1995 Annual Report

A Molecular Orbital Approach to Molecular Design Code Number:

S95-22

Program Leader: P.1.s:

Harold S. Freeman (NCSU)

Graduate Students:

G. C. Lickfield, M. J. Drews (Clemson), M. B. Polk, S. Kumar (Georgia Tech.), D. Hinks (NCSU) T. Chen, X. Hu (Georgia Tech.), L. Cleveland, J. Lye (NCSU)

Objectives The Objective is two-fold: 1) To demonstrate the utility of molecular orbital (MO) theory as a viable approach a) to the design of state of the art dyes, fibers and chemical auxiliaries and b) to enhancing the competitiveness of the U.S. textile industry; and 2) to pool the heretofore individual efforts and knowledge of several NTC investigators, who are interesied in the use of M.O.-based computational chemistry, to jointly set in place a fundamental approach to the design of a variety of organic compounds of commercial importance to the textile industry.

Abstract Separate modeling systems have been investigated and set up to model various properties of dyestuffs, chemical auxiliaries, and polymers at the molecular level. Preliminary results show promise, particularly for modeling the fate of photo-excited dyes, finding stable ground state dye molecular geometries, modeling PET oligomer mobility and modeling supercritical carbon dioxide mobilities. In addition, we believe that estimation of the octanol/water partition coefficient can provide a better understanding of the toxicity of certain groups of dyes, as initial work on a series of anthraquinone dyes showed a good correlation between the predicted octanol/water partition coefficient and the measured butanol/water partition coefficient. Our preliminary work constitutes a tentative first step towards the development of standard protocols to be implemented in future molecular design work.

National Textile Center Annual Report: August 1995

283

s95-22,2

Introduction Technological advances have placed powerful computers on the desk top of bench chemists, bringing inexpensive computer aided molecular modeling of physical and chemical properties into the chemistry laboratory. Examination of the pertinent chemical and biological literature shows that molecular modeling has been used frequently in the pharmaceutical industry, to find promising new drug candidates, with a great deal of success. Often, the approach is to match the molecular ‘shape’ of a compound with a sensitive region of an enzyme such that the compound will interfere with the function of the enzyme. In the case of dye structures, a subtly different approach is needed, as many of the properties of dyes depend upon the manner in which the compound interacts with light. Such interactions necessitate the use of quantum mechanical calculations. Over the years, many scientists have made contributions to quantum mechanical methods for the prediction of reaction pathways; thus, a number of algorithms have been developed that can be used to explain observations arising from studying simple compounds. The different methods may be divided into two types; semi-empirical and ab initio methods. Of the two types, semi-empirical methods are the most readily applicable to dye molecules because of the way in which the numerous orbital interactions in a dye molecule are handled with a modest demand on computer processing power. The most recently developed semi-empirical methods are known as AM1 and PM3, although many more are at our disposal. Ab initio (from the beginning) methods place a prohibitively high demand upon computing power and memory to be generally useful for the routine modeling of all but the simplest of molecules. Expedient utilization of this technology will enhance the competitiveness of the U.S. textile industry by:

0

Fast elimination of non-viable compounds for textile applications before expensive synthetic work begins (as an alternative to the traditional empirical ‘make it and see’ approach ).

ii)

Reduction of the development times for new dyes, polymers, and chemical auxiliaries - streamlined research.

284

National Textile Center Annual Report: August 1995

s95-22,3 iii)

Enhanced understanding and rationalization of quantum chemical events through visualization of such events.

iv)

Reduction of laboratory chemical inventory.

Molecular modeling technology is being applied to previously synthesized dye structures, with the intent of designing speciality dyes using the knowledge that we have gained. Specific areas of interest include the computer-aided design of lightfast dyes, nonmutagenic dyes, high substantivity dyes, and novel chromophores. Computer prediction of properties such as intermolecular interactions, electronic spectra, solubility, molecular orbital energies and shape, reaction pathways and transition states are playing a key role in the development of this aspect of the project. Utilization of systems tailored towards the modeling of the internal structure of PET fibers (Biosym and Cerius2-Molecular Simulations) enable factors such as the migration of PET oligomers in the fiber to be modeled. Finding potential docking sites of oligomers on modeled fiber surfaces which could act as nucleation points for oligomer crystal growth, represents a first step toward understanding oligomer crystallization in PET. A second avenue of exploration involves modeling carbon dioxide and oligomer mobility through the polymer matrix. Further investigations of novel polymer and copolymer blends give promise of new materials possessing novel rheological and mechanical properties.

