 2I523 is
illustrated in Fig. 2. From the transmission spectra reported
in Fig. 1a it appears that the NO inlet causes changes in the
� � 1
1
2
400–2000 cm and the 1650–1200 cm ranges. The first
range is shown in more detail in Fig. 1b and will be dis-
cussed below; the second range is typical for carbonate-like
species and these absorption bands will not be discussed
further. The main point to be stressed here is that no bands
�
1
are observed in the 1900–1700 cm range, typical of the
molecular NO stretching and no changes are produced by
the NO interaction on the overall IR transmission pro-
file. We recall here that it has been shown previously (13)
that on n-type oxidic semiconductors the interaction with
molecules of high electron affinity produces an increase in
the IR transmission as a consequence of electron trans-
fer from populated donor sites of the defective oxide to
the acceptor molecule. This phenomenon is evident mainly
in strongly reduced samples, i.e., those with large devia-
tions from stoichiometry. The sample examined in Fig. 1a
is very mildly reduced. However, in spite of the absence
of adsorbed NO molecules and the absence of electronic
effects due to the interaction with the oxidic support, reac-
tion products of the CO–NO interaction are immediately
evident. In Fig. 1b, curve 2, it is shown that in the first 15 min
of interaction at room temperature bands relating to reac-
�
1
tion products at 2350 and at 2260–2230 cm , assignable to
CO2 and N2O, are produced. Moreover, broad absorption
�
1
at 2180–2220 cm appears; at the same time a band in the
�
1
CO stretchingregion at 2120cm develops, while the bands
at lower frequencies, 2100–2070 cm , decrease. The broad
absorption at 2180–2220 cm , irreversible on outgassing at
RT, can be assigned on the basis of literature data (14, 15)
to isocyanate species, while the 2100–2070 cm bands are precovered with CO by exposure to 10 mbar CO (curve 2), and after
assignable (see Part I (1)) to carbon monoxide adsorbed on
three-dimensional copper particles. After prolonged inter-
action (Fig. 1b, curve 3) the main effect is a strong intensifi-
cation of the band at 2120 cm . After heating to 373 K, this
band strongly decreases in intensity and shifts to 2115 cm
�
1
�
1
FIG. 1. FTIR spectra of NO interaction on sample 2C423 precovered
with CO. (a) Full range transmission spectra of the clean sample (curve 1),
�
1
CO/NO interaction for 15 min (curve 3). (b) Absorption spectra in the
�
1
2
400–2000 cm range after 10 mbar CO inlet (curve 1) and after CO/NO
interaction for 15 min (curve 2) and 3 h (curve 3).
�
1
On sample 2I523 the same interactions (Figs. 2a and 2b)
do not produce IR-detectable reaction products, i.e. CO2
and N2O or NCO and carbonate-like adsorbed species; also
�
1
:
�
1
at the same time the broad absorption at 2220–2180 cm
is diminished and the CO2 absorption band shifts to 2350 after prolonged interaction, only a minor increase in the
�
1
cm (not shown for sake of brevity).
IR transmission is observed (Fig. 2a, curve 3, and Fig. 2b,

1
44
BOCCUZZI ET AL.
and 2114 cm 1, increasing in intensity by NO interaction,
can be assigned to CO adsorbed on copper atoms oxidized
by the NO dissociation. The increase in the intensity can be
explained by taking into account that the absorption coef-
�
+
ficient of the stretching of CO molecules adsorbed on Cu
sites is higher than that of CO molecules adsorbed on Cu0
sites (16) or possibly that the surface exposed copper sites
produced by the oxidation with NO can increase. The dis-
�
1
appearance of the 2190 cm band, assigned to CO weakly
adsorbed on Ti4 sites, is mainly related to a decrease in
the CO partial pressure in the cell in the expansion into a
larger volume during the NO interaction.
