A study of Cu/ZnO/Al2O3 methanol catalysts prepared by flame combustion synthesis

Journal of Catalysis 218 (2003) 67–77

www.elsevier.com/locate/jcat

A study of Cu/ZnO/Al₂O₃methanol catalysts prepared

by ﬂame combustion synthesis

J.R. Jensen,¹T. Johannessen, S. Wedel, and H. Livbjerg ^∗

Interdisciplinary Research Centre for Catalysis (ICAT), Department of Chemical Engineering, Technical University of Denmark, DK-2800 Lyngby, Denmark

Received 25 September 2002; revised 17 January 2003; accepted 20 January 2003

Abstract

The ﬂame combustion synthesis of Cu/ZnO/Al O catalysts for the synthesis of methanol from CO, CO , and H is investigated. The

2

3

2

oxides are generated in a premixed ﬂame from the acetylacetonate vapours of Cu, Zn, and Al mixed with the fuel and air prior to combustion.

The ﬂame-generated powder is examined by X-ray powder diffraction, determination of the speciﬁc surface area by the BET method,

determination of the copper dispersion in the reduced catalyst by a novel N O method, transmission electron microscopy, and testing of the

2

catalytic properties in a catalytic microreactor. A low peak temperature and quench cooling of the ﬂame tend to increase the dispersion of the

phases and the speciﬁc surface area of the particles. Properties of both the ternary composition, the three binary compositions, and the pure

oxides are discussed. The calculation of simultaneous phase and chemical equilibrium is used in the assessment of the phase composition

2

of the particles. The speciﬁc surface area varies from 100 m /g or a little below for samples without Al to several hundred m /g for the

respective compositions of pure Al O and ZnAl O . Copper dispersion after reduction varies from 1.8 to 14.1%. A ternary catalyst with

2

3

2 4

the composition of Cu:Zn:Al = 45:45:10 has the highest catalytic activity of all samples tested. This catalyst is also very selective and stable

toward thermal deactivation. The role of the individual catalyst components in the optimal catalyst is discussed.

Keywords: Aerosols; Nanoparticles; Flame synthesis; CuO; ZnO; Al O ; Methanol; Catalyst

2

3

1. Introduction

tribution. Jensen et al. [5] showed how to produce ZnAl₂O₄

spinel with a high surface area by co-combustion of volatile

Zn- and Al-acetylacetonates. The combination of mixing

at the molecular level and an extreme supersaturation for

the metal-oxide molecules formed by combustion ensures

the formation of a spinel phase with a high surface area

during the few milliseconds of residence time in the high-

temperature zone. The traditional wet methods which apply

(co-)precipitation are far more difﬁcult to control and re-

quire subsequent calcination, which is accompanied by an

unavoidable loss of surface area.

Small nanoparticles of metal oxides can be generated by

ﬂame combustion of volatile precursors. The so-called ﬂame

combustion synthesis, already widely used for materials like

carbon black, fumed silica, and titania [1,2], provides an

alternative route to catalyst manufacture as recently demon-

strated by Stark et al. [3] for V₂O₅/TiO₂composite particles

with excellent catalytic properties in the SCR process and by

Johannessen and Koutsopoulos [4] for a Pt/TiO₂catalyst for

SO₂oxidation.

This investigation expands the zinc-aluminate synthesis

by the methods of [5] to the ﬂame synthesis of the ternary

CuO/ZnO/Al₂O₃samples for the synthesis of methanol.

The Cu/ZnO/Al₂O₃catalyst system was chosen since it is

known to be structure sensitive [6]. A novel synthesis route

may lead to unexpected and unique catalyst properties. Fur-

thermore, the system is rather well investigated, which is

advantageous for the characterisation of a new route of syn-

thesis.

Flame synthesis is particularly effective for the genera-

tion of those material structures, which are inherently im-

portant in heterogeneous catalysis, i.e., high speciﬁc surface

areas for composite materials and nanostructured phase dis-

*

Corresponding author.

E-mail address: hl@olivia.kt.dtu.dk (H. Livbjerg).

Present address: Haldor Topsøe A/S, DK-2800 Lyngby, Denmark.

1

0021-9517/03/$ – see front matter

doi:10.1016/S0021-9517(03)00047-2

68

J.R. Jensen et al. / Journal of Catalysis 218 (2003) 67–77

The mechanisms of ﬂame aerosol synthesis

During combustion in the ﬂame zone the precursor

species are decomposed at high temperatures. The prod-

ucts are released as vapour molecules, which form particle

seeds by nucleation, when they become supersaturated dur-

ing cooling of the ﬂue gas. The particle seeds grow by

condensation of supersaturated vapours and by coagulation

following interparticle collisions. Concomitantly with parti-

cle growth, processes within the single particles alter their

shape toward compact forms (sintering) and rearrange the

phase composition toward stable crystalline phases by crys-

tal nucleation and growth. A special case arises when fast

combustion reactions generate molecules into a state of ex-

treme supersaturation, which is often followed by collision-

limited growth of particles, leading ultimately to very high

speciﬁc surface areas. For more volatile species, condensa-

tion occurs after the temperature eventually has reached the

saturation point. In multicomponent mixtures, the individual

components may condense consecutively, each in its own

temperature range.

The speciﬁc surface area, particle morphology, and crys-

talline structure of the particles can be altered by changing,

e.g., the ﬂame peak temperature, the high-temperature zone

residence time, and the precursor concentration. These con-

ditions can be varied within wide limits for a given burner

and further expanded by employing different burner types

and geometries. Pratsinis [1], Jensen et al. [5], and Jensen [7]

provide further details about ﬂame synthesis and references

to the relevant literature.

2. Experimental

Fig. 1. Flame reactor system with fuel, air, and precursor feed lines.

