Rapid-scanning frequency-domain fluorometer withpicosecond time resolution

Frank V. Bright and Gary M. Hieftje

A new frequency-domain fluorometer is described which is capable of operating over a frequency rangebetween 10 MHz and 1 GHz. The instrument employs a mode-locked Ar-ion laser as an 82-MHz repetition-rate excitation source and an inexpensive extracavity acoustooptic light modulator for secondary-frequencygeneration. The instrument is very rapid and can acquire, in '10 s, the entire frequency spectrum necessaryto determine excited-state lifetimes. Importantly, these frequency spectra are characterized by signal-to-noise ratios of at least 20 for moderate fluorescence signals (10-20 uA detector current). With this newinstrument, multiple exponential decays can be easily resolved, and single exponential decays can bedetermined at a precision (standard deviation) of better than 20 ps.

1. Introduction

The determination of fluorescence lifetimes is tradi-tionally carried out by one of two techniques. Thefirst, utilizing the impulse response, is a time-domainexperiment in which a pulse of light excites the fluores-cent sample under study. The resulting fluorescencedecay is then monitored directly using fast detectionelectronics.'-5 The second approach is the frequency-domain analog of the impulse experiment. In thiscase, high-frequency (MHz-GHz) sinusoidally modu-lated radiation is used to excite the sample. Followingthis excitation, the phase shift and amplitude demodu-lation of the fluorescence are determined.6-10 By mea-suring the phase shift and demodulation factors, onecan easily determine the luminescence lifetime.6' 3

The first frequency-domain fluorometer was de-scribed in 1927 by Gaviola and exhibited a precision of0.5 ns.14 Since this earliest frequency-domain fluo-rometer, several multifrequency instruments have alsobeen constructed. The first such instruments wereconstructed by Hauser and Heidt15 and further modi-fied by Haas and Hauser,16 but neither demonstratedthe ability to resolve complex multiexponential de-

When this work was done both authors were with Indiana Univer-sity, Chemistry Department, Bloomington, Indiana 47405; F. V.Bright is now with SUNY-Buffalo, Department of Chemistry, Buf-falo, New York 14214.

Received 2 February 1987.0003-6935/87/173526-04$02.00/0.t 1987 Optical Society of America.

cays. In 1983, Gratton and Limkemanl7 and, in 1985,Lakowicz and Maliwal'8 described multifrequency flu-orometers that are capable of resolving multiexponen-tial decays. Most recently, Lakowicz et al.'9 an-nounced a variant of the above instruments thatoperates at frequencies up to 2 GHz. In all these cases,however, the data collection time is at least 1 h; thiscollection time depends directly on the number offrequencies employed.

In this paper we describe a continuation of our workemploying the harmonic content of a mode-locked la-ser for the determination of fluorescence lifetimes.20-26

Specifically, we employ a mode-locked Ar-ion laser asthe excitation source with an extracavity acoustoopticlight modulator for generation of frequencies over theentire range from 10 MHz to 1 GHz. This approach issimilar to that employed by Gratton and co-workers,27

but differs significantly in that our frequency-domaindata collection time is only 10 s (worst case SNR = 20)for the entire frequency spectrum (dc to 1 GHz). Fur-thermore, the collection time is independent of thenumber of frequencies employed. In addition, theinstrument exhibits adequate resolving powers formono-, bi-, and triexponential decays if the compo-nents are reasonably well separated in decay time.For example, fluorophores of equal intensity and withnominal luminescence lifetimes of 0.80, 3.70, and 15.10ns can be successfully resolved with the instrument.For this resolution, required data-collection time is 3min, a factor of at least 20 faster than other frequency-domain approaches. The new instrument is able alsoto determine luminescence lifetimes which are signifi-cantly longer (15.10 ns) than the interpulse separation(12.2 ns) of the laser source.

3526 APPLIED OPTICS / Vol. 26, No. 17 / 1 September 1987

BD Sample

< , ] - _- JM2

Mono.

