Вы находитесь на странице: 1из 16

NEW

Formaldehyde synthesis from methanol over silver catalysts

Abstract

The formaldehyde synthesis from methanol was investigated over a polycrystalline


silver catalyst at temperatures up to 993 K. Water was added to the feed (water ballast
process) like in the commercial BASF process. The conversion of methanol and the
selectivity to formaldehyde appeared to increase with respect to the methanol ballast
process with no added water. A long-time experiment was carried out lasting over more
than 300 days time-on-stream. While the methanol conversion does not change
significantly, there is a pronounced change in hydrogen and CO 2 selectivity. A most
noteworthy observation is that over months of operation, the width of the temperature
region where formic acid is formed increases in a linear manner. Finally, interrupting the
oxygen supply for a few hours led to a temporary deterioration of the product selectivity
after oxygen re-admittance. All observations may well be interpreted in the framework of
the commonly discussed silveroxygen chemistry with its three different oxygen
species (O, O and O).

Keywords
Methanol oxidation;
Oxidative dehydrogenation;
Formaldehyde;
Silver catalyst;
Catalyst deactivation;
Oxygen species

1. Introduction

With an ever increasing production of (2.52.7)10 7 t per year worldwide in 2000 for the
production of ureaphenolic, melamine and acetal resins, formaldehyde is one of the
worlds most important chemicals [1]. It may be synthesized by either dehydrogenation
(Eq. (1)) or partial oxidation (Eq. (2)) of methanol:

equation(1)
equation(2)

Two important routes are prominent in the industrial formaldehyde


production [2], [3], [4], [5] and [6]. The first synthesis route is performed over an electrolytic
silver catalyst at air-lean conditions, which has been commercialized at the beginning of the
20th century. The second route, industrialized since the 1950s, is by the oxidation in excess air
over a ferric molybdate catalyst. Other routes have not yet been commercialized, e.g. the
sodium catalyzed dehydrogenation of methanol to anhydrous formaldehyde [7] and the partial
oxidation of methane over silica catalysts [8]. With about 55% of the industrial production
capacity in western Europe being based on the silver-catalyzed route in 2000 [1], it plays an
important role. Its advantages are the relatively low investment cost, the high yield and the
stable production.
In commercial production, the strongly exothermic reaction is carried out under adiabatic
conditions [2]. A selection of parallel and consecutive reactions is given in Table 1. They may
inevitably lead to many byproducts, e.g. carbon dioxide, carbon monoxide, formic acid,
hydrogen, coke, etc. The combination of the reactions yields a rather complex reaction network.
Table 1.
Some possible secondary reactions in the system with the components formic acid, hydrogen,
carbon dioxide and carbon monoxide, coke, etc. [2] and [4]
Number Reaction HR,973 K (kJ/mol)
(3) CH2OCO+H2 +12
(4) 676
(5) CH2O+O2CO2+H2O 519
(6) 314
(7) CH3OHC+H2O+H2 31
(8) CO+H2C+H2O 136
(9) CO+H2OCO2+H2 35
Various versions of the conventional silver-catalyzed process are adopted commercially,
described in many different patents [9], [10], [11], [12], [13] and [14]. Two main variations are
used industrially. The first process (hereafter referred to as methanol ballast process) in which
only air and pure methanol are fed is used in numerous companies (e.g. ICI and Degussa).
Features are an incomplete conversion of methanol and its distillative recovery. BASF uses
another important process (hereafter referred to as water ballast process) feeding extra water
with the reactant mixture, thereby achieving a complete conversion of methanol [2], [9] and [15].
However, water addition is limited by the requirement of the final products strength and of the
additional water needed for tail-gas scrubbing. Normally, the commercial production is carried
out with a H2O/CH3OH molar ratio of 40/60 or 0.67 [2]. Due to its large heat capacity, water
vapor may remove a great deal of reaction heat, thereby both preventing detrimental
overheating as well as sintering of the catalyst. Moreover, because addition of water vapor helps
to burn away the coke on the catalyst surface, it is reported for the industrial production that the
life time of Ag catalyst in the water ballast process is significantly longer than that in the
methanol ballast process [5]
In the present study, both methanol and water ballast processes were investigated
comparatively in the same laboratory-scale setup. The influence of water addition on
catalyst life time, formaldehyde selectivity and generation of byproducts was studied.

