Professional Documents
Culture Documents
A R T I C L E I N F O A B S T R A C T
Article history: In economy nowadays, methanol is already a key compound widely employed as building block for pro-
Received 26 October 2015 ducing intermediates or synthetic hydrocarbons, solvent, energy storage medium, and fuel. This status
Accepted 5 June 2016 is expected to last in the near future or even improve to the point of making this compound a central
Available online 5 July 2016
participant in the worldwide economic landscape. For these reasons, every improvement to its produc-
tion process, in terms of energy savings, optimization, etc., has potential to promote relevant economic
Keywords:
benefits. Methanol production comprises three main steps: preparation of syngas, methanol synthesis
Modeling
and downstream separation. This paper aims at reviewing technologies and procedures for modeling and
Optimization
Kinetics optimizing the second aforementioned phase: the synthesis reactor. Specifically, we focus on packed-
Transport phenomena bed units, which represent the most widespread technology. In the manuscript, we are going to describe
Catalyst deactivation and compare both steady-state and dynamic reactor models as well as analyze typical assumptions and
implementation schemes. The kinetics of methanol synthesis is also reported in detail due to a long debate,
present in the literature, concerning the real carbon source for methanol, the nature of the active sites
and the effect of their morphology and oxidation state.
© 2016 Elsevier Ltd. All rights reserved.
Contents
1. Introduction ........................................................................................................................................................................................................................................................... 72
1.1. History ........................................................................................................................................................................................................................................................ 72
1.2. Impact of methanol on energy and environment ....................................................................................................................................................................... 73
1.3. Methanol production ............................................................................................................................................................................................................................ 73
1.3.1. Traditional chemistry ........................................................................................................................................................................................................... 73
1.3.2. Some alternatives to methanol production other than syngas ............................................................................................................................. 75
1.3.3. Green chemistry .................................................................................................................................................................................................................... 76
2. Existing technologies ......................................................................................................................................................................................................................................... 78
2.1. Gas phase technologies ........................................................................................................................................................................................................................ 78
2.1.1. Adiabatic reactors .................................................................................................................................................................................................................. 78
2.1.2. Isothermal reactors ............................................................................................................................................................................................................... 79
2.1.3. Gas phase fluidized bed converter .................................................................................................................................................................................. 82
2.2. Liquid phase technologies .................................................................................................................................................................................................................. 82
2.3. Membrane reactors ............................................................................................................................................................................................................................... 83
2.4. One-step technologies ......................................................................................................................................................................................................................... 84
3. The catalyst ............................................................................................................................................................................................................................................................ 84
4. Reactor modeling ................................................................................................................................................................................................................................................. 84
4.1. Kinetics ...................................................................................................................................................................................................................................................... 84
4.2. Transport phenomena around and inside the catalyst ............................................................................................................................................................. 88
4.3. Catalyst deactivation ............................................................................................................................................................................................................................ 88
* Corresponding author. Politecnico di Milano, Dipartimento di Chimica, Materiali e Ingegneria Chimica “Giulio Natta”, Centre for Sustainable Process Engineering Research
(SuPER), Piazza Leonardo da Vinci, 32, 20133 Milano, Italy. Fax: +39 (0)2 2399 3280.
E-mail address: giulia.bozzano@polimi.it (G. Bozzano).
http://dx.doi.org/10.1016/j.pecs.2016.06.001
0360-1285/© 2016 Elsevier Ltd. All rights reserved.
72 G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105
1. Introduction manufacturing. Initially, it was used for lighting, cooking and heating
purposes but it was quickly replaced by more economical fuels. On
Large amounts of energy must be consumed to preserve human- the other hand, methanol was increasingly required by the chem-
ity’s actual living standards. Oil, natural gas and coal, still constitute our ical industry. In 1905, Sabatier proposed the first synthetic pathway
principal energy sources and offer the raw materials for producing a for producing methanol, which implied reacting CO and H2. Based
large variety of derivatives. Unfortunately, these resources are limited on this earlier discovery, the Badische Anilin und Soda Fabrik (BASF)
and not renewable on the human time scale, thus we will probably run patented a syngas-based methanol production process, where syngas
out of coal, oil and natural gas within 3 centuries [1]. For this reason, was supplied via coal gasification. This process required a zinc/
all feasible alternatives must be investigated to seek feasible and long- chromium oxide catalyst as well as high temperature and pressure
term solutions. Besides the problem of power generation, an important (300–400 °C and 250–350 atm). It was first deployed in Leuna,
question is/will be how to store and efficiently use energy. The utili- Germany, in 1923. The following technology developments aimed
zation of hydrogen is an option that is being discussed worldwide. at reducing the pressure and temperature levels in order to improve
However, its storage and transportation raises serious (safety) prob- process economics. Thanks to the invention of the steam reform-
lems, and no infrastructures are available. A feasible alternative could ing of methane, which allowed producing a purer syngas, a more
be liquid methanol. Methanol can serve as convenient energy storage active Cu/ZnO catalyst could be employed, thus decreasing the
medium, easily transportable fuel, solvent and building block for pro- process temperature and pressure to about 300 °C and 100 atm. This
ducing intermediates and synthetic hydrocarbons, also including significant improvement was proposed by ICI (Imperial Chemical
polymers and single-cell proteins. Therefore, it might be a key com- Industries) in 1966. Few years later, Lurgi developed a process with
pound in the global economy of the future. even lower operating pressure and temperature (230–250 °C and
The methanol economy [2], based on green-methanol synthe- 40–50 atm). Subsequent developments further improved the process
sis pathways, has been proposed in contrast to the hydrogen such that, nowadays, the production of methanol from carbon dioxide
economy, which requires a deep change in energy storage and trans- can be considered a mature technology. In fact, the actual metha-
portation means. Methanol has an octane number of 113 and its nol selectivity is over 99.8% with an energy efficiency of around 75%.
energy density is about half of that of gasoline (by volume). The blend Most of the current research efforts are now targeted to finding new
of 10%/90% methanol/gasoline can lead to an octane up to 130. Pure ways to synthetize methanol using both diverse origin carbon
methanol engines can reach efficiency close to 43% and maintain dioxide/hydrogen feeds and different chemistries, e.g. the direct ox-
it above 40% in a wide speed and load range [1]. Methanol can favor idation of methane.
the transition from fossil fuels to renewable energies. Later on, we Nowadays, methanol is used as primary feedstock for a large
will briefly comment on green processes for methanol production. variety of chemicals. Among the most important, we can mention
Other than the traditional production pathway via synthesis gas, formaldehyde (it consumes about 70% of the methanol produced
methanol could be manufactured in different and new ways. The worldwide), methyl-tert-butyl ether (MTBE, 20%), acetic acid and
carbon source may be natural gas (in this case, methanol could be dimethyl ether. In addition, a variety of intermediates employed in
directly produced at the gas well by direct oxidative transforma- manufacturing many products of our daily life derive from meth-
tion) or CO2, which can be recovered from industrial sites and, anol: paints, resins, silicones, adhesives, antifreeze, plastics, and so
eventually, from the atmosphere. This second production pathway on. Methanol is also used as transportation fuel in addition to gas-
would allow mitigating global warming due to the increasing pres- oline, and in the future, it will certainly play an increasing role in
ence of greenhouse gases in the atmosphere. All energy sources can such field. In fact, the emissions (hydrocarbons, NOx, SOx and par-
be exploited for methanol production, thus the “Methanol Economy” ticulate) associated with it are reduced. However, before being able
offers a feasible means of using and storing all sources of energy to use pure methanol as fuel, we still need to solve issues like metal
(renewable, atomic, etc.). At this point, it is clear how important meth- corrosion (especially, aluminum, zinc and magnesium can be at-
anol may be for our future economy. Since improvements in methanol tacked) and/or cold engine start, which is related to the absence of
synthesis process are still feasible, there is great research interest highly volatile compounds. In any case, adding a certain percent-
in this field. New production technologies have been proposed after age of methanol to gasoline improves the performance of internal
the first process start-up [3], which are principally aimed at reduc- combustion engines (ICE), thus mixtures of methanol and gaso-
ing investment costs (relevant for methanol production). The line are already being marketed. In China, starting from 2009,
modeling of methanol manufacturing process, and particularly the national fuel blending standards of M85 (85% methanol, 15% gas-
reactor model, allows improving energy savings and optimizing oline) and M100 (100% methanol) went into effect, and M15 standard
the production itself. Therefore, this review focuses on modeling is in adoption. In Shanxi province alone, there are more than 1000
strategies and optimization techniques for the methanol synthesis petrol stations, which are equipped to sell M15, and other 40 M85-
reactor. M100 refueling points are present. The Shanxi government plans
to expand rapidly the number of vehicles using methanol and the
1.1. History number of petrol stations. Therefore, China appears to play a leading
role in the methanol market and, consequently, the global metha-
Methanol was first isolated by Robert Boyle in 1661 via wood nol demand is growing rapidly as shown in Fig. 1.
distillation while its chemical composition was first discovered by The development of fuel cell vehicles (FCV) will certainly involve
Dumas and Peligot in 1834. At that time, production volumes the use of methanol as hydrogen carrier. Direct methanol fuel cells
were still small, e.g. 10–20 L per ton of wood treated for charcoal (DMFCs) are under development for a number of possible
G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105 73
60
• production of synthesis gas (syngas)
• conversion of the synthesis gas into methanol
• distillation of the reactor effluent
40
Different technologies are under investigation but cannot be
implemented at the industrial scale yet due to unsatisfactory
20 efficiency.
Currently syngas (a mixture of H2, CO and CO2) is mainly produced
by reforming of natural gas. However, it can be potentially synthesized
0 by reforming or partial oxidation of different carbon-based mate-
rials, e.g. petroleum, coal, petroleum coke, heavy oil, asphalt or biogas
2000 2005 2010
[6,7]. Economic considerations typically suggest which raw mate-
Fig. 1. Global methanol industry demand forecast [4]. rials are more suitable for use, even though long-term availability,
political considerations, energy consumption and environmental
applications including low power DMFCs for portable devices, such aspects play an important role in the final decision as well.
as smartphones and laptops, DMFCs for automotive systems and high The composition of syngas is usually characterized by the stoi-
power DMFCs for large-scale electricity production. However, con- chiometric number S given by the ratio of two quantities: the
siderable technological improvements are still necessary in order difference between the moles of hydrogen and CO2 and the sum-
to achieve higher efficiency for large DMFCs. mation of the moles of CO2 and CO:
Fig. 2. Worldwide methanol production processes (Source: ICI, now Johnson Matthey C + 1 2O2 ↔ CO (7)
[8]).
C + H 2O ↔ CO + H 2 (8)
is carried out at about 200–300 °C because the catalyst is active only In the homogeneous gas-phase oxidation process, methane and
inside this temperature range. Thermodynamic equilibrium limi- oxygen react at high pressure (30 to 200 atm) and temperature (200
tations and catalyst deactivation seriously affect methanol production. to 500 °C). It has been shown that 70–80% selectivity can be achieved
In particular, the decay in the catalyst activity takes to unavoid- with 8–10% conversion in optimized conditions and by using cold
able unsteady process operation. flames at 450 °C, 65 atm and less that 5% O2 content. Other studies
A good catalyst should remain active for several years (up to 4). suggest a selectivity to methanol of 30–40% at 450–500 °C and 30–
The deactivation of Cu/Zn/Al2O3-based catalysts is usually due to 60 atm, with a conversion of 5–10%. At high pressures, radical
chemical poisoning and thermal sintering. In general, any of sulfur reactions are predominant, thus the effect of the catalyst is limited.
compounds, chlorine and heavy metals represents a catalyst poison. Most of the methanol produced is rapidly oxidized or decom-
However, due to preemptive feed treatment and purification, sin- posed to formaldehyde, carbon oxides and formic acid. If the resulting
tering is often the only source of loss of activity. Sintering is a solid- mixture is post processed with proper technologies, the methanol
state transformation, which takes place at high temperatures and content can be increased [17]. This post-processing can be per-
is promoted by water vapor. This said, methanol converters are de- formed in different ways. However, it is highly desirable to produce
signed based on the estimated average catalyst activity and should methanol directly from methane, thus avoiding any additional steps.
be able to operate at their maximum capacity even at the so- In order to achieve higher selectivity, lower temperatures are pref-
called “end-of-run” condition. erable (less than 250 °C). Current catalysts are not sufficiently active
Improvements to the process of methanol synthesis can derive at low temperatures and new types of catalysts are therefore nec-
from the invention of new, more active and more resistant cata- essary. This problem has suggested investigating liquid phase reactions,
lysts as well as novel reactor designs. which can be conducted at milder temperature. Such reactions involve
the use of superacids, whose acidity is million times stronger than
concentrated sulfuric acid. At ambient temperature and in liquid phase,
1.3.2. Some alternatives to methanol production other than syngas methane can react with H2O2 in presence of superacids and produce
The first process we want to describe is that for producing meth- methanol with high selectivity. The reaction mechanism implies in-
anol via methyl formate. This route has been proposed a century serting a protonated hydrogen peroxide (H3O2+) molecule into the C—H
ago by Christiansen [15]: bond of methane. In such acidic environment, methanol is proto-
nated (CH3OH2+) and protected from further oxidation. However, H2O2
CH 3OH + CO ↔ HCOOCH 3 (15) is not suitable for use in large-scale production facilities and superacids
are very expensive. Therefore, the search for cheaper peroxides and
HCOOCH 3 + 2H 2 ↔ 2CH 3OH (16) superacids is still ongoing [1].
Methanol can be also synthesized in a two-step process involv-
CO + 2H 2 ↔ CH 3OH (17) ing methyl-halides as intermediates (the halogen behaves like a
catalyst and is not consumed in the reaction process). Methyl chlo-
It consists of two reaction steps. The first is the methanol
ride or bromide (CH3Cl, CH3Br), catalytically produced from methane,
carbonylation reaction, which occurs in liquid phase and requires
can be hydrolyzed to methanol and HCl or HBr, which can be
homogeneous catalysis that is usually provided by sodium or po-
reoxidized [18]:
tassium methoxide. The subsequent reaction, which involves
hydrogen, can occur in both liquid and gas phases and requires CH 4 ⎯X⎯
2
→ HX + CH 3 X ⎯H⎯⎯
2O
→ CH 3OH + HX (22)
copper-based catalysts. Patents (from Mitsui Petrochemicals,
Brookhaven National Laboratories and Shell) claim that their tech- 2HX ⎯1⎯⎯
2O2
→ X 2 + H 2O (23)
nology allows producing methanol from syngas at 80–120 °C and
10–50 atm instead of 200–300 °C and 50–100 atm. The main problem X = Cl, Br (24)
with this pathway is the presence of water and CO2 in the syngas,
which react with sodium methoxide and produce undesired Recycling the hydrogen halides generated in the first reaction
byproducts. Therefore, further separation steps are necessary. Some is essential for the economics of the process. The reoxidation of hy-
catalysts have been found, which are more resistant to these poisons drogen chloride is possible but remains technologically difficult,
[16], but further improvements are still needed for the technolo- especially at industrial scale. On the contrary, air readily oxidizes
gy to be mature for industrial application. hydrogen bromide to bromine.
Extensive research is also being conducted in manufacturing The chlorination of paraffins was discovered by Dumas in 1840
methanol directly from selective oxidation of methane without the and is a very old and well-known substitution reaction that is widely
need for syngas. The reaction path is reported below: adopted in industry. It comprises several thermally or photochemi-
cally initiated free radical reactions. The main drawback of this route
CH 4 + 1 2O2 → CH 3OH (18) is its low selectivity:
CH 4 + O2 → CH 2O + H 2O (19) CH 4 ⎯X⎯
2
→ CH 3 X ⎯X⎯
2
→ CH 2 X 2 ⎯X⎯
2
→ CHX 3 ⎯X⎯
2
→ CX 4 (25)
Coal
Chemicals, synthetic
hydrocarbons and
Heavy oil their products
Tar sands Existing technologies
Shale oil Technologies to be developed
Fig. 3. Methanol from fossil fuel resources (re-arranged from Olah et al. [1]).