SA MCA

IBM-PC DS

Fig. 1. Schematic diagram of the rapid frequency-scanned fluo-

rometer. Abbreviations represent: MLD, mode-locker driver; M1and M2, mirrors; AOM, acoustooptic modulator; DU, acoustoopticmodulator driver unit; Ramp, 500 Hz voltage ramp (0-1 V dc); BD,

beam dump; Mono., double monochromator; D, photomultiplier

tube; SA, spectrum analyzer; MCA, multichannel averager; DS,digital oscilloscope.

11. Experimental

A. Reagents

Rhodamine 6G (rhodamine 590, Exciton ChemicalCo.), rose bengal (Eastman Kodak), and rubrene (Al-drich) were used without further purification. A 1.00mM stock solution of each fluorophore was preparedby dissolving the appropriate amount of each in 100mliter of solvent. Absolute ethanol (AAPER Alcohol& Chemical Co.) was used as the solvent for rhodamine6G and rose bengal, and benzene (Mallinkrodt) em-ployed for the rubrene. It is important to note thatrubrene is a suspected carcinogen and appropriate careshould be employed in handling it. The solutions weresonicated for 30 min to ensure complete mixing. Thestock solutions were then diluted 1:103 with solvent, toyield 1.00 ,uM working solutions. For all results pre-sented here the solutions were degassed (dry N2 for 10min), and all measurements were carried out in cappedquartz cuvettes (Spectrocell Co.).

For all lifetimes reported here we employed a scat-tering suspension of uniform latex spheres in water forgeneration of the reference frequency spectrum.

B. Instrumentation

Figure 1 shows a schematic diagram of the fluores-cence-lifetime instrument used in this study; all com-ponents were rigidly mounted on an optical table (1.2X 3.7 m, Unidex). The instrument utilizes a mode-locked Ar-ion laser (model 171 laser, model 342 mode-locker, and model 452 mode-locker driver, SpectraPhysics, Inc.). The laser was operated at the 514.5-nmlaser line with an average output power of 110 mW andat a mode-locker frequency of 40.9790 MHz. The

incident beam, consisting of a train of pulses at arepetition rate of -82 MHz, was directed to the samplevia mirrors Ml and M2. Prior to exciting the sample,the laser beam passes through an acoustooptic lightmodulator (AOM; model AOM-30, IntraAction Corp.),which is driven by an impedance-matched acoustoop-tic deflector driver (DU; model DE-30X, IntraActionCorp.). The frequency at which the acoustoopticmodulator is driven is controlled by an analog input tothe acoustooptic deflector driver and is linearly relatedto the voltage applied at its analog input. For ourpurposes a linearly swept voltage ramp from a simplefunction generator (model 1600 function generator,Krohn-Hite Corp.) allows us to quickly sweep (scantime 2 ms) the modulator frequency. The effect is tofill in the 82-MHz holes in the frequency spectrum.2 0-2 7

In essence, this arrangement produces an amplitudemodulation of the mode-locked pulse train from thelaser; in the frequency domain, the effect is to generatesecondary frequencies (i.e., sidebands)27 that occur atthe sum and difference between the pulse-train fre-quency (82 MHz) and that of the modulator.27 Specif-ically, the frequency composition of the primary pulsetrain consists of a fundamental at 82 MHz, and har-monics occur at 82-MHz intervals. By amplitude-modulating the laser output at 30 MHz, we introducesecondary frequencies at +30 MHz on each of thefundamentals.

The resulting fluorescence was collected at an angleof 90° to the incident beam by a double monochroma-tor (Mono.; model 1680 Spectramate, Spex, Inc.) witha spectral bandpass of 9.0 nm. This wide bandpasswas chosen to provide a high signal-to-noise ratio. De-tection of the resulting signal was achieved with aphotomultiplier (D; model 31024, RCA, Inc.) poweredat a biasing voltage of -3500 V dc.

The photomultiplier output was configured differ-ently for ac and dc measurements. For dc detectionthe output from the photomultiplier was simply con-nected to a picoammeter (model 414S, Keithley, Inc.),which recorded the resultant scatter or fluorescencesignal. To record the high-frequency component ofthis signal, the output from the photomultiplier wassent to a microwave spectrum analyzer (SA; model640A-1, Polarad Electronics, Inc.) whose output wasconnected to a multichannel signal averager (MCA;model 1080 instrument computer, model SD-81/2 12-bit digitizer, and model SW-80 sweep controller, Nico-let, Inc.). The averaged frequency spectrum was thensent to a digital oscilloscope (DS; model 3091, Nicolet,Inc.) which permitted serial transfer to an IBM-PCthrough an RS-232 port. Typically, two data files aretransferred to the IBM-PC. The first is a lX scan, andthe second is a 10OX scan which amplifies the weakersecondary frequencies. Data collection and regressionwere all performed with an interactive nonlinear least-squares algorithm (Marquardt) similar to that de-scribed elsewhere.'7 -9, 2 4 ,2 7

1 September 1987 / Vol. 26, No. 17 / APPLIED OPTICS 3527

E

a

C:

0

0

-0

Frequency (MHz)

Fig. 2. Alternating-current amplitude vs frequency spectrum for514.5-nm mode-locked laser radiation scattered through the instru-ment. A scattering suspension of uniform latex spheres was em-ployed. This spectrum was collected during a 5-s averaging period.The large spikes on the left and right sides of the plot correspond tothe 82- and 164-MHz pulse-frequency components and are saturat-ed. The intermediate components have been generated by the

extracavity modulator.

111. Results and Discussion

Figure 2 shows the portion of a typical frequencyspectrum between the 82- and 164-MHz fundamentalpulse-train frequencies (large spikes). The intermedi-ate frequencies occur at the sum and difference withthe extracavity modulator (AOM; Fig. 1) frequency.The voltage ramp was swept at a 500-Hz repetition rateto generate these secondary frequencies, and a collec-tion time of 5 s was employed. The photomultipliertube current in this case was 15 MA.

Figures 3-5 show the variation in demodulation fac-tor with modulation frequency for rose bengal, rhoda-mine 6G, and rubrene, respectively. The solid linerepresents the best fit for a single Lorentzian (corre-sponding to a single exponential decay) for each of thefluorophores. Single exponential lifetimes deter-mined from these plots (average of ten replicate mea-surements + standard deviation) of 0.79 I 0.01,3.62 +0.01 and 15.10 I 0.02 ns for rose bengal, rhodamine 6G,and rubrene, respectively, agree well with values of0.80 ns,5 3.70 ns2 5, and 15 ns5 reported elsewhere. Ta-ble I shows also that a biexponential decay, if assumed,does not appreciably affect the goodness-of-fit withthese plots, as evidenced by a x2 (chi-squared) test.

In an effort to more completely characterize the newinstrument, a 1:1:1 (intensity ratio) mixture of rosebengal, rhodamine 6G, and rubrene was prepared andevaluated. The collection time in this case was 3 minover the frequency region of interest. Table I showsthe results for the best-case fit for monoexponential(X = 102), biexponential (X2 = 9.54), and triexponen-tial (X2 = 1.67) decay curves. Only the triexponentialdecay yields a reasonable x2 value. Although this x2

value is quite far from the ideal of 1.00, most of thisdeviation can be attributed to instrument noise whichis clearly frequency dependent (Fig. 2). Furthermore,a fourth decay parameter does not significantly im-

Rose BengaI

Frequency MHz)

Fig. 3. Demodulation factor vs frequency spectrum for 0.5 AM rosebengal in ethanol. Circles represent data points, and the solid line isthe best-fit Lorentzian curve corresponding to a fluorescence life-time of 0.79 ns. Collection time is 10 s with 12-bit resolution at 565

nm.Rhod. i n 6G

E

1U 1u 1 000

Frequency MHz)

Fig. 4. Demodulation factor vs frequency spectrum for 0.1 Mrhodamine 6G in ethanol. Circles represent data points, and thesolid line is the best-fit Lorentzian curve corresponding to a fluores-cence lifetime of 3.62 ns. Collection time is 8 s with 12-bit resolution

at 550 nm.RfRubrn

E

e,

10 100 1000

Fr eque ncy CMH )

Fig. 5. Demodulation factor vs frequency spectrum for 2.5 ALMrubrene in benzene. Circles represent data points, and the solid lineis the best-fit Lorentzian curve corresponding to a fluorescencelifetime of 15.10 ns. Collection time is 10 s with 12-bit resolution at

545 nm.

3528 APPLIED OPTICS / Vol. 26, No. 17 / 1 September 1987

1.0 I . I I I

0.8 _

0.6

0.4 - I-1

0.2 40.0 n - 2 ta-- L

0.4 I0.3 - i-

0.2 -

0.1 1

Table 1. Fluorescence Lifetimes Determined with the New Frequency-Domain Fluorometer

Species Exponential i- (ns) fA r2a (ns) f2 T3a (ns) x2

Rose bengal Single 0.79 - 1.12Rhodamine 6G Single 3.62 - - 1.28Rhodamine 6G Double 3.53 0.81 3.75 - 1.09Rubrene Single 15.10 - - 1.42Rubrene Double 15.15 0.94 14.98 - 1.371:1:1 mixtureb Single 6.94 - - 1021:1:1 mixtureb Double 2.56 0.71 19.31 0.29 - 9.541:1:1 mixtureb Triple 0.93 0.37 3.23 0.29 15.88 1.67

a Indicates not possible in the fitting procedure.b Intensity ratios.fi and f 2 are fractional concentrations of the total fE

prove the fit (X2 = 1.54). Importantly, the lifetimesrecovered with this triple exponential decay are within11% of those expected from the individual monoexpon-ential decays.

Finally, a single-frequency scanned cross-correla-tion instrument1 3 1 7-19 exhibits better signal-to-noise(SNR) ratios than does our rapid-frequency scanned

instrument. This conclusion was quantitativelyreached by Ramsey and Hieftje,20 who discussed all thepossible SNR advantages. However, if data collectionmust be rapid (e.g., in chromatography and/or for pho-toactive materials), SNR must always be sacrificed.Indeed, a single-frequency scanned cross-correlationinstrument cannot make such measurements fastenough.

This work was supported in part by the Office of Naval Research, by the National

Science Foundation through grant CHE 83-20053, and by The Upjohn Company.

References

1. J. N. Demas, Excited State Lifetime Measurements (Academic,New York, 1983).

2. J. B. Birks, Photophysics of Aromatic Molecules (Wiley, NewYork, 1970).

3. W. R. Ware, "Transient Luminescence Measurements," in Cre-ation and Detection of the Excited State, A. A. Lamola, Ed.(Marcel Dekker, New York, 1971), Vol. 1, Part A, p. 213.

4. W. R. Ware, "Kinetics of Fluorescence Decay: An Overview," inTime-Resolved Fluorescence Spectroscopy in Biochemistryand Biology, R. B. Cundall and R. E. Dale, Eds. (Plenum, NewYork, 1983), p. 23.

5. I. B. Berlman, Handbook of Fluorescence Spectra of AromaticMolecules (Academic, New York, 1971).

6. R. D. Spencer and G. Weber, "Measurements of SubnanosecondFluorescence Lifetimes with a Cross-Correlation Phase Fluoro-meter," Ann. N. Y. Acad. Sci. 158, 361 (1969).

7. J. B. Birks and D. J. Dyson, "Phase and Modulation Fluorome-ter," J. Sci. Instrum. 38, 282 (1961).

8. A. Muller, R. Lumry, and H. Kokubun, "High PerformancePhase Fluorometer Constructed from Commercial Subunits,"Rev. Sci. Instrum. 36, 1211 (1965).

9. J. R. Mattheis, G. W. Mitchell, and R. D. Spencer, "Phase-Resolved Nanosecond Spectrofluorometry: Theory, Instru-mentation, and New Applications of Multicomponent Analysisby Subnanosecond Fluorescence Lifetimes," in New Directionsin Molecular Luminescence, D. Eastwood, Ed. (ASTM STP822, Philadelphia, 1983), p. 50.

10. R. K. Bauer and A. Balter, "A Method of Avoiding Wavelength-Dependent Errors in Decay-Time Measurements," Opt. Com-mun. 28, 91 (1979).

11. E. R. Menzel and Z. D. Popovic, "Picosecond-resolution fluores-cence lifetime measuring system with a cw Laser and a Radio,"Rev. Sci. Instrum. 49, 39 (1978).

12. L. B. McGown and F. V. Bright, "Analytical Applications ofPhase-Resolved Fluorescence Spectroscopy," Anal. Chem. 56,1400A (1984).

13. E. Gratton and B. Barbieri, "Multifrequency Phase Fluorome-try Using Pulsed Sources: Theory and Applications," Spectros-copy 1, 28 (1986).

14. E. Gaviola, "Phase Fluorometer," Z. Physik, 42, 853 (1927).

15. M. Hauser and C. Heidt, "Phase Fluorometer with a Continu-ously Variable Frequency," Rev. Sci. Instrum. 46, 470 (1975).

16. H. P. Haar and M. Hauser, "Phase Fluorometer for measure-ment of Picosecond Processes," Rev. Sci. Instrum. 49, 632(1978).

17. E. Gratton and M. Limkeman, "A Continuously Variable Fre-quency Cross-correlation Phase Fluorometer with PicosecondResolution," Biophys. J. 44, 315 (1983).

18. J. R. Lakowicz and B. P. Maliwal, "Construction and Perfor-mance of a Variable Frequency Phase-Modulation Fluorome-ter," Biophys. Chem. 21, 61 (1985).

19. J. R. Lakowicz, G. Laczko, and I. Gryczynski, "A 2-GHz PhaseFluorometer," Rev. Sci. Instrum. 57, 2499 (1986).

20. J. M. Ramsey and G. M. Hieftje, "Signal-to-Noise Consider-ations in Fluctuation Analysis Spectroscopic Techniques," inNew Directions in Molecular Luminescence, D. Eastwood, Ed.(ASTM STP 822, Philadelphia, 1983), p. 82.

21. G. M. Hieftje, G. R. Haugen, and J. M. Ramsey, "New Methodfor the Determination of Luminescence Lifetimes by Using theBeat Noise of a cw Laser as a Multifrequency Modulated Sour-ce," Appl. Phys. Lett. 30, 463 (1977).

22. G. M. Hieftje and G. Horlick, "Correlation Methods in theChemistry Laboratory," Am. Lab. 13, 76 (1981).

23. G. M. Hieftje, J. M. Ramsey, and G. R. Haugen, "New Laser-Based Methods for the Measurement of Transient ChemicalEvents," in New Applications of Lasers to Chemistry, G. M.Hieftje, Ed. (ACS Symposia Series 85, Washington, 1978), Chap.8.

24. F. V. Bright, D. A. Wilson, and G. M. Hieftje, "Determination ofSubnanosecond Fluorescence Lifetimes with a UHF TelevisionTuner and a cw or Mode-Locked Laser," Appl. Spectrosc. 40,683 (1986).

25. F. V. Bright, C. A. Monnig, and G. M. Hieftje, "Rapid Frequen-cy-Scanned Fiber-Optic Fluorometer Capable of Subnanose-cond Lifetime Determinations," Anal. Chem. 58, 3139 (1986).

26. F. V. Bright, C. A. Monnig, and G. M. Hieftje, "A New Frequen-cy-Domain Fluorometer for the Rapid Determination of Pico-second Rotational Correlation Times," J. Appl. Phys. 61, 8(1987).

27. J. R. Alcala, E. Gratton, and D. M. Jameson, "A MultifrequencyPhase Fluorometer using the Harmonic Content of a Mode-Locked Laser," Anal. Instrum. 14, 225 (1985).

1 September 1987 / Vol. 26, No. 17 / APPLIED OPTICS 3529