Pathways in Rotaxane Molecular Shuttles

(1)

Bachelor Thesis Chemistry

Pathways in Rotaxane Molecular Shuttles

by

Stan Papadopoulos

26 August 2017

Studentnumber

10722718

Research Institute

Van ’t Hoff Institute for Molecular Sciences

Research Groups

Molecular Photonics &

Computational Chemistry

Supervisors

Prof. dr. A.M. Brouwer &

Dr. ir. B. Ensing

Daily Supervisor

A. Tiwari MSc.

(2)

Acronyms

bam benzylic amide macrocycle CMD Constrained MD CV collective variable DABCO 1,4-diazabicyclo[2.2.2]octane FF Force Field FPT Freeze-Pump-Thaw MD Molecular Dynamics ni naphtalimide

PES Potential Energy Surface PrCN butyronitrile

RESP Restrained Electrostatic Potential succ succinimide

(3)

Abstract

A rotaxane is a molecular machine consisting of multiple docks on a chain around which a macrocycle is trapped by the bulky end stations, allowing only translational motion, which in this specific case happens upon excitation. A computational study was performed to probe the shuttling mechanisms in hydrogen-bonded [2]-rotaxanes followed by a short experimental study.

The computational part was conducted with Molecular Dynamics, because it allows us to follow the system over time making it possible to observe the shuttling, of which the mechanism likely changes with chain length. The longer the chain, the more flexible and easily bent it can become whereas a short thread cannot bend much, therefore not allowing for harpooning in a small system; the focus was on C5 and C16 - the rotaxanes as depicted in Scheme 1 with the number being the amount of carbons between the two stations - due to time constraints. From equilibration simulations it is shown that in the neutral state the ring favours the succinimide station and upon excitation followed by charge transfer this preference switches, meaning that the ring is now docked at the naphthalimide station. Moreover, in the charged/excited state the thread is mostly encountered in a bent configuration, because the ring can hydrogen-bond to the succinimide station in addition to the naphthalimide station to which it is already bonded.

Free energy plots also showed that in the neutral state the preferred station is the succinimide one, which is supported by experiments. The energy curves from the charged state gave mixed results: for C5 the favoured dock changed from succinimide to naphthalimide, but in C16 nothing changed significantly, which might be due to the end-to-end restrictions of the constrained runs that inhibited bending and optimal formation of hydrogen bonds. Furthermore, all barriers derived from the energy curve were approximately 4-5 kcal/mol, which equals the strength of two weak hydrogen bonds.

In the simulations, two mechanism of shuttling were observed. The random-walk happened in C5 and in C16 the shuttling occurred via harpooning. From this we might assume that the mechanism changes depending on the length of the chain.

(4)

Populair wetenschappelijke samenvatting

Rotaxanen zijn moleculaire systemen die arbeid kunnen verrichten, in het geval van rotaxanen is dit via een shuttlebeweging. Ze zijn te vergelijken met een halter met een ring die om het handvat heen zit en niet kan ontsnappen door de grote uiteindes. De ring kan hierom alleen maar heen en weer bewegen tussen beide uiteindes. Aan deze uiteindes bevinden zich stations waar de ring kan ankeren. Normaliter is de ring altijd gebonden aan ´e´en station, maar wanneer het andere station negatief geladen wordt, is het gunstiger voor de ring om daar te ankeren, waardoor de ring zich naar het negatief geladen station beweegt. Een toepassing zou een moleculaire schakelaar zijn en afhankelijk van waar de ring zich bevindt kan je zeggen dat de schakelaar ”uit” of ”aan” staat.

De vraag is echter, wat voor mogelijkheden heeft de ring om zich langs het handvat te verplaat-sen? Deze kwestie wordt onderzocht met behulp van een computationele methode die berekent hoe het molecuul zal bewegen. Hierdoor kunnen we ons systeem in de tijd volgen en observeren hoe de shuttlebeweging verloopt. Als eerste is een ”random walk” voorgesteld, wat inhoud dat de ring zich willekeurig heen en weer over het handvat beweegt en toevallig bij het andere eind terechtkomt. Een nieuw voorstel is gedaan die zegt dat de beweging vergelijkbaar is met het werpen van een harpoen en het touw waar aan het vastzit terugtrekt. Bedoeld wordt dat de twee uiteindes naar elkaar toekomen en dat de ring bindingen kan maken met het andere station waar het aan vast blijft zitten en het initi¨ele station, dat zich terugtrekt, verlaat. Uit de berekeningen blijkt dat, ongeacht de grootte van de rotaxaan de ring altijd gestationeerd is bij het beginstation, maar dat na het toevoegen van lading de ring aan de andere kant van de halter is geankerd. De manier van bewegen verschilt wel afhankelijk van hoe lang het handvat is. Bij de korte handvatten verloopt het via een ”random walk”, maar bij het grotere systeem werd geobserveerd dat het shuttlen gebeurde volgens het harpoenmechanisme.

