* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download leading, peer-reviewed scientific journal
Survey
Document related concepts
Rutherford backscattering spectrometry wikipedia , lookup
3D optical data storage wikipedia , lookup
Ultraviolet–visible spectroscopy wikipedia , lookup
Phase-contrast X-ray imaging wikipedia , lookup
Nonlinear optics wikipedia , lookup
Optical aberration wikipedia , lookup
Silicon photonics wikipedia , lookup
Photon scanning microscopy wikipedia , lookup
Nonimaging optics wikipedia , lookup
Ellipsometry wikipedia , lookup
Surface plasmon resonance microscopy wikipedia , lookup
Retroreflector wikipedia , lookup
Dispersion staining wikipedia , lookup
Birefringence wikipedia , lookup
Transcript
Icarus 268 (2016) 156–171 Contents lists available at ScienceDirect Icarus journal homepage: www.journals.elsevier.com/icarus Polarimetry of moonlight: A new method for determining the refractive index of the lunar regolith Andrew Fearnside a,⇑, Philip Masding b, Chris Hooker c a No. 2 The Cobbles, Boots Green Lane, Allostock, Knutsford, Cheshire WA16 9NG, UK Imber House, Vale Road, Bowdon, Cheshire WA143AQ, UK c 6 High Street, Didcot, Oxfordshire OX11 8EQ, UK b a r t i c l e i n f o Article history: Received 31 March 2015 Revised 10 October 2015 Accepted 30 November 2015 Available online 29 December 2015 Keyword: Polarimetry Moon, surface Regoliths Mineralogy a b s t r a c t We present a new method for remotely measuring the refractive index of the lunar regolith, using polarised moonlight. Umov’s Law correlates the polarisation (Pmax) of scattered moonlight and the albedo (A) of the scattering lunar regolith. We discuss how deviations from this correlation have previously been linked to the so-called ‘Polarimetric Anomaly Parameter’, (Pmax)aA, which was proposed by Shkuratov and others as being related to variations in regolith grain size. We propose a reinterpretation of that parameter. We develop models of light scattering by regolith grains which predict that variation in the refractive index of regolith grains causes deviations from Umov’s Law. Variations in other grain parameters such as grain size and degree of space weathering do not produce this deviation. The models are supported by polarimetric measurements on powdered terrestrial materials of differing refractive index. We derive a simple formula to express the relationship between refractive index and the deviation from Umov’s Law and apply it to telescopic measurements of regions of the lunar surface. We show that the Aristarchus Plateau and the Marius Hills regions both comprise materials of unusually low refractive index. These results are consistent with recent estimates of the mineralogy of those areas. Picard and Peirce craters, in Mare Crisium, are shown to contain material of low refractive index similar to highland regions, as has been suggested by earlier studies of these craters. Ó 2015 Elsevier Inc. All rights reserved. 1. Background Moonlight is linearly polarised at almost all lunar phase angles. The degree of linear polarisation is defined as P = (I\ I||)/(I\ + I||) where I\ and I|| are the intensities of the components of moonlight with electric field vectors resolved perpendicular (\) or parallel (||) to the scattering plane. This quantity varies with lunar phase angle and reaches a peak (Pmax) at lunar phases typically lying between about 95° and about 110°. This peak value correlates strongly with the albedo (A) of the lunar surface region in question. The correlation, known as Umov’s Law, is linear when displayed on logarithmic axes and may be described by an equation of the form: log10(A) + a1log10(Pmax) = a2. Dollfus and Bowell (1971) applied this regression relation to observations of over a hundred small regions across the lunar surface. In the visual spectral range of light, they found that a1 = 0.72 and a2 = 1.81. Deviations from Umov’s Law caught the interest of Shkuratov, Dollfus and others (Shkuratov, 1981; Dollfus, 1998; Shkuratov and Opanasenko, 1992; Geake ⇑ Corresponding author. E-mail address: [email protected] (A. Fearnside). http://dx.doi.org/10.1016/j.icarus.2015.11.038 0019-1035/Ó 2015 Elsevier Inc. All rights reserved. and Dollfus, 1986; Novikov et al., 1982) who studied this law in relation to granular terrestrial materials and lunar regolith samples alike. When applied to moonlight, it was found that departures from Umov’s Law occurred as a ‘polarisation excess’ when associated with the ejecta of some bright young craters, and as a ‘polarisation deficit’ notable on the Aristarchus Plateau (so-called ‘Wood’s Spot’) and the Marius Hills, for example. It was suggested that a ‘polarisation excess’ was caused by anomalously large median grain sizes within the granular material being observed. The ‘Polarimetric Anomaly Parameter’: ðPmax Þa1 A was derived as a means of quantifying departures from Umov’s Law and it was suggested that variations in this parameter quantified variations in the median grain size of the lunar regolith. Shkuratov proposed that ‘polarisation deficit’ might result from grains of anomalously low albedo causing a breakdown of Umov’s Law, though later work suggested that lower median grain sizes might be the cause. The question of how to interpret ‘polarisation excess’ and ‘polarisation deficit’ has received very little attention since. Indeed, more recent work (Shkuratov et al., 2007; Hines et al., 2008; Jung et al., 2014) on lunar polarimetry has assumed the correctness of this interpretation of the ‘Polarimetric Anomaly Parameter’. A. Fearnside et al. / Icarus 268 (2016) 156–171 In this paper we address this question. We propose a new interpretation of ‘polarisation deficit’ and ‘polarisation excess’. This new interpretation is that departures from Umov’s Law are the result of variations in the refractive index of lunar regolith grains. This allows a new method for determining the refractive index of the lunar regolith remotely. 