Category Archives: Fluorescence

Density Keyword in Excited State Calculations with Gaussian


I have written about extracting information from excited state calculations but an important consideration when analyzing the results is the proper use of the keyword density.

This keyword let’s Gaussian know which density is to be used in calculating some results. An important property to be calculated when dealing with excited states is the change in dipole moment between the ground state and any given state. The Transition Dipole Moment is an important quantity that allows us to predict whether any given electronic transition will be allowed or not. A change in the dipole moment (i.e. non-zero) of a molecule during an electronic transition helps us characterize said transition.

Say you perform a TD-DFT calculation without the density keyword, the default will provide results on the lowest excited state from all the requested states, which may or may not be the state of interest to the transition of interest; you may be interested in the dipole moment of all your excited states.

Three separate calculations would be required to calculate the change of dipole moment upon an electronic transition:

1) A regular DFT for the ground state as a reference
2) TD-DFT, to calculate the electronic transitions; request as many states as you need/want, analyze it and from there you can see which transition is the most important.
3) Request the density of the Nth state of interest to be recovered from the checkpoint file with the following route section:

# TD(Read,Root=N) LOT Density=Current Guess=Read Geom=AllCheck

replace N for the Nth state which caught your eye in step number 2) and LOT for the Level of Theory you’ve been using in the previous steps. That should give you the dipole moment for the structure of the Nth excited state and you can compare it with the one in the ground state calculated in 1). Again, if density=current is not used, only properties of N=1 will be printed.

Orbital Contributions to Excited States


This is a guest post by our very own Gustavo “Gus” Mondragón whose work centers around the study of excited states chemistry of photosynthetic pigments.

When you’re calculating excited states (no matter the method you’re using, TD-DFT, CI-S(D), EOM-CCS(D)) the analysis of the orbital contributions to electronic transitions poses a challenge. In this post, I’m gonna guide you through the CI-singles excited states calculation and the analysis of the electronic transitions.

I’ll use adenine molecule for this post. After doing the corresponding geometry optimization by the method of your choice, you can do the excited states calculation. For this, I’ll use two methods: CI-Singles and TD-DFT.

The route section for the CI-Singles calculation looks as follows:

%chk=adenine.chk
%nprocshared=8
%mem=1Gb

#p CIS(NStates=10,singlets)/6-31G(d,p) geom=check guess=read scrf=(cpcm,solvent=water)

adenine excited states with CI-Singles method

0 1
--blank line--

I use the same geometry from the optimization step, and I request only for 10 singlet excited states. The CPCP implicit solvation model (solvent=water) is requested. If you want to do TD-DFT, the route section should look as follows:

%chk=adenine.chk
%nprocshared=8
%mem=1Gb

#p FUNCTIONAL/6-31G(d,p) TD(NStates=10,singlets) geom=check guess=read scrf=(cpcm,solvent=water)

adenine excited states with CI-Singles method

0 1
--blank line--

Where FUNCTIONAL is the DFT exchange-correlation functional of your choice. Here I strictly not recommend using B3LYP, but CAM-B3LYP is a noble choice to start.

Both calculations give to us the excited states information: excitation energy, oscillator strength (as f value), excitation wavelength and multiplicity:

Excitation energies and oscillator strengths:

 Excited State   1:      Singlet-A      6.3258 eV  196.00 nm  f=0.4830  <S**2>=0.000
      11 -> 39        -0.00130
      11 -> 42        -0.00129
      11 -> 43         0.00104
      11 -> 44        -0.00256
      11 -> 48         0.00129
      11 -> 49         0.00307
      11 -> 52        -0.00181
      11 -> 53         0.00100
      11 -> 57        -0.00167
      11 -> 59         0.00152
      11 -> 65         0.00177

The data below corresponds to all the electron transitions involved in this excited state. I have to cut all the electron transitions because there are a lot of them for all excited states. If you have done excited states calculations before, you realize that the HOMO-LUMO transition is always an important one, but not the only one to be considered. Here is when we calculate the Natural Transition Orbitals (NTO), by these orbitals we can analyze the electron transitions.

For the example, I’ll show you first the HOMO-LUMO transition in the first excited state of adenine. It appears in the long list as follows:

35 -> 36         0.65024

The 0.65024 value corresponds to the transition amplitude, but it doesn’t mean anything for excited state analysis. We must calculate the NTOs of an excited state from a new Gaussian input file, requesting from the checkpoint file we used to calculate excited states. The file looks as follows:

%Oldchk=adenine.chk
%chk=adNTO1.chk
%nproc=8
%mem=1Gb

#p SP geom=allcheck guess=(read,only) density=(Check,Transition=1) pop=(minimal,NTO,SaveNTO)

I want to say some important things right here for this last file. See that no level of theory is needed, all the calculation data is requested from the checkpoint file “adenine.chk”, and saved into the new checkpoint file “adNTO1.chk”, we must use the previous calculated density and specify the transition of interest, it means the excited state we want to analyze. As we don’t need to specify charge, multiplicity or even the comment line, this file finishes really fast.

