1. Introduction

For several decades, fluorescence has served as the primary optical tool with which the structure and chemical environment of biomolecules has been probed [1]. Incorporated into a diagnostic process, however, fluorescence spectroscopy often suffers from poor signal-to-noise ratios, and the low degree of optical coherence in the emitted radiation complicates or precludes the retrieval of phase information. The recent realization of lasing in several biomolecules, including green fluorescent protein [2–4], luciferin [5], and riboflavin [6,7], has introduced optical sources having considerable promise as biosensors [8] or simply as biocompatible emitters [9]. By integrating optically-active biomolecules with a resonator, the nonlinearity of a biolaser operating near threshold can be leveraged so as to increase significantly (relative to fluorescence spectroscopy) the sensitivity in detecting light-biomaterial interactions, examining molecule-solvent chemistry, or imaging cells in vivo. Despite the demonstration of stimulated emission in several luminescent biomolecules over the past four years, realizing the potential of coherent biomolecular optical sources as in situ probes will require detailed characterizations of these oscillators with respect to mode structure, temporal history, and polarization behavior. We report here such results for the biomolecular laser based on flavin mononucleotide (FMN), a derivative of vitamin B₂. More than 20 longitudinal modes, resolved for the first time, provide a frequency comb against which precise measurements of FMN-solvent interactions can be made. Partially substituting glycerol for water as the solvent reduces the FMN laser pump energy threshold to < 30 µJ/mm², more than a factor of 4 smaller than that observed when water alone is the solvent. The increased efficiency and gain bandwidth of the laser are attributed to oxygen-enhanced deactivation of the excited FMN triplet species and, thus, a significant reduction in the distributed absorption coefficient, α. The sensitivity of the spectral, polarization, and temporal behavior of the FMN laser to the identity of the solvent attests to the potential of this biomolecular oscillator as a biochemical and biomedical diagnostic.

Co-enzymes such as FMN and flavin adenine dinucleotide (FAD) are closely related biomolecules that share the same isoalloxazine ring and differ only in the moiety terminating the ribityl chain from which riboflavin draws its name [10]. For FMN and riboflavin, the species occupying the terminal hydroxyl position in the ribityl chain is a hydrogen atom and a phosphate group, respectively [11]. Isoalloxazines are known to be the chromophores responsible for the bioluminescence of marine bacteria, and the fluorescence quantum yield for FMN (23%) is similar to that for other flavins [6, 12].

2. FMN spectral narrowing and threshold behavior

In the present experiments, solutions of riboflavin 5′-monophosphate sodium salt hydrate (FMN) (purchased from Sigma Aldrich, F2253) were prepared in either deionized water or glycerol mixtures, and 100 µL of the solution was placed onto the surface of a flat laser mirror having a reflectivity at 355 nm and 570 nm of 5% and >99.7%, respectively. In the case of FMN/glycerol mixtures, samples were vortexed for 20 min., followed by sonication for 30 minutes to ensure homogeneous mixing. A second flat mirror, having the same transmission and reflectance spectra as the first, was positioned 375 µm above the lower reflector with a spacer. A frequency-tripled Nd:YAG laser system generated 10 ns FWHM, 355 nm pulses that were focused and directed through the upper mirror (and normal to the mirror surface) with a lens having a focal length of 50 mm. Fluorescence and laser emission produced by the FMN solution was separated from back-scattered pump radiation by a dichroic beamsplitter and detected by either a photodiode (risetime < 1.5 ns) or a 0.75 m spectrograph having a gated, intensified CCD array at the exit plane. The dispersion of the spectrograph/CCD detection system is 1.03 nm/mm in first order, and absolute energies of the FMN laser pulses were measured with a calibrated silicon detector and a bandpass filter having peak transmission at 570 nm.