2. The plan of this paper We begin in Section 3 by presenting a simple 2-dimensional (2D) model of light scattering from a lunar regolith grain. We show how this model may be used to interpret the structure of a correlation obeying Umov’s Law, including a consistent interpretation of both a ‘polarisation excess’ and a ‘polarisation deficit’. We then test this simplistic 2D model by considering a 3-dimensional (3D) multi-grain regolith model and show that it fully supports our interpretation of ‘polarisation excess’ and ‘polarisation deficit’. Next, in Section 4, we describe polarimetric experiments conducted on terrestrial grain samples and lunar regolith simulants. We show that these experiments confirm our new interpretation of ‘polarisation excess’ and ‘polarisation deficit’. Finally, in Section 5, we apply our interpretation to the polarimetry of moonlight. We show how polarisation measurements can be applied to determine the refractive index of the lunar regolith. The Aristarchus Plateau and the Marius Hills are considered in detail, in order to allow a comparison with the work of Shkuratov (1981) who considered these areas in particular. Mare Crisium is also discussed in terms of its stratigraphy as revealed by observational results presented here. 3. Theory Our simple 2D model represents an idealised regolith grain in isolation. This model assumes that inter-grain scattering of light has a negligible influence upon the polarisation properties of moonlight. Our 3D model takes account of inter-grain scattering and we will show that these two models provide a mutually consistent interpretation of Umov’s Law. tive index of the grain core was based on values for pyroxene and olivine stated by Hiroi and Pieters (1994), which include the necessary wavelength dependency. For npFe, the optical properties of iron were taken from Johnson and Christy (1974). By applying Maxwell–Garnett effective medium theory, we define: 3/ðK Fe K 2 Þ=ðK Fe þ 2K 2 Þ K1 ¼ 1 þ 1 /ðK Fe K 2 Þ=ðK Fe þ 2K 2 Þ 3.1.1. Grain structure A lunar regolith grain is represented as a core of uniform material of diameter D as shown schematically in Fig. 1. This core carries a uniform coating of space-weathered core material defining a shell of thickness d. The shell consists of a proportion (/) of nano-phase metallic iron beads (npFe) embedded in the same material as the core. The effective refractive index of the spaceweathered layer is represented by m1 = n1 + ik1 and is calculated using the refractive index (m2 = n2 + ik2) of the core material and the refractive index (mFe = nFe + ikFe) of the npFe beads. The refrac- ð1Þ Here, KFe = (nFe + ikFe)2, K2 = (n2 + ik2)2 and K1 = (n1 + ik1)2 from which m1 is obtained. The absorption coefficients of the grain core and space-weathered layer, for light of wavelength k, are defined as a2 ¼ 4pk2 =k and a1 ¼ 4pk1 =k respectively. The absorption coefficient (a2) for the grain core was derived from graphs presented by Nimura et al. (2008) concerning Apollo 16 soil samples. This type of model has also been used by Nimura to represent the upper layer of the lunar regolith (Nimura et al., 2008). We calculate the Fresnel amplitudes ^rjj;? ðhÞ for reflected light ab polarised perpendicular (\) or parallel (||) to the scattering plane at each internal and external grain boundary, where h is the angle of incidence of a ray traversing between medium ‘‘a” and medium ‘‘b”. Transmission and reflection coefficients were then calculated, and applied in a manner suggested by Hapke (2005) to derive an expression for the proportion of light scattered by the grain at a given phase angle (g), and the degree of linear polarisation (P(g)) of that scattered light. To do this, we calculated the single scattering albedo (w) of the grain in terms of the total scattering (sab) and transmission (tab) activities using angle-averaged values R ðr jj;? ¼ 2 j^rjj;? ðhÞj2 cos h sin hdhÞ of the Fresnel amplitudes. These ab ab are identified in Table 1 in terms of boundaries between different media at or within the grain. The reader is referred to schematic Figs. 2 and 3 presented by Nimura et al. (2008) which succinctly show the light scattering processes being modelled here. The total single scattering albedo is given by w = (w\ + w||)/2. We calculated the degree of linear polarisation according to the definition provided by Hapke (Chapter 14, Eq. (14.3) (Hapke, 2005)) in terms of bidirectional reflectance as: PðgÞ ¼ 3.1. 2D model 157 ½X ? ðg=2Þ X jj ðg=2Þ ½X ? ðg=2Þ þ X jj ðg=2Þ þ 2½wH2 Y ð2Þ where g/2 is the angle of incidence and reflection. Here, the term Y is the angle-averaged value of (X\(h) + X||(h)). The terms X\(h) and X||(h) are the back-scattering activity expressed using anglex x dependent (i.e. not angle-averaged) reflectances ðRab ðhÞ ¼ j^rab ðhÞj2 ). These are defined in Table 1 in terms of reflectances at boundaries between different media at and within the grain, and they account for back-scattered light from the space-weathered layer both externally at the grain surface and also internally, having performed one round-trip through the grain core and back. The angle hT = arcsin(sin (g/2)/m2) is the internal angle at which an incident light ray enters the grain core from within the space-weathered layer. It is also the angle at which it subsequently strikes the space-weathered layer from within the core before leaving the grain at an exit angle of g/2. The H-function, defined by the well-known radiative transfer theory of Chandrasekhar, is defined as: 1 pffiffiffiffiffiffiffiffiffiffiffiffi 1 1 Hðl; wÞ ¼ 1 1 1 w l r 0 þ 1 r 0 r 0 l ln 1 þ ; 2 l pffiffiffiffiffiffiffiffiffiffiffiffi l ¼ cosðg=2Þ; r0 ¼ 2=ð1 þ 1 wÞ 1 ð3Þ Fig. 1. The 2D model regolith grain. Thus, the regolith grain is defined in terms of five parameters. We restricted the value of grain parameters to lie within ranges comparable to values identified for the fine fraction of regolith samples returned to Earth by the Apollo and Luna missions. The parameter ranges are shown in Table 2. 