After doing this last calculation, we use the new checkpoint file “adNTO1.chk” and we format it:

formchk -3 adNTO1.chk adNTO1.fchk

If we open this formatted checkpoint file with GaussView, chemcraft or the visualizer you want, we will see something interesting by watching he MOs diagram, as follows:

We can realize that frontier orbitals shows the same value of 0.88135, which means the real transition contribution to the first excited state. As these orbitals are contributing the most, we can plot them by using the cubegen routine:

cubegen 0 mo=homo adNTO1.fchk adHOMO.cub 0 h

This last command line is for plotting the equivalent as the HOMO orbital. If we want to plot he LUMO, just change the “homo” keyword for “lumo”, it doesn’t matter if it is written with capital letters or not.

You must realize that the Natural Transition Orbitals are quite different from Molecular Orbitals. For visual comparisson, I’ve printed also the molecular orbitals, given from the optimization and from excited states calculations, without calculating NTOs:

These are the molecular frontier orbitals, plotted with Chimera with 0.02 as the isovalue for both phase spaces:

The frontier NTOs look qualitatively the same, but that’s not necessarily always the case:

If we analyze these NTOs on a hole-electron model, the HOMO refers to the hole space and the LUMO refers to the electron space.

Maybe both orbitals look the same, but both frontier orbitals are quite different between them, and these last orbitals are the ones implied on first excited state of adenine. The electron transition will be reported as follows:

If I can do a graphic summary for this topic, it will be the next one:

NTOs analysis is useful no matter if you calculate excited states by using CIS(D), EOM-CCS(D), TD-DFT, CASSCF, or any of the excited states method of your election. These NTOs are useful for population analysis in excited states, but these calculations require another software, MultiWFN is an open-source code that allows you to do this analysis, and another one is called TheoDORE, which we’ll cover in a later post.

Natural Transition Orbitals (NTOs) Gaussian


The canonical molecular orbital depiction of an electronic transition is often a messy business in terms of a ‘chemical‘ interpretation of ‘which electrons‘ go from ‘which occupied orbitals‘ to ‘which virtual orbitals‘.

Natural Transition Orbitals provide a more intuitive picture of the orbitals, whether mixed or not, involved in any hole-particle excitation. This transformation is particularly useful when working with the excited states of molecules with extensively delocalized chromophores or multiple chromophoric sites. The elegance of the NTO method relies on its simplicity: separate unitary transformations are performed on the occupied and on the virtual set of orbitals in order to get a localized picture of the transition density matrix.

[1] R. L. Martin, J. Chem. Phys., 2003, DOI:10.1063/1.1558471.

In Gaussian09:
After running a TD-DFT calculation with the keyword TD(Nstates=n) (where n = number of states to be requested) we need to take that result and launch a new calculation for the NTOs but lets take it one step at a time. As an example here’s phenylalanine which was already optimized to a minimum at the B3LYP/6-31G(d,p) level of theory. If we take that geometry and launch a new calculation with the TD(Nstates=40) in the route section we obtain the UV-Vis spectra and the output looks like this (only the first three states are shown):

Excitation energies and oscillator strengths:

Excited State 1: Singlet-A 5.3875 eV 230.13 nm f=0.0015 <S**2>=0.000
42 -> 46 0.17123
42 -> 47 0.12277
43 -> 46 -0.40383
44 -> 45 0.50838
44 -> 47 0.11008
This state for optimization and/or second-order correction.
Total Energy, E(TD-HF/TD-KS) = -554.614073682
Copying the excited state density for this state as the 1-particle RhoCI density.