(5)

1 Introduction

Molecular machines have recently been a topic of great interest, as the 2016 Nobel Prize in Chemistry was awarded to Ben Feringa, Jean-Pierre Sauvage and James Fraser Stoddart for the design and synthesis of these machines. They are a class of molecules that consist of at least one component which produce quasi-mechanical work in response to (photo)chemical stimuli.1Numerous of these nanomachines have already been synthesized due to the recent developments in supramolecular chemistry, but this research will only be about one type of machinery, the shuttle based on rotaxanes. The first synthetic shuttle was discovered by Stoddart et al., which was based also a rotaxane.2It is a mechanically interlocked molec-ular architecture in which a dumbbell shaped molecule is threaded through a macrocycle (Scheme 1). The ring is trapped, since both ends of the thread consist of docks, end-groups which are larger than the diameter of the macrocycle and to which the ring can bond, thus allowing only translocation and pre-venting dislocation. The applications for this process are particularly relevant to nanotechnology if used as a molecular switch for nanoscale electronic components, as it has been shown that a rotaxane-based device can be used to store data by electrochemically switching the states of the rotaxane.3

Rotaxanes based on hydrogen-bonding interactions (Scheme 1) have been studied for a considerable time.4–7The benzylic amide macrocycle (bam) is usually docked at the succinimide (succ) station, be-cause it can nestle itself between two peptide moieties to form four hydrogen bonds. Upon photoinduced excitation of the system an electron is donated by DABCO, which creates the naphtalimide (ni) radical anion thereby changing the relative binding affinity, because the electron density on the carbonyl moi-eties will be increased as a result. The macrocyle then shuttles to the charged station and docks there. After charge recombination, the ring will move to its original station.

Scheme 1. Reaction scheme of the molecular shuttle. In the stable (top left), the system is activated by a photochemical electron transfer, producing the ni radical anion (green). The bam ring (red) then leaves the succ station (blue), to which it is bound initially and binds to the ni radical anion.

(7)

adjusting the properties of the docks, it is important to understand the mechanism of the translocation. The first proposal was that the ring unbinds from its initial succ station and then moves along the thread via a random walk until it encounters the reduced ni, which then traps the ring.4 However, an alternate mechanism should be considered in which the ring makes one or more hydrogen bonds to the final station before it unbinds from the initial one.6 If the random walk mechanism is correct the shuttling rate should be independent of the acceptor station and only depend on the length of the alkyl chain, since the ring does not necessarily know what this station looks like, but it does take longer to reach the other site if the thread is longer. However, it has been proven that the acceptor station does have an impact in the shuttling rate, pointing towards the fact that the random walk might not be the actual shuttling mechanism.6

The alternate pathways will be investigated for different lengths of the hydrocarbon chain by means of Molecular Dynamics (MD) simulations. Different chain lengths will likely influence the preferred mechanism of shuttling, because the end stations can more easily approach each other by bending if the thread is somewhat longer, thus promoting harpooning. For a short chain this will not be energetically favourable, so most likely a random walk will happen there. If the alkyl chain becomes too long, though, the odds of the two station coming together will be small, since an incredible amount of conformations is possible relatively few of which composed of a bent state, and in this case the shuttling would comprise of two parts: the first will be the ring leaving the station and then shuttling will occur via harpooning.

In addition, the activation energies for the shuttling will be determined experimentally by means of Transient Absorption (TA) measurements, which is a type of time-resolved spectroscopy. A pulsed laser promotes a fraction of the molecules to an electronically excited state and a weak probe pulse is sent thereafter with a certain time delay.8 A decay curve can be obtained by plotting λmax versus the

time. The Eyring equation could then be used to calculate the energy barrier with the rate obtained from experiments.9

2 Computational Methods

Completely soluted systems as large as the [2]rotaxanes cannot be studied using ab initio MD, which uses DFT, which is why Force Field MD simulations are necessary as they are computationally less demanding. With Force Field MD the equilibrium and transport properties of a classical many-body system are computed.10Classical is meant in the sense that the assumption is made that the atoms obey the laws of classical mechanics.

To calculate all these properties Newton’s equations of motion are solved until the time averages of these have equilibrated.10 The atoms in the system each have an initial position and velocity and from those and the Force Field (FF), forces can be calculated. These forces induce an acceleration with which new velocities can be calculated; these can be used to determine the new positions (Scheme 2).