158 A. Fearnside et al. / Icarus 268 (2016) 156–171 Table 1 The single scattering albedo (wx) and back-scattering activity (Xx) defined in terms of scattering and transmission activities at boundaries within the grain. The three different media are signified by the following subscripts: ‘‘0” = vacuum; ‘‘1” = space-weathered layer; ‘‘2” = grain core. The polarisation states are denoted by the symbol x in which x = \ denotes the polarisation perpendicular to the scattering plane and x = || denotes the polarisation parallel to the scattering plane. Boundary location Single scattering albedo wx ¼ Scattering (vacuum M shell) x s01 Transmission (vacuum M shell) Back-scattering activity x 1s x ea2 D þt x s x t x e2a2 D s01 ð 21 Þ 01 21 21 x x x x X x ðg=2Þ ¼ S01 ðg=2Þ þ T 01 ðg=2ÞS21 ðhT ÞT 21 ðhT Þe2a2 D x x a2 D x ea2 D t 21 e ð1s21 Þð1t01 Þ x x r x e2a1 d ð1r10 Þð1r01 Þ 12 x ¼ r 01 þ x r x e2a1 d 1r ð 10 12 Þ x x ðg=2ÞÞ R21 ðhT Þe2a1 d ð1R01 x x ðg=2ÞR12 ðhT Þe2a1 d Þ ð1R01 x x ðg=2ÞÞð1R21 ðhT ÞÞea1 d ð1R01 x T 01 ðg=2Þ ¼ 1R x x ðg=2ÞR21 ðhT Þe2a1 d Þ ð 01 x x ðhT ÞÞð1R01 ðg=2ÞÞea1 d ð1R21 x T 21 ðhT Þ ¼ 1R x x ðg=2ÞR21 ðhT Þe2a1 d Þ ð 01 x x ð1r01 Þð1r12 Þea1 d x r x e2a1 d ð1r10 Þ 12 x x ð1r21 Þð1r10 Þea1 d ¼ 1r x r x e2a1 d ð 10 12 Þ x x r x e2a1 d ð1r21 Þð1r12 Þ 10 x ¼ r 21 þ x r x e2a1 d 1r ð 10 12 Þ x t01 ¼ Transmission (shell M core) x t21 Scattering (shell M core) x s21 2 x x ¼ R01 ðg=2Þ þ S01 x x ðhT Þ ¼ R21 ðhT Þ þ S21 2 x x ðhT ÞÞ R01 ðg=2Þe2a1 d ð1R21 x x ðg=2ÞR21 ðhT Þe2a1 d Þ ð1R01 -5 x 10 18 -0.4 (a) -0.5 -0.6 14 -0.8 12 -0.9 10 -1 8 Increasing D or α, φ d 6 -1.3 -1.4 4 -1.5 -1.6 -1 0.04 -0.8 0.02 -0.9 0 -1 -0.02 -1.1 Increasing n -1.2 -0.04 -1.3 -0.06 -1.4 -0.08 -1.5 2 -0.8 -0.6 -0.4 -0.2 0 0.2 log10 (B) 0.06 -0.7 log10 (Pmax) log10 (Pmax) -0.7 -1.2 (b) -0.5 -0.6 -1.1 0.08 -0.4 16 -1.6 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 -0.1 log10 (B) Fig. 2. Umov plots calculated according to the 2D grain model. (a) The effects of a variation in grain diameter. Here, grain core diameter (D) intervals are indicated by hue. The colour side bar indicates core diameter in metres. The same type of graphical structure is reproduced by variations in space weathering parameters (/, d) and in core absorption coefficient (a2). (b) The effects of a variation in grain core refractive index (n2). Data of common refractive index lies within the plot along a common longitudinal band, indicated by hue. The colour side bar indicates refractive index as a deviation relative to n2 = 1.68. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) With the optical wavelength and absorption coefficient fixed, eighteen different values were used for each of the remaining four variable grain parameters spanning each parameter range. This produced 184 different combinations of grain parameters, representing 184 different grains. We calculated the maximum degree of linear polarisation (Pmax) of light scattered from each one of these different grains, and the brightness of light scattered at the phase angle at which Pmax occurred. This information was then plotted upon on a scatter graph with logarithmic axes, collectively for all of the 184 different grains, and we found that the expected linear correlation of Umov’s Law was reproduced as is shown in Fig. 2. Fig. 2(a) and (b) shows how a variation through the range of values of each of the five grain parameters influences the structure of the Umov plot for the model regolith grains. Fig. 2(a) shows that the effect of increasing the value of any one of the mineral core diameter (D), the thickness (d) of the space-weathered layer, the proportion (/) of npFe present within the space-weathered layer, and the absorption coefficient (a2) of the grain core, is to reduce the brightness of scattered light and increase the maximum polarisation. The result is to extend the scatter of data points within the plot linearly along the direction of correlation. This is due to the grain becoming more opaque. Increased opacity reduces the proportion of light emerging from within the grain following internal scattering processes that tend to de-polarise the light. However, the proportion of relatively strongly polarised light resulting from specular surface reflection is not reduced in this way. Consequently, the overall degree of polarisation tends to rise as the grain parameters are varied to increase grain opacity. In contrast, as Fig. 2(b) shows, a variation in the value of the real part of the refractive index (n2) of the mineral core material has the effect of simply shifting, wholesale, the linear distribution of data points in a direction transverse to the usual direction of correlation. This shift is generally uniform along the entire length of the distribution of data points such that a linear correlation is preserved. This is analogous to the effect expected of the Polarimetric Anomaly Parameter, (Pmax)aA, proposed by Shkuratov and others. In the remainder of this paper, we refer to this shift in the Umov plot as a Polarisation Anomaly to distinguish it from the Polarimetric Anomaly Parameter, and we discuss the relationship between these two quantities further. The Polarisation Anomaly is strongly correlated to the real component (n2) of the refractive index of the grains associated with it. An Umov-type scatter plot may now be viewed as being composed of a succession of adjacent, parallel bands of data as shown in Fig. 2 (b). Each data band has the same unique Polarisation Anomaly along its length and each is populated by data points derived from grains sharing the same refractive index (real component). However, grains populating any one band differ from each other in all other grain parameters (D, d, /, a) over a range of values. Fig. 2(b) illustrates this graphically by identifying each such band with a different colour denoting the refractive index. Thus, by using the Polarisation Anomaly, a variation in grain composition can be identified independently of the otherwise confusing effects of grain size, opacity and, most notably, space weathering. 