2.1. Flame reactor setup

addition of the vapours of the acetylacetonate of copper,

zinc, and aluminum, respectively. The precursor powders are

distributed on horizontal trays encased in thermostated alu-

minum boxes with sufﬁcient gas contact to ensure saturation

of the efﬂuent gas with precursor vapours. The vapour con-

tent in the precursor lines is thus independent of the ﬂow rate

of carrier gas and can be controlled by the temperature of the

sublimation unit. The vapour pressures of the precursors are

determined from the precursor weight loss for a given vol-

ume of carrier gas and are shown in Fig. 2 as a function

of temperature. Any desired ternary particle composition is

obtainable by the proper combination of saturator tempera-

tures and carrier gas ﬂow rates. Temperatures in the range

of 90–180 ^◦C have been adequate for the present experi-

ments. The copper precursor has the lowest vapour pressure

and its saturation unit was operated solely at its maximum

temperature of 180 ^◦C. Above 200 ^◦C the precursor starts to

decompose. All gas lines containing precursor vapours are

heated to 200 ^◦C to prevent condensation.

CuO/ZnO/Al₂O₃samples are prepared by combustion

synthesis from the volatile precursors consisting of the

acetylacetonates of copper, zinc, and aluminum. The ﬂame

reactor has been designed to obtain particles with large spe-

ciﬁc surface areas. A high surface area is favoured by a low

precursor vapour pressure [5], which is obtained in this re-

actor by the combustion of a homogeneous gas in which N₂,

O₂, and H₂(or CH₄) are well mixed with the volatile cata-

lyst precursors prior to ignition. The reactor is also provided

with means for the optional quench cooling of the ﬂame by

the injection of jets of cold air downstream from the ignition

zone. The particles therefore are exposed to the peak tem-

peratures only as long as necessary for the generation of the

desired molecular structures, which reduces the effect of sin-

tering of the coagulated particles [8]. Details of the reactor

setup are shown in Fig. 1.

The reactor feed is mixed from separately controlled

ﬂows of fuel (H₂or CH₄), air, and three separate lines with

nitrogen carrier gas containing the precursor vapours. Each

precursor line passes a sublimation unit for the controlled

Prior to combustion all feed lines are intimately mixed

in a swirl mixer. Combustion occurs in a ﬂat ﬂame stabi-

J.R. Jensen et al. / Journal of Catalysis 218 (2003) 67–77

69

Table 1

Flame operating conditions

Flame

1

2

3

4

(CH ) (CH , quench) (H ) (H , quench)

4

2

3

c

Air (dm /min)

10.1

0.986

0

2.15

–

10.1

0.986

0

5.2

2.15

–

10.1

0

1.71

0

–

1.24

995

10.1

0

1.71

5.2

–

3

c

Methane (dm /min)

Hydrogen (dm /min)

Quenched air (dm /min)

Oxygen/methane ratio

Oxygen/hydrogen ratio

3

c

a

3

c

1.24

954

◦

b

Peak temperature ( C)

1464

1398

a

The quench jets are positioned 12.5 mm above the ﬂame arrestor.

Measured peak temperature corrected for radiation loss.

b

c

◦

Fig. 2. The vapour pressures of Cu, Al, and Zn acetylacetonate as a function

of temperature. The vapour pressures are determined from the weight loss

of the saturator units due to sublimation of the precursors. The sublimation

enthalpies for Al- and Zn-acetylacetonate are estimated based on the linear

correlation between 1/T and the vapour pressure.

All ﬂows are given at 20 C, 1 atm. The carrier gas through the Al-,

3

Cu-, and Zn-(acac) saturators varies between 0 and 1 dm /min, depending

on the desired composition.

lized by a ﬂame arrestor—a stainless-steel plate of diameter

36 mm perforated by numerous small holes. A cylindrical

quartz tube of length 200 mm and inner diameter of 67 mm

surrounds the ﬂame and shields it from intrusion of the am-

bient air.

A ring, which can be positioned at any given height above

the ﬂame, is placed inside the quartz shield and holds the

nozzles for quench cooling of the ﬂame by cold air. The de-

sign of the quench-cooling device is very critical since the

nozzles must give the jets of cold air a slightly upward and

off-axis direction to create a stabilizing swirl that prevents

the complete distortion of the ﬂame, which would otherwise

occur [8]. This special conﬁguration enables us to quench

cool directly above the ﬂame front without disturbing the

ﬂame below the quenching level.

Fig. 3. Flame temperature proﬁles measured along the center line of the

burner for the selected standard test conditions (cf. Table 1), which are de-

signed to vary systematically with respect to peak temperature and cooling

rate.

Inhalation of small particles is potentially health haz-

ardous, so for safety reasons the entire setup is placed in a

fume hood.

cooling are compared in the photos of Fig. 4. The green light

emitted by Cu above approximately 1000 ^◦C illustrates the

changes in the high-temperature zone, which extends high

above the ﬂame front without quench cooling (a) but is con-

ﬁned to the region below the ring during quench cooling (b).

It is principally the temperature–time history and the

concentration of precursor vapours of the reacting ﬂow in

the ﬂame which determine the properties of the efﬂuent

aerosol particles. Both sets of ﬂame conditions can be altered

within wide limits to achieve advantageous particle prop-

erties. Table 1 summarizes the experimental design for the

thermal ﬂame conditions, which are altered systematically

by switching on or off the quench cooling and by a change

from a lean hydrogen mixture to a rich methane fuel mixture.

The corresponding axial temperature proﬁles for the four se-

lected conditions are shown in Fig. 3. They are measured

with a platinum–platinum/10% rhodium thermocouple with

0.3-mm free wires and a welding bead with diameter 1.3 mm

and corrected for radiation inﬂuence [7]. The four different

conditions in Table 1 and Fig. 3 combine a signiﬁcant vari-

ation of the peak temperature—high for methane and low

for hydrogen—with a similarly signiﬁcant variation in the

ﬂame-cooling rate—a high rate with quench cooling and low

rate without. The methane ﬂame with and without quench

The powder produced in the ﬂame is collected on a

47-mm polycarbonate ﬁlter with 1-µm pores through which

the efﬂuent gas is withdrawn by a vacuum pump after being

cooled to 70 ^◦C. The particles are subjected to the following

examinations:

The BET surface area is measured by multipoint nitro-

gen adsorption (Gemini 2630, Micromeritics) at 77 K. An

equivalent primary particle diameter can be estimated from

the BET surface S_Aby

6

d_BET

=

,

ρ_pS_A

where ρ_pis the solid phase density.