(8)

Scheme 2. Visual concept of a MD algorithm. The arrowhead points towards the physical quantity that can be calculated using the quantity/quantities displayed at the origin of the arrow.

A Force Field is one of the main necessities of MD and is a set of functions and parameters, derivable from experiments or quantum mechanical calculations, which describe all molecular interactions to calculate the Potential Energy Surface (PES) (Equation 2.1).

V =XVbonds+ X Vangles+ X Vtorsion+ X Vvdw+ X Vcoul (2.1)

The first three terms of the equation denote the bonded interactions and the latter two are nonbonded ones. A great number of FF’s have been made, most optimized for specific systems and the one used in this study is an OPLS FF, which is optimized for liquid simulations and since it has been used in several other computational studies on rotaxanes.11–14

The simulations were set up by creating a topology file of an acetonitrile solvated cubic box with one rotaxane inside in which all the parameters are defined according to the OPLS FF with the Desmond package.15 This data file was converted to a topology file usable by LAMMPS, another MD package which was used to carry out the simulations.16 _{LAMMPS was chosen over Desmond, because it can}

be combined with PLUMED, which allowed us to bias our simulations and perform Constrained MD (CMD).17

Figure 1. Naphthalimide molecule as used in the Gaussian charge calculations.

(9)

a topology file needs to be created with extra charge on the naphthalimide. It was necessary to separate the station from the rest of the rotaxane for the charge calculation, because in reality the charge is localized only on that part. The only downside to this is that an extra hydrogen needed to be added to comply to covalency which was lost due to separation and this atom will also obtain charge. That charge was put on the carbon which was attached to the supplementary hydrogen. The calculations were done with the Restrained Electrostatic Potential (RESP) method, which comprises of two steps: first, the electrostatic potential is calculated with DFT, followed by assigning partial charges to every atom to best reproduce this DFT potential.18The charges were then replaced in a copy of the topology file from the neutral systems to create the topology files for the charged rotaxanes.

With these converted files as a basis for the calculations, NPT-simulations were first performed to determine the optimal box size by keeping the amount of particles, pressure and temperature constant, while varying the volume of the box (Appendix 1). Once the box size converges it is considered equi-librated and the average of all the sizes from that point onward is chosen as the optimal box size. We then chose the frame with a box size closest to that average and used that as the starting frame for the NVT-simulations in which the box size is kept constant instead of the pressure. This method is compu-tationally less demanding, since the box size is kept the same, if not additional equations of motion are needed to take the periodic boundary conditions into account as is the case with NPT.

NVT-simulations are performed to let the system sample the PES and observe what changes occur in the configuration of the rotaxane, but since the photoinduced translation of the macrocycle to the second station occurs with a rate of 1.35 · 106s−1and the simulations are conducted on nanosecond scale it is considered a rare-event, which is why CMD is used.4Moreover, CMD is used to create a free energy profile, which is not possible with regular MD, because there is no control over the reaction coordinates. With CMD a collective variable (CV) is used to bias a simulation by applying a restraining potential, which is achieved by constraining the output of the CV, a mathematical expression of an observable, to a certain value.19 I =X i∈ chain atoms i · w(i) (2.2) w(i) = e −γ·r(i) P je−γ·r(j) (2.3)

For this project a new CV was designed as shown in equation 2.2, enabling us to track the position of the ring on the chain in which the lowercase i represents every atom defined in the chain and w(i) is a weight function which increases in value if the atom i is closer to the ring. The weight function is defined as a fraction in which in the numerator is an exponential term specific to one atom and in the denominator a summation of those exponential terms for all atoms, resulting in a relative weight. The r(i)term in the weight function is the absolute distance of the atom to the geometrical centre of the ring, so if that distance is closer the weight will naturally be higher; the other term in the exponent, γ, is the weight factor, which influences how much the weight changes per distance unit and for this project 3 was benchmarked to be an optimal value. The output I of this CV is the index of the ring, which is equivalent to the number assigned to an atom on the chain, but the index can also have a non-integer

(10)

value, meaning that the centre of the macrocycle is between two atoms. For instance if the ring is exactly between index 12 and 13, the weights of those indices will both be 0.5 while the rest of the indices will have a weight of approximatily 0; the resulting summation will be 12 ∗ 0.5 + 13 ∗ 0.5 = 12.5, which is the correct index.