Finally, Periana investigated the use of Iodine in 2002 [19] and partial oxidation of biomass is attractive but presents technical prob-
this pathway appears to be of interest even though no apparent pro- lems, so generally two steps are necessary. In the first step, biomass
gress has been reported so far. Therefore, even if this route seems is heated up to 400–600 °C with low oxygen content in order to avoid
to be promising, its yield and selectivity make it noncompetitive complete combustion. Pyrolysis mainly occurs and the gases pro-
with traditional syngas-based strategies for methanol production. duced in the process are essentially CO, H2, CH4, CO2, H2O and volatile
A summary of the current technologies for producing metha- organics. A solid residue, called charcoal, is generated and ac-
nol and those still under development is shown in Fig. 3: counts for about 10–25% of the original fuel mass. In the second
step, charcoal and O2 react at 1300–1500 °C producing CO. The re-
sulting syngas still requires some purification steps before being used
1.3.3. Green chemistry
in methanol production facilities but the concentration of sulfur and
Green chemistry aims at synthesizing methanol entirely from
heavy metals is lower than that usually observed in the syngas origi-
renewable resources. A possible pathway is provided below:
nating from coal. Methanol production plants are typically large-
CO2 + 3H 2 ↔ CH 3OH + H 2O (27) scale due to economic reasons, thus approximately 1.5 Mt/y of biomass
should be needed to produce methanol at a rate of 2500 t/d (most
Hydrogen may originate from water dissociation by electroly- of the existing plants insure such production capacity). Therefore,
sis, where the source of electricity should be renewable as well (wind, biomass should be collected over vast areas of land and trans-
biomass, solar, tides, waves, etc.). Other H2 sources can be biomass ported over long distances. Consequently, it has been proposed to
pyrolysis or steam/oxygen gasification processes and the reform- convert it into an easy-to-transport liquid by means of fast pyrol-
ing of biomass-derived products (for instance ethanol, methane, etc.). ysis. Fast pyrolysis takes place at 400–600 °C with very low residence
Other routes are still under investigation. One of them implies pro- times, about 2 s, as to obtain maximum yield for liquid tar (70–
ducing biological H2 by means of microorganisms activated by 80%), called “biocrude”, and minimum amount of the other gaseous
sunlight while another is a water splitting process assisted by photo- and solid products, which can be partially used to drive the process.
semiconductor catalysts [20]. CO2 could be collected from industrial In fact, the combination of temperature and residence time regu-
waste or even from the atmosphere. In the long term, our depen- lates the distribution of the pyrolysis products among gaseous, liquid
dence on fossil fuels and on oil-producing countries would probably and solid substances. Higher temperature and longer residence times
cease to exist. However, for this to happen, methanol should be pro- increase biomass conversion to gas while lower temperatures and
duced in large amount and this target is difficult to achieve only longer vapor residence times favor the production of charcoal. Fast
by means of renewable resources. pyrolysis requires very high heating rates, such that the biomass
Waste biomass can be a potential feedstock for producing meth- has to be ground down to particles smaller than 3 mm. Short vapor
anol. If unused, it accumulates and can lead to socio-economic residence times minimize secondary reactions and the rapid removal
problems. However, we can exploit it as renewable resource for pro- of char minimizes the cracking of vapors. Vapors must then be cooled
ducing energy and as raw material for synthesizing chemicals, e.g. down rapidly to recover the bio-oil. Part of the char could also be
methanol. Biomass is every material produced by any kind of life ground and added to the biocrude, thus forming a slurry.
forms, vegetal or animal. Therefore, it includes wood, wood waste, All biomass sources can be gasified toward methanol produc-
agricultural crops, municipal solid waste, animal waste, aquatic plants tion but low-moisture materials are the most suitable for such purpose
and algae. To some extent, it is the same as going back to the origins and provide efficiencies up to 55%. Several demonstration projects
and producing methanol from wood, but more efficiently. Most of are under investigation around the world [22]. In addition, black
the technologies for methanol production that use biomass as feed- liquor originating from paper manufacturing is an interesting feed
stock are similar to those adopted when fossil fuels are exploited for gasification and methanol production. This holds true mainly in
as raw materials. Biomass is usually dried and ground to yield par- small countries, like Finland and Sweden, where paper mills are wide-
ticles of small and uniform size. Moisture should not overcome 15– spread. In fact, these countries could satisfy up to 50% of their motor
20%. The gasification of biomass generates a gas mixture containing fuel demand this way. Fig. 4 shows how to use biomass for meth-
H2, CO, CO2 and H2O. During the gasification process, oxygen burns anol production while preventing the CO2 concentration in the
a part of the biomass to generate heat [21]. The combustion gases atmosphere from increasing, thus mitigating greenhouse effects.
(CO2 and water) react with the remaining part of the biomass leading In the future, the increase in biomethanol demand will require
to CO and H2. No external heating source is necessary. The “one step” larger and larger quantities of biomass, thus large-scale cultivation
G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105 77
Fig. 4. The carbon cycle of biomethanol (rearranged from Olah et al. [1]).
of energy-devoted plants will be necessary. This is a feasible sce- purification steps. The CO2 can partially react with the excess hy-
nario for all countries with low population density like United States, drogen coming from methane steam reforming in order to produce
Australia, or Brazil, where large part of the territory could be used methanol. Another source for biogas is landfill, which is widely
to grow energy-devoted plants. Unless surplus land is available, adopted by municipalities. However, the amount of methanol coming
energy-devoted crop cultivation would compete with food and from this source is limited. The third generation of biomass com-
fiber production. For instance, limited to the European area, biomass prises algae and their derivatives and seems very appealing for a
is expected to be able to cover only up to 10–15% of the energy variety of reasons: algae are not food-competing and can be pro-
demand. Aspects like nutrients for plant growth and rotation of cul- duced much faster than other land-based cultures such as corn and
tivations must also be accounted for. Moreover, proper crops must soy; finally, they can be harvested continuously upon production.
be identified, which require minimum amount of fertilizers to grow. Aquatic plants and organisms show considerable potential because
This is essential to avoid increasing N2O emissions significantly (N2O water covers more than 70% of the Earth’s surface. Once collected,
is 296 times worse than CO2 as a greenhouse gas). Biotechnology they can be processed to form methane by anaerobic digestion and
and genetic modifications will boost the yield in crop cultivation. then methanol. Algae consists both of macroalgae, up to 60 m length,
Another option is to produce methanol from biological pro- and of microscopic organisms named microalgae. Further re-
cesses using CO2 waste as raw material. Certain bacteria, called search is expected to provide additional knowledge about such
methanotrophs, survive by extracting carbon and energy from aquatic organisms and to find a way to optimize the conversion of
methane. They selectively convert methane to methanol by using solar energy into biomass by means of algae. At present, the pro-
oxygen and an enzyme, called methane monooxygenase (MMO), duction of energy from algae is too expensive and not competitive
which operates at ambient temperature and pressure in aqueous with respect to fossil fuels or other type of biomass.
solution. Two types of MMO have been identified: the first con- The most promising approach to converting CO2 to hydrocar-
tains copper and is difficult to isolate and characterize while the bons is the catalytic or electrochemical hydrogenation to methanol.
second includes a pair of iron atoms and seems more appealing. Re- Efficient catalysts are available and are very similar to those used
search has been conducted to find a way of designing simple catalytic to produce methanol from syngas. Many methods for producing
systems that mime the MMO behavior. However, their practical ap- methanol from CO2 are known [24,25]. The factor that limits the
plication to the production of methanol has proven to be difficult. large-scale application of such process is the availability of CO2 and
In fact, MMO requires oxygen as oxidant and the presence of a re- H2. While carbon dioxide is widely available in exhaust sources (flue
ducing agent, known as NADH. In model experiments, H2O2 is used gases of power plants using fossil fuels, industrial plants, and so on)
while, in practical applications, hydrogen should be used as reduc- and in the atmosphere, hydrogen is currently generated from fossil
ing agent, thus making the process less economically attractive. sources, thus its availability is limited. A possible option is to de-
Methanol can be also produced from biogas. Most mammals compose methane and produce hydrogen and carbon. Then, we could
produce a flammable gas termed “biogas” during food digestion. synthesize methanol from recycled CO2. However, the potential
Biogas originates from the anaerobic decomposition of organic ma- success of this route strongly depends on the availability of natural
terial performed by anaerobic bacteria. It is a waste product for those gas. The solid carbon formed as byproduct could be used for dif-
microorganisms and is composed mainly of CH4 (55–75% depend- ferent applications, e.g. as pigment for ink and paints, in tires, as
ing on composition and efficiency) and CO2 but includes traces of filler for road construction, and so on. Combustion is obviously
other elements, like H2S. The oil crisis in 1970s renewed the inter- excluded in order to avoid further CO2 production. Thermal decom-
est in biogas production, where the feedstock is usually sewage position occurs when methane is heated up to temperatures above
sludge, municipal organic wastewater and animal dung. The envi- 800 °C in absence of oxygen. This process is at its first stages and
ronmental impact due to the disposal of such waste can be reduced significant research is still necessary to get to a mature and effi-
by adopting anaerobic digestion [23]. In large-scale operations, the cient technology for commercial applications. It could be a promising
feedstock can be fed to the digester in continuous fashion: the gas alternative to produce methanol at zero CO2 emissions. This tech-
leaves the unit from a gas outlet on top and the sludge, which is nology would allow using natural gas resources, including
recovered, can be used as soil fertilizer or conditioner. At present, unconventional ones such as methane hydrates. Water electroly-
biogas is mainly used for producing heat or electricity but could be sis is also an alternative hydrogen source for methanol production.
directly used for the manufacturing of methanol after proper The required electricity can originate from atomic, solar, hydro, wind,
78 G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105
2. Existing technologies
Fig. 5. (a) Adiabatic reactor with direct cooling; (b) adiabatic reactor with indirect heat exchange; (c) reactor with external cooling.
G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105 79
Gas outlet
Gas inlet
Dome
Perforated spherical
segments
Compensator
Catalyst
Gas inlet
Gas inlet
The first commercial scale plant (520 t/d) using the Mitsubishi con-
verter is located in Niigata (Japan) and has demonstrated to be
successful. A new plant, associated with higher production capacity
(2500 t/d) is under construction in Saudi Arabia. Projects for facilities
with capacities of 5000–10,000 t/d are currently under investigation.
Casale SA developed an Isothermal Methanol Converter (IMC)
in which the heat exchange surface is made of plates surrounded
by the catalyst (Fig. 13). A cooling fluid, which can be fresh syngas,
water or some other liquid, flows inside the plates. The size of the
converter has potentially no limit and its construction is simple
because it implies manufacturing a conventional vessel including
the catalyst bed and the plates. A tight control of the temperature
profile inside the mass of catalyst is achieved. This allows the con-
verter to track the optimal reaction rate locus. The reactor can be
designed with a mixed axial–radial flow configuration. The size and
number of pieces of equipment and pipes are minimized and the
energy consumption and size of the front end are optimized as well.
Handling the loading/unloading of the catalyst is easy and thanks
to the good temperature control, the average life of the catalyst lasts
longer. The catalyst temperature is controlled by changing the satu-
Fig. 12. Mitsubishi Superconverter [36].
rated steam pressure according to the process requirements. The
heat of reaction is entirely recovered to produce steam. In the axial–
radial configuration, the gas only crosses one catalyst bed and
The gas is fed at the bottom of the vessel through flexible tubes experiences a pressure drop smaller than one bar. The recycle ratio
and flows upward inside the inner shell of those tubes. Once the can be adjusted to minimize syngas consumption independently of
end of such tubes is reached, it turns downward and flows into the the level of activity of the catalyst.
annular space where the catalyst is loaded. The cooling water flows Davy Process Technology is currently licensing different and pe-
outside the tubes. The mixture of methanol and unreacted syngas culiar technologies for methanol synthesis (to some extent, the Casale
comes out of the bottom of the vessel. Group is following the same business strategy). An Axial Steam-
The temperature profile in the catalyst bed decreases from the Raising Converter (A-SRC) employs a simple shell-and-tube design
inlet section to the outlet section because of the heat exchange with with axial counter-current flow. Compressed syngas enters the A-SRC
the fresh gas. This solution insures high conversion per pass (about and moves downward inside a number of catalyst-filled tubes. Water
14% methanol is produced) because it allows maximizing the reaction from the steam drum enters the base of the vessel, vaporizes upon
rate. This reactor configuration also allows safe operation and pos- absorbing the heat of reaction and keeps the temperature under control
sesses high mechanical stability [37]. (Fig. 14a). An alternative configuration is the radial steam-raising
(R-SRC) converter (Fig. 14b), where syngas moves radially across a by designing larger-scale facilities for methanol production thanks
fixed catalyst bed on the shell-side of a reactor. As the syngas exo- to novel technologies. A fluidized bed converter was developed
thermically reacts to form methanol, tube-side water absorbs the heat where 50–60 μm catalyst particles are fluidized by the fresh gas fed
of reaction and generates to steam. JM Davy’s R-SRCs are particular- at the bottom. The temperature is controlled by using cooling pipes
ly useful for large-capacity methanol production processes. Finally, that allow to recover the heat of reaction at high pressure. The pos-
the Tube-Cooled Converter (TCC) uses its feed stream to cool down sibility of easily replacing the catalyst in operation allows constant
the catalyst bed in which the methanol synthesis reaction takes place and stable operating conditions.
(Fig. 14c). The syngas moves down through the catalyst bed, thus trans-
ferring heat to the fresh feed flowing up inside the tubes. 2.2. Liquid phase technologies
2.1.3. Gas phase fluidized bed converter Liquid phase processes for producing methanol might repre-
In 1993 the New Energy and Industrial Technology Develop- sent the future of methanol industry because they allow efficiently
ment Organization (NEDO) and the Petroleum Endowment Center removing the heat of reaction and enforcing tight temperature
(PEC) in Japan started a project aimed at lowering production costs control. Methanol synthesis technologies using three phase
G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105 83
(gas–solid–liquid) systems have also been proposed. In particular, process, operated in liquid phase, claimed to be able to achieve con-
notice that gas-phase fixed-bed reactors (GPFBR) cannot be the versions over 90%, thus avoiding to recycle the unreacted synthesis
system of choice when the feedstock is coal-derived synthesis gas gas. AMOCO [41], since 1995, tried to develop a process operating
because its low H/C ratio promotes coking phenomena [39]. To over- at low pressure and temperature. In 1998, a mini-pilot unit (50 ml
come the problem, trickle bed reactors (TBR) and slurry converters pressured vessel) was built at AMOCO Naperville R&D Center but
have been proposed. In particular, TBRs combine the advantages of problems were encountered. In particular, the system provided in-
both slurry and fixed bed reactors and appear to be more effi- sufficient recovery of low-quality heat but also required additional
cient. In TBRs the catalyst is still laid out in a fixed bed fashion but capital costs for the catalyst containment and the size of the reactor,
a co-current gas–liquid stream flows though the bed. This type of which seemed to be larger compared to fixed bed units due to the
reactor insures higher heat removal efficiency than fixed bed re- lower achievable reaction rate. Nowadays no industrial application
actors and direct use of coal-derived synthesis gas is possible. of this technology is known.
Moreover, high gas space velocity can be reached. In slurry and TBR Catalytica Advanced Technologies, USA, declared to have found
reactors, methanol is absorbed by the liquid phase and removed from a catalyst capable of converting methane directly into methanol [42].
the reaction system, thus allowing higher conversions due to the They proposed a liquid phase oxidative process that uses mercury(II)
mass action law. Therefore, the entity of the recycle is reduced with compounds and concentrated sulfuric acid. The letter acts as the
respect to conventional fixed bed reactors. In particular, a gas phase oxidant while mercury is the catalyst. Methane could be con-
process typically requires a recycle ratio of 5:1 while a liquid phase verted to methanol with 50% yield and 85% selectivity. However,
process operates with recycle ratios of 1:1 to 2:1. mercury and sulfuric acid are not environmentally friendly and sub-
Unfortunately, trickle bed and slurry reactors have not been used sequent efforts were devoted to investigating a stable, active and
for methanol production so far because of both modeling and scale- selective platinum complex ((2,2,-bipyridimine)PtCl2) for the oxi-
up issues. In fact, a priori scale-up of trickle bed and slurry reactors dation of methane to methyl esters. The new process yielded
from lab-scale units and data is not possible. Therefore, these tech- conversions to methyl bisulfate up to 72%. Methyl-bisulfate, must
nologies are still under investigation. be subsequently hydrolyzed to sulfuric acid and methanol.
Just to provide a few examples, Air Products developed a slurry A co-current slurry phase process, operating at lower tempera-
liquid phase technology (LPMEOH™) [40]. The fine catalyst par- tures (100–200 °C), was finally proposed by Pittsburgh University.
ticles were suspended in inert mineral oil (Fig. 15). This technology In this process, two methanol molecules are synthesized via the
allows better temperature control and faster mass transfer as well hydrogenolysis of methyl formate, which originates from the
as lower operating and capital costs compared to traditional tubular carbonylation of one methanol molecule. The global process is equiv-
fixed-bed reactors. The heat and mass transfer is very efficient and alent to the reaction of H2 and CO to give methanol. Conversions
the catalyst can be easily replaced. In 1997 this technology was up to 90% and selectivity up to 98% were achieved. However, methyl
adopted in a commercial methanol facility at the Eastman Chem- formate was the limiting reactant at the adopted operating condi-
ical’s coal gasification complex (Kingsport, Tennessee, USA) and tions (170 °C – 50 bar) and this is the reason why the process was
demonstrated to be able to produce methanol at a rate of 300,000 never scale up.
l/d. The process can operate in continuous fashion or discontinu- As a final consideration, liquid phase technologies show higher
ously. An internal heat exchanger allows heat transfer from the catalyst deactivation rates with respect to gas phase processes
mineral oil and enforces the temperature control. A higher conver- [43,44]. Other than sintering of copper crystallites, further deacti-
sion rate is achieved and the recycle is significantly reduced. vation phenomena may occur. For example, carbon formation and
Brookhaven National Laboratory (BNL) in NY (USA) developed a subsequent coking of the catalyst might happen either by the
methanol synthesis process (via the methyl formate reaction pathway) Boudouard reaction (Lee, p. 27 [44]) or via the decomposition of the
operating at lower pressures (<5 MPa) and low temperatures. The hydrocarbon oil used to suspend the catalyst. Alumina leaching may
be also important (Lee, p. 175 [44]). Finally, the catalyst deactiva-
tion can be triggered by carbonyl compounds (such as Fe(CO)5 and
Ni(CO)4), which attack and block active sites. COS was also found
to be a poisoning agent.
decrease in process costs. Currently, no industrial scale applica- CuO ranges between 20% and 80%, ZnO ranges between 15% and
tion of this type of technology is known. 50% and Al2O3 ranges from 4% to 30%. Additives, for instance MgO,
can also be present.
2.4. One-step technologies Magnesia (MgO) is a promoter and acts as catalyst stabilizer. It
was firstly adopted in the catalyst series Katalco (formerly ICI), par-
As discussed above, methanol can be produced with different tech- ticularly in Katalco 51-7. Later on, it has been incorporated in
nologies mainly based on two stage fixed-bed tubular reactors [49], subsequent variants named Katalco 51-8 and Katalco 51-9. It insures
even though other routes seem to be feasible [50]. According to the higher and more stable catalytic activity, thus allowing operation
final production target, methanol synthesis can be coupled with other at lower temperatures and favoring the reaction thermodynamics
syntheses such as the production of formaldehyde [51], dimethyl and the carbon efficiency of the process. It also helps to slow down
ether [52], acetic acid [53] or olefins [54]. These latter compounds thermal sintering and is beneficial to the catalyst selectivity. Now-
are usually manufactured by means of a series of processes [55] adays, catalyst selectivity is over 99% [60,61].
although more and more frequently the so-called “one-step” tech- Most authors believe that metallic copper is the main active com-
nologies are gaining importance. Such technologies rely on new ponent [62,63], while ZnO acts as a supporter and Al2O3 is a promoter.
catalysts as well as catalyst blends capable of favoring the simul- ZnO is also supposed to enhance the dispersion of copper par-
taneous (in a single reactor) synthesis of methanol and one or more ticles and stabilizes the copper crystals. However, it is disputed
of its derivatives. The main benefit is the possibility of shifting the whether the active site is Cu dissolved in ZnO [64] or metallic copper
equilibrium of the methanol synthesis reaction toward the prod- dispersed on the ZnO surface [65,66]. The catalyst activity has been
ucts due to the in situ consumption of methanol itself. The typical shown to correlate quasi-linearly with the exposed area of metal-
case is the co-production of methanol and dimethyl ether, which lic copper [65,66]. This statement is also supported by in situ X-ray
initially was considered only as a by-product of methanol synthe- photoelectron spectroscopy [67] and surface X-ray diffraction studies
sis. Nowadays, it is an interesting compound when the syngas [68]. However, Fujitani and Nakamura [69] showed conflicting results.
feedstock originates from renewable resources (e.g. biomass) [11]. By means of surface science techniques, they found that ZnO forms
Dimethyl ether (DME) is a good alternative to diesel fuel because it the active site other than dispersing Cu particles. Also, Ostrovskii
has a high cetane number, produces smoke-free combustion thanks [70] reported on ZnO being an active site for methanol synthesis.
to its high oxygen content and quickly vaporizes in the engine because The methanol synthesis reaction is an example of a structure in-
of its low boiling point. Traditionally it is produced starting from sensitive catalytic reaction, in which the activity is wholly dependent
methanol in a dehydration reactor [56]: on the total exposed copper area and not affected by the structure
of the crystallites. Variations in the CuO:Al2O3 ratio have a remark-
2CH 3OH ↔ CH 3OCH 3 + H 2O (28) able effect on the relative activity, as shown in accelerated life tests
[71].
