The Design of Chiral Nanoporous Materials by Computer Modelling
GR/K90463
Final Report
Principal Investigator: Dr P.M. Rodger
*
, University of Reading
Co-Investigator: Dr P.C.H. Mitchell, University of Reading
Research Worker: Dr A.J. Ramirez-Cuesta, University of Reading
Overview
This project was a computer modelling study of a new class of chiral material, based on layered aluminium phosphates (AlPOs). One material in particular, [Co(tn)
3
]Al
3
P
4
O
16
(GTex2; tn is 1,3diaminopropane), was notable in that it could be synthesised with enantiomerically pure crystals from a racemic mixture of the chiral template molecule, [Co(tn)
3
]
3+
.
1
The synthesis of these materials represented an important new development, as it opened up the possibility of using specifically chiral materials to optimise a range of asymmetric syntheses and separation processes. We sought, therefore, to understand how the chirality was engendered and stabilised in the final material. Through understanding the factors involved in stabilising the chiral AlPOs, we would hope to be able to predict what chiral features could be introduced into other intercalated materials to begin to develop materials with a range of different chiral nanostructures.
We have achieved our aims. Simulations using a range of different models for the intercalate — progressing from very simple spherical models to full atomistic models incorporating hydrogen bonding — have enabled us to determine that the chiral structure is stabilised by a combination of shape-dependent packing forces within the intercalate layers and hydrogen-bonding interactions between the intercalate and the AlPO layers Without either type of interaction, a much less stable material results, and it is doubtful that the macroscopic chiral domains observed during the crystallisation of GTex2 would still occur. An important result was that the hydrogen-bonding interactions were water-mediated, with PO…H
2
O…NH
2
hydrogen bonding trimers playing an important role in determining the mechanical stability of the materials. This structural role for water is reminiscent of many biological materials. In related and subsequent work we shall develop this theme of the role of water in stabilising inorganic structures and biominerals. In summary, we have found that the chirality in these AlPOs arises from a complementarity between the chiral templates and the AlPO structures, with water molecules alleviating any mismatch between the two structural units. Packing effects between the intercalates, and the size and charge-density of the intercalates, is then critical in determining whether enantiomeric ( e.g.
GTex2), or racemic ( e.g.
GTex1: [Co(tn)
3
]Al
3
P
4
O
16
, where en is 1,2diaminoethane) crystals form.
One of our concerns when designing this research project was to relate our modelling closely to experimental data. We decided that the full vibrational spectrum of the material was a very stringent test of the quality of the modelling, and so we have measured the full vibrational spectra of these materials by inelastic neutron scattering (INS) in the TFXA/TOSCA spectrometers at RAL ISIS. Materials (those we were modelling) were supplied by Dr Angus
Wilkinson (Georgia) whose group had synthesised them. We thus developed a most successful collaboration. Our simulations have been able to reproduce quantitatively INS energy transfers and scattering intensities. Indeed, we have shown that our methods are predictive, with our
*
Current Address: Department of Chemistry, University of Warwick, Coventry, CV4 7AL
simulated INS spectrum for GTex2 being published prior to measuring its INS spectrum. This is a very important development, because it has meant that we have been able to make a much more detailed assignment of the INS spectrum than has previously been possible, and so has allowed us to maximise the information available from INS. This is despite the fact that the chiral AlPOs are complex materials for INS, with two distinct, but coupled, sources of proton motion arising from the template hydrogens and the water. From our experiments, we have been able to identify a coupling between the AlPO phonon modes and template vibrational modes that is also correlated with the water motions; this indicates that water mediates the AlPO / template coupling and thus confirms experimentally the structural role for water identified in our simulations.
Aims and Objectives
As identified in the original proposal, the main aims were (in order of importance):
(1) To use computer modelling techniques to elucidate the interactions which control templating effects in the synthesis of chiral AlPOs.
(2) To explain the different templating effects of [Co(en)
3
]
3+
and [Co(tn)
3
]
3+
on the layer structure of AlPOs — the former generates stable chiral nanopores but gives a racemic crystal while the latter merely perturbs an achiral layered structure yet forms enantiomerically pure crystals.
(3) To develop force-fields and modelling methods that describe accurately the known chiral
AlPOs and the chiral nano-environments they give rise to.
(4) To demonstrate the viability of computer design in devising new materials with tailored chiral nano-environments.
In particular, several specific objectives were identified:
(a) To model the structure and behaviour of [Co(en)
3
Al
3
P
4
O
16
.3H
2
O] ( 1 ) and
[Co(tn)
3
Al
3
P
4
O
16
.2H
2
O].( 2) ;
(b) To predict the limiting composition for which these structures will remain stable;
(c) To identify the intermolecular interactions and molecular properties responsible for the chiral templating effect; and
(d) To identify templating agents to synthesise chiral AlPOs with designed nanopores.
