Measurement of Luminescence Decays: Methods and Instrumentation

From Lumipedia
Revision as of 19:10, 9 September 2017 by Admin (talk | contribs)

Jump to: navigation, search


by Dr. Mark Sulkes [[1]]

Department of Chemistry, Tulane University, New Orleans, Louisiana 70118, USA.


Developments in electronics, PMT technology, and excitation light sources have contributed to improvements in luminescence decay detection. The result in recent years has been more precise determination of lifetimes, particularly in the sub-ns regime, at often drastically reduced costs. The most important factor in cost reduction, also affording some enhanced capabilities, has been the development of numerous LED and laser diodes, spanning a wide wavelength range, that can provide sub-ns pulses and can be driven even to GHz frequencies. The cost of these light sources is typically a small fraction of conventional laser systems. The result in the lowest cost regime (waveform digitizer combined with LED/laser diode excitation sources) is a rather good and potentially quite inexpensive system for determining strong emission luminescence lifetimes from multi-ns to µs. This review is particularly concerned with time and frequency domain methods that are capable of more precise lifetimes in the ns to the sub-ns regime. These capabilities have been available for several decades at fairly significant costs. The advent of LED/laser diode excitation sources has greatly lowered the cost of these capabilities and even afforded some performance enhancements, particularly in the frequency domain. Powerful commercial instrumentation is available at much-reduced cost. User implemented or improved, systems at a further reduced cost are increasingly feasible.


1 Introduction

The measurement of luminescence decays as a function of time provides fundamental information for an enormous range of applications and systems in chemistry and biology.[1-3] Continuing technical developments have made possible precise luminescence decay measurements, even on sub-ns timescales, at increasingly lower cost. Partly, this is due to the availability of faster electronics at a decreasing cost. However, the single most important factor has been the development of LEDs and laser diodes that can be driven with narrow pulses in time and at high frequencies - often replacing expensive laser systems. The result has been powerful commercial systems at lower cost. There are also growing opportunities for surplus, homemade, and home modified components with excellent performance at a far lower cost.[4] The other limiting factor is detector - photomultiplier (PMT) - performance. Ongoing PMT developments have brought about improved performance at lower prices. In addition, past technical publications have shown how to get greatly enhanced performance from common and inexpensive models.

This review of instrumentation developments will mainly consider the time domain method of time correlated single photon counting (TCSPC)[1,2] and the frequency domain method of phase modulation fluorometry[1,2]; some other methods and equipment will be mentioned more briefly.


2 History

Apparently, the first luminescence lifetime measurements were reported in 1926, using phase fluorometry—ns timescale decays obtained by measuring phase delays (albeit lifetime values with somewhat limited precision).[5] Continuing early work employed phase modulation fluorometry. Lakowicz recounts some early developments [1]; Klein records various experimental benchmarks in the method from the 1920s to 1980s.[6]

Time domain measurements advanced with the development of instrumentation and methods for recording luminescence as a function of time. Here are some benchmarks:

  • first photomultiplier tubes (the 1930s)[7]
  • first oscilloscopes (the late 1940s)[8]
  • Polaroid cameras (the late 1940s) → oscilloscope photographs of decays
  • time gated boxcars for recording transients (beginning late 1950s to 1960s)[9]
  • sampling/signal averaging methods (beginning in early1960s- the early 1970s)[10]
  • first TCSPC (1971)[11]
  • µs timescale digitizer recording of decay waveforms (beginning in the late 1970s)
  • ns timescale: optical recording of waveforms (Tektronix 7912, the late 1970s)[12]
  • ns timescale digitizer recording of decay waveforms (beginning in the mid-1980s)
  • streak cameras for ns and ps decays (beginning in the early 1980s)[13]
  • laser based ultrafast upconversion detection, to fs (beginning in late 1980s) [1,14,15]

At present, commercial instrumentation is available for both time domain and frequency domain luminescence decay measurements. Although they have different operating principles, both kinds of instruments have similar performance quality criteria for light excitation sources and light detection.


