Supporting Information:
Computational Design of Molecular Probes for
Electronic Pre-Resonance Raman Scattering
Microscopy
Jiajun Du,
†
,
‡
Xuecheng Tao,
†
,
‡
Tomislav Beguˇsi ́c,
∗
,
†
and Lu Wei
∗
,
†
†
Division of Chemistry and Chemical Engineering, California Institute of Technology,
Pasadena, California 91125, USA
‡
Contributed equally to this work
E-mail: tbegusic@caltech.edu; lwei@caltech.edu
S-1
1 Details of ab initio simulations
In the quantum-chemical calculations in ORCA, we used an efficient geometry pre-optimization
with a tight-binding semi-empirical potential
S1,S2
and included the molecule-solvent inter-
action (DMSO as solvent in our experiments) through an implicit polarizable continuum
model.
S3
Results from density functional theory (DFT) and time-dependent density func-
tional theory (TD-DFT)
S4
schemes at the hybrid functional level (PBE0/cc-pVDZ) are re-
ported. PBE0
S5,S6
was tested for the calculation of electronic transition energies and dipole
moments of the Rhodamine dyes
S7
and, independently, our benchmark calculations confirm
that the Raman intensity predictions are insensitive on whether pure
S8
or range-separated
S9
density functional is used instead. We also performed benchmark calculations to validate
the robustness of the electronic structure calculations with respect to basis set sizes and sol-
vent environments. The convergence criteria of the self-consistent field steps in all reported
calculations was set to ‘very tight’, meaning that an energy change lower than 10
−
9
Hartree
between the last two iterations was required. It is worthwhile mentioning that in Fig. 3 and
Table 3 of the main text, as well as in Table S1 below, we report the vibrational frequencies
without a conventional scaling factor correction, which is why they might differ from the
experimental frequencies by about +4%. Such small and systematic errors in the frequency
calculations do not significantly alter the simulated Raman intensities, which are the main
focus of this work. For efficiency, in PPCy-8a/PPCy-10a, we replaced the octyl groups by
the ethyl groups, assuming that the absorption and Raman properties are not strongly in-
fluenced by the length of the alkyl chain. The optimized structures and the orbitals were
visualized with the Avogadro software.
S10,S11
Table. S1 reports the computational quantities that were used for epr-SRS intensity
calculations, including vertical excitation wavelength, detuning frequency, transition dipole
strength, Raman mode displacement, Raman mode frequency, and finally the epr-SRS signal
strength.
S-2
Table S1: Computational quantities that were used for epr-SRS intensity cal-
culations. From left to right, the columns present the wavelength of the ex-
perimental absorption maximum (
λ
max
), vertical excitation wavelength (
λ
eg
,
exp
,
estimated from
λ
max
using eq 10 of the main text), detuning frequency (
δ
=
ω
eg
,
exp
−
ω
pump
), transition dipole strength (
|
μ
|
, as computed with the quantum
chemistry method), Raman mode displacement (
∆
̄
ν
, eq 6 of the main text), Ra-
man mode frequency (
ω
( ̄
ν
)
, as computed with the quantum chemistry method),
and finally the theoretical epr-SRS signal strength computed using the DHO
model (
I
theory
).
λ
pump
= 860 nm
epr-SRS probes
λ
max
/nm
λ
eg
,
exp
/nm
δ
/cm
−
1
|
μ
|
/au
∆
̄
ν
ω
( ̄
ν
)
/cm
−
1
I
theory
/RIE
9CN-MARS2222
790
771
1338
6.31
0.146
2348
∗
940
∗
MARS2228
760
740
1877
6.10
0.175
2354
381
MARS2225
760
743
1823
6.12
0.146
2352
284
MARS2231
744
726
2152
5.91
0.176
2358
215
MARS2237
700
679
3102
5.55
0.205
2364
70
9CN-MARS2238
690
668
3342
5.50
0.208
2366
55
MARS2233
735
717
2310
5.84
0.174
2359
159
9CN-MARS2240
675
654
3669
5.42
0.209
2368
38
MARS2190
731
716
2335
5.50
0.154
2291
89
Bodipy-alkyne
666
640
3993
6.04
0.063
2314
4
λ
pump
= 838 nm
epr-SRS probes
λ
max
/nm
λ
eg
,
exp
/nm
δ
/cm
−
1
|
μ
|
/au
∆
̄
ν
ω
( ̄
ν
)
/cm
−
1
I
theory
/RIE
MARS2228
760
740
1572
6.10
0.175
2354
∗
620
∗
MARS2231
744
726
1847
5.91
0.176
2358
321
MARS2237
700
679
2797
5.55
0.205
2364
114
FC10
694
665
3101
4.44
0.134
2364
11
PPCy-8a
690
662
3162
5.21
0.012
2320
0
PPCy-10a
692
657
3289
5.25
0.031
2341
1
λ
pump
= 842 nm
epr-SRS probes
λ
max
/nm
λ
eg
,
exp
/nm
δ
/cm
−
1
|
μ
|
/au
∆
̄
ν
ω
( ̄
ν
)
/cm
−
1
I
theory
/RIE
PADBP-9
699
661
3248
7.22
0.148
2310
89
MARS2190
731
716
2086
5.50
0.154
2291
83
C-MARS2190
696
681
2816
5.24
0.184
2297
54
C-MARS2143
696
681
2807
5.43
0.178
2257
51
* Normalized with respect to the experimentally measured epr-SRS intensity. See Fig. 2 in the
main text for details.
S-3
2 Experimental details
2.1 epr-SRS spectro-microscope setup
For the laser with a fundamental wavelength of 1031.2 nm:
An integrated laser
(picoEMERALD, Applied Physics and Electronics, Inc.) is used as a light source for both
pump and Stokes beams. It produces 2 ps pump (tunable from 770 nm – 990 nm, band-
width 0.5 nm, spectral bandwidth
∼
7 cm
−
1
) and Stokes (also called probe, 1031.2 nm,
spectral bandwidth 10 cm
−
1
) beams with 80MHz repetition rate. Stokes beam is modulated
at 20 MHz by an internal electro-optic modulator. The spatially and temporally overlapped
pump and Stokes beams are introduced into an inverted multiphoton laser scanning mi-
croscopy (FV3000, Olympus), and then focused onto the sample by a 25X water objective
(XLPLN25XWMP, 1.05 N.A., Olympus). Transmitted pump and Stokes beams are collected
by a high N.A. condenser lens (oil immersion, 1.4 N.A., Olympus) and pass through a band-
pass filter (893/209 BrightLine, 25 mm, AVR Optics) to filter out Stokes beam. A large area
(10
×
10 mm) Si photodiode (S3590-09, Hamamatsu) is used to measure the remaining pump
beam intensity. 64 V DC voltage is used on the photodiode to increase saturation thresh-
old and reduce response time. The output current is terminated by a 50 Ω terminator and
pre-filtered by an 19.2-23.6-MHz band-pass filter (BBP-21.4+, Mini-Circuits) to reduce laser
and scanning noise. The signal is then demodulated by a lock-in amplifier (SR844, Stanford
Research Systems) at the modulation frequency. The in-phase X output is fed back to the
Olympus IO interface box (FV30-ANALOG) of the microscope. Image acquisition speed is
limited by 30
μ
s time constant set for the lock-in amplifier. Laser powers are monitored
through image acquisition by an internal power meter and power fluctuation are controlled
within 5 % by the laser system. 16-bit grey scale images are acquired by Fluoview software.
The epr-SRS spectra are acquired by fixing the Stokes beam at 1031.2 nm and scanning the
pump beam through the designated wavelength range point by point. 10 mM aqueous EdU
sample is used as a standard to give RIE (Relative Intensity to EdU) of different molecule
S-4
probes. The 1 mM dye solution (DMSO) is used to acquire the epr-SRS spectra. To mini-
mize possible photobleaching, a relatively low power (20 mW on sample for pump laser and
30 mW on sample for Stokes laser) is used.
For the laser with a fundamental wavelength of 1064.2 nm:
A similar laser system
(picoEMERALD, Applied Physics and Electronics, Inc.) but with a different fundamental
wavelength of 1064.2 nm (80 MHz repetition rate) and 6 ps pulse width was used. The
intensity of the Stokes beam was modulated by a built-in EOM at 8 MHz. The other set-up
is the same as the previous 1031.2 nm fundamental wavelength laser system.
2.2 Absorption and fluorescence spectra measurements:
UV-Vis absorption spectra were recorded on a Varian Cary 500 UV-Vis Spectrophotometer
(Agilent). The 50
μ
M dye solution (DMSO) with 1 mm cuvette was used to acquire the
absorption spectra. In Figure S6, 10
μ
M dye solution (DMSO) with 5 mm cuvette was used
to get the absorption spectra. Fluorescence spectra were measured on a Shimadzu RF-6000
spectrofluorophotometer using a quartz microcuvette (Starna 18F-Q-10-GL14-S) with the
same dye solution (10
μ
M DMSO solutions). The obtained spectra were self-normalized.
2.3 Uncertainty characterization of epr-SRS measurement:
To examine the uncertainty of epr-SRS measurements, we performed epr-SRS measurements
of replicate samples with the 1031.2 nm fundamental laser system, we randomly chose nitrile
dye MARS2231 and alkyne dye MARS2190 and C-MARS2190 for characterization, whose
epr-SRS intensity spanned across different levels. The results (Figure S1) showed that the
variation of each measurement is quite small: within 10 %.
S-5
MARS2231
MARS2190
C-MARS2190
0
100
200
300
400
RIE
Figure S1: Uncertainty characterization of epr-SRS measurements of three dyes. Data is
shown as mean
±
SEM. 4–6 replicates were used for each dye.
3 Connection between short-time and Albrecht’s A-
term formulas
0
ξ
opt
1
2
3
ξ
=
s
/δ
1
2
3
I
(
̄
ν
),ST
0
→
1
/
I
(
̄
ν
),Albrecht
0
→
1
Figure S2: Ratio between short-time and Albrecht A-term Raman intensities as a function
of
ξ
=
s/δ
.
S-6
From eq 11 of the main text it follows that
I
( ̄
ν
)
,
ST
0
→
1
/I
( ̄
ν
)
,
Albrecht
0
→
1
=
1
ξ
4
Z
∞
0
te
−
t
2
/
2+
it/ξ
dt
2
= 1
, ξ
→
0
(1)
In other words, the Albrecht expression can be considered as a large detuning (
δ
≫
s
) limit
of the short-time formula. As shown in Figure S2, the dependence of
I
( ̄
ν
)
,
ST
0
→
1
/I
( ̄
ν
)
,
Albrecht
0
→
1
on
ξ
is not monotonic. Rather, an optimal value at around
ξ
opt
≈
0
.
43 is found. However, we note
that
s
, defined as
s
2
=
P
ν
s
2
ν
, depends also on the Raman mode of interest (labeled ̄
ν
in the
main text) and, specifically, that
s > s
̄
ν
. Therefore, if
ξ
̄
ν
=
s
̄
ν
/δ > ξ
opt
, the displacements
of all other modes would ideally be zero to minimize the difference
ξ
−
ξ
opt
. Otherwise, if
ξ
̄
ν
< ξ
opt
, other modes would ideally be displaced as to tune the value of
ξ
close to the
optimal one. Table S2 lists the values of
ξ
for molecules presented here.
Table S2: Values of
ξ
=
s/δ
for epr-SRS probes studied in this work.
epr-SRS probes (860 nm)
ξ
epr-SRS probes (
∼
840 nm)
ξ
9CN-MARS2222
0.44
MARS2228
0.41
MARS2228
0.35
MARS2231
0.36
MARS2225
0.33
MARS2237
0.28
MARS2231
0.31
FC10
0.27
MARS2237
0.25
PPCy-8a
0.21
9CN-MARS2238
0.24
PPCy-10a
0.24
MARS2233
0.29
PADBP-9
0.28
9CN-MARS2240
0.22
MARS2190
0.28
MARS2190
0.25
C-MARS2190
0.23
Bodipy-alkyne
0.21
C-MARS2143
0.22
4 Duschinsky and Herzberg-Teller effects
To analyze the potential Duschinsky and Herzberg-Teller effects, we modeled the absorption
spectra of the MARS2231, MARS2233, MARS2237, and C-MARS2190 probes.
First, the absorption spectra were simulated within the vertical gradient (VG) model,
which is a version of the DHO model, i.e., uses the same force-constant matrix for the ground
S-7
12000
14000
16000
18000
20000
22000
24000
0.0
0.5
1.0
1.5
MARS2231
Exp
VG
AH
AH without Duschinsky
12000
14000
16000
18000
20000
22000
24000
0.00
0.25
0.50
0.75
1.00
MARS2233
12000
14000
16000
18000
20000
22000
24000
0.00
0.25
0.50
0.75
1.00
MARS2237
12000
14000
16000
18000
20000
22000
24000
Wavenumber / cm
−
1
0.00
0.25
0.50
0.75
1.00
C-MARS2190
Normalized intensity
Figure S3: Experimental and simulated electronic absorption spectra of MARS2231,
MARS2233, MARS2237, and C-MARS2190. Simulations were based on the vertical gra-
dient (VG) method, adiabatic Hessian (AH, including Duschinsky rotation) method, and
AH method without Duschinsky rotation (by setting Duschinsky matrix
J
to identity, op-
tion
USEJ FALSE
in Orca). All simulations were performed using the Orca
ESD
module,
S12
at room temperature (298.15 K), and within the Condon approximation for the transition
dipole moment. To simplify the comparison of the lineshape, the simulated spectra were
scaled and shifted to match the experiment at the 0-0 transition. A Gaussian broadening
function with
σ
= 280 cm
−
1
was used for VG and “AH without Duschinsky” simulations,
while
σ
= 150 cm
−
1
was used for AH simulations.
S-8
and excited electronic states, and within the adiabatic Hessian (AH) model, in which the
excited-state geometry is optimized and new vibrational modes and frequencies are evalu-
ated there. Specifically, the VG model neglects any changes in frequencies or normal-mode
(Duschinsky) rotation effects, whereas the AH model fully accounts for these effects. Im-
portantly, both of these models neglect anharmonicity of the true potential energy surfaces.
Figure S3 shows that the VG model is more reliable and robust for the studied molecules
due to the incorrect overestimation of the Duschinsky coupling between the excited-state
modes in the AH model (compare AH with “AH without Duschinsky”). Although the AH
approach is more accurate for MARS2233, it yields very inaccurate absorption spectra for
the other three molecules. Hence, the results justify our choice of the VG model over the
more expensive and less robust AH harmonic model.
Second, we turn to the non-Condon effects due to the dependence of the transition
dipole moment on the atomic coordinates. The first-order, Herzberg-Teller correction to the
Condon approximation accounts for the linear dependence of the transition dipole moment
on the coordinates and is often sufficient to estimate the degree of non-Condon effects. This
correction is typically necessary only in weakly allowed or symmetry-forbidden transitions. In
Figure S4, we compare the simulations within Condon and Herzberg-Teller approximations
with the experimental spectrum of MARS2231 and C-MARS2190, demonstrating that the
Herzberg-Teller effect is negligible.
5 Analysis of the 0-0 transition energy
Here, we motivate our interpretation of the absorption maximum as the 0-0 vibronic transi-
tion, as described in the Computational details of the main text. For molecules with a large
Stokes shift, the absorption maximum frequency (
ω
max
) typically coincides with the vertical
excitation frequency (
ω
eg
), which, in those cases, is very far from the 0-0 transition frequency
(
ω
0
−
0
). Fig. S6 clearly shows that the molecules considered in this work do not fall into this
S-9
12000
14000
16000
18000
20000
22000
24000
0.0
0.2
0.4
0.6
0.8
1.0
MARS2231
Exp
Condon
Herzberg-Teller
12000
14000
16000
18000
20000
22000
24000
Wavenumber [cm
−
1
]
0.0
0.2
0.4
0.6
0.8
1.0
C-MARS2190
Normalized intensity
Figure S4: Experimental and simulated electronic absorption spectra of MARS2231 and C-
MARS2190 (only the lowest-energy electronic transition is modeled), where the simulations
were based on the VG model and used either Condon or Herzberg-Teller approximation for
the transition dipole moment. See Fig. S3 for further details.
category, since their Stokes shifts are only modest (
<
900 cm
−
1
). For molecules with such
small Stokes shifts, the absorption maximum frequency is much closer to the 0-0 transition
frequency than to the vertical excitation frequency. Still, their 0-0 transitions do not neces-
sarily correspond to the respective absorption maxima. A more accurate estimate would be
to use the frequency of intersection of absorption and emission spectra, which follows from
the theory of vibronic transitions. However, this approach would require not only access
to absorption, but also emission spectra. In addition, some prospective probes might not
fluoresce due to a fast nonradiative relaxation, which would complicate the determination
S-10
of the 0-0 transition energy. For these reasons, we decided to use the absorption maximum
frequency as a reasonable estimate for the 0-0 transition energy. Here, we show that this
specific choice does not significantly affect our results.
For convenience, let us denote by “Method A-F” the approach in which we use the
intersection of absorption and fluorescence spectra to determine the 0-0 transition and by
“Method A” the approach in which only the absorption maximum is used. Table S3 shows
the detunings computed using these two methods to determine the 0-0 vibronic transition
frequency
ω
0
−
0
and using eq 10 of the main text to obtain the DHO-based estimate of the
vertical excitation energy. We then used these two sets of detunings to compute the Raman
intensities, whose values reflect the fact that the A-F detunings are red-shifted compared to
the Method A detunings. However, when comparing to experimental values, all values are
scaled to the highest Raman intensity (MARS2228). Fig. S5 shows that both methods result
in a similar agreement between theory and experiment. In contrast to Method A-F, Method
A allows us to model all molecules on the same footing, including those for which we might
not have access to fluorescence spectra.
Table S3: Comparing epr-SRS intensities from two simulation protocols, one
using the absorption maximum frequency (Method A) as the 0-0 transition fre-
quency, the other using the intersection of absorption and fluorescence spectra
(Method A-F) as the 0-0 transition.
Method A
Method A-F
epr-SRS probes
pump/nm
δ
/cm
−
1
I
A
/arb.u.
δ
/cm
−
1
I
A
−
F
/arb.u.
I
A
/I
A
−
F
MARS2228
838
1572
2.24e+05
1435
3.11e+05
0.72
MARS2231
838
1847
1.16e+05
1651
1.70e+05
0.68
MARS2237
838
2797
3.22e+04
2576
4.27e+04
0.75
PPCy-10a
838
3289
3.21e+02
3062
4.09e+02
0.78
C-MARS2190
842
2816
1.83e+04
2473
2.83e+04
0.65
C-MARS2143
842
2807
1.95e+04
2544
2.71e+04
0.72
MARS2190
842
2086
4.12e+04
1758
7.28e+04
0.57
S-11
MARS2228
MARS2231
MARS2237
PPCy-10a
C-MARS2190
C-MARS2143
MARS2190
0
200
400
600
epr-SRS Intensity / RIE
(normalized)
Experiment
Method A
Method A-F
Figure S5: Comparing simulated intensities from method A/A-F to the measured values.
Both methods result in a similar agreement between theory and experiment.
6 Dipole strengths: Simulation vs. experiment
Experimental dipole strengths were determined according to
S13
μ
2
exp
=
3
ℏ
cε
0
π
Z
∞
−∞
σ
(
ω
)
ω
dω,
(2)
where the experimental absorption cross section was obtained from the molar extinction
coefficient using
σ
(
ω
) = ln(10)
ε
(
ω
)
/N
A
.
The results (Table S4) show that the simulated transition dipole moments are reliable as
they only differ from the experimental values by approximately a constant factor.
S-12
Figure S6: Normalized absorption and fluorescence spectra of seven representative epr-SRS
probes. The intersection wavelength between absorption and emission sepctra is labeled on
each graph.
S-13
Table S4: Magnitudes of the experimental and simulated transition dipole mo-
ments in atomic units.
epr-SRS probes
|
μ
exp
| |
μ
sim
|
C-MARS2190
3.01
5.24
MARS2190
3.26
5.50
9CN-MARS2238
3.35
5.50
9CN-MARS2240
3.54
5.42
MARS2237
3.79
5.55
MARS2231
3.89
5.91
MARS2228
3.57
6.10
PPCy-10a
3.40
5.25
References
(S1) Grimme, S.; Bannwarth, C.; Shushkov, P. A robust and accurate tight-binding quan-
tum chemical method for structures, vibrational frequencies, and noncovalent interac-
tions of large molecular systems parametrized for all spd-block elements (Z= 1–86).
J. Chem. Theory Comput.
2017
,
13
, 1989–2009.
(S2) Bannwarth, C.; Ehlert, S.; Grimme, S.; Tight-Binding, B. P. S.-C. Quantum Chemical
Method with Multipole Electrostatics and Density-Dependent Dispersion Contribu-
tions.
J. Chem. Theory Comput.
2019
,
15
, 1652–1671.
(S3) Barone, V.; Cossi, M. Quantum calculation of molecular energies and energy gradients
in solution by a conductor solvent model.
J. Phys. Chem. A
1998
,
102
, 1995–2001.
(S4) Petersilka, M.; Gossmann, U.; Gross, E. Excitation energies from time-dependent
density-functional theory.
Physical review letters
1996
,
76
, 1212.
(S5) Perdew, J. P. Density-functional approximation for the correlation energy of the in-
homogeneous electron gas.
Phys. Rev. B
1986
,
33
, 8822.
S-14
(S6) Adamo, C.; Barone, V. Toward reliable density functional methods without adjustable
parameters: The PBE0 model.
J. Chem. Phys.
1999
,
110
, 6158–6170.
(S7) Zhou, P. Why the lowest electronic excitations of rhodamines are overestimated by
time-dependent density functional theory.
Int. J. Quantum Chem.
2018
,
118
, e25780.
(S8) Becke, A. D. Density-functional exchange-energy approximation with correct asymp-
totic behavior.
Phys. Rev. A
1988
,
38
, 3098.
(S9) Iikura, H.; Tsuneda, T.; Yanai, T.; Hirao, K. A long-range correction scheme for
generalized-gradient-approximation exchange functionals.
J. Chem. Phys.
2001
,
115
,
3540–3544.
(S10) Avogadro: an open-source molecular builder and visualization tool.
Version 1.93.0. http://avogadro.cc/.
(S11) Hanwell, M. D.; Curtis, D. E.; Lonie, D. C.; Vandermeersch, T.; Zurek, E.; Hutchi-
son, G. R. Avogadro: an advanced semantic chemical editor, visualization, and anal-
ysis platform.
J. Cheminform.
2012
,
4
, 17.
(S12) Neese, F.; Wennmohs, F.; Becker, U.; Riplinger, C. The ORCA quantum chemistry
program package.
J. Chem. Phys.
2020
,
152
, 224108.
(S13) Lewis, J. E.; Maroncelli, M. On the (uninteresting) dependence of the absorption and
emission transition moments of coumarin 153 on solvent.
Chem. Phys. Lett.
1998
,
282
, 197–203.
S-15