Despite a century of industrial production, still no comprehensive insight into the silver-
catalyzed formaldehyde synthesis is available. It has been pointed out for the methanol
oxidation that the high yields and selectivities may seem to render any attempt to study
the reaction with the aim to improve it superfluous, but that, considering the potential
market demand of formaldehyde, even small improvements will be economically
interesting [16]. This is certainly one of the reasons that quite some research work is
devoted to this system.

One approach is the surface science approach where the interaction of reactants and
well-defined single crystal surfaces is studied under high vacuum conditions. In order to
move to industrially more relevant conditions, some studies have been carried out with
polycrystalline electrolytic silver at atmospheric pressure and temperatures up to 930 K.
From the body of literature, a mechanistic model begins to emerge that may explain a
number of experimental observations. A brief outline will be given in the next
paragraphs.

It is generally accepted that understanding well the interaction of oxygen with silver is a
prerequisite to explain the mechanism of total reaction
routes [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29] and [30]. Three
different oxygen species O, Oand O are assumed to form during the
reaction [31] and [32]. Molecular oxygen dissociates on the silver and forms the weakly
bound atomic surface oxygen species Owhich benefits especially the formation of
HCOOH and the complete oxidation of products to H2O and CO2. Deng et al. even
supposed that molecularly adsorbed oxygen acts as a precursor and transforms to
atomic oxygen in this process [33]. With similar atomic dimensions (Ag: 60 , O: 62 ),
oxygen may dissolve in the silver lattice. This O cannot participate directly in the
reaction. The strongly bound oxygen species O is formed from O, when the latter
segregates from the bulk to the surface. O exists predominantly in the uppermost layer
of silver catalyst and tends to only catalyze the dehydrogenation of methanol. The
oxygensilver chemistry also seems to influence the adsorption of methanol which has
been found to hardly occur on an oxygen-free silver surface [34]. Under reaction
conditions, the silver is constantly exposed to oxygen, part of which will penetrate into
the crystal lattice of silver and form the oxygen species of O , O and O species that are
depleted by the corresponding reactions, which shows a dynamic equilibrium [35].

The population is a strong function of parameters and pre-treatment. The three oxygen
species have, e.g. stability regions at different temperatures. Another factor is the
dynamic restructuring of the silver: SEM images reveal that the morphology of silver
catalyst changes strongly during the reaction [32], [36], [37], [38] and [39]. Millar et al.
observed that pinholes appear firstly predominantly in the vicinity of initial surface
defects and spread gradually from the region of defects to the entire silver surface [40].
This reaction-induced morphological restructuring reinforces the formation of grain
boundary defects over the catalyst, which enables more oxygen to penetrate into the
silver lattice and, in turn, to intensify the reaction [30], [31], [41], [42] and [43].

Recent studies have focused on using silver supported catalysts [44], on elucidating the
mechanism with sophisticated vacuum experiments, such as temporal analysis of
products (TAPs) [45] or on optimizing the thermal management [46]. The present
experimental work addresses three industrially relevant issues that have been little
studied in the past:

The investigation of the water ballast process with H2O/CH3OH of 02.0;



(Detrimental) formic acid formation as a function of control parameters and time;

Long-time behavior (10 months) of conversion and selectivities.
Two byproducts that were closely monitored were carbon dioxide and formic acid.
Carbon dioxide, one of the major byproducts, was monitored with time-on-stream to
gauge the catalyst activity during a long-time operation. Formic acid is a strongly
corrosive byproduct and an undesired contaminant of the final product. It is, therefore,
also strictly controlled in commercial operation (normally <0.01 wt.% in the product [2]).
Before the onset of catalyst deactivation, the formation of formic acid is restricted to only
a limited temperature range. After deactivation, normally 49 months, formic acid is
formed at all temperatures, which forces a shutdown and c atalyst replacement in industrial
production.