3.2. Calibration: refractive index from polarisation anomaly Here we use the results of the 2D grain model to derive a simple formula for use in calculating the refractive index of a regolith 159 A. Fearnside et al. / Icarus 268 (2016) 156–171 -0.6 (a) Small Large (c) -0.7 -0.8 log10 (Pmax) -0.9 -1 -1.1 -1.2 -1.3 -1.4 -1.5 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 log10 (B) -0.6 -0.6 (b) -0.7 Low High 2 lines -0.8 -0.9 log10 (Pmax) log10 (Pmax) -0.8 -1 -1.1 -0.9 -1 -1.1 -1.2 -1.2 -1.3 -1.3 -1.4 -1.4 -1.5 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 Low Medium High (d) -0.7 2.8 -1.5 2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 log10 (B) log10 (B) Fig. 3. (a) A small portion of a simulated regolith comprising a loose pack of grains. A single ray path incident upon (from left) and exiting from (to right) the pack is shown. Dashed lines indicate the boundary between a host grain core and its space-weathered shell layer. (b–d) Umov plots for 144 different simulated regoliths spanning different combinations of structural and optical parameters. (b) This Umov plot indicates data points associated with two of the three different values of refractive index of regolith grain (relatively ‘‘low” n = 1.58, and relatively ‘‘high” n = 1.78). Confidence intervals (dashed lines) are shown for each refractive index data group. These define the 2r confidence limits in the least-squares linear regression calculated for the group in question. (c) This Umov plot indicates data points associated with ‘‘small” (D = 2.4 lm) and ‘‘large” (D = 4.8 lm) average grain sizes. A similar effect is achieved by ‘‘small” (a = 3386 m1) or ‘‘large” (a = 5080 m1) absorption coefficient. (d) This Umov plot indicates data points associated with ‘‘low”, ‘‘medium” and ‘‘high” values (/ = 0.005, 0.015, 0.025) of npFe content within the space-weathered layer coating grains. Table 2 Structural and optical parameters defining the simulated regolith grain. Grain parameter Range Core diameter (D) Weathered layer thickness (d) Refractive index (real part, n2) Absorption coefficient (a2 = 4pk2/k) Proportion (/) of npFe 10–200 lm 10–220 nm 1.58–1.78 5080 m1 (k = 500 nm) 0.01–0.03 according to the Polarisation Anomaly it displays on an Umov plot. This formula is applicable to materials with refractive indices n2 in the range: 1:5 6 n2 6 2:0. This range covers the expected range of refractive indices of the lunar regolith. The single scattering albedo (w) of a particle can be defined in terms of its scattering efficiency (QS) and its extinction efficiency (QE) as w = QS/QE. For closelyspaced particles, QE 1 and so w QS. This can be expressed in terms of the reflection coefficient Se for surface reflection of externally incident light, the reflection coefficient Si for surface reflection of internally incident light and the internal transmission factor h for light within the particle, as QS = Se + h(1 Se)(1 Si) (1 Sih)1. As a first approximation, the surface scattered component Se of the reflection coefficient represents wholly polarised light whereas the internally scattered component h(1 Se)(1 Si) (1 Sih)1 represents wholly unpolarised light, depolarised by internal scattering processes. The degree of polarisation, P, of the scattered light may then be reduced to P Se/QS Se/w. Taking log- arithms of this expression gives log10(P) = log10(Se) log10(w). We assume that the albedo (A) of a particulate surface is proportional to the back-scattering activity of the particles forming it. A comparison of the expressions for back-scattering activity (Xx) and the single scattering albedo (wx) of Table 1 reveals a close relationship between the two, in which the latter includes the former together with forward scattering terms. With the benefit of a hindsight knowledge of Umov’s Law, we propose that this relationship can be expressed in the form wx X cx2 , where c2 is a constant. This allows us to write wx Ac2 . Our logarithmic expression then takes the functional form of Umov’s Law: log10(P) = c1 c2log10(A) in which c1 = log10(Se). The Polarisation Anomaly, Dpol, being the perpendicular separation between two parallel such lines, is then obtained by: 1 Sð2Þ Dpol ¼ log eð1Þ b Se where b ¼ ! ð4Þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 þ c22 . The terms SðjÞ e ðj ¼ 1; 2Þ are the surface reflection coefficients for a grain of refractive index mj. For irregular particles with a complex refractive index mj = nj + kj with 1:2 6 nj 6 2:0 and kj 1 following both Hapke (Hapke, 2005) and Shkuratov (Shkuratov, 1981), we may write, approximately: SðjÞ e ¼ b1 ðnj 1Þ2 ðnj þ 1Þ2 þ b2 ð5Þ 160 A. Fearnside et al. / Icarus 268 (2016) 156–171 Hapke suggested that b1 = 1.0 and b2 = 0.05. However, Shkuratov suggests that surface reflection may be refined by setting b1 ¼ b1 g; G; FF0 to be a phase coefficient of surface porosity, connected with the coefficient of volumetric porosity. The term F represents the total area of microsurfaces of a regolith specularly reflecting towards the observer from a given region of area F0. The term G allows for the departure of the Fresnel reflection indicatrix from a d-function due to micro-roughness at grain