+
The copper loading and the pretreatments of the samples
in the experiments reported in Figs. 1 and 2 are quite simi-
lar; the different reactivity must be related to the different
nature of the exposed copper sites in the two samples, as ev-
ident from the characterization work (1). It can be proposed
that on C samples NO molecules are dissociated on metallic
copper sites exposed at the surface of the stepped particles
and that the produced atoms adsorbed on the metallic sites
react with CO and NO giving rise to the products CO2, N2O,
and NCO; the absence of sites able to dissociate NO on the
2
I523 sample inhibits all the reactions. A confirmation of
this hypothesis comes from the study of the same interac-
tion on C samples with a higher copper loading and reduced
at different temperatures.
On sample 4C573 (Fig. 3a) the NO interaction on pread-
sorbed CO produces in a few minutes a significant in-
crease in the overall transmission spectra; moreover, quite
�
1
strong bands are produced in the 2400–2000 cm and 1650–
1
�
1
200 cm regions, while no bands are observed also in this
case in the NO stretching region. The increase in the IR
transmission is an indication that on this sample NO dis-
sociation occurs also on the reduced titania, giving rise to
an electron transfer reaction from the solid to the adsorbed
species, as already discussed (6, 13). In Fig. 3b it is clearly
�
1
shown that the bands at 2103 and 2071 cm gradually de-
crease and are almost completely depleted in 9 min of con-
�
1
tact, that a broad absorption at 2220 cm grows, and also
�
1
that the band at 2350 cm , already assigned to CO2, grows
Fig. 3b, curves 2–4); only in the last spectrum are compo-
nents at 2234–2264 cm , due to N2O, observed (Fig. 3b,
curve 5). Thus, there is a significant difference between the
(
FIG. 2. FTIR spectra of NO interaction on sample 2I523 precovered
with CO. (a) Full range transmission spectra of the clean sample (curve 1),
precovered with 10 mbar CO (curve 2), and after CO/NO interaction for
�
1
�
1
1
1
5 min (curve 3). (b) Absorption spectra in the 2400–2000 cm range after experiments on the 2C and the 4C sample in the rate of
0 mbar CO inlet (curve 1) and after CO/NO interaction at RT for 15 min
formation and in the relative intensity of N2O species in re-
spect to the other reaction product, the CO2. By increasing
(
curve 2) and at 473 K for 15 min (curve 3).
the copper loading and the reduction temperature (Fig. 4)
curve 2). The small increase in the IR transmission can be the same kind of CO–NO interaction leads to quite simi-
taken as an indication that small amounts of NO are dissoci- lar features. The main differences are again in the relative
ated at the surface of defective titania containing embedded intensities of the absorption bands; in particular the CO2
copper atoms or possibly on a surface layer of nonstoichio- to N O band ratio is very high, while the broad absorp-
2
�
1
metric copper titanate. By heating in the reaction mixture tion at 2220–2180 cm is quite weak. As already stated,
an increase in the intensity of the CO absorption band at spectra show that the relative amounts of CO , N O, and
114 cm is noticed (Fig. 2b, curve 3). The bands at 2120 NCO in the CO–NO interactions are strongly dependent
2
2
�
1
2

CHARACTERIZATION AND ACTIVITY OF Cu/TiO2 CATALYSTS, II
145
different product formation. As regards the metal loading
effect, it is evident that at high metal loading, CO2 forma-
tion is favored over N2O and NCO formation; this point
will be discussed below. Moreover, it appears that: (i) on
the samples prepared by wet impregnation, NCO species
are not formed at all after either mild or strong reductive
pretreatments (Figs. 2 and 5); (ii) on the sample prepared by
FIG. 3. FTIR spectra of NO interaction on sample 4C573 precov-
ered with CO. (a) Full range transmission spectra of the sample precov-
ered with 10 mbar of CO (curve 1) and after different CO/NO interaction
times: 3 min (curve 2), 6 min (curve 3), 9 min (curve 4), 15 min (curve 5).
