DOI: 10.1002/cctc.201402907

Full Papers

This is the peer reviewed version of the following article: ChemCatChem 2015, 7, 928 – 935, which has been published in
final form at http://dx.doi.org/10.1002/cctc.201402907. This article may be used for non-commercial
purposes in accordance with Wiley Terms and Conditions for Self-Archiving.

Structure, Activity, and Deactivation Mechanisms in
Double Metal Cyanide Catalysts for the Production of
Polyols
Neyvis Almora-Barrios,[a] Sergey Pogodin,[a] Luca Bellarosa,[a] Max García-Melchor,[a]
Guillem Revilla-López,[a] Miquel García-RatØs,[a] Ana BelØn Vµzquez-García,[b]
Pedro Hernµndez-Ariznavarreta,[b] and Nfflria López*[a]
the subsequent use of these polymers, in particular to foam
stability. Despite the wide industrial interest in DMCs, there are
only a few experimental studies and a complete lack of theoretical research of their structure, activity, and performance.
The present work is thus the first attempt to describe the
nature of the active site, the main polymerization mechanism,
and two potential origins for the high-weight tails from a theoretical standpoint by analyzing three crucial steps in the polymerization process.

Polyether polyols are used widely in the plastic and coating industries in the form of polyurethanes. The polymerization of
epoxides can be catalyzed by double metal cyanides (DMCs),
Zn3[Co(CN)6]2. These catalysts were first reported in the 1960s
by General Tire Inc. and provide products with excellent technical features, which are better than those that result from traditional alkaline polymerization as side reactions are alleviated.
However, DMC-catalyzed polymerization is not free of drawbacks as high-molecular-weight side products (1–3 wt %) can
form in the propylene process. These tails are detrimental to

Introduction
The polymerization of C2 and C3 epoxides to polyols has been
performed traditionally by adding a base, typically KOH, but
side reactions with isomerizations to allyl alcohol,[1] detrimental
to production, were observed. In the 1960s, General Tire Inc.[2]
described a new class of double metal cyanide (DMC) compounds, Zn3[Co(CN)6]2, which are active catalysts in this process. The crystal structure of the parent compound was reported as Zn3[Co(CN)6]2·12 H2O,[3] and the catalyst was claimed to
be heterogeneous in nature.[4] Thereafter, these materials have
been optimized by different companies and have become the
catalyst of choice for the production of polyether polyols.
DMCs are able to polymerize epoxides, namely, propylene
oxide, to lead to high-quality polymers that present the following characteristics: 1) the weight distribution is narrower than
that with base-induced polymerization, 2) the polymers exhibit
a low amount of unsaturations, and 3) the process prevents
the side reactions that are typical in alkaline polymerization.
The typical characteristics of the polyols made with DMC cata-

lysts are very high molecular weights and average hydroxyl
functionalities of 500–50 000 Da.[2, 3] As DMCs are very active,
the amount of catalyst needed is rather small.
The degree of crystallinity plays a fundamental role in the
activity of the catalyst. The higher the degree of amorphicity,
the most active the catalyst.[5] To achieve a significant amount
of amorphization, the synthesis of the catalysts besides Zn and
Co salts included complexing agents (CAs) to induce amorphicity during growth.[6] CA modifiers include ethanol, isopropyl alcohol, n-butyl alcohol, sec-butyl alcohol, tert-butyl alcohol, isobutyl alcohol, and their mixtures,[5] but also ketones, esters
amides, ureas,[2a, 7] ionic liquids,[8] and glyme (dimethoxyethane)[9] have been proposed. The reaction shows an induction period that was suggested to be caused by some degree
of ligand replacement and structural changes.[9]
Since the initial formulation, several synthetic processes
have been proposed,[10] for instance, the reaction of an aqueous solution of ZnCl2 and other secondary metal salts such as
K3Fe(CN)6 with CA. Zhang et al.[6] investigated the role of ZnCl2
and different CAs in the polymerization of propylene oxide
and described a potential mechanism for the reaction. They
found that excess ZnCl2 was ultimately responsible for the
polymerization. However, free ZnCl2 was not active by itself.[6]
Therefore, in the active catalyst, ZnCl2 was suggested to be incorporated to the lattice in the form of CNÀZnÀCl terminations, which were identified as dormant active sites. An indepth characterization was presented by Chen and Chen,[11]
which included tert-butyl alcohol and a co-complexing agent
polyether with electron spectroscopy for chemical analysis

[a] Dr. N. Almora-Barrios, Dr. S. Pogodin, Dr. L. Bellarosa,
Dr. M. García-Melchor, Dr. G. Revilla-López, Dr. M. García-RatØs,
Prof. N. López
Institute of Chemical Research of Catalonia, ICIQ, Av.
Paisos Catalans 16, 43007 Tarragona (Spain)
E-mail: nlopez@iciq.es
[b] A. B. Vµzquez-García, P. Hernµndez-Ariznavarreta
REPSOL Technology Center
Ctra. Extremadura Km. 18, 28935 Móstoles (Spain)
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cctc.201402907.