X-ray diffraction (XRD) is obtained with a Cu-K_αra-

diation diffractometer (Philips PW1820/3711). The diffrac-

70

J.R. Jensen et al. / Journal of Catalysis 218 (2003) 67–77

Cu after a correction for a slow diffusion of O into the bulk

material. The measurements are made at ambient pressure

with a constant ﬂow of gas through a ﬁxed bed of cata-

lyst particles by switching the feed gas composition from

pure helium to 2% N₂O in helium. The efﬂuent gas passes

through a freeze trap with liquid nitrogen, which selectively

removes unreacted N₂O from the gas. Nitrogen evolved by

the oxidation reaction can then be determined accurately by

a calibrated thermal conductivity detector. Prior to the mea-

surements the samples are reduced in pure hydrogen for 1 h

at 220 ^◦C. The method is suitable for the small amounts of

catalyst of the present investigation and is applied to the par-

ticles in situ in the catalytic test reactor described below. It

was proved that the N₂O treatment does not change the cat-

alytic activity or the Cu dispersion. The N₂O treatment is

made at 90 ^◦C with 2% (mol/mol) N₂O in He. Pretreatment

of the sample, reduction, etc., is made according to the pro-

cedures described in [7].

(a)

(b)

Fig. 4. Photograph of the ﬂame reactor which illustrates the changes in

the high-temperature zone during quench cooling. The ﬂame contains Cu,

which emits green light above approximately 1000 C. (a) Methane–air

◦

ﬂame and (b) quenched methane–air ﬂame.

2.2. Catalyst tests

tograms can be measured expediently directly on the particle

ﬁlter cake subject to a correction for a weak background

signal from the ﬁlter material. Crystallite dimensions in dif-

ferent crystal directions were estimated from line broad-

ening using the Scherrer equation [9]. The peak widths at

half height were estimated by ﬁtting the peaks to a multiple

Gaussian distribution using a least-squares method. When

the crystallite size gradually becomes smaller, line broad-

ening eventually dominates the diffractogram, which hence

cannot be resolved reasonably. In this case the material ap-

pears X-ray amorphous, although in many cases more or less

perfect nanocrystalline particles and phases may presumably

be present. Further details of the XRD analysis are shown

in [7].

Transmission electron microscopy (TEM) is used for par-

ticles sampled directly from the hot gas by diffusional and

thermophoretic²deposition on 300-µm mesh Ni TEM grids

with carbon ﬁlm. The grid is inserted manually by means of

pincers for 5–10 s at the center line of the ﬂame approxi-

mately 100 mm above the diffuser. A Philips 300-kV TEM

was used to image the particles and to measure the overall

chemical composition by EDAX (energy-dispersive X-ray

analysis).

Measurements of the catalytic properties are made with

the catalytic microreactor shown in Fig. 5. A small amount

of powder sample (∼ 20 mg) collected from the ﬂame gen-

erator is transferred to a Pyrex glass capillary tube with an

internal diameter of 2 mm so that it forms a small bed held

in place by quartz wool plugs at one end of the tube. A con-

striction made in the wall of the capillary tube stabilizes the

innermost wool plug, positioned at the gas outlet from the

bed (cf. Fig. 5)

When a catalyst is tested, its capillary tube is mounted in

a cylindrical recess sealed by Viton o-rings at the upper part

of the reactor assembly. When the reactor is reassembled,

the capillary tube is encased in a cylindrical steel jacket. The

feed gas mixture enters through the cylindrical annulus be-

tween the capillary tube and the surrounding steel jacket in

downward ﬂow and passes the catalyst bed, leaving the reac-

tor in upward ﬂow internally in the capillary tube. The steel

jacket ﬁts into a hole bored in a thermostated cylindrical

block of aluminum–bronze so that one end, containing the

catalyst, is held isothermal due to the high heat conductiv-

ity of the bronze, while the other end protrudes outside the

bronze block and is kept at a lower temperature to protect

the Viton o-ring seals and inlet and outlet gas tubing con-

nections. The reactor operation is trouble-free for catalyst

temperatures up to 400 ^◦C.

Copper dispersion³is measured with the N₂O method

in which N₂O is assumed to react with the surface copper

atoms selectively by the reaction 2Cu + N₂O → Cu₂O +

N₂[10–12]. We have developed a novel adaption of the N₂O

method by which the transient evolvement of N₂in a mi-

croreactor is used to compute the concentration of surface

Due to the miniature catalyst bed and the high heat con-

ductivity of the bronze block, the reactor and the catalyst

bed stay effectively isothermal, which has been conﬁrmed

by measurements of the gas exit temperatures with thin ther-

mocouples inserted into the capillary tube.

The reactor material is nickel-free steel. The high-tempe-

rature zone of the steel jacket is furthermore lined inter-

nally with copper. These materials are necessary to avoid

formation of volatile carbonyls, i.e., Ni(CO)₄and Fe(CO)₄,

2

Thermophoresis is the mechanism by which particles are transported

in a ﬂuid with a thermal gradient, i.e., from high temperature to low tem-

perature. Therefore, particles are readily deposited on a cold surface—the

TEM grid—when it is inserted in a hot gas containing the metal oxide par-

ticles.

3

The dispersion of the Cu crystallites is deﬁned as the ratio of surface

Cu atoms to total Cu atoms.

J.R. Jensen et al. / Journal of Catalysis 218 (2003) 67–77

71

Fig. 5. Catalytic microreactor for testing of catalyst powder. All external units including pressure and ﬂow controllers, active carbon trap, gas sampling valve,

and the gas chromatograph are omitted from the drawing.

from CO in the feed gas, which pollute the catalyst with Ni

and Fe.

The GC is calibrated with gas standards of known com-

position.

