Leaf scattering polarimetry and chlorophyll fluorescence

A number of spectropolarimetry measurements on leaves using white light and narrow band illumination have been reported in recent years because of interest in signatures for remote sensing of exoplanet life. Chlorophyll fluorescence occurs in white light illuminated experiments and is known to be polarised and so might contaminate surface scattering measurements. A displacing filter experiment was performed on the leaves of two common plants, Ficus benjamina and Chamaedorea elegans , to estimate the contribution of chlorophyll fluorescence in the spectropolarimetry of leaf scattering and transmission. Chlorophyll fluorescence is present as a measurable polarised component of up to 15% of scattered and transmitted light for the leaf samples. Other effects emerged which indicate that in vivo measurements on leaves are subject to the responsive nature of the leaves to changes in colour and intensity of the incident light. Crown Copyright © 2019 Published by Elsevier Ltd. This is an open access article under the CC BY license. ( http://creativecommons.org/licenses/by/4.0/ )


Background
Spectropolarimetry has been used to explore linear and circular polarised scattering from a variety of algae and plant leaves to measure possible chiral scattering effects that could potentially be used to identify the presence of photosynthesis in the optical signature of exoplanets [1][2][3][4][5][6] . There are a number of persistent issues amongst the current measurements including the large variety of plants and algae used by various experimenters, the lack of agreed standards, and differing techniques and apparatus to measure nominally the same set of Stokes and Mueller matrix components of the scattered or transmitted light. This short article attempts to address one aspect of the comparison of spectropolarimetric experiments with one another, namely the presence of polarised fluorescence in the scattered light from the biological sample and the effect this may have on the precise measurement of the Stokes parameters and scattering matrix coefficients.

Experimental configuration
A high-precision spectropolarimeter [4] was used to measure the transmitted or scattered light from a leaf or calibration object. The sample was illuminated with a quartz halogen lamp (Thorlabs QTH10) with an added Lyot depolarizer to ensure only unpolarised light is incident on the sample. A 10 nm narrow band filter (e.g. Thorlabs FB680-10) is placed alternately between the sample and the light source or between the sample and the polarimeter. In the E-mail address: w.e.martin@herts.ac.uk first case the configuration is optically similar to the laser spectropolarimetry in refs. [3,4,5b] with narrow band illumination. In the second case the white light illuminates the sample and the narrow band filter defines the wavelength of the polarimeter measurement similar to refs [1,2,5a,6] . Differences in the scattered or transmitted light between the two measurements will give an indication of how narrow band and white light polarimetry techniques on leaves and other chlorophyll containing biological material differ.
The experimental configuration for scattering measurements is shown in Fig. 1 . The narrow band filters were held in a fixture that allowed precise positioning and orientation to be maintained when moving from between the sample and the polarimeter head (a, after) to between the sample and the illumination source (b, before). Beam stops and shielding were arranged to ensure there was minimal stray light and to confine the illumination to an 8 mm diameter spot on the filter/sample. The stop diameter and the detector apertures combine to a solid angle for scattered light collection of ~0.02sR. The stop positions are indicated in Fig. 1 . Fig. 2    process. The samples were illuminated only during the measurements as will be discussed later.
Scattering measurements were taken with the polarimeter at 45 °to the beam centre line and with the sample plane normal to the illumination: The incidence angle between the incident light and the surface normal was θ = 0 °, the observation angle was ϕ = 45 °For transmission measurements θ = 0 °, ϕ = 180 °.

Samples and calibration
The incident light source had a residual linear polarization fraction of ~0.01 over the lamp emission range and this was reduced to an average of ~0.001 over the 640-740 nm measurement range with the use of the Lyot depolarizer. A few of the dichroic filters used in the measurements had detectable birefringence ( < ~.02 λ). The axial and azimuthal orientation of the filters was rigidly fixed so that no rotation or tilt occurred when the filters were moved between the two measurement positions. Any residual filter birefringence cancels to very low values as will be seen in the data graphs. The approximate average irradiance on the samples for the narrow band illumination was 3 × 10 − 4 W/cm 2 ; the irradiance for the white light illumination was ~3 × 10 − 2 W/cm 2 .
Calibration of the scattering measurements was accomplished using a Spectralon [15] standard as a uniform Lambertian scatterer. An opal glass diffuser [Edmund 46-167] was used as a consistent reference for the transmission measurements. The latter may have been effected somewhat by residual strain birefringence. The calibration measurements are presented with the leaf measurements.
The experiment is simple in concept but the narrow band and white light intensities are different and it is known that leaves change their optical properties in response to the incident light intensity and the illumination history [8 , 10-12] . Some data arising from the illumination history will be presented in the Results section.

