Becker W. Advanced Time-Correlated Single Photon Counting Techniques

Подождите немного. Документ загружается.

7.2 Optical Systems 285

filters, and interference filters also depends on the ray angle. Therefore the effec-

tive NA is changed if a filter is placed in the convergent or divergent beam before

or behind the fibre.

7.2.6 Reflections in Optical Systems

In any optical system light is reflected or scattered at the sample, the lens surfaces,

filters, monochromator slits, lens holders, and/or other passive design elements. In

an optical system for visual use, the reflections cause loss of contrast and ghost

images. In an optical system for TCSPC, the reflections cause distortions of the

effective IRF. As long as the reflections are the same for the recording of the sig-

nal and the recording of the IRF, the result is simply that the recorded curve looks

ugly. However, usually the reflections are different at the signal wavelength and at

the wavelength of the IRF recording. The variation of the effective IRF can cause

noticeable errors in the determination of fluorescence decay times or other sample

parameters.

The problem of reflections is most critical if highly reflective surfaces are in-

volved. Optical elements to be considered are interference filters, reflective ND

filters, and photomultiplier cathodes. Problems can occur by backreflection of

excitation light into the sample, and by multiple reflection in the detection path.

Reflection of excitation light can be noticeable among highly scattering sam-

ples and in cases where the excitation beam is directed into the back of the sample.

In Fig. 7.24, left, an interference filter or a metal ND filter is placed in the parallel

beam between two lenses in the detection light path. For an interference filter this

is generally correct, because it needs a collimated beam to work correctly. How-

ever, the excitation light blocked by the filter is reflected back into the beam path

and focused into the sample. The reflected excitation light results in a secondary

excitation pulse. A similar situation can occur with reflection at the slit brackets of

a monochromator (Fig. 7.24, right).

LensLens

Sample

Interference Filter

metal ND filter

Detector

LensLens

Sample

Monochromator

Slit

Excitation Excitation

Fig. 7.24 Reflections from a highly reflective surface back into the sample

The relative size of the secondary pulse depends on the filters, on the absorp-

tion and scattering properties of the sample, and on the width of the monochroma-

tor slit. The IRF of the system is therefore not necessarily constant.

Backreflection of excitation light into the sample can be reduced by placing an

absorptive filter in the beam path between the interference filter and the sample.

286 7 Practice of TCSPC Experiments

Absorptive filters do not need a collimated beam. This filter can be tilted or placed

in the divergent beam between the cuvette and the lens; reflections at the absorp-

tive filter are not then focused back into the sample.

Two typical examples of reflections at reflective surfaces in the detection light

path are shown in Fig. 7.25. The worst case is that of reflections between two

highly reflective surfaces, such as two interference filters or metal ND filters

(Fig. 7.25, left). Strong reflections can also occur between the cathode of a PMT

and an interference filter in front of the detector (Fig. 7.25, right). However, re-

flections are also noticeable between a highly reflective surface and a normal glass

surface.

LensLens

Sample

Interference Filters

metal ND filters

Detector

LensLens

Sample

Interference Filter

metal ND filter

PMT

Cathode

Fig. 7.25 Reflections between two highly reflective surfaces (left) and between a highly

reflective surface and the PMT cathode (right)

Some extremely unfortunate situations can also occur in systems with several

parallel detection paths. Often light reflected back from one detection path can

enter the other one. A typical example is the „T geometry“ often used for anisot-

ropy experiments, see Fig. 7.26.

LensLens

Sample

Monochromator

Lens Lens

Monochromator

Polariser Polariser

Fig. 7.26 Reflection from one detection path into the other in a dual detector system

Any light reflected at the polarisers is directed into the opposite detection path.

Reflection at an uncoated surface is about 5%. This 5% reaches the opposite de-

tector after a few ns and causes a step in the recorded decay function. A second

reflection can appear from the monochromator slits if the polarisers are oriented

parallel to one another. The only remedy is to tilt the polarisers or to place them in

the convergent beam between the lenses and the monochromators. Even then some

reflection from the lenses is usually detectable.

7.2 Optical Systems 287

Some cases of reflections in the IRF of a TCSPC system are shown in

Fig. 7.27. Reflections between two similar interference bandpass filters placed

2 cm from each other are shown in the upper left graph. The upper right graph

shows a reflection between the cathode of an MCP-PMT, and an interference filter

placed 3.7 cm in front of it. The reflection between an absorptive ND filter di-

rectly in front of the PMT and an interference filter 14 cm in front of it are shown

in the lower left graph. The undistorted IRF is shown in the lower right graph.