Nickel carbonyl already forms at ambient temperature

from the nickel in the stainless steel feed lines and the ﬂow

controllers. The feed gas must be puriﬁed in an active car-

bon ﬁlter to avoid a clearly discernible, gradual decline of

the selectivity due to nickel poisoning of the catalyst (cf.

Fig. 6).

The following standard test procedure is followed: The

standard reaction conditions are 220 ^◦C, 16 bar, and approx-

imately 20 mg of catalyst loaded into the reactor. The feed

ﬂow rate is 25 (N cm³)/min. Initially, the sample is reduced

by increasing the temperature (20 ^◦C/min) from room tem-

perature to the reaction temperature while ﬂushing the reac-

tor with the feed gas (40 (N cm³)/min).

The sample initially forms a packed bed within the glass

capillary tube. Due to the small particle size, a pressure drop

of several bars builds up to force the gas through the bed.

The sample, however, shrinks on reduction by the reaction

mixture during heating of the reactor, and the bed turns into

a coherent, highly porous pellet with a narrow annular slit

between the pellet and the glass wall for gas passage. The

pressure drop subsequently becomes negligible and the re-

actor in fact is transformed into a single-pellet reactor with a

The feed gas is drawn from a cylinder with a synthesis gas

of ﬁxed composition: 10% CO, 10% CO₂, and 80% H₂with

an impurity level less than 0.002%. The pressure of the re-

actor is controlled by a two-stage reduction valve at the feed

cylinder. The ﬂow of gas through the reactor is controlled by

a ﬂow controller for constant downstream pressure (Brooks

8842 with needle no. 1) positioned downstream from the re-

actor. The reactor pressure is monitored by manometers after

the reduction valve and before the ﬂow controller, respec-

tively.

The product gas stream leaving the reactor is heated to

110 ^◦C to avoid condensation of products. Downstream from

the ﬂow controller, at ambient pressure, the ﬂow rate is mea-

sured by an optical/digital bubble ﬂow meter.

The composition of the efﬂuent gas is analysed by a Shi-

madzu GC-17A gas chromatograph. A 6-way gas sample

valve with a 100-µl sample loop injects a gas sample into

the carrier gas stream in which the organic compounds and

the ﬁxed gases are separated on two capillary columns and

detected by a ﬂame ionization detector and a thermal con-

ductivity detector, respectively.

72

J.R. Jensen et al. / Journal of Catalysis 218 (2003) 67–77

Fig. 6. Long-term test of catalyst activity and selectivity for catalyst “A”

(cf. Table 2). The ordinate denotes the differential rates under the standard

conditions: 220 C, 16 bar, approximately 20 mg of catalyst and 10.0% CO,

Fig. 7. X, T diagram for the methanol synthesis (cf. footnote 4).

◦

in the gas:

10.0% CO , and 80% H . The ﬁgure also illustrates the inﬂuence of nickel

2

poisoning of the catalyst before the nickel carbonyl trap was installed.

CO₂+ 3H₂ꢀ 3CH₃OH + H₂O,

CO + H₂O ꢀ 3CO₂+ H₂.

cylindrical pellet with a diameter of 1.8 mm and a length of

20–30 mm.

Both are exothermic, reversible reactions. In the X, T di-

agram of Fig. 7, both reactions are at equilibrium on the

equilibrium curve, which for control of the chemical analy-

sis is seen to be nicely approached by the measurements at

high temperature. The activity measurements for the cata-

lyst samples are all made at 220 ^◦C and a space velocity of

approximately 20 (N cm³)/(g_cats). At these conditions, X

is less than 2.5% and far below the equilibrium curve for all

samples tested. The observed rate is hence the differential (or

initial) rate measured at 16 bar and a composition approx-

imately equal to the feed gas composition. The measured

activity in Table 2 is shown as the molar rate of methanol

production per gram of unreduced sample, or as the turnover

frequency (TOF) deﬁned as the molar rate of methanol pro-

duction per mole of surface Cu.

The reactor is then kept at the standard conditions for 24 h

and it is ascertained that the efﬂuent gas composition has

stabilized before the run is disrupted.

After the activity test, the copper dispersion is measured

by N₂O titration, which is initiated by ﬂushing the reactor

with pure hydrogen for 1 h at 220 ^◦C and 1 atm. The reac-

tor is then cooled to 90 ^◦C, at which temperature the N₂O

titration is made as described in [7].

Fig. 7 shows the conversion X⁴for the carbon oxides as

a function of temperature and space velocity for sample A

of Table 2, which is the most active of the catalysts tested.

The following two reactions describe the chemical changes

4

The degree of conversion is deﬁned as X = F /F

where F

total molar ﬂow rates of

is

Mf

CO

We assessed the effect of pore diffusion on the measured

activity by estimating values of the effective diffusivities and

the efﬂuent methanol molar ﬂow rate and F

CO

CO + CO in the reactor feed.

2

Table 2

Correlations between ﬂame conditions, chemical composition, speciﬁc surface area, Cu dispersion, and catalytic properties

Catalyst

a

A

B

C

D

E

Flame

H , quench

2

H , quench

2

H , quench

2

CH , quench

4

CH

4

Atomic ratio Cu:Zn:Al

BET surface area (m /g)

45:45:10

123

8.2

50:50:0

113

8.8

4.4

3.8

50:0:50

195

5.9

–

1.8

45:45:10

93

11.2

4.3

9.7

23.3

10.6

50:50:0

66

15.1

15.9

3.4

2

BET equiv. particle size (nm)

ZnO crystallite dimensions (nm)

Copper dispersion (%)

b

3.4

8.9 (14.1)

c

2

Cu surface area (m /g)

21.4

11.6

9.7

27

5.6

59

8.7

31

d

Estimated Cu particle size (nm)

Activity (µmol/(g s))

3.66

7.1

19.2

1.61

6.8

19.2

0.27

2

20.7

1.99

3.5

19

1.07

5.1

16.5

cat

−3 −1

TOF (10

3

s

)

SV (cm /(g s)) (st. T , P )

cat

a

3

◦

In the quenched ﬂames, the quenched-air ﬂow is 6.8 dm /min (25 C, 1 atm).

b

c

d

Dimension of the ZnO crystallites estimated using the Scherrer equation.