surfaces. While we do not implement such detailed methods here, we do generalise Eq. (5) such that b1 and b2 are free parameters. By solving Eq. (5) for n2 (i.e. j = 2) and substituting for Sð2Þ using Eq. (4), e one may write an expression for the refractive index of grains in terms of their Polarisation Anomaly (Dpol) relative to an Umov scat^2 : ter line for other grains of a reference refractive index n 1=2 bDpol b1 b2 þSð1Þ þ2b3 b1 Seð1Þ 10bDpol b2 e 10 ^2 Þ ¼ ; n2 ðDpol ;c2 ; n b1 þb2 Seð1Þ 10bDpol qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ^ 2 1Þ ðn 2b1 ^ 2 Þ2 1þb3 g b b3 fðn ð6Þ Sð1Þ þb2 þ 3 e ¼ b1 ^ ðn2 þ1Þ ^ 2 þ1Þ2 ðn Here we have generalised the result by adding a free parameter b3. When b3 = 1 we find Seð1Þ to be defined as Eq. (5), as expected. In order to derive values of b1, b2 and b3 we calculated Dpol for many values of n2 using our 2D model, and then fitted Eq. (6) to the results. The values of b1, b2 and b3 were adjusted to optimise this fit. We found that a good fit only occurs when b3 0 when applied to the range of n2 being considered. We found excellent agreement when n2 was restricted to the range 1:5 6 n2 6 2:0, with b1 = 0.9568, b2 = 0.0178 and b3 = 0. Eq. (6) then takes the succinct form: bDpol b1 b2 þ Sð1Þ e 10 ^2 Þ ¼ ; n2 ðDpol ; c2 ; n bDpol b1 þ b2 Sð1Þ e 10 Sð1Þ e ¼ b1 ^ 2 1Þ ðn þ b2 ^ 2 þ 1Þ ðn ð7Þ ^ 2 associated With knowledge of a reference refractive index n with a reference Umov line of measured gradient c2, one can use ^ 2 Þ of grains associEq. (7) to derive the refractive index n2 ðDpol ; c2 ; n ated with a parallel, second Umov line offset from the first by a Polarisation Anomaly Dpol. This method will be used extensively in Sections 4 and 5 of this paper to derive the refractive index of granular terrestrial and lunar materials. Fig. 5 shows the curve: n2(Dpol) calculated using the 2D grain model for a reference refractive ^ 2 ¼ 1:526. Together with this curve is shown a second index of n curve (solid line) corresponding to Eq. (7). 3.3. Three-dimensional (3D) model The 2D grain model is, of course, limited in its scope not merely because it is only 2D but also because it represents a grain in isolation. In order to overcome these limitations we developed a more realistic 3D multi-grain model of the lunar regolith which does include the effects of grain-to-grain scattering of light. 3.3.1. Model basics Each of the grains in a given multi-grain model regolith was composed of an irregularly-shaped polygonal core of uniform mineral material coated with a uniform space-weathered layer. The refractive indices and absorption coefficients of these polygonal grains were calculated in the same way as was done for the grain of the 2D model. We employed geometrical optics to trace the paths of many light rays propagating through a collection of many such irregular grains representing a regolith. In doing so, we computed the phase curves for the brightness and for the degree of linear polarisation of light scattered by many different types of such regoliths each distinguished by the average size and/or composition of their grains. The state of polarisation of each light ray was represented by its Stokes vector, which was calculated using the methods described by Muinonen et al. (1996). We define the Stokes vector of a ray as M = [I, Q, U, V] in general. Rays of light initially incident upon the regolith are un-polarised so that MI = [1, 0, 0, 0]. At each intercept between a light ray and a regolith grain boundary, the Stokes vectors for the scattered and transmitted components of light rays are calculated using MS = R K MI and MT = T K MI respectively. The matrices R, T and K are the Mueller matrices for reflection, for transmission, and a rotation matrix K = K(u): 2 1 0 0 0 0 0 0 1 3 6 0 cosð2uÞ sinð2uÞ 0 7 7 6 KðuÞ ¼ 6 7 4 0 sinð2uÞ cosð2uÞ 0 5 The rotation matrix allows for a change of scattering plane at each intercept. All rays are initially constrained to a single 2D plane whereby the rotation matrix applies a rotation angle (u) of zero radians. Fig. 3 shows a small part of a typical 2D pack of grains in a simulated regolith, and the 2D path of a single ray which enters the regolith at an incidence angle of 80°. Eventually it escapes back to space having passed through many grains. To approximate the 3D scattering of the light rays, a random rotation ui was applied to each one of j line segments of a light ray between successive grain boundary intercepts, using rotation matrix K(ui), (i = 1, 2, . . .j), piece-wise for the full 2D path taken by a light ray. This technique thereby applied j successive rotations to the segments of the 2D path of a light ray from incidence into the regolith to emergence from it. This enforced a change of scattering plane at each intercept of the ray with a grain boundary while preserving the Fresnel amplitudes previously calculated along the 2D path. The random rotation angles ui were applied up to 20 times (j = 20) along the path segments of each emergent ray path so that a single emergent ray produced 20 distinct 3D ray paths. The phase angle (g) for each 3D ray path was defined as the angle between the incoming and emergent ray. Emergent rays were grouped into bins according to phase angle, with each bin covering 1° of phase angle across the range from 0° to 160°. An average Stokes vector M = [I, Q, U, V] for all the rays in each bin was calculated to determine the average intensity and polarisation for that range of phase angle. qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi The linear polarisation was defined by P ¼ Q 2 þ U 2 =I. The irregular polygonal shapes of the grains within a model regolith were varied randomly and a minimum 0.1 lm gap was imposed between grain vertices, whereas other physical properties were kept constant for