Evaluation of various semi-empirical methods The CAChe implementations of MOPAC and ZINDO place eight semi-empirical methods at our disposal, namely: CNDO/l, CNDO/2, INDO/l, IND0/2, MNDO, MIND0/3, AMl, PM3 and lNDO/S (for spectroscopic calculations only), and we have begun to assess which method(s) are most appropriate for the modeling of synthetic dyes. The different methods are listed above in chronological order, the CNDO methods being the oldest and PM3 being the most recent and most rigorous of all of the methods. We have used these methods to correlate known, measurable physical properties of dyes, with predicted values arising from our modeling system. Such properties include color, solubility and bond distances.

National Textile Center Annual Report: August 1995

285

9X-22,4 Preliminary results involving comparisons between predicted and x-ray measured bond lengths in two dye molecules (C.I. Disperse Yellow 86 and C.I. Disperse Red 167) allow us to draw the following conclusions: Mechanical methods (commonly used in the pharmaceutical industry’s modeling

i)

effort) are inadequate for predicting the bond lengths and molecular shape of stable conformations of dyes. This is thought to be at least in part due to stereo-electronic effects, which become very important in highly conjugated structures such as dyes, and these effects are not considered in mechanical methods. ii)

AM1 and PM3 are the two most reliable methods for modeling dyes. AM1 seems

to be more reliable than PM3 in cases of dyes containing a conjugated amino group. These methods give the lowest mean errors and strongest correlation with experimental values of bond length. Figure 1 below demonstrates the correlation between experimental and predicted @Ml) bond lengths.

2.0

A

AM1 Geometry R2 = 0.9769 End Error = 0.1723

1.5

Av. Error = 1 .I 809%

1.0 1.0 1.5 2.0 Bond lengths (angstrom) from x-ray data.

Figure 1.

286

Correlation between AMI Predicted bond lengths and x-ray bond distances

National Textile Center Annual Report: August 1995

s95-22,5 As an example, Table 1 below gives the coefficients of determination for correlations between the measured (x-ray) bond lengths and those predicted by the various method for the azo dye C.I. Disperse Red 167. In addition to the eight semi-empirical methods, a mechanical method (MM2) was also included in the study. Table 1.

Summary of mean errors associated with predictions of bond lengths from various computational methods for C.I. Disperse Yellow 86. Method Used

Mean Error (%)

Method Used

Mean Error (%)

MM2

2.38 %

MIND0/3

2.37 %

CNDO/l

1.97 %

MNDO

1.97 %

cNDo/2

1.61 %

AM1

1.18 %

INDOll

1.65 %

PM3

1.16 %

INDO/

1.65 %

Figure 2 below compares the x-ray crystal structure of Disperse Yellow 86 with the minimum energy structure given by PM3, after a geometry optimization in CAChe MOPAC 6.0. Clearly, the two structures are very similar.

Figure 2.

Comparison of the x-ray crystal structure (left) of C.I. Disperse Yellow 86 with the

PM3 minimum energy structure (right).

National Textile Center Annual Report: August 1995

287

S95-22,6 This aspect of our study enables us to determine which parameterization is the most suitable for geometry optimizations of different dye classes, which is most suitable for prediction of light absorption characteristics, and to determine the dependence of geometry on the light absorption characteristics of certain dye classes.

MO evaluation of hydroxybenzotriazole stabilizers Our investigations into factors contributing to the stabilization of disperse dyes led us to begin to study the fate of benzotriazole molecules following absorption of light. As an initial experiment, it seemed logical for us to model a known and well-characterized structure such as Tinuvin 326 (l), a commercial photostabilizer:

‘CH, (1) Tinuvin 326 The results of this experiment are now being used to account for the behavior of a group experimental disperse dyes containing a built in benzotriazole stabilizer moiety (See fig-

R=H,Me

R = H, Me

Figure 4. Examples of dyes containing an hydroxybenzotriazole stabilizer group

288

National Textile Center Annual Report: August 1995

s95-22,7 Hydroxybenzotriazole photostabilizers absorb ultraviolet radiation, and are believed to dissipate this energy via a proton transfer mechanism. (see figure 5). This reaction pathway has been modeled using CAChe MOPAC.

hv U.V.

I so

+++Ce,.,

tautomerism

so

‘CH,

Figure 5. Proposed energy dissipation mechanism for hydroxybenzotriazole photostabilizers

R

R = H, CHs (2) Preliminary modeling work suggests that: The stabilizer residue (2) built into the dyes in figure 5, where R=H is susceptible to 1) attack by electrophiles at the phenyl carbon atom para to the amino linkage (see *) when the molecule is in the ground state, whereas the commercial stabilizer is not. This was ascertained by using a frontier electron orbital weighting method once frontier electron densities had been obtained from MOPAC AM1 .

