T H E
J O U R N A L
O F
PHYSICAL CHEMISTRY Registered in U. S. Patent Office 0 Copyright, 1976, by the American Chemical Society
VOLUME 80, NUMBER 4 FEBRUARY 12,1976
Phatochemlcal Ion Formation in Lumiflavin Solutions S. Gwyn Ballard,* David C. Maurerall, The Rockefeller University, New York, New York 10021
and Gordon Tollin Department of Chemistry, University of Arizona, Tuscon, Arizona 8572 I (Received June IO, 1975)
We have studied ion formation in pulse-irradiated lumiflavin solutions using both a sensitive photoconductance instrument with a time resolution of 0.3 bus, and conventional flash photolysis with optical detection. Free ions are observed only in solvents of high dielectric constant. Their formation and decay are complex processes influenced by the presence of electron donors and, most dramatically, by solvent pH. Experimental observations are interpreted in terms of a primary ionization process involving electron transfer in a triplet flavin dimer or aggregate, followed by a number of fast proton transfer reactions. Extrinsic electron donors (indoles) undergo electron transfer to a nonaggregated triplet species. An attempt is made to relate the kinetics of species observable through absorbance measurements to those of the ionic species measured conductimetrically. Conductance has been shown to be an extremely sensitive method for studying photochemical ionization, and particularly the acid-base behavior of transient photoproducts. These reactions may be almost invisible to optical detection; the ability of absorbance measurement to detect neutral species, however, makes the two methods highly complementary. The reactions described in this paper occur over an excess acid or base concentration range of a few micromolar, and measurable kinetic changes occur as a resllt of acid/base levels of a few parts per billion.
Introduction The lowest triplet state of flavin (isoalloxazine) is known to accept electrons from a variety of organic compounds.lY2 Indole derivatives3 are of particular importance as donors because of the recent demonstration4v5that a tryptophan side chain is part of the coenzyme binding site in certain flavoenzymes (i.e., the flavodoxins). It has been shown that both the flavin singlet6 and triplet7 excited states are highly quenched in these proteins, and it is possible that the quenching mechanism involves electron transfer to the indole ring. Flavins have also been implicated in several photobiological phenomena,s-ll giving additional importance to an understanding of isoalloxazine photochemistry. Inasmuch as the primary photoproducts of electron transfer from an organic donor to the flavin triplet state will be ion radicals, electrical conductance measurements should provide a useful means of monitoring the photoreaction, provided that solvent conditions permit uncorrelation of the first-formed ion pairs. Furthermore, the flavin radical is known to have a pK near neutrality (8.4 for lumi-
flavin),12J3 and so conductance measurements might also reveal proton transfer reactions which occur subsequent *to the initial electron transfer. With these considerations in mind we have undertaken a flash photolysis study of the electron transfer reactions between the triplet and ground states of lumiflavin, and also between the lumiflavin triplet state and indole. Optical and conductimetric detection methods were employed. Both organic (acetonitrile, butyronitrile, and dichloromethane) and aqueous solvents were used in order to demonstrate dielectric effects; the former also provide aprotic environments, thereby decreasing the rates of proton transfer reactions such that they become distinct from the primary electron transfer processes. The results of these experiments have provided some new insights into the chemistry of the flavin triplet state, and have demonstrated the utility of electrical measurements in the study of fast protonic equilibria in photochemical reactions. Kinetic conductance techniques have played an important role in the elucidation of a number of photophysical processes involving creation or annihilation of charge carriers. These include electron ejection from mole341
342
cu1es,14-17 thermal and infrared stimulated detrapping of electrons trapped in polarizable matrices1a21,2s and their recombination with parent cations or scavenger^,^^,^^ secondary ionization phenomena following pulse radiolysis of fluid s o l ~ t i o n s ,and ~ ~ ~ionic ~ ~ ~photodissociation ~ ~ of charge-transfer c ~ m p l e x e s . ~ ~ - ~ ~ In general these studies have employed relatively crude photoconductance apparatus, consisting essentially of a resistive divider network containing the cell, and polarized by large continuously applied dc voltages. While fields of many kilovolts per centimeter are tolerable in studies of solid samples, this is not the case with solutions. The redox potentials of organic solutes are not likely to be more than a few volts; electrode voltages exceeding these potentials will cause undesirable electrolysis. The apparatus employed in this study avoids large dc polarizing voltages and demonstrates that great sensitivity can be achieved without them, using low-noise operational electronics and signal averaging. An alternative approach is alternating-current conductimetry,21s28though this technique has neither the sensitivity nor the speed of the pulsed dc method employed in the present study. Conductance methods have been very little applied to monophotonic ionization processes25-27,29~30 which have, however, been studied by various investigators using conventional flash p h o t ~ l y s i s . ~ In l - ~such ~ studies the kinetic absorbance and conductance methods nicely compliment each other. Conductance is very much more sensitive, has equally fast response (potentially into the nanosecond region), and is specific for charged species. Absorbance gives more information on the identity of transient products, though some information can also be gained from measurement of ion mobility. Because of the large values of this quantity for H+ and OH-, particularly in water, the conductance method is especially suited to the measurement of fast protonic equilibria.