�
1
(
(
b) Absorption spectra in the 2400–2000 cm range after CO interaction
curve 1) and after different contact times with CO/NO: 3 min (curve 2),
6
min (curve 3), 9 min (curve 4), 15 min (curve 5).
on different factors, viz. the extent of reduction of the sup-
ported metal and of the support, the metal loading, and the
preparation method of the samples.
On the basis of the illustrated data some observations can
be made on the effect of the metal loading and the prepa-
FIG. 4. FTIR spectra of NO interaction on sample 8C673 precovered
with CO. (a) Full range transmission spectra of the clean sample (curve 1)
and after CO/NO interaction for 5 min (curve 2) and 10 min (curve 3).
� 1
(
b) Absorption spectra in the 2400–2000 cm range of sample 8C523
precovered with CO (curve 1) and of 8C673 after CO/NO interaction for
ration method on the activity and the selectivity toward the 5 min (curve 2) and 10 min (curve 3).

1
46
BOCCUZZI ET AL.
4
% sample. Isocyanate species can be adsorbed on the oxi-
dic support and on copper sites in different oxidation states.
On oxidic supports their stability is primarily determined by
the chemical and electronic properties of the support (14a);
the absence of isocyanate species on the Cu/TiO2 samples
prepared by wet impregnation was interpreted in a previ-
ous paper (6) as due to a destabilization of these species
on the reduced, highly conductive titania. However, in the
C sample examined in this work, absorptions in the spec-
tral region characteristic of NCO species are observed also
on the strongly reduced samples. Therefore the previously
made hypothesis that isocyanate species were lacking on
Cu/TiO2 as a consequence of the high conductivity of re-
duced titania appears questionable; it is more likely that the
formation and stabilization of these species depends on the
oxidation state of copper and on the nature and structure
of the borderline region between metal and support. Re-
cently Solymosi and Bansagi (14b) studied the adsorption
ofHNCO on H–ZSM-5and on Cu-containingZSM-5, SiO2,
and Al2O3 catalysts. Their data clearly show that the loca-
tion of the absorption bands due to NCO species bonded to
Cu depends on the oxidation state of the copper, i.e., 2230–
� � 1 +
1
0
2
240 cm for Cu –NCO, 2200–2210 cm for Cu –NCO,
�
1
2+
and 2180–2185 cm for Cu –NCO. On the basis of these
�
1
literature data the broad absorption in the 2220–2180 cm
range can be assigned to different Cu and Ti–NCO species
coexisting on the surface.
Looking at the transmission section of the previously il-
lustrated experiments, it appears that the transmission pro-
files of the samples reduced at different temperatures differ
substantially as regards the shape. These differences, as al-
ready discussed previously (1, 6), are due to the presence in
the most strongly reduced samples of almost free electrons
in the reduced titania, absorbing/reflecting the IR radiation.
The C673 and 1673 samples show at the end of the reductive
treatment a very low residual transparency; on both of them
the admission of NO causes an immediate and significant
increase in the IR transmission (Figs. 4a, 6a, and 5a).
After a long time of contact a substantial increase of a
�
1
component in the CO absorption band at 2130 cm is ob-
served on the 2I673 sample (Fig. 5b) and on 2C673 (Fig. 6b).
FIG. 5. FTIR spectra of NO interaction on sample 2I673 precovered
with CO. (a) Full range transmission spectra of the clean sample (curve
�
1
On the 2I673 sample the increase of the band at 2130 cm
,
assigned to CO adsorbed on isolated Cu surface sites, up to a
1
1
), after interaction with 10 mbar of CO (curve 2), and after inlet of
0 mbar of NO and 5 min of interaction (curve 3). Absorption spectra in value six times the initial one, and the decrease of the bands
�
1
� 1
4+
the 2400–2000 cm range after different CO/NO interactions: 10 mbar of
CO (curve 1), inlet of 10 mbar of NO and 5 min interaction (curve 2), 30
min interaction (curve 3), after heating at 373 K (curve 4), after heating
at 473 K (curve 5).
at 2186–2208cm , assigned to CO adsorbed on Ti surface
sites, indicate that the reoxidation by NO of the reduced ti-
tania leads, on the wet impregnated samples, to an increase
�
1
of exposed copper sites. The increase of the 2130 cm
band is significantly smaller on the sample prepared by the
chemisorption hydrolysis a broad absorption in the NCO chemisorption–hydrolysis method, showing, before NO in-
stretching region is observed at all the examined copper teraction, the bands at 2103 and 2071 cm assigned to CO
loadings (2–8% ) and after all the reductive pretreatments adsorbed on Cu sites exposed on small copper particles.