Catalyst A has the Cu dispersion 14.1 when it is not exposed to synthesis gas prior to the N O titration.

Based on overall copper content, Cu surface area and a density of solid copper of 8920 kg/m .

2

3

J.R. Jensen et al. / Journal of Catalysis 218 (2003) 67–77

73

calculating the effectiveness factor for the actual reaction

3.1. Calculation of chemical and phase equilibrium

conditions, treating the catalyst bed as a cylindrical catalyst

particle with a diameter of 2 mm [7]. The calculated effec-

tiveness factor is between 0.9 and unity for all tests, which

proves that the activity measurements are made under condi-

tions of insigniﬁcant inﬂuence of pore diffusion. The effect

of external diffusion resistance can also be neglected due to

the narrow space available for gas ﬂow in the cylindrical an-

nulus surrounding the particle.

As an a priori assessment of the volatility and phase com-

position of the particles in the ﬂame, we have calculated

the simultaneous chemical and phase equilibria for the given

overall ﬂue gas composition.

The calculations are made by the method of Michel-

sen [14], which minimizes the total Gibbs free energy of

a multicomponent, multiphase system at constant pressure

and temperature subject to the conditions of a given content

of elements. The gas is assumed ideal. All condensed phases

are assumed immiscible so that each condensed component

forms a separate pure solid or liquid phase. In the calcu-

lations we have attempted to include an exhaustive list of

all possible vapour phase and condensed components which

can be formed by the given elements in the ﬂame. The ther-

mochemical properties needed for the calculations are taken

from published data sources [7]. Examples of equilibrium

calculations are shown in Fig. 8 for the synthesis conditions

of the ternary sample.

3. Results and discussion

Preliminary tests of the ﬂame-generated ternary Cu/ZnO/

Al₂O₃catalyst prove that they are active and selective

for methanol synthesis. The atomic ratio of Cu:Zn:Al =

45:45:10 represents a ﬂat optimal activity for the ﬂame-

generated catalysts. The ﬂame conditions, however, affect

the observed activity signiﬁcantly.

To clarify the role of each of the three constituent ele-

ments, we shall compare the ternary catalyst with the pure

oxides and the three binary systems with respect to struc-

ture and catalytic properties for different ﬂame conditions.

Table 2 shows the samples which were prepared and tested

with Cu as one of the components. Only those samples have

potential for use as methanol catalysts. The activity per unit

surface area is very low for both pure copper and pure ZnO,

but in combination with a number of speciﬁc oxides and in

particular ZnO, copper exhibits a strongly enhanced activ-

ity due to the so-called synergic promotion [13]. Table 2

includes the standard ternary catalyst and the Cu/ZnO bi-

nary catalyst, both for two different ﬂame conditions and a

single sample of the Cu/Al₂O₃binary catalyst. Studies of

ﬂame-generated ZnO/Al₂O₃particles are made by Jensen

et al. [5] and particle properties for the pure oxides are stud-

ied by Jensen [7] and Jensen et al. [5].

The calculations ignore both surface energy effects,

which are potentially important for the stability of nanoparti-

cles, and kinetic and diffusional effects, which may delay the

actual attainment of equilibrium. However, in spite of these

shortcomings, equilibrium calculations may still point to

the likely identity of crystalline phases and volatile species

formed at different temperatures and at least approximately

identify the temperature ranges for the phase transforma-

tions. The limitations of the equilibrium approach is illus-

trated, though, by the fact that we have never identiﬁed the

stable form of alumina, α-alumina, in the ﬂame-generated

nanoparticles. Evidently, the nucleation or stability of this

alumina form is hindered in the small particles. α-Alumina

is therefore not included in the calculations.

At the particle formation temperatures from 600 to

1500 ^◦C, water has no inﬂuence on the thermodynamic sta-

Fig. 8. Equilibrium distribution of elements in the CuO/ZnO/Al O system between solids and an ideal gas phase as a function of temperature. The solid

2

3

phases are shown in part A whereas the gas phase species are shown in part B. In this calculation α-Al O is excluded from the possible phases. The moles of

2

3

copper in the phases containing copper are normalized by the total moles of copper. The copper-containing phases therefore add up to 1. The zinc-containing

species are normalized the same way. Phases containing aluminum only add up to unity at temperatures where none of the aluminum is bound in Cu- and

Zn-containing species.

74

J.R. Jensen et al. / Journal of Catalysis 218 (2003) 67–77

work. For ZnO the attainable surface areas are much smaller,

i.e., in the range of 30–50 m²/g. In XRD ZnO is wurtzite.

Particles of the binary ZnO/Al₂O₃composition are char-

acterized by the formation of the spinel phase of zinc alu-

minate ZnAl₂O₄, which for any overall composition appears

to be formed to the full extent possible. The speciﬁc surface

area increases considerably compared to that of pure ZnO

when the overall composition contains even small amounts

of Al₂O₃and quite high and thermally stable ZnAl₂O₄sur-

faces can be made in the premixed ﬂame. Equilibrium cal-

culations [7] substantiate that the ZnAl₂O₄spinel is the

most stable phase below approximately 2000 ^◦C (provided

α-Al₂O₃is excluded from the calculation). The readiness

and high yield with which ZnAl₂O₄is formed in a single

step from a homogeneous gas mixture proves the potentiality

of the premixed ﬂame for the synthesis of composite mate-

rials with high surface areas. The contribution of ZnAl₂O₄

to an increased surface area and thermal stability is a likely

explanation of the beneﬁcial role of alumina in the methanol

synthesis catalyst.