. Experimental

Catalytic partial oxidation of methanol was carried out over an electrolytic silver powder
(Bayer AG, 20 mesh) in a tubular fixed-bed reactor. With the help of two HPLC pumps,
methanol and water were pumped separately into the two different vaporizers (kept at
363 and 393 K, respectively) which were set up in series. Assisted by N2 as a carrier
gas, evaporation yielded partial pressures with a standard deviation of 1.5%. Oxygen
was added behind the second vaporizer and the mixture flowed through the catalyst bed
from the top of the reactor. Table 2 gives the range and standard settings of all relevant
operating parameters. In this paper, yields and selectivities are expressed with respect
to the excess component CH3OH and not with respect to oxygen

Table 2.
Standard settings and operation range in the laboratory-scale setup
Standard setting Operation range
Temperature (K) 823 473993
Residence time (s; at 823 K) 0.085 0.060.45
GHSV (h1; at 298 K) 1.5 104
WHSV (h1; at 298 K) 440
CH3OH fraction in the feed (vol.%) 17.5
O2 fraction in the feed (vol.%) 7
CH3OH/O2 molar ratio 2.5
H2O/CH3OH molar ratioa 0.67 02.0
a
In the methanol ballast process, no water was fed and the other conditions were the same as
those in the water ballast process. The space velocity was kept constant by varying the N 2 flow
accordingly.

The reactor was made of a 50 cm long Al2O3 tube and had an inner diameter of 10 mm.
The blank activity proved negligible. The reactor was filled with quartz chips ( 1 mm)
before the catalyst bed to improve the mixing and pre-heating of the feed. The 3 cm
long Ag catalyst bed was strongly diluted with quartz chips (Ag/quartz: 0.1 g/3 g) in
order to weaken the local heating effects.

Heating was provided by an electric heat coil controlled by a Eurotherm temperature


controller, which monitored the temperature with the help of a thermocouple, located
directly in the middle of catalyst bed through a capillary Al 2O3 tube. At the given
parameter settings, no isothermal operation was possible (see Section 4). The
temperatures given in this paper are those located on the centerline, at an axial position
of approximately 1 cm into the catalyst bed (T1). In order to prevent the product
formaldehyde from decomposing at the high temperature, the reaction was carried out
with a very short residence time of <0.086 s. Moreover, the product mixture was rapidly
quenched to about 413 K subsequently with a cooling jacket behind the catalyst bed.
The quench downstream caused a strong axial temperature gradient: the temperature
on the centerline 2 cm downstream of T1, just before the quench (T2), was about 42 K
higher. This strong gradient is the reason that, compared with the results of Panzer [31],
different runs may exhibit relative shifts on the temperature axis even if their qualitative
behavior is identical.

Analysis of the products was made by online gas chromatography (HP 5890 series II)
with two TCD detectors. Qualitative and quantitative separation of both high and low
boiling point compounds was possible in this setup with a mol-sieve (5A, HP) and a
polar capillary column (Poraplot-Q, HP). The separation of formaldehyde and water is
relatively difficult, but was attained in this study within a reasonable accuracy (see error
bars in Fig. 3). The carbon balance is about 1005% with an acceptable margin of error.
For the temperature variation, a heating rate of 0.16 K/min was selected. Combined with
a sampling period of 45 min and a sampling time of seconds, it may be safely assumed
that the values obtained at this extremely slow heating rate are representative of the
steady state values with a temperature resolution of 7.5 K.
3. Results
3.1. Influence of operating parameters
3.1.1. Influence of reaction temperature

Fig. 1 depicts conversions and yields as a function of temperature. A light-off


temperature was observed at about 570 K. CO2 displayed a maximal yield at the
relatively low temperature of 575 K and then dropped off with temperature. The yield of
formaldehyde increased gradually with temperature and reached a maximum at about
923 K, which corresponds well with the commercial operation temperature. The abrupt
decrease of the formaldehyde yield above 923 K was accompanied by a yield increase
of CO and H2, suggesting a gas phase decomposition of formaldehyde to CO and H 2 at
the high temperature. Formic acid appeared only in a limited temperature region
(approximately 570850 K) and could not be observed in the high temperature region
before the deactivation of catalyst (see Section 3.3).