�
1
(
from 423 to 673 K); (iii) the relative intensity of this ab- In order to explain these differences it can be proposed
sorption in respect to CO2, CO and N2O is maximum on the that in the wet impregnated samples a nonstoichiometric

CHARACTERIZATION AND ACTIVITY OF Cu/TiO2 CATALYSTS, II
147
favored by the presence of anionic impurities at the surface
of titania.
Summarizing the previously reported results, on the C
samples mildly reduced NO is dissociated on copper small
particles, as shown by the depletion of the bands at 2103 and
�
1
2
070 cm , not accompanied by the appearance of bands in
the NO stretching region. The adsorption of NO on single
crystal copper (100) and (110) surfaces studied by EELS
revealed that at temperatures above 100 K the adsorption
is dissociative and that all species containing nitrogen are
desorbed or dissociated below room temperature (17). On
samples reduced more severely, both I and C, the inlet of
NO produces a very strong increase in the IR transparency,
explained by an electron transfer process of free electrons
from the reduced titania to the oxygen atoms produced in
the NO dissociation. There is, therefore, an indication that
NO dissociation can occur on the examined catalysts on two
different kinds of sites, on copper small particles, and on
defective reduced titania.
The high activity in the CO–NO reaction observed on
C Cu/TiO2 samples already at RT is interesting. It was
important from both a fundamental and a practical point
of view to ascertain whether this high activity is a tran-
sient phenomenon, typical only of the freshly reduced cat-
alysts and significant only in a first reaction run, or whether
it is a permanent phenomenon. In Fig. 7, a comparison
is presented of the absorption spectra produced by three
FIG. 6. FTIR spectra of NO interaction on sample 2C673 precovered
with CO. (a) Full range transmission spectra of the sample precovered
with CO (curve 1) and after CO/NO interaction for 1 min (curve 2), 15 min
(
2
curve 3), and 15 min at 473 K (curve 4). (b) Absorption spectra in the
400–2000 cm range of the sample precovered with CO (curve 1) and
�
1
after CO/NO interaction for 5 min (curve 2), 10 min (curve 3), 15 min
curve 4), 15 min at 373 K (curve 5), and 15 min at 473 K (curve 6).
(
surface layer of copper titanate or, alternatively, a defect-
rich amorphous layer of composition TiOx, incorporating
copper atoms and small clusters, is formed during the re-
ductive treatments. This phase, by NO interaction, can be
reoxidized to TiO2, giving rise to the exposure at the surface
of greater amounts of copper sites. As already proposed in
the discussion of the HRTEM and TPR data presented in
FIG. 7. FTIR absorption spectra of different CO/NO interactions on
sample 4C523:10 mbar of NO on a freshlyreduced sample precovered with
0 mbar of CO (curve 1); readsorption of 10 mbar of CO and of 10 mbar
of NO, after outgassing of the first interaction (curve 2); readsorption, first
of 10 mbar of NO and then of inlet of 10 mbar of CO, after outgassing at
1
Part I (1), the formation of the amorphous layer could be RT of the previous gas phase (curve 3).