3.4. CuO/Al₂O₃system (sample C of Table 2)

The phase stability of this system resembles that of the

ZnO/Al₂O₃system. Equilibrium calculations show that ei-

ther copper(II) aluminate (CuAl₂O₄) or copper(I) aluminate

(Cu₂Al₂O₄) form stable solid phases for temperatures be-

low 2200 ^◦C [7]. The formation of CuAl₂O₄is conﬁrmed

by XRD and seems to take place to the full extent possi-

ble for the given precursor composition [7]. Due to the low

vapour pressures of the aluminates, the ﬂame-synthesized

particles consistently have high speciﬁc surface areas, which

is the case indeed for the sample C subjected to catalytic

test in Table 2. Evidently copper aluminate is an advan-

tageous precursor for highly dispersed Cu because disper-

sions in the order of 25% was measured on samples of the

CuO/Al₂O₃particles after being cautiously reduced by He-

diluted H₂[7]. Therefore, the CuO/Al₂O₃sample at ﬁrst

sight appears highly promising for Cu-catalyzed reactions

and was indeed proven very active for the dehydrogenation

of ethanol to acetaldehyde and ethyl acetate [7]. Unfortu-

nately, the high Cu surface area rapidly deteriorates when

exposed to elevated hydrogen partial pressures and in partic-

ular to the methanol reactant mixture [7]. In Table 2, sam-

ple C, in spite of a high BET surface area, has the lowest Cu

dispersion. Although Cu/Al₂O₃shows methanol synthesis

activity, the Cu-based turnover frequency is clearly inferior

to that of the ZnO-containing catalysts. The overall catalytic

activity of sample C is therefore poor and the CuO/Al₂O₃

sample is of little interest for methanol synthesis.

Fig. 9. X-ray powder diffractograms for the ﬁve samples of Table 2.

bility of the particles, and it is neglected in the equilibrium

calculations.

3.2. Identiﬁcation of phases

Fig. 9 shows the powder diffractograms for samples A–

E of Table 2. The ZnO phase gives the strongest diffraction

signal. In fact, almost all the identiﬁable peaks can be re-

ferred to ZnO. The same difference in signal intensity is seen

for the binary samples. Only for sample E, separate CuO

peaks are detectable in the diffractogram. The dispersion of

the different phases increases when the ﬂame temperature is

decreased and when alumina is added to the sample.

3.3. ZnO/Al₂O₃system

We have investigated the premixed-ﬂame synthesis of

particles of this binary composition quite thoroughly [5,7].

For the pure oxides ZnO and Al₂O₃the speciﬁc surface area

of the generated particles can be maximized by decreasing

the ﬂame temperature or decreasing the precursor vapour

pressure. For Al₂O₃, surface areas in excess of 400 m²/g

can readily be obtained. The crystal structure is γ -alumina or

X-ray amorphous depending upon the primary particle size.

Particles synthesized in ﬂames are usually crystallized in

metastable phases such as γ - and δ-aluminas [1]. The struc-

tures of δ- and γ -alumina are related in that the long-range

order has increased in δ-alumina. Further heat treatment of

the δ alumina leads to formation of the stable α-phase. How-

ever, only the γ -phase was detected with certainty in this

3.5. CuO/ZnO system (samples B and E of Table 2)

Equilibrium calculations for this binary system [7] indi-

cate that the solid phases are formed almost simultaneously

J.R. Jensen et al. / Journal of Catalysis 218 (2003) 67–77

75

from the gas species Zn(g) and CuO(g) when the ﬂame tem-

perature is cooled below approximately 1000^◦C. The stable

solids are ZnO, CuO (T < 600 ^◦C) and Cu₂O (T > 600 ^◦C),

of which ZnO is identiﬁed in the particles by XRD (Fig. 9)—

CuO in sample E and vestigially in sample B.

There are no stable mixed oxide phases, but there seems

to be a certain, although limited, solubility of CuO in

ZnO [13], but basically the oxide phases of the two metals

remain segregated either in separate particles or in composite

particles with possibly only two different crystallites grown

together.

For pure CuO the speciﬁc surface area is only moderately

high and comparable to that of pure ZnO [7]. However, the

binary particles tend to have surfaces larger than those of

the pure oxides, probably because the immiscibility of the

two oxide phases blocks for sintering at the contact points

between unlike particles [7].

Two binary samples were included in the tests shown in

Table 2. Both samples have a composition of CuO:ZnO =

1:1 but were synthesized at different temperatures and there-

fore subjected to different degrees of sintering. Sample E

was synthesized in a high-temperature ﬂame and has a lower

BET surface area (66 m²/g) than sample B (113 m²/g), syn-

thesized in a low-temperature ﬂame.

Fig. 10. Transmission electron micrographs of particles of samples A and E

from Table 2, which shows the inﬂuence of aluminum and ﬂame temper-

ature on the phase dispersion. Sample A is generated at a low ﬂame tem-

perature with quenching and 10% aluminum and sample E at high ﬂame

temperature without quenching and no aluminum. In sample A the particles

form dendritic clusters of coagulated very small primary particles. Sample E

consists of separated individual crystals of ZnO (faceted, regular) and CuO

(irregular). The chemical composition of the individual particles in sample

A cannot be resolved due to limitations in resolution of the instrument.

Evidently, the higher temperature principally affects the

ZnO phase of sample E because the ZnO crystallite size in-

creases disproportionately more than the BET surface area

decreases (Table 2). The CuO phase is probably less af-

fected by sintering as evidenced by the comparable Cu dis-

persion (after reduction) for samples B and E. The growth

of ZnO crystallites in the high-temperature ﬂame possibly

occurs subsequent to the initial particle formation as a sec-

ondary transport of vapours from small to large particles

(Ostwald ripening). This was observed for the ZnO/Al₂O₃

system under similar conditions with ZnO in excess of

that needed for the formation of aluminate [5]. In the low-

temperature ﬂame the secondary growth of ZnO crystallites

is suppressed, not due to the lower peak temperature but

rather to the faster cooling rate of the quenched hydrogen

ﬂame.

The high dispersion of the phases and the high speciﬁc

surface area for sample B emphasize the importance of con-

trolling the peak temperature of the ﬂame to the lowest ad-

equate level to avoid unnecessary sintering. The Scherrer

crystallite size for ZnO is less than the equivalent particle

size based on the BET area, which indicates that each parti-

cle consists of several crystallites grown together. For sam-

ple E there are identiﬁable XRD peaks for CuO (cf. Fig. 9).