National Textile Center Annual Report: August 1995

289

S95-22,8 The ground state structure of Tinuvin 326 (1) has a planar.benzotriazole ring. After 2) the molecule is excited, the central ring nitrogen pyramidalizes, thus bringing the phenolic hydrogen in close proximity to the ring nitrogen. In the excited state, the preferred conformation of Tinuvin 326 delivers the phenolic proton within 1.9 angstroms of the adjacent benzotriazole ring nitrogen atom, and excitation of the corresponding moiety in the experimental dyes (figure 5) leaves the proton about 2.1 angstroms away from the ring nitrogens. Upon studying the light-induced changes in geometry, we found, in the case of the 3) Tinuvin 326 (l), that 136 KcaVmole is dissipated by proton transfer alone. The energy difference between the ground and excited states is 28 Kcal/mole after the proton has transferred from the phenolic oxygen to an adjacent benzotriazole ring nitrogen. In the case of the stabilizer moiety (2) incorporated into the experimental dyes, the 4) total energy dissipated by the same mechanism accounts for 126 KcaVmole, but the smallest difference in energy between the ground and excited states is larger (38 Kcal/mole), suggesting that transfer from the excited to the ground state may be less efficient.

Polymer modeling studies. A silicon Graphics Power Indigo2 workstation has been purchased and installed. Two separate molecular modeling packages (Biosym and Cerius2-Molecular Simulations) are currently being evaluated in terms of model generation and predictive capabilities. The study involves the construction of (PP, PS, PET, PEO) polymer chains and generating amorphous cells from these polymers, predicting their Cohesive Energy Densities (C.E.D.) and phase diagrams for the individual polymers and their blends, and then comparing these predictions to published results.

Prediction of dye solubilities and genotoxicity For the series of anthraquinone dyes (3a-h) shown below, for which toxicity data are known, the CAChe 1ogP regression model was used to estimate octanol/water partition coefficients. These calculated values were found to correlate well with the measured butanol/water coefficient, as shown in Figure 6 below. Interestingly, our group has previously shown a correlation between aqueous solubility and toxicity of these dyes. These results highlight the relevance of solubility and partition coefficients (which we have shown we can estimate) to the toxicity of a compound. The estimated water/octanol partition coefficient is now established as a useful consideration for us when designing nontoxic dyes.

290

National Textile Center Annual Report: August 1995

s95-22,9

0

NH2

,Q ’-\

3

3

Rl

R2

0s

0

R,

OH

(3) 3g

3h

H

Cl

H

H

H

H

H

Me

Br

H

CMe,

3a

3b

3c

3d

3e

RI

H

Me

H

H

R2

H

H

Me

R3

H

H

Cl

3f

*naphthoxy instead of phenoxy

Anthraquinone dyes 3a-h

Exptl IogP

1 -2 I. 2

I

I

I

I

I

I

3

I

I

,

ID

4

Predicted IogP

Figure 6. Correlation of the predicted octanol/water partition coefSicient with the measured propanol/water partition coeflcient for dyes 3a-h

National Textile Center Annual Report: August 1995

291

s9522, 10

Migration and diffusion in PET Based on X-ray analysis, crystal structure models for the cyclic trimer of PET have been generated (Figure 7). Various 2-D PET surface models have also been generated to examine possible docking sites which can serve as potential nucleation sites for oligomer crystal growth. In addition, molecular mechanics and energy minimization are being used to calculate surface binding energies of the structures for the general classes of surfactant hydrophobic groups with both the oligomer crystals and the PET surface structures. Finally, simulation tools have been used to predict phase diagrams for binary mixtures of the cyclic trimer in various solvents. These predicted results have been correlated with published experimental results. Similar calculations are being performed using several of the surfactant hydrophobe structures. Molecular models representative of both amorphous and crystal cells of PET are being generated to examine the solubility and diffusion of supercritical carbon dioxide via molecular mechanics and dynamics. Following a recent publication work is planned involving theoretical calculations, novel polymer synthesis and physical testing.

Conclusion The necessary equipment has been installed and preliminary investigations are pav-

ing the way to developing new protocols for the design of textile related compounds. Encouraging preliminary results have culminated in two publications: 1) D. Hinks, J. Lye, H. S. Freeman, “Computer-Aided Dyestuff Design”, Book of Papers AATCC International Conference, October 1995. (in Press) 2) R. Pachter, W. W. Adams, X. Hu, M. Polk, S. Kumar, “Theoretical Calculations on rigid and semi-flexible Polymer Systems”, MRS Abstracts of the Spring 1994 meeting.

292

National Textile Center Annual Report: August 1995