Whenever a bifunctional catalyst is available that includes CuO/
ZnO/Al 2 O 3 for methanol synthesis and γ-Al 2 O 3 for methanol 4. Reactor modeling
dehydration, reaction (28) takes places together with the metha-
nol synthesis and water gas shift in the same reaction environment As mentioned in Section 2, the most common technology for pro-
(the same reactor). Reaction (28) is limited by the thermodynamic ducing methanol is the Packed-Bed Reactor (PBR) or, more
equilibrium, but this limitation is not a significant drawback as it specifically, the shell-and-tube boiler-reactor, which was separate-
was for the methanol synthesis reaction. Finally, this DME synthe- ly proposed by different industrial players. PBRs or fixed-bed reactors
sis reaction is exothermic, occurs without change in the number of are commonly used in the chemical industry [72,73]. In the case
moles and is not affected by pressure. It is favored at low temperatures. of methanol and of most processes based on strongly exothermic
reactions, reactors are designed for efficient cooling [3] and are often
3. The catalyst tubular and carry solid catalyst particles on a single side (usually
the tube side). The chemical reactions involved in the chemical
Since the CuO/ZnO/Al2O3 catalyst is the most commonly adopted process take place on the catalyst active surface. PBRs are usually
in industry [57], that is the only catalyst we are going to analyze. the equipment of choice for two principal reasons: they take to
Its composition can vary depending on the manufacturer as dis- higher conversions per weight of catalyst compared to other con-
played in Table 1. verters and they insure efficient heat removal. The observed
conversion depends on the amount of catalyst rather than on the
reactor volume. The reactor modeling requires the characteriza-
Table 1 tion of the process kinetics, of the transport phenomena occurring
Examples of compositions of the catalysts from different manufacturers. inside and around the catalyst, of the catalyst deactivation mecha-
Manufacturer Cu wt% Zn wt% Al wt% Other Ref. nisms and, eventually, the resolution of the mass and energy
IFP 45–70 15–35 4–20 Zr-2-18 [58] balances. Therefore, all of these aspects will be addressed in the next
Synetix (formerly ICI) 20–35 15–50 4–20 Mg [58] chapters. Then, steady state and dynamic models will be analyzed.
BASF 38.5 48.8 12.9 Rare Earth [58]
oxide -5 4.1. Kinetics
Shell 71 24 12 [58]
Süd Chemie 65 22 31 [58]
Dupont 50 19 17 [58] Kinetic modeling is fundamental for the design and optimiza-
United Catalysts 62 21 17 [58] tion of chemical reactors. A detailed knowledge of the reaction
Haldor Topsoe MK-121 >55 21–25 8–10 [58] mechanism leads to a better description of the process, thus al-
Mitsubishi Gas Chemical 63.6 33.4 3 [59]
lowing to optimize it and ensure larger profits. The reaction kinetics
Company
Ammonia Casale 30 50 3 Cr (16) [59]
and the mechanism involved in the production of methanol are still
Lonza 40 20 Zr (40) [59] open issues. Many kinetic studies have been published, but they differ
AIST, RITE 45.2 27.1 4.5 Zr (22.6) [59] in terms of feed gas composition and operating conditions, thus
Si (0.6) conclusions concerning the kinetics and the mechanisms cannot be
YYK Corporation 76.3 11 12.7 [59]
drawn yet. Later on, we will focus on the kinetics of the methanol
G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105 85
CO/CO2/H2 mixtures are used. Dybkjaer [85] suggested a rate ex- show up in two different overall reactions. Therefore, the model si-
pression based on a dual-site Langmuir–Hinshelwood mechanism: multaneously provides two different concentrations for some
H2 is adsorbed, dissociated and reacts with the adsorbed CO2. In this intermediates like formyl and methoxy species. Notwithstanding
model H2 and H2O compete for one specific active site, while CO these limitations, the rate expressions of Graaf et al. are the most
and CO2 adsorb on a second, different active site. commonly used in reactor modeling and fit well the experimental
In 1985 and after a detailed study, Liu et al. [83] proposed that data they generated with a spinning basket reactor at 210–245 °C
the hydrogenation of CO2 is the main source of methanol in absence and 1.5–5 MPa. They demonstrated that methanol originates si-
of water, at low temperature and at low conversion. However, in multaneously from CO and CO2. Moreover, they compared their model
presence of water and at high temperature and conversion, meth- with those proposed by Seyfert and Luft [84], Villa et al. [82], Klier
anol is mainly produced from CO hydrogenation. Villa et al. [82] et al. [57], and Dybkiaer [85] by fitting the parameters of such models
considered CO as the only source of methanol and included the water with their experimental data. Based on this analysis, they showed
gas shift reaction in their kinetic scheme, thus recognizing its im- their model reproduces more accurately the experimental data.
portance in the overall reaction mechanism. In 1987, Kuczynski et al. In 1991, Skrzypek et al. [90] proposed a reaction mechanism
[104] used Villa’s expression for fitting experimental data related based only on the hydrogenation of CO2 and on the water gas shift
to low pressure synthesis from CO and H2 (CO mole fraction from reaction. In fact, they performed experiments with a CO and hy-
0.1 to 0.54, 3–9 MPa, 483–545 K). drogen feed at different concentrations and they did not find any
Takagawa and Ohsugi [86] performed experiments with CO/H2, production of methanol. The same results were previously achieved
CO 2 /H 2 and CO/CO 2 /H 2 mixtures at 4–10 MPa and 483–573 K. by Kagan et al. [109]. On the contrary, they observed that metha-
However, some carbon dioxide was present in the experiments with nol was generated with high selectivity when feeds containing only
CO/H2. In contrast to the results of Klier et al. [57], they did not find CO2 and H2 were employed. Therefore, they assumed that the role
any optimum CO2/CO ratio. They were not able to demonstrate the of CO was restricted to participating in the WGS reaction and re-
promoting effect of CO2 in the lower CO2/CO range but affirmed that moving water from the reaction environment. Thus, CO was
all the three reactions are fundamental in methanol synthesis. supposed to have a beneficial effect on the synthesis reaction. They
However, they declared their results to be in line with those of Klier proposed their own rate expressions that captured the trend of the
et al. [57] and Liu et al. [83]. reaction rate (observed experimentally and including a maximum)
Chincen et al. [65] demonstrated CO2 to be the main reactant in as a function of CO2 and hydrogen concentrations in the feed. Finally,
methanol synthesis by using 14C-labeled reactants. they used a deactivated catalyst to produce the pool of measure-
Shack et al. [88] showed that CO2 inhibits CO hydrogenation but ments on which the model parameters were fitted. However, because
not its own hydrogenation to methanol. These results suggest that the effect of deactivation was not taken into account, the model pre-
CO and CO2 hydrogenation occur on separate sites. Starting from dicts reaction rates that are too small.
this evidence, separate sites for adsorption of CO and CO2 were Similar results were achieved by Millar et al. [110] and by Ras-
assumed. The experiments of Liu et al. [83], carried out with 18O mussen et al. [63]. Millar et al. demonstrate that Cu is the active
labeled carbon dioxide, supported this hypothesis. The identifica- site where the hydrogenation to methanol takes place when CO2/H2
tion of the active sites involved in the methanol synthesis generated feeds are used.
controversies too. McNeil et al. [89] reported on all the hypoth- Askgaard et al. [91] proposed a kinetic model based on param-
eses presented in the literature about the possible (different) active eters evaluated by means of gas phase thermodynamics for the
sites on which the CO and CO2 hydrogenation could take place. They chemisorption process. This is an extension to a previous model pre-
proposed a phenomenological model for methanol production in- sented in 1992 by Ovesen [62,111]. This approach was necessary
volving both the hydrogenation of CO and CO2 and including also to overcome the large pressure and temperature difference between
an inhibition effect of CO2 on CO hydrogenation. They also justi- the operating conditions of lab reactors and industrial converters.
fied the presence of a maximum in methanol production as a They also performed experiments with Cu single crystal and surface
function of the ratio between CO and CO2 in the feed. Two differ- studies. A mechanism involving 16 elementary steps was pro-
ent sites were assumed for the adsorption of CO and CO2: Cu1+ for posed. Askgaard et al. accounted for the only hydrogenation of CO2
CO and Cu0 for CO2. Hydrogen was assumed to adsorb on ZnO. The and water gas shift reaction. The RDS of the methanol synthesis was
inhibition effect of CO2 was explained with the absorption of CO2 identified as the formate hydrogenation to methoxide and oxide.
itself on Cu1+ sites, which are the only ones active for CO hydroge- Their calculated rates showed good fitting to the experimental data,
nation. They proposed a rate expression for methanol production upon adaptation to the actual process operating conditions. They
based on 6 parameters fitted based on their experimental data. The also identified that the methanol synthesis reaction undergoes self-
main difference of the proposed expression with respect to that of inhibition at temperatures lower than 500 K because stable and
Klier et al. [57], from which they started their study, is that the hy- abundant formate species block most of the available active sites.
drogenation of CO2 is included and its inhibition effect on CO The model of Askgaard et al. shows a systematic deviation from ex-
hydrogenation is accounted for. Moreover, their rate expression perimental data at small or large pCO/pCO2 ratio. Complex calculations
covers also the case in which CO2 in not present in the feed, where are necessary to evaluate the model parameters.
it has been showed that methanol is still produced. The microkinetics In 1996,Vanden Bussche and Froment proposed a mechanistic
of the process is also proposed in their paper. They identify the model for methanol production based on a detailed reaction scheme
methoxy hydrogenation as Rate Determining Step (RDS) for meth- that assumes CO2 as the main carbon source [92]. Temperature, pres-
anol production from CO, under the assumption that the methoxy sure and gas phase composition effects are adequately accounted
species is the Most Abundant Reaction Intermediate (MARI). On the for in their model. All the parameters included in their rate equa-
other hand, formate was considered as the MARI for methanol pro- tions were evaluated by solving a fitting problem. Based on their
duction from CO2 and its hydrogenation was assumed the RDS for model, both hydrogen and carbon dioxide undergo dissociative ad-
this process. sorption on copper active sites. Carbonate species are produced,
Graaf et al. [87,108] suggested that CO/CO2 and H2/water could suddenly hydrogenated to bicarbonates and subsequently reduced
adsorb on two different active sites. They proposed that methanol to Cu formate, methoxy species, formaldehyde, and, eventually, to
could be generated via successive hydrogenation of CO and CO2 and methanol. The RDS is the hydrogenation of formate. The mecha-
the WGS reaction followed a formate-based path. However, their nistic nature of the model allows careful extrapolation outside of
kinetic expressions do not take into account that some intermediates the experimental conditions with regard to which the model
G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105 87
parameters are computed. This model accurately predicts the fraction of Cu(110) facet under reducing environment, which is in
maximum conversion rate at 2% CO2. contradiction with what was suggested by Hansen et al. [115]. TEM
Ovesen et al. proposed a dynamic microkinetic model for meth- in combination with a MicroElectroMechanical System (MEMS) will
anol synthesis based on surface measurements [112]. They started enable atomic resolution microscopy at ambient pressure and pos-
from the model of Askgaard et al. [91] (a steady state model) and sibly clarify which type of surface is active at the usual synthesis
applied the so-called Wulff construction to evaluate the surface and conditions employed in industry. Grabow and Mavrikakis mainly
interface free energies in order to clarify the changes in particle mor- addressed CO2 hydrogenation as a mean for CO2 fixation but they
phology. The change in the number of active sites as a function of also included several other CO and CO2 hydrogenation pathways,
the operating conditions is included in their model. They sug- the WGS reaction and the reactions of formation of some byproducts
gested that the observed modification in particle morphology is and intermediates, such as formic acid (HCOOH) and formalde-
related to a modification in the number of oxygen vacancies at the hyde (CH2O). No assumption on RDS was adopted to calculate the
Zn–O–Cu interface. For their model, the number of active sites model parameters, which came from DFT. The proposed model ac-
depends on the conditions of the gas phase mixture in contact with counts for 49 elementary steps is a mean-field microkinetic model.
the catalytic surface. When larger CO/CO2 ratios are employed, the This is justified because adsorbed spices are well-mixed and dis-
number of available active sites increases on the (110) surface plane, tributed on the catalyst surface and diffusion limitations are reduced
where methanol synthesis is faster. This type of model allows better at those temperatures at which the methanol synthesis takes place.
reproduction of the experimental data (taken from Graaf et al. [87]) In this paper, an accurate description of the elementary steps is given
with respect to static microkinetic models. Moreover, it gives better and the model parameters are fitted to the experimental data of Graaf
explanation of the behavior of the system in transient conditions. et al. [87]. The sensitivity of the catalytic activity to the WGS rate
The system dynamics seems to justify the apparently conflicting re- for CO/CO2/H2 mixtures is small but larger WGS rates consume CO
action orders reported in the literature. In fact, under dynamic and reduce the adsorption of CO2. The beneficial effect of CO is found
conditions, different reaction orders are deduced depending on the to be related to the removal of OH*/H2O* (* indicates adsorbed
oxidizing or reducing environment. Once again, metallic Cu appears species) via the WGS reaction, thus shifting the CO2 hydrogena-
to be the active phase in the methanol synthesis. However, the in- tion toward the production of methanol. Simulations with different
teraction with ZnO is crucial in promoting the uniform dispersion feed compositions show that the carbon source, from which meth-
of the Cu particles on the support. Ovesen’s model is more consis- anol originates, depends on the operating conditions and a general
tent with experimental data than Askgaard’s model is. rule does not exist. Finally, the model of Grabow et al. shows some
Kubota et al. [93] proposed another kinetic model where CO2 was discrepancies with regard to the data of Graaf et al. for CO-rich feeds,
the principal responsible for methanol generation. They found their even though those data were extrapolated out of the experimen-
equations to be reasonably in accordance with experimental data tal range. This can be related to the structural changes in the catalyst
coming from a test plant. Setinc and Levec [94] proposed a kinetic under reducing environment.
scheme for the methanol synthesis in liquid phase where the meth- Peter et al. [116] compared three different kinetic models: two
anol yield is proportional to both CO2 and CO concentrations. global kinetic models and a microkinetic model. The first two are
Rozovskii and Lin [95] tested their kinetic models on two dif- a power low kinetics and a semi-empirical Langmuir–Hinshelwood–
ferent gas phase compositions, where one was rich in CO2 and the Hougen–Watson model taken from Vanden Bussche and Froment
other was rich in CO. They found their scheme to be accurate when [92]. The microkinetic model is that of Ovesen et al. [112]. All the
applied to a CO2 rich mixture but not in the other case. models have demonstrated to be accurate in fitting experimental
Lim et al. [96] proposed a kinetic model based on the analysis data. The global models describe more accurately the kinetics at the
of 48 elementary reactions, which takes into account the most likely adopted experimental conditions and are more useful for reactor
rate determining steps (RDS). They found that the surface reac- modeling. On the other hand, the microkinetic model takes into
tion of methoxy species is the RDS for CO hydrogenation while the account the changes in the catalyst morphology, which can be crucial
hydrogenation of a formate intermediate is the RDS for the water in methanol synthesis and are still under investigation. The model
gas-shift reaction. They claimed to have explained the effect of CO2 requires further improvements, i.e. high pressure in situ studies to
on the process for methanol production for the first time and added enlighten the synergistic effects of ZnO and Cu, which at present
the reaction of synthesis of DME to their scheme. This was neces- are not included in Ovesen’s model.