ChemCatChem 2015, 7, 928 – 935

928

 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
(ESCA), IR spectroscopy, XRD, SEM, and X-ray photoelectron
spectroscopy (XPS) experiments performed to characterize
them. From these experiments the active site was identified as
Zn2+.[10]
Typically, the polymerization is conducted in a conventional
batch reactor in which the catalyst and initiator are introduced.
Then the reactor is set to the desired temperature (usually between 70 and 150 8C) and a portion of the epoxide is added.
The sign that the catalyst has been activated is a decrease in
pressure. At this point, the epoxide is fed to maintain the internal pressure from 0.7 to 4 bar.[8, 12] The activity and induction
period with alkoxy alcohols was studied,[13] which showed
a sharp decrease in the crystallinity from the stable
Zn3[Co(CN)6]2·12 H2O if CA was included. The type of CA is a crucial factor in the polymerization, which includes induction periods and polymerization rate. Low-polarity CAs were proposed
as the best candidates for commercialization. Kim et al.[14] identified that the type and amount of oxygen in the first coordination sphere of Zn seemed to play an important role in the
polymerization as the Zn atoms are the true active sites.[15]
They also identified a fast exchange between dormant and
active sites and a reaction mechanism (Scheme 1). Moreover,
they illustrated how the coordination of different CAs has
a big effect on the activity and initiation periods.[14] Finally,
a mechanistic study that included a mathematical model for

the reaction network was proposed in Ref. [16] for the polymerization process.
Similar DMC catalysts have been suggested for the copolymerization of epoxides with CO2, which expands the catalytic
scope of DMC catalysts. For instance, sol–gel synthesis was
also applied in the synthetic process of DMC and then tested
to copolymerize styrene oxide and CO2.[17] In this case, a mechanism was put forward and the active system was identified as
a semicrystalline Zn3[Co(CN)]6 phase with excess Zn. CAs were
also investigated in semibatch copolymerizations of cyclohexene oxide and CO2.[8, 18] In all cases, the choice of CA was identified as the most important single factor to activate CO2 incorporation to the polymer. In addition, the synthesis of nanosized DMC catalysts was performed to achieve synthesis control,[18, 19] and the obtained materials were tested for hydrogen
storage.[20]
The intrinsic properties of DMC catalysts are not understood
at the atomic level, mainly because of the difficulties encountered if computational methods are applied to the study of
these catalysts. Perhaps the most important hurdle is that related to the intrinsic nature of the catalyst.[21] It is well known
that if the catalyst is prepared as a single-crystalline phase, the
DMC is poorly active and CAs have to activate the catalyst. If
a different CA is used, the degree of crystallinity of the material
is reduced, which indicates that the amorphous phase is the
truly active phase. In addition to these structural problems, the
reaction has not been well characterized yet, and the nature of
the active site still remains under debate. Herein, we present
the first theoretical study of the polymerization mechanism, as
well as the role of different CAs (Cl, OH, OEt, OtBu). Furthermore, we provide a qualitative description of the reaction network and pay special attention to the first reaction steps and
the origin of high-weight polymers with the aim to provide
a molecular understanding of this very complex process.

Results and Discussion
Crystalline structures for prototypical DMC catalysts
Amorphous materials retain the first coordination sphere and
reduce the degree of ordering in the second coordination
sphere.[22] Thus, theoretical simulations applied to the crystalline structure of DMC can help to identify the most common
coordination patterns in the amorphous structure and also
some of the medium-range structures that can provide valuable information on the active site, reaction mechanism, and
deactivation routes. Therefore, the models presented here
allow us to clarify the active site patterns needed for the reaction to occur. The DMC structure presents a hexagonal unit
cell,[23] in which the octahedral Co unit acts as a spacer for the
tetrahedral Zn polygons (Figure 1). The loss of secondary order
would only mean that the octahedra and tetrahedra would not
be as ordered as pictured here but they would still have the
same nature.
In our case, we have taken the parent compound in the
native DMC structure to reoptimize the cells. A full replacement of the Cl anions in the lattice by the CAs OH, OEt, and

Scheme 1. Reaction mechanism adapted from Ref. [14].