Results and Achievements
The project has been successful in achieving these aims and objectives. At the same time, we have been able to make important new developments in using molecular modelling to interpret inelastic neutron scattering (INS) so that the combined methodology now provides a much more powerful tool for characterising and understanding the properties of new materials. This last was an unexpected development of the simulation program, is clearly at the international forefront of
INS, and has already generated considerable interest within the INS community.
2
F
ORCE
F
IELDS
Successful new force fields were developed during this project. Two approaches to AlPO force fields were already available in the literature—both for materials of composition AlPO
4
. The first was a three-body potential with polarisability and formal charges due to Henson and
Gale.
3
Given the use of formal charges, this was in principle transferable to materials of different stoichiometry, but our early simulations showed that it did not give a good representation of the chiral AlPOs. The second was a pair potential developed by van Beest et al.
,
4
and used partial charges fitted to ab initio cluster calculations. Due to the replacement of the Al by a Co
3+
site in the chiral AlPOs, this potential was not directly transferable to our materials, but it was found that a simple rescaling of the charges was able to reproduce the experimental structure and properties with good accuracy. This potential had the added advantage of simplicity, as it did not require the use of 3-body terms and thus did not require the additional computational time
involved in examining all possible atom-triplets. Attempts were made to extend the modified van Beest potential by adding polarisability (using the shell model), but this was not found to improve the accuracy of the AlPO potential.
Parameters for the cobalt-complex templates were derived using standard Lennard-Jones parameters (CHARMm force field) supplemented with ESP-derived charges determined by fitting to the electron density for the isolated template molecule calculated using density functional theory methods (ADF and Dgauss programs). These combined potentials gave a good representation of both structural and dynamic properties for the AlPOs. Initially, a standard SPC potential was used for the water, but in view of the highly ionic environment it was found necessary to adopt a polarizable water model. In view of the key structural role found for water in these materials, it is not surprising that a high quality water potential was required. Full details of the potential are given in ref. 5; the quality of this potential can be seen from the range of properties predicted, as described below and in refs 2, 5 and 9.
6
P
ROPERTY
P
REDICTIONS
Initial tests of the various potentials were made by comparing with the observed X-ray crystal structures.
7
The final potential gave a very good representation of the crystal structure: interatomic distances were predicted within a few hundredths of an Ångstrom, and unit cell parameters were on average within 0.5% of the experimental values in constant stress simulations. The refined potentials were then tested against spectroscopic data. Our refined potential was initially found to give a good representation of the IR spectrum of these materials, particularly in the region of the P–O stretch. It was then used to predict the inelastic neutron scattering spectrum of GTex2; this prediction was subsequently tested experimentally and found to be highly accurate. The potentials were also used in a simultaneous MD/INS study of GTex1 and again found to give a good representation of the experimental data. We concluded that the potential refined within this work gave a quantitatively accurate representation of the experimental properties.
We note that the range of properties considered in this study goes beyond those normally considered in MD studies of materials. In particular, the comparison with INS is unusual, and the quality of our agreement with the INS data particularly pleasing. As will be developed below, this quality of agreement has enabled us to generate a very detailed assignment of the
INS spectra. Indeed, we have developed a much more detailed assignment than any previous work, and done this for a material with considerable complexity. This achievement is not just of scientific merit, but has important applications as well. For example, in the present study it has enabled us to provide convincing experimental proof to support the observation first made from our molecular modelling: that water plays a key role in determining the strikingly different chiral behaviour of the materials GTex1 and GTex2.
ORIGINS OF CHIRALITY
An important aim of this project was to elucidate the origins of chirality in these new AlPOs, and to identify the key factors associated with stabilising them. In this we have had considerable success. The origins of chirality were investigated by two classes of simulation. In the first, a series of different template models were used to identify which features of the template were most significant. Three levels of complexity were used: (i) a single spherical charged site; (ii) a helical model, in which the chiral shape of the template was reproduced, but none of its hydrogen bonding capabilities; and (iii) a fully atomistic model that incorporated both the packing and the hydrogen bonding interactions. We found that the first model lead to a major distortion of the crystal structure, even in constant volume simulations (which do not permit the unit cell to change). Model (ii) was substantially more successful, giving a good reproduction of the crystal structure in constant pressure simulations (which allow for anisotropic expansion of contraction of the unit cell, but no change in its shape), but a poor representation of the crystal
structure in constant stress simulations (which place no constraints upon the unit cell shape or dimensions). It was not until both the shape and the hydrogen bonding interactions were included that quantitative agreement with the experimental structure was obtained in the most demanding simulations—the constant stress simulations. It is thus clear that both the shape and the hydrogen bonding potential of the template is important in determining the structure of these materials.