sary to justify the negative values of CO2 conversion that came up Park et al. [98,117] extended the work of Lim et al. [96,118] and
in their experiments. Based on their analyses, the existence of CO2 presented a kinetic model based on a three-site adsorption (Cu+1,
in the reactor feed increases the methanol yield. In fact, a small Cu0 and ZnO) where they also assumed the typical RDSs of com-
amount of CO2 reduces the average rate of the WGS reaction, thus mercial catalysts. The kinetic parameters were computed by fitting
the increasing content of water inhibits the production of DME. the rate equations to 188 experimental sets of data collected at dif-
Therefore, as they proposed in a subsequent paper [113], the pro- ferent operating conditions. They argued that when both CO and
duction of methanol can be optimized by selecting the proper feed CO2 hydrogenation occurs, CO conversion is affected by tempera-
composition. Because they used a Cu/ZnO/Al2O3/Zr2O3 catalyst, it ture, pressure and space velocity changes while CO2 conversion in
would be interesting to repeat the same studies with a Cu/ZnO/ only marginally influenced by these factors, thanks to the WSG re-
Al2O3 catalyst. action. Moreover, conversion is deeply influenced by the CO–
Grabow and Mavrikakis [114] developed an extended hydrogen ratio. The contribution of each reaction shows to be related
microkinetic model. They confirmed methanol to be produced by to the content of CO as well as the maximum methanol concen-
both from CO and CO2 hydrogenation and argued that CO2 is the tration. In particular, the maximum methanol production can be
source of 2/3 of the produced methanol at industrial operating con- attained by optimal selection of the temperature and the CO frac-
ditions. They derived model parameters by using Density Functional tion. Increases in space velocity have negative effects while an
Theory (DFT) calculations performed on the Cu(111) surface. Sub- increase in pressure, which is related to higher reactants concen-
sequently, they fitted the same parameters to published experimental tration, proves to be beneficial. Moreover, studies with different
data measured at industrial operating conditions [87]. They con- particle sizes were performed in order to identify the importance
cluded that a more exposed plain like Cu(100), Cu(110) or Cu (210) of internal diffusion limitations. The presented model shows a dual
can offer a better model surface for studying the active sites in- contribution of CO and CO2 to the production of methanol. If coupled
volved in the methanol synthesis. Ovesen et al. [112] found a larger with a model for evaluating the effectiveness factor, this set of rate
88 G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105
equations allows modeling and designing large-scale reactors quite including Knudsen diffusion. The Wilke model uses the binary Fick’s
easily. law substituting the binary diffusivity with multicomponent
The rate equations of all of the presented kinetic models are sum- diffusivity. It is a simplification of the Maxwell–Stefan equation in
marized in Table 3 along with the experimental conditions under the Fickian form. The Dusty gas model can be simplified by using
which such models were derived/validated. the Wilke–Bosanquet model. This latter comprises the Wilke model
Given the reported analysis, both CO and CO2 seem to be in- and the Bosanquet formula (1944), which evaluates the diffusivity
volved in the synthesis of methanol. It could be interesting to as the reciprocal of the summation of the reciprocal of bulk and
compare the behavior of the different proposed kinetic models when Knudsen diffusion coefficients. Dusty gas and Wilke–Bonsanquet
applied to a real reactor model. In fact, the Graaf et al. and Vanden models include Knudsen diffusion while Maxwell–Stefan and Wilke
Bussche and Froment rate equations are the only ones extensively models neglect it by considering only bulk diffusion. Knudsen dif-
used in reactor modeling. The problem of the kinetic model of Graaf fusion coefficients are related to the mean pore diameter of the
et al. is that in their complex reaction mechanism, based on 48 re- pellets and independent of pressure, whereas bulk diffusion is in-
actions, different intermediate species are present in two different dependent of the mean pore diameter and inversely proportional
reaction pathways. The result is that two different concentrations to pressure. Knudsen diffusion prevails for Knudsen numbers (ratio
of the same product are predicted. Thus, the mechanism pro- of the molecular mean free path and the pore diameter) over 10.
posed by Vanden Bussche and Froment seems to be more convincing. Usually commercial methanol plants are operated in such a way that
Knudsen diffusion is negligible as Knudsen number is below 1. Thus
4.2. Transport phenomena around and inside the catalyst we can disregard this phenomenon at the pellet level. In practice,
if the catalyst mean pore diameter exceeds 10 nm, Knudsen diffu-
The mass transport limitations inside the catalyst are signifi- sion has minor influence on diffusive mass transport [121]. However,
cant especially for the low-pressure methanol synthesis and when if it is significant, then a combined bulk and Knudsen diffusion model
the CuO/ZnO/Al2O3 catalyst is used. For 4.2 mm catalyst pellets, which should be adopted because this affects significantly the predicted
is the size usually adopted, Seyfert and Luft [84] reported an ef- reactor conversions. The same authors [120,122] compared the mass
fectiveness factor of 75% at 538 K and 80 bar. Graaf et al. [108] and mole based models in order to test the equations consistency.
evaluated effectiveness factors ranging from 32% to 92% depend- For methanol synthesis, the deviations of the two approaches were
ing on the temperature and on the component and observed a less than 4% [120]. Rout et al. [122] found a discrepancy of 2%
reduction in the effectiveness factor for increasing temperature. between mass and mole based models. They suggested that this dis-
The problem of intra-particle diffusion inside the catalyst used crepancy might depend on a contradiction between the species
in the production of methanol has been tackled in several contri- boundary condition and total continuity equation in the case of
butions [84,108,119–122]. The pellets are usually supposed to have molar-based models. They also showed that the two so-called rig-
an isotropic pore size distribution as well as an isotropic distribu- orous models, dusty gas and Maxwell–Stefan, insure slightly better
tion of tortuosity, porosity, and thus conductivity. Lommerts et al. agreement between mass and molar based pellet equations. Exist-
[119] investigated the different diffusion models proposed for the ing simulations [120–122], in agreement with Lommerts et al. [119],
evaluation of the catalyst efficiency. They analyzed five principal also pointed out that pressure [130], temperature variations and
approaches: viscous flow within the catalyst particles are irrelevant and have
no effect on the reactor performance.
1. The dusty-gas model [123,124]; The use of parallel or random pore models is discussed by Rout
2. The modified or extended Stefan–Maxwell equations [125–127]; et al. [122]. The parallel pore model assumes that the catalyst par-
3. The classic multi-component pore diffusion (constant diffusion ticle has cylindrical pores with plane surface [131]. In the random
coefficients) combined with convective mass transport; pore model, the pellet is disassembled and reassembled, thus forming
4. The classic multi-component pore diffusion model (constant dif- macro and micro structures [132]. Mass transport is supposed to
fusion coefficients); originate from a combination of diffusion fluxes through the micro
5. A model based on Thiele modulus and “first-order” kinetics for and macro regions. Rout et al. could get good agreement with ex-
CH3OH and H2O production. perimental data by using the parallel model, thus this is the model
recommended for reactor simulations. They also suggested using
Models 1 and 2 implicitly account for convective mass trans- the Maxwell–Stefan model to simulate intra-particle diffusion but
port while the others take into account it explicitly. Models 1, 3, 4 Solvik and Jacobsen [120] concluded that Wilke and Wilke–
and 5 allow for Knudsen diffusion. By analyzing the predictions of Bosanquet models are a good enough approximation for the
these models via comparison to experimental data by Seyfert and simulation of methanol synthesis reactors. This has to do with re-
Luft [84], one can notice that convective mass transport, due to the ducing model complexity and computational time. Based on this
consumption of gas phase reactants during the methanol synthe- review, it is evident that no conclusions can be drawn on which
sis, is negligible compared to mass diffusion. The same conclusions model is more accurate and reliable. However, the simpler model
apply to pressure and temperature gradients inside the particle. In based on Thiele modulus is commonly adopted for online
addition, Lommerts et al. [119] concluded that it is acceptable to optimization/control/simulation purposes.
apply the simplest concept of Thiele modulus to account for intra-
phase diffusion and surface reaction phenomena. In fact, such simple 4.3. Catalyst deactivation
model is more easily incorporated into methanol reactor models and
less consuming in terms of computational time. Moreover, by com- The effective average life of the CuO/ZnO/Al2O3 catalyst depend
paring such simple model to more complex ones, such as the dusty on its stability over time, thus the formulation has to be designed
or the Stefan–Maxwell models, we find acceptable deviations in the to stabilize the copper agglomerates on the catalytic surface at the
values of both methanol and water effectiveness factors. usual operating conditions of the process. Thermal sintering and
Jacobsen’s group deeply analyzed the flux models from the point poisoning are the principal phenomena responsible for catalyst de-
of view of both pellets and reactor modeling [120–122]. They studied activation. Poisons such as sulfur, sometimes iron and nickel
the intra-particle multicomponent diffusion by comparing Maxwell– carbonyls can be introduced into the reactor with fresh syngas. Sulfur
Stefan, Dusty gas, Wilke [128] and Wilke–Bosanquet [129] models. content has to be as low as 5 ppm or even 1 ppm, if possible. Chlo-
The dusty model is an extension of the Maxwell–Stefan model ride impurities can promote catalyst sintering as well. Because the
G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105 89
Table 3
Equations rates.
2–8 MPa k ( fCO f H22 − fCH3OH K eq ) Schermuly and Luft, 1977 [77]
rCO→CH3OH =
483–538 K (1 + AfCO + Bf H2 + CfCH3OH + DfCO2 )2
7.5 MPa ⎛ 1 pCO ⎞
rCO→CH3OH = krCO→CH3OH ⎜ 1 + eq
−3
(
K CO K H2 2 pCO pH2 2 − pCH3OH K reqCO→CH3OH ) Klier at al., 1982 [57]
498–523 K ⎝ kredox pCO2 ⎟⎠ (1 + K CO pCO + K CO2 pCO2 + K H2 pH2 )
⎛ 1 pCH3OH pH2O ⎞
rCO2 →CH3OH = krCO2 →CH3OH ⎜ pCO2 − eq ⎟
⎝ K rCO2 →CH3OH pH3 2 ⎠
f f − fCO f H2O K *3
rRWGS = CO2 H2
M2
N/A A1 A2 A30.5 [ fCO2 f H2 − fCH3OH f H2O ( fH2 K eq,CO )] Dybkjaer, 1985 [85]
rCO2 →CH3OH = 1 2
rCO2 →CH3OH =
k3K CO2 ⎡⎣ fCO2 f H322 − fCH3OH f H2O ( fH3 2K 3eq )⎤⎦
2
3–9 MPa ⎡ p
pH2 2 pCO2 − rCOCH2 →3OH
pH2O ⎤ Skrzypek et al., 1991 [90]
460–550 K ⎢ K eq CH3OH pH2 ⎥
rCO2 →CH3OH = k1K H2 2 K CO2 ⎢ ⎥
⎢ (1 + K H2 pH2 + K CO2 pCO2 + K CH3OH pCH3OH + K H2O pH2O + K CO pCO ) ⎥
3
⎢⎣ ⎦⎥
⎡ pCO pH2O
pH2 pCO2 − rRWGS ⎤
⎢ K eq ⎥
rRWGS = k2K H2 K CO2 ⎢ ⎥
⎢ (1 + K H2 pH2 + K CO2 pCO2 + K CH3OH pCH3OH + K H2O pH2O + K CO pCO ) ⎥
2
⎢⎣ ⎥⎦
32
0.1–0.4 MPa ⎛p ⎞ ⎛ pCO2 ⎞ 2 Askgaard et al., 1995 [91]
r+ = k−11K 5−3 2K 8−1K 9 K 10 K 11 ⎜ H2 ⎟ θ*
483–563 K ⎝ po ⎠ ⎝⎜ p ⎠⎟o
1 pCH3OH pH2O 2
r− = k−11K 5−3 2K 8−1K 9 K 10 K 11 θ*
K G pH3 22 po1 2
rCO2 →CH3OH = r+ − r−
k′ p ⎡1 − K * (
p p )⎥⎦
p p ⎤ H2O CO
⎢⎣ 1 CO2 3
= CO2 H2
(1+ ( K K K )( p ) + K p + K p )
rRWGGS
K H2O p H2O
H2 H2 H2O H2O
8 9 H2 H2
(continued on next page)
90 G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105
Table 3 (continued)
4.9 MPa
473–548 K rCO2 →CH3OH =
kM { pCO2 pH2 − pCH3OH pH2O
[ K M p2H ]2
} Kubota et al., 2001 [93]
[1 + K CO 2
pCO2 + K H2O pH2O ]
2
rRWGS =
kR { pCO2 − pCO pH2O
[ K R pH2 ] }
[1 + K CO 2
pCO2 + K H2O pH2O ]
5 MPa krCO→CH3OH K CO K H2 2 K CH ,CO ( pCO pH2 2 − pCH3OH K PrCO→CH3OH ) Lim et al., 2009 [96]
rCO→CH3OH =
523–553 K (1 + K CO pCO ) (1 + K H0.25 pH0.25 + K H2O pH2O )
krCO2 →CH3OH K CO2 K H2 K CH ,CO2 ( pCO2 pH3 2 − pCH3OH pH2O K PrCO2 →CH3OH )
rCO2 →CH3OH =
(1 + K CO2 pCO2 ) (1 + K H0.25 pH0.25 + K H2O pH2O ) pH2 2
k K K 0.5 ( p p − pCO pH2O K PrRWGS )
rRWGS = rRWGS 0.5CO20.5 H2 CO2 H2
(1 + K H2 pH2 + K H2O pH2O ) (1 + K CO2 pCO2 ) pH0.25
5–9 MPa
rCO→CH3OH =
kr′CO→CH3OH K CO ⎡⎣ fCO f H02.5 − fCH3OH ( fH0.5K Pr
2 CO →CH3OH )⎤⎦ Park et al., 2014 [98]
503–613 K (1 + K CO fCO ) (1 + K H0.25 f H02.5 + K H2O f H2O )
kr′CO2 →CH3OH K CO2 ⎡⎣ fCO2 f H12.5 − f H2O fCH3OH ( f H12.5K PrCO2 →CH3OH )⎤⎦
rCO2 →CH3OH =
(1 + K CO2 fCO2 ) (1 + K H0.25 f H02.5 + K H2O f H2O )
syngas used as feedstock is purified, usually the sintering of Cu par- life. Sintering rates increase for increasing temperature and in pres-
ticles is considered the only source of loss of catalytic activity. This ence of water. Due to the highlighted mechanism, sintering is related
is also suggested by Chincen et al. [133]. The fresh catalyst is black to the absolute Melting Point Temperature (MPT; copper = 1356 K).
because the Cu particles are small while the sintered catalyst is pink. However, two other temperatures have to be accounted for: the Huttig
Sintering results from crystallite growth and has two effects: it temperature (0.3*MPT [K]; copper = 407 K) and the Tamman tem-
reduces the surface area of the catalyst and causes loss of support perature (0.5*MPT [K]; copper = 678 K). At the Huttig temperature,
area. In the case of methanol synthesis, the main mechanisms of surface atoms located at any lattice imperfection start dissociating
crystallite growth are related to crystallite and atomic migrations. and diffusing on the surface. At the Tamman temperature, bulk atoms
The crystallite migration comprises entire crystallites moving on the become mobile. Therefore, sintering starts at the Huttig tempera-
support surface, colliding and coalescing. In the atomic migration, ture and becomes faster and faster up to the Tamman temperature.
metal atoms move on the support surface and join larger crystal- Traces of chlorine in the feed of the methanol reactor promote sin-
lites. Atomic migration is favored at lower temperatures than tering by reacting at about 230 °C with the active metal/metal oxide
crystallite migration. In any case, the global mechanism of sinter- and producing a highly mobile copper chloride phase having a Tamman
ing is rather complex and involves several elementary steps [134]: temperature of 79–174 °C. These temperatures can justify why Tohji
et al. [135] indicated that sintering starts over 227 °C, Supp [136] reports
– dissociation and emission of metal atoms from metal crystallites copper to recrystallize at temperatures over 270 °C and Roberts found
– adsorption and trapping of metal atoms on the surface that copper crystals begin growing at 250 °C. Therefore, we can con-
– diffusion of metal atoms across the surface clude that sintering starts at 190–227 °C and becomes faster and more
– metal or metal oxide particle dispersion severe at 270–300 °C.
– surface wetting by metal particles The sintering of copper crystallites takes to a reduction in the
– metal particle nucleation catalyst active surface area with a resulting decrease in activity. Sun
– coalescence of metal particles et al. [137] studied the effect of the feed composition on the cata-
– capture of atoms by metal particles lyst deactivation. They performed experiments both with differential
– liquid formation and finite conversion reactors and with different feeds: CO2/H2 and
– metal volatilization CO/H2. For differential conversion reactors, they found that CO/H2
– splitting of crystallites feeds cause rapid catalyst deactivation because of the reducing nature
– metal atom vaporization of CO while CO2/H2 feeds practically do not induce any deactiva-
tion. The previous conclusions are flipped at finite conversion
The type of chemical process and the corresponding operating con- conditions. In particular, CO/H2 feeds insure slower catalyst deac-
ditions influence which of the abovementioned steps happen in real tivation in virtue of the presence of reaction products (such as CO2
G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105 91
produced by the WGS reaction) that slow down the over-reduction da1 kd1pCO2
of Cu and ZnO due to CO. Reversely, CO2/H2 feeds cause faster de- = −a1d1 2
of 2–5%. This is because of the optimal oxidation state of copper dt 1 + K CO2 pCO2 + K CO2 K 1H 2K HCO2 pCO2 p1H22
under such conditions. Higher CO2 concentrations lead to higher de-
A pseudo-homogeneous reactor model including this hybrid
activation rates. This seems to occur due to the production of H2O
kinetic scheme was simulated for 3 months by Løvik [147]. At the
that disaggregates the catalyst matrix and reduces the surface of
copper crystallites [43,139,140]. end of the run, a2 was small and a1 was nearly zero, thus this coupled
kinetic-deactivation model overestimates the rate of deactivation
The catalyst deactivation due to sintering can be modeled with
a power law expression (PLE): (typically, a catalyst for methanol production lasts for 3–4 years).
da
= −kd (T ) a n (32)
dt 4.4. Steady-state models
Table 4
Deactivation models.
Exp. conditions Reactor type Deactivation model Kinetic constant Author, year Ref.
[h−1]
250 °C Liquid phase bubble column da kd,1 = 2.1⋅ 10−4 Roberts et al., 1993 [143]
= −kd ,1a
5.27 MPa dt
240 °C Gas phase Berty Reactor da ⎛ f ⎞
m
kd,2 = 1.5 ⋅ 10−2 Kuechen and Hoffmann, 1993 [144]
5 MPa = −kd ⎜ CO ⎟ a 2 = −kd ,2a 2
dt ⎝ fCO2 ⎠
(m is not reported)
250 °C Liquid phase pilot plant da kd,3 = 2.1⋅ 10−4 Cybulski, 1994 [145]
= −kd ,3a
dt E = 91270 [ J mol]
240 °C Fixed bed da kd,4 = 8.78 ⋅ 10−3 Skrzypek et al., 1994 [132]
= −kd ,4a5
7 MPa dt (data from Bart and Sneeden, 1987)
250 °C Slurry auto clave reactor da kd,5 = 2.4 ⋅ 10−3 Skrzypek et al., 1994 [132]
= −kd ,5a5
5.27 MPa dt (data from Roberts et al., 1993)
250 °C Liquid phase bubble column da kd,6 = 2.7 ⋅ 10−4 Skrzypek et al., 1994 [132]
= −kd ,6a1.5
5.27 MPa dt (data from Roberts et al., 1993)
250 °C Gas phase Berty Reactor da kd,7 = 4 ⋅ 10−4 Sahibzada et al., 1997 [134]
= −kd ,7a n
5.27 MPa dt
(n equal to 10 at start and 1 after 10 h)
92 G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105
Table 5
Constitutive equations of the reactor model for methanol synthesis. Molar and mass quantities are in capital and lower case letters, respectively; ̃ and ̂ define quantities
referring to moles or mass, respectively.