ChemCatChem 2015, 7, 928 – 935

www.chemcatchem.org

929

 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers

Figure 1. DMC crystal structure. Color code: tetrahedral Zn sites (green),
octahedral Co centers (cyan), C (cyan), N(blue), O (red) and H (white) atoms.
The CA in the present case is represented by the terminal red O in the OH
groups.

Figure 2. Surface energies, g, for the different CAs and the lowest two
energy facets shown in Figure 3.

OtBu were examined. Upon replacement, all the bulk cells
were reoptimized (Table 1). The structure is kept with some
minor changes to adapt the small OH or the bulky OtBu cen-

Table 1. Cell parameters and the most relevant ZnÀZn distances in the
DMC crystal with different CAs. All distances in Š.
CA

A

b

c

ZnÀZn

ZnÀZn

Cl
OH
OEt
OtBu

7.524
7.530
7.530
7.499

7.526
7.534
7.561
7.562

18.295
17.111
16.967
17.049

4.381
3.754
3.721
3.739

6.515
6.525
6.502
6.511

ters, but in all cases the separation between the Zn tetragonal
coordination spheres and the octahedral configurations of the
robust Co(CN)6 spacers is maintained. The cell parameters observed upon CA replacement are very similar to those of the
parent compound. The main changes occur along the c direction for which even a 5 % change is reported for CA = OEt. If
replaced, the shortest ZnÀZn contacts are reduced to 3.7 Š,
from an initial value of 4.4 Š, and the second ZnÀZn sphere is
maintained at a distance of 6.5 Š.
A cut along the lowest index surfaces was performed for all
these bulk materials with full substitution. For DMC, both
(0 0 1) and (1 0 0) surfaces show the lowest energies. The (0 0 1)
surface exposes Zn atoms at the external part with Zn atoms
in the outermost positions spaced by Co(CN)6 motifs (Figures 2
and 3). These layers are bonded to the next through Cl atoms.
The (1 0 0) surface is more open, and the Zn tetrahedral motifs
are exposed on the surface, however, they are not spaced by
Co as in (0 0 1) but linked by Cl (or the CA). Therefore, the
(0 0 1) surface can be described as isolated single sites, whereas
close ZnÀZn contacts appear for the (1 0 0) surface.
Surface energy calculations for the two main facets are presented in Figure 2. Our results indicate that the (0 0 1) surface,
in which Zn atoms are isolated, is more stable regardless of
the CA. The lowest energy required to create a surface corresponds to very bulky capping agents as expected from the
steric hindrance. Electronic effects also play a role, and linkers
ChemCatChem 2015, 7, 928 – 935

www.chemcatchem.org

Figure 3. Lowest energy surfaces of DMC with different CAs: a) Cl, b) OH,
c) OEt, d) OtBu. Color code: Co centers (large gray/blue) Zn centers (green),
Cl (yellow), C (cyan), N (blue), O (red) and H (white) atoms.

with oxygen fragments stabilize the lowest energy facet significantly. Relative energies, which might affect the ratio between
930

 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
isolated and close Zn centers are also affected significantly by
the CA. For instance, the equilibrium nanocrystal shows
a more extended (1 0 0) facet for EtO as the ratio between the
surface energies of (1 0 0) and (0 0 1) is more favorable.

the ability of Zn atoms to show coordination spheres that
range from four to six.[25] Although the adsorption energy of
the epoxide is weak, the high pressure of the epoxide in the
experiments (4 bar) justifies this adsorption step. Once coordinated, the epoxide is activated and one of the carbon atoms
of the epoxide suffers a nucleophilic attack by the terminal OH
group, which results in the formation of a five-membered ring
intermediate (Figure 4). This can be opened easily to give

Reaction profile for DMC catalysts
Once the most likely structure for the two surfaces was assessed, we next investigated the mechanism for the reactions
shown in Scheme 1 for both isolated Zn atoms (0 0 1) and the
close Zn surface (1 0 0). Ethylene oxide was considered as
a polymerization unit. Qualitatively, our path follows the mechanism illustrated by Kim et al.,[14] which is summarized in
Scheme 1 for the CA = OH catalyst. The reaction profiles can
be found in Figure S1, in which the data for the reactions summarized in Table 2 and the adsorption energies of incoming
epoxides are taken into account.