Excited State 2: Singlet-A 5.5137 eV 224.86 nm f=0.0138 <S**2>=0.000
41 -> 45 -0.20800
41 -> 47 0.24015
42 -> 45 0.32656
42 -> 46 0.10906
42 -> 47 -0.24401
43 -> 45 0.20598
43 -> 47 -0.14839
44 -> 45 -0.15344
44 -> 47 0.34182

Excited State 3: Singlet-A 5.9254 eV 209.24 nm f=0.0042 <S**2>=0.000
41 -> 45 0.11844
41 -> 47 -0.12539
42 -> 45 -0.10401
42 -> 47 0.16068
43 -> 45 -0.27532
43 -> 46 -0.11640
43 -> 47 0.16780
44 -> 45 -0.18555
44 -> 46 -0.29184
44 -> 47 0.43124

The oscillator strength is listed on each Excited State as “f” and it is a measure of the probability of that excitation to occur. If we look at the third one for this phenylalanine we see f=0.0042, a very low probability, but aside from that the following list shows what orbital transitions compose that excitation and with what energy, so the first line indicates a transition from orbital 41 (HOMO-3) to orbital 45 (LUMO); there are 10 such transitions composing that excitation, visualizing them all with canonical orbitals is not an intuitive picture, so lets try the NTO approach, we’re going to take excitation #10 for phenylalanine as an example just because it has a higher oscillation strength:

%chk=Excited State 10: Singlet-A 7.1048 eV 174.51 nm f=0.3651 <S**2>=0.000
41 -> 45 0.35347
41 -> 47 0.34685
42 -> 45 0.10215
42 -> 46 0.17248
42 -> 47 0.13523
43 -> 45 -0.26596
43 -> 47 -0.22995
44 -> 46 0.23277

Each set of NTOs for each transition must be calculated separately. First, copy you filename.chk file from the TD-DFT result to a new one and name it after the Nth state of interest as shown below (state 10 in this case). NOTE: In the route section, replace N with the number of the excitation of interest according to the results in filename.log. Run separately for each transition your interested in:

#chk=state10.chk

#p B3LYP/6-31G(d,p) Geom=AllCheck Guess=(Read,Only) Density=(Check,Transition=N) Pop=(Minimal,NTO,SaveNTO)

0 1
--blank line--

By requesting SaveNTO, the canonical orbitals in the state10.chk file are replaced with the NTOs for the 10th excitation, this makes it easier to plot since most visualizers just plot whatever set of orbitals they read in the chk file but if they find the canonical MOs then one would need to do some re-processing of them. This is much more straightforward.

Now we format our chk files into fchk with the formchk utility:

formchk -3 filename.chk filename.fchk
formchk -3 state10.chk state10.fchk

If we open filename.fchk (the file where the original TD-DFT calculation is located) with GaussView we can plot all orbitals involved in excited state number ten, those would be seven orbitals from 41 (HOMO-3) to 47 (LUMO+2) as shown in figure 1.

Figure 1. Canonical orbitals involved in the 10th excited state according to the TD-DFT calculation

If we now open state10.fchk we see that the numbers at the side of the orbitals are not their energy but their occupation number particular to this state of interest, so we only need to plot those with highest occupations, in our example those are orbitals 44 and 45 (HOMO and LUMO) which have occupations = 0.81186; you may include 43 and 46 (HOMO-1 and LUMO+1, respectively) for a much more complete description (occupations = 0.18223) but we’re still dealing with 4 orbitals instead of 7.

Figure 2. Natural Transition Orbitals for Phenylalanine. Orbital 44 (particle) and Orbital 45 (hole) exhibit the largest occupations for Excited State No. 10

The NTO transition 44 -> 45 is far easier to conceptualize than all the 10 combinations given in the canonical basis from the direct TD-DFT calculation. TD-DFT provides us with the correct transitions, NTOs just paint us a picture more readily available to the chemist mindset.

NOTE: for G09 revC and above, the %OldChk option is available, I haven’t personally tried it but using it to specify where the excitations are located and then write the NTOs of interest into a new chk file in the following way, thus eliminating the need of copying the original chk file for each state:

%OldChk=filename.chk
%chk=stateN.chk

NTOs are based on the Natural Hybrid orbitals vision by Löwdin and others, and it is said to be so straightforward that it has been re-discovered from time to time. Be that as it may, the NTO visualization provides a much clearer vision of the excitations occurring during a TD calculation.

Thanks for reading, stay home and stay safe during these harsh days everyone. Please share, rate and comment this and other posts.

Calculation of Intermolecular Interactions for Sensors with Biological Applications


Two new papers on the development of chemosensors for different applications were recently published and we had the opportunity to participate in both with the calculation of electronic interactions.

A chemosensor requires to have a measurable response and calculating either that response from first principles based on the electronic structure, or calculating another physicochemical property related to the response are useful strategies in their molecular design. Additionally, electronic structure calculations helps us unveil the molecular mechanisms underlying their response and efficiency, as well as providing a starting point for their continuous improvement.