a given regolith simulation. Thus, all grains in a given simulated regolith shared the same thickness of spaceweathered outer layer, npFe concentration, average grain size and core refractive index. The upper layers of the lunar regolith are believed to be very loosely arranged and form a so-called ‘fairy-castle’ structure. Accordingly, a fraction of grains were randomly deleted from the grain pack in order to simulate an open structure. Relatively smaller grains populated the upper layers of the model regolith as is expected of the lunar regolith (Papike et al., 1981). Those smaller grains are expected to dominate the optical properties of the regolith surface. Indeed, our simulations confirm that incoming light is quickly absorbed and tends not to penetrate far into the regolith. We simulated 4000 distinct regoliths, and traced 360,000 input light rays through each one of those regoliths. Each simulation started by illuminating a simulated A. Fearnside et al. / Icarus 268 (2016) 156–171 regolith with a set of 10 un-polarised light rays. The paths of the light rays were traced through the regolith until they either emerged from its illuminated surface or were attenuated to an intensity below 1% of their initial intensity. The process was repeated for each new incidence angle up to incidence angles of 80°. To investigate the effect of different grain sizes and properties, 144 different regoliths were simulated representing all of the different combinations of the parameters listed in Table 3. 3.3.2. Umov’s Law Fig. 3 shows three Umov plots derived from our 3D model. In each of Fig. 3(b)–(d), we indicate how different data points are associated with different grain parameters. In particular, Fig. 3(b) shows different data points according to the refractive index of the grain cores of the simulated regolith. It is clear that the 3D model has reproduced the banding structure predicted by the 2D model discussed above. It shows that Polarisation Anomaly is governed by refractive index differences. A variation in other grain parameters of the simulated regolith do not produce this banding pattern. For example, increasing either the average grain size or the npFe content of grain coatings simply moves data points along (to the upper left) the Umov correlation line, not across it. This can be seen in Fig. 3(c) and (d) and is also predicted by our 2D model. 3.3.3. Linear regression and error analysis We have calculated regression lines for each of the two linear groups of data points presented in Fig. 3(b). These are indicated in Fig. 3(b) in terms of their associated confidence intervals. The confidence intervals appear as pairs of lines bounding an associated linear data group. They define a 2r interval of confidence in the least-squares regression line calculated for the data group in question. One can see that the confidence intervals for the two data groups do not overlap. This indicates a significant level of confidence that the single refractive index value assigned to each linear data group can be well represented by a simple linear regression line. While our error analysis employs 1r confidence intervals, these confidence intervals are too small to be visible on the scale of Fig. 3(b) and so we present 2r confidence intervals there for illustrative purposes only. A 1r confidence interval in a linear regression upon either data group of Fig. 3(b) corresponds to an interval in refractive index, as measured from Polarisation Anomaly, of about ±0.0026. 3.3.4. Conclusions The 3D regolith model confirms the relationship between Polarisation Anomaly and refractive index predicted by the simple 2D grain model. Both confirm that Polarisation Anomaly is independent of the effects of grain size variation. This appears to be contrary to the interpretations of Shkuratov and of Dollfus regarding Umov’s Law and the ‘Polarimetric Anomaly Parameter’. Recall that ‘polarisation excess’, ‘polarisation deficit’ and the ‘Polarimetric Anomaly Parameter’ that quantifies them, continue to be interpreted as a measure of variations in the median value of regolith Table 3 Structural and optical properties defining simulated regolith grains. The three absorbtion coefficient values (a = 4pk/k) correspond to wavelengths of 500 lm, 600 lm and 750 lm. A total of 144 different combinations of the above parameter values were used. Grain property Value Mean grain size (lm) Minimum grain separation (lm) Weathered layer thickness (nm) Refractive index (real part, n) Absorption coefficient (a = 4pk/k, m1) Proportion (/) of npFe 0.6, 1.2, 2.4, 4.8 0.1 100 1.58, 1,68, 1.78 5080, 4233, 3386 0.005, 0.01, 0.015, 0.025 161 grain size. We now propose a new interpretation. That is, these effects are all caused by variation in the bulk refractive index of the lunar regions being observed. Regions showing a ‘polarisation deficit’ (Aristarchus Plateau, Marius Hills) do so because they comprise regolith grains having notably lower refractive index. Similarly, regions showing a ‘polarisation excess’ do so because they comprise regolith grains having notably higher refractive index. We shall discuss both matters further below with reference to new polarimetric measurements of the Moon. However, before that, we discuss a direct experimental test of our predictions regarding Polarisation Anomaly. 4. Experiment While two separate models each predicted that Polarisation Anomaly is caused by variation in the refractive index of regolith grains, we resolved to test these theories directly against experiments. 