Their Scherrer crystallite size is 18.9 nm [7]. Thus, for sam-

ple E, generated at a high temperature, the crystallite sizes of

ZnO and CuO and the BET-equivalent size are all three in the

range of 15.1–18.9 nm, which seems to further illustrate the

effects of excessive sintering in the high-temperature ﬂame.

This is also seen in Fig. 10 (right) where large, faceted crys-

tals of ZnO appear on the TEM image. After reduction of

sample E, the equivalent copper diameter is increased to

31 nm, so reduction of CuO to Cu for this catalyst is accom-

panied by an increase in size, probably by the coalescence

of adjacent particles.

The reduction of the CuO/ZnO samples (samples B

and E) yields Cu dispersions considerably higher than that

of the CuO/Al₂O₃sample (sample C) in spite of the fact

that the BET surface area of the latter sample is much larger.

Together with the higher turnover frequency of catalysts B

and E compared to catalyst C, this allows us to conclude

that ZnO stabilizes the Cu crystallites formed by reduction

and simultaneously modiﬁes them chemically to be more

catalytically active. The latter observation conﬁrms a signif-

icant synergic promotion of ZnO in the ﬂame-synthesized

catalyst. The cause of the intensely investigated synergy be-

tween Cu and ZnO in the low-pressure methanol catalysts is

as yet unsolved and is still open to the widely different in-

terpretations (cf., e.g., [6,13,15]). These can be grouped into

two classes: one in which the activity is related to the ex-

posed surface area of the Cu crystallites, although the copper

surface is modiﬁed by the presence of ZnO, and the other in

which the active sites are Cu⁺cations dissolved in the ZnO

crystals and the activity therefore is not related directly to

the Cu surface area as determined by the N₂O method, but

rather to the surface of ZnO. Catalysts B and E have very

similar copper dispersions while the ZnO crystallite size is

much larger for catalyst E. Since the copper-based turnover

frequency is only slightly smaller for catalyst E than for cat-

alyst B, the activity obviously does not depend on the ZnO

surface, which is vastly different for the two catalyst sam-

ples.

76

J.R. Jensen et al. / Journal of Catalysis 218 (2003) 67–77

3.6. CuO/ZnO/Al₂O₃system (samples A and D of Table 2)

drogen ﬂame. This can be explained by a secondary growth

of the ZnO crystals in the ternary system equivalent to that of

the binary case as discussed above. The turnover frequency

is moderately lower for the catalyst from the methane ﬂame

for both the binary and the ternary catalyst, which may be

caused by the larger ZnO crystallite size. It explains why

catalyst D has the poorer activity of the two ternary cata-

lysts, in spite of the fact that they have approximately the

same copper dispersion.

The contribution of aluminum and a low ﬂame temper-

ature in the creation of a high dispersion of the phases is

evident from Fig. 10, which compares the TEM images of

samples A and E.

The ternary samples were synthesized in a high-tempera-

ture ﬂame (D) and a low-temperature ﬂame (A) with an

overall composition of Cu:Zn:Al = 45:45:10. These two

catalysts have the highest activity of all. The copper disper-

sion is approximately 10% for both reduced samples but the

turnover frequency is higher for the catalyst synthesized at

the low temperature, which thus yields the highest activity

of the ﬁve catalyst samples.

Fig. 8 shows the calculated phase distribution at equilib-

rium for the overall composition of the ternary sample under

the conditions in the ﬂames, again with α-alumina excluded

from the calculations. At temperatures above 1300 ^◦C the

aluminates ZnAl₂O₄or ZnAl₂O₄+ CuAl₂O₄are the only

stable solid phases (Fig. 8A). Between 1300 and 1700^◦C

ZnAl₂O₄is the most stable solid phase. Due to the molar

ratio of zinc, copper, and aluminum (Zn:Cu:Al = 3:3:4) in

the equilibrium calculation, CuAl₂O₄is not present at equi-

librium since all aluminum is bound as ZnAl₂O₄. Above

1700 ^◦C CuAl₂O₄appears in the equilibrium distribution in

that Zn is vaporized. Below 1200 ^◦C the volatility becomes

negligible (Fig. 8B), and the oxides of Zn, Cu(II), and Cu(I)

appear as additional, segregated solid phases (Fig. 8A). This

provides the following likely mechanism for the gas-to-solid

transformation in the ﬂame: Immediately upon combustion

solid ZnAl₂O₄starts to form from the highly supersaturated

Al species in the gas. This leads to dendritic aggregates of

small coagulated primary aluminate particles (cf. Fig. 10,

left). As the ﬂue gas cools, the saturation points for the

pure oxides of Zn and Cu are reached and the oxides start

to condense, however in a more gradual way, controlled by

the cooling rate. The oxides presumably nucleate and grow

at the large surface of the initially formed aggregates of

ZnAl₂O₄. The ﬁnal pelletized sample thus attains a struc-

ture with a ZnAl₂O₄skeleton with alternating crystallites of

CuO and ZnO intimately mixed along its surface.

3.7. Catalyst stability

Catalyst A of Table 2 in addition to the standard activity

test has been subjected to further tests as shown in Figs. 6

and 7. The stability of the catalyst was measured over a

4-day period, during which the reactor was operated contin-

uously at the standard test conditions. The differential rate

of reaction as a function of time is shown in Fig. 6. The ﬁg-

ure also shows the effect of Ni-carbonyl poisoning of the

catalyst which gave rise to installment of the active carbon

trap. Without the trap installed, the activity slowly decreases

and the undesired methane production increases. With the

trap installed, the activity after an initial increase over 24 h

remains constant for the whole test period. The test was con-

cluded by a 4-h rise of the temperature to 290 ^◦C, which

has no effect on the activity measured subsequently at the

standard temperature. Fig. 6 also demonstrates the excellent

selectivity of the catalyst. Methane, which is the only de-

tectable by-product of the reaction, is produced in minute

amounts unless the catalyst is contaminated by Ni.

