Tuning the Vibrational Relaxation of CO Bound to Heme and Metalloporphyrin Complexes

Article abstract of DOI:10.1021/jp961418f

Picosecond mid-infrared pump-probe experiments were used to investigate vibrational relaxation (VR, which here denotes loss of excess vibrational energy) of CO bound to synthetic heme and porphyrin complexes with different metal atoms (M = Fe, Ru, Os) and different proximal ligands (imidazoles and pyridines).Isotope effects of 13CO vs 12CO and solvent effects were also studied.A remarkable correlation between the carbonyl vibrational lifetime and the carbonyl vibrational frequency ν_CO is observed.The vibrational lifetime decreases as ν_CO decreases.The lifetime-frequency correlation is consistent with a linear reaction between carbonyl vibrational relaxation rate and ν_CO.Hemes and porphyrins show similar lifetime-frequency correlations, but the absolute value of the VR rate in meso-tetraphenylporphyrin complexes is slightly faster than in protoporphyrin IX dimethyl ester heme complexes.The predominant VR process is shown to be intramolecular transfer from CO to heme vibrations, rather than intermolecular transfer from CO to solvent vibrations.The intramolecular process occurs by anharmonic coupling via ?-bonding between CO and the metalloporphyrin and heme.In metalloporphyrin and heme complexes, changes in back-bonding to CO simultaneously affect both CO frequency and the strength of anharmonic coupling, accounting for the observed lifetime-frequency correlation.Increasing back-bonding lowers the CO frequency and increases the anharmonic coupling, shortening the vibrational lifetime.Similar lifetime-frequency correlations are observed in wild-type and mutant heme proteins.It is possible to continuously tume the vibrational relaxation rate of CO over a range spanning about a factor of 4, by systematic modification of the chemical structure of the heme or porphyrin complex to which it is bound.

Full text of DOI:10.1021/jp961418f

J. Phys. Chem. 1996, 100, 18023-18032
18023
ARTICLES
Tuning the Vibrational Relaxation of CO Bound to Heme and Metalloporphyrin Complexes
Jeffrey R. Hill, Christopher J. Ziegler, Kenneth S. Suslick, and Dana D. Dlott*
School of Chemical Sciences, UniVersity of Illinois at Urbana-Champaign, 505 S. Mathews AVenue,
Urbana, Illinois 61801
C. W. Rella
Hansen Experimental Physics Laboratory, Stanford UniVersity, Stanford, California 94305
M. D. Fayer*
Department of Chemistry, Stanford UniVersity, Stanford, California 94305
ReceiVed: May 16, 1996; In Final Form: August 7, 1996X
Picosecond mid-infrared pump-probe experiments were used to investigate vibrational relaxation (VR, which
here denotes loss of excess vibrational energy) of CO bound to synthetic heme and porphyrin complexes
with different metal atoms (M ) Fe, Ru, Os) and different proximal ligands (imidazoles and pyridines).
Isotope effects of 13CO vs 12CO and solvent effects were also studied. A remarkable correlation between the
carbonyl vibrational lifetime and the carbonyl vibrational frequency νCO is observed. The vibrational lifetime
decreases as νCO decreases. The lifetime-frequency correlation is consistent with a linear relation between
carbonyl vibrational relaxation rate and νCO. Hemes and porphyrins show similar lifetime-frequency
correlations, but the absolute value of the VR rate in meso-tetraphenylporphyrin complexes is slightly faster
than in protoporphyrin IX dimethyl ester heme complexes. The predominant VR process is shown to be
intramolecular transfer from CO to heme vibrations, rather than intermolecular transfer from CO to solvent
vibrations. The intramolecular process occurs by anharmonic coupling via π-bonding between CO and the
metalloporphyrin or heme. In metalloporphyrin and heme complexes, changes in back-bonding to CO
simultaneously affect both CO frequency and the strength of anharmonic coupling, accounting for the observed
lifetime-frequency correlation. Increasing back-bonding lowers the CO frequency and increases the
anharmonic coupling, shortening the vibrational lifetime. Similar lifetime-frequency correlations are observed
in wild-type and mutant heme proteins. It is possible to continuously tune the vibrational relaxation rate of
CO over a range spanning about a factor of 4, by systematic modification of the chemical structure of the
heme or porphyrin complex to which it is bound.
I. Introduction
surroundings. In broad terms, there are three ways this might
occur,4 as illustrated schematically in Figure 1. VR might be
a purely intramolecular process, involving energy transfer from
the CO vibration to heme or porphyrin vibrations, via the
metalloporphyrin-CO bonds. We will call this process CO-
to-porphyrin transfer. There are both σ- and π-orbitals involved
in metalloporphyrin-CO bonding; the π-bonds are formed by
back-bonding, i.e., back-donation from the metalloporphyrin dπ
and pπ orbitals to the π*-antibonding orbitals of CO.8 Alter-
natively, carbonyl VR might be a purely intermolecular process,
involving energy transfer from the CO vibration to solvent
vibrations, via nonbonded interactions. We will call this process
CO-to-solvent transfer. The CO-to-porphyrin or CO-to-solvent
processes might also involve excitations of the low-frequency
collective vibrations of the solvent (instantaneous normal modes
or solvent phonons9). These processes may be distinguished
by systematically varying the porphyrin structure, the solvent,
the temperature, etc. The series of complexes studied here are
variations on the canonical heme protein structure found in
myoglobin, hemoglobin, and peroxidase, where the porphyrin
is iron(II) protoporphyrinate IX (protoheme), the proximal axial
ligand is the imidazole group of the amino acid histidine, and
In order to better understand the relationships between
molecular structure and vibrational dynamics, we have inves-
tigated the rates of vibrational relaxation (VR) of carbon
monoxide (CO) bound to the active site of synthetic heme and
porphyrin complexes. A brief communication of some of this
work appeared recently.1 Here VR is used to denote Vibrational
energy relaxation, the loss of energy from vibrationally excited
CO to the rest of the system. VR rates were determined using
picosecond mid-infrared (mid-IR) pump-probe experiments.
We focus our attention on heme and metalloporphyrin com-
plexes, because well-developed techniques of synthetic chem-
istry and genetic engineering permit systematic control over the
structure, both in these synthetic analogs of heme proteins and
in heme proteins themselves. In addition, these studies are
expected to provide additional insight into how heme protein
structure influences molecular dynamics at the active site of a
protein.2-6
Loss of vibrational energy from CO involves anharmonic
coupling7 between the carbonyl vibrational fundamental and its
X Abstract published in AdVance ACS Abstracts, November 1, 1996.
S0022-3654(96)01418-9 CCC: $12.00 © 1996 American Chemical Society

18024 J. Phys. Chem., Vol. 100, No. 46, 1996
Hill et al.
Figure 2. Some representative pump-probe data obtained on a
porphyrin complex, Os(TPP)(CO)(L) in CH2Cl2 solvent. TPP denotes
5,10,15,20-tetraphenylporphyrin. The smooth curves through the data
are fits to single-exponential functions with time constants of 14.8 and
10.9 ps for proximal ligands L ) 4-cyanopyridine and 2-methylimid-
azole. All the decay constants obtained in this work are given in Table
1.
Figure 1. Schematic diagram of CO bound to a heme or porphyrin
complex in a solvent. M denotes a metal atom. Loss of vibrational
energy from CO excited by a mid-IR picosecond laser pulse might
involve intramolecular CO-to-porphyrin vibration processes, intermo-
lecular CO-to-solvent vibration processes, or a mixed process involving
intra- or intermolecular vibrations and solvent phonons. CO-to-
porphyrin processes must depend on metal-to-CO bonding. M-CO
bonding involves both σ- and π-orbitals. The M-CO π-bonds are
formed by back-donation of electrons from the metalloporphyrin
macrocycle to antibonding π* orbitals of CO.
detail in ref 11. For brevity, we will use the following
abbreviations throughout: coproporphyrinate I tetraisopropyl
ester ) COPRO, 5,10,15,20-tetraphenylporphyrin ) TPP,
protoporphyrinate IX dimethyl ester ) PHDME, and protopor-
phyrinate IX ) PPIX. Fe porphyrins and free-base porphyrins
were purchased from Aldrich Chemical. Fe(PHDME)(Cl) and
Fe(TPP)(Cl) were dissolved in CH2Cl2, and a 2-fold excess of
the appropriate proximal ligand was added. The Fe was reduced
with sodium dithionite, and CO was added. The Ru and Os
porphyrin carbonyls were synthesized from free-base porphyrins
using Ru3(CO)12 and Os3(CO)12.12,13 Crystals of the Ru and
Os porphyrin carbonyls are stable in air. These crystals could
be dissolved in various solvents containing an excess of the
desired proximal ligand. Fourier transform IR (FTIR) spectra
were measured to verify purity and to determine νCO. All the
compounds used here had a single dominant conformer, as
evidenced by the appearance of a single carbonyl stretching
transition (e.g., see Figure 6a). The 13CO forms of Ru(COPRO)
and Os(COPRO) were synthesized from the 12CO compounds,
by irradiating solutions of the 12CO carbonyls under an
atmosphere of 13CO with visible light. FTIR was used to verify
that the exchange of 13CO for 12CO was complete (cf. Figure
6a).
the “solvent” is the surrounding protein.10 We varied the
structure of the complex using different metal atoms M in an
isoelectronic series (M ) Fe, Ru, Os), different porphyrins
(which are the ligands cis to the CO), and different proximal
(trans) ligands. In addition, experiments were performed using
different solvents and isotopic substitution with 13CO.
In previous work, mid-IR pump-probe experiments were
used to investigate VR of CO bound to wild-type myoglobin
(Mb),2-6 mutant myoglobins,3 and a heme complex containing
either Fe, Ru, or Os.1,11 These experiments suggested the
possibility of a correlation between νCO, the fundamental
frequency of the carbonyl stretching transition, and the VR
lifetime. The observed trends were rationalized by proposing
the dominant mechanism of VR to involve intramolecular
anharmonic coupling from CO to heme, via the heme-CO
π-bonds, rather than the heme-CO σ-bonds. We argued that
the dominant mechanism was “through π-bond coupling” rather
than “through σ-bond coupling”.1,11 Structural factors that
increase back-bonding decrease both νCO and the VR lifetime
by increasing the through π-bond coupling.1-3,11 In the prior
study, only three variants of a single heme complex11 were
studied, so our proposed correlation was speculative but not
definitive. In this work, we have investigated a large number
of heme and porphyrin complexes, and we find a remarkable
correlation between the vibrational lifetime and the carbonyl
stretching frequency.1 By modifying the complex structure,
through changing the heme or porphyrin, the metal atom, or
the proximal axial ligand, it becomes possible to tune the
vibrational relaxation lifetime of CO over about a factor of 4.
Never before has virtually continuous control of a vibrational
lifetime been observed in a condensed phase polyatomic
molecule.1
Pump-probe measurements were performed at the Stanford
Free Electron Laser Center. Details of the free electron laser
(FEL), the experimental apparatus, and an explanation of
pump-probe experiments are presented in ref 4. The pump-
probe experiment measures the vibrational lifetime of the CO
stretching mode,1,2,14-17 which for the compounds studied here
lies in the 1880-1980 cm-1 range (5.05-5.38 µm).
III. Results
All the pump-probe decay data obtained in this study were
well fit by a single-exponential function of the form S(t) ) S0
exp(-t/T1), where T1 is the lifetime of vibrationally excited (V
) 1) CO bound to porphyrin. For example, Figure 2 shows
some representative pump-probe data obtained on a porphyrin,
Os(TPP)(CO)(L) in CH2Cl2 solvent, with two different proximal
ligands (L). Table 1 lists all the vibrational lifetimes measured
in this work. The error bounds on the values of T1 are estimated
at ((1-2) ps. Also given in the table are the carbonyl stretching
II. Experimental Section
The preparation of the heme and porphyrin compounds used
here followed standard literature methods, similar to those for
preparing the coproporphyrinate I compounds described in more
frequencies. The errors in frequency, estimated at (1 cm-1
,

Tuning the Vibrational Relaxation of CO
J. Phys. Chem., Vol. 100, No. 46, 1996 18025
TABLE 1: Vibrational Lifetimes and Vibrational Frequencies of CO Bound to Porphyrin Compounds. The Solvent Is CH2Cl2
unless Otherwise Noted
protoporphyrin IX
tetraphenyl porphyrin (TPP)
dimethyl ester (PHDME)
carbonyl
vib relaxation
lifetime T1 (ps)
carbonyl
vib relaxation
lifetime T1 (ps)
freq (cm-1
)
freq (cm-1
)
Fe Compounds
pyridine
4-aminopyridine
1975
1969
32.5
35.6
1968
1963
43.3
44.5
Ru Compounds
pyridine
1943
1938
1948
1936
1936
1958
15.6
16.6
18.3
16
15
16.5
1934
1930
1943
1927
1927
1946
17.3
12.3
18.5
15.5
19.2
13
4-aminopyridine
4-cyanopyridine
imidazole
2-methylimidazole
triphenylphosphine
Os Compounds
pyridine
1912
1903
1918
1899
1897
1928
12.6
12.2
14.8
11.1
10.9
10.7
1898
1891
1903
1887
1887
1893
10.6
12
12.9
11.6
11.6
12.4
4-aminopyridine
4-cyanopyridine
imidazole
2-methylimidazole
triphenylphosphine
Solvent Effect on Ru(TPP)(CO)(pyridine)
CH2Cl2
CHCl3
CCl4
CH3CCl3
1943
1937
1948
1949
1949
15.6
15.8
19
16.1
11.5
dibutyl phthalate (DBP)
Isotope Effects
Ru(COPRO) (pyridine)(13CO)
Os(COPRO) (pyridine)(13CO)
1880
1860
19.5
12.9
Figure 4. Carbonyl vibrational relaxation (VR) rate versus carbonyl
stretching frequency νCO for a series of protoheme dimethyl ester
compounds M(PHDME)(CO)(L) in CH2Cl2 solvent. The proximal
ligands L ) pyridine (O), imidazole (4), 2-methylimidazole (b),
4-aminopyridine (9), 4-cyanopyridine (0), and triphenylphosphine (g).
The dashed line is a least-squares fit, neglecting the triphenylphosphine
data.
Figure 3. Carbonyl vibrational relaxation (VR) rate versus carbonyl
stretching frequency νCO for a series of tetraphenylporphyrin complexes
M(TPP)(CO)(L) in CH2Cl2 solvent. The metal atom M is Fe, Ru, or
Os as indicated. The proximal ligands L ) pyridine (O), imidazole
(4), 2-methylimidazole (b), 4-aminopyridine (9), 4-cyanopyridine (0),
and triphenylphosphine (g). The dashed line is a least-squares fit,
neglecting the triphenylphosphine data.
are not instrument-limited. Instead, they arise from uncertainties
in determining the peak frequencies of transitions with spectral
widths in the 10-20 cm-1 range.
decreases as νCO decreases. As discussed later, the two
complexes with triphenylphosphine ((C6H5)3P) as the axial
ligands, denoted by stars in Figure 3, are exceptions to the
observed correlation. Neglecting these two data points, the data
in Figure 3 are well fit by a linear relationship, at least over
this frequency range. The dashed line in Figure 3 is obtained
using linear least-squares fitting, excluding the triphenylphos-
phine data.
a. TPP and PHDME. In Figure 3, we plot the VR rate
versus carbonyl stretching frequency for TPP complexes. These
data refer to M(TPP)(CO)(L) in CH2Cl2 solution. We used three
different metal atoms (M ) Fe, Ru, Os) and six proximal
ligands. A total of 14 different compounds were studied. The
data on these compounds in Figure 3 show a high degree of
correlation between the VR rate and the carbonyl stretching
frequency νCO. The VR rate increases as νCO decreases. This
relationship will be referred to as a lifetime-frequency cor-
relation. The lifetime (inverse of the VR rate constant)
Figure 4 shows the PHDME data. These data refer to
M(PHDME)(CO)(L) complexes in CH2Cl2 solution. A strong
correlation between VR rate and νCO is again observed with
the exception of the compounds with triphenylphosphine

18026 J. Phys. Chem., Vol. 100, No. 46, 1996
Hill et al.
Figure 5. Solvent effects on the carbonyl vibrational relaxation (VR)
rate of Ru(TPP)(CO)(pyridine). The solvents were CCl4 (b), CHCl3
(9), CH2Cl2 (O), CH3CCl3 (0), and dibutylphthalate (f). The dashed
line is fit to the TPP data in Figure 3.
Figure 6. Isotope effect data on a coproporphyrin compound,
M(COPRO)(CO)(pyridine). (a) Infrared spectra of 12CO and 13CO bound
to Ru(COPRO)(pyridine), showing a sizable isotope frequency shift.
(b) The pump-probe decays for both 12CO and 13CO species are shown.
The decays are virtually indistinguishable, showing isotopic substitution
has no appreciable effect on the VR rate.
proximal ligands. Neglecting the triphenylphosphine data, linear
least-squares fitting was used to determine the dashed line in
Figure 4.
In both TPP and PHDME, the vibrational lifetimes are quite
sensitive to the value of νCO. Over a range (∼1900-1980 cm-1
)
where νCO changes by ∼4%, the CO vibrational lifetimes range
over a factor of ∼4 (the longest lifetimes are ∼45 ps and the
shortest ∼11 ps).
In comparing the TPP and PHDME data, the slopes of the
best-fit lines through the data are identical to within experimental
error; however, the intercepts are slightly different. In our
analysis, this difference appears to be statistically significant.
At a given frequency, the VR rates of TPP compounds are on
average about 10% faster than the VR rates of PHDME
compounds.
b. Solvent Effects. The effects of different solvents on
Ru(TPP)(CO)(pyridine) were studied. The solvents were CCl4,
CHCl3, CH2Cl2, CH3CCl3, and dibutyl phthalate (DBP). There
are small CO frequency shifts in the different solvents. Relative
to the 1943 cm-1 value in CH2Cl2 solvent, which was the solvent
used in all the other experiments, νCO is blue-shifted by 5-6
cm-1 in CCl4, CH3CCl3, and DBP, and νCO is red-shifted 6 cm-1
in CHCl3 (see Table 1). The VR lifetimes in three chlorinated
hydrocarbons CHCl3, CH3CCl3, and CH2Cl2 were almost
identical (Figure 5), ∼16 ps. In CCl4 the lifetime is slightly
longer, 19 ps. The effects on VR of these four solvents are
smaller than changing the porphyrin from TPP to PHDME. In
contrast, DBP has a noticeable influence on the VR rate. In
DBP, the VR lifetime of 11.5 ps is ∼40% shorter than in
CH2Cl2. As shown in Figure 5, there is no apparent correlation
between the relatively small solvent-induced frequency shifts
and changes in the VR lifetimes.
c. Isotope Effects. The effects of substituting 13CO for 12CO
are shown in Figures 6 and 7. We studied M(COPRO)(CO)-
(pyridine) in CH2Cl2 solvent, where M ) Ru, Os. Figure 6
shows data on the Ru compound. The mid-IR spectrum in
Figure 6a shows that 13CO produces a substantial frequency
red shift. The pump-probe data in Figure 6b show the pump-
probe decays of the 12CO and 13CO compounds are virtually
indistinguishable, in spite of the substantial frequency shift.
Figure 7 is a plot of the isotope effect data on Ru and Os
compounds versus νCO. Based on the dependence of VR rate
on νCO seen in Figures 3 and 4, the CO frequency reduction
due to 13CO would be expected to roughly double the VR rate.
Figure 7 shows instead that the effect of 13CO on VR rate is
Figure 7. Plot of isotope effects on carbonyl VR using 12CO and 13CO
forms of Ru(COPRO)(CO)(pyridine) and Os(COPRO)(CO)(pyridine)
in CH2Cl2 solution. Isotopic substitution induces a substantial frequency
shift but has almost no effect on VR rate.
quite small, at most a few percent. In fact, any differences in
VR rates between 12CO and 13CO compounds seen in Figure 7
are likely within the errors of our measurement. Therefore, there
is essentially no isotope effect on the VR rate.
IV. The Vibrational Relaxation Process
A brief discussion of the VR process will prove useful in
interpreting our results. In pump-probe experiments,4
a
picosecond duration mid-IR pulse is tuned to the fundamental
vibrational transition of CO (the V ) 0 to V ) 1 transition, in
the 1860-1980 cm-1 range). The excitation produced by the
pulse is essentially localized on CO. This CO oscillator may,
to a good degree of accuracy, be treated as a two-level system
because the mid-IR pulses are not resonant with the V ) 1 to V
) 2 transition. The frequency bandwidth of the pumping pulses
is less than the vibrational anharmonicity5 of ∼25 cm-1
.
The excited CO oscillator loses its excess vibrational energy
via anharmonic coupling with the surrounding medium. This
coupling is weak7 in the sense that the period of oscillation (∼20
fs) is far less than the typical lifetime (∼20 ps). The VR rate

Tuning the Vibrational Relaxation of CO
J. Phys. Chem., Vol. 100, No. 46, 1996 18027
depends on the density of states in the medium at the frequency
assisted VR processes. A small molecule such as CHCl3 has a
sparse density of vibrational states at νCO. Unless there is an
unlikely coincidence, there will be no energy-conserving
pathways for CO to CHCl3. For vibrational energy transfer to
molecules with sparse state densities, at least one solvent phonon
from the phonon continuum is required to assure energy
conservation.
VR is expected to be most efficient when the quantum number
mismatch with the final state is minimal.7 The states closest in
energy to νCO with the smallest quantum number mismatch are
those consisting of two higher frequency vibrations in the ∼1000
cm-1 range, or two vibrations plus one phonon. For an extreme
example, states consisting of 20 ∼100 cm-1 phonons are
unlikely to be well coupled to the CO oscillator due the poor
quantum number mismatch.
νCO and the strength of coupling to those states. The combined
effects of these two quantities, state density and coupling
strength, are conveniently combined in the concept of the force
correlation function.7,18
The vibrational modes of the medium produce a broad
spectrum of structural fluctuations that act on the system. In
this case, the system is the CO oscillator. The force correlation
function describes the magnitudes of all fluctuating forces
exerted by the medium on the system.7,18 The rate of vibrational
energy loss from CO is proportional to the rate of transitions
from V ) 1 to V ) 0. This rate is proportional to the Fourier
component of the force correlation function at νCO
.
In the classical mechanical treatment of the force correlation
function,18 the medium induces transitions only when the
medium’s vibrations are excited, for example when the medium
is at finite temperature. Recently, a fully quantum mechanical
treatment of the force correlation function was developed.7 The
quantum mechanical treatment shows it is not necessary for the
medium’s vibrations to be thermally populated for VR to occur.
In the quantum mechanical treatment, the nonradiative relaxation
of the carbonyl oscillator two-level system is seen to be
analogous to the more familiar problem of a radiative transition
driven by a radiation field. VR in the absence of thermal
population of the medium’s vibrations is analogous to spontane-
ous emission, where radiative relaxation occurs even in the
absence of a radiation field. The rates of spontaneous VR
processes are independent of temperature and are nonzero even
at zero temperature. VR induced by thermally excited vibrations
of the medium is analogous to stimulated emission in a radiative
system. The rates of the VR stimulated emission processes
increase with increasing temperature. The quantum mechanical
treatment is important because temperature-dependent studies
of VR in heme proteins2 have shown the quantum mechanical
process analogous to spontaneous emission dominates, even at
ambient temperature.
V. Discussion
A. Summary of Relevant Experimental Results. The most
significant features of our results are (1) the striking correlation
between VR lifetime and the carbonyl stretching frequency νCO
,
as well as the precise degree to which the vibrational lifetime
can be tuned by controlling heme structure, (2) the small but
systematic porphyrin effect in which PHDME heme complexes
generally have longer VR lifetimes than TPP porphyrin com-
plexes, (3) the insensitivity of the VR rate to changes of solvent,
and (4) the lack of an isotope effect (13CO for 12CO) on the
vibrational lifetime, despite the substantial frequency shift
induced by isotopic substitution. A small number of exceptions
to the lifetime-frequency correlation and solvent insensitivity
are also observed, as discussed below.
B. Lifetime-Frequency Correlation. Our data show that
changes in νCO are correlated with changes in the VR lifetime.
The VR lifetime decreases as νCO decreases. Other workers
have previously correlated other quantities with νCO
. For
example, correlations between νCO and νFe-C were described
by Li and Spiro,20 who concluded that carbonyl frequency
shifting was primarily controlled by changes in back-bonding.
An increase in back-bonding simultaneously weakens the C-O
bond and strengthens the Fe-C bond. Correlations between
νCO and 13C and 17O chemical shifts were reported by Guo et
al.,21 who concluded that changes both in carbonyl frequency
and chemical shift parameters were caused by changes in the
extent of back-bonding.21,22
As the carbonyl frequency changes, the change in the VR
lifetime might be caused by either of two factors:7 (1) a change
in the number of available states or (2) a change in the
anharmonic coupling between CO and the surrounding medium.
The isotope effect experiments (as discussed in detail in section
V.E.) clearly show that the vibrational lifetime changes are due
to changes in the anharmonic coupling. Because changes in
νCO and the vibrational lifetime are affected by changing the
structure of the porphyrin, it is strongly indicative that the
coupling which is significant is the coupling between CO and
porphyrin, rather than the coupling between CO and solvent.
The solvent effect experiments (see section V.F.) confirm this
to be the case for chlorinated hydrocarbon solvents. Thus, the
lifetime-frequency correlation occurs because porphyrin struc-
tures that change νCO similarly change the coupling between
CO and porphyrin, and this coupling is the predominant factor
that controls the rate of vibrational energy flow from CO to
porphyrin.
The VR rate of CO reflects the fluctuating forces exerted by
the medium on CO at a high frequency of ∼2000 cm-1. These
high-frequency motions are predominantly associated with
higher frequency intramolecular vibrational modes of the
medium, as opposed to lower frequency structural or confor-
mational transitions.4 CO is an unusual ligand in that its
fundamental vibration falls into a frequency range where there
are no other fundamental vibrations of either porphyrin or
solvent. Vibrational energy transfer from CO to either porphyrin
or solvent thus involves states which are combinations or
overtones of fundamental vibrations or states which are com-
binations of vibrations and solvent phonons. The solvent
phonons are a continuum of lower frequency collective states,9
which range from zero frequency to a cutoff in the few hundreds
of cm-1 range.
In polyatomic molecules, the density of vibrational states,
consisting of fundamentals, overtones, and combinations, in-
creases with increasing energy. At lower energies, the state
density is discrete or sparse. At higher energies, the state density
becomes large and essentially continuous. The frequencies
where these sparse and dense regions occur depend on the size
of the molecule. A large molecule such as porphyrin will have
19
very many vibrational states at frequency νCO
,
so there are a
large number of possible energy-conserving pathways for
carbonyl VR, involving CO-to-porphyrin vibrations. Vibrational
energy transfer to molecules with continuous state densities do
not require phonons. Transfer processes involving vibrations
and simultaneous creation or annihilation of phonons are
possible as well. These latter processes are termed phonon-
C. σ- and π-Bonds. Since we see the VR lifetime to be
well correlated with νCO, a brief discussion of factors which
affect νCO is now presented. A vast literature exists concerning
carbonyl vibrational frequencies and carbonyl bonding in

18028 J. Phys. Chem., Vol. 100, No. 46, 1996
Hill et al.
metal-carbonyl systems, which we will not attempt to sum-
marize here. It is well-known that carbonyl stretching frequen-
cies can be affect by cis or trans ligand effects.23 In the heme
complexes used here, cis refers to the tetradentate porphyrin
ligand and trans to the monodentate proximal ligand.
Returning to Figure 1, we note that metal-CO bonding
involves both σ- and π-bonds. Cis and trans ligands might
affect either or both types of bonding. Changing σ- and
π-bonding affects νCO in different ways. Effects of changing
π-bonding have been discussed extensively by many authors,
e.g., refs 20-22 and 24-26, whereas effects of changing
σ-bonding are ordinarily less significant and not discussed
much. We need to consider σ-bonding here to interpret
deviations from the lifetime-frequency correlation seen in our
data.
ments to reveal vibrational relaxation processes which depend
on anharmonic coupling. Prior to the development of infrared
pump-probe techniques, essentially nothing was known about
M-CO anharmonic coupling.
In a previous paper,11 we argued that vibrational energy
transfer from excited CO to heme principally involved through
π-bond coupling, rather than through σ-bond coupling. The
argument was based on three data points, corresponding to the
heme compounds M(COPRO)(CO)(pyridine), where M ) Fe,
Ru, Os, in CH2Cl2 solvent. It was known that νCO decreases
with increasing mass of the M atom. The heavier M atoms
have more polarizable and more spatially extended valence
electrons, which increase the extent of back-bonding.28,29 Our
finding was that the VR lifetime decreased as the mass of the
M atom increased.11
A simple and useful way of understanding these effects is to
consider an isolated linear triatomic molecule M-C-O.23,27 In
Herzberg’s book,27 analytical solutions are presented for the
vibrational frequencies in terms of the atomic masses and bond
force constants. To model the heme or porphyrin complexes,
M should be considerably heavier than C or O, and the C-O
force constant kCO should be considerably larger than the M-C
force constant kMC. In this case, the ν3 vibration (in Herzberg’s
notation) is primarily a C-O stretching vibration,27 with a small
amount of coupling with the M-C oscillator. In a harmonic
oscillator, the frequency is proportional to (k/µ)1/2, where µ is
the reduced mass. Thus, it is approximately correct to view
changes in νCO as arising from changes in either k or µ.
The decrease in VR lifetime with increasing mass of the metal
atom is exactly opposite what one expects if through σ-bond
coupling were dominant.11,30 From the viewpoint of classical
mechanics, which presents a simple and intuitive picture of this
process, the CO oscillator loses energy by driving the M-C
oscillator. When the M atom is made heavier, the M-C
oscillator is shifted further off-resonance from the CO oscillator,
the amplitudes of M atom oscillations decrease, and the extent
of through σ-bond coupling is decreased.6,11 From the viewpoint
of the mechanical force correlation function model in either the
classical18 or quantum pictures,7 making the M atom heavier
makes the modes associated with the heavier atom lower in
frequency, resulting in smaller Fourier coefficients of the force
correlation function at νCO. In either description, heavier metal
atoms must reduce through σ-bond coupling. Because heavier
M atoms decreased the VR lifetime, the through σ-bond
coupling mechanism can be convincingly ruled out.11
Structural or environmental factors which affect π-bonding
change the extent of back-bonding from the metalloporphyrin
dπ and pπ orbitals to the antibonding π* orbitals of CO.8
Increased back-bonding decreases kCO, which tends to decrease
ν
CO. Decreased back-bonding has the opposite effect.
Although the primary effect of changing back-bonding is to
The evidence against through σ-bond coupling suggested that
through π-bond coupling was dominant. More significantly,
through π-bond coupling could be invoked to explain the
direction of the observed relation between lifetime and fre-
quency.11 Increased back-bonding would be expected to
decrease the VR lifetime by increasing through π-bond coupling,
while lowering νCO, as observed.11 The lifetime-frequency
correlations seen in Figures 3 and 4 thus provide compelling
evidence for the dominance of through π-bond coupling in the
relaxation of vibrationally excited CO bound to heme.
change the force constant kCO, there is a secondary effect which
may be viewed as a reduced mass effect. Increasing back-
bonding, which reduces kCO, simultaneously increases kMC, and
vice versa.20 When kMC increases, the reduced mass of the CO
oscillator seems to increase, because a stronger M-C bond
means CO oscillations involve increasing displacements of the
heavier M atom. The reduced mass effect of changing π-bond-
ing opposes the force constant effect. The net effect of
increasing back-bonding is as follows: there is a decrease in
E. Isotope Effects on Vibrational Relaxation. The isotope
effect experiments reveal a feature of the lifetime-frequency
correlation which is crucial in understanding the mechanism of
carbonyl VR. As shown in Figures 6 and 7, substituting 13CO
for 12CO in heme compounds causes a substantial change of
νCO but no change in the VR lifetime. There is a negligible
isotope effect on VR. A 13CO experiment on Mb-CO gave
the same result.3 The isotope effect experiments clearly
demonstrate the lifetime-frequency correlation does not depend
on the absolute value of νCO, but instead on the induced shift
of νCO caused by changing the chemical structure and thus the
extent of back-bonding.
νCO due to the decrease in kCO, but the decrease in νCO is slightly
offset by an increase due to increasing µ. For example, using
the equations in ref 27, it can be shown that reducing kCO in
FeCO to produce a 70 cm-1 decrease in νCO (e.g., from 1970
to 1900 cm-1), the reduced mass effect offsets the decrease in
ν
CO by about 5 cm-1. The reduced mass effect is not discussed
extensively in the literature, probably because it is a perturbation
to the dominant force constant effect.
A similar reduced mass effect can result from changing the
σ-bonding. Some ligands can affect the σ-bonding between M
and C.23 Increasing the strength of the M-C bond increases
the reduced mass of the CO oscillator, decreasing νCO. The
important feature of σ-bonding needed for subsequent discus-
sions is that changes in σ-bonding induced by different ligands
can affect νCO without significantly changing the back-bonding
between M and CO. σ-bond effects induced by ligands thus
provide a possible mechanism for changing νCO without
affecting through π-bond anharmonic coupling.
It is conceivable that the vibrational lifetime could depend
on the absolute value of νCO in such a way that decreasing νCO
decreased the lifetime. This situation could arise only if changes
in the lifetime depended on changes in the density of states.3-4,7
Such a dependence might exist, for instance, if the density of
heme vibrational states at νCO increased with decreasing νCO
.
A similar dependence might be expected if there existed an
energy mismatch between the carbonyl oscillator and some
specific heme vibration or vibrations located at somewhat lower
energy. Then decreasing νCO would decrease the energy gap
for this process, thereby decreasing the vibrational lifetime. In
D. Anharmonic Coupling through σ- and π-Bonds. There
have been extensive prior studies of M-CO bonding and the
effects on νCO
study is the ability of picosecond infrared pump-probe experi-
23
.
The new and unique feature of the present

Tuning the Vibrational Relaxation of CO
J. Phys. Chem., Vol. 100, No. 46, 1996 18029
either case, 13CO VR would have a significantly shorter lifetime
than 12CO VR, which is not observed. The isotope experiments
show that changes in the vibrational lifetime do not depend on
the absolute value of νCO, which rules out the possibility that
the lifetime-frequency correlation is due to the structure of the
must affect the anharmonic coupling between CO and porphyrin.
It is difficult to see how this proposed solvent effect mechanism
could explain the data in Figure 5. The solvent effect on VR
cannot simply be a change in the through π-bond anharmonic
coupling between CO and porphyrin, because that would result
in a lifetime-frequency correlation which is not observed. There
are some solvents (e.g., CHCl3 and CH3CCl3) where νCO differs
somewhat (by 12 cm-1 in this case), but the VR rates are
identical. In CH3CCl3 and DBP, νCO is identical, but the VR
rates are quite different. For the solvent effect to occur by
affecting the rate of process i, the solvent would somehow have
to affect the anharmonic coupling in a new way which does
not introduce a correlated frequency shift. We believe this to
be unlikely.
density of heme states near νCO. The lifetime-frequency
correlation must therefore depend on changes in the anharmonic
coupling between CO and heme or porphyrin, induced by
changing the structure of the complex.
The lifetime-frequency correlation is thus understood as
follows. Changing porphyrin substituents (e.g., the metal atom
and proximal ligands) changes the extent of back-bonding. An
increase in back-bonding lowers νCO. An increase in back-
bonding simultaneously increases through π-bond anharmonic
coupling, which decreases the CO vibrational lifetime.
F. Solvent Effects on Vibrational Relaxation. The solvent
effect experiments address the relative significance of intra-
molecular Vs intermolecular VR processes in the lifetime-
frequency correlation data in Figures 3 and 4, where CH2Cl2
was the solvent. The observations which need to be discussed
are the essentially identical (∼16 ps) VR lifetimes of the Ru
complex in CH2Cl2, CHCl3, and CH3CCl3, the slightly longer
(19 Vs 16 ps) VR lifetime in CCl4, and the significant shortening
(∼40%) of the VR lifetime in DBP.
Process ii is a solvent phonon-assisted transfer from CO to
porphyrin. Changing the solvent changes both the solvent
phonon densities of states and the strength of anharmonic
coupling between the porphyrin complex and the solvent
phonons. There are several reasons why we expect (but cannot
yet prove positively) that this process is not significant in the
porphyrin-CO complexes. First, phonons are not really neces-
sary for the CO-to-porphyrin process, due to the large state
density of porphyrin vibrations at νCO. Second, the phonon
densities of states of CHCl3 and CH3CCl3 are different from
that of CH2Cl2, but the VR lifetime is the same in all three of
solvents. Third, we know from temperature-dependent studies2
that this type of phonon-assisted CO-to-porphyrin process is
not significant in Mb, which is quite similar to the porphyrin
complexes we have studied.
To understand solvent effects, we must consider the four
general mechanisms for loss of vibrational excitation from CO
in a solvated porphyrin complex:
CO f porphyrin vibrations
CO f porphyrin vibrations ( solvent phonon(s) (ii)
CO f solvent vibrations (iii)
(i)
Process iii is an intermolecular transfer from CO to solvent
vibrations, without phonons. We will discuss DBP separately
from the chlorinated hydrocarbons. These latter solvents are
all rather small molecules, and they have a sparse density of
vibrational states near νCO. Using tabulated vibrational spectral
data,31 one can estimate the density of states near νCO to decrease
in the order CH3CCl3 > CCl4 > CHCl3 > CH2Cl2. CH3CCl3
of course has the greatest density of vibrational states. CCl4 is
next because it has four low-frequency C-Cl stretching
vibrations. The latter two solvents have high-frequency C-H
stretching vibrations which are too high in energy to interact
with the CO oscillator. Due to the sparsity of solvent vibrational
states at νCO, it would be an unlikely coincidence if any solvent
vibrational states were precisely resonant with the carbonyl
oscillator. Thus, the likelihood of CO-to-solvent vibration
transfer process iii (without phonons) is quite small in the
chlorinated hydrocarbons. Furthermore, even if by a series of
coincidences this process were possible in all four of these
solVents, because the vibrational densities of states are very
different, VR would be very different in each solvent, which is
not observed.
CO f solvent vibrations ( solvent phonon(s) (iv)
Unlikely processes such as CO f (many solvent phonons) are
not considered above. Processes i and iii involve energy transfer
from the vibrationally excited CO to vibrational states of the
porphyrin or the solvent. Processes ii and iv are phonon-assisted
versions of processes i and iii. For any of these processes to
occur, the solvent or porphyrin must have at least one state (i.e.,
a combination or overtone of vibrations and/or phonons) at
frequency νCO, and there must be anharmonic coupling between
the CO stretching fundamental and the porphyrin or the solvent.
It is straightforward to approximately determine the density of
vibrational states of porphyrin or solvent using vibrational
spectra or normal mode calculations, but the coupling is difficult
to determine a priori, since anharmonic coupling depends on
the finer details of the potential surface.7
In processes ii and iv, a plus sign indicates a phonon is emitted
and a minus sign indicates a phonon is absorbed. Phonon
absorption occurs only at finite temperature, where the phonons
are thermally populated. Because phonon energies range from
zero frequency to a frequency cutoff in the few hundred cm-1
range,9,32 the phonon population changes greatly in the 0-300
K range (at 300 K, kT ∼ 200 cm-1). Temperature-dependent
studies of VR2 can be used to determine whether the dominant
VR process is phonon-assisted. If phonons are involved, the
rate of the VR process will change significantly in this
temperature range.
Process iv is a phonon-assisted intermolecular transfer.
Phonon assistance permits energy transfer from excited CO to
solvent vibrations lying within a few hundred cm-1 of νCO. Thus,
process iv is the only way energy can be transferred from CO
to a small solvent molecule which has a sparse vibrational
density of states at νCO. All other things remaining equal, the
larger the value of the phonon cutoff frequency in a solvent,
the more solvent vibrational states can participate in CO-to-
solvent vibration transfer.
Process i is a purely intramolecular CO-to-porphyrin transfer.
The porphyrin density of states at νCO is very large,19 so process
i can occur even without the solvent. Process i is the process
which we believe dominates the VR of CO in porphyrin
complexes and which is responsible for the lifetime-frequency
correlation. For the solvent to affect the rate of process i, it
As above, energy transfer from CO to DBP can occur by
process iv. However, the DBP solvent is a much larger
molecule than the chlorinated hydrocarbon solvents. DBP
molecules have a dense manifold of vibrational states at νCO
.
Thus, energy transfer from vibrationally excited CO to DBP

18030 J. Phys. Chem., Vol. 100, No. 46, 1996
Hill et al.
vibrations can also occur by process iii, which is not possible
in the chlorinated hydrocarbon solvents.
In light of the above discussion, we can offer some reason-
able, but not definitive, interpretations of our solvent effect data.
There is no evidence for significant CO-to-solvent processes
(either processes iii or iv) in CH2Cl2, the solvent used to
establish the lifetime-frequency correlation. Otherwise, the
carbonyl VR rate would be significantly different in the different
chlorinated hydrocarbon solvents, because their state densities
at νCO are very different. The slightly longer VR lifetime in
CCl4 does suggest the possibility that a small fraction of the
VR in CH2Cl2 (as well as CHCl3 and CH3CCl3) could be due
to intermolecular vibrational energy transfer, most likely by
process iv. A CO-to-CH2Cl2 process which was competitive
with the CO-to-porphyrin process could account for the 19 to
16 ps lifetime reduction in going from CCl4 to CH2Cl2, if the
CO-to-CH2Cl2 process had a lifetime of ∼100 ps at room
temperature. If this relatively inefficient CO-to-solvent process
does exist in CH2Cl2, there are two possibilities for explaining
why it is absent (or is perhaps even more inefficient) in CCl4.
Either the relatively low-frequency solvent phonon cutoff in
CCl4 reduces the number of energy-conserving pathways for
CO-to-solvent transfer (notice here it would be the phonon
spectrum of CCl4 which mattered, not the vibrational spectrum),
or the nonpolar CCl4 interacts less strongly with the porphyrin-
CO complex than other, more polar chlorinated hydrocarbons
solvents.
Figure 8. Comparison of synthetic heme compound data from this
work to myoglobin-CO (Mb-CO) protein data from refs 2 and 3.
The designations A0 and A1 refer to coexisting conformers of wild-
type Mb-CO. H64L and H64V are mutants where the distal histidine
is replaced by leucine or valine. H93G(Im) is a mutant where the
proximal histidine is replaced by glycine, with added imidazole as the
proximal ligand. V68N is a mutant where valine-68 is replaced by
asparagine. The Mb data should be compared to Fe(PHDME), which
is practically identical to protoheme in Mb. The Mb VR rates are clearly
greater than in synthetic PHDME heme complexes. The greater VR
rates in Mb are attributed to differences in solvent effects between
PHDME in CH2Cl2 compared to protoheme inside the protein matrix.
The protein data also suggest a lifetime-frequency correlation similar
to that seen in synthetic heme and porphyrin complexes in Figures 3
and 4.
The significant effect of DBP on the VR rate seems to be
indicative of a CO-to-DBP transfer process. This intermolecular
process may (iv) or may not (iii) involve phonons. The
alternative to this interpretation is to postulate some mechanism
for DBP to substantially change the CO-to-porphyrin anhar-
porphyrin perimeter quite distant from CO. In previous works
20
where correlations were found between νCO and νFeC or
between νCO and 13C coupling constants,21 these correlations
were said to apply equally well to heme proteins with PPIX
and PHDME and TPP complexes. Systematic differences
between these different porphyrin structures were not observed
(or perhaps were not resolved). Besides resolving a clear
difference between TPP and PHDME (Figures 3 and 4), as
shown in the next section, we can also resolve clear differences
between PHDME and heme proteins (cf. Figure 8).
monic coupling without any effect on νCO. The relative
importance of phonons could be investigated by studying the
temperature dependence of VR in DBP. The presumed ability
of DBP to participate in intermolecular VR processes is a result
of the large state density at νCO, but it is also worth noting the
possibility of relatively strong nonbonded interactions between
the aromatic phenyl groups of DBP and porphyrin.
More work needs to be done to better understand questions
raised by the solvent effect data. For example, it would be
interesting to study a series of compounds in DBP to see whether
a lifetime-frequency correlation exists in DBP solutions. It
will also be interesting to expand the range of solvents used in
this study and to perform temperature-dependent experiments
to investigate the role of solvent-heme interactions in VR.
G. Additional Features of Through π-Bond Coupling.
The through π-bond model accounts for the most significant
features of our data, namely, the existence of a lifetime-
frequency correlation, and the direction of the correlation, i.e.,
lower frequency means shorter lifetime. However, without some
embellishments, the simplest form of this model does not
account for two features of our data: (1) Both PHDME heme
complexes and TPP porphyrin complexes show lifetime-
frequency correlations, but the absolute values of the lifetimes
in PHDME and TPP are different (the lifetimes are somewhat
longer in the heme). (2) All the imidazole and pyridine
complexes in Figures 3 and 4 show a lifetime-frequency
correlation, but the triphenylphosphine complexes deviate
significantly from this correlation.
The triphenylphosphine data in Figures 3 and 4 appear to be
exceptions to the lifetime-frequency correlation. Something
similar is seen in the frequency-chemical shift correlation data
reported21 by Guo et al. There the heme proteins P450 and
chloroperoxidase21 seem to be exceptions to their correlation.
In our work and in their work, the systems which evidence these
correlations all have proximal ligands which form metal-
nitrogen bonds (histidine in the proteins, imidazoles, and
pyridines in the synthetic heme complexes). The exceptions
have quite different proximal bonding. There are metal-
phosphorus bonds in the triphenylphosphine porphyrin com-
plexes and metal-sulfur bonds in P450 and chloroperoxidase
proteins, where histidine is replaced by cysteine. What appear
to be exceptions to these correlations may not be exceptions at
all. It is conceivable that a series of complexes with metal-
phosphorus bonds or metal-sulfur bonds might themselves
show a lifetime-frequency correlation. For example, the
triphenylphosphine data are suggestive of a lifetime-frequency
correlation for metal-phosphorus bonded systems which has
only a slightly different slope and intercept from the correlation
for metal-nitrogen bonded systems in Figures 3 and 4.
There are two orthogonal ways of looking at deviations from
the lifetime-frequency correlation. For example, consider the
triphenylphosphine data in Figure 3. If these data points are
taken to lie aboVe the line, then triphenylphosphine is causing
the VR rate to increase without affecting νCO. On the other
The fact there exists a small but discernible systematic
difference in the lifetime-frequency correlation between TPP
and PHDME is interesting. It shows that carbonyl VR rates
do not depend entirely on the structure at the active site, but
rather these rates can be influenced by substituents on the

Tuning the Vibrational Relaxation of CO
J. Phys. Chem., Vol. 100, No. 46, 1996 18031
hand, if these data points are taken to lie to the right of the
line, then triphenylphosphine is causing νCO to increase without
affecting the VR rate. These two possibilities require quite
different interpretations, and either or both possibilities might
be involved in causing the observed deviations. In the former
case, triphenylphosphine is not affecting νCO, but it is providing
additional fluctuating forces on CO to increase the VR rate.
The through π-bond coupling remains the same, but the three
phenyl groups provide additional state density for the VR
process. In the latter case, triphenylphosphine is causing a shift
in νCO without affecting the magnitude of the force correlation
function at CO. This could occur, for instance, if triphenyl-
phosphine shifted νCO by changing the M-CO σ-bonding. As
discussed in section V.C, ligands which affect the M-C
σ-bonding can affect νCO without affecting the π-bonding and
the through π-bond anharmonic coupling.
of mutant heme protein VR, we argued that CO to protein
vibrational energy transfer was not likely to be significant.3 If
specific interactions between CO and the protein were respon-
sible for carbonyl VR, then a rather complicated dependence
of the VR rate on the details of the protein structure would be
expected to result. Instead, a very simple linear lifetime-
frequency correlation is observed.3,4
VI. Summary and Conclusions
A large number of heme and porphyrin complexes were
synthesized using different metal atoms M (M ) Fe, Ru, Os)
and different proximal ligands (imidazoles and pyridines). A
remarkable correlation between carbonyl frequency νCO and
carbonyl vibrational relaxation (VR) lifetime in these complexes
in CH2Cl2 solvent was found, using mid-IR pump-probe
experiments to directly measure the rate of energy loss from
vibrationally excited CO. The vibrational lifetime decreases
as νCO decreases. The observed correlation is consistent with
a linear relation between vibrational decay rate and νCO. The
fact that structural changes of the complexes strongly affect VR,
whereas a series of different chlorinated hydrocarbon solvents
do not, indicates the predominant VR process in CH2Cl2 solvent
is an intramolecular CO-to-porphyrin vibrational energy transfer
rather than an intermolecular CO-to-solvent transfer. Isotope
effect experiments show that the relationship between lifetime
and frequency occurs because the frequency shift of νCO is
related to the strength of anharmonic coupling between CO and
porphyrin. Experiments with different metal atoms (Fe, Ru,
Os) show the most important anharmonic coupling involves the
M-CO π-bonding. Vibrational energy relaxation of CO occurs
mainly by through π-bond anharmonic coupling to heme or
porphyrin.
The through π-bond anharmonic coupling model satisfactorily
explains the most significant finding in this work, the lifetime-
frequency correlation. Other, more subtle features of our data,
however, require more study. First, there is a systematic
difference in the lifetime-frequency correlation between a
heme, PHDME, a porphyrin, TPP, and myoglobin (PPIX plus
protein). At a given value of νCO, the VR lifetime is shortest
in Mb, longer in TPP, and longest in PHDME. Second, a
proximal ligand which makes a metal-phosphorus bond does
not obey the same lifetime-frequency correlation seen with
imidazoles and pyridines which make a metal-nitrogen bond.
We do not yet know whether these more subtle features involve
interactions which affect the frequency without much affecting
the lifetime, which affect the lifetime without much affecting
the frequency or a combination of both.
In the same spirit, we now consider the PHDME vs TPP data
(see the dashed lines in Figure 5). For a given value of νCO
,
TPP compounds on average have faster VR rates. This might
be interpreted to mean that the phenyl substituents on TPP
compounds increase the fluctuating forces on CO to increase
the VR rate without affecting νCO. Alternatively, TPP could
affect νCO, perhaps by changing the M-CO σ-bonding, without
significantly increasing the fluctuating forces on CO. Resolving
these questions will require additional studies on more heme
and porphyrin compounds or an independent probe of the extent
of σ- and π-bonding in each compound.
H. Proteins and Synthetic Hemes. Our synthetic heme
complexes provide new insights into pump-probe experiments
on Mb-CO.2,3 In Figure 8 we compare Mb VR data from refs
2 and 3 to the PHDME and TPP data from this work. Shown
in Figure 8 are data2 on two different conformers of wild-type
horse heart Mb, denoted A0 and A1. These conformers coexist
in protein solutions. They are thought to arise from different
states of the distal histidine.22 Also shown are data3 for several
myoglobin mutants, including V68N, where valine-68 is re-
placed by asparagine. This substitution is interesting because
it produces a particularly large shift25 in νCO, almost as large as
substituting Os for Fe.
The data in Figure 8 suggest the existence of a lifetime-
frequency correlation for proteins,3,4 similar to the synthetic
heme complexes. Because PPIX in Mb is only very slightly
different from PHDME (in PHDME, two carboxylate groups
on the perimeter of PPIX are esterified), the Mb data in Figure
8 should be compared to the PHDME data, rather than the TPP
data. In comparing Mb to PHDME, it is clear the protein VR
lifetimes at a given νCO are noticeably shorter. Solvent effects
appear to be the most likely explanation for the systematically
faster VR in proteins compared to PHDME. It does not seem
likely that the systematic difference could be due to the minor
differences between PHDME and PPIX.
One of the goals of our work is to develop a systematic
understanding of the relationships between VR and molecular
structure. Over many years of study, it was generally found
that even relatively small structural changes in molecules may
have large and presently unpredictable effects on VR. The CO-
porphyrin systems we are studying are unique in this respect:
by changing the structure in well-defined and well-controlled
ways, we are able to essentially continuously tune the CO
vibrational relaxation rate over a substantial range spanning
about a factor of 4.
In Mb, the solvent is a protein matrix with polar amino acid
residues in the heme pocket (e.g., the distal histidine). In our
PHDME experiments, the solvent was CH2Cl2. We know in
principle that solvent effects are capable of increasing the VR
enough to account for the Mb to PHDME differences seen in
Figure 8, because these differences are comparable to the effects
of changing CH2Cl2 to DBP, as seen in Figure 5. It is well-
known that electrostatic interactions between residues in the
heme pocket and heme-CO can account for the different νCO
values in heme proteins.22,24-26 The question arises whether
the protein affects the VR rate simply by affecting the back-
bonding and the through π-bond coupling between heme and
CO or whether specific interactions between protein and CO
permit CO-to-protein vibrational energy transfer. In our study
Acknowledgment. This research was supported by the
Office of Naval Research, Biology Division, through Contract
NOOO14-95-1-0259 (D.D.D. and J.R.H.). D.D.D. and J.R.H.
acknowledge additional support from National Science Founda-
tion Grant DMR-94-04806. K.S. and C.Z. acknowledge support
from National Institutes of Health Grant PHS 5RO1-HL-25934.
C.Z. thanks the National Science Foundation for fellowship

18032 J. Phys. Chem., Vol. 100, No. 46, 1996
Hill et al.
(13) Che, C. M.; Poon, C. K.; Chung W. C.; Gray, H. B. Inorg. Chem.
1985, 24, 1277.
(14) Heilweil, E. J.; Cavanagh, R. R.; Stephenson, J. C. Chem. Phys.
Lett. 1987, 134, 181. Heilweil, E. J.; Cavanagh, R. R.; Stephenson, J. C. J.
Chem. Phys. 1988, 89, 230.
(15) Tokmakoff, A.; Sauter, B.; Fayer, M. D. J. Chem. Phys. 1994, 100,
9035.
(16) Tokmakoff, A.; Zimdars, D.; Sauter, B.; Francis, R. S.; Kwok, R.
S.; Fayer, M. D. J. Chem. Phys. 1994, 101, 1741.
(17) Tokmakoff, A.; Urdahl, R. S.; Zimdars, D.; Frances, R. S.; Kwok,
A. S.; Fayer, M. D. J. Chem. Phys. 1995, 102, 3919.
(18) Velsko, S.; Oxtoby, D. W. J. Chem. Phys. 1980, 72, 2260.
(19) Levy, D. Laser Spectroscopy of Cold Gas-Phase Molecules. In
Annu. ReV. Phys. Chem.; Rabinovitch, B. S., Ed.; Annual Reviews, Inc.:
Palo Alto, CA, 1980; Vol. 31; pp 197.
(20) Li, X. Y.; Spiro, T. G. J. Am. Chem. Soc. 1988, 110, 6024.
(21) Park, K. D.; Guo, K.; Adebodun, F.; Chiu, M. L.; Sligar, S. G.;
Oldfield, E. Biochemistry 1991, 30, 2333.
(22) Oldfield, E.; Guo, K.; Augspurger J. D.; Dykstra, C. E. J. Am. Chem.
Soc. 1991, 113, 7537.
(23) Braterman, P. S. Metal Carbonyl Spectra, Academic Press: London,
1975.
(24) Ray, G. B.; Li, X.-Y.; Ibers, J. A.; Sessler, J. L.; Spiro, T. G. J.
support. The work at Stanford University was supported by
the Medical Free Electron Laser Program, through the Office
of Naval Research, by Contract NOOO14-91-C-0170. M.D.F.
acknowledges additional support from National Science Foun-
dation Grant DMR-93-22504. We thank Michael M. Rosenblatt
for synthesizing the COPRO complexes. We thank Prof. Alan
Schwettman and Prof. Todd Smith and their research groups at
the Stanford Free Electron Laser Center for making it possible
to perform these experiments.
References and Notes
(1) Dlott, D. D.; Fayer, M. D.; Hill, J. R.; Rella, C. W.; Suslick, K. S.;
Ziegler, C. J. J. Am. Chem. Soc. 1996, 118, 7853.
(2) Hill, J. R.; Tokmakoff, A.; Peterson, K. A.; Sauter, B.; Zimdars,
D.; Dlott, D. D.; Fayer, M. D. J. Phys. Chem. 1994, 98, 11213.
(3) Hill, J. R.; Dlott, D. D.; Rella, C. A.; Peterson, K. A.; Decatur, S.
M.; Boxer, S. G.; Fayer, M. D., J. Phys. Chem. 1996, 100, 12100.
(4) Hill, J. R., Dlott, D. D.; Rella, C. W.; Smith, T. A.; Schwettman,
H. A.; Peterson, K.; Kwok; A.; Rector, K.; Fayer, M. D. Biospectroscopy
1996, 2, 277.
(5) Hochstrasser, R. M. Proc. SPIEsInt. Soc. Opt. Eng. 1992, 1921,
16.
(6) Owrutsky, J. C.; Li, M.; Locke, B.; Hochstrasser, R. M. J. Phys.
Chem. 1995, 99, 4842.
(7) Kenkre, V. M.; Tokmakoff, A.; Fayer, M. D. J. Chem. Phys. 1994,
101, 10618.
(8) Cotton, F. A.; Wilkinson, G. AdVanced Inorganic Chemistry, 5th
ed.; Wiley-Interscience: New York, 1988.
(9) Seeley, G.; Keyes, T. J. Chem. Phys. 1989, 91, 5581. Xu, B.-C.;
Stratt, R. M. J. Chem. Phys. 1990, 92, 1923. Wu, T. M.; Loring, R. F. J.
Chem. Phys. 1992, 97, 8568.
(10) Antonini, E.; Brunori, M. Hemoglobin and Myoglobin in their
Reactions with Ligands; North-Holland: Amsterdam, 1971.
(11) Hill, J. R.; Dlott, D. D.; Fayer, M. D.; Peterson, K. A.; Rella, C.
W.; Rosenblatt, M. M.; Suslick, K. S.; Ziegler, C. J. Chem. Phys. Lett.
1995, 244, 218.
Am. Chem. Soc. 1994, 116, 162.
(25) Balasubramanian, S.; Lambright, D. G.; Boxer, S. G. Proc. Natl.
Acad. Sci. U.S.A. 1993, 90, 4718. Decatur, S. M.; Boxer, S. G. Biochem.
Biophys. Res. Commun. 1995, 212, 159.
(26) Springer, B. A.; Sligar, S. G.; Olson, J. S.; Phillips, G. N., Jr. Chem.
ReV. 1994, 94, 699.
(27) Herzberg, G. Molecular Spectra and Molecular Structure, II.
Infrared and Raman Spectra; Van Nostrand Reinhold: New York, 1945;
pp 173-4.
(28) Schick, G. A.; Bocian, D. F. J. Am. Chem. Soc. 1984, 106, 1682.
(29) Kim, D.; Su, Y. O.; Spiro, T. G. Inorg. Chem. 1986, 25, 3993.
(30) Benjamin, I.; Reinhardt, W. P. J. Chem. Phys. 1989, 90, 7535.
(31) Meier, W.; Schrader, B. Raman/IR Atlas of Organic Compounds;
Verlag Chemie: Dortmund, 1974.
(32) Moore, P.; Tokmakoff, A.; Keyes, T.; Fayer, M. D. J. Chem. Phys.
1995, 103, 3325.
(12) Barley, M.; Becker, J. Y.; Domazetis, G.; Dolphin D.; James, B.
R. Can. J. Chem. 1987, 61, 2389.
JP961418F