Fig. 1.
Variation of the operating temperature in the water ballast process with 225 days time-on-stream.

3.1.2. Influence of residence time

Methanol conversion and the selectivities to formaldehyde and hydrogen were


determined at different residence times (0.060.45 s; Fig. 2). The higher the residence
time was, the more methanol was converted (open triangles). However, the longer
residence time was not beneficial for the formaldehyde formation: its selectivity
decreased apparently under the longer residence time, which may be partly due to the
fast decomposition of formaldehyde in the gas phase to H 2 and CO at high operation
temperatures. The H2 selectivity did increase with residence time, albeit not to the extent
that the formaldehyde selectivity decreased.

Fig. 2.
Variation of the residence time (823 K).

3.1.3. Influence of molar ratio of H2O/CH3OH in the feed

The influence of water vapor in the reaction gas on the formaldehyde selectivity was
estimated. Water vapor content was varied in the region of H 2O/CH3OH molar ratio of 0
2.0. The space velocity was kept constant by varying the N 2 flow accordingly. This led to
a constant CH3OH/O2 molar ratio. Each result was an average over a 15 h lasting
stationary test. Fig. 3 indicates that the conversion of methanol increased with the
H2O/CH3OH molar ratio, however, the selectivity to formaldehyde passed through a
maximum around a H2O/CH3OH molar ratio of about 0.75, which corresponds basically
well with the above-mentioned molar ratio of 0.67 in industrial formaldehyde
manufacture (indicated by the vertical dashed line). Because of the experimental error
in the formaldehyde detection, the experiment was reproduced at different feed
concentrations, supporting the conclusions reported above. It is also shown in Fig.
3 that the selectivity to CO2 decreased with the molar ratio of H2O/CH3OH. The more
water vapor was fed in the reaction gas, the less CO2 was detected
Fig. 3.
Selectivity to HCHO and CO2 vs. the molar ratio of H2O/CH3OH (823 K; CH3OH, 8.5%; O2, 3.5%)

3.2. Comparison of water and methanol ballast processes

In order to estimate the effect of water vapor on the reaction, during the long-time water
ballast study some experiments were carried out where methanol ballast conditions
were ensured. The two modes of operation were, therefore, studied at identical space
velocity in the same setup. The gradual decrease of catalyst activity (see Section 3.3)
was slow compared to the time elapsed between the two experiments carried out
subsequently and could be neglected.

In comparison with the methanol ballast process, the water ballast process displayed a
higher conversion of methanol and selectivity to formaldehyde. Fig. 4 shows the
methanol conversion rose by 10% and the formaldehyde selectivity increased by 4% at
823 K, if adding water vapor to the reaction gas. The selectivity to CO 2, the product of
total oxidation, was obviously suppressed in the water ballast process as compared to
the methanol ballast process.
Fig. 4.
Comparison between the water and methanol ballast process.

In addition, the selectivity to formic acid was also clearly smaller in the water ballast
process. Fig. 5 shows the result of two experiments carried out one immediately after the other.
The selectivity to formic acid in the water ballast process was almost 30% lower at 620 K than
that in the methanol ballast process.

Fig. 5.
Comparison of HCOOH selectivity between the water and methanol ballast process (lines indicate a
moving average over a 10 K temperature interval).
3.3. Aging behavior

In order to gauge the long-time behavior, conversions and selectivities were measured
for over 300 days time-on-stream. After a short initialization, both methanol conversion
and formaldehyde selectivity were basically constant (steady state performance). The
latter may have displayed a slight decrease ( Fig. 6). A much more pronounced change
was observed for the selectivities to CO2 and H2 (Fig. 7). S(CO2) increased gradually
over the entire experiment. S(H2) decreased from 14 to 8% during the first 100 days. It
then dropped to about 56%, where it remained basically constant for the remainder of
the experiments.

Fig. 6.
The life time of silver catalyst in the water ballast process (823 K)

Fig. 7.
Selectivity to CO2 and H2 vs. time-on-stream in the water ballast process (823 K).
Fig. 7.
Selectivity to CO2 and H2 vs. time-on-stream in the water ballast process (823 K).