Table 2. Initiation (I 2) and propagation (P) steps for the DMC catalyst
with CA = OH and both (0 0 1) and (1 0 0) surfaces from coadsorbed reactants. DE is the reaction energy and Ea is the corresponding activation
barrier. All energies in kJ molÀ1.
Reaction

Surface

DE

Ea

I2
I2
P
P
I2-CA

(0 0 1)
(1 0 0)
(0 0 1)
(1 0 0)
(1 0 0)

À64
À145
À54
À57
À43

148
87
156
107
61

Initiation steps:
HOR0 ÀZnþHOR $ HORÀZnþR0 OH

ðI1Þ

HORÀZnþEO $ HOCHÀCH2 ORÀZn

Figure 4. A representation of the initial (coadsorbed reactants) and final
state of (I 2) reactions on (0 0 1) and (1 0 0) surfaces of DMC. The color code is
the same as in the previous figures.

ðI2Þ

a four-coordinated Zn with a growing chain on top, which is
the starting point for the propagation step. Reaction on both
surfaces is exothermic by approximately 60 kJ molÀ1 in the
(0 0 1) surface and close to 150 kJ molÀ1 for the (1 0 0) one. Although the reaction is hindered by a significant barrier of approximately 150 kJ molÀ1 on the (0 0 1) surface, the energy
demand is much lower on the open (1 0 0) facet (Table 2). Thus,
the (1 0 0) facet is more active that the corresponding (0 0 1)
facet.

Propagation steps:
HOCHÀCH2 ORÀZnþEO $ HOðCHÀCH2 OÞ2 RÀZn

ðPÞ

Termination steps:
HOðCHÀCH2 OÞn RÀZnþR00 OH $ EÀOHþZnÀOR00

ðTÞ

In which HORÀZn is the catalyst with the growing polymer,
HOR is the initiator, EO is ethylene oxide, and R’’OH is the terminating agent. As the whole list of reactions is complex and
some of them are performed many times with practically no
energy variation (P) we will concentrate on a few key steps to
understand the activity of the process. The similar character of
the propagation steps in chain polymerization is well established.[24]

Propagation: HOCHÀCH2ORÀZn+EO$HO(CHÀCH2O)2RÀZn
(P)
The propagation step is repeated with little variation during
most of the polymerization process. In the present work, we
have concentrated only on the formation of the dimer of the
polyol as our interest concerns only the most salient features
of this particular step. A summary of the structures and energies is presented in Table 2 and Figure 5. Again, for the propagation step, the adsorption of the epoxide in the Zn atom in
which the chain grows is slightly exothermic by 10–
25 kJ molÀ1. The propagation is exothermic if evaluated in both

Initial step: HORÀZn+EO$HOCHÀCH2ORÀZn (I 2)
The main features of the initiation step are presented in
Table 2. Epoxide adsorption is quite weak and dominated by
ChemCatChem 2015, 7, 928 – 935

www.chemcatchem.org

931

 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
tants to coordinate to Zn, we analyzed the empty levels that
correspond to Zn as they are responsible for the catalyst Lewis
acidic character. These levels are lower in energy, thus electron
transfer (or sharing) with the incoming OR groups is very effective. This observation is in close agreement with the experimental determination that Zn atoms linked to O atoms are responsible for the high activity observed in the DMC catalysts.[14] Therefore, the (1 0 0) surface is more likely to activate
the molecules and polymerize them because the number of O
groups is larger than that in (0 0 1). These results are in agreement with earlier experimental results.[12] Thus CAs that contain
oxo groups are well suited to improve the activity of the catalyst. A particular example is N-containing ligands that generate
weaker Lewis acids as the crystal field splitting generated is
smaller (for instance if SCN and OHÀ groups are compared).
This activation does not affect further polymerization steps as
in all cases the growing polymer is bound to the Zn active site
through an OR group and it is not removed from the active
site until it reaches a high weight, up to 50 000 Da. However,
the changes caused by CAs with a higher O content also have
their own drawbacks as we will see later.
Figure 5. A representation of the initial (coadsorbed reactants) and final
state of (P) reactions on (0 0 1) and (1 0 0) surfaces of DMC. The color code is
the same as in previous figures.