1 Energy balance p dT
Mc NR
mix
Aint dz
d
( )
= π int U (Tshell − Tbulk ) + ρcat (1 − ε b ) ∑ − ΔH rjeac η j R j
Aint j
1 − ε b ⎞ ⎞ u2ρ gas ⎛ 1 − ε b ⎞
= − ⎜ 1.75 + 150 ⎛⎜
1 Momentum balance dP ⎛
dz ⎝ ⎝ Re ⎟⎠ ⎟⎠ d p ⎜⎝ ε b3 ⎟⎠
2 Mass balance m dω i NR
= ρcat (1 − ε b ) ∑ MWiυi, jη j R j
Aint dz j
ˆ pmix dTbulk NR
2 Energy balance mc
Aint dz
d
(
= π int U (Tshell − Tbulk ) + ρcat (1 − ε b ) ∑ − ΔĤˆ reac
Aint j
j )
η j rj
1 − ε b ⎞ ⎞ u2ρ gas ⎛ 1 − ε b ⎞
= − ⎜ 1.75 + 150 ⎛⎜
2 Momentum balance dP ⎛
dz ⎝ ⎝ Re ⎟⎠ ⎟⎠ d p ⎜⎝ ε b3 ⎟⎠
1 − ε b ⎞ ⎞ u2ρ gas ⎛ 1 − ε b ⎞
= − ⎜ 1.75 + 150 ⎛⎜
3 Momentum balance dP ⎛
dz ⎝ ⎝ Re ⎠⎟ ⎠⎟ d p ⎝⎜ ε b3 ⎠⎟
1 − ε b ⎞ ⎞ u2ρ gas ⎛ 1 − ε b ⎞
= − ⎜ 1.75 + 150 ⎛⎜
4 Momentum balance dP ⎛
dz ⎝ ⎝ Re ⎠⎟ ⎠⎟ d p ⎝⎜ ε b3 ⎠⎟
(2) The pseudo-homogeneous models based on mass balances. As- formulated in terms of mass fractions, once can get two additional
sumption (9) is removed while assumption (8) is still enforced. benefits: a conservative description of reactive systems with vari-
(3) The heterogeneous models based on molar balances. As- able number of moles (this is the case of methanol synthesis); and
sumption (8) is removed but assumption (9) is still in place. the overall molar conservation equation is no longer necessary. The
(4) The heterogeneous models based on mass balances. Both as- recent literature presents a number of contributions aimed at pro-
sumptions (8) and (9) are removed. posing models for the methanol synthesis reactor based on mole
balances [154–156]. These authors state that the assumptions in-
As already mentioned, intra-phase mass transport phenomena troduced in such models are reasonable because the methanol yield
are modeled by means of the catalyst effectiveness factor, which is per pass is smaller than 7%. However, those assumptions might be
generally calculated from the modified Thiele modulus [150–152], risky when it comes to estimate the hydrogen fraction. Other re-
even though different options have been discussed in Section 4.2. searchers propose to use mass balances [148,157] and compare the
The pseudo-homogeneous formulation can be employed in order predictions originated from the two different types of models. The
to reduce the computational effort required by heterogeneous molar fraction of methanol calculated via molar balances are in-
models. Notice that by neglecting both concentration and temper- evitably underestimated compared to those computed via mass
ature gradients between the gas and the solid phase, solid and gas balances. In fact, there is no apparent reason for assuming that any
phase mass and energy balances, which were previously derived variation in molar flows due to chemical reactions is negligible. Very
separately, can be combined. likely, this erroneous belief derives from a misunderstanding dating
In fact, pseudo-homogeneous formulations allow completely back to 1990s. Some well-established reaction engineering books
changing the nature of the mathematical problem to address. Het- (i.e. [73,158]), which several authors referred to, proposed ex-
erogeneous models lead to systems of differential-algebraic equations, amples of models of heterogeneous reactors based on molar balances.
which are stiff by definition [153]. Instead, pseudo-homogeneous However, this was just because the trivial examples used as case
models lead to ordinary differential equation systems, which are easier studies included only equimolar reactions. These models are no longer
to solve. Moreover, it is well-known that by employing mass balances valid for the methanol synthesis.
G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105 93
The difference between heterogeneous and pseudo-homogeneous temperature zone inside the catalyst bed. The full computation re-
models is essentially the presence/absence of gradients between gas quired less than half an hour. Their simulations also show that the
and solid phases. The simplification introduced in pseudo- reverse water gas shift reaction occurs mainly close to the reactor
homogeneous models can be justified every so often by the high inlet. The concentration of carbon dioxide decreases while it is con-
gas velocity adopted in industrial reactors. In fact, these operating verted to methanol, carbon monoxide and water. At the same time,
conditions take to very thin boundary layers around the catalyst the gas temperature increases, thus generating a slight increase in
pellets. For the current industrial operating conditions, Manenti et al. gas velocity close to the reactor inlet. A difference between the tem-
[148] demonstrated that the pseudo-homogeneous model pro- perature of gas and catalyst is also observed. However, note that
vides predictions for temperature and composition profiles that the inlet gas velocity they used is about 0.3 m/s in while other
overlap those of the heterogeneous model. In addition, Rezaie and authors mention spatial velocities around 0.7 m/s [162] or even
co-workers [155] performed the same type of comparison and higher [163] (related to industrial operating data). For such higher
achieved similar results showing that the pseudo-homogeneous gas velocities no temperature difference between the solid and the
model slightly overestimates the temperature hot spot. This is not gas phase is observed.
due to neglecting the species diffusivity in the boundary layer. In Another important aspect to consider is that the most common
fact, the presence of transport limitations around the particle would configuration of the methanol synthesis reactor comprises two stages:
yield opposite effects. The explanation for that resides in thermal the water-cooled section, which is that previously described, and
exchange issues. The temperature profile of the synthesis reactor the gas-cooled section, where the fresh syngas entering the reactor
is particularly sensitive to the inlet feed flowrate. In particular, the is preheated thanks to the hot effluent of the water-cooled section.
position of the hot spot significantly shifts upon changes in the feed All the aforementioned assumptions and considerations are still valid
flow. Analogously, a feed flow reduction leads the reacting system for the gas-cooled section of the two-stage configuration and the
toward the thermodynamic equilibrium, thus the molar fraction of same applies to the reported models. There is only the need for
methanol increases and the throughput of the reactor drops. Studies adding an additional conservation equation, i.e. the energy balance
conducted by Rezaie and co-workers [155] highlighted that upon for the cooling gas that flows though the shell side of the gas-
reducing the inlet gas flowrate, the behavior of the two phases (solid cooled section [164]:
and gas) drastically changes. Consequently, the results coming from
pg dT
Mc D
pseudo-homogenous and heterogeneous models no longer overlap. = −π i U shell (T − Tshell ) (34)
This is because when the inlet flowrate is larger, the gas flowing Ac dz Ac
through the reactor heats up slower and the methanol yield de-
creases. On the other hand, the feed temperature does not affect Rahimpour and Lotfinejad assessed the effectiveness of the coun-
significantly the reactor behavior, even though there is a connec- tercurrent configuration of the gas-cooled section of the two-
tion between the operating temperature of the steam reformer and stage reactor [165]. They assumed that both the shell and the tube
the final methanol yield, at least from a process viewpoint [159]. side could be modeled as one-dimensional plug flow equipment,
Thus, the tube side is suitably designed to cope with even strong neglected the intra-phase diffusion in the catalyst pellets, ne-
oscillations in the inlet temperature. In industrial best practice, the glected the axial heat dispersion, and assumed the gas mixture to
feed temperature is adjusted to control or partially compensate for be ideal. As shown in their article (Fig. 16), the total production of
catalyst deactivation [155,160]. The rate of heat transfer toward the methanol is slightly larger for the countercurrent configuration. The
boiling water flowing outside the tubes affects methanol conver- aforementioned considerations have general validity as they apply
sion in such a way that the optimal performance of the reactor is to different syngas sources and different PBR types. Other reactor
a function of the shell side temperature: high temperatures reduce configurations deserve special models. For instance, Arab et al. [166]
the maximum achievable conversion due to the thermodynamic analyzed the impact of arranging the solid catalytic phase as either
equilibrium while low temperatures reduce the rate of reaction. Since a packed bed or monoliths. For this study, the reactor feed com-
the shell side is practically a steam generator, its temperature can prised only CO2 and H2. The production of methanol via PBRs and
be controlled by adjusting the steam pressure with a roof valve. From reactors exploiting catalyst monoliths (monolithic reactors, Fig. 17)
former analyses, it derives that the reactor behavior can be accu- was analyzed by modeling the heat and mass transfer phenomena
rately described via heterogeneous models but can be also properly
predicted via pseudo-homogeneous models, when both feed flowrate
and shell temperature and are reasonably close to nominal condi-
tions. A qualitative validity range is also proposed [148].
Interesting research work was presented by Petera et al. [161],
which introduces a powerful method for solving multi-dimensional
differential systems thanks to a universal coupling of macro- and
micro-geometries represented by a two-mesh finite element frame-
work. The micro-mesh refers to the catalyst particle while the macro-
mesh is applied to the catalyst bed. They developed a numerical
algorithm for solving multidimensional heterogeneous reactor
models and applied it to a methanol reactor. Their model includes
all of the four dimensions previously mentioned: axial and radial
reactor coordinates, catalyst particle radius and time. In this way,
they could compute the concentration and temperature profiles in
the axial and radial directions as well as inside the catalyst parti-
cle. The effect of incidental crushing of the catalyst was simulated
and showed the birth of a hotter hot spot due to the reduction in
particle diameter. In addition, the buoyancy effect was accounted
for, which relates to the increase in temperature and consequent
decrease in density. This combination of these phenomena is ex-
pected to take to an irreversible dynamics generating a high Fig. 16. Co-current/counter-current comparison for the two-stage synthesis reactor.
94 G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105
commercial tools forces to model the methanol synthesis reactor 4.5. Dynamic models
by means of predefined models and the unit of choice is usually the
Gibbs reactor. Only few articles mention the possibility of integrat- Dynamic modeling is traditionally more complex to imple-
ing external, more detailed, models into such commercial suites (i.e. ment and more time consuming than steady-state modeling. This
using CAPE-OPEN directives) [176]. is why it is less common to find detailed models for the dynamics
Table 6
Dynamic models. Heterogeneous PBR by Rezaie et al. [155] (model 1); homogeneous PBR by Rezaie et al. [155] (model 2); homogeneous model by Manenti et al. [177]
(model 3)).
where εS void fraction of catalyst; Ct total concentration (mol/m3); a catalyst activity; yi molar fraction in the fluid phase; η catalyst effectiveness factor; yiS, mole fraction of
component i in the solid phase; ρB density of the catalytic bed (kg/m3); ri, reaction rate of component i (mol/kg/s); kgi, mass transfer coefficient for component i (m/s); ĉ pS
specific heat of the solid at constant pressure (J/g); TS temperature of the solid phase (K); T bulk gas phase temperature (K); ΔHrienthalpy variation of the reaction i; aVspecific
surface area of catalyst pellet (m2/m3); hf gas–solid heat transfer coefficient (W/m2/K); M total molar gas flow rate (mol/s); Ac cross section area of each tube; c pg specific
heat of the gas at constant pressure (J/mol); Di tube inside diameter (m); Ushell overall heat transfer coefficient between coolant and process streams (W/m2/K); Tshell tem-
perature of coolant stream (K).
and control of methanol synthesis reactors in the literature. Sim- synthesis reactor. From a conceptual viewpoint, the easiest way to
ilarly to steady-state modeling, the majority of the contributions proceed implies discretizing the axial profile and then solve a series
address the dynamics modeling of PBRs while only very few authors of ordinary differential equation or differential-algebraic equation
try to assess the dynamic performances of other reactor configu- systems. Parvasi and co-workers [179,180] detailed the procedure
rations such as membrane or slurry reactors. Rezaie et al. [155] to initialize numerically the solution of such a system but anyway
proposed a first paper aimed at comparing different dynamic models very misleading results can be easily achieved due to numerical issues.
for the PBR and test their capability of fitting industrial data (col- They used a first order upwind discretization scheme to approxi-
lected under decaying catalyst performance). Table 6 reports the mate the convective terms and a centered difference scheme for the
pseudo-homogeneous and heterogeneous models (model (1) and diffusive terms. Manenti et al. [177] solved the system using the Back-
model (2)) analyzed by Rezaie et al. and model (3), which is based ward Euler method, which is typically implemented in commercial
on mass balances and was proposed by Manenti et al. [177,178]. It suites, and the trend is reported in Fig. 21. It is worth noting that,
is important to note that the first two models, based on molar bal- when the system is subject to a perturbation, the model is predict-
ances, should be implemented in terms of mass balances (see Section ing a discontinuity in the dynamic trends, which has no physical
4.4). The dusty gas model [124] and kinetic scheme by Graaf et al. meaning. In fact, we are assuming the gas flow inside each single
[87] were used. According to Jackson’s formulation [123] the im- tube of the PBR to be experience turbulent motion. This infeasible
plementation of any model for the methanol synthesis could benefit result originates from the stiffness of the system of equations that
from using elemental balances rather than component balances. This easily leads to numerical divergence. Moreover, even though it is
logic was applied by Manenti et al. and proved to be correct. neglected in steady-state models, it is apparently essential to account
A common problem of dynamic models is their numerical so- for axial dispersion in dynamic models. In fact, by doing so, the orig-
lution and the corresponding convergence issues. A finite-difference inal hyperbolic partial differential equation system is transformed
approach is usually adopted to discretize the partial differential into a parabolic partial differential equation system and the non-
equation system deriving from the dynamic modeling of the methanol physical fluctuations are somehow removed.
G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105 97
Fig. 21. Effect of flux segregation assumption and Forward Euler integration method. The trends deal with three snapshots at 16 s, 20 s, and 80 s after the step-
perturbation has occurred [177]. The existence and validation of dynamic models led to studies of process control for the methanol synthesis reactor and process (see Section
5.2).
Other reactor configurations are being studied in the litera- It is worth noting that the proposed model does not enforce total
ture, even tough not systematically. For example, Rahimpour’ group mass conservation because the total mass flowrate in the retentate
is investigating a membrane reactor. Parvasi et al. proposed a hy- and permeate sections are not constant with z. This is because there
drogen perm-selective membrane reactor, whose model has to exists mass transfer between the two phases. A continuity equa-
account for the hydrogen permeation through the palladium mem- tion should be added to the system to make the model consistent.
brane (Table 7). The hydrogen permeation has to be also calculated However, since the molecular weight of hydrogen is small, the error
in the model. If the membrane is composed by pure palladium, the induced might be small as well.
model equations has been proposed by Itoh [181] while if the mem- Wu and Gidaspow proposed hydrodynamic models for slurry re-
brane is palladium-silver, the model equations were proposed by actors [187]. In particular, they implemented a two-dimensional
different authors [182,183]. Other authors [184] proposed similar hydrodynamic model of the Air Products slurry bubble column reactor
water-perm-selective membrane reactors and studied also differ- for producing methanol from syngas. Their model is the first to compute
ent membrane reactor configurations [185,186]. the gas and the particle holdups as well as the particle rheology.
Table 7
Perm-selective membrane reactor model [179].
Reaction side Mass balance, all components except hydrogen ∂yi M ∂yi N R
ε BC t =− + ∑ ηrj ρBa i = all except hydrogen
∂t Ac ∂z j =1
∂y H N R α
Reaction side Mass balance, hydrogen
ε BC t
∂y H
∂t
=−
M
Ac ∂z
+ ∑ ηrj ρBa + H
j =1 Ac
( pHt − pHr )
Reaction side Energy balance ∂T M ∂T 3
π Di
ε BC t c pg =− + ηρBa ∑ ri (− ΔHi ) + U shell (Tshell − T ) +
∂t Ac ∂z i =1 Ac
π Do
+ Utube (Ttube − T )
Ac
∂y H α H
Tube side Mass balance, hydrogen
Ct
∂y H
∂t
=−
M
Ac ∂z
+
Ac
( pHr − pHt )
Tube side Energy balance ∂T M ∂T π Do
C t c pg =− + Utube (Ttube − T )
∂t Ac ∂z Ac
αH is the hydrogen permeation rate constant. It is not clear in the reference if the balances are referring to the atomic or molar hydrogen.