Fig. 7.
Selectivity to CO2 and H2 vs. time-on-stream in the water ballast process (823 K).

Fig. 8 shows the selectivity to formic acid as a function of temperature for various times-
on-stream. No formic acid was detected during the first 100 days. After this period,
formic acid appeared, but only in a limited temperature region of 600690 K
with S(HCOOH)<0.05. However, this region grew with time, such that after 8 months
(240 days), formic acid could be observed over a wide range of temperatures (580
870 K). In order to visualize the slow extension of formic acid formation into the high
temperature region, we define an upper (Tupper) and a lower (Tlower) temperature limit such
that these two bound the region where formic acid could be detected in the outlet
with S(HCOOH)>0.01. The two temperatures are plotted in Fig. 9 versus time-on-
stream. The upper temperature increased linearly to 870 K during the first 8 months with
a correlation coefficient of R2>0.97. After this time, the shift into the higher temperature
region stopped and the upper temperature stayed constant with time. The lower
temperature limit decreased with time-on-stream, but in a less pronounced manner and
without a discontinuity. The two trends made the region of HCOOH formation broader
with time
Fig. 9.
Upper and lower detection temperature of HCOOH vs. time-on-stream

To study the nature of this deactivation, the operation was interrupted once during the
time while the upper limit still changed (180 days time-on-stream) and once when it was
constant (280 days time-on-stream). The interruption consisted of 4 weeks during which
the catalyst bed was kept under flowing N 2 at 523 K. In both cases, after the
experimental conditions had been restored, no marked discontinuities were observed
(Fig. 9).

3.4. Oxygen-free treatment

For more insight into the interaction between oxygen and silver during the partial
oxidation of methanol, the catalyst was temporarily exposed to an oxygen-free
atmosphere. Starting from steady state, the oxygen feed was interrupted for 5 h.
Methanol, water vapor and nitrogen kept flowing over the catalyst. After this, oxygen
(and hence the normal feed composition) was switched back on. The exit gas
composition was averaged over 5 h following the oxygen-free treatment. Fig. 10a shows
that, at three different temperatures (623, 773 and 823 K, respectively), the conversion
of methanol and the yield of CH2O decreased strongly after oxygen-free treatment as
compared to steady state. In addition, the yields of carbon dioxide and formic acid had
increased while the hydrogen yield was lower. The decrease of the ratio
of S(CH2O)/S(CO2) after oxygen-free treatment is depicted in Fig. 10b.

Fig. 10.
Influence of oxygen-free treatment on: (a) the conversion and the yield; (b) the ratio of S(CH2O)/S(CO2) as
compared to steady state in the water ballast process
Discussion

In order to classify the results, it should firstly be noted that laboratory-scale operation
was neither adiabatic nor isothermal. Like in [25], a pronounced temperature gradient
was measured in the catalyst bed. When discussing the results, it should be kept in
mind that the temperatures given are those measured at T1 and used for the
temperature control. Rather than the absolute values of T1, the trend should be
considered. The excellent isothermicity of catalytic wall reactors as reported by
Redlingshefer et al. [47] has also been used for the methanol oxidation to
formaldehyde [31]. However, more development is required for the system prior to
publication.

A second caveat is that even if we will discuss the reported findings in the framework of
the state-of-the-art of silveroxygen chemistry, it should be obvious that due to the
absence of spectroscopic evidence, this paper cannot contribute much to the body of
evidence. However, the phenomenological approach taken here may indeed give
suggestions for new experiments. As a matter of fact, transferring promising results with
fiber-optic in situ spectroscopy [48] to this reaction system is currently underway.
Raman spectroscopy seems to be an appropriate method [49].

The comparison of water and methanol ballast processes indicates that an appropriate
water vapor addition plays a great role in the reaction. Bao et al. [35] found the oxygen-
induced restructuring process can be greatly enhanced by adding a small amount of
water vapor to O2, which makes the formation of oxygen species more favorable. An
evident increase of CH3OH conversion shown in Fig. 4 supports this hypothesis. As well
as reinforcing the conversion of methanol, water vapor can suppress effectively the total
oxidation and reduce consequently the generation of CO 2 shown in Fig. 3 ; Fig. 4.
Various tentative explanations can be given: water vapor may block specific surface
sites at which total oxidation takes place; it may help burn away the coke on the catalyst
surface to expose more active sites.