Fig. 7.27 Reflections in optical systems. Upper left to lower right: Reflection between two

similar interference filters placed 2 cm from each other, reflection between the PMT cath-

ode and an interference filter, reflection between an absorptive ND filter and an interfer-

ence filter, true IRF with all sources of reflection removed. BHL600 picosecond diode

laser and R3809U MCP PMT, time scale 500 ps/div, intensity scale logarithmic from 10 to

100,000 counts

In practice any design is a compromise between reflections, efficiency, filter

performance, and filter fluorescence. Reflections can be directed out of the detec-

tion path by tilting the critical part, or by placing it in a nonparallel part of the

beam. It must be decided whether the resulting change in the characteristics of the

filter can be tolerated or not. Often slight changes in the beam geometry have a

large effect, especially if aperture and field stops are used in the right places; see

below.

288 7 Practice of TCSPC Experiments

7.2.7 Baffles, Aperture Stops and Field Stops

Even in perfectly designed optical systems, some light is scattered and reflected at

optical surfaces or beam stops. The stray light must be kept out of the detection

light path. Therefore a good optical system has a number of circular stops and

baffles that block unwanted light from the detector. Stray light suppression can be

compared to the shielding of an electronic system. Unfortunately its importance is

similarly underestimated, so that cheap telescopes and microscopes for children

often have better stray light suppression than scientific optical systems.

As a rule of thumb, no parts of the housing, lens holders, or the inside of optical

tubes should be visible from the active area of the detector. This can be achieved

by circular stops that restrict either the field of view of the detector (field stops) or

the effective aperture of a beam (aperture stops). The term „baffle“ is often used;

baffles are cylindrical or conical tubes or circular stops that keep off unwanted

light from the detector [71].

Two examples of baffles are shown in Fig. 7.28. In mirror systems with a

folded beam path, light entering at a large angle from the optical axis light can

often reach the detector directly. An example is the Cassegrainian telescope shown

left; a conical baffle reduces the stray light considerably. A similar situation can

occur if a lens is placed far from a detector (Fig. 7.28, right). Light passing outside

the lens can reach the detector unless it is blocked by a baffle.

Baffle

Mirror

Detector

DetectorBaffle

Lens

Fig. 7.28 Baffles in optical systems

To make a baffle efficient, the inner surface must be blackened and reflection at

grazing incidence must be suppressed. This can be achieved by vanes on the inner

surface or by machining a thread into the baffle. If this is not possible there is a

very efficient trick used by amateur astronomers: Spray the surface with black

paint and let it dry. Then spray it once more and spread sugar on it. Let it dry

again and wash the sugar off. The result is a structured surface that is almost free

of reflection at grazing incidence. Of course, also the inner side of the housing and

all other nonoptical surfaces should be painted with matte black paint. If these

rules are obeyed, there is a considerable reduction of daylight leakage and arte-

facts from scattered excitation or scattered fluorescence light.

Another way to suppress stray light is through field stops and aperture stops. A

field stop is placed in a conjugate image plane of the excited spot in the sample.

An aperture stop is placed most efficiently in the image plane of another aperture,

which can be another stop, a lens, or a mirror. Two examples are shown in

Fig. 7.29. A lens, L1, focuses a laser beam into a sample. The light emitted by the

7.2 Optical Systems 289

sample is collected by the same lens, L1, diverted by a dichroic beamsplitter, and

focused on a detector by another lens, L2. The system is a standard setup of

TCSPC lifetime microscopy, where L1 is the microscope lens. The problem in

such systems is laser light scattered at the edge of the microscope lens, at the mi-

croscope lens itself, and at the dichroic beamsplitter. The stray light can be re-

duced by a field stop in the conjugate sample plane, A, as shown in Fig. 7.29, left.

Detector

A: Conjugate

sample

plane of

Sample

Laser

Detector

A: Conjugate

sample

B: Conjugate

plane of

Sample

Laser

Light distribution

in conjugate

sample plane

Light Distribution

(scattering)

Fig. 7.29 Field and aperture stops

An extreme example of a field stop is confocal detection, where the stop is a

pinhole that blocks all light that does not come from a diffraction-limited spot in

the sample.

A field stop in the conjugate sample plane may not be feasible if the sample is

highly scattering. In this case the light emitted from the surface of the sample

cannot be focused into plane A. Nevertheless, the stray light can be considerably

reduced by an aperture stop. L2 projects an image of the microscope lens into

plane B. A stop in plane B blocks the light scattered at the edge L1, and a consid-

erable fraction of the light scattered at the lens surface and the dichroic mirror.