Routes to deactivation: Origins of the formation of highweight tails
A side problem of DMC catalysts in polyether polyol polymerization might be related to the high activity reported. Together
with the narrow polymer distribution there is a certain amount
of high-weight polymers. These tails prevent the formation of
stable foams, one of the major applications of polyols. Highmolecular-weight polymers were identified as 2–3 % of the
weight of the total sample.[12] These species have a surfactantlike effect that alters the solubility and produces the phasingout of the polyurethane polymers in the isocyanate-polyol
foam synthesis. The tail was more abundant if ZnÀOH groups
were present.[12] The same patent identifies the interaction
with a protic acid (such as acetic acid) for some time as able to
abate the problem.[12] The acid was found to replace the
ZnÀOH vibrational fingerprints with ZnÀacetate ones. In that
case, the use of a solvent that was aqueous in character was
found to be positive. The topology of site isolation is important in the patents,[26] and the presence of EO in a fraction of
the polymer (i.e., a reduction of the amount of PO) was also
claimed to be positive to avoid the formation of high-weight
tails. The nature of the tails, either hydrophobic or hydrophilic,
also had a role.[26]
The formation of such tails might have two explanations
with a covalent (bonding) origin: 1) presence of multiple active
sites and 2) the ramification of the polymer by following the
paths described in the previous section. Given the nature of
the Zn active sites in the (1 0 0) surface, we have identified two
potential routes that can alter the behavior described for
normal polymer growth. The first is explained by the fact that
the surface Zn atoms in the (1 0 0) expose the OH linker between the Zn atoms to the surface so they can be easily incorporated into the polymer. To analyze this possibility we have
investigated the possibility that reaction I 2 takes with the CA

the (0 0 1) and (1 0 0) facets. The corresponding activation energies differ significantly, more than 150 kJ molÀ1 on (0 0 1) and
again a much lower energy demand on the (1 0 0) facet, an approximately 50 kJ molÀ1 smaller barrier. As the (P) step is basically repeated in homopolymerization, the step presented here
corresponds to a generic (P) profile.
To understand the reason for the lower activation energies
found for the (1 0 0) surface, we analyzed its electronic structure (Figure 6). The calculated Bader charges show no significant differences for the surface atoms of both facets (Tables S1
and S2). As the activity is controlled by the ability of the reac-

Figure 6. Projected density of states for the two surfaces of the DMC catalyst
with CA = OH. Continuous blue lines correspond to Zn atoms on the surface,
dotted lines correspond to the bulk, and grey ones are the total. The Fermi
level is marked by the vertical black line.

ChemCatChem 2015, 7, 928 – 935

www.chemcatchem.org

932

 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
al, this chemical interconnection of the chains leads to multiple
heads that polymerize in a single chain.
There is yet an alternative potential origin for the highweight fraction polyol chains related to the physical (noncovalent) interaction between different chains that leads to chain
entangling. This process is physical and thus it can only be presented here with qualitative arguments. The polymers grow in
the isolated site Zn centers in the catalyst, which is quite diluted, and in principle they might seem as isolated chains once
the catalyst is activated. However, if step (P) is repeated a few
times the growing chains that did not interact previously
might be not isolated anymore. Indeed, the simultaneous formation of multiple molecular chains on near active sites leads
to the formation of a polymer brush with a dense interweaving
of the polymer backbones. For those, numerous intersections
of the backbone trajectories effectively create topological entanglements of the growing molecules, not easy to break apart
once the produced polymers are far from the catalyst. Actually,
estimations in the framework of reptation theory by Doi and
Edwards[27] suggest that the diffusion coefficient of the center
of mass of a polymer chain in a melt of many other chains is
proportional to DG / NÀ21À2a, in which N0 is the length of the
0
polymer chain in terms of Kuhn segments, 1 is the mass concentration of polymer, and a is approximately 0.60/0.75. Although an accurate quantitative characterization of the effect
on the discussed polymerization reaction is challenging, the
way to avoid it is clear: the larger the distance between the
active sites with polymers that are growing simultaneously, the
fewer entanglements are formed in the product (Figure 8). This
is in agreement with earlier reports.[18]

Figure 7. A representation of the initial (coadsorbed reactants) and final
state of the alternative (I 2-CA) reaction on (1 0 0) surfaces of DMC. The color
code is the same as in previous figures.