4.1. Goniometer measurements Using a goniometer, polarisation phase curves were determined for each of 17 powdered samples of three different materials of known refractive index (real component, n2). In the goniometer a monochromatic video camera (model DMK21AU04.AS) with focussing lens, was mounted at a fixed angular position relative to each sample. A white LED lamp mounted at the far end of a rotating arm illuminated the sample from a fixed distance at variable angles of illumination. The angle subtended by the lamp arm and the line-of-sight of the video camera defined the phase angle (g) of the illuminated sample. Samples were imaged through a linear polarising filter and a green pass-band filter, having an 80 nm wide pass band centred upon 540 nm. The lamp emissions were found to be un-polarised. A set of still images of each sample was recorded when illuminated by the lamp at a common specified phase angle. Each still image within a set corresponded to a different angular position of the linear polarising filter. The sample brightness (I) at a given phase angle was defined as the average pixel value within a sample image. This changed with changing polarising filter angular position (h), conforming very well to Malus’ Law which takes the form I = Acos2(h + u) + B, in which A and B are constant and u is an arbitrary phase angle. From this the degree of linear polarisation was calculated as P = A/(A + 2B). For each sample material, a sequence of such polarisation measurements was made at different phase angles spanning a phase curve, and from that phase curve the maximum degree of polarisation (Pmax) was determined for that sample. The brightness of a given sample was determined as the measured reflection from the sample at the minimum possible phase angle of the goniometer. This minimum phase was the same in all cases and was chosen to minimise the effects of grain shadowing within a sample. The three sample materials were black soda lime glass, silicon carbide and JSC-1a lunar regolith (mare) simulant (Joy, 2014). Each sample material was sieved to separate the grains into different size ranges as listed in Table 4. Refractive index values for the soda lime glass and the silicon carbide samples were obtained from the literature (Rubin, 1985; Shaffer, 1971), whereas the refractive index for JSC1a was directly measured using the Becke line method (Bowles, 2014). Each such value of refractive index is in respect of green light of wavelength 540 nm. While silicon carbide and soda lime glass are not especially lunar-like, they have been chosen because of their uniform composition, known chemistry and widely differing refractive indices relative to each other. They provide controllable end-member materials with which to test our theoretical models which should 162 A. Fearnside et al. / Icarus 268 (2016) 156–171 Table 4 Sample grain size ranges. Material n2 1st size 2nd size 3rd size 4th size 5th size 6th size Soda lime glass 1.526 455–355 lm 300–250 lm Silicon carbide 2.671 F80 (165 lma) F220 (63 lma) JSC-1a 1.603 355–300 lm 180–150 lm 125–90 lm 7th size 212–180 lm 180–150 lm 150–120 lm 125–90 lm 90–75 lm F400 (22 lma) F600 (15 lma) – – – 90–75 lm 75–45 lmb <45 lmb – a Note: ‘‘F” values for silicon carbide are according to U.S. Department of Commerce Commercial Standard CS 271-65 for which mean, maximum and minimum particle diameters (D) are as follows. F80: Dmean = 165 lm, Dmax = 292 lm, Dmin = 102 lm. F220: Dmean = 63 lm, Dmax = 102 lm, Dmin = 20 lm. F400: Dmean = 22 lm, Dmax = 45 lm, Dmin = 11 lm. F600: Dmean = 15 lm, Dmax = 35 lm, Dmin = 9 lm. b Note: Refractive index of these grains possibly less than 1.603 due to differing composition. Not confirmed. apply to any granular material, not just lunar regolith. Of course, JSC-1a is widely regarded as being a good regolith simulant and is included for comparison. However, it is not clear that the composition (and therefore refractive index) of JSC-1a is uniform as a function of grain size. Below we suggest that it may vary. 4.2. Umov’s Law The value of maximum polarisation and sample brightness was determined for each sample, and the result is plotted as an Umov plot in Fig. 4. Fig. 4 clearly shows the linear trend line associated with Umov’s Law. Regression lines are shown for the data associated with each of the three materials of Table 4. The regression lines for the soda lime glass and for the silicon carbide are separated and approximately parallel. In both of these samples, the data associated with the largest grains occupy the upper left parts of the regression line (lesser brightness, greater polarisation) whereas the data associated with ever smaller grains occupies increasingly the lower right portions (greater brightness, lesser polarisation). These features of the data are fully in accordance with both the 2D and the 3D regolith models (see Figs. 2 and 3). Most notably, the measured transverse separation between the regression lines for soda lime glass and silicon carbide has a mean value of 0.288. With refractive ^ 2 ¼ 1:526 for soda lime glass and n2 = 2.671 for silicon indices of n carbide, at a wavelength of light of 540 nm, the difference between these two refractive indices is 1.145. Our 2D grain model predicts that such a refractive index difference should manifest itself in the form of a Polarisation Anomaly of Dpol = 0.28. This is in close agree- Fig. 4. Umov plot for peak polarisation vs. brightness of all samples of soda-lime glass, silicon carbide and JSC-1a lunar simulant samples of all grain size ranges. Confidence intervals (1r) are indicated for the silicon carbide and soda lime glass data (dashed lines) and the corresponding upper and lower extremes (Dmax, Dmin) of the measurement error interval in Polarisation Anomaly (Dpol). ment with the measured value of Dpol = 0.288. Furthermore, the measured transverse separation between the regression line for soda lime glass and the cluster of data points for the largest four samples of JSC-1a, is Dpol = 0.048. The 2D grain model predicts that this Polarisation Anomaly