FMN emission spectra, recorded at low resolution (Δλ = 1 nm) and with water as the solvent, are presented in Fig. 1. For a pump (355 nm) energy fluence of E_p = 61 µJ/mm², a fluorescence continuum having a width of 50 nm (FWHM) is observed but stimulated emission peaking at λ ≈570 nm appears when E_p reaches 115 µJ/mm². Self-absorption by FMN appears to be responsible for the red shift of the laser spectrum relative to peak fluorescence. The emergence of laser radiation from FMN is accompanied by spectral narrowing and the collapse of the spontaneous emission background. As the pump energy fluence is increased further, the laser output intensity grows rapidly, the spectral width is 7 nm when E_p = 232 µJ/mm², and the structure superimposed onto the laser spectra of Fig. 1 is due to partially-resolved longitudinal modes. The appearance in Fig. 1 of a clear pump energy threshold for this optical oscillator is confirmed by measurements of the laser output energy with a calibrated detector. As illustrated in Fig. 2, extrapolation of the data for FMN:H₂O solutions (represented by the black symbols (o)) to zero output energy shows the threshold pump energy to be 110 ± 10 µJ/mm², which is in agreement with Fig. 1.

 

Fig. 1 FMN fluorescence and laser emission spectra recorded at low resolution (Δλ = 1 nm) as the 355 nm pump energy fluence E_p was increased from ~61 µJ/mm² to 232 µJ/mm². Laser threshold occurs for E_p ~115 µJ/mm², and the observed spectral width for the FMN laser is ~7 nm FWHM. For clarity, the intensity of the 61 µJ/mm² (fluorescence) spectrum was increased by an order of magnitude. The inset illustrates the chemical structure of FMN, and all of the spectra presented here were acquired with 10 mM FMN:water solutions.

Download Full Size | PDF

 

Fig. 2 Dependence of the relative FMN laser pulse energy (expressed in nJ) on the pump energy fluence for both FMN/water and FMN/glycerol/water solutions. All of these FMN solutions are 10 mM, and data are presented for water (o), 1:3 (•) and 1:2 ($\square$) glycerol/water solutions. The inset is a generalized diagram of the experimental arrangement, and note that the ordinate is logarithmic.

Download Full Size | PDF

Similar data have also been acquired when glycerol is the solvent and representative results are given in Fig. 2. All of the FMN solutions are 10 mM and the solid red circles (•) represent data obtained for 1:3 glycerol solutions (1 part 10 mM FMN and 2 parts glycerol) whereas the open green squares ($\square$) are associated with 1:2 (1 part 10 mM FMN and 1 part glycerol) solutions. The contrast between the laser performance of FMN in water and glycerol is stark. Note, for example, that the threshold energy fluence for FMN:glycerol solutions is more than a factor of 4 lower for 1:3 (glycerol/water) solutions than that measured when water alone is the solvent. Furthermore, the threshold value of E_p for 1:2 mixtures (~45 µJ/mm²) is approximately twice that for the 1:3 solutions, or precisely the value expected if the pump fluence is inversely proportional to the glycerol content in the solutions. Conversely, removing glycerol entirely and diluting FMN in water alone results in the laser deteriorating rapidly, as shown by the 5 mM and 10 mM FMN:water data of Fig. 3. It should also be mentioned that the slope efficiencies for the glycerol data of Fig. 2 are considerably higher than the values measured for FMN:H₂O lasers (Fig. 3). Specifically, the maximum slope efficiency for the 1:3 glycerol/water data of Fig. 2are is more than a factor of six larger than that observed in the absence of glycerol.

 

Fig. 3 Data similar to those of Fig. 2, comparing 5 mM and 10 mM FMN:water solutions.

Download Full Size | PDF

Much of the credit for the dramatically improved performance of this visible (λ ≈570 nm) oscillator when glycerol is the solvent should be attributed to replenishing of the gain medium by molecular oxygen. This presumption is consistent with early studies on flavin photochemistry which revealed that “…the yellow color of riboflavin solutions faded when exposed to light but was partially restored upon readmission of oxygen” [13]. Upon dispensing a 100 µl droplet of a FMN:glycerol:water mixture onto the bottom mirror of our plane-parallel resonator, it was observed that air bubbled through the pipette tip as a consequence of forward pipetting, resulting in the formation of bubbles in the suspension which was facilitated by the low surface tension of glycerol [14]. Previous studies have indicated that the photoexcitation of FMN in the 320-400 nm wavelength region (UVA) culminates in the excited molecule residing in the metastable triplet state, owing to an intersystem crossing similar to that in dyes [15]. However, in the presence of oxygen, the FMN triplet species is deactivated, as expressed by the reaction [10,16]:

3. Temporal narrowing of FMN laser

All previous reports of stimulated emission in biomolecules have regarded data analogous to those of Figs. 1-3 as sufficient evidence of lasing. However, temporal narrowing is an equally important indicator of the onset and decay of optical oscillation in pulsed systems, and Fig. 4 shows the normalized pump and FMN emission waveforms observed for three values of pump energy fluence. Both waveforms were recorded for each pump laser shot by a 2.5 GHz bandwidth oscilloscope and matched photodiodes. The inset to panel (b) of Fig. 4, which gives the response of the two photodiodes to the same pump laser pulse, demonstrates the reliability of this experimental approach. It should also be mentioned that the undulations on the pump waveforms are the result of beating between longitudinal modes because the oscillator for this particular pump laser was not injection-seeded.

 

Fig. 4 Normalized pump (black) and FMN:H₂O laser (red) waveforms for three values of the 355 nm pump energy fluence (E_p): (a) 34 µJ/mm², (b) 178 µJ/mm², and (c) 268 µJ/mm². In (a), the laser is below threshold but, in (b) and (c), temporal narrowing of the output pulse is evident. The inset to (b) shows the superposition of two laser pump waveforms of the same pulse, recorded separately by two matched photodetectors. The inset to (a) illustrates that the decay of the experimental fluorescence waveform (shown in red) is well-represented by a single, decaying exponential (τ = 6.0 ns, black line). In panel (c), the temporal behavior of the falling portion of the laser signal is displayed on a semi-logarithmic scale, demonstrating that the laser intensity decays according to a double exponential, ${\text{Ae}}^{-\text{t}/{\tau }_1}+{\text{Be}}^{-\text{t}/{\tau }_2}$, where A and B are constants. The lifetimes τ₁ and τ₂ are found to be 1.73 ns and 1.29 ns. Note the change in the scale of the abscissa between panel (a) and that of panels (b) and (c).

Download Full Size | PDF

When the pump energy fluence is below threshold (Fig. 4(a); E_p = 34 µJ/mm²), the fluorescence waveform exhibits a breadth of 8.1 ns FWHM. Owing to the significant overlap between the pump and fluorescence waveforms of Fig. 4(a), accurately determining the fluorescence temporal decay constant (τ⁻¹) necessitated the deconvolution of the two emission traces. Doing so with commercial software (FluorTools Decay Fit 1.3) yields a single decay constant of τ⁻¹ = (6.0 ± 0.5 ns)⁻¹, which is consistent with the known radiative lifetime for the S₁ manifold of riboflavin [17]. If, however, E_p is increased to 178 µJ/mm² (Fig. 4(b)), the temporal width of the FMN emission pulse drops abruptly to 2.4 ns because of rapid depopulation of the S₁ manifold of states by stimulated emission. The onset of lasing is delayed by 7 ns with respect to the beginning of the pump pulse but peak intensity is reached 1.5 ns after the system passes threshold. As illustrated in Fig. 4(c), at still higher values of E_p (268 µJ/mm²) the laser pulse width decreases to 1.75 ns, and the rapid rise (<1.5 ns), but exponential decay, of the output intensity are reminiscent of the behavior of a Q-switched oscillator. The inset to panel (c) of Fig. 4 shows the laser intensity decays according to a double-exponential: ${\text{Ae}}^{-\text{t}/{\tau }_1}+{\text{Be}}^{-\text{t}/{\tau }_2}$where A and B are constants. The best fit of this expression to the data yields τ₁ = 1.3 ns and τ₂ = 1.7 ns. The latter of these is within 2% of the value given by the expression for the photon lifetime of the resonator: ${\tau }_p=\frac{2nd}{c[1-R_1{R}_2]}=1.67$ns where d (the mirror spacing) is 375 µm, n is the refractive index of the gain-medium (1.333 for water at 570 nm and 293 K) and R₁ and R₂ are the reflectivities of the two cavity mirrors at the laser wavelength(s) (R₁ = R₂ = 99.9%). Furthermore, since can also be written as $\frac{Q}{{\omega }_o}$ where ${\omega }_o$ is the center radian frequency for the oscillator (3.3x10¹⁵ sec⁻¹), the resonator Q is found to be 5.6x10⁶. Before leaving this section, we note that temporally resolving the laser output intensity provides additional information for biolasers that previously was inaccessible or difficult to estimate. For example, the time interval between the arrival of the pump pulse and the onset of lasing (nominally 7.5 ns in Fig. 4(b) and 4(c)) provides a convenient estimate of the gain coefficient given by g_ol ≈30 where l is the total gain length corresponding to 7.5 ns. The result, g_o ≈14%-cm⁻¹, is a conservative value because it does not account for distributed loss (such as triplet absorption) in the gain medium.

4. Impact of glycerol on FMN laser spectra

FMN laser spectra were also recorded at spectral resolutions considerably higher than that of Fig. 1 and an example, representative of those obtained throughout these experiments, is given in Fig. 5, for which the spectral resolution is 0.02 nm. When E_p = 300 µJ/mm², the gain coefficient of this oscillator is sufficiently high (estimated above to be ~0.14 cm⁻¹) that more than 20 longitudinal modes are above threshold and literally scores of measurements found the free spectral range (FSR) to be 0.325 ± 0.005 nm, where the uncertainty represents one standard deviation. This is precisely the value expected for the FSR (≡ λ²/2nd) when λ = 570 nm, d = 375 µm and n = 1.333 for H₂O at 570 nm. As exhibited by the consistent asymmetry of the FMN laser mode profile of Fig. 5 and, specifically, the departure of specific longitudinal mode intensities from a continuous, symmetric spectrum decaying monotonically on either side of line center, inhomogeneous broadening appears to contribute significantly to the FMN gain profile. This, in turn, suggests that the laser mode profile will be sensitive to the FMN molecule’s environment, a conclusion borne out by laser spectra such as those presented in Fig. 6. Adding glycerol, for example, to the 10 mM FMN/water solutions of Figs. 1, 3-5 (in a 1:3, glycerol:H₂O ratio) results in a shift of the spectrum by more than 1 nm to shorter wavelengths (shown in blue in Fig. 6). Furthermore, the FSR of the FMN/glycerol/water spectrum is 0.302 ± 0.006 nm, which is within 1.3% of that expected from the glycerol and water indices of refraction [18]. It must be emphasized that significant differences from the spectrum of Fig. 5 (FMN solvated in water) are evident for glycerol concentrations much smaller than that in Fig. 6 but the 66% v/v glycerol/FMN mixture (blue, Fig. 6) was chosen so as to make the distinctions obvious.

 

Fig. 5 FMN:H₂O laser spectrum for which the pump energy density and spectrometer resolution were 300 µJ/mm² and 0.02 nm, respectively. The free spectral range (FSR ≡ λ²/2nd) where λ is the center wavelength, n is the refractive index of the gain medium, c is the speed of light, and L is the spacing between the resonator mirrors) of 0.325 ± 0.005 nm is consistent with λ = 570 nm, L = 375 µm and n = 1.333 for water.

Download Full Size | PDF

 

Fig. 6 Laser spectra representative of those recorded for FMN/water/glycerol solutions. A pronounced shift of the spectrum to the blue is observed when 10 mM FMN solutions are diluted with glycerol in a 1:3 volume ratio. For the FMN/glycerol spectrum (shown in blue), the longitudinal mode separation (FSR = 0.302 ± 0.006 nm) is consistent with n = 1.429 for glycerol at 570 nm, and the appearance of transverse mode structure is evident. For comparison, a laser spectrum for a 10 mM FMN/water solution is shown in red, and the inset is a magnified comparison of the FMN:H₂O and FMN:glycerol:H₂O spectra in the 567.5-569.6 nm region.

Download Full Size | PDF

From a spectral perspective, another interesting effect of introducing glycerol to the gain medium is the surprising appearance of transverse mode structure, as evidenced by the weak features on the blue side of the most intense longitudinal mode peaks of Fig. 6. The frequencies of the q^th longitudinal mode of the mn^th Hermite-Gaussian transverse modes of an optical resonator are given by the relation [19]:

We should also note that, although the spectral resolution available for these experiments (Δλ = 0.02 nm; resolving power ~280,000) is beyond that reported in the past for biolasers, the resolution is insufficient to observe the spectral widths of the individual longitudinal modes of Figs. 5 and 6. That is, the Q value of 5.6x10⁶ mentioned earlier implies a longitudinal mode linewidth of ~10⁻⁴ nm which is two orders of magnitude lower than that accessible at present.

5. Polarization characteristics of FMN laser emission

The impact of glycerol on FMN is further clarified by measurements of the polarization of the laser output beam. A polarizing beamsplitter (denoted PBS) served to resolve the s (vertical) and p (horizontal) polarization components of the output intensity, and the relative contributions of each component to the laser output are shown by the waveforms of Fig. 7 for E_p = 182 µJ/mm². Because the Nd:YAG pump is p-polarized, the FMN laser is preferentially polarized horizontally but, for the 10 mM FMN/water solution (panel (a)), the peak intensity for the vertical polarization is ~58% of that for the p-polarized signal. This characteristic of the laser radiation can be expressed in terms of the ellipticity or polarization (P) of the beam, where P is given by: $P=\frac{I_p-I_s}{I_p+I_s}$. Defining P according to the peak values of the intensities I_s and I_p, the polarization of the FMN laser in the absence of glycerol is 0.44 ± 0.08. When glycerol is added to the gain medium, however, the increase in solution viscosity is known to inhibit the rotation of FMN [20]. The result is the suppression of the s-component of the polarization and an FMN laser output that more closely reflects the polarization characteristics of the pump. As illustrated in panel (b) of Fig. 7, the addition of glycerol results in P rising to 0.78 ± 0.08 and, interestingly, the FMN laser pulsewidth decreases from 2.8 ns in Fig. 7(a) to < 2.0 ns when glycerol is introduced to the gain medium solution.

 

Fig. 7 FMN laser waveforms for which the s and p-polarized components of the output radiation (shown in black and red, respectively) are resolved by a polarizing beamsplitter (PBS): a) 10 mM FMN in water; b) 1:3 ratio of glycerol and 10 mM FMN in water. The insets to a) and b) are schematic diagrams of the experimental arrangement. All of the data were acquired for a pump energy fluence of 182 µJ/mm².

Download Full Size | PDF

The results of Fig. 7 can be described qualitatively by the Perrin equation [1] that relates the fluorescence lifetime (τ) of a spherically-symmetric fluorophore, rotating in solution, to its optical correlation time (θ):

where r is the fluorescence anisotropy (calculated from the polarization P) and r_o is the intrinsic anisotropy (i.e., in the absence of depolarization). The rotational correlation time is given by the expression:

where η is the viscosity of the solution, and V is its volume. Although Eqs. (3) and (4) are in qualitative agreement with the waveforms of Fig. 7, this fluorescence-based model is not strictily applicable to the present laser experiments in which the optical field intensity builds up quickly from the noise (spontaneous emission) by stimulated emission. Indeed, the observed increase in anisotropy in Fig. 7(b) (resulting from the introduction of glycerol to the solution) is larger than that expected in fluorescence experiments for which P must be ≤ 0.5.

6. Conclusion

In summary, the temporal, spectral, and polarization characteristics of an FMN biomolecular laser have been examined in detail. When the energy fluence of the 355 nm pulsed pump laser exceeds ~115 µJ/mm² (with water as the solvent) for a plane-parallel resonator having a mirror separation of 375 µm, stimulated emission is observed over a ~7 nm wide spectral region and peak intensity is generated at 570 nm. The threshold pump energy fluence falls by more than a factor of 4 when glycerol is added to the solvent. Also, the shot-to-shot stability of the laser spectrum is excellent and pulses as short as 1.75 ns are emitted by the laser. The measured spontaneous emission lifetime of FMN (6.0 ns) and cavity photon lifetime (1.7 ns) are in agreement with previously published values or calculations. The longitudinal mode structure of the FMN laser has been observed with a spectral resolution of 0.02 nm, and >20 modes are above threshold when water is the solvent and E_p = 300 µJ/mm². Owing to the available spectral and temporal resolution, the influence on an FMN laser of its chemical environment can be detected readily. The output of the FMN laser is slightly p-polarized when water is the only solvent, but the addition of glycerol suppresses rotational diffusion of the molecule which, in turn, increases the degree of horizontal polarization of the laser. These optical properties underscore the potential of FMN and other biomolecular lasers as probes for a broad range of biochemical and biomedical diagnostics, molecule-molecule interactions, and in vivo imaging of cells.