instead of the initiating agent, and the results are shown in
Table 2 and Figure 7.
As seen from the results, the incorporation of the CA to the
chain is relatively easy as the process is exothermic and the
barrier is low. The computed barrier might be exceedingly low
if we consider that the CA employed in the model is the smallest OH group (compared to typical OEt or OtBu groups). Moreover, this path shows that the (1 0 0) surface might be destabilized by the opening of the ZnÀOÀZn bridge. Therefore, the
growing polymer would have two different positions from
which the propagation steps could take place simultaneously.
Unfortunately, the identification of the CA incorporated is experimentally impossible because of the high weight of the
polymer.
In addition, the incorporation of the CA into the polymer
has an extra effect as the neighboring Zn atoms can be more
exposed to the reaction mixture and other polyols can grow in
the nascent active Zn. Thus atoms in the bulk of the material
can become active centers, which would result in a higher activity. This observation explains the activation step needed for
the catalyst to be more active. In principle, this result is positive in terms of production. The second effect corresponds to
the interference between different sites and it also affects
open surfaces in which the Zn atoms are separated by a linker
such as the (1 0 0) surface. Given the distribution of the Zn
atoms on this surface, it is possible to form a polyol with more
than one active head. We present the summary of one of
these potential paths in Table 2.
From the process described in Table 2, it can be seen that
the cross-interaction between two different sites is indeed possible. In this case, the reaction is again exothermic and the
stress of the OÀCÀCÀO chain can be accommodated easily in
the final structure. The reaction shows a slightly larger activation barrier than that reported in previous sections. However,
under reaction conditions T = 130 8C and 4 bar, epoxide pressure cannot be discarded. As we have mentioned, the specific
concentration of the nonisolated Zn sites cannot be specified
as the Zn coordination sphere and the ZnÀO bonds are both
very labile.[25] Thus, it is possible that although the ZnÀZn
atoms are not interconnected through the catalyst lattice but
happen to be physically close because of the amorphization of
the solid or the polymerization-induced breaking of the materiChemCatChem 2015, 7, 928 – 935

www.chemcatchem.org

Figure 8. Surface growth of polymer chains at low and high density of
active sites.

Moreover, the interaction with different chains will also be
affected by the polarity, pH and ionic force of the media in
which the polymer grows. In this case, whether EO or PO is
the polymerization monomer plays a fundamental role in the
degree of extension of the chain.
Design rules for DMC and modulators to prevent highweight tails
The theoretical modeling described here identifies the key factors of polyols production in DMC materials. First, the Zn
active sites are as isolated as possible and present coordination
spheres with ligands other than N. O atoms as linkers seem especially suited to adjust the properties of Zn from an electronic
point of view. However, ZnÀO structures are labile and thus
933

 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
by projector augmented wave (PAW) frozen cores,[28c, 30] and valence monoelectronic states were expanded in plane waves with
kinetic cut-off energy of 450 eV. The k-point sampling was set to
be denser than 0.66 ŠÀ1 for bulk calculations.[31] The crystal structure of the DMC was obtained from Ref. [23]. For all the CA-modified systems, all the Cl atoms in the structure were replaced by the
CA (OH, Cl, OEt, OtBu) and the bulk was reoptimized with the
same set up. The two surfaces with low Miller indices, (0 0 1) and
(1 0 0), were also investigated by a k-point density of 0.40 ŠÀ1.
These surfaces were modeled as periodically repeated slabs that
contain two layers of Co(CN)63À and four layers of Zn2+ that were
interleaved by a 10 Š vacuum. The dipole moments that arise as
a result of the asymmetry of the lattice were then removed following standard dipole corrections. To keep the stoichiometry of the
material if the slabs were cut, capping OH groups were added.
The equations for the surface energy were modified to represent
the energies from the equations: C12N12Zn4Co2(OH)2+H2O!
C12N12Zn4Co2(OH)2ÀOH(0 0 1)+1/2 H2 ; 2 C12N12Zn4Co2(OH)2+2 H2O!
2 C12N12Zn4Co2(OH)2À4 OH(1 0 0)+2 H2, which is related to water
and hydrogen reservoirs. Large and variable OH contents
have been found for several of these materials as demonstrated
by IR spectroscopy.[10, 18, 32] The surface energies were then obtained with a one side relaxation scheme as follows g =
(EunrelaxÀslabÀNbulkEbulkÀNOHEH2O+0.5 NOH EH2)/2 A+(ErelaxÀslabÀEunrelaxÀslab)/
A. In which EunrelaxÀslab and ErelaxÀslab are the unrelaxed and relaxed
slab energies, Ebulk is that of the bulk, Nbulk is the number of formula
units, NOH is the number of hydroxyl fragments, EH2O, EH2 are the reference gas-phase energies, and A is the surface area. The dipole
moments that arise because of the asymmetry of the lattice were
then removed following standard dipole corrections. A (2 ” 1) supercell was employed to describe the reactivity on the (0 0 1) facet,
and a (1 ” 1) supercell was used for the larger (1 0 0) facet. Although
in principle the unit cells might seem small, there are two technical
issues that justify our selection: 1) the Zn atoms are rather separated in an isolated site fashion and the unit cells are rather large because of the presence of spacers, 2) the polymerization takes place
in a very dense environment as the pressure of the epoxide is approximately 4 bar and the polymer grows to cover the catalyst surface completely, thus the coverage under reaction conditions is
one chain per active site. The location of the transition states was
identified by employing the climbing image version of the nudged
elastic band method (CI-NEB)[33] with an optimization threshold of
0.05 eV. The saddle nature of the transition state structures was obtained by the diagonalization of the numerical Hessian generated
by 0.02 Š steps.