The study shows an initialization phase to ignite the silver catalyst, which was also
observed by Nagy et al. [32]. This is necessary not only in the methanol ballast process
but also in the water ballast process, which was also observed in the industrial plants
from some hours to several days. The in situ Raman studies of Millar et al.
demonstrated that subsurface OH(a) species were formed very rapidly in a well-
defected Ag sample, whereas extremely slowly over the smooth surface of a fresh Ag
sample [50]. It is assumed that the initial activation of silver catalysts is due to the
diffusion of OH(a) species and the consequent formation of defected sites.

A notable trend is that S(CO2) increased with time whereas S(H2) decreased (Fig. 7).
This may point to an increase of the number of weakly bound atomic O on the silver
surface with time. The degree of total oxidation is commonly attributed to the surface
adsorbed atomic O. While this is particularly true in the low temperature region, after
extended times it may also hold in the high temperature region, where O plays the
prominent role in the dehydrogenation of methanol.

HCHO is normally unstable at high temperatures, being easily decomposed to H 2 and


CO or, in the presence of oxygen, even totally oxidized to CO 2 and H2O. It is also
assumed that HCHO could be further oxidized to HCOOH in the presence of the weakly
bound surface atomic O[31]:

As mentioned above, the influence of O became more and more notable with time,
HCOOH was consequently formed in a wider temperature region and even at high
temperatures after several months of operation (Fig. 8). Because the O species plays
an important role mainly in the lower temperature region, it is reasonable that the
amount of HCOOH increased particularly at low temperatures around 600 K with time-
on-stream. The reason that no formic acid was detected above 870 K is that this is the
decomposition temperature and does not reflect a discontinuity in the deactivation
process.

The experimental results after oxygen-free treatments confirmed this hypothesis well on
another aspect. The dynamic equilibrium of the O-development is disturbed during
treatment with N2 gas only. Since oxygen species could not evolve well during nitrogen
treatment, the dominant oxygen species in the subsequent oxidative reaction would
initially be the weakly bound atomic O, which shows high oxidative power and helps
promoting the formation of carbon dioxide and formic acid. With the help of in situ
Raman spectroscopy, Pettinger et al. discerned a gradual re-development of
O formation over the Ag catalyst when it was exposed again to O 2 after a 30 min
treatment with pure N2[51]. As a result, treatment under oxygen-free conditions followed
by re-admission of oxygen should temporarily lead to an increase of O -catalyzed
formation of carbon dioxide and formic acid as well as a decrease of O -catalyzed
hydrogen formation as compared to the steady state. This is indeed found (Fig. 10a).
Consequently, the oxygen-free treatment should lead to an Ag catalyst on which the
steady state distribution of oxygen species does not fully develop immediately. Hence, a
temporary decrease of methanol conversion and CH2O yield would be expected after
oxygen-free treatment. This is also supported well by Fig. 10a. A decrease of the ratio
of S(CH2O)/S(CO2) appeared after the treatment under oxygen-free condition, especially
with almost 50% at low temperature where O is normally developed well and plays a
dominant role (shown in Fig. 10b).

Conclusions

The silver-catalyzed synthesis of formaldehyde from methanol was studied at near


industrial condition. The reaction proves sensitive to some operating parameters, e.g.
reaction temperature, residence time and molar ratio of H 2O/CH3OH.

Methanol can be converted more completely and the selectivity to formaldehyde


increased if water is added to the reaction gas. In addition, an appropriate amount of
water vapor decreases the total oxidation in the reaction system and suppresses the
generation of formic acid. Besides removing a great deal of heat to avoid the sintering of
the catalyst, water vapor may also block total oxidation sites, or burn off coke, to
improve the silver activity significantly.

The amount of carbon dioxide, the main byproduct, grew with time-on-stream, revealing
a gradual enhancement of oxidation in the reaction system. Another major byproduct,
formic acid, was first formed in a narrow temperature region. However, the region grew
wider for more than 10 months.

Вам также может понравиться