3.8. Future work

This structure combines the beneﬁcial features discussed

above for the CuO/ZnO and CuO/Al₂O₃samples and the

ZnO/Al₂O₃system. It ensures an intimate contact between

Cu and ZnO and a large and stable surface area of ZnAl₂O₄,

which binds Al, so that the formation of copper aluminate

with its poor catalytic properties is prevented. The copper

particle size after reduction is close to the BET-equivalent

diameter for samples A and D. Evidently, the ternary cata-

lyst, to a higher degree than the other catalysts, protects the

copper particles from growing during reduction from CuO

to Cu.

The theory, that Cu_xO nucleates independently of the

other species at temperatures well below the peak tempera-

tures of the ﬂame, is corroborated by the fact that the copper

dispersion is almost independent of the ﬂame conditions for

both the CuO/ZnO (samples B and E of Table 2) and the

CuO/ZnO/Al₂O₃catalyst sample (samples A and D). In

both the binary and the ternary sample, however, the ZnO

crystallite size is larger in the methane ﬂame than in the hy-

Future work on this catalyst may include an attempt to

optimize the catalytic properties of the prepared samples

by adjustment of the ﬂame synthesis conditions. This study

would also require a more careful control of the reduction

process in that the activity of the ﬁnal catalyst presumably is

very sensitive to the reduction conditions.

4. Conclusion

Flame combustion synthesis has been proven to be a ver-

satile method for producing metal-oxide catalysts with large

speciﬁc surface areas.

The speciﬁc surface area, structure, and phase compo-

sition of the product materials are signiﬁcantly affected by

the ﬂame conditions and the burner design. Thus, immedi-

ate alterations in the properties of the efﬂuent particles can

be effected by simple adjustments of the operating condi-

tions. There are many possibilities for optimizing product

J.R. Jensen et al. / Journal of Catalysis 218 (2003) 67–77

77

properties and developing materials, tailormade for speciﬁc

purposes.

of catalyst, or catalytic membranes. Both applications are

presently under investigation in our laboratory.

The premixed ﬂame, in which fuel, air, and several dif-

ferent volatile precursor species are all mixed together prior

to combustion, provides a homogeneous temperature envi-

ronment for the particle synthesis since all precursor species

traverse the high-temperature zone. The premixed ﬂame is

shown to be ideally suited for the synthesis of nanostructured

mixed oxide phases like, e.g., spinels, which are produced in

a single step, which means that no further, surface reducing

thermal treatment is necessary. The material properties ob-

tainable by combustion synthesis are particularly attractive

for catalytic materials.

It is shown in this study that ﬂame combustion synthesis

can generate a Cu/ZnO/Al₂O₃catalyst for methanol syn-

thesis with very high activity, selectivity, and stability. The

study also demonstrates how the versatility of the ﬂame reac-

tor can be used to affect fast, systematic changes in catalyst

composition and structure, which is useful in the study of

how the structure of a catalyst affects its catalytic behaviour.

Flame synthesis of catalysts may ﬁnd a more widespread

use in catalytic research due to the versatility of the method.

It is, as yet, quite untested in large-scale catalyst manufac-

ture. In the laboratory, the process presently appears rather

expensive and has a rather low production rate. However, to-

day some materials (TiO₂, carbon black, SiO₂) are already

manufactured in large quantities by industrial ﬂame synthe-

sis. An optimized scale-up of the laboratory process possibly

may lead to competitive manufacturing processes also for

catalytic materials.

Acknowledgment

This work was funded by the Danish National Research

Councils.

References

[1] S.E. Pratsinis, Prog. Energy Combust. Sci. 24 (1998) 197.

[2] M.S. Wooldridge, Prog. Energy Combust. Sci. 24 (1) (1998) 63.

[3] W.J. Stark, K. Wegner, S.E. Pratsinis, A. Baiker, J. Catal. 197 (1)

(2001) 182.

[4] T. Johannessen, S. Koutsopoulos, J. Catal. 205 (2002) 404.

[5] J.R. Jensen, T. Johannessen, S. Wedel, H. Livbjerg, J. Nanoparticle

Res. 2 (4) (2000) 363.

[6] J.B. Hansen, in: G. Ertl, H. Knözinger, J. Weitkamp (Eds.), Handbook

of Heterogeneous Catalysis, Vol. 4, Wiley–VCH, Weinheim, Germany,

1997, p. 1856, Chapt. 3.5.

[7] J.R. Jensen, Flame synthesis of composite oxides for catalytics appli-

cations, PhD thesis, Department of Chemical Engineering, Technical

University of Denmark, 2001.

[8] J.P. Hansen, J.R. Jensen, H. Livbjerg, T. Johannessen, AIChE J. 47

(2001) 2413.

[9] H.P. Klug, L.E. Alexander, X-Ray Diffraction Procedures—For Poly-

crystalline and Amorphous Materials, Wiley, New York, 1974.

[10] G.C. Chinchen, C.M. Hay, H.D. Vandervell, K.C. Waugh, J. Catal. 103

(1987) 79.

[11] J.W. Evans, M.S. Wainwright, A.J. Bridgewater, D.J. Young, Appl.

Catal. 7 (1983) 75.

[12] S. Sato, R. Takahashi, T. Sodesawa, K. Yuma, Y. Obata, J. Catal. 196

(2000) 195.

It is also possible that ﬂame combustion may ﬁnd a spe-

cial application in the manufacture of, e.g., catalysed hard-

ware, i.e., surfaces of process equipment with thin coatings

[13] K. Klier, Adv. Catal. 31 (1982) 243.

[14] M.L. Michelsen, Fluid Phase Equilib. 53 (1989) 73.

[15] A.Y. Rozovskii, G.I. Lin, Kinet. Catal. 40 (6) (1999) 773.

Article Doi

A study of Cu/ZnO/Al2O3 methanol catalysts prepared by flame combustion synthesis

DOI: 10.1016/S0021-9517(03)00047-2

Source and publish data:

Authors:

Article abstract of DOI:10.1016/S0021-9517(03)00047-2

Full text of DOI:10.1016/S0021-9517(03)00047-2

Products guided by the article

Relevant to this article

Hot Product