7.2.8 Detectors

A detailed description of detectors and their performance in TCSPC applications

is given under Chap. 6, page 213. Detector problems related to special TCSPC

applications are also discussed under Chap. 5, page 61. The following paragraph

deals with some practical issues related to single photon detectors and their use in

TCSPC.

7.2.9 Choosing the Detector

The choice of the detectors is dictated by the required time resolution, spectral

range and sensitivity, tolerable dark count rate, detector area, available count rates,

ruggedness, and possible budgetary constraints. The paragraphs below attempt to

290 7 Practice of TCSPC Experiments

answer some frequently asked questions about the selection of the detector and its

accessories.

How fast a Detector do I need?

The transit-time spread of the detector determines the instrument response func-

tion (IRF). As a rule of thumb, a single-exponential lifetime can be measured with

good efficiency down to the full width at half maximum (FWHM) of the IRF. For

shorter lifetimes the efficiency degrades rapidly, i.e. more photons are needed to

obtain the same lifetime accuracy. Nevertheless, lifetimes 10 to 100 times shorter

than the IRF width can be measured. The practical limit is given by the IRF stabil-

ity, which is of the order of 1 ps (see Fig. 7.36 and Fig. 7.37, page 298). Conse-

quently, single-exponential decay functions can be measured with medium speed

detectors, such as the Hamamatsu TO

8 PMTs, or H5783 photosensor modules

(see Fig. 6.39, page 250).

For multiexponential decays the situation is more complex and depends on the

ratio of the lifetimes and the ratio of the amplitude coefficients. If the ratio of the

lifetimes to each other is on the order of 10 and the ratio of the amplitudes close to

1, the components can easily be resolved even if the short one is hidden within the

IRF. Lifetimes components closer than 1:1.5 are generally hard to resolve. The

situation becomes almost hopeless if two components shorter than the IRF width

have to be resolved. Therefore, the detector should be faster than at least the sec-

ond fastest decay component. Moreover, to resolve multiexponential decays it is

helpful to have a clean IRF without secondary peaks and bumps. Complex decay

functions are a strong argument for using an MCP PMT.

A crucial point of detector selection is whether or not an accurate IRF can be

recorded in the given optical system. IRF recording is often a problem in micro-

scopes or other systems that use the same beam path for excitation and detection.

Reflection and scattering makes it difficult to record an accurate IRF in these

systems. In two-photon microscopes the detector may not even be sensitive at the

laser wavelength, or the laser wavelength may be blocked by filters. If an accurate

IRF is not available, lifetimes much shorter than the detector IRF cannot be relia-

bly deconvoluted. The rule of thumb is to use a detector with an IRF width shorter

than the shortest lifetime to be measured.

Another point to be considered is the pulse width of the light source and the

pulse dispersion in the optical system. Multimode fibres or fibre bundles used at

high NA can easily add a few hundred ps to the IRF widths. It is, of course, not

necessary to use a detector that has an IRF width shorter than 30

50% of the pulse

dispersion of the optical system.

Which cathode Version is the Best?

The most common cathode types for PMTs are the bialkali, the multialkali, the

extended multialkali, and the GaAs and GaAsP cathodes. Typical curves of the

cathode radiant sensitivity are given in Fig. 6.16, page 230. The selection of the

cathode is often a tradeoff between red and NIR sensitivity and dark count rate.

The extended red multialkali cathode and the GaAs and GaAsP cathodes re-

quire cooling. Coolers are often bulky and expensive, imposing constraints on the

7.2 Optical Systems 291

optical design. A convenient solution is PMT modules with internal peltier cool-

ers, such as the Hamamatsu H7422.

The spectral transmission characteristics of the optical system are a frequently

neglected issue. A system containing flint-glass elements turns nontransparent at

wavelengths below 360 nm. In two-photon microscopy or two-photon FCS, the

laser blocking filter limits the spectral range on the long- wavelength side. A mul-

tialkali PMT detecting through 3 mm BG 39 glass delivers almost the same spec-

tral response as a bialkali PMT (see Fig. 5.91, page 156).

Single-photon avalanche photodiodes are often praised for their high quantum

efficiency. It is correct that these detectors have a peak quantum efficiency of

almost 80%. Nevertheless, APDs are far from being a panacea. The spectral effi-

ciency curve is that of a silicon photodiode, and peaks between 700 and 800 nm.

In the visible range the efficiency is lower. The crossover point for the efficiency

curve of a GaAsP cathode is at about 500 nm (see Fig. 6.17, page 231). Moreover,

the detector area is less than 0.1 mm

, compared to 50 to 100 mm

of a PMT. The

efficiency of the APD can be exploited only if the light can be focused into the

active area. This is definitely not the case for diffuse optical tomography and two-

photon microscopy of thick tissue.

When comparing the sensitivity of different detectors, please note that there are

different specifications in use (see Sect. 6.2.6, page 229). The „cathode radiant

sensitivity“ is the cathode current for a given incident power at a given wave-

length. It is usually given as a plot versus wavelength. The cathode radiant sensi-

tivity is directly related to the cathode quantum efficiency. Both the cathode radi-

ant sensitivity and the quantum efficiency can be used to compare different

detectors. However, measurement of the radiant sensitivity or quantum efficiency

is extremely difficult, and some variance in the specifications is therefore un-

avoidable. Moreover, different detectors of the same type may differ slightly in

sensitivity, especially at the long wavelength end of the spectral range. The effec-

tive radiant sensitivity and quantum efficiency of a detector may also be lower

than the value specified for the cathode. Not all photoelectrons are transferred into

the dynode system of a PMT and not all electron-hole pairs trigger an avalanche in

an single photon APD.

The „cathode luminous sensitivity“ is the cathode current per watt incident light

power from a tungsten lamp operated at 2,856 °C. The cathode luminous sensitiv-

ity is the integral of the product of the cathode radiant sensitivity and the lamp

spectrum. Because the lamp has its emission peak in the NIR the cathode lumi-

nous sensitivity lets the sensitivity of NIR-sensitive cathodes appear higher than it

actually is.

The „anode luminous sensitivity“ is the anode current per watt incident light

power from a tungsten lamp operated at 2,856 °C. It is the product of the cathode

luminous sensitivity and the PMT gain. Because almost any gain can be obtained

by changing the supply voltage, the anode luminous sensitivity cannot be used to

compare different PMTs.

Does cooling Increase the Sensitivity of a PMT?

Cooling does not increase the quantum efficiency of a PMT. It does, however,

reduce the dark count rate. A cooled detector does not count more signal photons,

292 7 Practice of TCSPC Experiments

but delivers less background pulses. If the signal count rate is of the order of the

background rate or below, cooling does increase the signal to-noise ratio, espe-

cially if the acquisition time is not limited. However, before you install a cooler,

make sure that the background is not caused by afterpulsing, leakage of daylight,

filter fluorescence, or electrical noise pickup.

Can dark counts be suppressed by increasing the Discriminator

threshold?

The pulse height distribution of the dark pulses is almost the same as for the pho-

ton pulses. Increasing the threshold reduces the dark counts and the signal counts

by the same ratio. Please note that this applies only to dark pulses originating from

the detector itself. Counts caused by electrical noise pickup can, of course, be

suppressed. However, the better solution is correct shielding of the detector.

Does higher Preamplifier gain increase the Sensitivity?

A preamplifier cannot make two photons from one. Higher gain does therefore not

directly increase the sensitivity. However, an increase of sensitivity can be ob-

tained if the amplitude of a fraction of the single-photon pulses is too small to be

detected by the discriminator of the TCSPC device. Moreover, with the preampli-

fier, the PMT can be operated at a lower gain, which results in less afterpulsing

and less signal-dependent background. Reducing this background improves the

statistics of the data and therefore the accuracy of a curve fitting procedure.

7.2.10 Quick Test of PMTs

In a TCSPC system it may be necessary to test whether or not a PMT is working.

A simple test can be made by a general-purpose oscilloscope. To withstand a pos-

sible accident like a discharge in a damaged tube or voltage divider, the oscillo-

scope should have a maximum input voltage of several hundred V. Nevertheless,

before you apply a high voltage to the PMT, you should make sure that the cable

connections are not damaged and that there is a reliable ground return path. Do not

connect to or disconnect the PMT cable from the oscilloscope when the high volt-

age is on. Please see Sect. 7.6, page 315.

To run the test, switch the oscilloscope input to 1 M

:, DC and select an input

voltage range of 10 mV per division or less and a time base of 10 µs per division.

Activate the trigger for the oscilloscope channel being used and select the „norm“

trigger mode and a trigger level of –5 to –10 mV. Make sure that there is

no light

on the PMT. Then start to increase the operating voltage of the PMT. For PMT

modules with internal high voltage generator, switch on the power supply and

increase the gain control voltage. When the PMT operating voltage approaches

80% of the maximum value you should see the first dark pulses of the PMT. The

pulses are negative and have a sawtooth shape, with a steep leading edge and an

slow exponential trailing slope. Of course, this pulse shape is not the true shape of

the single-photon pulse delivered by the PMT. It is simply the result of the RC

time-constant formed by the sum of the PMT anode capacitance (5 to 10 pF), the

cable capacitance (typically 80 pF/cm), and the oscilloscope input capacitance (10

7.2 Optical Systems 293

to 30 pF) in conjunction with the total load resistance (1 M

: parallel with a resis-

tor possibly connected to ground inside the PMT housing):

fall

= (C

anode

+ C

cable

+ C

inp

)

load

(7.5)

The average amplitude of the pulses is

inpcableanode

pmt

peak

CCC





(7.6)

fall

= fall time constant of the observed signal

anode

= anode capacitance of the PMT

cable

= cable capacitance, approximately 100 pF per meter

inp

= input capacitance of the oscilloscope

pmt

= PMT gain at the used operating voltage

load

= Total load resistance

e = elementary charge, 1,602

-19

For a gain of 10

and a total capacitance of 100 pF, the average pulse amplitude

is 1.6 mV. The amplitude of the largest pulses is 5 to 10 times higher than the

average amplitude. Therefore you should see the pulses when the gain approaches

a few 10

. If you give light to the PMT, the pulse rate increases and the signal

turns into a more or less continuous signal. However, be careful with the light

intensity. Any intensity that you can

see is far beyond the safe level for a PMT

operated at high gain.

Pulses of a Hamamatsu R5600 miniature PMT, a Hamamatsu R3809U MCP,

and an Amperex XP2020UR PMT are shown in Fig. 7.30.

Fig. 7.30 Test of the basic function a PMT with an oscilloscope. R5600 (left,

load

= 250 k:, V = –980 V, 5 mV/div), R3809U (middle, R

load

= 1 M:, V = –3.0 kV,

2 mV/div) and XP2020RU (right, R

load

= 500 k:, V = –2.9 kV, 200 mV/div)

The different amplitude of the pulses is the result of the different gain of the de-

tectors (please note the different input voltage range of the oscilloscope). The

different slopes are mainly a result of the different load resistance.

294 7 Practice of TCSPC Experiments

7.2.11 Signal-Dependent Background

All photon counting detectors suffer more or less from afterpulsing (see

Sect. 6.2.8, page 233). Afterpulsing occurs on the time scale of a few microsec-

onds. In high-repetition-rate TCSPC experiments, the afterpulses of many signal

periods accumulate and deliver a considerable background level.

Figure 7.31 shows recordings of a fluorescence signal (top) and of the dark

counts (bottom) for a cooled H5773

20 photosensor module (left) and an

R3809U

50 MCP-PMT (right).

Fig. 7.31 Signal-dependent background. Fluorescence signals and dark counts recorded by

a cooled H577320 PMT module (left) and a R3809U50 MCP-PMT. The background of

the fluorescence decay measurements is substantially higher than the dark count level

Both detectors were used with a 20 dB, 1.6 GHz preamplifier. The H577320

was operated at a gain control voltage of 0.9 V, the R3809U

50 at an operating

voltage of –3 kV. The pulse repetition rate was 50 MHz, the count rate approxi-

mately 400 kHz, the acquisition time 40 seconds. For both detectors the back-

ground of the fluorescence measurement is considerably higher than the dark

count level. However, the ratio of the fluorescence counts to the background

counts is more than 10 times better for the R3809U

50 than for the H577320.

This result is in agreement with the low afterpulsing probability found for the

R3809U

50 (see Fig. 6.32, page 245).

Figure 7.31 shows that the dynamic range of fluorescence decay measurements

can be severely limited by afterpulsing. For a given detector, the afterpulsing

probability and thus the background can be reduced by reducing the gain. The

reduced detector gain must be compensated for by a higher preamplifier gain or a

reduced discriminator threshold. However, loss of time resolution and increase of

differential nonlinearity set a lower limit to the detector gain.

Another way to increase the dynamic range is to reduce the pulse repetition

rate. The downside of this solution is the increase of pile-up distortion or acquisi-

tion time.

It should be noted that filter fluorescence can cause a signal-dependent back-

ground that is almost indistinguishable from the background caused by afterpuls-

ing. Therefore, if background is a problem, the optical system should also be