< Fconstr>= 1 n − a n X i=a κ(Ii− Iconstr) (2.4) ∆A = Z I 1 < Fconstr> dI (2.5)

With the CV implemented in PLUMED, multiple CMD simulations were set up for C5 and C16, both in the neutral and charged state and with an end-to-end constraint to create the PE curves along the reaction coordinate. In every run the CV was constrained to one integer index, ranging from 1 to n + 6 with the first integer index the nitrogen of the succ station closest to the biphenyl moiety and the final one being the carbon before the nitrogen of the ni station. The timespan of each simulation was 10 ns, which was assumed to be long enough for the forces to be equilibrated. This fact was made visible by plotting the indices versus time to verify if they were fluctuating around the set value (Appendix 2). From those indices the average force during the simulation can be calculated as described in equation 2.4 with a force constant κ of 100 kcal/mol/ ˚A2. Then, with integration according to equation 2.5 the PE curves can be created. In some cases, such as C5 and C5-1, the resolution of these curves was not sufficient around the minima, which is why CMD runs with broken indices around the minima were performed to smoothen the graph.

3 Experimental Methods

3.1 Procedure

A UV/VIS-spectrum was measured of a 100 mM C5 solution in butyronitrile (PrCN). 1,4-diazabicyclo[2.2.2]octane (DABCO) (113.60 mg, 1.01 mmol) was then dissolved in PrCN (10 mL) and diluted three times before adding it (1 mL, 33mM) to the C5-solution (2 mL). Another UV/VIS-spectrum was measured. The Freeze-Pump-Thaw (FPT) procedure was executed 3 times on that solution and then another five times after adding the stirring bar. Again an absorption spectrum was measured. Finally TA measurements were conducted and another UV/VIS-spectrum of the same sample was mea-sured afterwards to check for degradation.

The same procedure was executed for the C32 system, except that the DABCO solution was reused, instead of prepared again.

(11)

4 Results & Discussion

4.1 Computational

Four variants of the rotaxane were studied using MD, but only the smallest two (C5 & C16) were continued with, because the larger two would take too long to fully analyze due to the sheer simulation time. The numbers refer to the number of carbons on the chain between the two stations.

4.1.1 NVT-simulations

Figure 2. Histogram of the distance (R) between the stations, defined as shown in Figure 3, and the respective distance of the ring to each station as obtained from a NVT-simulation of C5 without a bias potential.

(12)

From the NVT-simulations histograms can be obtained like in Figure 2 to quantify certain observables of the calculation. The three curves correspond to the distance between the geometrical center of the ring and the ni station (taken as the nitrogen), the ring and the succ station (taken as the first nitrogen with respect to the diphenyl moiety) and the distance between the two stations (same atoms are used) (see Figure 3 for indications). An intuitive choice for the succ station would be the center of the two middle carbon atoms, but the geometrical center of the ring is sometimes also situated left of that. A histogram created with those definitions would result in a cut-off graph, whereas defining the succ station as the first nitrogen does result in a histogram which nears 0 at the beginning. Instead of a normalized histogram as is usually the case, it was opted to produce one with the frequencies, since this would actually quantify what configuration is preferred from the area under each curve. Because these distances are calculated from single atoms, one should carefully interpret the position of the ring. Furthermore, at larger distances the frequencies for the bam-station distances are non-zero, which is due to the periodic boundary conditions. If the rotaxane crosses the boundary, the ring and opposing dock are suddenly situated at the edges of the box and are ’far apart’. This issue is encountered in all histograms. It can be concluded from Figure 2 that after equilibration the bam ring is docked at the succ station and not at the ni one. Furthermore, the end-to-end distance (blue line) shows two bands, one representing the stretched state and the other the bent state (Figure 4 & Figure 5 respectively).

Figure 4. C5-rotaxane in a stretched state obtained from a 20 ns NVT-simulation. The hydrogens are omitted for clarity.

Figure 5. C5-rotaxane in a bent state obtained from a 20 ns NVT-simulation. The hydrogens are omitted for clarity.

(13)

Figure 6. Histogram of the distance (R) between the stations, defined as shown in Figure 3, and the respective distance of the ring to each station as obtained from a NVT-simulation of C5 with the negatively charged ni (C5-1) without a bias potential.

When Figure 6 is compared with Figure 2, it is apparent that after equilibration, the tables have turned, as expected, meaning that the ring is now docked at the ni station. Moreover, if the succ-ni distance is compared, it becomes clear that in the charged state the thread slightly favours a bent configuration. This is likely due to the fact that that the ring now has 4 free carbonyl groups that can make hydrogen bonds with the hydrogens of the succinimide station by bending the thread a bit. As seen in Figure 2, the reverse situation will not occur as much in the neutral state, because the carbonyl groups of the naphthalimide are not strong enough as hydrogen bond acceptors.

(14)

Figure 7. Histogram of the distance (R) between the stations, defined as shown in Figure 3, and the respective distance of the ring to each station as obtained from a NVT-simulation of C16 without a bias potential.

