Fluorescence Lifetime Measurement via a Radionuclide-Scintillation

Fluorescence Lifetime Measurement via a. Radionuclide-Scintillation Light Source and Analog. Cross Correlation. Daniel L. Burden, Steven E. Hobbs, and...
0 downloads 0 Views 173KB Size
Download PDF
Anal. Chem. 1997, 69, 1936-1941

Fluorescence Lifetime Measurement via a Radionuclide-Scintillation Light Source and Analog Cross Correlation Daniel L. Burden, Steven E. Hobbs, and Gary M. Hieftje*

Department of Chemistry, Indiana University, Bloomington, Indiana 47405

β-Emitting 90Sr is used with a plastic scintillator to produce excitation-light pulses for fluorescence lifetime analysis. This light source is less expensive, more compact, and much more reliable than traditionally employed excitation sources such as lasers or pulsed flash lamps. The pulse train from this light source varies randomly in amplitude and time. Cross-correlation signal analysis is ideal for such a source because, unlike other time domain techniques, cross correlation takes complete advantage of its random nature. Here we report on the construction of an instrument and the methods employed to make fluorescence lifetime measurements via the new source and an analog correlation processor. Although the light intensity of the scintillator-based excitation source is comparatively low, an adequate signal level can be generated. The fluorescence lifetimes of three fluorophores are measured with a 1-mCi radionuclide to demonstrate a lifetime range from less than 1.5 to 28 ns. Long-lifetime measurements require an extra calibration step in order to compensate for delay cable energy loss. The light collection efficiency of the current instrument was found to be undesirably low; improvements in the instrument optics are suggested that will increase the collection efficiency and enhance the detection capability. The fluorescence (excited-state) lifetime of an analyte species can be measured by a variety of alternative techniques.1,2 Common to all is some form of pulsed or modulated excitation light source. Flash lamps, steady-state arc lamps, and lasers are the most commonly used sources. Flash lamps are typically pulsed by fast, high-voltage electronics and are ordinarily used to measure lifetimes in the time domain via time-correlated single-photon counting (TCSPC). Pulsed lasers can be used in a similar fashion.3 Steady-state arc lamps can be employed with components such as Pockels cells or Debye-Sears devices to produce a sinusoidally modulated excitation beam.4,5 Lifetimes are then determined in the frequency domain by monitoring the fluorescence phase shift and demodulation at various frequencies. Alternatively, lasers can * Reprint requests: e-mail, [email protected]; FAX, (812) 855-0958. (1) Demas, J. N. Excited State Lifetime Measurements; Academic Press: New York, 1983. (2) Lakowicz, J. R. Principles of Fluorescence Spectroscopy; Plenum Press: New York, 1983; Chapter 3. (3) O’Connor, V. D.; Phillips, D. Time-correlated Single Photon Counting; Academic Press: Orlando, FL, 1984; Chapter 3. (4) McGown, L. B.; Bright, F. V. Anal. Chem. 1984, 56, 1400A-1415A. (5) Mattheis, J. R.; Mitchell, G. W.; Spencer, R. D. In New Directions in Molecular Luminescence; Eastwood, D., Ed.; ASTM Special Technical Publication 822; ASTM: Philadelphia, PA, 1983; pp 50-64.

1936 Analytical Chemistry, Vol. 69, No. 10, May 15, 1997

be employed in a repetitively pulsed or continuous-wave mode to introduce multiple modulation frequencies in either a simultaneous or a quasi-simultaneous fashion. With this last approach, frequency domain signal processing has become popular in order to capitalize upon the broad frequency bandwidth of these sources.6-10 For all existing light sources and measurement techniques, however, the physical size, cost, and complexity of the instrumentation prohibit time-resolved fluorescence from being more commonly employed in chemical problem solving. Since the early reports of correlation fluorometry,11 crosscorrelation methods have been simplified considerably. These techniques can eliminate much of the cost and complexity incurred by the more common time and frequency domain signalprocessing electronics.12 Cross correlation produces information about the coherence (or similarity) between the frequency components common to both the modulated excitation and fluorescence wave forms.13 In the time domain, cross correlation is expressed as

1 Tf∞ 2T

cxy(τ) ) lim

∫

T

x(t)y(t ( τ) dt

-T

where cxy(τ) is the cross-correlation function between the timevarying fluorescence signal x(t) and the time-varying excitation wave form y(t), and τ is their relative displacement in time. The time domain statement accurately reflects the manner in which cross correlation is implemented instrumentally. Multiplication of the wave forms is accomplished by an inexpensive radio frequency or microwave mixer, common in radio and microwave electronic equipment. Integration of the product is performed by a simple low-pass filter. Such basic and ubiquitous components allow the correlation processor to be constructed for just a few hundred dollars. Additionally, cross correlation lends itself nicely to fluorescence lifetime measurements that employ randomly modulated excitation (6) Alcala, R. J.; Gratton, E.; Jameson, D. M. Anal. Instrum. 1986, 14 (3&4), 225-250. (7) Bright, F. V.; Hieftje, G. M. Appl. Opt. 1987, 26, 3526-3529. (8) Hieftje, G. M.; Haugen, G. R.; Ramsey, J. M. Appl. Phys. Lett. 1977, 30, 463-466. (9) Lakowicz, J. R.; Laczko, G.; Gryczynski, I. Rev. Sci. Instrum. 1986, 57, 2499. (10) Mitchel, G.; Swift, K. In Time-Resolved Laser Spectroscopy in Biochemistry II; Lakowicz, J. R., Ed.; SPIE-The International Society for Optical Engineering: Bellingham, WA, 1990; Vol. 1204, Part 1, pp 270-274. (11) Spencer, R. D.; Weber, G. Ann. N.Y. Acad. Sci. 1969, 158, 361-376. (12) Ramsey, J. M.; Hieftje, G. M.; Haugen, G. R. Appl. Opt. 1979, 18, 1913. (13) Horlick, G.; Hieftje, G. M. Contemp. Top. Anal. Clin. Chem. 1978, 3, 153216. S0003-2700(96)01130-4 CCC: $14.00

© 1997 American Chemical Society

sources.14 Random modulation produces an excitation wave form that has the appearance of noise. That is, the intensity fluctuations occur at random time intervals and with random amplitude. Such excitation modulation is intrinsically produced by a radionuclidescintillation light source due to the inherently stochastic nature of the radioactive disintegration process. A host of practical advantages accompany the use of such a light source. In comparison to the traditional means of excitation, the new source is extremely inexpensive. The source for this work was constructed for just over $500. Also, unlike bulky flash lamps, arc lamps, and lasers, the radionuclide-powered scintillator can be made extremely compact. In addition, the light source requires no external power and needs very infrequent maintenance. Thus, many of the conventional light source shortcomings are eliminated. Such a light source was first investigated by Ross to increase the precision of absorption photometry measurements.15 Using γ-emitting 65Zn and a liquid scintillator, Stadelmann utilized TCSPC to measure the excited-state lifetime of fluorescent dye compounds.16 Later, Hobbs and Hieftje demonstrated a compact, more intense, β-powered light source for lifetime measurements which also employed TCSPC.17 However, TCSPC has several drawbacks. The electronics are costly and bulky, and the attainable signalto-noise ratio is limited by the random nature of the light source. In addition, TCSPC inherently wastes signal information since it requires that no more (and usually far less) than a single fluorescence photon be detected for each pulse of the excitation source. The present work employs electronic cross correlation for lifetime determinations by means of the radionuclide scintillator. The combination of low cost and compact size makes this approach particularly appealing. Furthermore, the elegant manner in which cross correlation takes complete advantage of the random modulation provides the unusual satisfaction of working with, rather than against, a noisy signal. EXPERIMENTAL SECTION The radioactive isotope employed was β-emitting 90Sr (28.3-yr half-life). The daughter product, 90Y, is also a β emitter and decays rapidly (3-h half-life). In equilibrium, the two isotopes produce a continuous distribution of β energies up to 2.3 MeV with an average of 1 MeV. All lifetime measurements were taken with a 1-mCi (37.7 Mdps) capsule purchased from Isotope Products (Burbank, CA). Pilot U, a poly(vinyltoluene) plastic scintillator that has an emission maximum at 391 nm, was purchased from NE Technologies (Monmouth Junction, NJ). When used with 90Sr, this scintillator is specified to produce an average of 10 000 photons/MeV.18 The instrument diagram is shown in Figure 1. The plastic scintillation material was machined into a rectangular bar of dimensions 1 cm × 1 cm × 5 cm. Following mild heating, the bar was bent to nearly 90°. After being shaped, the material was polished until optically transparent and chemically plated with silver to direct the isotropically emitted scintillation photons to either of two PMTs. The bent scintillator was then placed within a lead block, which also contained the sample cell. The 90Sr (14) Dorsey, C. C.; Pelletier, M. J.; Harris, J. M. Rev. Sci. Instrum. 1979, 50, 333-336. (15) Ross, H. H. Anal. Chem. 1966, 38, 414-420. (16) Stadelmann, H. R. J. Lumin. 1970, 3, 143-151. (17) Hobbs, S. E.; Hieftje, G. M. Appl. Spectrosc. 1995, 49, 15-19. (18) Clark, D. Nucl. Instrum. Methods 1974, 117, 295-303.

Figure 1. Analog cross-correlation instrument diagram. Scintillation signals are amplified and sent to the local oscillator of the radiofrequency (RF) mixer. Fluorescence signals are sent to the RF port of the same mixer. The output taken from the DVM (digital voltmeter) as a function of delay is the cross-correlation function.

capsule was placed on top of the scintillator, halfway from each end. This arrangement protected the PMT photocathodes from Bremsstrahlung X-rays created by β-particle deceleration in the scintillation material. The high-energy X-rays penetrate the plastic but are stopped by the surrounding lead. The mirrored, inner surface of the scintillator bar directed the scintillation light to a silver-plated glass sample cell at one end and around a curve through an optical low-pass filter to a PMT at the other. Fluorescence from the sample cell was observed with a second PMT through an optical high-pass filter at 90°. Both PMTs were purchased from Hamamatsu (R2693) and were chosen for their fast response time (1.2-ns rise time) and relatively large photocathode area. Each PMT was powered at -900 V with eight of the nine available amplification stages equivalently biased using 50-kΩ resistors. The last amplification stage was not used in order to keep the average anode current below the 100-µA maximum rating and to improve the overall time response of the instrument. The signal from the scintillation PMT was delayed by a combination of a line stretcher (air-dielectric coaxial transmission line, GenRad Model 874-LTL) and fixed lengths of RG58U cable. The line stretcher was continuously variable over a 1.5-ns range. When used in conjunction with discrete cable segments of 1.5-ns increments, a continuous range of delay values was accessible. The delayed signals were amplified by a 20-dB, 0.1-1000-MHz amplifier (Mini-Circuits, ZFL-1000LN) and sent to the local oscillator (L) port of a 7-dBm double-balanced radio-frequency mixer (Mini-Circuits, ZFM-2). The fluorescence signal was sent to the reference-frequency (R) port of the same mixer. Both the L and R inputs of the mixer were responsive from 1 to 1000 MHz. Thus, information below 1 MHz was lost. The output of the intermediate-frequency port (I) provided a signal proportional to the product of the two inputs over a frequency range of 0-1000 MHz. The output from the mixer was sent to a low-pass electronic filter (Krohn-Hite, Model 3342) set to a cutoff frequency of 1 Hz to perform the integration. The resulting dc signal was amplified by 40 dB and read by a digital voltmeter. The delay-producing cable segments were inserted in the scintillation channel. The exact temporal offset between the Analytical Chemistry, Vol. 69, No. 10, May 15, 1997

1937

scintillation and fluorescence channels was determined by disconnecting the radio-frequency mixer inputs and measuring the difference in pulse arrival times with a two-channel digital oscilloscope (Tektron, TDS520) that is capable of 40-ps resolution. Jitter in the PMT transit time degraded the attainable delay resolution to (50 ps. The instrument response was measured without a high-pass optical filter in front of the fluorescence PMT and by using a solvent blank in the mirrored sample cell. Scintillation light was directed to the sample cell, where it was subsequently reflected to the photocathode of the fluorescence PMT. The reflective coating on the scintillator and the sample cell served to contain and direct scintillation rays of high-incidence angles as well as increase the excitation path length inside the sample cell. Pilot U is specified to have a pulse width of 1.2 ns and a refractive index of 1.58. Although the volume of the coated scintillator (1 cm × 1 cm × 5 cm) and cylindrical sample cell (1-cm diameter × 2 cm) increased the optical path of the reflected scintillation photons, the distance was not large enough to cause any measurable excitation pulse broadening with respect to the response time of the detectors (3-ns fwhm). The measured correlation function is a convolution of the sample impulse response (the fluorescence decay curve) with the autocorrelation function of the scintillation flashes (the instrument response). A convolute-and-compare algorithm19 was used to extract the exponential luminescence decay of the sample from the cross-correlation data. That is, a single exponential was assumed to model the fluorescence decay and was iteratively convolved with the autocorrelation function of the instrument response for comparison to the measured correlation function. The exponential decay producing a minimum difference between the measured and calculated curves gave the value of the excitedstate lifetime of the sample. The routine was implemented in LabVIEW 3.0 and with a Macintosh IIfx computer (68030 processor operated at 40 MHz). LabVIEW 3.0 is convenient to use because of its built-in convolution object (subvirtual instrument), which simplifies the programming task. The sum of the squared residuals was minimized by a Golden Section Search algorithm. This one-dimensional algorithm determines a function minimum with the fewest possible iterations.20 In addition, this search algorithm is concise and easier to implement than the more sophisticated and more commonly used simplex optimization routine. The processing times that resulted were 1-2 min, depending on the number of data points collected. To demonstrate the viability of the new technique, fluorescence lifetimes of several molecules were determined. Coumarin 460, 2,5-bis(5′-tert-butylbenzoxazolyl-2′)thiophene (BBOT), and decacyclene were obtained from Exciton, Sigma, and Aldrich, respectively. The overlap of the analyte excitation and scintillator emission bands, the analyte emission wavelengths, and the cutoff wavelength of the optical high-pass filter were different for each lifetime measurement. The specific conditions for each experiment can be found in the corresponding figure captions. All solutions were degassed with nitrogen for 20 min prior to use, and lifetime measurements were made in triplicate to determine (19) Ware, W. R. Creation and Detection of the Excited State; Marcel Dekker: New York, 1971. (20) Press, W. H.; Flannery, B. P.; Teukolshy, S. A.; Vetterling, W. T. Numerical Recipes: The Art of Scientific Computing; Cambridge University Press: New York, 1986; pp 277-282.

1938

Analytical Chemistry, Vol. 69, No. 10, May 15, 1997

Figure 2. Lifetime measurement of 1 mM coumarin 460 in ethanol: excitation maximum, 373 nm; emission maximum, 445 nm; highpass filter cutoff, 435 nm. (-•-) Interpolated instrument response; (b) measured cross correlation; (- -) calculated lifetime, 3.4 ( 0.1 ns. Literature lifetime, 3.3 ns.21

precision. Due to the low intensity of the scintillation source (∼200 nW average power), solutions of relatively high concentration were employed. Pulse height analysis of the scintillation flashes was performed by adding a 30-pf capacitor in parallel to each of the last three amplification stages of the PMT dynode chain. In combination with the voltage divider current of 2 mA, the capacitors serve to maintain a constant dynode potential under large instantaneous currents and preserve amplification linearity. Output from the PMT was sent to a linear current preamplifier (Stanford Research Systems, SR570) set to an upper bandwidth of 1 MHz. The preamplifier was used to shape the nanosecond anode pulses for binning via a PC plug-in multichannel analyzer card (EG&G ORTEC, 916A). Due to the microsecond time constant of the preamp/MCA combination, a 0.1-µCi 90Sr source (Central Scientific) was utilized. This activity provided a low disintegration rate, which avoided the registration of erroneously large pulse sizes due to multiple radionuclide decays occurring within the preamp/ MCA time constant. RESULTS AND DISCUSSION In order for a convolute-and-compare calculation to function properly, data from equally spaced delay intervals must be used. However, it was extremely time consuming to adjust the line stretcher to obtain equivalent delay intervals. Instead, data were taken at delay times without concern for exact temporal spacing. An example of the measured values is illustrated by the dots in Figure 2. A LabView 3.0 cubic spline interpolation routine was used to artificially “enhance” the data resolution and to provide data points spaced at equivalent intervals for the convolute-andcompare routine. After data interpolation, the baselines of the instrument response and cross-correlation data were set to zero. Postcollection baseline subtraction was necessary because the instrumentation was not capable of electronic zero-offset adjustment. The proper baseline correction was determined by measuring the cross-correlated signal at a delay that was long enough to ensure that both the cross-correlation and the instrument response curves had returned to their baseline values. Because of the extreme sensitivity of the convolute-and-compare calculation to the zero

Figure 3. Lifetime measurement of 250 µM BBOT in ethanol: excitation maximum, 372 nm; emission maximum, 434 nm; high-pass filter cutoff, 455 nm. (s) Interpolated instrument response; (- - -) interpolated cross correlation; (- -) calculated lifetime, 1.5 ( 0.1 ns. Literature lifetime, 1.51 ns.22

offset, a delay more than 7 times longer than the lifetime of the analyte was employed. Following baseline correction, the crosscorrelation data set was normalized to the signal maximum of the instrument response and sent to the convolute-and-compare routine. Figure 2 shows a cross-correlation lifetime determination of coumarin 460. The instrument response, measured cross correlation, and the result of the minimized convolute-and-compare calculation are all displayed. The collected data set of 23 points (dots) and the corresponding interpolation values (lines) are shown for each curve. Both the instrument response and the cross-correlation measurements were interpolated with 500 points, resulting in a data spacing of 0.05 ns. Further resolution enhancement slightly improved the agreement of the lifetime calculation to that found in the literature but increased the processing time and required excessive computer memory. Interpolation with 500 points was chosen to optimize the tradeoff between accuracy and computational convenience. Baseline correction measurements were made at a 25-ns delay. The difference between the data at negative delays and the calculated function is most likely caused by electrical reflections which occur in the particular cable segments used to produce the points. Noise at negative delays is also visible in the instrument response curve. Because the peak value of the correlation function is considerably less than the peak value of the instrument response, the relative noise in the correlation function is amplified when the two curves are normalized. As can be seen, however, the noisy points do not appear to significantly alter the accuracy of the fit. The resulting lifetime of 3.4 ( 0.1 ns is in close agreement with other measurements in the literature.21 Figure 3 illustrates the time resolution of the instrument for the excited-state lifetime measurement of BBOT. A set of 30 data points was collected for each curve, but for simplicity, only the interpolated values are shown. The zero offset value was determined at 30 ns. As in the coumarin 460 measurement, 500 points were used for both the instrument-response and crosscorrelation interpolations. This resulted in ∼0.06-ns data point spacing before applying the deconvolution algorithm. The cal(21) Yaowu, H.; Changjiang, M.; Dafan, Z.; Wang, W. Yingyong Huaxue 1987, 4, 53-57.

culated value of 1.5 ( 0.1 ns is in agreement with the literature,22 even though this lifetime is well below the fwhm of the instrument response. The fundamental limitation of short-lifetime measurements is dictated by reproducibility, regardless of the speed of the detectors or the excitation pulse. Ideally, a perfectly stable instrument response could be used to measure a fluorescence decay time that approached the duration of a delta function. The lowfrequency stability of radionuclide-powered light sources23 enhances the ability of the instrument to measure fast decay times. Unlike conventional light sources, which are subject to lowfrequency intensity variation caused by power fluctuations or electrode erosion, the high activity (27-ns interpulse spacing) and long half-life (28.3 yr) of the 90Sr ensure that the time-integrated (∼1 s) light intensity is essentially constant. As a result, both the instrument response and the correlation function are steady and can be deconvoluted with good precision and accuracy, allowing very fast decay times to be measured. In this regard, cross correlation takes special advantage of the nature of the light source. Although fluorescence lifetime extraction is dependent upon the inherent high-frequency noise of the light source, the correlation technique allows the decay information to be gathered at low frequencies where the light source noise is virtually insignificant. The imprecision that does exist in the measurement is a consequence of low-frequency drift (e1 Hz) in the signal-processing electronics, not of slow variation in the light source intensity. As is evidenced in Figure 3, the ease with which the BBOT cross-correlation function can be distinguished from the instrument response indicates that the electrically introduced noise in the signal is small. The clear difference between these two curves also indicates that the shortest measurable excited-state duration for this version of the instrument is well below 1 ns. The use of scintillators with shorter decay times, detectors with faster response times, or improved stabilization of the signal-processing electronics could lower this limit even further. The measurement of long fluorescence lifetimes required additional steps in order to produce accurate results. Because the scintillation signal in this experiment was delayed with segments of RG-58U cable, determination of long lifetimes necessitated the use of lengthy segments. In the case of the decacyclene measurement shown in Figure 5 (28-ns lifetime), segments in excess of 75 ft (23 m) were needed to produce crosscorrelation values to 145 ns. However, even this delay does not meet the seven-time-constant rule of thumb found to be necessary for baseline offset measurement in coumarin 460 and BBOT; the proper cable length is in excess of 100 ft (30 m). Because such a delay cable would have been difficult to manage, the correct baseline offset was determined by software convolution of the instrument response with the proper 28-ns exponential decay.24 The decacyclene data were then adjusted to the offset value predicted by the convolution at a delay of 130 ns. Although this method is practically unattractive because it requires a priori knowledge of the fluorescence lifetime, it would not have been necessary if a sufficiently long, but unwieldy, cable had been employed to measure the zero offset. (22) Renliang, X.; Mitchell, W. A. J. Photochem. Photobiol. 1991, 57, 351-360. (23) Jones, K.; Malcome-Lawes, D. J. J. Chromatogr. 1988, 441, 387-393. (24) Berlman, I. B. Handbook of Fluorescence Spectra of Aromatic Molecules, 2nd ed.; Academic Press: New York, 1971; pp 397.

Analytical Chemistry, Vol. 69, No. 10, May 15, 1997

1939

Figure 4. Pulse height distributions in delay cables of various lengths indicating substantial power loss to the cable: (- - -) 6.5, (- -) 41, and (-) 123 ns. Integration time, 60 s.

Another complication for long-lived excited-state measurement was that the calculated lifetime was significantly shorter than the accepted value found in the literature. Such a difference would usually suggest the existence of fluorescence quenching; however, systematic error caused by the instrument is another possible cause. To investigate the source of this discrepancy, pulse height analysis of the electrical signals traversing different cable lengths was performed. The results revealed significant pulse attenuation for long cables (see Figure 4). That is, for a given distribution of pulse sizes (amplitudes), short delay cables (6.5 ns) delivered an unskewed distribution to the mixer. However, transmission through long cables skewed the distribution toward low pulse heights. This shaping is indicative of electrical power loss to the cable during pulse transmission. As a result, the mixer output at long delays was artificially low and the calculated lifetime was erroneously short. This problem is completely correctable, however, simply by compensating for the cable attenuation following data collection. A calibration curve was generated by plotting the average pulse height vs cable length. Normalization of these data to the average pulse height of the shortest cable transformed the data into attenuation correction factors and, when plotted vs cable length, resulted in a nearly linear (r ) 0.998) working curve. Multiplication of the cross-correlated data points by the appropriate correction factor yielded the proper decay curve and lifetime calculation. Thus, signal lost to the delay cables, and not fluorescence quenching, is the only major cause of the discrepancy. Figure 5 illustrates a decacyclene cross-correlation measurement after application of the power loss correction factors. A set of 50 points was collected for each curve. Interpolation using 500 data points gave an accurate result of 28.1 ( 0.1 ns. Detection limits were not determined with this instrument because of poor photon collection efficiency. The number of photons detected per scintillation can be estimated by comparing the pulse height spectrum of the scintillation flashes to that of the pulse height spectrum produced by single photons. Figure 6 illustrates the number of photons per pulse that were detected in the scintillation channel. Figure 6a shows the average pulse height of single photoelectrons generated by the scintillation PMT. The data were collected via exposure of the PMT to ambient room light through an aperture to give an average count rate of 1400 counts/s. Since the waiting time between events of a random process (photon arrival the PMT) is exponentially distributed, a count rate of 1400 counts/s ensures that the probability of 1940 Analytical Chemistry, Vol. 69, No. 10, May 15, 1997

Figure 5. Lifetime measurement of decacyclene after correction for power loss to the delay cables: (-) interpolated instrument response; (- -) interpolated cross correlation; (- - -) calculated lifetime, 28.1 ( 0.1 ns. Literature lifetime, 28 ns.24

Figure 6. Pulse height analysis of output from the scintillation PMT: (a) pulse height distribution of single photons; (b) pulse height distribution of scintillation flashes. The integration time was 3600 s. By assuming a 20% photocathode quantum efficiency, an average of just 15 photons/pulse is calculated to strike the photocathode.

detecting multiple photon events is low. Using a current amplifier time constant of 1 µs, the probability of detecting single-photon events is greater than 99.9%. In addition, this count rate was chosen to match the count rate of the scintillator bar activated by the 0.1-µCi source. Thus, the probability of detecting temporally overlapped events in both experiments is nearly equal, and a legitimate comparison of the two pulse height spectra can be made. The pulse distribution generated by the scintillator bar activated by the 0.1-µCi source (Figure 6b), gave a count rate of

1450 counts/s; however, the average pulse amplitude is only 2.6 times larger. By assuming a 20% photocathode quantum efficiency for the R2693 PMT25 over the emission wavelengths of Pilot U (380-470 nm), 50% of the pulses striking the PMT are calculated to consist of 15 photons or less, while the remaining 50% range in size from 16 to slightly over 200 photons. These figures are far below what is expected based upon the manufacturer’s specifications for the scintillator output (average of 10 000 photons/pulse with 90Sr). The poor collection efficiency suggested by the above calculation is most likely caused by a combination of several factors. A small amount of silver oxide, co-deposited on the scintillator during the silver-plating procedure, could attenuate light incident upon the surface of the scintillator. Additionally, the distance between the end of the scintillator bar and the scintillation-detecting PMT was ∼1 cm, reducing the solid angle of collection. Internal and external reflections at each optical interface (air/scintillator, air/ filter, air/PMT) also decrease the light transmission. Since the scintillation rays are not collimated, reflection losses greater than the nominal 4%/interface are expected. The same attenuation mechanisms occur in the fluorescence channel, and to a greater extent. Due to geometrical constraints, ∼2 cm separates the fluorescence sample cell from the fluorescence PMT. Given the poor collection efficiency in both channels, a detection limit measured with the current instrument would not be meaningful. Such optical losses are usually reduced by bringing the components into direct contact with each other and by coupling each interface with a refractive index-matching solution or gel. This practice is common in the field of nuclear physics where the number of optical components is few and in which the scintillator/PMT geometry is relatively straightforward. Preliminary measurements made with 90Sr and Pilot U optically coupled to a PMT photocathode indicate that an average of over 5000 photons/pulse is detectable. Revision of the present instrument design to allow tight geometric arrangement and optical coupling (25) Photomultiplier Tube Products Catalog; Hamamatsu, Shizuoka, Japan, 1994; pp 80.

of adjacent components should reduce light losses and make detection limit measurements more indicative of the true signal generation capability of the light source. These measurements will be the subject of a future communication. CONCLUSIONS Manual interchange of the delay cables, while acceptable for the present demonstration, is labor intensive, requiring ∼20 min for the collection of a single curve. Consequently, future improvements will include automation of the delay and signal acquisition functions to eliminate the time required for manual interchange of the cables. The fluorescence lifetime measurements demonstrated herein were made using between 23 (coumarin 460) and 50 data points (decacyclene). With a 1-Hz electronic filter (1-s integration time), computer-controlled data acquisition, and a multiplexed analog delay network, the cross-correlation curve for coumarin 460 could be acquired in ∼23 s, while roughly 50 s would be required for decacyclene. This modification would significantly simplify the data collection procedure, making more routine lifetime determinations practical. Lastly, exposure to the relatively active radiation source is easily avoided. Because the 90Sr source is tightly encapsulated, there is no risk of radioisotope escape and consequent contamination. Adequate shielding from the high-energy β particles can be provided by just a few centimeters of Lucite or by several millimeters of lead. Moreover, manipulation of the capsule can be done safely as long as the handling time is minimized and the pellet is held ∼1 ft away from the experimenter’s body and appendages. ACKNOWLEDGMENT This research was supported by the National Institutes of Health, Grant GM 53560. Received for review November 7, 1996. February 3, 1997.X

Accepted

AC961130M X

Abstract published in Advance ACS Abstracts, April 1, 1997.

Analytical Chemistry, Vol. 69, No. 10, May 15, 1997

1941