Experimental Section Acetonitrile and butyronitrile are prepared by two successive distillations of MCB chromatoquality solvent stored over calcium hydride, the first from PzO5 and the second from activated Linde 3A molecular sieve. Water is doubly distilled in glass, and dichloromethane (Fischer spectroscopic grade) used without further purification. Indole is purified by recrystallization from methanol-water, followed by vacuum desiccation. Perchloric acid and tetramethylammonium hydroxide (Eastman) are used without purification and added in the form of millimolar solutions in the appropriate solvent by means of a micropipet. Lumiflavin is prepared according to the method described in ref 34. Solutions are prepared a t concentrations from 2.5 to 50 pM and vigorously purged with prepurified Nz gas (Ohio) before irradiation (residual 0 2 concentration ca. 0.5 pM). Conductivity experiments use repetitive 10-ns pulses of uv light (3371 A) from a Molectron UVlOOO nitrogen laser a t peak powers up to 1 MW. Irradiated solution volume is 0.2 ml in the center of the conductance cell. The platinum electrodes are polarized a t up to 10 V cm-l immediately prior to the laser discharge, the polarizing voltage being maintained throughout the measurement period (up to 100 ms), then reversed for an approximately equal time to minimize solute electrolysis and bulk transport effects. Precise balancing of these two periods holds the dark conductance constant to better than mho over extended periods of The Journal of Physical Chemistry, Vol. 80, No. 4, 1976
S. G. Ballard, D. C. Mauzerall, and G. Tollin
time. Photoionization is measured via the accompanying small change in cell conductance; typical ion drift currents in the present experiments are to A, though averaging techniques permit considerably smaller currents to be measured. Approximate detection limit is M univalent ions of molecular weight 100 with a response time of 300 ns, M with a response time of 50 ps. Provision is made to compensate the large cell currents involved with aqueous solutions. The apparatus also automatically zeroadjusts before each laser pulse, thereby referring the photoconductance signal to an initial zero baseline without ac coupling. Output is a voltage analogue of the ion drift current. It is digitized, averaged if necessary, and displayed via an X-Y recorder for subsequent analysis. The apparatus also provides a continuous readout of the total dc cell conductance, obtained by sampling immediately before each flash. This permits establishment of standard prephotolysis conditions, and allows continuous monitoring of the extent of irreversible photoionization throughout the experiment. The entire apparatus is contained in an elaborate Faraday cage, with temperature constant to approximately 1 "C. All operational circuitry is powered via stabilized supplies based on Ni-Cd cells, and the digitization and averaging electronics powered via an ultraisolation transformer. Trigger pulses between the laser and the central controller are transmitted through ultrafast optical isolators, to minimize coupling of the laser discharge noise into the signal processing equipment. A block diagram of the conductance apparatus is shown in Figure 1; it is to be described in detail elsewhere.35 A description of the optical detection flash photolysis instrument is to be found in ref 2.
Results and Discussion Flash Photolysis with Conductimetric Detection. A . Lumiflauin Photoionization i n the Absence of Extrinsic Donors. i. Neutral Solutions. For the purposes of this discussion, neutral solutions are defined as those to which no acid (HC104) or base (NMe40H) has been added. Figure 2 shows the kinetic behavior of conductance of a laser-irradiated 50 pM lumiflavin solution in acetonitrile over the entire time scale from ion formation to complete decay. Details of the rise kinetics are shown in Figure 3, in which the dashed curve represents the response of the instrument to a delta function current input (obtained by substituting a p-i-n photodiode for the conductance cell, and attenuating the laser pulse appropriately). Kinetics of growth of the ion current in the first few tens of microseconds after pulse irradiation are accurately first order, though deviations are observed when indole is present (cf. Figure 8, section B). Amplitude of the ion signal is proportional to the laser power, but t1/2 varies neither with laser power nor with lumiflavin concentration over the range 2.5 to 50 pM. Thus, the simplest mechanism compatible with first-order kinetics, namely
lumiflavin lumiflavin ground state triplet state
ion pair
uncorrelated ions
cannot be correct. In acetonitrile and butyronitrile the signal amplitude is a slowly saturating function of [LG] over the range studied, which cannot be accounted for by optical screening. More linear behavior is seen in water, but the risetime is shorter by a factor of 2, and the signal amplitude
Photochemical Ion Formation in Lumiflavin Solutions
x-Y
343
+ High. a p e d dIglt Ia0 r
Amadnp
recorder
computer
FARADAY CAOE
I
Readoul
Md*enon UVlOOO
Flgure 1. Block diagram of the photoconductance apparatus. The laser is pulsed at approximately 10 Hz, digitization being performed in realtime. A typical data set involves the averaging of 32 or 64 shots at 300-ns resolution or 4 shots at 50-ps resolution.
5
0
L 0
I
2
3
4
5
6
7
a
1 9
IO
IIrnDCd
Figure 2. Retrace of photoconductance behavior of 50 pM lumiflavin solution in "neutral" acetonitrile. Data show the average of 4 shots, and is not additionally smoothed. The single-shot signal-to-noise ratio is approximately 50.
smaller by a factor of 15 (see Figure 9). No ion formation is seen in dichloromethane. These observations are interpreted as being due to monophotonic ionization of a preformed dimer or more complex aggregate; oxygen quenching strongly suggests that the species involved is a triplet state. According to this
IO
15
20
25
30
t(pec)
Figure 3. Detail of the radical ion current growth kinetics, average of 32 pulses. The dotted curve is the single-shot response of the instrument to a delta function current input (10-ns laser flash, detected by p-i-n photodiode, response 1 ns).
scheme the primary photochemical events can be summarized as
!I
hv
LG'
(L.+L.-)
+ hi L.+ + L.1.3x10k-1
Lr (free triplet; does not undergo electron transfer) The Journal of Physical Chemistry, Vol. BO, No. 4, 1976
344
S.G. Ballard, 0.C.Mauzerall, and G. Tollin
Solutions of the flavins in water have been previously shown to contain aggregated ~ p e c i e s , and ~ ~ -it~ seems ~ reasonable to invoke similar species in acetonitrile solutions. We suggest that the dimer or aggregate concentration is in fact considerably lower in water than in acetonitrile or (particularly) butyronitrile due to smaller hydrog,en-bonding interactions between the flavin molecules in aqueous systems. If electron transfer occurs only in the aggregated excited flavin, then the variation of signal amplitude butyronitrile > acetonitrile >> water appears to have a ready explanation. The observed shorter risetime of ion signal in water implies a more rapid uncorrelation of the ion pairs, presumably due to reduced Coulombic interaction between the radical cation and anion components. The dimer theory is further supported by the fact that although optical flash photolysis shows a triplet species with lifetime over 100 k s (see below), primary photoionization occurs with a t 1 / 2 of less than 10 us. This implies that there must be two triplet state populations, one of which is “free” and accounts for the 100-ks signal in the absorbance measurements, and one which is part of the aggregate and has a much shorter lifetime (because of the electron transfer reaction). Strong additional evidence for two distinct triplet species comes from the behavior of the system in high acid and in the presence of indole, described below. If the aggregate hypothesis is correct in this case, the observation of rapid and abruptly terminated primary photoionization implies a tight aggregate into which the “free” triplet species cannot penetrate significantly. If this were not the case, ion formation would occur over the full 100-w~lifetime of the free triplet. In other words, aggregation appears to be primarily a property of the ground state flavin molecules only. In a solution of 50 MMlumiflavin, assuming that the triplet yield is 0.5, we estimate that the laser generates approximately 5 WMflavin triplet (both forms). Photogenerated concentrations of L a + and L.- can be estimated from the conductance equation
where K is the increment of cell conductance (mho) due to photoionization; c h is the corresponding concentrations of radical cation and anion; A+ = XL.+ XL.-, where XI,.+ and XL.- are the equivalent conductances of L.+ and La- at concentration c* in the solvent employed, and a t the prevailing temperature. For the purposes of this study the limiting values (i.e., a t infinite dilution) are employed, and denoted by X o ~ . + and X o ~ . - . s is the cell constant, obtained by calibration with tetraethylammonium perchlorate, whose limiting conductance in acetonitrile is accurately known.39a The equivalent conductances of a large number of ionic species in acetonitrile and many other nonaqueous solvents have been studied and tabulated in the literature. A comprehensive compilation of this data, together with source references and some theoretical discussion is contained in ref 39b. For the purpose of this study, the following limiting equivalent conductances in acetonitrile are used:
+
Ion
AQ, mho cm2 mol-’
Ion
L.+ L.-
35 35
H+ OH-
A‘, mho cm2 mol-’ 100 120
X o ~ +in CH3CN is taken directly from ref 39b (Appendix 5.12.9, p 678). X o ~ . + and X o ~ . - in CH3CN are estimated by The Journal of Physical Chemistry, Vol. SO,No. 4, 1976
extrapolation of mobility data for large organic ions to the radii of L.+ and La-. XOOH- in CH3CN has not been recorded in the literature surveyed. The value employed in this study (120 mho cm2 mol-l) is estimated on the basis of extrapolation of data for halide anions in acetonitrile to the radius of OH-. Support for this value is provided by the asymptotic approach to unity of the parameter p (see below and Figure 5 ) . The observation that the t1/2 of the subsequent bimolecular recombination is invariably much greater than the ion current risetime implies that essentially all L-+L.- pairs uncorrelate. The conductance equation gives [La+] and M when [LT] = 5 KM, indicating that [La-] = 6.5 X about 13%of the total triplet population is involved in the electron transfer reaction. Decay of the primary photosignal does not apparently obey simple first- or second-order kinetics, though we have not undertaken rigorous analysis of the observed transients. The t 112 of the decay is approximately 250 I.LS in the true neutral or slightly basic solutions (50 KM flavin, 500 kW laser) and we attribute it primarily to the process of reverse electron transfer:
+
kz
L*+ La- +2LG
AXo = -{X’L.+
+ A’L.-]
= -70
can be calcuated from the decay data: at 20 *C it has a value of approximately 6 X lo9 M-l s-l. It should be pointed out that if the k2 reaction were the sole relaxation process, then true second-order kinetics would be seen. The tl/2 of the decay does vary approximately as the inverse of the laser power, however, indicating that the k 2 reaction is the dominant process in situations where there is little or no “undershoot” (see below). It appears to correspond to the rapid transient phase in the flash photolysis signal measured at 560 nm (see below). In acidic and basic solutions other processes compete with k2, resulting in an intriguing kinetic complexity shown in the summary diagram (Figure 4)and described in the section acid-base behavior. ii. Acid-Base Behavior. An important feature of the socalled neutral solution is that its recovery kinetics show an “undershoot” phase, i.e., the conductance falls below its initial value and undergoes final relaxation in the positive direction. This undershoot is small in the neutral case, but in the presence of acid or base (added as perchloric acid or tetramethylammonium hydroxide) it dominates the conductance kinetics. In suitable conditions several reversals occur (Figure 4). These effects are not explicable purely in terms of the k2 reaction, but appear to involve alternative pathways of reaction of Le+ and L.- with excess OH- or H30+. These reactions are assigned rate constants k3 and k4, and both give rise to large negative conductance changes (Axo): ks 4x109 L.++OH- + L.+H2O AXo = -{ X o ~ . + X’OH-} = -155 kq 4x109 L.- + H 3 0 + -+ LH.+H,O AXo -(X’L.X’H~O+] = -135 122
-
+
-
+
The dramatic sensitivity of the observed kinetics to acid and base illustrates the responsiveness of the conductimetric method to reactions involving H30+ and OH-, because of the high mobilities of these ions compared to the large organic radical ions. Partition of L-+ between the k 2 and h3 reactions is described empirically by the parameter p, and
Photochemical Ion Formation in Lumiflavin Solutions
i
J
2 p M H+
345
to a very low level. We attribute this growth of conductivity to absorption of atmospheric water. Finally, deliberate addition of water to ca. 10 mM has no discernable effect on the above kinetics, suggesting that water is already present in excess”. hs has,-a value of approximately 2 X lo9 M-l s-l (calculated from t l l z ) , and an associated positive conductance change precisely that required for reattainment of the baseline. A similar situation exists when base is in excess. In this case a fraction p of Le+ is diverted into the k3 reaction, resulting in the same fraction of La- remaining unreacted via k2. The final relaxation is now attributed to the process ((a
A I -
/ J
7 . 5 p M H+
1 0 . 5 p M H+ Figure 4. Variation of the secondary conductance behavior of pulse irradiated lumiflavin solutions with acid/base levels of the solvent. The “2 pM H+” condition corresponds to the data of Figure 3, Le., complete time scale of each trace is ca. 10 ms.
that of La- between k 2 and k 4 by the parameter a. The observed kinetic curves can be resolved into a positive conductance contribution due to k1 and negative contributions due to k2 plus k3 or kd, each weighted according to the mobilities of the species involved. Decay of the positive component is via k2 and either k3 or k 4 . Decay of the negative component is more complex. In acid, diversion of a fraction a of L.- into the kg reaction implies that the yield of the k2 reaction is reduced by the factor (1 - a ) . Consequently a fraction a of L.+ remains unreacted via k2, and the final relaxation to the baseline can be attributed to the process
in which k6 has an approximate value of 3 X lo9 M-l s-l. It is assumed that water is in sufficient concentration (>lo0 pM) that the k6 step is rate limiting. The final rapid regeneration of LG and OH- by reaction of L- with water is associated with a positive conductance change exactly that required to restore the initial conditions. The value of kg coincides closely with that determined optically; k6 was not obtained by conventional flash photolysis, however, because of the difficulty of detecting L.-. The above scheme (summarized in Figure 6) appears to describe simply and elegantly the remarkable symmetry of the acid and base kinetics. Variation of a and p as a function of solution acid/base levels is shown in Figure 5. In calculating a it is assumed that the amplitude ($o+) of the initial positive transient is constant in all cases and equal to its value in the limit of slow inversion. The apparent truncation in acidic solutions (Figure 4) is an artifact of the slow digitization timebase employed in recording this data. The equation for $o+ is $o+ = k(XoL.t
+ Xo~.-) = 70k (k is a scaling constant)
(i)
The amplitude of the final relaxation phase, and hence the undershoot, is proportional to the extent of the k5 reaction, and hence to a. The appropriate equation is $0-
= ka(XoH30+- XoL.t] = 65ak
(ii)
Simultaneous solution of (i) and (ii) gives a = 1.08($0-/$0+)acid
(iii)
The internal consistency of this simple description can be seen from the fact that the total signal amplitude must be equal to the sum of the k 2 and k 4 components This reaction and others to be described below require that an excess (i.e., greater than a few tens of micromolar) of water be present in the acetonitrile solvent. There are several independent indications that this is a justified assumption. Direct measurement of water content of nitriles prepared by the method employed in this study has been performed by several investigators (e.g., ref 42); in general levels not much less than millimolar are to be expected. We find that the conductivity of freshly prepared acetonitrile increases rapidly on standing in the laboratory atmosphere, and this phenomenon is particularly dramatic in samples pretreated by electrolytic cleaning to reduce water content
$total
= k(1 - a)70
+ k(a)35
(iv)
+
(iv) clearly equals (i) (ii). In the calculation of p eq i is still applicable to $o+. The undershoot is given in base by $0-
- ~OL.-]
= k@(XooH-
= 85pk
(VI
Simultaneous solution of (i) and (v) gives
P = 0.82($0-/’&0+)base
(vi)
The above description is adequate for all levels of base up to 10 pM, and for acid up to approximately 1.5 pM. It will be noted that the form of Figure 5 strongly suggests that the so-called “neutral” solution is, in fact, somewhat acidic. The Journal of Physical Chemistry, Vol. 80, No. 4, 1976
S.G. Ballard, D. C. Mauzerall, and G. Tollin
346
vin are partially protonated in the ground state. Thus, a possible mechanism for generating the large positive signal in high acid is a reaction of LGH+ with the “free” triplet, also protonated, according to the scheme
o( o r , B
1.0
f I
0.8
0.6
0.4
MHC104 acid
rM H+
2
0
2
< & >
6
4
AXo = (XoH30+- XOLH+]
8
pM
NMe40H
6
8
,uM OH
base ( a d d e d )
2
4 acid
0
,