Results
The scattered light measurements from the two types of leaves, Ficus benjamina and Chamaedorea elegans , or ' ficus ' and 'palm' respectively are given below. The 'a' values are with the filter after the sample -wide band illumination; the 'b' values are for the filter placed before sample -narrow band illumination.

Total scattering and transmission
The raw data for the total scattered light and transmitted light from the samples is given in Figs. 3 and 4 . The 'a' and 'b' measurements are very similar for a given leaf type but the ficus and palm leaves have dissimilar scattering and transmission curves. The behaviour of interest begins to emerge when differences between the a and b measurements are calculated. The three sigma error bars are smaller than the data symbols in these graphs. The curves joining the data points are smooth splines for these and all other graphs and are not physical.

Normalised scattering and transmission differences
The normalised scattering and transmission differences are calculated as follows from the raw measurements: where Rx and Tx are respectively the measured total scattering and transmission intensity measurements from the averaged output of the detectors in Fig. 1 .
These values are the difference in the light scattered from the sample or transmitted by the sample at a given wavelength depending on whether the illumination is broad or narrow band. Figs. 5 and 6 below show the result of measurements on the two leaf types and on the calibration samples. Also included in Fig. 5 is a result from a simple model of the scattering to be discussed later. Fig. 5 also shows a scaled fluorescence spectrum derived from a 405 nm laser diode substituted for the lamp illuminating a Ficus benjamina leaf in the 'a' configuration. This spectrum is similar to that observed by others [12,13] and uses the same configuration as the scattering measurements.
The scattering, ra-rb, measurements contain spectral features that are similar to a fluorescence spectrum. The ra-rb difference should be principally the difference in fluorescence efficiency between broadband, 'a', and narrow band, 'b' illumination. However, the greater total light flux on the leaves in the 'a' configuration can result in changes in both scattering and fluorescence emission due to adaptation [8,10,11] so the net effect is likely to be complex. Transmission difference measurements are less clear with large differences between the two leaf types and relatively large variations in the opal glass transmission measurements which are reproducible.

Stokes components and matrix coefficients
The calculation of the Stokes and Mueller Matrix coefficients follows straightforwardly from the modulated signals detected with the lock-in amplifier. With unpolarised illumination the Mueller matrix collapses to a single column vector with only m 11 , m 21 , m 31 , and m 41 as coefficients. Using this property simplifies the calculations [see Ref 3 , (2) where v(n ω) is the ac voltage from detectors at the appropriate multiple of the PEM modulation frequency ω, I dc is the dc volt-  [3] . These are coefficient values, i.e., normalised to the overall signal levels, and are insensitive to the small differences in overall scattering or transmission levels found in the previous section. The circular scattering values, Stokes V and m 41 , are significantly larger than the linear scattering, Stokes Q and m 21 in the chosen scattering geometry. Fig. 7 also shows an inset with the polarised fluorescence emission components resulting from a 532 nm laser substituted for the lamp in Fig. 1 . The ficus polarisation coefficients do not have significant wavelength variations or phase changes and this will be discussed in a later section.
The 'b' measurements in Fig. 8 can be compared to the recent extensive Mueller matrix transmission measurements of Patty,  ±3 σ error bars are visible or are smaller than the data symbols. The small inset shows fluorescence emission coefficients for ficus with a 532 nm laser replacing the lamp in Fig. 1 in the 'a' configuration. et al. [5b] . Leaves from maize and maple were measured with an imaging spectropolarimeter able to differentiate between veins and surfaces. Fig. 8 indicates that average transmission coefficients of ficus and the thinner palm leaves are different with m 41 larger than m 21 in both types. In [5b] the magnitudes of the maple and maize coefficients are similar to ficus when averaged but the maize m 21 coefficients are larger than the m 41 for most of the 640-720 nm range. Fluorescence effects are likely to be small in [5b] but not entirely absent similar to Fig. 8 .
Taking the a-b differences, e.g. m 41 = m 41 a-m 41 b, as for the total scattering and transmission coefficients gives the following graphs in Figs. 9 and 10 . The scattered light polarisation coefficient differences are quite clear and indicate that wide band versus narrow band illumination results in significant differences in  the scattering coefficients m 41 and m 21 . Both "unpolarised to circular scattering differences" ( m 41 ) and "unpolarised to linear scattering differences" ( m 21 ) are present with the linear coefficient differences being an order of magnitude smaller. The transmission coefficient differences are much smaller for both leaves and are relatively free of wavelength dependent features.

Temporal and illumination variability
A shutter was used to control the exposure of the samples to the illumination source and was opened only during measurements. A typical run of data for a single wavelength consisted of nine sets of five measurements of the Stokes components which were then averaged and the standard deviation calculated for the total scattering or transmission V and Q components. Fig. 11 shows a typical total scattering measurement (Ra) illumination history for a palm sample at 670 nm to 690 nm with the filter after the sample, i.e. the sample is illuminated by the full spectrum of the tungsten halogen source during the shutter open times. The temporal changes at shorter and longer wavelengths were much smaller than that at the peak fluorescence of ~680 nm. The narrow band (Rb) data shows negligible temporal variations during the illumination periods at all wavelengths.
The light scattered and emitted from the sample has at least two time dependent components, a longer decay time and a shorter recovery time. The longer quenching or decay time con-  stant is approximately 30 s. and the recovery time is approximately 15 s. This behaviour has been studied by many authors [10][11][12] and the decay times are consistent with quenching and recovery of leaf chlorophyll fluorescence in a dark adapted sample.

Discussion
There are clear differences between the broad-band illumination ('a' or filter after sample) and narrow band illumination ('b' or filter before sample) in both scattering and transmission from the leaves. Assuming that broadband illumination will contain the maximum fluorescence, the difference between ra and rb should reveal the fluorescence spectrum directly. The result of the mea-surement was that the difference between the 'a' and 'b' total scattering revealed ra > rb at wavelengths at longer than 680 nm and ra < rb at wavelengths shorter than 680 nm. The difference was about 15% in either case.
A simple model was used to examine this behaviour. The 'a' measurement with the filter after the sample is assumed to be of the form where Ia( λ) is the measured (scattering) intensity, Tf( λ) is the filter transmission, Ip( λ) is the normalised lamp spectrum (see Fig. 2 ), R( λ) is the leaf reflectivity and F( λ) is the fluorescence intensity. C is a constant related to fraction of light as fluorescence, i.e., the to-tal fluorescence emission with broadband illumination at the given unfiltered illumination level. The 'b' measurement is assumed to be of the form (4) where Ib( λ) is the measured (scattering) intensity with the filter before the sample, and D is a small constant.
Using the data in Fig. 2 for Ip( λ) and Tf( λ), the fluorescence data for 405 nm illumination of ficus from Fig. 5 for F( λ), and C = 0.45, D = 0.2 produces the curve of Ia-Ib in Fig. 5 . While the fit to the experimental data in Fig. 5 is not particularly good at longer wavelengths, the major features are captured and the conclusion is that leaf fluorescence accounts for up to 15% of the light from broadband leaf scattering. The principal assumptions are that the leaf reflectivity, R( λ), is approximately constant with respect to illumination intensity and the fluorescence excitation efficiency is approximately constant with wavelength. Note that the latter assumption implies some fluorescence will be present even with narrow band illumination as long as absorption is present. This means that the a-b measurements here represent a lower bound to the actual fluorescence content of the light detected from broad band illumination. The use of more complex functions for D involving leaf absorption or reflectivity [13] do not give a significantly better fit to the data. The physical significance of D can be assigned to a small constant difference in R( λ) between the two different illumination levels of the 'a' and 'b' measurements. This difference is apparent in the data of Fig. 3 but it is not particularly obvious.
The polarisation coefficient measurements are consistent with previous narrow band measurements of leaves with this apparatus [3] which show large circular scattering coefficients at wavelengths shorter than 700 nm, a zero crossing in the scattering coefficient at ~710 nm, and a further rise and fall with a phase reversal at λ~710 nm. Differencing the 'a' and 'b' scattering coefficients reveals curves, Fig. 9 , showing that both the V ( m 41 ) and Q ( m 21 ) scattering (plus fluorescence) are positive for λ < 670 nm and negative for λ > 670 nm. It is posited that polarised chlorophyll fluorescence is principally responsible for the scattered light V ( m 41 ) and Q ( m 21 ) a-b curve shapes.
In a separate measurement in the Fig. 1 apparatus 'a' configuration with 532 nm or 405 nm illumination, excitation that is completely blocked by the filters, the pure fluorescence from a ficus leaf in Fig. 3 gives the curves in the inset in Fig. 7 for fluorescence polarisation coefficients. There is no evidence of a polarisation sign change in V or Q in this fluorescence signal. These measurements at very high intensity (~0.1 W/cm 2 ) revealed differences between the emission spectra with 405 nm and 532 nm illumination and between different sam ples. Tem poral variations with laser illumination are also much larger than seen in Fig. 11 'a' measurements at all wavelengths. See reference [12] for examples of these factors. Further detailed measurements of the fluorescence polarisation emissions versus scattering dependencies are needed at varying illumination levels and on other samples.
The transmission measurement ta-tb differences, Figs. 6 , 8 , and 10 , were less useful because of smaller signal to noise ratios in general and poor performance from the transmission standard. The chlorophyll fluorescence content in leaf transmission light is also subject to self-absorption and the complex absorption and emission properties of the antenna compounds [13,16] as well as additional (multiple) scattering due to internal structures. There is however an indication of transmitted fluorescence in the thinner palm leaves in the form of a negative sign in ta-tb at wavelengths < 690 nm in Fig. 6 .
Temporal variations in the measured scattering and transmission with broadband illumination point to another consideration in spectropolarimetric measurements of chlorophyll containing material: Illumination magnitudes and history may be significant to the measurement. The principal effects on the measurements in this paper are seen in Fig. 11 where broadband illumination quenches the 670 nm to 690 nm light from a palm leaf in the initial measurement sequence. Also shown are sequences from the narrow band 'b' measurements at 670 nm and 690 nm for this leaf. Comparing the 690a and 690b sequences, the 'a' broadband illumination quenching results in a ~2% reduction in the average light from the leaf. The 'b' sequence shows no quenching but the scattered light on average is about 12% less (little fluorescence is present). Temporal variations at all other 'b' wavelengths are also unobservable. Although not specifically recorded, repeat measurements after moderate time intervals (20-60 min) on individual leaves indicate the scattered and fluorescence light from the leaf returns to the initial dark adapted level so that no permanent damage is apparent.

Conclusions
The original premise of this work was that chlorophyll fluorescence would have an effect on the Stokes scattering measured using narrow band light sources [3,4,5b] and broadband sources with monochromators or filters [1,2,5a,6] . This has been shown to be the case in scattered light measurements. Future spectropolarimetric measurements, particularly with sun illuminated vegetation, should carefully consider the contribution of polarised chlorophyll fluorescence to the direct scattering from the leaf surfaces. This will be collection geometry dependent in many experiments, particularly field measurements in remote sensing [5a,6,9,12,17,18] . Partially polarised fluorescence contributes as much as 15% of the measured scattered light between 640 nm and 740 nm in the measurements presented here.
The results in Figs. 5 and 9 indicate that polarised fluorescence is a major component in the variation in scattering coefficients between wide band and narrow band illumination. Previous measurements [3,4] using narrow band illumination have consistently failed to find the 'chiral-like' spectral forms for circular scattering observed in white light experiments [1,2,5a,6] . It is proposed here that a significant portion of chirality signatures previously observed in leaves and bacteria in white light experiments may arise from chlorophyll fluorescence and not direct chiral elastic scattering [14] from chlorophyll containing structures in the biological material. Multiple scattering from structures at or just below the leaf surface will also contribute additional angle dependent circular polarisation components to add to the illumination history dependent polarised fluorescence emissions. The light interaction process is complex to disentangle for all configurations and additional detailed circular polarisation analysis of chlorophyll fluorescence emissions and scattering from in vivo samples will benefit future studies.
The conclusions regarding the sensitivity of spectropolarimetric measurements to chlorophyll chirality for use in remote sensing of exoplanet life should take account of the presence of polarised fluorescence. Fluorescence polarisation is dependent on the chiral properties of chlorophyll and possible chlorophyll analogues and is a strong signature in its own right in scattered light from leaves.

Declaration of Competing Interests
None.
that have made it more readable. The two Reviewers have made very useful suggestions that have improved the presentation. Support from the UK Science and Technology Facilities Council , Grant ST/R0 0 06598/1 is gratefully acknowledged.

Supplementary material
Supplementary material associated with this article can be found, in the online version, at doi: 10.1016/j.jqsrt.2019.106760 .