98 G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105
Fig. 23. Systematic staging optimal design applied to the two-stage water-cooled/gas-cooled PBR. The traditional case is the best industrial practice; the 3rd case is the
systematic staging applied to the methanol yield; the 4th case is the systematic staging accounting for the steam generation in the shell side of the water-cooled stage.
measurement of the most relevant information is performed and exothermic. Shahrokhi and Baghmisheh confirmed that well-
measures are sent to a dedicated server located inside the distrib- tuned PID loops provide faster control actions and smaller overshoot
uted control system (DCS). Before using the measures for control in terms of reactor temperature profile (the overshoot is the maximum
and optimization purposes, a crucial step is to purge them from pos- deviation from the original set point). Based on these results, they
sible stochastic errors and from possible outliers (or gross errors) redefined the optimal temperature profile deriving from the optimal
as well [198,199]. This step is essential to prevent any misleading design and suggested a “constrained optimal temperature” for safer
operation of the automatic control due to temporary or persisting operation as reported in Fig. 24. More recently, advances have been
wrong measurement. Such a purification procedure goes under the proposed to insure improved operation of methanol synthesis
name of process data reconciliation [200,201]. Reconciled data can reactors/plants: the plantwide process control for the overall meth-
be used for control purposes and, subsequently, for online optimi- anol production plant (Luyben [193]) and the nonlinear model
zation purposes like in a stiff pyramid of decisions where the predictive control (MPC) for the only PBR (Manenti et al. [206,207]).
information flows upward and actions flow downward. Kordnejad Luyben proposed a plantwide control system based on PID loops
et al. [202] proposed a real-time optimization (RTO) based on simu- for the overall process including the synthesis section and the pu-
lated plant data to optimize the operational performance of a PBR rification section (Fig. 25). The philosophy is not to have any
reactor. They measured the temperature of the tube side outflow uncontrolled battery limit and to saturate all the degrees of freedom
and used the shell-side temperature as unique optimization vari- of the plant with the minimum amount of control loops. The sim-
able. In this way, they improved methanol yield by 15%. Such a plest control scheme is the best control scheme [208].
methodology originates from the best industrial practice, i.e. the use Following the process flow, Luyben proposes a flow controller
of steady-state models. While in the past optimization was based on the inlet syngas stream. Then, the syngas is compressed in the
on steady models and was completely independent of control in the rotating units K1 and K2 in between of which the gas flow is cooled
operational management pyramid [203], nowadays control and op- down by a utility exchanger. The temperature controller regulates
timization are completely integrated in the so-called integrated the inlet temperature to K2 by manipulating the coolant flowrate.
optimal control or model predictive control (MPC) [204] and dynamic A pressure controller is also applied to the rotating shaft of K2 to
real-time optimization (DRTO) [205], where one unique dynamic ensure the pressure on the downstream line is in line with its
model is used to handle all the operational targets. Unfortunately, setpoint. The compressed syngas goes to the Feed Effluent Heat
due to the complexity of the methanol reactor, there is lack of con- Enchanger (FEHE), where it is preheated by the hot effluent of the
tributions and advances in process control and optimization (based methanol synthesis reactor. The FEHE has no degrees of freedom
on the new dynamic-simulation-based approach) concerning such for control and its target is to recover as much energy as possible
process unit. It is well-known that the typical PBR for methanol syn- (energy-saving process intensification). Since the inlet tempera-
thesis could evolve dynamically along 4 different coordinates: axial ture of the reactor is a key parameter, it is controlled via a
and radial directions, catalyst radius, and time [178]. Shahrokhi and temperature loop that manipulates the utility fluid of a heat ex-
Baghmisheh [156] applied different control techniques to the Lurgi- changer. It is worth saying that, in contrast to what Fig. 25 shows,
type one-stage water-cooled methanol synthesis reactor. They the utility exchanger usually heats up the process stream (its cor-
modeled both the tube side storing the catalyst and the shell side responding arrow should point down, thus indicating that the utility
comprising the steam drum unit. Then, they applied the proportional– fluid is cooling down). The reactor effluent is pre-cooled in the FEHE
integral (PI) and the proportional–integral–derivative (PID) techniques and its temperature is refined in a utility exchanger, where a tem-
to manage the system when feed flow perturbations and shell-side perature controller is managing the corresponding coolant flow rate.
pressure changes occurred. Their target was to assess the optimal After the cooling step, a good portion of methanol condenses and
operating conditions of an industrial reactor subject to typical fluc- is recovered in a flash drum separator. The drum aims at separat-
tuations in the feedstock (tube-side) and in the pressure/temperature ing the unreacted syngas from condensable compounds, thus
of the boiling water (shell-side), which are more conservative than methanol and water. The syngas is recompressed and recycled back
those coming from the optimal design. During transient operation, the reactor. Since recycles introduce an additional degree of freedom
there could be any temperature hot-spot along the catalyst bed, which in the control paradigm [209], a control loop is also associated with
is not identified via steady-state studies. This might accelerate the this process stream. Moreover, in order to prevent any accumula-
catalyst sintering and deactivation as well as favor the so-called tion of by-products due to the gas recycle, a vent is always present.
methanation reaction (CO + 3 H2 → CH4 + H2O), which is extremely Here, it is managed via a composition controller.
100 G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105
Fig. 24. Constrained optimal temperature accounting for PID actions in smoothening perturbations.
On the other hand, the liquid accumulated in the drum is trans- comprises: (I) two loops controlling the liquid levels of reflux drum
ferred to a second flash drum separator to refine the recover and reboiler; (II) one loop taking care of the pressure on the last
additional syngas. The liquid level in the first and second drums is tray; (III) another loop controlling the inlet flowrate (possibly jointly
managed by level controllers that manipulate the liquid outflows. to the reflux flowrate); (IV) and two loops managing the column
The second drum is also equipped with a gas recycle line that goes temperatures on the first and last tray (sometimes different trays
back to the reactor and is managed via a pressure controller. Since are selected). In this specific case, one single temperature control
a gas purge is already performed on the recycle stream leaving the loop is enough to manage the temperature profile along C1. In ad-
first drum, it is unnecessary to insert another one and a potential dition, it is important to say that the temperature of the inlet flow
control loop is avoided. The liquid flow exiting the second drum is is a key parameter and is usually refined with a utility exchanger.
transferred to the distillation column C1 for methanol/water sep- In the Luyben scheme, the second flash drum separator is sup-
aration. The minimum number of control loops for a distillation posed to already predispose the temperature for proper operation
column is usually six [210]. The conventional control scheme of the distillation column. The liquid level controllers applied to the
reflux drum and the reboiler define the flowrates of methanol (the The first quadratic term ensures that the controlled variables stay
final product) and water, respectively. The last outlet stream is the as close as possible to their respective setpoints. The second qua-
gas effluent from the partial condenser, which is recycled back to dratic term suggests desired values for the manipulated variables
the reactor after recompression. This process line is managed by a (usually related to energy saving policies). Finally, the third qua-
flowrate controller as well. Another advance to the control/ dratic term is the so-called anti-ringing that aims at preventing
optimization of PBRs for producing methanol was recently proposed possible close-loop instabilities. In practice, it smoothens the optimal
by Manenti and Bozzano [207]. It does not relate to the whole plant trajectory of the manipulated variables, thus avoiding repeated fluc-
scale, but it has to do with the control of the methanol converter tuations over time. All the aforementioned terms are represented
and introduces a technique called optimal control or MPC. The MPC as vectors of differences properly weighted (w). The weighting co-
has been invented at the end of 1970s in its linear formulation efficients allow balancing the order of magnitude of each term and/
[211–213] and counts many applications worldwide [214]. In the or hierarchically rank them. Despite its benefits, the application of
last decades, nonlinear MPC [215,216] has proven to be more ef- MPC to real life systems is still an open issue when the dynamic
fective than its linear counterpart because model mismatch is system model has a distributed nature, as PBR does. Some authors
minimized (all chemical plants exhibit nonlinear dynamics). The are working in this direction on other systems [219] but general
basic idea is to exploit the prediction potential of dynamic models methodologies are still unavailable. The reason for this has to do
to foresee the future behavior of the plant and thus select and apply with the computational time required to solve optimization problem
the right/better control action in advance. In other words, the MPC and the issue of handling the solution of partial differential equa-
runs the dynamic process model several times to build different tions automatically. Both of these are still challenging problems.
control trajectories and the control actions associated with the best Manenti and Bozzano [207] proposed an MPC framework applied
trajectory are selected and implemented onto the system (servo- to the methanol synthesis reactor in order to simultaneously control
mechanism, regulation and hybrid problems can be efficiently the dynamic temperature profile and the axial position of the hot-
handled). Since real plants are continuously subject to distur- spot. These goals were achieved by manipulating the shell-side
bances, the calculation procedure is iterated every sampling time temperature and the recycle flowrate. Thanks to this control scheme,
in a so-called receding horizon fashion [217,218]. The MPC is for- two major benefits are ensured: the reactor is optimally con-
mulated as a weighted multivariable constrained nonlinear trolled and the position of the temperature hot-spot is managed to
programming problem with a quadratic objective function subject prevent the catalyst from sintering. An approach similar to MPC is
to the differential or differential-algebraic equations of the dynamic called DRTO [220,221]. The difference is that MPC solves a control
model and to specific bounds on the states and the manipulated problem via a quadratic objective function while DRTO refers to eco-
variables as well: nomic aspects and relies on an economic objective function.
Moreover, the degrees of freedom of MPC are the manipulated vari-
⎧ k+hp k +hp −1
⎫ ables whereas those of DRTO are the setpoints of the controlled
⎪ ∑ ω j ( x j − x j ) + ∑ ω1,l (ul − ul ) ⎪
set 2 tar 2
⎪ j =k+1 l =k ⎪ variables. It means that the DRTO is able to suggest operating con-
min
u(k )…. u(k +hc −1)
⎨ k +hp −1
⎬ ditions that optimally track the market trends (e.g. the methanol
⎪ ⎪
∑ ( (i ) (i −1)
)
2
⎪ + ω 2,l u l − ul ⎪ grade production) without modifying the conventional control loops
⎩ l =k ⎭ (instead MPC overwrites these conventional loops). Rahimpour and
subjject to: co-workers [222,223] applied the DRTO to a membrane reactor and
⎧ x min
j ≤ x j ≤ x max
j
a spherical-bed reactor for methanol synthesis proposing optimal
⎪ min state/manipulate variable trajectories. Kordabadi and Jahanmiri [224]
⎪ul ≤ ul ≤ ul
max
⎨ (i),min (40) extended the concept of DRTO (usually adopted on the short-
⎪Δul ≤ Δul(i) = ul(i) − ul(i−1) ≤ Δul(i),max term) to the identification of optimal trajectories related to the
⎪Convolution Model catalyst bed deactivation. Fig. 26 reports the optimal trend they found
⎩
for the shell side temperature and the related profile of the tube
where u stands for the manipulated variables; x stands for con- side temperature during 1400 days of operations. Clearly, as the cat-
trolled variables, ω stands for the weight of each single term; hp alyst loses activity, the DRTO raises the shell temperature
stands for prediction horizon; k stands for the number of sam- progressively. Fuad et al. [225] have also implemented the newest
pling times; set and tar stand for setpoint and target, respectively. techniques for DRTO included in the Interior Point OPTimizer (IPOPT)
102 G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105
[203] for the long-term optimization of methanol synthesis subject concerning operations management studies but the industrial ten-
to catalyst deactivation. dency to integrating different production processes to save energy
and capital cost (e.g. methanol synthesis and biomass gasifica-
6. Conclusions tion) will stimulate research in such field.
Another expected development is related to the number of
Methanol is a very important compound. The development of process parameters to design/optimize in methanol plants. The lit-
processes capable of producing it from residues and waste is a hot erature is proposing self-consistent cases, which consider single units
research topic worldwide. Methanol is also expected to be a po- or parts of the methanol synthesis process. Others are considering
tential solution to the partial deployment of fossil source-based simplified optimization procedures to improve the design of the
economies. Moreover, it is a recognized energy carrier that is whole plant. The perspective is to move toward an overall plant op-
better than other alternatives in terms of transportation, storage timization based on detailed models, which includes process yield,
and reuse. CO2 conversion, steam co-generation, pre-heating/pre-cooling energy
Technologies to produce it are mature and of different nature. intensification, and so on.
Improvements are expected from new processes based on slurry Beyond all the aforementioned concepts, there is still strong lack
and membrane reactors. In addition, heat removal and recovery is of experimental and industrial data concerning methanol-related
a potential field of innovation for single and multi fixed-bed reactor production systems.
configurations. In every case, new or improved catalysts for meth-
anol production are likely to be discovered in the near future. These
catalytic materials are expected to allow lowering the tempera- References
ture and pressure required for the methanol synthesis reaction. While
pressure is still useful to favor the reaction that is limited by the [1] Olah GA, Goeppert A, Surya Prakash GK. Beyond oil and gas: the methanol
thermodynamic equilibrium, the current temperature of the meth- economy. Weinheim, Germany: Wiley-VCH; 2006.
[2] Olah GA. Beyond oil and gas: the methanol economy. Angew Chem Int Ed Engl
anol synthesis process is over the Huttig temperature, thus promoting 2005;44:2636–9.
the catalyst sintering. [3] Lange J-P. Methanol synthesis: a short review of technology improvements.
Since the operating conditions of current industrial processes are Catal Today 2001;64:3–8.
[4] Zeep. <http://zeep.com/market-opportunities/market-development-and
rather different from those of several decades ago, the scientific com- -growth/>; 2015 [accessed 21.04.15].
munity should start expanding the validity range of the current [5] Agency IE. World energy outlook. Paris: 2004.
available kinetic models. In addition, new kinetic models will be nec- [6] Pino I, Vita A, Laganà M, Recupero V. Hydrogen from biogas: catalytic tri-
reforming process with Ni/LaCe O mixed oxides. Appl Catal B 2014;148-
essary, which account for morphological modifications in the catalyst
149:91–105.
structure due to different operating conditions. These new models [7] Vita A, Cristiano G, Italiano C, Pino I, Specchia S. Syngas production by methane
will clarify the contribution of the single mechanisms through which oxy-steam reforming on Me/CeO2 (Me = Rh, Pt, Ni) catalyst lined on cordierite
methanol is synthesized (CO and CO2 routes). This will be of inter- monoliths. Appl Catal B 2015;162:551–63.
[8] Matthey J. www.matthey.com/>; 2015.
est for the relevant problem of CO2 reuse. Kinetic models both for [9] Kochloefl K. Ertl G, Knozinger H, Weitkamp J, editors. Steam reforming.
methanol synthesis and for similar systems, which are able to Handbook of heterogeneous catalysis. Weinheim, Germany: Wiley-VCH; 1997.
account for the dynamic modifications in the catalyst structure in [10] Pavone A. Mega methanol plants. Report No 43D, process economics program.
Menlo Park, California: SRI Consulting; 2003.
operation, are going to spread. [11] Ahrenfeldt J, Thomsen TP, Henriksen U, Clausen LR. Biomass gasification
Transport phenomena have been studied in detail and under- cogeneration – a review of state of the art technology and near future
stood. Moreover, there seems to be a certain convergence in the perspectives. Appl Therm Eng 2013;50:1407–17.
[12] Manenti F. CO2 as feedstock: a new pathway to syngas. Comput Aided Chem
literature between the need of academia for complex models and Eng 2015. p. 1049–54.
that of industry for simpler models. This is going to take to rea- [13] Manenti F., Pierucci S., Molinari L. Process for reducing CO2 and producing
sonable compromises between accuracy of the model predictions syngas. WO 2015/015457 A1. 2015.
[14] Bassani A, Pirola C, Maggio E, Pettinau A, Frau C, Bozzano G, et al. Acid Gas
and computational efforts. to Syngas (AG2S™) technology applied to solid fuels gasification: a new way
Steady-state models for fixed-bed reactors are very well con- to reduce H2S and CO2 emissions by improving Syngas production. Appl
solidated, even though there is still the need for further refinements, Energy 2016;doi:10.1016/j.apenergy.2016.06.040.
[15] Christiansen JA. Method of producing methyl alcohol from alkyl formate. US
e.g. the conversion of molar balances to mass balances to improve
Patent 1,302,011. 1919.
the prediction accuracy. Conversely, other reactor configurations have [16] Marchionna M, Lami M, Raspolli Galletti AM. Synthesizing methanol at lower
not been exhaustively studied yet. This is the case of slurry and mem- temperature. Chemtech 1997;27.
brane reactors. The scientific community is expected to explore all [17] Olah GA, Prakash GKS. Efficient and selective conversion of methane to
methanol. US patent No 60 671-650. 2005.
main chemical and physical phenomena governing these very prom- [18] Olah GA, Gupta B, Farina M. Selective monohalogenation of methane over
ising and emerging technologies. supported acid or platinum metal catalysts and hydrolysis of methyl halides
A different picture emerges for dynamic models because the dis- over gamma-alumina-supported metal oxide/hydroxide catalysts. A feasible
path for the oxidative conversion of methane into methyl alcohol/dimethyl
tributed nature of the fixed-bed methanol reactor makes their ether. J Am Chem Soc 1985;107:7097.
development rather cumbersome. According to the future need for [19] Periana RA, Mironov O, Taube DJ, Gamble S. High yield conversion of methane
producing methanol from biomass and biogas, significant variabil- to methyl bisulfate catalyzed by iodine cations. Chem Commun (Camb)
2002;2376.
ity in the feedstock, in the syngas composition and in the operating [20] Navarro RM, Sanchez-Sanchez MC, Alvarez-Galvan MC, del Valle F, Fierro JLG.
conditions will be experienced. This will soon demand detailed a Hydrogen production from renewable sources: biomass and photocatalytic
priori analyses unavoidably based on dynamic models. The main opportunities. Energy Environ Sci 2009;2:35–54.
[21] Manenti F, Adani F, Rossi F, Bozzano G, Pirola C. First-principles models and
problem is still the complex numerical solution of partial differ- sensitivity analysis for lignocellulosic biomass-to-methanol conversion process.
ential equations. Moreover, the assumptions required to model the Comput Chem Eng 2015;doi:10.1016/j.compchemeng.2015.05.012.
dynamics of the methanol synthesis reactor differ from those in- [22] Cassedy ES. Prospects for sustainable energy. A critical assessment. Cambridge:
University Press; 2000.
troduced for steady-state modeling. The future development of
[23] Nyns E-J. Methane. Ullmann’s Encyclopedia of industrial chemistry, vol. 21.
reliable and possibly low computational demanding dynamic models Weinheim, Germany: Wiley-VCH; 2003. p. 599.
as well as the improvements in computing infrastructures will allow [24] Xiaoding X, Moulijn JA. Mitigation of CO2 by chemical conversion: plausible
applying concepts like model predictive control and dynamic op- chemical reactions and promising products. Energy Fuels 1996;10:305.
[25] Saito M, Murata K. Development of high performance Cu/ZnO-based catalysts
timization, already successfully adopted elsewhere, to the complex for methanol synthesis and water-gas shift reaction. Catal Surv Asia
methanol reactor as well. The literature lacks of contributions 2004;8:285.
G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105 103
[26] Tijm PJA, Waller FJ, Brown DM. Methanol technology development for the new [59] Wernicke H-J, Plass L, Schmidt F. Methanol generation. In: Bertau M,
millennium. Appl Catal A Gen 2001;221:275. Offermanns H, Plass L, Schmidt F, Wernicke H-J, editors. Methanol: the basic
[27] Pinto A, Rogerson PL. Impact of high fuel cost on plant design. Chemengng chemical and energy feedstock of the future. 2014.
Progress 1977;73:95–100. [60] Moulijn JA, Makkee M, van Diepen A. Chemical process technology. Chichester.:
[28] LeBlanc JR, Schneider RV. Cape horn Methanol: reliable performance in a Wiley; 2001.
remote location. World Methanol Conference, Vancouver, British Columbia, [61] Hansen JB. Ertl G, Knozinger H, Weitkamp J, editors. Methanol synthesis.