In the first paper, CdTe Quantum Dots (QD’s) are used to visualize in real time cell-membrane damages through a Gd Schiff base sensitizer (GdQDs). This probe interacts preferentially with a specific sequence motif of NHE-RF2 scaffold protein which is exposed during cell damage. This interactions yields intensely fluorescent droplets which can be visualized in real time with standard instrumentation. Calculations at the level of theory M06-2X/LANL2DZ plus an external double zeta quality basis set on Gd, were employed to characterize the electronic structure of the Gd³⁺ complex, the Quantum Dot and their mutual interactions. The first challenge was to come up with the right multiplicity for Gd³⁺ (an f⁷ ion) for which we had no experimental evidence of their magnetic properties. From searching the literature and talking to my good friend, inorganic chemist Dr. Vojtech Jancik it was more or less clear the multiplicity had to be an octuplet (all seven electrons unpaired).

As can be seen in figure 1a the Gd-N interactions are mostly electrostatic in nature, a fact that is also reflected in the Wiberg bond indexes calculated as 0.16, 0.17 and 0.21 (a single bond would yield a WBI value closer to 1.0).

PM6 optimizations were employed in optimizing the GdQD as a whole (figure 1f) and the MM-UFF to characterize their union to a peptide sequence (figure 2) from which we observed somewhat unsurprisingly that Gd³⁺interacts preferently with the electron rich residues.

This research was published in ACS Applied Materials and Interfaces. Thanks to Prof. Vojtech Adam from the Mendel University in Brno, Czech Republic for inviting me to collaborate with their interdisciplinary team.

The second sensor I want to write about today is a more closer to home collaboration with Dr. Alejandro Dorazco who developed a fluorescent porphyrin system that becomes chiefly quenched in the presence of Iodide but not with any other halide. This allows for a fast detection of iodide anions, related to some gland diseases, in aqueous samples such as urine. This probe was also granted a patent which technically lists yours-truly as an inventor, cool!

The calculated interaction energy was huge between I⁻ and the porphyrine, which supports the idea of a ionic interaction through which charge transfer interactions quenches the fluorescence of the probe. Figure 3 above shows how the HOMO largely resides on the iodide whereas the LUMO is located on the pi electron system of the porphyrine.

This research was published in Sensors and Actuators B – Chemical.

New paper in JPC-A


As we approach to the end of another year, and with that the time where my office becomes covered with post-it notes so as to find my way back into work after the holidays, we celebrate another paper published! This time at the Journal of Physical Chemistry A as a follow up to this other paper published last year on JPC-C. Back then we reported the development of a selective sensor for Hg(II); this sensor consisted on 1-amino-8-naphthol-3,6-disulphonic acid (H-Acid) covalently bound to a modified silica SBA-15 surface. H-Acid is fluorescent and we took advantage of the fact that, when in the presence of Hg(II) in aqueous media, its fluorescence is quenched but not with other ions, even with closely related ions such as Zn(II) and Cd(II). In this new report we delve into the electronic reasons behind the quenching process by calculating the most important electronic transitions with the framework laid by the Time Dependent Density Functional Theory (TD-DFT) at the PBE0/cc-pVQZ level of theory (we also included an electron core potential on the heavy metal atoms in order to decrease the time of each calculation). One of the things I personally liked about this work is the combination of different techniques that were used to assess the photochemical phenomenon at hand; some of those techniques included calculation of various bond orders (Mayer, Fuzzy, Wiberg, delocalization indexes), time dependent DFT and charge transfer delocalizations. Although we calculated all these various different descriptors to account for changes in the electronic structure of the ligand which lead to the fluorescence quenching, only delocalization indexes as calculated with QTAIM were used to draw conclusion, while the rest are collected in the SI section.

jpca

Thanks a lot to my good friend and collaborator Dr. Pezhman Zarabadi-Poor for all his work, interest and insight into the rationalization of this phenomenon. This is our second paper published together. By the way, if any of you readers is aware of a way to finance a postdoc stay for Pezhman here at our lab, please send us a message because right now funding is scarce and we’d love to keep bringing you many more interesting papers.

For our research group this was the fourth paper published during 2014. We can only hope (and work hard) to have at least five next year without compromising their quality. I’m setting the goal to be 6 papers; we’ll see in a year if we delivered or not.

I’d like to also take this opportunity to thank all the readers of this little blog of mine for your visits and your live demonstrations of appreciation at various local and global meetings such as the ACS meeting in San Francisco and WATOC14 in Chile, it means a lot to me to know that the things I write are read; if I were to make any New Year’s resolutions it would be to reply quicker to questions posted because if you took the time to write I should take the time to reply.

I wish you all the best for 2015 in and out of the lab!

%d bloggers like this: