Laser photochemistry of oxygen. Application to studies of the absorption spectra of dissolved oxygen molecules analyses of oxygen photonics and dosimetry of laser radiation in biomedical experiments. © 2017 Journal of Biomedical Photonics & Engineering.

. This paper summarizes the data of our lab on the rates of photooxygenation of singlet oxygen traps upon direct laser excitation of oxygen in air-saturated organic solvents and water. Methods of application of these data to calculation of absorbance ( A ) and molar absorption coefficients ( ε ) in the maxima of the main oxygen absorption bands (1273, 765 and 1070 nm) are discussed. The most accurate results were obtained from comparing the photooxygenation rates upon porphyrin-photosensitized and direct excitation of oxygen molecules. It is shown that ε 1273 is not sensitive to the presence of heavy atom (bromine) in solvent molecules and markedly decreases on going from non-polar solvents to water being proportional to the radiative rate constants obtained from the quantum yields of singlet oxygen phosphorescence at 1274 nm. The coefficient ε 765 markedly increases in the presence of bromine. In solvents lacking heavy atoms the 1.5-2-fold increase of ε 765 was observed on going from non-polar solvents to water. Simultaneously, the ratios ε 1273 / ε 765 are changed from (7-10)/1 in non-polar solvents to 1.5/1 in water. The value of ε 1070 obtained in carbon tetrachloride is shown to be about two orders smaller than ε 1273 in the same solvent. The results are important for both


Introduction
It has been known since the seminal papers of R.
Mulliken that oxygen molecules have triplet ground state and two relatively low-lying singlet levels [1,2]. Fig. 1 shows energy diagram of the major electronic transitions in monomeric oxygen molecules in rarified gas as follows from measurements of the absorption spectra in Earth atmosphere and in the gas phase at atmospheric pressure [3][4][5][6]. These transitions are highly forbidden therefore the absorption bands of monomeric oxygen molecules are so weak that they normally cannot be measured using conventional spectrophotometers under ambient conditions.
Reliable absorption spectra were obtained in the gas phase and solutions under high oxygen pressure of 50-140 atm and in liquid oxygen. However, under these conditions dimerization of oxygen molecule occurs, therefore the absorption spectra correspond mainly to dimols (O2)2. In the dimol spectra, the IR maxima analogous to those of monomols are shifted toward shorter wavelengths, and a series of additional maxima appear in visible region. Relative intensities of the spectral peaks of dimols are strongly different from those in the gas phase at atmospheric pressure [6][7][8][9][10][11] (Fig. 2). It was reported in the classic papers by Evans [7] and Matheson's group [12,13] that excitation of oxygen dimols dissolved in organic solvents at oxygen pressure 130-140 atm. leads to formation of the 1 g singlet oxygen, which causes oxygenation of added organic singlet oxygen traps, aminoacids, proteins and other molecules. Based on these experiments, Ambartzumian and his collaborators advanced an idea that direct laser excitation of intrinsic dimol and monomol oxygen molecules in living cells might be a reason for biological action of laser radiation in visible and infrared [14,15]. It was also suggested that laser excitation of oxygen might be applied to destruction of cancer cells in vitro and in animal and human organisms.
These papers caused heated discussions. Apparently, biological action caused by laser excitation of oxygen molecules should be directly proportional to the rate of 1 O2 production by laser radiation, which in turn, depends on laser power (Ilas) and the fraction of laser light absorbed by oxygen molecules (1-10 -A ox ), where Aox is oxygen absorbance (Aox << 0.01). Hence: Biological effect ~ ~ Ilas (1-10 -A ox) = Ilas × 2.3 × Aox. (1) Although biological and therapeutic action of red and IR lasers has been reported many times ([15-19] and refs therein) there is no crucial evidence showing that this action under ambient conditions and moderate (therapeutic) laser power is always due to oxygen excitation. Thus, reliable information on absorption coefficients of oxygen molecules under normal conditions is of great importance for both understanding of oxygen photonics in natural environment and mechanistic analyses of laser action and dosimetry of IR laser radiation in biomedical experiments.
Several years ago our group started work on mimicking biological effects of laser radiation in solutions of singlet ( 1 Δg) oxygen traps in organic solvents and water at normal pressure and temperature. It was established that monomeric oxygen molecules dissolved in these solvents work as photosensitizers of oxygenation reactions. Kinetic analyses of the reaction rates allowed us to develop methods for calculation of absorbance (optical density) for oxygen absorption bands under natural conditions. The present paper summarizes major results of our group [20-28].

Experimental procedure
For sample irradiation, diode lasers LAMI Gelios ("Surgical innovation technologies", Moscow) were applied with the maxima of the emission bands at 1267-1270, 1064-1070 and 760-762 nm and the half bandwidth 4-12 nm. Laser light was focused into a fiber light guide with of 1 mm diameter. The eliciting light power, was controlled by the power meter Ophir ORION-TH with the sensor head 20C-SH (Israel). Solutions were irradiated in the 10 mm rectangular quartz cell under normal pressure and room temperature. Volume of the solutions was 1.5 ml. Diameter of the irradiated spot was 5-8 mm.
Tetracene, (Aldrich Chemistry, >98%), 1,3diphenylisobenzofuran (DPIBF) and uric acid (Acros Organics) were used as singlet oxygen traps. Tetraphenylporphyrin (TPP) (Aldrich, Chemistry) was employed in certain experiments as photosensitizers of singlet oxygen production. Absorption spectra were measured using a Hitachi U-3400 (Japan) and SF-56 (LOMO Spektr, St. Petersburg) spectrophotometers. The rate of 1 O2 formation was calculated from the rate of photobleaching of the main absorption maxima of the traps (474 nm for tetracene, 410-414 nm for DPIBF, 530 nm for rubrene and 290 nm for uric acid) under irradiation by IR lasers.
It was shown that irradiation of air-saturated solutions of the traps in organic solvents and water by IR lasers of moderate power (0.03-2 W) with the emission maxima at 1270, 1070 and 765 nm causes oxygenation of the traps, which is accompanied by bleaching of their main absorption bands. The oxygenation rates were measured as follows. At first, the absorbance (optical density) (A0) in the main absorption maximum of a trap was recorded. In most experiments, solutions with A0 = 0.8-1.5 in the 1 cm cell quartz were employed. Then, the solutions were irradiated during 1-30 min, shaken, and the absorbance (Af) was measured again. The photoreaction rates (Vlasr, ΔA/min) were calculated from measuring the illumination time (t) needed to decrease the absorbance so that ΔA = A0 -Af = 0.05 -0.1. As a result, not more than 15% overall trap content was oxidized during irradiation. The oxygenation rates were then normalized to power of laser radiation in einstein/s (see details in refs [20-28]).

Mechanistic analyses
The photooxygenation action spectra were measured using tunable lasers. These spectra indicate that oxygenation of the traps occurred due to direct laser excitation of oxygen molecules. In CCl4, C6F6 and ethanol the main IR absorption maximum was found at 1273 nm. The width (FWHM) of this band was 17-18 nm (106 cm -1 ) [20,21,23] (Fig. 3). This maximum is shifted by 5 nm to longer wavelengths compared to the absorption maxima of monomeric oxygen in rarified gas ( Fig. 1). More recently these results were confirmed by The action spectra in the dark red region were reported for the first time in papers of our group [24,26] (Fig. 3). In CCl4, C6F6 and ethanol the maximum was found at 765 nm, at slightly longer wavelengths than in the rarefied gas (Fig. 1). The band width was estimated to be 8-9 nm (~130 cm -1 ). Recently Ogilby's group in Denmark confirmed main results of our studies using luminescence technique and extended measurements to several additional solvents. In the luminescence experiments of this group the excitation maxima varied within 763-772 nm and FWHM -within 6-17 nm. In agreement with our data in CCl4 they observed the excitation maximum at 765 nm with the bandwidth of 7 nm (120 cm -1 ) [31]. Apparently, parameters of the action spectra characterize the major absorption bands of dissolved oxygen under ambient conditions.
In organic solvents the oxygenation of the traps was inhibited by singlet oxygen quencher -β-carotene, in water and alcohols -by sodium azide. The photoreaction was not observed after purging with argon or nitrogen. After saturation with pure oxygen the photoreaction rates increased 5-fold compared to airsaturated solutions [20-24]. Thus, evidence was obtained that under ambient conditions, monomeric oxygen molecules work as photosensitizers of oxygenation reactions similarly to oxygen dimols appearing at high oxygen pressure. The mechanisms of the photoreactions correspond to the following scheme:  Fig. 4. No light emission was observed upon excitation at 740 nm. Addition of acetone (50%) caused the decrease of luminescence intensity by more than two orders. Under oxygen purging the luminescence intensity increased (more detailed paper is in preparation). The coefficient α corresponds to degree of overlapping of the action spectrum with the spectrum of laser radiation. It was estimated from the photooxygenation action spectra (Fig. 3) and laser emission spectra using the "Origin" program:

Principles of data analysis
where A(ν) is a function characterizing normalized excitation spectrum of oxygen (Fig. 2) in energy units, in which A varies from 0 to 1; and I(ν) is a function characterizing the normalized spectrum of laser radiation, where I varies from 0 to 1. Apparently, if the wavelength of laser light exactly corresponds to the maximum of the oxygen absorption band, α is equal to 1. Usually, the emission maxima of diode lasers, were shifted to shorter wavelengths compared to the oxygen absorption maxima, therefore α varied from 0.2 to 0.9 (see [27,28] for details).
Eq. 2 allows for accurate determination of the oxygen absorption coefficients corresponding to the maxima of oxygen absorption bands using experimental values for the oxygenation rates under laser irradiation:

Absorbance at 1273 nm
The first paper, in which A1273 was estimated using Eq. 4 and air-saturated carbon tetrachloride as a solvent, was published in 2004 [25]. The data were obtained from measurement of the oxygenation rates of tetracene upon excitation of dissolved oxygen at 1273 nm by tunable forsterite laser. The authors arrived to the values: A1273  7.210 -6 , ε1273  0.003 M -1 cm -1 , 1273 10 -23 cm 2 , which served us as a reference during several years. Because of technical limitations higher precision could not be reached at that time. Then, accurate measurements of the relative rates of DPIBF oxygenation in different solvents were carried out. The results allowed analysis of the solvent dependence of the oxygen absorption coefficients. The first paper on this subject was published in 2005 [22]. Later, more detailed papers followed [23,25]. In particular, it was established that in accord with the Einstein law, the molar absorption coefficients corresponding to the oxygen absorption band at 1273 nm in series of organic solvents were proportional to the radiative rate constants (kr) for this electronic transition obtained from measurement of the quantum yield of singlet oxygen phosphorescence (n is the refractive index) [22-25]: ε1273 ~ kr/n 2 .

(5)
However, in the papers mentioned above the proportionality was not revealed in water and alcohols. We proposed that this effect might be due to problems in kinetic analyses of the data [26,27]. Indeed, in the first papers we assumed that the rate constant (kox) for the reaction of the trap with 1 O2 and molar absorption coefficients of traps (εtrap), which were used for determination of the trap concentrations, did not depend on solvents. In order to solve this problem an experimental procedure was developed, which excluded εtrap and kox from calculations. To reach this goal, the rates of oxygenation of the traps were compared upon direct and photosensitized oxygen excitation at equal trap concentrations. In this case, the following equation was obtained [26-28]:

Amax = (Vlasr/Vpsr)(Ipsex/αIlas) Ф(1-10 -Aps )/2.3, (6)
where Vpsr is the rate of photosensitized oxygenation of the traps; Ilas and Ipsex are the incident photon flux in einstein L -1 s -1 for light applied to excitation of oxygen and photosensitizer respectively; Ф is the yield of singlet oxygen generation by pigments photosensitizers (TPP and TPPS were employed in our experiments). For calculations with Eq. 6, the photooxygenation rates were measured in the ΔA/t units, where ΔA is a change of absorbance in the maxima of the absorption spectra of the traps and t is time (minutes) of laser irradiation. Instead, one has to know accurate values of Ф for , in which we averaged current and previous results of our group for four solvents: CCl4, acetone, ethanol and water. The ways to estimate coefficients  were not found at that time therefore it was suggested that   1 in all experiments. It was obtained using this assumption, DPIBF and Eq. 6 that absorbance of oxygen at 1270 nm in CCl4 is 1.5 times higher than we reported previously [21]. Besides, the relative values of the molar absorption coefficients for this band were proved to be proportional to the radiative rate constants (kr) of singlet oxygen also in water and ethanol. Deviation from the proportionality in water and alcohol reported in the preceding papers [22,23,25] was shown to be due to the strong increase of kox for DPIBF in these media compared to organic solvents having no OHgroups (see [26,28] for details).
Papers of 2003-2012 solved principal experimental problems arising upon measurement of the absorption coefficients for dissolved oxygen. As a result, estimation of the numerical values for these coefficients in different environment was carried out [24-26]. As knowledge of the oxygen absorption coefficients is of fundamental importance for oxygen photonics, subsequent experiments were aimed at increasing the accuracy of our measurements.
For this goal, diode lasers having relatively narrow spectral band (4-6 nm) were employed. Coefficients  were calculated from the action spectra of trap oxygenation (Fig. 2) and spectra of laser radiation (in energy units) using Eq. 2 (see [27,28] for details). Two methods of analysis were applied. According to one of them, A1273 was carefully measured in CCl4 using Eq. 6 and relative rates of oxygenation of three traps -DPIBF, tetracene and rubrene upon photosensitized and direct oxygen excitation [28]. It was established that for all traps the relative oxygenation rates upon photosensitized and direct oxygen excitation normalized to power of exciting radiation and the fraction of light absorbed by porphyrins were ~10 4 : (Vpsr /Vlasr)/(Ipsex/αIlas) 10 4 .
Hence, if 100% exciting light is absorbed by porphyrins, the efficiency of photosensitized oxygen excitation is 10 4 times higher than the efficiency of direct oxygen excitation. In real experiments porphyrin absorbed not more that 10% exciting light, therefore relative efficiency of photosensitized oxygen excitation was by one order smaller, therefore the ratio corresponding to Eq. 7 is about 10 3 [28].
In addition, it was found that A1273 in CCl4 is almost two times greater than the value reported in ref.
[21] ( Table 1). The difference was shown to be mainly due to the fact that the real value of the molar absorption coefficient for tetracene was almost two times smaller than the literature value applied in the previous work [21] (see [28] for details). The obtained A1273 in CCl4 was then used for correction of numerous results of our previous measurements of A1273 in other organic solvents, which were estimated using Eq. 4 and CCl4 as the reference solvent [28] ( Table 1). As mentioned above, these results were obtained upon direct laser excitation of oxygen from relative rates of DPIBF oxygenation in CCl4 and other organic solvents having no OH groups in their molecules.
According to the second method, the rates of DPIBF oxygenation were compared upon direct and photosensitized oxygen excitation in each solvent. The   Table 3 The radiative rate constants (kr) and the phosphorescence quantum yields (Φr) for triplet-singlet transitions in oxygen molecules dissolved in carbon tetrachloride estimated from the experimentally obtained ε1273 and ε765 (Table 1) using Eq. 8.  Table 1 is supported by apparent correlation of their relative values with the relative values of kr/n 2 for the 1 Δg-state of singlet oxygen in different solvents (Table 2). Thus, the data of Table 2 indicate that one and the same mechanism is responsible for the solvent dependence of both kr calculated from the quantum yield of singlet oxygen phosphorescence at 1270 nm and ε1273 obtained by methods of laser photochemistry. Both coefficients are about 1000 times greater than those in rarefied gas (see [21] for refs). It follows from Minaev's theoretical considerations that due to spinorbit coupling in O2 molecules and perturbations caused by their collisions with solvents, the a 1 Δg ← Х 3 gtransition which is purely magnetic in the rarefied gas becomes allowed as electric dipole in solutions [34,35].
Hence, one can estimate the absolute value of kr from the obtained 1273 using the equation for the electric dipole transitions (so called, Kravetz equation). The simplest form of this equation was provided by Strickler and Berg [36]: kr = n 2 2.88×10 -9 ν 2 Δν εox g1/g2, where n is the solvent refractive index, ν is the transition frequency in cm -1 , Δν is the half-width of the oxygen absorption band (FWHM), εox is the molar absorption coefficient of oxygen in M -1 cm -1 and g1 and g2 are multiplicity of the initial and final states, g1/g2 = 3/1 (see also [26] and refs therein). Table 3 shows the results of application of Eq. 8 to the parameters of oxygen obtained in CCl4 (Table 1) and compares these results with the experimentally measured values obtained by different groups. Taking into account an approximate character of Eq. 8 and scattering of the experimental numbers, correlation between the calculated and measured kr is satisfactory.

Absorbance at 765 nm
The first paper, in which A765 was estimated in airsaturated carbon tetrachloride was published in 2007 [24]. Tetracene was used as the singlet oxygen trap and the tunable titan-sapphire laser was applied for oxygen excitation. The maximum of the photooxygenation action spectrum was observed at 765 nm (Fig. 3); A765 was roughly estimated to be 3.5 times smaller than A1273 and ε765  10 -3 M -1 cm -1 was obtained [24]. Later, similar measurements were performed using DPIBF as a singlet oxygen trap. Values of A765 were estimated in carbon tetrachloride, acetone, ethanol and water. It was established that ε765  10 -3 M -1 cm -1 in these solvents [26].
After further improvement of the measurement procedure and data analyses with accounting for coefficients  (see previous section) it was found that ε1273 is almost two times greater than that obtained previously and the ratio (ε1273/ε765) was equal to ~7 in CCl4. A slight increase of ε765 with the increase of solvent polarity was also reported [27,28]. The most recent studies were performed based on Eq. 6 and comparison of the rates of DPIBF oxygenation upon direct and photosensitized oxygen excitation. In these experiments we used a new diode laser with the emission maximum at exactly 765 nm and 2.5 nm bandwidth. Several additional solvents were studied. Some results are presented in Table 1 and Fig.  5. It is seen that the ratio (ε765/ε1273) varies within 0.1-0.15 in non polar solvents, increases with the increase of solvent polarity to about 0.3 in acetone and ethanol and to 0.75 in water. Exclusion is 1-bromohexane, which despite low polarity, shows relatively high ε765. This effect can be attributed to the presence of heavy atom (bromine), which is known to enhance the 1 Σg + (v=0)  3 g -(v=0) transition [9,31]. Based on these values of kr, the quantum yield of 1 O2 phosphorescence at 765 nm can be estimated. Table 3 indicate that even in CCl4 the phosphorescence yield of this emission is very low. In other organic solvents and water the lifetime of 1 Σg + is known to be much smaller. For instance, it was estimated that Σ  30 ps in ethanol and 6.5 ps in water [38,39].
Therefore, the quantum yield of phosphorescence in these solvents should be by ~4 orders smaller than in CCl4, being about 10 -11 .

Absorbance at 1070 nm
The absorption coefficient for the band at 1073 nm corresponding to the transition 1 Δg (v=1) ← 3 g -(v=0) was reported in Refs. [26,28], in which oxygenation of tetracene and DPIBF solutions were studied under irradiation by the Nd-Yag laser (1064 nm) or by powerful diode lasers (1060 nm). As a result, we arrived to the ratio A1273/A1073 = 60±15. Fig. 6 illustrates relative intensities of the major oxygen absorption bands in CCl4. Comparison with Fig. 2 shows that the ratio A1073/A1273 obtained in our experiments is 30-40 times smaller than in the absorption spectra of oxygen dimols (O2)2 (see also [9,11,26]). Judging by the ratio A1073/A1273, the concentration of oxygen dimols is very low under normal conditions being less than 3%. However, the ratio A1073/A1273 is a mirror image of the intensities of the major bands in the phosphorescence spectrum of monomeric singlet oxygen at 1274 and 1590 nm I1590/I1274 (see refs in [26,28]).

Conclusions
Thus, application of the methods of laser photochemistry allowed measurement of the absorption coefficients for major absorption bands of monomeric oxygen molecules dissolved in organic solvents and water under ambient conditions. This information has never been reported before our studies. Knowledge of these coefficients is obviously of basic importance for photonics of triplet and singlet oxygen because it provides information on the electronic structure of oxygen molecules and their interaction with environment. Recently, major results of our studies were confirmed by other researchers [29][30][31]. Although at this state of research numerical values reported by different groups sometimes do not quite coincide (only in acetone identical values were obtained by all authors) the reported values are still of the same order of magnitude. The difference does not exceed the factor of 3. As mentioned in the present paper and previous papers of our group, the values reported by our group also passed through certain evolution due to improvement of measurement techniques and data processing. As mentioned above, to obtain correct numbers one should take into account many parameters, determination of which is not always simple.
Biomedical importance of this research is also apparent. Our experiments demonstrate that direct laser excitation of oxygen molecules is a real natural process, which can be detected using modern laser and spectroscopic techniques. However, the effectiveness of this process is very low, because the absorption coefficients of oxygen have a very small value. The rate of photosensitized singlet oxygen production is 10 3 -10 4 times greater than the rate of direct oxygen excitation. Therefore, it is difficult to expect that under moderate laser power allowed for photodynamic and laser therapy direct excitation of free oxygen dissolved in cell structures causes appreciable destruction of living tissues. It was shown recently that reasonable rates of cell damage can be observed under laser power of 200 W/cm 2 that greatly exceeds the excitation power allowed for medical applications (usually < 200 mW/cm 2 ) [40].
However, low-power laser radiation might influence enzyme-bound oxygen molecules whose concentration is much higher. The data are reported that singlet oxygen produces signaling effect, triggers expression of antistress genes and apoptosis and stimulates immune system response [41][42][43][44]. In conclusion it is worth noting that according to our results, the lasers with the wavelength 765 nm are probably more appropriate for oxygen excitation in biomedical systems than lasers emitting at 1273 nm.