are prone to be modified, added to the polymer as ligands,
and impose a large dynamicity to the active center, which
opens more active sites along the reaction of centers with
more O ligands. However, these can be more likely to interchange among the ligands and the reactant and can lead to
some degree of multiple head growth. It has been proposed in
Ref. [12] that a protic acid can alleviate the formation of highweight tails. By revising our data, we can infer the reason
behind the positive role of carboxylate acids. Acids can protonate CA, which breaks the short ZnÀZn contacts that we have
described as detrimental because of covalent interactions.
Moreover, as the CAs are labile they can be replaced by the
carboxylic conjugated bases. These linkers are also labile[25] but
they are less susceptible to be transferred to the growing
polyol because of the dihapto nature of these linkers. In addition, protons can interact with the OH groups of the polyol to
make the growing polymer more “linear” and reduce the probability of chain entangling through noncovalent mechanisms.

Conclusions
We have shown the most salient atomic features of double
metal cyanide catalysts in the polymerization of epoxides by
means of theoretical simulations based on DFT. Even if the
complexity of the system is far beyond the abilities of the
state-of-the-art modeling techniques, the investigation of three
key steps in the initialization and the first propagation steps
can shed light on the process. Indeed, although the degree of
amorphization is crucial to long-term operation, the local geometrical patterns are reasonably well reproduced in the simulations of the optimized crystals. Therefore, the most exposed
surfaces contain two types of active centers: one with two Zn
atoms at a short distance and a second for which the ZnÀZn
contacts are in the range of approximately 7 Š. The activity of
isolated atoms with a coordination of Zn depends on the coordination sphere. N-containing ligands establish a crystal field
that allows interaction with incoming epoxides, but O-containing ligands improve activity. Therefore, Zn atoms surrounded
by at least two O-containing groups are needed as active sites.
However, this configuration imposes a large degree of lability
on the structure and can cause the formation of chains with
multiple reactive elongation sites because of the chemical interference between the centers. Physical phenomena such as
intercrossing cannot be discarded based on catalyst topology
either, although this can be addressed by using surfaces with
more spaced catalytic centers. Finally, the addition of protic
acids can lead to a lower amount of high-weight tails as they
can cleave the redundant or new OH groups among Zn atoms
and smooth the lability of the catalysts. Simultaneously, proton
addition can lead to less convoluted polymers that are less susceptible to noncovalent interactions.

Acknowledgements
We would like to thank REPSOL for supporting this research.
Keywords: cobalt · density functional theory · double metal
cyanides · polymerization · reaction mechanisms · zinc
[1] S. Seneker, N. Barksby, Am. Chem. Coun. 1996, 305.
[2] a) H. R. Johnston, US3278459, 1966; b) M. Jack, US3404109, 1968; c) J.
Milgrom, US404109, 1968; d) R. J. Herold, R. A. Livigni, Polym. Prepr. Am.
Chem. Soc. Div. Polym. Chem. 1972, 13, 545; e) A. Livigni, R. J. Herold,
O. C. Elmer, S. L. Aggarwal in Polyethers, Vol. 6, American Chemical Society, Washington, DC, 1975, pp. 20 – 37; f) R. J. Herold, Macromol. Synth
1974, 5, 9.
[3] D. F. Mullica, W. O. Milligan, G. W. Beall, W. L. Reeves, Acta Crystallogr.
Sect. B 1978, 34, 3558 – 3561.
[4] B. Le-Khac, EP 19970919340, 2001.