Moving on to C16, from Figure 7 it is clear that it is entirely different from C5 after equilibration regarding the configuration of the thread. The ring is docked nicely on the succ station, but the end-to-end-distance is a wider spread, which shows that the thread explores many configurations without a clear preference for either a bent or stretched state. This distribution of configurations changes when the rotaxane is reduced, which is shown in Figure 8.

(15)

Figure 8. Histogram of the distance (R) between the stations, defined as shown in Figure 3, and the respective distance of the ring to each station as obtained from a NVT-simulation of C16 with the negatively charged ni (C16-1) without a bias potential.

There is now a minor preference for the bent configuration in the excited state, which becomes apparent from the fact that there is now only one distinct peak of the succ-ni distance at a small value, even though it is still able to extend itself. This shift in tendency is most likely because the carbonyl groups of the ring are free to make hydrogen bonds with the succ station while being docked at ni, forcing a bent structure. A similar occurrence will not be extraordinarily stable in the neutral state, because that would mean the the carbonyl moieties of the ni station would have to make hydrogen bonds. As is known, in the neutral state, it is not a strong hydrogen bond acceptor, thus the ring will not have a definitive preference to be bonded in such a way and form a bent configuration.6

4.1.2 Constrained MD

As explained in Section 2 CMD simulations were performed on C5 and C16 in the neutral and the charged state. For C5(-1) 10 ns simulations were carried out with each simulation constraining the ring along the atoms on the chain, the indices of which are shown in Figure 9; the same applies to C16(-1) (see Figure 12 for labeling). During each of those simulations the output of the CV, which is the index, is tracked during the entire simulation to determine if the forces have equilibrated (Appendix 2), which is necessary for creating the free energy plots that were constructed using the method descript in Section 2.

(16)

In all simulations the end-to-end distance was fixated at a distance the chain would be stretched because of time constraints of the project. This would increase the equilibration of the forces put onto the ring, as this would decrease the interactions the ring could have with both stations if the chain was bent, thus preventing extra external forces.

Figure 9. The index labeling of C5 as used in the CMD simulations. Each index represents one 10 ns CMD run in which the geometrical center of the ring is constrained to that index.

(17)

Figure 11. The free energy curve as obtained from the 13 CMD runs of C5-1.

Figure 10 is a prime example of a free energy plot in the neutral state, with the lowest minimum around the indices of the succinimide and the other, somewhat higher, minimum at the indices of the naphthalimide.5,20 One intriguing aspect of it is that the barrier for the shuttling even in the neutral state is approximately 4 kcal/mol, which is in the range of the breaking of a hydrogen bond.21,22 This implies that shuttling could happen with thermal fluctuations around room temperature even while not being excited, but this has not been reported yet as it is difficult to observe. It has been reported for similar systems that the barrier is approximately 3 times larger.4A reason would be that the restriction on the end-to-end distance is decreasing the thermodynamic barrier, since the entropy becomes lower as a result of a lower amount of possible configurations.

Upon reduction of the ni station the energy surface (along the reaction coordinate) changes in a way that the ring now favours the charged moiety in the sense that the minimum at the naphthalimide is lower in energy than the succinimide, but the barrier height is retained. The energy required to go back to the succ station is approximately 2 kcal/mol higher, but should not be adequate to withhold the macrocyle to shuttle back. Therefore it is likely that the end-to-end restriction distorts the energy landscape considerably. The reason being that the rotaxane cannot explore all its configurations leading to the fact that the ring cannot hydrogen bond to the opposing dock, which in turn would influence the forces.

(18)

Figure 12. The index labeling of C16 as used in the CMD simulations. Each index represents one 10 ns CMD run in which the geometrical center of the ring is constrained to that index.

(19)

Figure 14. The free energy curve as obtained from the CMD runs of C16-1.

The energy curve of C16 (Figure 13) looks similar to that of C5 (Figure 10) albeit more stretched out. Also in this case the barrier has a low value of 5 kcal/mol, enabling a shuttling motion rather effortlessly, but once on the other side the system will be in a metastable state due to the ni station not being a well defined minimum, because the barrier to go back is 1 kcal/mol; upon the ring’s arrival at the other side it should move back almost instantaneously to the succ dock. Moreover, there is a tendency to always go back to the succinimide, since the top of the barrier resides closer to the naphthalimide, allowing the ring to have more possibilities to move backwards.