1
48
BOCCUZZI ET AL.
successive CO–NO interactions, all performed on the same
C523 sample; the first one is on a freshly reduced sample reaction, both the molecules must be activated. The best
curve 1), the second one after outgassing at RT the previ- catalysts for NO reduction with CO are Pt group metals (7),
In order to have an efficient catalyst for the CO–NO
4
(
ous reaction mixture and by readsorption firstly of CO and but these metals are expensive. Previously quoted work (4,
then of NO (curve 2), and the last one after outgassing at 5) showed that reduced copper catalysts can also be active
RT the previous reaction mixture and readsorbing firstly in this reaction. The following steps can be proposed in a
NO and then CO (curve 3). It can be observed that in the general scheme:
second run of the reaction, with the reagents introduced
Cu
as in the first run, only limited differences are observed in
the reaction products and in the adsorbed species formed;
in particular the absorption band of the produced CO2 ap-
pears weaker, but stronger bands are observed at 2260–
0
CO��→ Cu –CO
(1)
ad
Cu
1
2
+
�
NO ��→ Nad + Oad → N2 + Cu O
(2)
(3)
2
�
1
2
150 cm and in the CO stretching region. These features
+
�
0
Cu O + CO → CO2 + 2Cu
2
indicate that in the second run the reaction proceeds with
almost the same rate as on the freshly reduced sample. In
the last interaction (curve 3) the absorption band of the pro-
TiOx
�
�
NO + e ���→ N ad + O
(4)
ad
�
1
duced CO2 is almost absent; in the 2260–2150 cm region
NO + Nad → N2O
Nad + Nad → N2
Nad + CO → NCOad
(5)
(6)
(7)
�
1
the band is narrower, showing a maximum at 2200 cm
;
�
1
and the components at 2260–2230 cm , assigned to N2O,
are completely lacking, as also are some components of
the absorption band related to NCO species. In particu-
lar, the bands at 2230–2240 cm and 2220–2210 cm are
almost completely absent, bands which, according to the
assignment of Solymosi and Bansagi (14b), relate to NCO
groups bonded to reduced copper sites, Cu and Cu . The
absorption band of the CO stretching appears significantly
broader and less reversible by outgassing. It is important to
notice that the absorption frequencies of CO adsorbed on
reduced isolated copper sites and on oxidized copper sites
are almost the same, the main difference between the two
species being their resistance to outgassing, as also reported
by other authors (16).
N2Oad → N2 + Oad.
(8)
�
1
� 1
Carbon monoxide, as shown by FTIR, is molecularly ad-
sorbed on small copper particles and on isolated copper
atoms or clusters, whereas nitric oxide does not appear to
be molecularly adsorbed at all; in fact no bands are pro-
duced in the NO stretching region. However, the presence
of oxidation products of CO and NO immediately after NO
inlet at RT on 2C473 (Fig. 1b) indicates that NO is disso-
ciated on copper small particles, according to reaction (2)
and, on the samples reduced at high temperature, also on re-
duced titania, following reaction (4). Cu(111) and Cu(100)
surfaces are well known for their ability to dissociate nitric
oxide even at temperatures as low as 80 K (17). Taking into
account that NO dissociation is probably the rate determing
step in the CO–NO reaction, the highest activity observed
0
+
From these experimental data it can be deduced that the
readsorption of NO before CO produces more oxidized
copper sites, less efficient in the reaction. The lower activity
of copper oxidized sites produced in the last interaction
compared to the fully reduced ones confirms our previous
results on the impregnated Cu/TiO2 samples (6) and those
of Hierl et al. on Cu/Al2O3 samples (4) that reduced copper
sites are needed for the CO–NO reaction.
on the C samples when mildly reduced appears related to
the high dissociative NO activity of copper metal particles
and, furthermore, on these samples, Cu sites are readily
0
restored by CO, following reaction (4). The increase ob-
served in the CO2/N2O band intensity ratios when increas-
ing the copper loading on the C samples is probably related
to a surface structure sensitivity of the selectivity for N2O
formation. In Part I (1) an increase in metal particle size
with increase of metal loading was inferred from an anal-
ysis of the CO absorption bands, and therefore an effect
of metal particle size can be suspected, favoring reaction
3
.4. Relationship between Surface Structure of the
Catalyst and Activity in the CO–NO Reaction
Two reactions are thermodynamically favored at 298 K
between CO and NO:
(
a) 2NO + 2CO → N2 + 2CO2
b) 2NO + CO → N2O + CO2.