3 Time Domain Methods

3.1 Methods other than TCSPC

Following excitation, luminescence intensity is recorded as a function of time. All of the methods in the benchmark list except for streak cameras normally employ PMT detectors. For the recording and averaging of analog signals, triggered waveform digitizers have essentially supplanted other methods. Given the current relatively low cost of a >300 MHz digitizer as new equipment and the extremely low cost in the used equipment market, the advent of pulsed LED/laser diode excitation sources makes this is a rather low cost method for obtaining an overall system response of several ns or less;[4] PMT transit time spread (TTS) generally is the limiting factor.

Streak cameras offer excellent capabilities but at high prices. Following the production of photoelectrons at a cathode, a streak camera then deflects the photoelections via a transverse electric field that increases in strength with time; the transversely deflected photoelectrons are amplified by microchannel plates before impacting on a phosphor.[13,16] The 2D images are then optically recorded. The use of CCD cameras allows for detection at a photon counting limit.1[16] Such a system is very powerful: It allows for acquisition of 2D (wavelength, time) data, with high sensitivity, down to ps. Ultimate resolution would be obtained with light excitation widths of a few ps to sub-ps, requiring expensive laser systems. Sub-ps resolution has been reported with a photon counting streak camera.[17] The limiting factor, aside from the cost of any such laser sources, is the cost of the detection system - well over 10[math]^{5}[/math] $US.

Laser based fluorescence upconversion detection has a time resolution limited by the laser pulses, potentially down to fs.[1,14,15] However, these methods are quite expensive and require specialized expertise.

3.2 Method of TCSPC

The potential advantages of TCSPC include high sensitivity, the absence of analog noise, and instrumental response functions that can be as low as several tens of ps. There are numerous discussions on the details of implementation, with applications.[18-22] The cost of a system with an instrumental response function of <100 ps has dropped drastically.

The method in its simplest (but not optimally configured) form can be understood as a “stopwatch” method: The appearance(detection) of the excitation pulse turns on the “stopwatch.” (This “start” pulse could come from detection of a beam split laser pulse by a fast photodiode. A sync signal from the driver circuit of an LED or laser diode could also generate a “start” pulse.) The excitation pulse populates the emitting state, causing luminescence photons to be emitted. The pulse from a single detected luminescence photon generates a “stop” pulse that turns off the “stopwatch”; the elapsed time on the “stopwatch” is recorded. Over many such recorded events, a histogram is built up corresponding to photon emission as a function of time. Observation of the luminescence rise time (or measurement with a scatterer) establishes the “t = 0” time. To prevent early photon emissions from having disproportionate weight, the number of detected photons per pulse (i.e. monitored photon count rate) must be limited to one per ~100 excitation pulses. To circumvent the problem of many “null stopwatch turn-ons,” the photon count “stop” pulse can be time delayed so that it becomes the “start” pulse; this is the normal mode of operation, whereby the histogram is recorded in “flipped” time order. Fast amplifiers and constant fraction discriminators are used to generate “start” or “stop” signals emanating from PMTs or photodiodes. Traditionally, a discrete time-to-amplifier converter module (TAC nuclear instrumentation module, NIM) was used as the “stopwatch.” TAC pulses within narrow amplitude bins corresponding to times aretabulated by a multichannel analyzer operating in pulse height analysis mode. Since the 2000s, integrated PC boards or standalone electronics have become available that carry out all of these functions. The dead time between countable detected photons is ~80 ns for the best currently available TCSPC electronics,[21] somewhat more for older TCSPC electronics. Boards are available that carry out various forms of multidimensional TCSPC based FLIM and other multidimensional methods (e.g. simultaneous multi-wavelength detection using a multi-anode PMT).[19,21,23] The growing “obsolescence” of the required NIM electronics modules (TACs, quad constant fraction discriminators) and multichannel analyzers has resulted in their increasingly low acquisition costs on internet used equipment venues—at least while they remain reasonably available.