Canada. 1991. Handbook of heterogeneous catalysis, vol. 4. Weinheim, Germany: Wiley-VCH;
[29] Toyo Engineering. <http://www.toyo-eng.com/jp/en/products/petrochmical/ 1997.
methanol>; 2015 [accessed 21.04.15]. [62] Ovesen CV. Kinetic modeling of reactions on Cu surfaces. [Ph D Thesis].
[30] Linde Engineering. <http://www.linde-engineering.com/en/process_plants/ Laboratory of Applied Physics, Technical University of Denmark, Lyngby. 1992.
hydrogen_and_synthesis_gas_plants/gas_generation/isothermal_reactor/ [63] Rasmussen PB, Holmblad PM, Askgaard T, Ovesen CV, Stoltze P, Norskov JK,
index.html>; 2015 [accessed 31.03.15]. et al. Methanol synthesis on Cu(100) from a binary gas-mixture of CO2 and
[31] Haid J, Koss U. Lurgi’s Mega-Methanol technology opens the door for a new H2. Catal Lett 1994;26:373–81.
era in down-stream applications. Stud Surf Sci Catal 2001;136:399–404. [64] Klier K. Methanol synthesis. Adv Catal 1982;31:243–313.
[32] Supp E, Quinkler RF. Meyers RA, editor. The Lurgi low-pressure methanol [65] Chinchen GC, Hay CM, Vandervell HD, Waugh KC. The measurement of copper
process. Handbook of synfuels technology. McGraw-Hill; 1984. p. 113–31. surface areas by reactive frontal chromatography. J Catal 1987;103:79–86.
[33] Spath PL, Dayton DC. Preliminary Screening – Technical and Economic As- [66] Chinchen GC, Waugh KC, Whan DA. The activity and state of the copper surface
sessment of Synthesis Gas to Fuels and Chemicals with Emphasis on the Potential in methanol synthesis catalysts. Appl Catal 1986;25:101–7.
for Biomass-Derived Syngas. NREL/TP-510-34929, <http://wwwnrelgov/ [67] Fleisch TH, Mieville RL. Studies on the chemical state of Cu during methanol
docs/fy04osti/34929pdf>; 2003. synthesis. J Catal 1984;90:165–72.
[34] Lurgi. <http://www.zeogas.com/files/83939793.pdf>; 2015 [ accessed 11.04.15]. [68] Clausen BS, Steffensen G, Fabius B, Villadsen J, Feidenhans’l R, Topsoe H. In
[35] Takase I, Niwa K. Mitsubishi (MGC/MHI) methanol process. CEER, Chem Econ situ cell for combined XRD and on-line catalysis tests: studies of Cu-based
Eng Rev 1985;17:24–30. water gas shift and methanol catalysts. J Catal 1991;132:524–35.
[36] Hawkins GB. http://www.slideshare.net/GerardBHawkins/methanol-flowsheets [69] Fujitani T, Nakamura J. The chemical modification seen in the Cu/ZnO methanol
-a-competitive-review>; 2015 [accessed 21.04.15]. synthesis catalysts. Appl Catal A Gen 2000;191:111–29.
[37] Fiedler E, Grossmann G, Kersebohm DB, Weiss G, Witte C. Methanol. Ullmann’s [70] Ostrovskii VE. Mechanisms of methanol synthesis from hydrogen and carbon
encyclopedia of industrial chemistry release. 6th ed. Wiley-VCH, Verlag GmbH oxides at Cu–Zn-containing catalysts in the context of some fundamental
& CoKGaA; 2003. problems of heterogeneous catalysis. Catal Today 2002;77:141–60.
[38] Davy Process Technology. <http://www.davyprotech.com/what-we-do/ [71] Fitzpatrick T, Hicks T. Catalysts with higher and more stable activity enable
licensed-processes-and-core-technologies/core-technologies/synthesis/ cost savings and boost output in methanol production, advances in methanol
specification/>; 2015 [accessed 22.03.15]. synthesis. Digital Refining. <www.digitalrefining.com/article/1000297>; 2010.
[39] Wang J, Anthony RG, Akgerman A. Mathematical simulations of the [72] Fogler HS. Elements of chemical reaction engineering. 4th ed. Prentice Hall;
performance of trickle bed and slurry reactors for methanol synthesis. Comput 2006 ISBN 0-13-047394-4.
Chem Eng 2005;29:2474–84. [73] Levenspiel O. Chemical reaction engineering. 3rd ed. USA: John Wiley and Sons;
[40] Heydorn EC, Diamond BW, Lilly RD. Commercial-Scale Demonstration of the 1999.
Liquid Phase Methanol (LPMEOH_) Process: Final Report. Prepared by Air [74] Leonov E, Karabaev MM, Tsybina EN, Petrishcheva GS. Kinet Katal
Products Liquid Phase Conversion Company for the US DOE National Energy 1973;14:970–5.
Technology Laboratory, <http://wwwostigov/scitech/servlets/purl/823132>; [75] Rozovskii AY, Lin GI, Liberov LG, Slivinskii EV, Loktev SM, Kagan YB, et al. Kinet
2003. Katal 1977;18:691.
[41] Marchionna M, Di Girolamo M, Tagliabue L, Spangler MJ, Fleisch TH. Natural [76] Rozovskii AY, Kagan YB, Lin GI, Slivinskii EV, Loktev SM, Liberov LG, et al. Kinet
gas conversion. Pert V. Stud Surf Sci Catal 1998;119. Katal 1975;16:810.
[42] Periana RA, Taube DJ, Gamble S, Taube H, Satoh T, Fujii H. Platinum catalysts [77] Schermuly O, Luft G. Methanol-syntheses im treibstrahlreaktor. Ger Chem Eng
for the high-yield oxidation of methane to a methanol derivative. Science 1977;1:222.
1998;280:560–4. [78] Denise B, Sneeden RPA. Hydrogenate CO2. Chemtech 1982;12:108–12.
[43] Kung HH. Deactivation of methanol synthesis catalysts – a review. Catal Today [79] Monnier JR, Apai G, Hanrakan HJ. Effects of CO2 on the conversion of H2/CO
1992;11:443–53. to methanol over copper-chromia catalysts. J Catal 1984;88:523–5.
[44] Lee S. Methanol synthesis technology. Florida.: CRC Press; 1990. [80] Bardet R, Thivolle-Cazat J, Trambouze Y. Hydrocondensation des oxydes de
[45] Iulianelli A, Pirola C, Comazzi A, Galli F, Manenti F, Basile A. Water gas shift carbone, à la pression atmosphérique, sur des catalyseurs Cu-ZnO-Al2O3.
membrane reactors. In: Basile A, Di Paola L, Hai FI, Piemonte V, editors. Influence de l’eau sur la formation du méthanol. Cr Hebd Séances Acad Sci
Membrane reactors for energy applications and basic chemical production, Sèr C 1984;299:423.
vol. 76. Elsevier, Woodhead Publishing Series in Energy; 2016. [81] Chinchen GC, Denny PJ, Parker DG, Short GD, Spencer MS, Waugh KC, et al.
[46] Sea B, Lee KH. Methanol synthesis from carbon dioxide and hydrogen using The activity of Cu-ZnO-AlrO. Methanol synthesis catalysts. Prepr Am chem
a ceramic membrane reactor. Reaction Kinetics Catal Lett 2003;80:33–8. Soc Diu Fuel Chem 1984;29:178–88.
[47] Gallucci F, Paturzo L, Basile A. An experimental study of CO2 hydrogenation [82] Villa P, Forzatti P, Buzzi-Ferraris GGG, Pasquon I. Synthesis of alcohols from
into methanol involving a zeolite membrane reactor. Chem Eng Process carbon oxides and hydrogen. 1. Kinetics of the low-pressure methanol
2004;43:1029–36. synthesis. Ind Eng Chem Process Des Dev 1985;24:12–19.
[48] Iliuta I, Larachi F, Fongarland P. Dimethyl ether synthesis with in situ H2O [83] Liu G, Willcox D, Garland M, Kung HH. The role of CO2 in methanol synthesis
removal in fixed-bed membrane reactor: model and simulations. Ind Eng Chem on Cu-Zn oxide: an isotope labeling study. J Catal 1985;96:251–60.
Res 2010;49:6870. [84] Seyfert W, Luft G. Untersuchungen zur Methnolsynthese im
[49] Rahimpour MR. A two-stage catalyst bed concept for conversion of carbon Mitteldruckbereich. Chem Ing Tech 1985;57:482–3.
dioxide into methanol. Fuel Process Technol 2008;89:556–66. [85] Dybkjaer I. Design of ammonia and methanol synthesis reactors. NATO
[50] Riaz A, Zahedi G, Klemeš JJ. A review of cleaner production methods for the conference on chemical reactor design and technology, Canada. 1985.
manufacture of methanol. J Clean Prod 2013;57:19–37. [86] Takagawa M, Ohsugi M. Study on reaction rates for methanol synthesis
[51] Thivasasith A, Sirijaraensre J, Khongpracha P, Warakulwit C, Jansang B, from carbon monoxide, carbon dioxide, and hydrogen. J Catal 1987;107:161–
Limtrakul J. Reaction mechanism of methanol to formaldehyde over Fe- and 72.
FeO-modified graphene. Chemphyschem 2015;16:986–92. [87] Graaf GH, Stamhuis EJ. Beenackers AACM. Kinetics of low-pressure methanol
[52] McBride K, Turek T, Güttel R. Direct dimethyl ether synthesis by spatial synthesis. Chem Eng Sci 1988;43:3185–95.
patterned catalyst arrangement: a modeling and simulation study. AIChE J [88] Shack CJ, McNeil MA, Rinker RG. Methanol synthesis from hydrogen, carbon
2012;58:3468–73. monoxide, and carbon dioxide over a CuO/ZnO/Al2O3 catalyst: I. Steady-state
[53] Mekala M, Goli VR. Comparative kinetics of esterification of methanol-acetic kinetics experiments. Appl Catal A Gen 1989;50:247–63.
acid in the presence of liquid and solid catalysts. Asia-Pac J Chem Eng [89] McNeil MA, Schack CJ, Rinker RG. Methanol synthesis from hydrogen, carbon
2014;9:791–9. monoxide and carbon dioxide over a CuO/ZnO/Al2O3 catalyst: II. Development
[54] Tian P, Wei Y, Ye M, Liu Z. Methanol to olefins (MTO): from fundamentals to of a phenomenological rate expression. Appl Catal 1989;50:265–85.
commercialization. ACS Catal 2015;5:1922–38. [90] Skrzypek J, Lachowska M, Moroz H. Kinetics of methanol synthesis over
[55] Zhu Y, Wang S, Ge X, Liu Q, Luo Z, Cen K. Experimental study of improved commercial copper/zinc oxide/alumina catalysts. Chem Eng Sci 1991;46:2809–
two step synthesis for DME production. Fuel Process Technol 2010;91:424– 13.
9. [91] Askgaard TS, Norskov JK, Ovesen CV, Stoltze P. A kinetic model of methanol
[56] Bercic G, Levec J. Intrinsic and global reaction rate of methanol dehydration synthesis. J Catal 1995;156:229–42.
over .gamma.-alumina pellets. Ind Eng Chem Res 1992;31:1035–40. [92] Vanden Bussche KM, Froment GF. A steady-state kinetic model for methanol
[57] Klier K, Chatikavanij V, Herman RG, Simmons GW. Catalytic synthesis of synthesis and the water gas shift reaction on a commercial Cu/ZnO/Al2O3
methanol from CO/H2: IV. The effects of carbon dioxide. J Catal 1982;74:343– catalyst. J Catal 1996;161:1–10.
60. [93] Kubota T, Hayakawa I, Mabuse H, Mori K, Ushikoshi K, Watanabe T, et al.
[58] Spath PL, Dayton DC. Preliminary Screening – Technical and Economic Kinetic study of methanol synthesis from carbon dioxide and hydrogen. Appl
Assessment of Synthesis Gas to Fuels and Chemicals with Emphasis on the Organomet Chem 2001;15:121–6.
Potential for Biomass-Derived Syngas. <wwwnrelgov/docs/fy04osti/34929pdf>; [94] Setinc M, Levec J. Dynamics of a mixed slurry reactor for the three phase
2015. methanol synthesis. Chem Eng Sci 2001;56:6081–7.
104 G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105
[95] Rozovskii AY, Lin GI. Fundamentals of methanol synthesis and decomposition. [127] Krishna R, Wesselingh JA. The Maxwell-Stefan approach to mass transfer. Chem
Top Catal 2003;22:137–50. Eng Sci 1997;52:861–911.
[96] Lim HW, Park MJ, Kang SH, Chae HJ, Bae JW, Jun KW. Modeling of the kinetics [128] Wilke CR. Diffusional properties of multicomponent gases. Chem Eng Prog
for methanol synthesis using Cu/ZnO/Al 2O3/ZrO2 catalyst: influence of carbon 1950;46:95–104.
dioxide during hydrogenation. Ind Eng Chem Res 2009;48:10448–55. [129] Bosanquet CH. Report BR-507. British TA. 1944.
[97] Grabow LC, Mavrikakis M. Mechanism of methanol synthesis on Cu through [130] Burghart A, Aerts J. Pressure changes during diffusion with chemical reaction
CO2 and CO hydrogenation. ACS Catal 2011;1:365–84. in a porous Pellet. Chem Eng Process: Process Intensif 1988;23:77–87.
[98] Park N, Park MJ, Lee YJ, Ha KS, Jun KW. Kinetic modeling of methanol synthesis [131] Prachayawarokorn S, Mann RF. Effects of pore assembly architecture on catalyst
over commercial catalysts based on three-site adsorption. Fuel Process Technol particle tortuosity and reaction effectiveness. Catal Today 2007;128:88–99.
2014;125:139–47. [132] Froment GF, Bishoff KB. Chemical reactor analysis and design. 2nd ed. Hoboken,
[99] Natta G. Synthesis of methanol. Catalysis. NY, USA.: Reinhold; 1955. p. 3. NJ, USA: John Wiley and Sons; 1979.
[100] Bakemeier H, Laurer PR, Schroder W. Development and application of a [133] Chincen GC, Denny PJ, Jenning JR, Spencer MS, Waugh KC. Synthesis of
mathematical model of the methanol synthesis. Chem Eng Prog, Symp Ser 1 methanol, 1., catalysts and kinetics. Appl Catal 1988;36:1.
1970;66:1–10. [134] Argyle MD, Bartholomew CH. Heterogeneous catalyst deactivation and
[101] Vedage GA, Herman RG, Klier R. Chemical trapping of surface intermediates regeneration: a review. Catalysts 2015;5:145–269.
in methanol synthesis by amines. J Catal 1985;95:423–34. [135] Tohji K, Udagawa Y, Mizushima T, Ueno A. The structure of the Cu/ZnO catalyst
[102] Agny RM, Takoudis CG. Synthesis of methanol from carbon monoxide and by an in situ EXAFS Study. J Phys Chem 1985;89:5671.
hydrogen over a copper-zinc oxide-alumina catalyst. Ind Eng Chem Prod Res [136] Supp E. Improved methanol process. Hydrocarbon Process 1981;3:71–5.
Dev 1985;24:50–5. [137] Sun JT, Ian S. Metcalfe, and mortaza sahibzada deactivation of Cu/ZnO/Al2O3
[103] Nappi A, Fabbricino L, Hudgins RR, Silveston PL. Influence of forced feed methanol synthesis catalyst by sintering. Ind Eng Chem Res 1999;3868–72.
composition cycling on catalytic methanol synthesis. Can J Chem Eng [138] Skrzypek J, Sloczynski J, Ledakowicz S. Methanol synthesis. Polish Scientific
1985;63:963–70. Publishing; 1994.
[104] Kuczynski M, van Ooteghem A, Westerterp KR. Methanol adsorption by [139] Ladebeck J. Improve methanol synthesis. Hydrocarbon Process 1993;89–91.
amorphous silica alumina in the critical temperature range. Colloid Polym Sci [140] Sahibzada M, Chadwick D, Metcalfe IS. de Pontes M, Espinoza RL, Nicolaides
1986;264:362–7. CP, Scholz JH, Surell M, editors. Methanol synthesis from CO2/H2 over
[105] Herman RG, Klier K, Simmons GW, Finn BP, Bulko JB. Catalytic synthesis of Pd-promoted Cu/ZnO/Al2O3 catalyst: kinetics and deactivation. Natural Gas
methanol from CO/H2. I. Phase composition, electronic properties, and Conversion IV. Studies in Surface Science and Catalysis Elsevier Science B V;
activities of the Cu/ZnO/M2O3 catalysts. J Catal 1979;56:407–29. 1997. p. 107.
[106] Mochalin VP, Lin GI, Rozovskii AY. Kinetic-model of the process of methanol [141] Bartholomew CH. Sintering kinetics of supported metals: new perspectives
synthesis on the Snm-1 catalyst. Khim Promst 1984;1:11. from a unifying GPLE treatment. Catal Today 1993;2:1–124.
[107] Malinovskaya OA, Rozovskii AY, Zolotarskii IA, Lender V, Lin GI, Matros YS, [142] Forzatti P, Lietti L. Catalyst deactivation. Catal Today 1999;52:165–81.
et al. Kinetic-model of methanol synthesis over a catalyst containing copper. [143] Roberts GW, Brown DM, Hsiung TH, Lewnard JJ. Deactivation of methanol
Kinet Katal 1987;28:851. synthesis catalyst. Ind Eng Chem 1993;32:1610–21.
[108] Graaf GH, Scholtens H, Stamhuis EJ, Beenackers AACM. Intra-particle diffusion [144] Kuechen C, Hoffmann U. Investigation of simultaneous reaction of carbon
limitations in low-pressure methanol synthesis. Chem Eng Sci 1990;45:773– monoxide and carbon dioxide with hydrogen on a commercial copper/zinc
83. oxide catalyst. Chem Eng Sci 1993;48:3767–76.
[109] Kagan YB, Rozovskii AY, Liberov LG, Slivinskii EV, Lin GI, Loktev SM, et al. Study [145] Cybulski A. Liquid-phase methanol synthesis: catalysts, mechanism, kinetics,
of mechanism of methanol synthesis from carbon monoxide and hydrogen chemical equilibrium, vapor-liquid equilibria, and modeling – a review. Catal
using radioactive carbon isotope C14. Dokl Akad Nauk SSSR 1975;224:1081–4. Rev Sci Eng 1994;36:557–613.