In view of these findings, we undertook a detailed analysis of the hydrogen bonding sites, paying particular attention to any possible hydrogen bonding network between them. From this analysis it emerged that water plays a key role in GTex2, and is responsible for mediating the hydrogen bonding network needed to stabilise the observed structure. In particular, most of the template–framework links are actually O phosphate
…H
2
O…H
2
N hydrogen bonding chains; indeed, all water molecules were found to participate in such hydrogen bonding trimers. A consequence of this is that the water is immobilised, and so no diffusion of water is observed in
GTex2. In this context it is interesting to note that it does not appear to be possible to dehydrate
GTex2 without changing the crystal structure. Although such water-mediated hydrogen bonds were also found in GTex1 (which forms the racemic crystal), not all the water molecules were involved and the hydrogen bonding network was weaker. We note that GTex1 can be made anhydrous.
The second class of simulation was designed to elucidate the origin of the chiral behaviour in the tn-derived materials. In this case we performed a series of simulations on the native compound, and on materials in which different patterns of templates had been replaced by their enantiomers.
8
This indicated that the spontaneous chiral segregation of template molecules is thermodynamically driven. The defect energy that arises when one template in the lattice is replaced by its enantiomer is 3.0 eV. From analogous simulations with different concentrations of enantiomeric defects, we have estimated that about 0.5 eV of this energy is associated with direct template-AlPO interactions, while the remaining 2.5 eV is due to interactions between templates within the intercalated layer. The larger contribution clearly arises from packing effects between the template molecules, and is a strong driving force to ensure that each layer is enantiomerically pure. The weaker effect (0.5 eV for template-AlPO interactions) is largely due to the effect of the substitution on the water. The template is accommodated in a chiral pocket in
GTex2, and when the chirality of the template and framework are mismatched, the water is no longer able to bridge the phosphate oxygen and template amine groups effectively. Although a weaker effect, 0.5eV is still substantial with repsect to RT , and is sufficient to explain the spontaneous segregation of template enantiomers into separate macroscopic domains.
We also performed the TGA thermal gravimetric analysis of both compounds, and we found that they exhibit a very different behaviour, the tn compound (chiral) has to be heated above 100 C to remove the water molecules in the interlayer space, and then shows evidence of other decomposition. The en material (racemic) eliminates all the water at 70 C, giving extra experimental evidence of the strength of the interaction between water and layer in the chiral segregation process.
M ATERIAL STABILITY AND ANALOGUES
From our early simulations it became apparent that these layered AlPOs were not at all stable to removal of some of the template molecules. The implication is that the integrity of the materials requires the presence of the template. The question then arises whether diffusion of molecules through the interlayer region is possible. We simulated the dynamics of water molecules and found that there is no diffusion within the interlayer spaces at normal temperatures. From this it was apparent that these particular materials were unlikely to have significant applications as chiral framework materials. It is, however, quite likely that they may form important materials as support phases — engendering chirality in associated coatings by virtue of their chiral structure. Such materials may, for example, be useful as support materials for the stationary
phase in chiral HPLC. In consequence, we decided not to pursue a search for analogues of the
GTex materials, and instead turned our attention to developing the combined MD/INS methodology, which had begun to show such promise for characterising new nanoporous materials.
I
NELASTIC
N
EUTRON
S
CATTERING
Neutron scattering probe the whole range vibrational motion accessible to molecular dynamics simulations, and thus are a particularly apposite test of the validity of simulations. We have made a study of two templated aluminophosphates (GTex1 and GTex2)
9,10
with both INS and
MD. A detailed interpretation of the observed vibrational spectra is not possible based only on the neutron data alone, but that information can be obtained by performing MD simulations and calculating neutron scattering properties from the resulting atomic trajectories.
11
When present,
INS spectra are completely dominated by scattering from hydrogen atoms; however, the motion of hydrogen atoms is superimposed upon that of the molecules and framework to which they are attached, and so the INS spectrum does contain information about the dynamics of the whole material. In our application, the INS is determined by the hydrogens associated with the water and template molecules. While we have analysed the complete spectrum in this way,
10
we have also been able to extract the specific contribution from water by performing both simulations and experiments on the chiral AlPOs and on the hydrated and dehydrated template salts
([Co(tn)
3
]Cl
3
and [Co(en)
3
]Cl
3
). For GTex1 the simulations and experiments were performed concurrently, and were found to give quantitative agreement. For GTex2, the simulations were performed (and published) first, and were subsequently found to be predictive of the experiments. Analysis of the water motion showed distinct environments for each water molecule in the unit cell, with the corresponding frequencies for water librational modes being significantly higher in GTex2 (the enantiomerically pure material). These results were entirely consistent with the structural role for water observed in the MD simulations and confirmed in the TGA analysis; the corresponding differences in binding-energy for the water are associated with the chiral segregation.