Experimental Section
The structural and catalytic properties of DMC catalysts were investigated by DFT using the VASP code.[28] The Perdew–Burke–Ernzerhof (PBE) functional was chosen.[29] Inner electrons were replaced

ChemCatChem 2015, 7, 928 – 935

www.chemcatchem.org

934

 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
[5] B. Le-Khac, US5470813, 1995.
[6] X.-H. Zhang, Z.-J. Hua, S. Chen, F. Liu, X.-K. Sun, G.-R. Qi, Appl. Catal. A
2007, 325, 91 – 98.
[7] a) R. J. Belner, US3427334, 1969; b) M. Jack, US3427256, 1969.
[8] I. Kim, K. Anas, S. Lee, C.-S. Ha, D.-W. Park, Catal. Today 2008, 131, 541 –
547.
[9] P. T. Bowman, H. R. Hinney, B. Le-Khac, EP 0 743 093 B1, 2001.
[10] S. Chen, P. Zhang, L. Chen, Prog. Org. Coat. 2004, 50, 269 – 272.
[11] S. Chen, L. Chen, Colloid Polym. Sci. 2004, 282, 1033 – 1038.
[12] M. K. Faraj, B. Le-Khac, W. Wang, US6063897, 2000.
[13] K. Lee, E. D. Murray, L. Kong, B. I. Lundqvist, D. C. Langreth, Phys. Rev. B
2010, 82, 081101.
[14] I. Kim, J. T. Ahn, C. S. Ha, C. S. Yang, I. Park, Polymer 2003, 44, 3417 –
3428.
[15] a) X. Liu, M. Kang, X. Wang, Hecheng Xiangjiao Gongye 2002, 14, 247;
b) X. Liu, M. Kang, X. Wang, Hecheng Xiangjiao Gongye 2002, 24, 147.
[16] H. Yin, Z. R. Chen, Chin. J. Chem. Eng. 2002, 10, 218 – 222.
[17] Y. Dienes, W. Leitner, M. G. J. Mueller, W. K. Offermans, T. Reier, A. Reinholdt, T. E. Weirich, T. E. Mueller, Green Chem. 2012, 14, 1168 – 1177.
[18] I. K. Lee, J. Y. Ha, C. Cao, D.-W. Park, C.-S. Ha, I. Kim, Catal. Today 2009,
148, 389 – 397.
[19] D. Du, M. Cao, X. He, Y. Liu, C. Hu, Langmuir 2009, 25, 7057 – 7062.
[20] S. S. Kaye, J. R. Long, J. Am. Chem. Soc. 2005, 127, 6506 – 6507.
[21] N. López, N. Almora-Barrios, G. Carchini, P. Błonski, L. Bellarosa, R.
García-Muelas, G. Novell-Leruth, M. García-Mota, Catal. Sci. Technol.
2012, 2, 2405 – 2417.

ChemCatChem 2015, 7, 928 – 935

www.chemcatchem.org

[22] D. Di Tommaso, R. I. Ainsworth, E. Tang, N. H. de Leeuw, J. Mater. Chem.
B 2013, 1, 5054 – 5066.
[23] J. C. Wojdeł, S. T. Bromley, F. Illas, J. C. Jansen, J. Mol. Model. 2007, 13,
751 – 756.
[24] P. J. Flory, Principles of Polymer Chemistry, Cornell University Press, New
York, 1953.
[25] L. Bellarosa, J. M. Castillo, T. Vlugt, S. Calero, N. Lopez, Chem. Eur. J.
2012, 18, 12260 – 12266.
[26] J. J. Lear, J. F. Pazos, O. D. Sloan, 6083420, 2000.
[27] M. Doi, S. F. Edwards, J. Chem. Soc. Faraday Trans. 1978, 74, 1789 – 1801.
[28] a) G. Kresse, J. Furthmuller, Phys. Rev. B 1996, 54, 11169 – 11186; b) G.
Kresse, J. Furthmuller, Comput. Mater. Sci. 1996, 6, 15 – 50; c) G. Kresse,
D. Joubert, Phys. Rev. B 1999, 59, 1758 – 1775.
[29] J. P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 1996, 77, 3865 –
3868.
[30] T. Bucko, J. Hafner, S. Leb›gue, J. G. ngyµn, J. Phys. Chem. A 2010, 114,
11814 – 11824.
[31] M. Methfessel, A. T. Paxton, Phys. Rev. B 1989, 40, 3616 – 3621.
[32] B. Le-Khac, EP 0 700 949 B1, 1999.
[33] G. Henkelman, B. P. Uberuaga, H. Jónsson, J. Chem. Phys. 2000, 113,
9901 – 9904.

Received: November 12, 2014
Revised: December 16, 2014
Published online on February 6, 2015

935

 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim