Experimental and computational electrochemistry of quinazolinespirohexadienone molecular switches - Differential electrochromic vs photochromic behavior

Article abstract of DOI:10.3762/bjoc.15.240

Our undergraduate research group has long focused on the preparation and investigation of electron-deficient analogs of the perimidinespirohexadienone (PSHD) family of photochromic molecular switches for potential application as "photochromic photooxidants" for gating sensitivity to photoinduced charge transfer. We previously reported the photochemistry of two closely related and more reducible quinazolinespirohexadienones (QSHDs), wherein the naphthalene of the PSHD is replaced with a quinoline. In the present work, we report our investigation of the electrochemistry of these asymmetric QSHDs. In addition to the short wavelength and photochromic long-wavelength isomers, we have found that a second, distinct long-wavelength isomer is produced electrochemically. This different long-wavelength isomer arises from a difference in the regiochemistry of spirocyclic ring-opening. The structures of both long-wavelength isomers were ascertained by cyclic voltammetry and ¹H NMR analyses, in concert with computational modeling. These results are compared to those for the symmetric parent PSHD, which due to symmetry possesses only a single possible regioisomer upon either electrochemical or photochemical ring-opening. Density functional theory calculations of bond lengths, bond orders, and molecular orbitals allow the rationalization of this differential photochromic vs electrochromic behavior of the QSHDs.

Full text of DOI:10.3762/bjoc.15.240

Experimental and computational electrochemistry of
quinazolinespirohexadienone molecular switches –
differential electrochromic vs photochromic behavior
Eric W. Webb, Jonathan P. Moerdyk, Kyndra B. Sluiter, Benjamin J. Pollock,
Amy L. Speelman, Eugene J. Lynch, William F. Polik and Jason G. Gillmore*
Full Research Paper
Open Access
Address:
Beilstein J. Org. Chem. 2019, 15, 2473–2485.
Department of Chemistry, Hope College, 35 East 12th Street,
Holland, MI 49422-9000, USA
doi:10.3762/bjoc.15.240
Received: 03 July 2019
Email:
Accepted: 10 October 2019
Published: 18 October 2019
Jason G. Gillmore* - gillmore@hope.edu
* Corresponding author
This article is part of the thematic issue "Molecular switches".
Guest Editor: W. Szymanski
Keywords:
cyclic voltammetry; density functional theory; heterocycles; molecular
switches; photochromic photooxidants; spirocycles
© 2019 Webb et al.; licensee Beilstein-Institut.
License and terms: see end of document.
Abstract
Our undergraduate research group has long focused on the preparation and investigation of electron-deficient analogs of the perim-
idinespirohexadienone (PSHD) family of photochromic molecular switches for potential application as "photochromic photooxi-
dants" for gating sensitivity to photoinduced charge transfer. We previously reported the photochemistry of two closely related and
more reducible quinazolinespirohexadienones (QSHDs), wherein the naphthalene of the PSHD is replaced with a quinoline. In the
present work, we report our investigation of the electrochemistry of these asymmetric QSHDs. In addition to the short wavelength
and photochromic long-wavelength isomers, we have found that a second, distinct long-wavelength isomer is produced electro-
chemically. This different long-wavelength isomer arises from a difference in the regiochemistry of spirocyclic ring-opening. The
structures of both long-wavelength isomers were ascertained by cyclic voltammetry and 1H NMR analyses, in concert with compu-
tational modeling. These results are compared to those for the symmetric parent PSHD, which due to symmetry possesses only a
single possible regioisomer upon either electrochemical or photochemical ring-opening. Density functional theory calculations of
bond lengths, bond orders, and molecular orbitals allow the rationalization of this differential photochromic vs electrochromic be-
havior of the QSHDs.
Introduction
Photochromic molecular switches, in which light initiates re- thermally or photochemically, have become ubiquitous in a
versible coloration of a short-wavelength isomer (SW) to a wide range of applications [1-5]. Typically, organic photochro-
long-wavelength isomer (LW), which fades back to SW either mism results from a spirocyclic ring-opening or other isomeri-
2473

Beilstein J. Org. Chem. 2019, 15, 2473–2485.
zation which results in increased conjugation. Electrochromism gating to photoinduced charge transfer (PICT) initiated pro-
is also of increasing materials relevance, e.g., for self-dimming cesses (Figure 1). PSHDs were promising for this, as their pho-
automotive mirror and aircraft window darkening applications tochromic reversion of LW back to SW proceeds purely ther-
[6-8]. In electrochromic applications, the color change is gener- mally, leaving the long-wavelength absorption available for
ally due to a change in the oxidation state of a conjugated bimolecular PICT. Moreover, the LW is sufficiently more
system. This change in the oxidation state is most often con- reducible that, even accounting for its lower excitation energy,
comitant with conformational and orbital occupancy changes, the LW excited state, LW*, is more reducible than SW*,
rather than any σ-bond-forming or breaking processes. The making LW the more potent photooxidant of the two. However,
viologens are perhaps the most ubiquitous example of small for practical use as photooxidants, the difference in the reduc-
molecule organic electrochromism [6,7,9].
tion potential between LW and SW would need to be increased
further, and LW would need to be more reducible to be of use
One example that combines photochromic and electrochromic in photooxidation of relevant substrates (e.g., Dewar benzenes,
behavior (the latter of an unusual sort) is the class of perim- quadricyclanes, or bishomocubanes as quantum amplified isom-
idinespirohexadienones 1 (PSHDs) whose synthesis, electro- erization substrates [11-15], or vinylcarbazole or alkoxystyrene
chemistry and UV–vis spectroscopy were reported by Minkin derivatives for radical cation cylcloaddition and polymerization
and co-workers [10] for 1a (Scheme 1). Electrochemically, they reactions [16-20]). We thus proposed the replacement of the
report observing a single, two-electron reduction peak and two naphthalene in 1a with a more electron-deficient quinoline ring.
distinct one-electron oxidation peaks upon scanning using Due to the saturated spirocyclic carbon insulating the dienone
cyclic voltammetry. They therefore proposed [10] that the elec- electrophore from the quinoline moiety in the SW form, we ex-
trolysis of 1a proceeds by an “ECE” (electrochemical-chemical- pected minimal change in the SW reduction potential relative to
electrochemical) mechanism (1a → 1a•− → 2a•− → 2a2− → the PSHDs, but a significant difference for the completely
2a•− → 2a) in which the dienone portion of the molecule conjugated LW isomer(s). Previously we reported the detailed
accepted the first electron, followed by a radical anion rear- synthesis of two novel quinazolinespirohexadienone (QSHD)
rangement to the long-wavelength isomer, whose radical anion photochromes 3a,b (Scheme 2) and their photochemical proper-
is so much easier to reduce that it immediately accepts a second ties as well as a proof of structure for the photochemically
electron at this potential; on the oxidative return wave the generated long-wavelength isomer (pLW) 4a,b (not 5a,b) [21].
subsequent oxidations of the LW dianion to its radical anion Herein, we report the electrochemistry of these QSHDs.
and then its neutral state are observed. Thus, in this unusual
system, electrochromism proceeds by the same sort of spiro-
cyclic ring-opening as the photochromic rearrangement but
occurs from the radical anion rather than a photoexcited state.
The reduction of the molecular switch necessary for elec-
trochromism is in a sense catalytic: the rearranged product is
reoxidized to a neutral LW isomer, which reverts thermally to
SW upon standing, just as it does when the LW is generated
photochemically.
Figure 1: Proposed gating of sensitivity to photoinduced charge
transfer by a photochromic photooxidant in which only LW is a compe-
tent photooxidant of the donor of interest.
Results and Discussion
Electrochemical analysis
Scheme 1: PSHD photochromism [10].
When we replicated cyclic voltammetry (CV) experiments on
PSHDs, we observed similar voltammograms for both 1a and
The PSHD system was of interest to us as a potential photo- 1b (Figure 2), consistent with the two-electron reduction and
chromic photooxidant that would add an additional level of two one-electron oxidation processes reported by Minkin for 1a
2474

Beilstein J. Org. Chem. 2019, 15, 2473–2485.
Scheme 2: QSHD photochromism [21].
Figure 2: Cyclic voltammograms of a) 1b before irradiation or electrolysis (solid blue), b) 1b/2b after 25 scans (dotted red), c) 1b/2b upon photolysis
of 1b (dashed green) scanned over a narrower potential window so as not to reduce 1b, and d) 1b/2b scanning the full potential window after photoly-
sis of 1b (dashed orange).
[10]. As expected, a growth of two reversible one-electron can be attributed at least in part to our use of half-peak poten-
reduction–oxidation peaks was observed upon multiple scans, tials for all irreversible peaks, while Minkin reported peak
representing the reduction and oxidation of the electrogener- potentials (our peak potentials were within 60 mV of Minkin’s
ated long-wavelength form 1b generated in situ. As expected, value). We found the parent compound 1b, whose electrochem-
either photolysis or multiple CV scans led to the same LW istry was not previously reported by Minkin, to be 50 mV less
reduction and oxidation peaks (within the error bars indicated in reducible than 1a (−1.68 V vs −1.63 V for 1a), in qualitative
Table 1). We also report an identical Eored of 2a as Minkin [10], agreement with our computational predictions [22]. Similarly,
though a 140 mV difference was found in the Eored of 1a, which 2b differed from 2a by only 20 mV (Table 1) where computa-
2475

Beilstein J. Org. Chem. 2019, 15, 2473–2485.
Table 1: Experimental and computational Eored of 1a,b and 3a,b and their LW isomers, reported in V vs SCE.
Electrochemical
Eored (eLW·−)
Photochemical
Eored (pLW·−) Eored (pLW)
Compd.
1a
Type
Eored (SW)a
Eored (eLW)
literatureb
exptl.c
predictedd
exptlc
predictedd
exptlc
predictedd
exptlc
−1.77
−1.628 0.031
−1.56
−1.681 0.013
−1.70
−1.632 0.030
−1.61
−1.325
−1.220 0.049
−0.85
−0.857 0.010
−0.89
−0.871 0.013
−0.90
−0.866 0.009
−0.88
−0.89
1b
3a
3b
−1.202 0.032
−1.26 0.11
−1.170 0.060
−1.029 0.035
−1.008 0.024
−0.865 0.005
−0.90
−0.730 0.005
−0.72
−1.631 0.017
−1.069ad
0.029
−0.843 0.012
−0.729 0.004
predictedd
−1.63
−0.86
−0.72
aIrreversible peak. Eored is reported as the half-peak potential Ep/2red (except in literature value for 1a). bReference [10]. Irreversible SW peak re-
ported as peak potential (Epred), not half-peak potential (Ep/2red). cExperimental values in CH3CN containing 0.1 M Bu4NPF6, standardized vs ferro-
cene/ferrocenium and corrected to vs SCE as in reference [23]; error bars = one standard deviation from the mean of at least 7 replicates.
dComputational Eored predicted using correlation 6 in reference [22] based on the energies of the corresponding ground-state and one-electron-
reduced species computed using B3LYP/6-31G(d) with implicit acetonitrile solvent using the CPCM solvent model with UAKS radii, on geometries op-
timized in the gas-phase; predictions of second reduction potentials are not possible by this method.
tions predicted a minimal difference. The electrogeneration of similar ECE mechanism to that of 1a is likely also occurring for
2b through repeated potential step bulk electrolysis of 1b in 3b.
acetonitrile-d3 yielded a sufficient quantity of LW to obtain a
1H NMR spectrum, which revealed identical chemical shifts as Cyclic voltammetric analysis of quinazolinespirohexadienone
those for the photogenerated 2b. First and second reduction (QSHD) 3b (Figure 3) was qualitatively similar to the parent
potentials were also the same, within error limits, for photogen- PSHDs 1a,b as expected based on structural similarity and
erated and electrogenerated 2b, as would be expected. This is computationally calculated molecular orbital diagrams
consistent with the excellent overlap of all four voltammo- (Figure 6 and Supporting Information File 1). Surprisingly, the
grams in Figure 2.
Eored of the electrochemically generated LW form of 3a was
more negative by 10 mV than that of 2a, even with the more
The ECE mechanism reported by Minkin for the elec- electron-deficient quinoline ring. Presumably this is because the
trochromism was further supported by bulk electrochemical ex- 5-position on the benzene ring of the quinoline, is the least
periments. With repetitive conditioning and scanning under withdrawing point of attachment, and the inductive with-
argon-deaerated conditions, the initially yellow 1b solution drawing properties of the quinoline nitrogen are far enough re-
turned to an orange-red color, which we hypothesized to be the moved from the electrophore to not cause any appreciable
LW dianion, 2b2−. Upon exposure to air this solution immedi- change in reduction potential. The other qualitative difference
ately turned green, the known color of the LW isomer 2b. The for 3b was the presence of two LW reduction peaks (LW →
addition of benzoquinone (a chemical oxidant) to the electro- LW•− → LW2−) on the first scan (Figure 3a, dotted red). How-
chemically reduced solution under argon gave similar results, ever, the observation of the two LW reduction peaks was
consistent with our hypothesis. When the yellow solution consistent with UV–vis spectroscopy that indicated a signifi-
remained open to the atmosphere during electrolysis, the cant amount of a LW isomer upon solvation, and repetitive
initially yellow solution turned green at first (presumably while electrochemical scans exhibited the anticipated growth of the
air oxidation of 2b2− to 2b could keep up with electrochemical LW reduction–oxidation peaks as more of the LW isomer was
reduction). With further electrolysis even these solutions turned generated electrochemically (Figure 3b, solid blue).
to the orange-red color observed for the deaerated solutions.
The solutions did become green upon prolonged exposure to Most surprisingly, when 3b was photolyzed to form 4b in solu-
air. This behavior seems indicative of insufficient transport of tion under similar conditions to our previous report [21] and
air into the cell through the small vents in the cell cap to replace then analyzed electrochemically (Figure 3c and d, dashed green
the oxygen being consumed during repeated electrolytic scans. and orange), four reduction (and oxidation) peaks in the region
Similar results were observed for solutions of 3b suggesting a of the LW isomer were observed. Two reduction–oxidation
2476

Beilstein J. Org. Chem. 2019, 15, 2473–2485.
Figure 3: Cyclic voltammograms of a) 3b (with trace 5b) before irradiation or electrolysis (solid blue), b) 3b + 5b after 25 scans (dotted red) showing
the growth of the two 5b reduction peaks, c) 3b (with trace 5b) after photolyzing but without reducing 3b in the electrolysis (dashed green) which
shows the two initial eLW 5b and two new pLW 4b reversible redox waves observed, and d) 3b + 4b + 5b observed by scanning the full potential
window after the photolysis of 3b (dashed orange).
couples matched the potential of the electrogenerated LW form single (likely two-electron) return oxidation peak was observed,
observed earlier while the other two peaks were shifted more possibly indicating a large overpotential for the oxidation of
positive by 60–110 mV indicating the presence of a third elec- 5a2−. It is possible that 3a undergoes only a one-electron reduc-
troactive species in addition to 3b and photogenerated 4b. tion and rearrangement of 3a → 3a•−→ 5a•− (without further
Given the asymmetric nature of 3b, two distinct options for reduction to 5a2−) and subsequently only one oxidation to 5a.
spirocyclic ring-opening exist, leading to 4b or 5b. Thus, we But this would make the second one-electron reduction (thought
postulated that the electrogenerated LW form was in fact 5b. to be 5a•− → 5a2−) observed on repeated cycling unexplainable.
A more likely explanation is that the two-electron ECE reduc-
Having found two distinct LW forms depending on generation tion of 3a still occurs to give 5a2− but that the slight electron-
by photolysis or electrolysis of SW 3b, we turned our attention donating nature of the additional methyl group destabilizes 5a•−
to whether a similar phenomenon was observed for 3a. Indeed, enough to require a substantial overpotential for reoxidation of
different redox peaks were observed in the same general LW 5a2− to 5a•−, such that it occurs at the same potential as oxida-
region for the CV of photolyzed versus electrogenerated LW tion of 5a•− back to neutral 5a. This could occur either sequen-
forms of 3a, consistent with the electrogenerated formation of tially at the same potential or in a single two-electron process.
5a compared against the known formation of 4a via photolysis. The latter explanation seems qualitatively in better agreement
The voltammograms of 3a (Figure 4) did however differ from with the observed voltammogram in Figure 4.
both 3b and 1b. A (presumably two-electron) reduction peak
was observed for 3a but only one return oxidation wave was ob- In terms of achieving more potent photooxidants through the
served. Upon repeated scanning the two one-electron reduc- exchange of the naphthalene ring of the PSHDs for the more
tions of 5a → 5a•− → 5a2− were observed, but still only a electron-deficient quinoline in the QSHDs, we had expected the
2477

Beilstein J. Org. Chem. 2019, 15, 2473–2485.
Figure 4: Cyclic voltammograms of a) 3a before irradiation or electrolysis (solid blue), b) 3a + 5a after 25 scans (dotted red) showing the growth of the
two eLW 5a reduction peaks, c) 3a after photolyzing but without reducing 3a in the electrolysis (dashed green) which shows the two different revers-
ible redox waves observed for pLW 4a, and d) 3a + 4a + 5a scanning the full potential window after the photolysis (dashed orange).
LW isomers to become more easily reducible, with minimal further supported through NMR. Previous work [21] had
change in reduction potential for the SW isomers. Indeed Eored conclusively shown through 1H NOE NMR spectroscopy that
of the SW isomers 1b, 3a, and 3b were the same within the the LW isomers resulting from photolysis of 3a and 3b were 4a
error, and only 50 mV more reducible than 1a. Comparing the and 4b, which open toward the more electron-deficient
Eored of pLW isomers (4a,b relative to 2a,b), a roughly 140 mV heteroaromatic ring of the quinoline and away from the R
difference is observed, with 4 being more reducible than 2 as group. Unfortunately, efforts to obtain a sufficient quantity of
expected. However, the difference in reduction potential be- the electrogenerated LW form of 3a (i.e., 4a or 5a) for defini-
tween PSHD and QSHD for eLW was much less. For 5b Eored tive 1H NMR spectra or NOE experiments were not successful
was 28 mV more positive than 2b indicating that 5b is a slightly However, both UV–vis and electrochemical measurements indi-
better oxidant than the parent PSHD. But Eored of eLW 5a was cated small amounts of a long wavelength form present in the
surprisingly shifted 10 mV more negative than PSHD LW 2a, dark immediately upon solvation (i.e., a thermal or solva-
meaning it was actually harder to oxidize. Ultimately Eored of tochromic LW form), particularly in 3a. The presence of a LW
2a, 2b, 5a, and 5b are essentially the same within error limits. isomer prior to irradiation or electrolysis was also consistent
Apparently, the quinoline is not nearly as electron withdrawing with the 1H NMR spectrum of a sample of 3a which, while
when linked through the benzo ring as when it is linked through known to be pure in the solid state [21]), in solution showed the
the heteroaromatic ring.
expected 3a chemical shifts but also smaller peaks (ca. 20%
relative to 3a) with similar splitting and chemical shifts as those
for photogenerated 4a (to which we initially erroneously attri-
Spectroscopic analysis
The potential for two distinct products from electrolysis or pho- buted them [21]). However, the frequencies for this solva-
tolysis as indicated through electrochemical analysis was tochromic LW were slightly but distinctly different from either
2478

Beilstein J. Org. Chem. 2019, 15, 2473–2485.
3a or photogenerated 4a (Figure 5a, e.g., consider protons n, g, pLW isomer 4a that grows in upon photolysis (Figure 3c,
and v). Signals for 4a began to grow in with even very brief dashed green; Figure 5b). Furthermore, modest NOE enhance-
photolysis (Figure 5b), intentionally taken to low conversion ments can be observed even on the small amounts of 5a present
(ca. 8% relative to 3a) to enable comparison to the solvat- in these solutions. These NOEs, while weak, did aid in
achromic LW (ca. 20% relative to 3a). Moreover, as the oxida- assigning the peaks as labelled in Figure 5b and demonstrate
tion and reduction peaks associated with eLW 5a (Figure 3b, that the blue-labelled eLW peaks in Figure 5 are indeed on the
dotted red) grow with increasing numbers of CV scans and same molecule, and that this is distinct from pLW 4a as studied
match those present initially in solution in small amounts ther- previously by NOE [21]. Thus, NMR, while not conclusive on
mally (Figure 3a, solid blue), we conclude the electrogenerated its own, was able to provide additional support for our struc-
and thermal LW forms are the same isomer 5a, while it is the tural assignment.
Figure 5: 1H NMR distinction between SW 3a, thermal/eLW 5a, and pLW 4a, in acetone-d6, as observed a) before and b) after photolysis.
2479

Beilstein J. Org. Chem. 2019, 15, 2473–2485.
UV–vis spectra taken of eLW solutions generated by bulk elec- Eored for all SW and LW structures in Table 1 is just 21 mV
trolysis in acetonitrile and acetonitrile-d3 were compared (27 mV root mean squared deviation), not far from the 14 mV
against photoirradiated (pLW) solutions. No difference was ob- mean standard deviation in the experimental measurements.
served in the UV–vis spectrum of the LW isomer prepared from This demonstrates that the computational correlation employed
either photolysis or electrolysis of solutions of 1b (as expected is useful in structural assignment and accurate within a stan-
due to symmetry). The λmax for the 2b photogenerated solution dard of error of the experimental reduction potentials of these
was at 574 nm in acetonitrile and acetonitrile-d3. The electro- spirohexadienones’ different constitutional SW and LW
generated 2b λmax at 573 nm in both solvents differed by only isomers.
1 nm, within error limits of our instrumentation. For 4a/5a the
λmax in acetonitrile-d3 was 539 nm for the photolyzed solution Density functional theory (DFT) calculated molecular orbitals
(4a) versus 529 nm for the electrolyzed (5a). Similarly, for 4b (MOs, e.g., Figure 6 and Supporting Information File 1) indi-
and 5b, the photolyzed λmax in acetonitrile was 558 nm versus cate that the LUMO (MO 106 for 3a) lies exclusively on the
549 nm for the electrogenerated and was 564 nm versus the dienone moiety both in 1b and in the quinazolinespirohexa-
electrogenerated 550 nm in acetonitrile-d3. The differences in dienone photochromes 3a,b. Thus, the dienone moiety is likely
the UV–vis spectrum of about 10 nm between eLW and pLW the electrophore in the SW in all PSHD and QSHD switches,
isomers of QSHDs 3 indicated a similar length for the conju- just as Minkin [10] specifically asserted for 1a. These diagrams
gated system but a significant enough difference to indicate dif- also suggested to us that QSHDs 3 might also undergo an ECE
ferent species. This is consistent with the formation of two mechanism similar to PSHDs 1, as we have now confirmed ex-
constitutional LW isomers that are structurally similar yet perimentally.
distinct, as the difference in absorbance would be expected to
be noticeable but not large. Observation of no difference in the
UV–vis spectrum of the LW isomer prepared from either pho-
tolysis or electrolysis of solutions of 1b (able to only form one
LW isomer, 2b) supports the formation of two different LW
isomers for 3a,b rather than attributing the small spectral
changes to the presence of electrolyte in the electrolyzed solu-
tions, to an interaction with air, or to a side reaction in solution.
The shorter wavelength for electrogenerated 5a,b may result
from slightly less planarity and decreased conjugation versus
the photogenerated 4a,b. The additional ortho-methyl group in
5b may exacerbate this sterically, and/or could contribute an
inductively donating effect.
Finally, the thermal reversion of eLW 5a,b was visually and
spectroscopically obvious to have begun within a few minutes,
Figure 6: HOMO (MO 105, red and blue) and LUMO (MO 106, green
and to be complete within 12–18 hours. This is similar to what
we previously reported for pLW 4a,b [21]. This comparatively
slow and purely thermal reversion is consistent with the need
for a thermodynamically unfavorable intramolecular (or sol-
and yellow) computed for 3a in its ground (S0) state. Similar results
were observed for the frontier orbitals of 1b and 3b (see Supporting
Information File 1).
vent or adventitious acid/base-mediated) proton transfer to Having conclusively demonstrated the differential regiospeci-
begin the reversion mechanism, as Minkin has described [10]. ficity of photochromic vs electrochromic ring-opening of the
Interestingly, reversion of either LW species to SW is observed QSHDs (Scheme 3) by experiment, we next sought to under-
immediately upon removal of solvent, which played a role in stand why SW would open differently upon excitation to SW*
complicating our analysis, as the LW species cannot be isolated vs reduction to SW•−. We turned to computation (of 3a, 4a, and
as solids.
5a) for an explanation. Due to large differences in the methods
for calculating excited and ground-states, the triplet T0 state
was used instead of the singlet S1 state as the photochemical
Computational analysis
As shown above, computationally predicted reduction poten- intermediate (SW*) for computational purposes. Since T0 is the
tials [7] were in very good agreement with those determined ex- lowest energy triplet state, it is amenable to a ground-state com-
perimentally (Table 1). The mean absolute deviation between putation. While the real SW* photochemical intermediate may
computational prediction and experimental measurement of likely be the first singlet excited state (S1), T0 and S1 possess
2480

Beilstein J. Org. Chem. 2019, 15, 2473–2485.
Scheme 3: Proposed mechanism for differential formation of pLW (4) and eLW (5) from SW (3).
equivalent orbital occupancy and differ only in their spin state. lowest unoccupied molecular orbital (LUMO), MO 106, of 3a
Our rationale for using T0 rather than S1 is that ignoring elec- are displayed in Figure 6. MO 105 allows for a direct compari-
tron-exchange interaction introduces substantially less error son of the T0 and D0 electron distributions, as these two states
than using different computational methods when comparing differ only in the population of this orbital. There is more
orbitals of SW* to those of SW and SW•−. Using S1 for SW* contribution from this MO to the N–C bond (hereafter desig-
would require unbalanced ground and excited-state calculations, nated the benzoN–C bond, cf. Scheme 3) of the nitrogen on the
e.g., time-independent and time-dependent density functional benzo ring of the quinoline moiety than the N–C bond of the
theories, or single-reference and multi-reference methods. How- nitrogen off the heterocyclic ring of the quinoline moiety (here-
ever, using T0 for SW* is a straightforward ground-state calcu-
lation, as are the calculations of D0 (the radical anion) and S0.
Clearly, as shown in Figure 7, the difference in reduction vs ex-
citation is the occupancy in what was formerly the HOMO of
SW. Because S0 is the common precursor to both T0 and D0,
analysis of the S0 MOs was used to explain differences in the
electron distributions that lead to two different products upon
either excitation to form SW* (yielding pLW 4) or reduction to
form SW•− (yielding eLW 5). This approach of using the S0
Figure 7: Frontier orbital occupancies of relevant electronic states of
MOs was validated computationally, as we found there is little
3a. Note: the photochemical excited state may be T0 (as pictured) or
change in the character and relative energies of the MOs among
S0, T0, and D0. The calculated isosurface generated for the S0
highest occupied molecular orbital (HOMO), MO 105, and
perhaps more likely S1 (inverted spin in molecular orbital 106 relative
to T0), but T0 is more computationally tractable and is equivalent to S1
for orbital occupancy arguments.
2481

Beilstein J. Org. Chem. 2019, 15, 2473–2485.
after referred to as the heteroN–C bond, cf. Scheme 3). This Table 2, heteroN–C has a greater bond order than benzoN–C in
asymmetrical MO contribution is consistent with experimental T0, while the relative magnitude of these bond orders is
observations. The absence of an electron from this orbital 105 reversed in D0. This is consistent with the formation of eLW 5a
in the case of T0 results in the preferential weakening of the from 3a•− D0 (breaking the heteroN–C bond) and pLW 4a from
benzoN–C bond over the heteroN–C bond. The formation of SW* 3a*, modelled as T0 (breaking the benzoN–C bond) as
pLW from SW (T0) results from the breaking of this weaker shown in the proposed mechanism in Scheme 3. Finally, bond
bond. More electron density in the benzoN–C bond in orbital lengths were derived from geometry optimizations for D0 and
105 might possibly suggest a stronger bond when the orbital is T0 with and without solvent and are also included in Table 2.
fully filled (as in D0), and could suggest that the heteroN–C There is an obvious inverse relationship between calculated
bond might be the more likely to rupture upon reduction to bond orders and bond lengths. For D0, a longer bond length is
SW•−.
observed for heteroN–C while a longer bond length is observed
for benzoN–C in T0. These results, regardless of solvent or gas-
It is common to think of thermal ring-opening as looking more phase considerations, are consistent with the discussion above.
like that of the biradical (T0 or S1) than of the radical anion In the cases of both D0 and T0, the longer bond is the one
(D0), so we might have presumed the structure of any thermal broken in the isomerization of SW 3a to eLW 5a and pLW 4a,
or solvatochromic LW isomer that occurred to have been that of respectively. Similar results are also observed for S0. Though
pLW (4). However, the same rationalization that more electron the differences are much more modest in S0, the heteroN–C
density in the benzoN–C bond in orbital 105 might possibly bond is computed to be both longer and weaker, as in D0, and
suggest a stronger bond relative to the heteroN–C bond when thus it makes sense that any thermal or solvatochromic LW is
the orbital is fully filled in S0 is how we rationalize our obser- the eLW isomer.
vation that it is eLW (5b) that occurs in the only case where we
see a purely thermal or solvatochromic cleavage. In turn we can Conclusion
then rationalize on the basis of sterics (R = Me vs H) that it is While moving from PSHD (7) to QSHD (3) increased ΔEored
logical that this thermal or solvatochromic rupture might only between SW and pLW isomers, and therefore capacity to gate
be observable in 3b (R = H) but not 3a (R = Me). This is in photoinduced charge transfer, by about 130 mV, the excited
good agreement with our experimental observations.
state reduction potential E*red of pLW 4 remains < +1.0 V vs
SCE, insufficient to oxidize most donor molecules of relevance
Our rationalization based on molecular orbitals was confirmed to materials applications. If our photochromic photooxidants are
by looking at both bond lengths and bond orders of the two to be effective in real systems, the pLW will need to be made
relevant N–C bonds in SW•− and SW*. Bond order calculations more electron deficient. Our computational predictions of the
provide a more quantitative explanation for the experimental reduction potential [22], unfortunately not completed until after
observation. Because bond stability correlates to bond order, it preparing the present system, were our first insight into the fact
is expected that the N–C bond which breaks upon forming the that electrochemical ring-opening was not yielding the struc-
LW product would possess a smaller bond order. The BO calcu- ture we had previously proven for the photochromic pLW. We
lations for intermediate states T0 and D0 indeed support this have thus demonstrated these computation’s practical utility in
prediction. Table 2 summarizes the results of BO calculations a real-world experimental situation, as well as their suitably
for both bonds in these states. Since geometry plays a critical high degree of accuracy. We are therefore hopeful they can help
role in the stability of bonds, only BO calculations done from guide our search for practical photochromic photooxidants, with
optimized geometries for each state are displayed. As shown in a more reducible pLW*.
Table 2: Computed bond lengths and bond orders of the relevant N–C bonds for photochromic (T0) and electrochromic (D0) ring-opening of 3a.
Bolded (longer and weaker) bond is the one that is cleaved in each case.
State
Bond
Bond order
Bond length (Å)
Consider for
S0
S0
benzoN–C
heteroN–C
0.9425
0.9346
1.478
1.478
thermal LW/
solvatochromism
D0
D0
benzoN–C
heteroN–C
0.8879
0.8738
1.508
1.513
electrochromism
photochromism
T0
T0
benzoN–C
heteroN–C
0.8143
0.8834
1.546
1.503
2482

Beilstein J. Org. Chem. 2019, 15, 2473–2485.
For now, our team of undergraduate researchers has conclu- pseudo-reference, the voltammetry was corrected to a ferrocene
sively demonstrated differential regiospecificity in the photo- external reference. Ferrocene solutions were cyclically scanned
chromic vs electrochromic spirocyclic ring-opening of these before and after each analyte experiment, beginning in the posi-
QSHD molecular switches to two different LW isomers. This tive (oxidative) direction over an 800 mV window roughly
may provide an interesting structural framework for molecular centered on the reduction of Fc/Fc+ redox couple at a scan rate
logic or other applications that complements the recently of 0.1–0.5 V/s, 10−4 A sensitivity, and 1 mV sample interval.
reviewed photoelectrochromic properties observed in a range of Internal resistance (iR) compensation was manually set to
photochromic families upon the electrochemical oxidation of 95–99.5% of the measured resistance so that the peak separa-
either their SW or pLW isomers [24]. We have also demon- tion for the reversible oxidation–reduction of ferrocene had a
strated that any modest amount of thermal (or solvatochromic) peak separation of ca. 60 mV and good peak shape without
coloration of the QSHDs is due to small amounts of the elec- entering oscillation. Typical iR compensation in acetonitrile
trochromic eLW isomer 5, rather than the photochromic pLW ranged from 150–230 Ω. Photochromic solutions were cycli-
(4) as we had previously surmised [21]. Finally, we have been cally scanned beginning in the negative (reductive) direction at
able to rationalize these results computationally on the basis of 0.1–0.5 V/s with 1 mV sample interval over an appropriate
bond lengths, bond orders, and molecular orbital occupancy.
range of potentials and current sensitivity to observe the redox
couples of either SW and LW or just LW isomers as desired,
without reaching the solvent breakdown limit. Reduction poten-
tials were taken as the half-peak potential of irreversible peaks
Experimental
Materials
Acetonitrile was of the highest HPLC grade (used as received or the midpoint of reversible peaks, standardized versus ferro-
or dispensed through a nitrogen-purged MBraun MBS-SPS cene (measured before and after each analyte sample), and
07-299 solvent purification system) or highest anhydrous grade corrected to versus SCE by adding 0.38 V [23]. At least seven
(used as received). Acetonitrile-d3 and acetone-d6 were used as replicates of each data point were obtained, with the mean value
received in 1 g ampules (Cambridge Isotope Labs). Ferrocene, reported with error bars indicating the standard deviation from
tetrabutylammonioum hexafluorophosphate (TBAH), and silver the mean among all replicates.
nitrate were of electrochemical grade and used as received.
Photochromes 1a,b were prepared according to the literature Potential step bulk electrolyses were performed on 3 mL argon-
[10]. Photochromes 3a,b were prepared as we previously re- purged acetonitrile or acetonitrile-d3 solutions of 2 mM analyte
ported [21].
and 0.1 M TBAH. The solution was placed in a glass cell pre-
pared by cutting the top off a 7 mL scintillation vial and
fireglazing the edges. The outer rubber of a 19/22 septum was
Instrumentation
Photochemical irradiations and UV–vis spectroscopy were per- removed with a razor blade and used to cap the vial. The rubber
formed as previously described [21] on 3 mL or 4 mL argon- had holes drilled for a reference and auxiliary electrode and a
purged acetonitrile solutions of 0.1 M TBAH supporting elec- slit cut for a 7 × 70 mm platinum mesh working electrode so
trolyte and 1–5 mM analyte, sufficient to attain signals much that the surface of the mesh faced the working and auxiliary
greater than background over the full potential window consid- electrodes. Teflon tubing was inserted through a pinhole in the
ered.
septum for an argon purge and vent. The setup used a CHI112
nonaqueous Ag/AgNO3 reference electrode. The auxiliary elec-
Cyclic voltammetry was performed on a CHI Model 604a elec- trode was formed by removing the silver wire from a CHI112
trochemical analyzer or a BAS epsilon e2 electrochemical nonaqueous reference electrode and replacing it with a rolled
analyzer. (The equivalent BAS or CHI cells and electrodes 7 × 70 mm piece of platinum mesh. The tube was filled with
could be used on either potentiostat interchangeably, with or supporting electrolyte. Electrolyses were run on 2 to 3 mL sam-
without the BAS C3 cell stand.) Solutions were placed in a ples, placing the platinum mesh electrode ca. 20 mm into the
glass cell and bubbled with argon to deaerate for 3–5 minutes. solution. To avoid the buildup of charge while maximizing cur-
Scans were taken under an argon blanket using a glassy carbon rent and efficiency, solutions were set to precondition before
working electrode, a platinum wire counter electrode, and a scanning the potential window with multiple repetitions run.
separately sparged Ag/AgNO3 (10 mM) nonaqueous reference Solutions of 1b were preconditioned at −2.5 V for 5 seconds
electrode isolated from the working compartment with a Vycor and 0 V for 6 seconds and 16–20 repetitions run. Solutions of
frit. In some cases, a bare Ag wire was used as a pseudo-refer- 3a,b were preconditioned at −2.2 V for 5 seconds and 0 V for
ence instead. The working, reference, and counter-electrodes 6 seconds with 16 repetitions. The electrolyzed solutions were
were arranged in a triangle through a Teflon cell cap. Rather opened to air or had a chemical oxidant (benzoquinone) added
than relying on either the Ag/AgNO3 reference or Ag wire to complete the reduction of the dianion to the LW isomer.
2483

Beilstein J. Org. Chem. 2019, 15, 2473–2485.
NMR spectroscopy was performed on samples 5 mm NMR (CPCM) as the self-consistent reaction field (SCRF) [37].
tubes (Wilmad) made of clear quartz (photolyzed samples) or It is from these implicit solvent-optimized geometries that re-
amber pyrex (dark samples) on a Varian Mercury or Bruker ported bond lengths were derived. Molecular orbital and bond
AvanceIII 400 MHz NMR. 1H NMR experiments on 5a were order calculations were subsequently performed on these
performed on an argon-purged, ca. 24 mM solution in acetone- geometries.
d6. 1D NOE spectra were collected using 400 manually inter-
leaved scans with a 4 s relaxation delay, targeting a single peak
per experiment, in a manner similar to that previously reported
Supporting Information
[21] for photogenerated 4a,b.
Supporting Information File 1
Additional figures and Z-matrices.
Computational modeling
[https://www.beilstein-journals.org/bjoc/content/
supplementary/1860-5397-15-240-S1.pdf]
Predicted ground-state reduction potentials were computed
based on the energy difference of the ground-state molecule
(S0) and its one electron-reduced (D0) radical anion and our
published correlation of this energy difference with experimen-
tal reduction potentials (vs SCE in acetonitrile) [22]. Gas-phase
geometries were optimized at the B3LYP/6-31G(d) level of
theory. Single point energies were computed at the same level
of theory with implicit acetonitrile solvent implemented
using the Conductor-like Polarizable Continuum Model
Supporting Information File 2
Complete data sets for experimental and computational
reduction potentials of all compounds.
[https://www.beilstein-journals.org/bjoc/content/
supplementary/1860-5397-15-240-S2.xlsx]
with UAKS radii. These computations were performed using Acknowledgements
the Gaussian 03 software package [25], implemented This work was funded by a Camille & Henry Dreyfus Founda-
through the WebMO [26] graphical user interface on the Curie tion Start-up Award (SU-04-040), a Cottrell College Science
cluster [27] in the Hope College Computational Science & Award from Research Corporation (CC6653), and a Faculty
Modeling Laboratory on a single node (a single 2.60 GHz AMD Early Career Development Grant (CHE-0952768) from the US
Opteron-252 processor with 8 GB RAM and 250 GB HD).
National Science Foundation (NSF). This work was also funded
in part by a grant to Hope College from the Howard Hughes
Calculations of bond length, bond order, and molecular orbitals Medical Institute through the Undergraduate Science Education
to rationalize the observed differential photochromic and elec- Program, which provided support for the Curie cluster [27] in
trochromic ring-opening of 1 to 2 and 3, respectively, were per- the Hope College Computational Science & Modeling (CSM)
formed on the Midwest Undergraduate Computation Chemistry Laboratory and to ALS (Computational Science & Modeling
Consortium (MU3C) cluster [28,29]. Computations were per- Scholars Program) and JPM (HHMI Research Scholars
formed on a single node (dual Intel X5650 CPU, with 6 cores Program). Additional computations were conducted on the
running at 2.66 GHz) using the Gaussian 09 [30] software MU3C [28,29] cluster, supported by Major Research Instru-
package implemented through the WebMO [26] graphical user mentation grants (CHE-0520704 and CHE-1039925) from the
interface. Restricted open-shell Hartree-Fock (ROHF) theory NSF, also housed in the Hope College CSM Laboratory.
[31] with the Becke 3, Lee, Yang, and Parr (B3LYP) hybrid Support from CSM Laboratory director Prof. Brent P. Krueger
functional [32-34] was used for geometry optimizations, molec- and former staff member Mr. Paul Van Allsburg is also grate-
ular orbitals, and bond orders calculations with the 6-31G(d) fully acknowledged. ALS acknowledges additional support
basis set [35] for open-shell species (T0 and D0), while conven- from a Jean Dreyfus Boissevain Scholarship awarded to the
tional B3LYP (with standard Hartree-Fock theory) was used for Hope College Chemistry Department by the Camille & Henry
closed-shell S0 calculations. The concerted use of ROHF and Dreyfus Foundation. EJL (home institution College of the
B3LYP provides a restricted open-shell B3LYP (DFT) method, Canyons, Santa Clarita, CA) acknowledges support by a
which was particularly important to obtaining good bond orders Research Experiences for Undergraduates Site award (CHE-
for open-shell (T0 and D0) species. Bond orders were calcu- 0851194) to the Hope College Chemistry Department from the
lated using the Natural Bond Order (NBO) 3.1 package [36] NSF.
contained within Gaussian 09. The geometry was first opti-
ORCID® iDs
mized for the S0 state in the gas phase. Based on this starting
point, geometries for S0, T0, and the one-electron reduced D0
states were then (re)optimized with an implemented
William F. Polik - https://orcid.org/0000-0003-0419-5135
Jason G. Gillmore - https://orcid.org/0000-0001-7506-3645
acetonitrile conductor-like polarizable continuum model
2484

Beilstein J. Org. Chem. 2019, 15, 2473–2485.
20.Roh, Y.; Gao, D.; Bauld, N. L. Adv. Synth. Catal. 2001, 343, 481–489.
doi:10.1002/1615-4169(200107)343:5<481::aid-adsc481>3.0.co;2-s
21.Moerdyk, J. P.; Speelman, A. L.; Kuper, K. E., III; Heiberger, B. R.;
Ter Louw, R. P.; Zeller, D. J.; Radler, A. J.; Gillmore, J. G.
J. Photochem. Photobiol., A 2009, 205, 84–92.
References
1. Crano, J. C.; Guglielmetti, R. J., Eds. Organic Photochromic and
Thermochromic Compounds; Plenum Press: New York, NY, USA,
1999; Vol. 1. doi:10.1007/b114211
2. Duerr, H.; Bouas-Laurent, H., Eds. Photochromism: Molecules and
Systems, revised ed.; Elsevier: Amsterdam, 2003.
3. Desvergne, J.-P.; Pozzo, J.-L., Eds. Proceedings of the fourth
International Symposium on Photochromism Photoswitchable
Molecular Systems and Devices. Mol. Cryst. Liq. Cryst. 2005, 430–431,
1–586. doi:10.1080/15421400590946073
4. Branda, N.; Tian, H., Eds. Organic Photoswitchable Multifunctional
Molecules and Materials. Dyes Pigm. 2011, 89, 193–336.
doi:10.1016/j.dyepig.2010.12.002
doi:10.1016/j.jphotochem.2009.04.011
22.Lynch, E. J.; Speelman, A. L.; Curry, B. A.; Murillo, C. S.;
Gillmore, J. G. J. Org. Chem. 2012, 77, 6423–6430.
doi:10.1021/jo300853k
23.Pavlishchuk, V. V.; Addison, A. W. Inorg. Chim. Acta 2000, 298,
97–102. doi:10.1016/s0020-1693(99)00407-7
24.Barachevsky, V. A.; Butenko, V. G. Russ. J. Gen. Chem. 2018, 88,
2747–2772. doi:10.1134/s1070363218120459
25.Gaussian 03, Revision D.01; Gaussian, Inc.: Wallingford, CT, 2004.
26.Schmidt, J. R.; Polik, W. F. WebMO Pro version 9.1, WebMO
Enterprise versions 10.1, 12.1, 14.0; WebMO LLC: Holland, MI, 2014;
available from http://www.webmo.net (accessed June 2019).
27.Curie Cluster. http://curie.chem.hope.edu/ (accessed June 30, 2019).
28.MU3C – Midwest Undergraduate Computational Chemistry
Consortium. http://mu3c.chem.hope.edu/ (accessed June 30, 2019).
29.Kuwata, K. T.; Kohen, D.; Krueger, B. P.; Polik, W. F.
Counc. Undergrad. Res. Quart. 2012, 32 (4), 9–14.
30.Gaussian 09, Revision A.1; Gaussian, Inc.: Wallingford, CT, 2009.
31.Roothaan, C. C. J. Rev. Mod. Phys. 1960, 32, 179–185.
doi:10.1103/revmodphys.32.179
32.Becke, A. D. J. Chem. Phys. 1996, 104, 1040–1046.
doi:10.1063/1.470829
33.Becke, A. D. Phys. Rev. A 1988, 38, 3098–3100.
doi:10.1103/physreva.38.3098
34.Lee, C.; Yang, W.; Parr, R. G. Phys. Rev. B 1988, 37, 785–789.
doi:10.1103/physrevb.37.785
35.Hariharan, P. C.; Pople, J. A. Theor. Chim. Acta 1973, 28, 213–222.
doi:10.1007/bf00533485
5. Favaro, G.; Irie, M., Eds. Special issue on Photochromism 1 and 2. J.
Photochem. Photobiol., C 2011, 12, 71–236.
doi:10.1016/j.jphotochemrev.2011.08.004
6. Monk, P. M. S.; Mortimer, R. J.; Rosseinsky, D. R. Electrochromism
and Electrochromic Devices; Cambridge University Press: Cambridge,
UK, 2007. doi:10.1017/cbo9780511550959
and references therein.
7. Baumann, K. L.; Lin, R.; Giri, P.; Franz, S. Protic-Soluble Organic
Electrochromic Compounds. U.S. Pat. Appl. US2017/0146880 A1, May
25, 2017.
8. Jarosz, T.; Gebka, K.; Stolarczyk, A.; Domagala, W.
Polymers (Basel, Switz.) 2019, 11, 273. doi:10.3390/polym11020273
and references therein.
9. Madasamy, K.; Velayutham, D.; Suryanarayanan, V.; Kathiresan, M.;
Ho, K.-C. J. Mater. Chem. C 2019, 7, 4622–4637.
doi:10.1039/c9tc00416e
10.Minkin, V. I.; Komissarov, V. N.; Kharlanov, V. A.
Perimidinespirocyclohexadienones. In Organic Photochromic and
Thermochromic Compounds; Crano, J. C.; Guglielmetti, R. J., Eds.;
Plenum Press: New York, NY, USA, 1999; Vol. 1, pp 315–340.
doi:10.1007/0-306-46911-1_9
36.Glendening, E. D.; Reed, A. E.; Carpenter, J. E.; Weinhold, F. NBO,
Version 3.1.
37.Tomasi, J.; Mennucci, B.; Cammi, R. Chem. Rev. 2005, 105,
2999–3094. doi:10.1021/cr9904009
and references therein.
11.Gillmore, J. G.; Neiser, J. D.; McManus, K. A.; Roh, Y.;
Dombrowski, G. W.; Brown, T. G.; Dinnocenzo, J. P.; Farid, S.;
Robello, D. R. Macromolecules 2005, 38, 7684–7694.
doi:10.1021/ma050348k
12.Kavarnos, G. J.; Turro, N. J. Chem. Rev. 1986, 86, 401–449.
doi:10.1021/cr00072a005
and references therein.
13.Borsub, N.; Kutal, C. J. Am. Chem. Soc. 1984, 106, 4826–4828.
doi:10.1021/ja00329a030
License and Terms
This is an Open Access article under the terms of the
Creative Commons Attribution License
14.Evans, T. R.; Wake, R. W.; Sifain, M. M. Tetrahedron Lett. 1973, 14,
701–704. doi:10.1016/s0040-4039(00)72438-5
15.Hasegawa, E.; Okada, K.; Ikeda, H.; Yamashita, Y.; Mukai, T.
J. Org. Chem. 1991, 56, 2170–2178. doi:10.1021/jo00006a039
16.Bauld, N. L. Hole and electron transfer catalyzed pericyclic reactions.
In Advances in Electron Transfer Chemistry; Mariano, P. S., Ed.; JAI
Press: Greenwich, CT, 1992; pp 1–66.
(http://creativecommons.org/licenses/by/4.0). Please note
that the reuse, redistribution and reproduction in particular
requires that the authors and source are credited.
The license is subject to the Beilstein Journal of Organic
Chemistry terms and conditions:
and references therein.
17.Bauld, N. L.; Gao, D.; Aplin, J. T. J. Phys. Org. Chem. 1999, 12,
808–818.
(https://www.beilstein-journals.org/bjoc)
The definitive version of this article is the electronic one
which can be found at:
doi:10.1002/(sici)1099-1395(199911)12:11<808::aid-poc207>3.0.co;2-
m
doi:10.3762/bjoc.15.240
18.Bauld, N. L.; Gao, D. Polym. Int. 2000, 49, 253–259.
doi:10.1002/(sici)1097-0126(200003)49:3<253::aid-pi352>3.0.co;2-p
19.Bauld, N. L.; Roh, Y. Tetrahedron Lett. 2001, 42, 1437–1439.
doi:10.1016/s0040-4039(00)02301-7
2485