In the case of the reduced C16 the energy plot barely changed (Figure 14), but the minima became less broad. Furthermore, in the proximity of the ni station (index 20) an extra peak has arisen, which distorts the smooth surface. No oddities were encountered in the trajectory and the forces were equili-brated. One thing to note is that a one dip in the index occurred (Figure 27, but since the calculated force is an average calculated over 106 frames obtained from a 10 ns simulation and this distortion equates only to 2% of the entire run the force should not be influenced significantly. Finally, no switch in dock-ing preference has been observed after adddock-ing the charge. Because of the histogram (Figure 8) it is valid to say that the charges were added correctly, since the ring shuttled to the other side via harpooning and stayed docked there. The same reason as C5-1 likely applies to C16-1 as well: due to the constraint on the end-to-end distance the state obtained from CMD is strained too much and is not as stable as it could have been without the restrictions. However it is possible for the bam ring to be docked at the reduced niin the stretched state proven by Figure 8, showing that the ring is docked at the ni station while also spending a considerable amount of time in stretched states.

(20)

4.1.3 Mechanism of shuttling

To probe the shuttling mechanism of the rotaxanes all these simulations were performed. The tendency to shuttle upon reduction is confirmed by the histograms and the energy plots (except the free energy curve of C16-1), but these are derived from equilibrated states and do not provide information about the mechanism. Only during the NVT-simulation of C5-1 and C16-1 shuttling was observed, C26-1 and C32-1 remained unchanged. The fact that in the larger systems no translocation took place during the equilibration simulation is understandable, since the simulation time was in the order of nanoseconds while in reality the shuttling takes microseconds.4Within the simulations the shuttling can be regarded as a rare-event, thus it is more likely to occur in the smaller systems, because the amount of configurations are less decreasing the entropic factor.

It might be concluded that the shuttling mechanism changes depending on the length of the thread. Shuttling in C5 went according to the random walk, likely because the thread is too short to bend without which harpooning cannot occur. In C16 first the bam ring left the station after which the thread assumed a bent configuration which resulted in harpooning. From these simulations it might be assumed that shorter chains prefer shuttling via a random-walk mechanism, whereas longer chains prefer harpooning.

4.2 Experimental

To study the shuttling of the hydrogen-bonded [2]rotaxanes TA measurements were conducted on the C5- and C32-systems. These had already been synthesized and isolated, thus they could immediately be dissolved in butyronitrile. This solvent was used first, instead of acetonitrile, which was used in the simulations, because the larger systems dissolve better in it. UV/VIS-spectra were taken after the solvation to determine if the correct transitions could be observed and if the absorbance (∼1.5) at λmax

(353 nm) was suited for the TA. This would indeed be the case after the addition of DABCO, because then the rotaxane sample would be diluted 1.5 times, thus decreasing the absorbance by a factor of 1.5 according to Lambert-Beer’s law. DABCO was added to act as an electron donor during the shuttling process to create the ni·− radical anion, which causes that station to be a strong H-bond acceptor.4The UV/VIS-spectrum of the solution after DABCO was added was similar to the previous one, but the absorbance had decreased to approximately 1. To prepare the sample for TA the solution needed to be degassed, since a radical is formed during the shuttling and if molecular oxygen (a biradical species) would be present, it would quench the triplet state and oxidize the radical anion if formed, thus inhibit the shuttling. After this step another UV/VIS-spectrum was measured to verify if nothing was out of the ordinary because of the FPT; this was the case.

(21)

Figure 15. Decay curve of the population of the ni radical anion in C32

From the TA-spectra (Appendix 3, Figure 28 & Figure 29) it would be possible to deduce the trans-lation by the shift of λmax over time, which has been proven to happen on a microsecond scale.4,23

However, if the corrected absorption at λmax is plotted a decay curve of the population of ni·− is

ob-tained(Figure 15), which shows that the population decreases much too rapidly for shuttling to occur. Moreover, the original shape of the spectrum is lost after multiple timesteps, which indicates that the sample might not be photostable anymore or has degraded, as it as an old sample. Indeed such was the case, since another UV/VIS-spectrum was taken and the absorbance at λmaxhad decreased.

5 Conclusion

A study has been done on the shuttling mechanism of hydrogen-bonded [2]rotaxanes. First of all, from the NVT-simulations we can conclude that in C5 and C16 naphthalimide becomes the favoured sta-tion after reducsta-tion as expected. Furthermore, the chain spends a considerable amount of time in bent configurations due to more hydrogen bonds being available if the ring is docked at naphthalimide, since hydrogen-binding with two stations is now available, which was not the case in the neutral state, because the hydrogen bonds with the naphthalimide were not strong enough.

Secondly, a collective variable has been developed to track the position on the ring, which allowed us to perform Constrained Molecular Dynamics on the rotaxane. Using this, free energy curves were obtained and in C5 confirmed the shift in preference; in C16 this could not be confirmed as the energy minimum at the naphthalimide dock in the reduced state did not become lower than the one at the succinimide station. The reason for it being the end-to-end distance restraint, which limited the hydrogen bonding. This was necessary for this project, however, because the project duration was a limiting factor and this approach would be the fastest way of converging the forces, which is necessary for producing an accurate energy plot. The barriers of shuttling could also be determined from the energy plot and they were rather small. In all cases (C5 & C16, neutral & charged) it was approximately 4-5 kcal/mol, which

(22)

is equivalent to two weak hydrogen bonds, and implies that shuttling is possible even in the neutral state due to thermal fluctuations even though barriers thrice as large have been reported in previous experiments. It is possible that the fixation of the end-to-end distance distorts the energy surface too much.

Next, the mechanism of the shuttling is shown to change depending on the chain length. Upon charge addition to the naphthalimide station a random walk mechanism was observed for C5, but in C16 a mixed mechanism took place. First the ring explored the energy landscape with a random walk when leaving the initial (succinimide) dock, continued by the chain bending, which allowed to translation to finish via harpooning. Thus shorter chains likely prefer random walk, whereas the longer the thread becomes, the more favourable harpooning gets until a certain length after which a mix will dominate the translational motion.

Finally, transient absorption measurements were also performed on C5 and C32 in butyronitrile. These were unfortunately inconclusive as the naphthalimide radical anion decayed within 1 microsec-ond, disabling the shuttling. This might be due to degradation of the rotaxanes as the samples were quite old.

6 Outlook

As mentioned in the beginning of the discussion, the two largest systems (C26 & C32) still need to be analyzed using constrained MD in order to create a potential energy curve, which was not done due to time constraints. In addition, the constrained MD which was performed had an end-to-end restriction, which is of course not natural. The next step would be to do the constrained MD without the fixation of the end-to-end distance. Another computational method would be metadynamics, which is another way to explore the energy landscape and is able to find minima along the reaction coordinate of the CVs. Finally, the original purpose of this project was also to compare the computational data to experimental data, such as the activation energies. To further validate the calculations it would be advised to redo the experiments for all systems, preferably in acetone as this was the solvent used in the simulations.

7 Acknowledgements

I thank Fred Brouwer and Bernd Ensing for allowing me to do a joint Bachelor Project, which made it even more interesting. I also want to thank them for the valuable discussions and talks. Furthermore, I want to thank Ambuj Tiwari for the daily supervision and for being a really nice guy who was always available for questions and problem solving. Next, I want to thank Michiel Hilbers for helping with the experimental/technical side of the project. Additionally, many thanks to Ferry, Rhea and Tamika for all the time we’ve spent helping each other out. Finally, I want to thank both the Computational Chemistry and Molecular Photonics Groups for the nice atmosphere, which made my time working on this project really enjoyable and made the time fly by.

(23)

8 Bibliography

[1] Ballardini, R.; Balzani, V.; Credi, A.; Gandolfi, M. T.; Venturi, M. Acc. Chem. Res. 2001, 34, 445–455.

[2] Anelli, P. L.; Spencer, N.; Stoddart, J. F. J. Am. Chem. Soc. 1991, 113, 5131–5133.

[3] Green, J. E.; Choi, J. W.; Boukai, A.; Bunimovich, Y.; Johnston-Halperin, E.; DeIonno, E.; Luo, Y.; Sheriff, B. A.; Xu, K.; Shin, Y. S.; Tseng, H.-R.; Stoddart, J. F.; Heath, J. R. Nature 2007, 445, 414.

[4] Brouwer, A. M.; Frochot, C.; Gatti, F. G.; Leigh, D. A.; Mottier, L.; Paolucci, F.; Roffia, S.; Wurpel, G. W. Science 2001, 291, 2124–2128.

[5] Panman, M. R.; Bodis, P.; Shaw, D. J.; Bakker, B. H.; Newton, A. C.; Kay, E. R.; Brouwer, A. M.; Buma, W. J.; Leigh, D. A.; Woutersen, S. Science 2010, 328, 1255–1258.

[6] Baggerman, J.; Haraszkiewicz, N.; Wiering, P. G.; Fioravanti, G.; Marcaccio, M.; Paolucci, F.; Kay, E. R.; Leigh, D. A.; Brouwer, A. M. Chem. Eur. J. 2013, 19, 5566–5577.

[7] Brouwer, A. M.; Jagesar, D. C.; Wiering, P. G.; Kay, E. R.; Leigh, D. A. ChemPhysChem 2016, 17, 1902–1912.

[8] Berera, R.; van Grondelle, R.; Kennis, J. T. Photosynth. Res. 2009, 101, 105–118.

[9] Mortimer, R. G.; Eyring, H. Proc. Natl. Acad. Sci. 1980, 77, 1728–1731.

[10] Frenkel, D.; Smit, B. Understanding molecular simulation: from algorithms to applications; Aca-demic press, 2001; Vol. 1; pp 63–84.

[11] Halstead, S. J.; Li, J. Mol. Phys. 2017, 115, 297–307.

[12] Talotta, C.; Gaeta, C.; Neri, P. Org. Lett. 2012, 14, 3104–3107.

[13] Bruns, C. J.; Li, J.; Frasconi, M.; Schneebeli, S. T.; Iehl, J.; Jacquot de Rouville, H.-P.; Stupp, S. I.; Voth, G. A.; Stoddart, J. F. Angew. Chem. Int. Ed. 2014, 53, 1953–1958.

[14] Zheng, X.; Sohlberg, K. J. Phys. Chem. A 2003, 107, 1207–1215.

[15] Bowers, K. J.; Chow, E.; Xu, H.; Dror, R. O.; Eastwood, M. P.; Gregersen, B. A.; Klepeis, J. L.; Kolossvary, I.; Moraes, M. A.; Sacerdoti, F. D. Scalable algorithms for molecular dynamics simula-tions on commodity clusters. Proceedings of the 2006 ACM/IEEE conference on Supercomputing. 2006; p 84.

[16] Plimpton, S. J. Comput. Phys. 1995, 117, 1–19.

[17] Tribello, G. A.; Bonomi, M.; Branduardi, D.; Camilloni, C.; Bussi, G. Comput. Phys. Commun. 2014, 185, 604–613.

(24)

[18] Bayly, C. I.; Cieplak, P.; Cornell, W.; Kollman, P. A. J. Phys. Chem. 1993, 97, 10269–10280.

[19] Fiorin, G.; Klein, M. L.; H´enin, J. Mol. Phys. 2013, 111, 3345–3362.

[20] Cheng, C.; McGonigal, P. R.; Stoddart, J. F.; Astumian, R. D. ACS Nano 2015, 9, 8672–8688.

[21] Grabowski, S. J. J. Phys. Chem. A 2001, 105, 10739–10746.

[22] Feyereisen, M. W.; Feller, D.; Dixon, D. A. J. Phys. Chem. 1996, 100, 2993–2997.

[23] Altieri, A.; Gatti, F. G.; Kay, E. R.; Leigh, D. A.; Martel, D.; Paolucci, F.; Slawin, A. M.; Wong, J. K. J. Am. Chem. Soc. 2003, 125, 8644–8654.

(25)

Appendix 1: NPT-simulations

Figure 16. Box length of a cubic unit cell as obtained from NPT-simulations on C5.

(26)

Figure 18. Box length of a cubic unit cell as obtained from NPT-simulations on C16.

(27)

Appendix 2: Constrained MD Indices

Figure 20. The output of the CV of every constrained MD-simulation of C5 (14 in total) was tracked during the entire simulation. The black lines represent the set value in the CMD run and the coloured lines are the actual indices from the simulations.

Figure 21. The output of the CV of some constrained MD-simulation of C5-1 (13 in total) was tracked during the entire simulation. The black lines represent the set value in the CMD run and the coloured lines are the actual indices from the simulations.

(28)

Figure 22. The output of the CV at index 11.5 of C5-1 to increase the resolution around the minimum.

Figure 23. The output of the CV at index 11.8 of C5-1 to increase the resolution around the minimum. The average index lies below the set value, since the ring was practically on the nitogren of the ni station and the instability of that position outweighed the force constant of the bias potential.

(29)

Figure 24. The CV output of the first half of the constrained MD-simulation of C16 was tracked during the entire simulation. The black lines represent the set value in the CMD run and the coloured lines are the actual indices from the simulations.

Figure 25. The CV output of the second half of the constrained MD-simulation of C16 was tracked during the entire simulation. The black lines represent the set value in the CMD run and the coloured lines are the actual indices from the simulations.

(30)

Figure 26. The CV output of the first half of the constrained MD-simulation of C16-1 was tracked during the entire simulation. The black lines represent the set value in the CMD run and the coloured lines are the actual indices from the simulations.

Figure 27. The CV output of the second half of the constrained MD-simulation of C16-1 was tracked during the entire simulation. The black lines represent the set value in the CMD run and the coloured lines are the actual indices from the simulations.

(31)

Appendix 3: TA-spectra

Figure 28. TA-spectra of C5

Pathways in Rotaxane Molecular Shuttles

Bachelor Thesis Chemistry