The rate of time-tagged photon counts is limited to ~10[math]^{-2}[/math] of the pulsed excitation repetition rate, making a MHz excitation rate highly desirable. For example, a 10[math]^{6}[/math] Hz rate would enable photon counting at ~10[math]^{4}[/math]Hz; the photon counting histogram will build up within a reasonable time. Formerly, such excitation rates were available only with expensive laser systems. Historically, a high-end TCSPC system often would employ a mode locked Ar[math]^{+}[/math] laser that synchronously pumped a dye laser, with ~5 ps output pulses cavity dumped at rates as high as 4 MHz. Alternatively and more recently, an ~800 nm “pulse picked” 80 MHz Ti: sapphire oscillator could be employed, with second or third harmonic generation (near UV or UV) of the picked pulses. In this period more modestly priced TCSPC systems also were available that employed flash lamps with relatively short pulses (~ 1 ns) at ~10[math]^{4}[/math] Hz rates. For lower pulse repetition rates, data acquisition times must be significantly longer to acquire decay histograms with satisfactory photon counting statistics.

At the same time, the need to limit the photon counting rate to ~10[math]^{-2}[/math] of the pulse excitation rate actually can offer some advantages, at least for sensitive detection. Under normal circumstances, neutral density filters may be needed to attenuate the detected signal. Given the excess signal, even relatively weak excitation pulses, such as ~60 ps pulses from an LED, would present no hindrance for TCSPC. Since there is no analog noise and photon counting statistics prevail, with sufficient time even very weak emission decays can be acquired with good precision (but beware the possibility of accompanying weak background emission). With higher photon flux pulses and high pulse repetition rates, the sensitivity can be quite extraordinary. Phase fluorometers are not as well suited for the measurement of extremely weak luminescence.

TCSPC is not well suited for the measurement of long lifetimes. For example, measurement of a 1 µs decay would require a spacing in time of excitation pulses of ~5 µs, corresponding to a ~10[math]^{5}[/math] Hz repetition rate. For these situations, multiscalar electronics modules are available that carry out photon counting within selectable time interval bins that can be as short as 65 ns.[24]

4 Frequency Domain Methods

If a luminescing chromophore is excited by an intensity modulated light source at a given sinusoidal angular frequency [math]\omega \, (=2\pi f)[/math] the emission will also be at angular frequency [math]\omega[/math]. However, the finite lifetime of the emitting state imposes a time delay on the driven emission; this time delay can be expressed as a phase shift [math]\varphi _{\omega }[/math]. As the driving frequency [math]\omega[/math] increases, the phase shift increases. If it is the case that the decay is single exponential, the single exponential lifetime [math]\tau[/math] is related to the phase shift by [math]\tan \varphi _{\omega }=\omega \tau[/math].[1] There is also a relationship between the driving frequency and the observed amplitude excursions of the luminescence. Suppose that the excitation intensity has ratio of peak-to-peak amplitude relative to mean intensity of b/a, where a is the time averaged intensity value and b is the maximum sinusoidal amplitude excursion from the mean value. The corresponding value for the driven emission, B/A, will be smaller. As the driving frequency ω increases, the B/A ratio will decrease; the amplitude excursions gradually wash out - washed out by the long timescale of the emission relative to the driving excursions. Thus the modulation [math]m_{\omega }=\frac{B/A}{b/a}[/math] decreases as [math]\omega[/math] increases. For a single exponential decay, [math]m_{\omega }=(1+\omega ^{2}\tau ^{2})^{1/2}[/math].[1] Phase shift and modulation detection typically is done by a cross correlation method, wherein the PMT signal is modulated by a small frequency offset (tens of Hz) and heterodyned with the excitation signal.[1]

At low driving frequencies, changes in phase shift and modulation values are relatively small and subtle (modulation → 1, phase shift → 0); at very high frequencies the values again depend less sensitively on ω (modulation → 0, phase shift → π/2). An optimum single measurement frequency, where phase and modulation vary most strongly with ω, would be a ω such that ωτ ≈ 1. In other words, good determination of shorter lifetimes requires higher frequencies. For example, a good frequency for a 1 µs lifetime determination would be f ≈ 0.16 MHz. For a 100 ps lifetime, a good frequency would be f ≈ 1.6 GHz. Note that, unlike TCSPC, long as well as short lifetimes (the latter as limited by apparatus frequency response) can conveniently be measured by the same apparatus.

For a single exponential decay, measurement at a single favorable frequency would suffice, at least in principle. However, side from an experimental need to find what an optimum frequency value actually was, if measurements took place over a range of ω then both [math]\varphi_{\omega}[/math] and [math]m_{\omega}[/math] could be globally fitted, resulting in a more precisely determined τ. What if the decay really was not single exponential? In that case, fitting at a single frequency would result in misleading values for τ. Fitting over a frequency range would enable fitting to a more realistic multiexponential model.

The excitation light intensity can be modulated electro-optically (Pockels cell). Laser beams are well suited to electro-optic modulation, though it is also possible with other cw sources. Modulation can take place only to ~200 MHz.[1] Considering this lower than desired value, the capabilities of LEDs/laser diodes stand out: driven LED/laser diode modulations to several GHz now are specified for some commercial products. Lakowicz and collaborators have exploited the properties of cavity dumped 4 MHz 5 ps pulse trains to extend measurements to higher frequencies still, given that the Fourier transform has significant contributions to tens of GHz.[25] Their measurements went to 10 GHz, limited at that point by the frequency response of the detector.

The detectors for monitoring excitation light and also emitted light must be able to respond to excursions for any given frequency ω. Monitoring of excitation light to GHz ranges can be done with a fast photodiode; a more sensitive detector, a PMT, generally will be required for emitted light.

Unlike frequency domain measurements, TCSPC has the ability to produce precise data for weak, even extremely weak, emission. However, Mizuno et al.[26] have recently reported a frequency domain instrument with 1 GHz response (LED excitation), where the luminescence waveform as a function of time was obtained using TCSPC with a microchannel plate photomultiplier (MCP-PMT).


5 Instrumentation

The components described are common to time and frequency domain instruments. Similar figures of merit apply in the two cases.

5.1 Dynode Based PMTs

Conventional PMTs are dynode based, whereas photoelectron gain is effected in individual parallel 6-12 µm diameter microchannels in an MCP-PMT.[27] TTS for a single photon anode PMT pulse is a key parameter associated with performance for both TCSPC and phase fluorometry (related to frequency response in the latter case): the smaller the TTS value, the better.

The TTS for typical dynode based PMTs is generally in the ~1-2 ns range, whereas it can be as low as ~20 ps for MCP-PMTs.[28] The corresponding frequency responses are several hundred MHz for dynode PMTs and multi-GHz for MCP-PMTs.[1] However, the price of an MCP-PMT can be more than an order of magnitude greater than a dynode-based PMT. In the absence of an MCP-PMT, it would be desirable to use a dynode-based PMT with the best TTS. In the 2000s, compact PMTs began to appear with TTS values ~0.3 ns, suitable for rather good performance in TCSPC.[29] At the time of this writing, a compact PMT with transit time spread of 0.2 ns is available for around $1000 US.[30] Micro (mm size) PMTs are now available as well, but they do not have small TTS values.

Based on past published research, it also is possible to obtain a considerably improved TTS for inexpensive side-on (1P28 or R928 style) PMTs, where the unmodified TTS generally approaches~1 ns. Beck improved performance by eliminating some dynodes, with higher voltage at the first stage and last stage to anode.[31] Subsequently, it was reported that increasing the potential from the cathode to first dynode combined with precise positioning of the luminescence on the photocathode resulted in a FWHM TTS of 160 ps.[32] Ware et al. obtained a TTS of 235 ps by increasing the gain on the first two and last two dynode stages, with optimized imaging of the luminescence on the photocathode.33Canonica et al. obtained a TTS of 112 ps.[34] As in the other cases, the most effective step was increasing the first stage dynode voltage; positioning of a small focused image on the photocathode had some effect as well, as did tweaking the capacitances on late dynode stages. Perhaps the ultimate set of improvements would be afforded by a combination of reducing the number of dynodes, increasing the gain of early and late stages, and precise imaging of the luminescence on the photocathode.

5.2 MCP-PMTs

The superior TTS of MCP-PMTs obviously recommends them; the best TCSPC systems, with ps or sub-ps excitation and MCP-PMTs, can have FWHM system response times of several tens of ps. The impediment is the price. Use of MCP-PMTs in TCSPC fortunately affords a long usable lifetime forthese expensive components. After a cumulative anode current of ~10[math]^{-1}[/math]C/cm[math]^{2}[/math], a typical MCP-PMT will have undergone a reduction in gain such that it is no longer usable.[35,36] (Recent production improvements have extended this figure by more than an order of magnitude.[35, 36]). Particularly relevant for tcspc is the Hamamatsu R3809U-50 series. When operated at -3000 V with an initial anode current of 100 nA, figures supplied by Hamamatsu [37] show that the relative output declines to 50% after ~1.3∙10[math]^{3}[/math] hours; the cumulative anode charge over the usable lifetime should then be on the order of 0.5 C. Given that the typical gain is 2∙10[math]^{5}[/math] at -3000 V,[38] the usable device lifetime in photon counts should be ~10[math]^{13}[/math] counts, neglecting dark current effects.

For perspective analog applications, total expected anode current should be assessed in order to decide whether it is cost effective to use an MCP-PMT; applications requiring excessive anode currents may not be desirable. The Lakowicz group developed frequency domain lifetime instruments with performance into the GHz range, requiring MCP-PMTs.[25, 39] Reported anode currents were in the range of 10-100 nA.[39] A usable device lifetime of multiple hundreds to multiple thousands of hours might be expected under these conditions.


5.3 LEDs and Laser Diodes as Excitation Sources

Multiple commercial vendors offer some or all of the following capabilities: Laser diodes or LEDs are available with discrete output wavelengths ranging from the UV, below 250 nm, to the near IR. Attainable pulsewidths, with the proper driver circuits, also available, can be as little as ~40 ps with repetition rates to 80 MHz. Various frequency modulation waveforms are possible, to at least 2 GHz. The LED and laser diode sources have long term lifetimes; the driver circuits are operationally are turnkey devices. These capabilities can be purchased in standalone form for use in a user constructed frequency or time domain systems; fully implemented commercial systems of both types are also available.

There are some extremely inexpensive user implemented possibilities. Driver circuits have been described that, with basic electronics building capabilities, are fairly inexpensive and easy to build, that should be capable of producingpulsewidths and repetition rates fairly close to those specified for commercial drivers.[4, 40] A number of extremely inexpensive high luminosity LEDs, from the near UV to near IR, are available on internet venues;[4] wavelengths available and luminosities steadily increase while prices decrease.

5.4 Mode Locked ps and fs Laser Systems

These laser systems offer high levels of performance for time and frequency domain systems, but at a high cost. For fs oscillators, the availability of solid state pump lasers affords near turnkey operation and long term reliability. However, the need for a pulse picker further contributes to cost. Historically, the staple of high-end TCSPC systems - or a 10 GHz frequency domain system that has been described[25] - was a mode locked Ar[math]^{+}[/math] ion laser synchronously pumped cavity dumped dye laser system. These systems offered a number of desirable features: a wavelength tunable ~ 5 ps output pulses with a high peak power (easy frequency doubling), cavity dumped repetition rates adjustable up to 4 MHz, and transform limited linewidths of ~ 1 cm[math]^{-1}[/math]. The relatively narrow linewidths - not available with fs pulses - enable excitation for TCSPC study of discrete vibronic structure that is observable in gas expansion cooled samples.[41, 42] Aside from high acquisition cost and need for frequent replacement of expensive Ar[math]^{+}[/math] plasma tubes, mode locking an Ar[math]^{+}[/math] laser was somewhat touchy. Since the 2000s, passively mode locked 80 MHz ps Nd:YAG lasers have become available, with either 355 nm or 532 nm output. These solid state lasers offer turnkey operation with a high level of reliability; they can serve as the optical pump for a synchronously pumped dye laser. Although sync pumped dye lasers are no longer offered commercially, they were produced for a number of years by both Coherent and Spectra Physics. As a result, the dye lasers and associated electronics periodically appear for sale on used equipment venues.

6 Acknowledgements

Support from the Tulane University Committee on Research is gratefully acknowledged.

7 References

[1] Lakowicz, J. R. Principles of Fluorescence Spectroscopy; 3rd ed.; Springer US, New York: 2006.

[2] Bernard, V. Molecular fluorescence: principles and applications; 2001.pp. 37.

[3] Valeur, B.; Berberan-Santos, M. N. Molecular fluorescence: principles and applications 2nd ed; 2nd ed.; John Wiley & Sons: 2012.

[4] Sulkes, M.; Sulkes, Z. Measurement of luminescence decays: High performance at low cost. American Journal of Physics2011,79 (11), 1104-1111.

[5] Gaviola, E. Ein fluorometer. Apparat zur messung von fluoreszenzabklingungszeiten. Zeitschrift fur Physik1927,42 (11-12), 853-861.

[6] Klein, U. K. A. Picosecond Fluorescence Decay Studied by Phase-Fluorometry and Its Application to the Measurement of Rotational Diffusion in Liquids. Arabian Journal for Science and Eng.1984,9 (4), 327-344.

[7] Lubsandorzhiev, B. K. On the history of photomultiplier tube invention. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment2006,567 (1), 236-238.

[8] Pereira, J. M. D. The history and technology of oscilloscopes. IEEE instrumentation & measurement2006,9 (6), 27-35.

[9] Ware, D.; Mansfield, P. High Stability``BoxcarIntegrator for Fast NMR Transients in Solids. Review of Scientific Instruments1966,37 (9), 1167-1171.

[10] Studer, M. C.; Wild, U. P.; Gunthard, H. H. Apparatus to measure fluorescence decay curves. Journal of Physics E: Scientific Instruments1970,3 (11), 847.

[11] Lamola, A. A.; Ware, W. R. Creation and Detection of the Excited State; 1 ed.; Marcel Dekker: 1971.

[12] Miller, E. K. Time-domain measurements in electromagnetics; Springer Science & Business Media: 1986.

[13] Nordlund, T. M. Streak Cameras for Time-Domain Fluorescence. In Topics in Fluorescence Spectroscopy: Techniques, Lakowicz, J. R., Ed.; Springer US: Boston, MA, 1999; pp 183-260.

[14] Chosrowjan, H.; Taniguchi, S.; Tanaka, F. Ultrafast fluorescence upconversion technique and its applications to proteins. FEBS J.2015,282 (16), 3003-3015.

[15] Xu, J.; Knutson, J. R. Ultrafast fluorescence spectroscopy via upconversion: applications to biophysics. Methods Enzymol.2008,450 (Fluorescence Spectroscopy), 159-183.

[16] Komura, M.; Itoh, S. Fluorescence measurement by a streak camera in a single-photon-counting mode. Photosynthesis Research2009,101 (2), 119-133.

[17] Stokkum, I. v.; Gobets, B.; Gensch, T.; Mourik, F. v.; Hellingwerf, K. J.; Grondelle, R. v.; Kennis, J. T. M. Sub-Picosecond Spectral Evolution of Fluorescence in Photoactive Proteins Studied with a Synchroscan Streak Camera System. Photochemistry and photobiology2006,82 (2), 380-388.

[18] O'Connor, D.; Phillips, D. Time-correlated single photon counting; Academic Press: 1984.

[19] Becker, W. Advanced time-correlated single photon counting applications; Springer: 2015.

[20] Kapusta, P.; Wahl, M.; Erdmann, R. Advanced Photon Counting; Springer: 2015.

[21] Becker, W. The bh TCSPC handbook; Becker & Hickl: 2014.

[22] Becker, W. Advanced time-correlated single photon counting techniques; Springer Science & Business Media: 2005.

[23] Gratton, E. Measurements of Fluorescence Decay Time by the Frequency Domain Method. In Perspectives on Fluorescence, Springer: 2016; pp 67-80.

[24] PicoQuant GmbH, time tagging electronics.

[25] Laczko, G.; Gryczynski, I.; Gryczynski, Z.; Wiczk, W.; Malak, H.; Lakowicz, J. R. A 10-GHz frequency-domain fluorometer. Review of Scientific Instruments1990,61 (9), 2331-2337.

[26] Mizuno, T.; Nakao, S.; Mizutani, Y.; Iwata, T. Photon-counting 1.0 GHz-phase-modulation fluorometer. Review of Scientific Instruments2015,86 (4), 043110.

[27] Bulter, A. Single-photon counting detectors for the visible range between 300 and 1,000 nm. In Advanced Photon Counting, Kapusta, P., Wahl, M., Erdmann, R., Eds.; Springer: 2014; pp 23-42.

[28] See Lakowicz, reference [1], Table 4.1 for representative values for dynode based PMts and MCP-PMTs.

[29] Re, R.; Contini, D.; Caffini, M.; Cubeddu, R.; Spinelli, L.; Torricelli, A. A compact time-resolved system for near infrared spectroscopy based on wavelength space multiplexing. Rev Sci Instrum2010,81 (11), 113101.

[30] Hamamatsu, Inc. model H10720/H1072.

[31] Beck, G. Operation of a 1P28 photomultiplier with subnanosecond response time. Rev. Sci. Instrum.1976,47, 537-541.

[32] Kinoshita, S.; Kushida, T. High performance, time-correlated single photon counting apparatus using a side-on type photomultiplier. Review of Scientific Instruments1982,53 (4), 469-472.

[33] Ware, W. R.; Pratinidhi, M.; Bauer, R. K. Performance characteristics of a small side-window photomultiplier in laser single-photon fluorescence decay measurements. Review of Scientific Instruments 1983,54 (9), 1148-1156.

[34] Canonica, S.; Forrer, J. Ã.; Wild, U. P. Improved timing resolution using small side-on photomultipliers in single photon counting. Review of Scientific Instruments1985,56 (9), 1754-1758.

[35] Conneely, T. M.; Milnes, J. S.; Howorth, J. Extended lifetime MCP-PMTs: Characterisation and lifetime measurements of ALD coated microchannel plates, in a sealed photomultiplier tube. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment2013,732, 388-391.

[36] Lehmann, A.; Britting, A.; Eyrich, W.; Uhlig, F.; Dzhygadlo, R.; Gerhardt, A.; Gotzen, K.; Hohler, R.; Kalicy, G.; Kumawat, H.; et al. Improved lifetime of microchannel-plate PMTs. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment2014,766, 138-144.

[37] Data supplied by Hamamatsu, Inc.

[38] Hamamatsu, Inc. R3809U-50 data sheet.

[39] Lakowicz, J. R.; Laczko, G.; Gryczynski, I. 2-GHz frequency-domain fluorometer. Review of Scientific Instruments1986,57 (10), 2499-2506.

[40] Uhring, W.; Zint, C. V.; Bartringer, J. A low-cost high-repetition-rate picosecond laser diode pulse generator. Proceedings of SPIE2004,5452, 583-590.

[41] Teh, C. K.; Sipior, J.; Sulkes, M. Spectroscopy of tryptophan in supersonic expansions: addition of solvent molecules. The Journal of Physical Chemistry1989,93 (14), 5393-5400.

[42] Philips, L. A.; Webb, S. P.; Martinez, S. J.; Fleming, G. R.; Levy, D. H. Time-resolved spectroscopy of tryptophan conformers in a supersonic jet. journal of the American Chemical Society1988,110 (5), 1352-1355.