[110] Millar GJ, Rochester CH, Waugh KC. An in situ high pressure FT-IR study of [146] Rahimpour MR, Fathikalajahi J, Jahanmiri A. Selective kinetic deactivation
CO2/H2 interactions with model ZnO/SiO2, Cu/SiO2and Cu/ZnO/SiO2 methanol model for methanol synthesis from simultaneous reaction of CO2 and CO with
synthesis catalysts. Catal Lett 1992;14:289–95. H2 on a commercial copper/zinc oxide catalyst. Can J Chem Eng 1998;76:753–
[111] Ovesen CV, Stoltze P, Nørskov JK, Campbell CT. A kinetic model of the water 61.
gas shift reaction. J Catal 1992;134:445–68. [147] Løvik JI. Modelling, estimation and optimization of the methanol synthesis
[112] Ovesen CV, Clausen BS, Schiøtz J, Stoltze P, Topsøe H, Nørskov JK. Kinetic with catalyst deactivation. [PhD Thesis]. 2001.
implications of dynamical changes in catalyst morphology during methanol [148] Manenti F, Cieri S, Restelli M. Considerations on the steady-state modeling
synthesis over Cu/ZnO catalysts. J Catal 1997;168:133–42. of methanol synthesis fixed-bed reactor. Chem Eng Sci 2011;66:152–62.
[113] Lim HW, Hye JJ, Park MJ, Kim HS, Bae JW, Ha KS, et al. Optimization of [149] Ravaghi-Ardebili Z, Manenti F. Unified modeling and feasibility study of novel
methanol synthesis reaction on Cu/ZnO/Al2O3/ZrO2 catalyst using generic green pathway of biomass to methanol/dimethylether. Appl Energy
algorithm: maximization of the synergic effect by the optimal CO2 fraction. 2015;145:278–94.
Korean J Chem Eng 2010;27:1760–7. [150] Graaf GH, Sijtsema PJ, Stamhuis EJ, Joosten GE. Chemical equilibria in methanol
[114] Grabow LC, Mavrikakis M. Mechanism of methanol synthesis on cu through synthesis. Chem Eng Sci 1986;41:2883–90.
CO 2 and CO hydrogenation. ACS Catal 2011;1:365–84. [151] Graaf GH, Stamhuis EJ, Beenackers A. Kinetics of low-pressure methanol
[115] Hansen PL, Wagner JB, Helveg S, Rstrup-Nielsen JR, Calusen BS, Topsøe H. Sci synthesis. Chem Eng Sci 1988;43:3185–95.
China Chem 2002;295:2053. [152] Lommerts BJ, Graaf GH, Beenackers A. Mathematical modeling of internal mass
[116] Peter M, Fightl MB, Ruland H, Kaluza S, Muhler M, Hinrichsen O. Detailed transport limitations in methanol synthesis. Chem Eng Sci 2000;55:5589–
kinetic modeling of methanol synthesis over a ternary copper catalyst. Chem 98.
Eng J 2012;203:480–91. [153] Buzzi-Ferraris G, Manenti F. Differential and differential-algebraic systems for
[117] Park N, Park MJ, Ha KS, Lee YJ, Jun KW. Modeling and analysis of a methanol the chemical engineer: solving numerical problems, vol. 4. Weinheim,
synthesis process using a mixed reforming reactor: perspective on methanol Germany: Wiley-VCH; 2014.
production and CO2 utilization. Fuel 2014;129:163–72. [154] Velardi SA, Barresi AA. Methanol synthesis in a forced unsteady-state reactor
[118] Lim HW, Park MJ, Chae HJ, Kang SH, Jun KW. Mathematical modeling for the network. Chem Eng Sci 2002;57:2995–3004.
Cu/ZnO/Al2O3 catalyzed synthesis of methanol. 8th World Congress of [155] Rezaie N, Jahanmiri A, Moghtaderi B, Rahimpour MR. A comparison of
Chemical Engineering: Incorporating the 59th Canadian Chemical Engineering homogeneous and heterogeneous dynamic models for industrial methanol
Conference and the 24th Interamerican Congress of Chemical Engineering 8, reactors in the presence of catalyst deactivation. Chem Eng Process Process
2009. Intensif 2005;44:911–21.
[119] Lommerts BJ, Graaf GH, Beenackers AACM. Mathematical modeling of internal [156] Shahrokhi M, Baghmisheh GR. Modeling, simulation and control of a methanol
mass transport limitations in methanol synthesis. Chem Eng Sci synthesis fixed-bed reactor. Chem Eng Sci 2005;60:4275–86.
2000;55:5589–98. [157] Abrol S, Hilton CM. Modeling, simulation and advanced control of methanol
[120] Solsvik J, Jakobsen HA. Modeling of multicomponent mass diffusion in porous production from variable synthesis gas feed. Comput Chem Eng 2012;40:117–
spherical pellets: application to steam methane reforming and methanol 31.
synthesis. Chem Eng Sci 2011;66:1986–2000. [158] Fogler HS. Elements of chemical reaction engineering. Englewood Cliffs, NJ:
[121] Solsvik J, Jakobsen HA. Multicomponent mass diffusion in porous pellets: Prentice-Hall, USA; 1992.
effects of flux models on the pellet level and impacts on the reactor level. [159] Park N, Park M-J, Ha K-S, Lee Y-J, Jun K-W. Modeling and analysis of a methanol
Application to methanol synthesis. Can J Chem Eng 2013;91:66–76. synthesis process using a mixed reforming reactor: perspective on methanol
[122] Rout KR, Hillestad M, Jakobsen HA. A numerical study of pellet model production and CO2 utilization. Fuel 2014;129:163–72.
consistency with respect to molar and mass average velocities, pressure [160] Løvik I, Hillestad M, Hertzberg T. Modeling and optimization of a reactor
gradients and porosity models for methanol synthesis process: effects of flux system with deactivating catalyst. Comput Chem Eng 1999;23:S839–42.
models on reactor performance. Chem Eng Res Des 2013;91:296–317. [161] Petera J, Nowicki L, Ledakowicz S. New numerical algorithm for solving
[123] Jackson R. Transport in porous catalysts. Elsevier; 1977. multidimensional heterogeneous model of the fixed bed reactor. Chem Eng
[124] Mason EA, Malinaukas AP. Gas transport phenomena in porous media: the J 2013;214:237–46.
dusty-gas model. Chem Eng Monograph 17. NY.: Elsevier; 1983. [162] Panahi PN, Mousavi SM, Niaei A, Farzi A, Salari D. Simulation of methanol
[125] Maxwell JC. On the dynamic theory of gases. Philos Trans R Soc 1866;157:49– synthesis from synthesis gas in fixed bed catalytic reactor using mathematical
88. modeling and neural networks. Int J Sci Eng Res 2012;3.
[126] Stefan J. Uber das gleichgewicht und die bewegung insbesondere die diffusion [163] Yusup S, Anh NP, Zabiri H. A simulation study of an industrial methanol reactor
van gasgemengen. Sitzber Akad Wiss Wien 1871;63:63–124. based on simplified steady-state model. Int J Res Rev Appl Sci 2010;5:213–22.
G. Bozzano, F. Manenti / Progress in Energy and Combustion Science 56 (2016) 71–105 105
[164] Manenti F, Leon-Garzon AR, Bozzano G. Energy-process integration of the [194] Mirvakili A, Rostami M, Paymooni K, Rahimpour MR, Moghtaderi B. Hydrogen
gas-cooled/water-cooled fixed-bed reactor network for methanol synthesis. looping approach in optimized methanol thermally coupled membrane reactor.
Chem Eng Trans 2013;1243–8. Int J Hydrogen Energy 2012;37:235–49.
[165] Rahimpour MR, Lotfinejad M. A comparison of co-current and counter-current [195] Cheng SH, Chen HJ, Chang H, Chang CK, Chen YM. Multi-objective optimization
modes of operation for a dual-type industrial methanol reactor. Chem Eng for two catalytic membrane reactors – methanol synthesis and hydrogen
Process 2008;47:1819–30. production. Chem Eng Sci 2008;63:1428–37.
[166] Arab S, Commenge JM, Portha JF, Falk L. Methanol synthesis from CO2 and [196] Kralj AK, Glavic P. Multi-criteria optimization in a methanol process. Appl
H2 in multi-tubular fixed-bed reactor and multi-tubular reactor filled with Therm Eng 2009;29:1043–9.
monoliths. Chem Eng Res Des 2014;92:2598–608. [197] Zhang Y, Cruz J, Zhang S, Lou HH, Benson TJ. Process simulation and
[167] Air Products and Chemicals, Inc Liquid phase methanol la porte process optimization of methanol production coupled to tri-reforming process. Int J
development unit: modification, operation, and support studies. Report to Hydrogen Energy 2013;38:13617–30.
Department of Energy (USA) for Contract No DE-AC22-90PC89865. 1991. [198] Bagajewicz MJ, Jiang Q. Gross error modeling and detection in plant linear
[168] Farsi M, Jahanmiri A. Application of water vapor and hydrogen-permselective dynamic reconciliation. Comput Chem Eng 1998;22:1789–809.
membranes in an industrial fixed-bed reactor for large scale methanol [199] Buzzi-Ferraris G, Manenti F. Outlier detection in large data sets. Comput Chem
production. Chem Eng Res Des 2011;89:2728–35. Eng 2011;35:388–90.
[169] Rahimpour MR, Elekaei H. Enhancement of methanol production in a novel [200] Romagnoli J, Sanchez MC. Data processing and reconciliation in chemical
fluidized-bed hydrogen-permselective membrane reactor in the presence of process operations. San Diego, USA: Academic Press; 2000.
catalyst deactivation. Int J Hydrogen Energy 2009;34. [201] Manenti F, Grottoli MG, Pierucci S. Online data reconciliation with poor
[170] Manenti F, Leon-Garzon AR, Ravaghi-Ardebili Z, Pirola C. Systematic staging redundancy systems. Ind Eng Chem Res 2011;50:14105–14.
design applied to the fixed-bed reactor series for methanol and one-step [202] Kordnejad M, Shokroo EJ, Shahcheraghi M. Real time optimization of shell and
methanol/dimethyl ether synthesis. Appl Therm Eng 2014;70:1228–1237. tube methanol reactor using evolutionary and genetic algorithms. Petroleum
[171] Rachford HH, Rice JD. Procedure for use of electrical digital computers in Coal 2013;55:322–9.
calculating flash vaporization hydrocarbon equilibrium. Trans AIME [203] Biegler LT, Zavala VM. Large-scale nonlinear programming using IPOPT: an
1952;4:195. integrating framework for enterprise-wide dynamic optimization. Comput
[172] Manenti F, Leon-Garzon AR, Ravaghi-Ardebili Z, Pirola C. Assessing thermal Chem Eng 2009;33:575–82.
energy storage technologies of concentrating solar plants for the direct [204] Manenti F. Considerations on nonlinear model predictive control techniques.
coupling with chemical processes. The case of solar-driven biomass Comput Chem Eng 2011;35:2491–509.
gasification. Energy 2014;75:45–52. [205] Biegler LT. An overview of simultaneous strategies for dynamic optimization.
[173] Feng W, Ji P, Chen B, Zheng D. Analysis of methanol production from biomass Chem Eng Process: Process Intensif 2007;46:1043–53.
gasification. Chem Eng Technol 2011;34:307–17. [206] Manenti F, Cieri S, Restelli M, Nascimento Lima NM, Zuniga Linan L, Bozzano
[174] Van-Dal ES, Bouallou C. Design and simulation of a methanol production plant G. Online feasibility and effectiveness of a spatio-temporal nonlinear model
from CO2 hydrogenation. J Clean Prod 2013;57:38–45. predictive control: the case of methanol synthesis reactor. Comput Aided Chem
[175] Kim WS, Yang DR, Moon DJ, Ahn BS. The process design and simulation for Eng 2012;867–71.
the methanol production on the FPSO (floating production, storage and [207] Manenti F, Bozzano G. Optimal control of methanol synthesis fixed-bed reactor.
off-loading) system. Chem Eng Res Des 2014;92:931–40. Ind Eng Chem Res 2013;52:13079–91.
[176] De María R, Díaz I, Rodríguez M, Sáiz A. Industrial methanol from syngas: [208] Stephanopoulos G. Chemical process control: an introduction to theory and
kinetic study and process simulation. Int J Chem Reactor Eng 2013;11:469– practice. Englewood Cliffs, New Jersey, USA: PTR Prentice Hall; 1984.
77. [209] Luyben ML, Luyben WL. Essentials of process control. 5th ed. Chemical
[177] Manenti F, Cieri S, Restelli M, Bozzano G. Dynamic modeling of the methanol engineering series. McGraw-Hill; 1997.
synthesis fixed-bed reactor. Comput Chem Eng 2013;48:325–34. [210] Manenti F. Natural gas operations: considerations on process transients, design,
[178] Manenti F, Cieri S, Restelli M, Lima NMN, Lamia ZL. Dynamic simulation of and control. ISA Trans 2012;51:317–24.
the Lurgi-type reactor for methanol synthesis. Chem Eng Trans 2011;379–84. [211] Morari M, Lee JH. Model predictive control: past, present and future. Comput
[179] Parvasi P, Khosravanipour Mostafazadeh A, Rahimpour MR. Dynamic modeling Chem Eng 1999;23:667–82.
and optimization of a novel methanol synthesis loop with hydrogen- [212] Cutler CR, Ramaker BL. Dynamic matrix control{ A computer control algorithm.
permselective membrane reactor. Int J Hydrogen Energy 2009;34:3717–33. AIChE 86th National Meeting, Houston, TX. 1979.
[180] Parvasi P, Rahimpour MR, Jahanmiri A. Incorporation of dynamic flexibility [213] Cutler CR, Ramaker BL. Dynamic matrix control{ A computer control algorithm.
in the design of a methanol synthesis loop in the presence of catalyst Joint Automatic Control Conference, San Francisco, California. 1980.
deactivation. Chem Eng Technol 2008;31:116–32. [214] Qin SJ, Badgwell TA. A survey of industrial model predictive control technology.
[181] Itoh N. A membrane reactor using palladium. AIChE J 1987;33:1576–8. Control Eng Pract 2003;11:733–64.
[182] Shu G, Grandjean BPA, Kaliaguine S. Methane steam reforming in a symmetric [215] Dones I, Manenti F, Preisig HA, Buzzi-Ferraris G. Nonlinear model predictive
Pd- and Pd-Ag/porous ss membrane reactor. Appl Catal 1994;119:305–25. control: a self-adaptive approach. Ind Eng Chem Res 2010;49:4782–91.
[183] Barbieri G, Maio FPD. Simulation of the methane steam reforming process in [216] Qin SJ, Badgwell TA. Allgower F, Zheng A, editors. An overview of nonlinear
a catalytic Pd-membrane reactor. Ind Eng Chem Res 1997;36:2121–7. model predictive control applications. Switzerland: Birkhauser; 2000.
[184] Farsi M, Jahanmiri A. Dynamic modeling of a H 2O-permselective membrane [217] Rawlings JB, Mayne DQ. Model predictive control theory and design. Madison,
reactor to enhance methanol synthesis from syngas considering catalyst WI, USA: Nob Hill Pub; 2009.
deactivation. J Nat Gas Chem 2012;21:407–14. [218] Henson MA, Seborg DE. Nonlinear process control. Upper Saddle River, NJ, USA:
[185] Farsi M, Jahanmiri A. Dynamic modeling and operability analysis of a dual- Prentice Hall PTR; 1998.
membrane fixed bed reactor to produce methanol considering catalyst [219] Christofides PD. Nonlinear and robust control of partial differential equation
deactivation. J Ind Eng Chem 2014;20:2927–33. systems: methods and applications to transport-reaction processes. Boston,
[186] Rahimpour MR, Rahmani H, Bayat M. Contribution to emission reduction of USA: Birkhäuser; 2001.
CO2 by a fluidized-bed membrane dual-type reactor in methanol synthesis [220] Biegler LT. Solution of dynamic optimization problems by successive quadratic
process. Chem Eng Process: Process Intensif 2010;49:589–98. programming and orthogonal collocation. Comput Chem Eng 1984;8:243–7.
[187] Wu Y, Gidaspow D. Hydrodynamic simulation of methanol synthesis in [221] Biegler LT. Technology advances for dynamic real-time optimization. Comput
gas-liquid slurry bubble column reactors. Chem Eng Sci 2000;55:573–87. Aided Chem Eng 2009;1–6.
[188] Grossmann IE, Biegler LT, Part II. Future perspective on optimization. Comput [222] Rahimpour MR, Behjati HE. Dynamic optimization of membrane dual-type
Chem Eng 2004;28:1193–218. methanol reactor in the presence of catalyst deactivation using genetic
[189] Hillestad M. A systematic generation of reactor designs I. Isothermal conditions. algorithm. Fuel Process Technol 2009;90:279–91.
Comput Chem Eng 2004;28:2717–26. [223] Rahimpour MR, Parvasi P, Setoodeh P. Dynamic optimization of a novel
[190] Hillestad M. A systematic generation of reactor designs II. Non-isothermal radial-flow, spherical-bed methanol synthesis reactor in the presence of
conditions. Comput Chem Eng 2005;29:1101–12. catalyst deactivation using Differential Evolution (DE) algorithm. Int J Hydrogen
[191] Hillestad M. Systematic staging in chemical reactor design. Chem Eng Sci Energy 2009;34:6221–30.
2010;65:3301–12. [224] Kordabadi H, Jahanmiri A. A pseudo-dynamic optimization of a dual-stage
[192] Montebelli A, Visconti CG, Groppi G, Tronconi E, Kohler S. Optimization of methanol synthesis reactor in the face of catalyst deactivation. Chem Eng
compact multitubular fixed-bed reactors for the methanol synthesis loaded Process: Process Intensif 2007;46:1299–309.
with highly conductive structured catalysts. Chem Eng J 2014;255:257–65. [225] Fuad MNM, Hussain MA, Zakaria A. Optimization strategy for long-term
[193] Luyben WL. Design and control of a methanol reactor/column process. Ind catalyst deactivation in a fixed-bed reactor for methanol synthesis process.
Eng Chem Res 2010;49:6150–63. Comput Chem Eng 2012;44:104–26.