Our MD simulation of INS spectra generated much interest and discussion after the results were presented in Grenoble at the ILL Workshop on modelling neutron scattering spectra
(December 1998). Although INS spectra have been simulated by MD calculations before, albeit for simpler materials,
12
there is some discussion about the theoretical range of validity of this approach. We have shown that the theory is applicable to our current systems, and are pursuing the question of general validity with colleagues at the ILL and ISIS and with Dr J-Chen Li at
UMIST. We expect collaborations to develop.
Conclusions
The project has been successful in achieving our aims and the specific objectives. Accurate force fields have been developed and shown to reproduce a range of structural and spectroscopic properties. These have then been used in model calculations to identify the key factors that stabilise the chirality in these materials. We were able to confirm experimental evidence that post-dated the original application and indicated that these materials were unstable to removal of any of the templating ions. One consequence of this instability is that the materials do not have immediate applications as molecular sieves — however, they may still be useful as chiral support materials, and so it is still of vital importance to characterise and understand the origins of the chirality in these systems.
One of the major achievements of this work has been to understand the mechanisms by which the chiral AlPO structures are stabilised. The chiral discrimination was shown to originate from a combination of the shape and hydrogen bonding sites associated with the transition metal complex used as a template in the material synthesis. More surprisingly, however, we found that water played a major structural role in stabilising the enantiomerically pure material
([Co(tn)
3
]Al
3
P
4
O
16
.2H
2
O) by forming bridging hydrogen bonds between the template molecule and the AlPO layer. Indeed, we found that a major contribution to the defect energy associated with introducing a template molecule of opposite handedness was from the consequent disruption of the O phosphate
…H
2
O…H amine
hydrogen bonded trimer. This structural role for water is analogous to that found in many biological materials, and raises interesting possibilities for alternative approaches to generating chiral nanoporous materials.
Another of the major achievements of this work has been the development of molecular simulation methods in concert with experimental INS to provide a very detailed characterisation of complex materials. In this we have extended the potential of INS well beyond its current state. We have shown how MD can be used both predictively and to provide a very detailed assignment of the spectrum. We have obtained a reliable interpretation for a nanoporous material with three identifiable chemical proton species (CH
2
, NH
2
and H
2
O), and with several distinguishable environments for each species. In the process we have shown how MD/INS can be used to obtain a great deal more information about new materials than was previously possible. The MD simulations also help us to identify the low energy phonon modes in our materials and we plan to develop this aspect of the work. In much atomistic modelling work any relation with experiment rarely goes beyond comparing calculated and X-ray crystal structures.
Our inclusion of INS spectra in our studies has provided a sensitive and synergistic interaction between modelling and experiment which we intend to pursue in further work.
References
1
Since the Co complex acts as a template around which the AlPO material is synthesised, we shall refer to the Co complexes as the template molecule and the AlPO as the framework or layer that forms around the template.
2
A.J. Ramirez-Cuesta, P.C.H. Mitchell A.P. Wilkinson, S.F. Parker and P.M. Rodger, invited paper, APS Proceedings of the Neutrons & Numerical Methods, ILL Grenoble, in press
3
J.D. Gale and N.J. Henson, J. Chem. Soc. Faraday Trans.
, 1994, 90 (20), 3175
4
B.W.H. Van Beest, G.J. Kramer and R.A. van Santen, Phys. Rev. Lett.
1990, 64 , 1955
5
A.J. Ramirez-Cuesta, P.C.H. Mitchell and P.M. Rodger, J. Chem. Comm. Faraday Trans.
,
1998, 94 , 2249
6
Due to space limitations, we can do no more than summarise our conclusions in this report and have not included the large amount of numerical data generated in this project. These data are in the process of being published and may also be accessed through the investigators.
7
D.A. Bruce, A.P. Wilkinson, M.G. White, A. Bertrand , J. Chem. Soc. Chem. Commun.
1995,
2059
8
A.J. Ramirez-Cuesta, P.C.H. Mitchell and P.M. Rodger, “A molecular dynamics analysis of a chiral and a non-chiral ALPO. The role of water in chiral segregation”, in preparation.
9
A.J. Ramirez-Cuesta, P.C.H. Mitchell, S. Parker, A. Wilkinson and P.M. Rodger, Chem.
Comm, 23 (1998) 2653.
10
A.J. Ramirez-Cuesta, P.C.H. Mitchell and P.M. Rodger, submitted, Phys, Rev. Letts
11
A.J. Dianoux, J.L. Sauvajol, G.R. Kneller and J.C. Smith, Journal of non-crystalline Solids,
472, 1994.
12
G.R. Kneller, W. Doster, M. Settles, S. Cussack and J.C. Smith, J. Chem. Phys., 1992,