(
a) with respect to reaction (b). Particle-size-dependent ef-
fects have been observed on Rh catalysts in the selectivity
for the CO–NO reaction (18) and studies of this reaction
(
The mechanismsand the pathwaysofthe two reactionshave over Rh(110) and Rh(111) have demonstrated that the se-
not been unambiguously determined; in particular, contro- lectivity in these reactions is strongly sensitive to the struc-
versy persists as to whether N2O and NCO are intermedi- ture of the Rh catalyst metal surface (19); in fact on the
ates in the N2 formation.
(110) surface there was a much stronger tendency toward

CHARACTERIZATION AND ACTIVITY OF Cu/TiO2 CATALYSTS, II
149
N2 production than on the (111) surface. It was suggested zation work (1) showed that well-formed three-dimensional
that on the (110) surface NO dissociation is less inhibited copper particles are present, while on I samples reduced at
and leads to greater steady-state concentration of adsorbed high temperature a reduced titania phase is formed. On
N atoms, thus producing more favorable conditions for N this basis the apparently conflicting data can be explained.
atom recombination, relative to the NO + N atom reaction Moreover, it was also shown that reduced copper sites are
to make N2O. A similar explanation can be also proposed also necessary for CO activation, and in fact the reaction
in our case.
activity decreases when oxidized copper is produced by ex-
On the C samples reduced at high temperature the de- posing the reduced samples to NO alone.
0
crease in the exposed Cu sites, produced by the decoration
with TiOx suboxides shown by in the IR experiments dis-
cussed in Part I (1), produces a decrease in the activity. This
indicates that on these samples the decrease in the number
of active sites for reaction (3) is not balanced by the increase
ofthe active phase for the NO dissociation accordingto path
ACKNOWLEDGMENT
The financial support of the Italian CNR, contribution 95.1036, is ac-
knowledged.
REFERENCES
(
4) on the TiOx suboxide.
On the I samples the Cu and TiO2 are interdispersed at an
1
2
3
. Boccuzzi, F., Chiorino, A., Martra, G., Gargano, M., Ravasio, N., and
Carrozzini, B., J. Catal. 165, 129 (1997).
. Fragale, C., Gargano, M., Ravasio, N., Rossi, M., and Santo, I., Inorg.
Chim. Acta 82, 157 (1984).
. Di Castro, V., Gargano, M., Ravasio, N., and Rossi, M., in “Preparation
of Catalysts V” (G. Poncelet, P. A. Jacobs, P. Grange, and B. Delmon,
Eds.), p. 95, Studies in Surface Science and Catalysis, Vol. 63. Elsevier,
Amsterdam, 1991.
. (a) Hierl, R., Urbach, H. R., and Kn o¨ zinger, H., J. Chem. Soc. Faraday
Trans. 88, 314 (1992); (b) Hierl, R., Urbach, H. R., and Kn o¨ zinger, H.,
in “Proceedings, 10th International Congress on Catalysis, Budapest,
atomic level already in the mildly reduced samples, possi-
bly as an amorphous copper titanate surface layer. The low
activity of these samples indicates that the isolated copper
atoms exposed at the surface of these samples are far less
active in NO dissociation than in the hydrogen dissociation,
discussed previously. By increasing the reduction tempera-
ture a reduced, defective, electron-rich, nonstoichiometric
titania phase is produced, as shown by the decrease in the
IR transparency observed, and this phase can be efficient
in the dissociation of NO following the reaction (4). On
these samples, by increasing the reduction temperature, in-
creasing amounts of almost free electrons are produced that
can be efficiently transferred to NO molecules and thereby
dissociate them, as a consequence of an electron transfer
from the reduced solid to an antibonding orbital of the NO.
This explains the higher activity of the high-temperature-
reduced I samples.
4
1992” (L. Guczi, F. Solymosi, and P. T e´ t e´ nyi, Eds.), p. 2697. Elsevier,
Amsterdam, 1993.
5. (a) Hall, W. K., and Valyon, J., Catal. Lett. 15, 311 (1992); (b) Iwamoto,
M., Yahiro, H., Tanda, K., Mizuno, N., Mine, Y., and Kagane, S.,
J. Phys. Chem. 93, 3727 (1991).
6
. Boccuzzi, F., Guglielminotti, E., Martra, G., and Cerrato, G., J. Catal.
46, 449 (1994).
1
7
. Kudo, A., Steinberg, M., Bard, A. J., Campion, A., Fox, M. A.,
Malhouk, T. E., Webber, S. E., and White, J. M., J. Catal. 125, 565
(
1990).
8. Harris, J., Faraday Discuss. Chem. Soc. 96, 1 (1993).
9
. (a) Cacciatore, M., 12th Intern. Symp. on Plasma Chem., Minneapolis
4
. CONCLUSIONS
(
1995), Abstracts p. 1405; Cacciatore, M., and Billing, G. D., Pure Appl.
Chem. 68, 1075 (1996). (b) Sun, Q., Xie, J., and Zhang, T., Surf. Sci.
38, 11 (1995).
A marked difference in activity is observed in the hy-
drogenation of 1,3-cyclooctadiene between the catalysts
prepared by a chemisorption–hydrolysis method (C) and
those prepared by wet impregnation (I), the former show-
ing turnover frequencies about 100 times greater than the
latter. This large difference in activity is mainly ascribed to
differences in the morphology of the copper species present
3
1
0. (a) Hayden, B. E., in “Dynamics of Gas Surface Interactions” (C. T.
Rettner and M. N. R. Ashfold, Eds.), p. 137. Royal Society of Chem-
istry, Cambridge, UK, 1991; (b) Michelsen, H. A., and Auerbach,
D. J., J. Chem. Phys. 94, 7502 (1991) and references therein.
1. Poelsena, P., Lenz, K. L., Bernasek, S. L., Vorheij, I. K., and Comsa,
G., Surf. Sci. 189/190, 431 (1987).
1
1
2. Polizzotti, R. S., and Ehrlich, G., J. Chem. Phys. 71, 259 (1979).
at the surface of the two different kinds of catalysts: there 13. (a) Boccuzzi, F., Morterra, C., Scala, R., and Zecchina, A., J. Chem.
Soc. Faraday Trans. 2 77, 2069 (1981); (b) Boccuzzi, F., Ghiotti, G.,
Chiorino, A., and Guglielminotti, E., Surf. Sci. 269/270, 514 (1992).
are three-dimensional stepped small particles of Cu in the
first case (C), very efficient in hydrogen dissociation, and
isolated copper sites embedded in an amorphous reduced
layer in the second case (I) samples. In the CO–NO reac-
1
4. (a) Solymosi, F., Volgyesi, L., and S a´ rk a´ ny, J., J. Catal. 54, 336 (1978);
b) Solymosi, F., and Bansagi, T., J. Catal. 156, 75 (1995).
5. Matyshak, V. A., and Krylov, O. V., Catal. Today 25, 1 (1995).
(
1
tion, C samples show a decrease in activity on increasing 16. Sheppard, N., and Nguyen, T. T., in “Advances in Infrared and Raman
Spectroscopy” (J. H. Clark and R. E. Hester, Eds.), Vol. 5, Chap. 2,
p. 67. Heyden, London, 1978.
7. Wendelken, J. F., Appl. Surf. Sci. 11/12, 172 (1982).
8. Oh, S. H., and Eickel, C. C., J. Catal. 128, 526 (1991).
9. Peden, C. H. F., Belton, D. N., and Schmieg, S. J., J. Catal. 155, 204
(1995) and references therein.
the catalyst prereduction temperature, while the opposite
behavior is observed on I samples. In this reaction the gen-
erally accepted rate determining step is the NO dissociation
and this can occur on small copper metal particles and on
reduced titania:on mildlyreduced C samplesthe characteri-
1
1
1