arises from a refractive index for JSC-1a of n2 = 1.59. This is very close to the independently-measured (Bowles, 2014) refractive index for JSC-1a of n2 = 1.603 for light of wavelength 540 nm. Table 5 summarises the comparisons between actual refractive indices and the predictions of the 2D grain model as applied to our goniometer measurements. Chemical analysis of JSC-1a (Orbitec, 2007) reveals that its main components are plagioclase feldspar and basaltic glass. The plagioclase component has a higher proportion of Al2O3 and a lower proportion of FeO2 than has the basaltic glass component. This will directly influence the refractive indices of these components. We suspect that the smallest grain-sized samples of JSC-1a in our experiments may have been dominated by lower-index plagioclase grains, whereas the larger grain-sized samples may have been dominated by higher-index basaltic glasses. The milling process by which JSC-1a is prepared is likely to preferentially enrich its finest fraction with grains of the relatively more friable plagioclase. This could explain why the samples of JSC-1a having smaller grains appear to have a refractive index comparable to that of our soda lime glass samples. Microscopic inspection of the sample grains revealed that the larger grain sizes contained an elevated proportion of highly vesiculated glass fragments not present in the smaller grain-sized samples. This vesicular structure is likely to be the cause of the clustering of data points associated with the larger grain-size samples in the upper left end of the Umov scatter plot for JSC-1a in Fig. 4. This is because, while the larger grain sizes may possess greater size, they also may possess greater voidage/ vesicles which reduce the effective optical depth/size of those grains, and also increase the probability of outward scattering of light from a grain before that light has fully traversed the grain. The effect may be to make the grains appear optically smaller than they are physically. Fig. 5 shows how the magnitude of the Polarisation Anomaly corresponds with a variation in the real part of the refractive index of a grain modelled according to the 2D and 3D regolith models and measured experimentally (Table 5) for soda lime glass, JSC1a and silicon carbide. The agreement is encouraging, especially across the range 1:5 6 n2 6 2:0. Here, we have chosen a value for ^ 2 ¼ 1:526 from which to deterthe reference refractive index of n mine Polarisation Anomaly and this is representative of the refrac- Table 5 Measured refractive index values and actual values (⁄measured according to the Becke line method). Material Dpol n2 (predicted) Error interval n2 (actual) SiC JSC-1a 0.288 ± 0.015 0.048 ± 0.015 2.74 1.593 2.62–2.94 1.57–1.61 2.671 1.603 ± 0.01⁄ A. Fearnside et al. / Icarus 268 (2016) 156–171 3 2.8 REFRACTIVE INDEX 2.6 2.4 2.2 2 1.8 1.6 1.4 0 0.05 0.1 0.15 0.2 0.25 0.3 POLARISATION ANOMALY Fig. 5. Calibration curves for predicting refractive index values n2 derived from ^ 2 ¼ 1:526. (1) The 2D Polarisation Anomaly relative to a reference refractive index n grain model (dashed line); (2) the analytical Eq. (7) (solid line); (3) the 3D grain model (triangles); (4) measured Polarisation Anomaly for soda lime glass, JSC-1a and silicon carbide (solid dots) at a wavelength of 540 nm. tive index of the soda lime glass powder employed in the present experiments. This is also, of course, close to the value of the refractive index of plagioclase which is abundant in lunar highland terrains. Terrestrial samples of plagioclase are known to have refractive indices which range in value from about 1.485 to 1.60 depending on composition and structure (e.g. fused, thetomorphic or crystalline). Pyroxenes typically range from 1.591 to 1.755 in refractive index. For olivines, the range is from about 1.651 to 1.869, while lunar ilmenite samples are known to have a refractive index (real part) of about 2.58. These values are found from a review of the literature shown in more detail in Table 6, and correspond to optical wavelengths of 550 nm. We use these ranges as a guideline for the likely values of refractive index we might expect to see in lunar regoliths of different mineralogy. This helps us to apply our predictions regarding Polarisation Anomaly to remote observations of the Moon, as we discuss in the following sections of this paper. Preliminary measurements applied to controlled mixtures of silicon carbide grains and soda lime glass grains (refractive index end members) indicate that the Polarisation Anomaly associated with the mixture changes in proportion to the relative contribution of the end members of the mixture. This is seen to occur both when the end members are of similar grain size, or of different grain size. This is in accordance with preliminary theoretical modelling and is a topic of on-going work. Its further discussion is beyond the scope of the present paper. Notably, when controlled mixtures of grains of the same refractive index (e.g. silicon carbide alone) but different grain size are employed (grain size end members), such a mixture produces no Polarisation Anomaly. In that case, each grain size end member is associated with a different position along the linear trend line of the same Umov plot and the corresponding position associated with the mixture moves along (but not away from) that trend line, between the endmember positions, in proportion to the relative contribution of the end members of the mixture. This is as expected from theoretical modelling described above and reinforces the conclusion that refractive index changes induce Polarisation Anomaly, whereas grain size changes alone do not. 163 shown in Fig. 4 for silicon carbide and soda lime glass. The number of data points in the Umov plots for each material is limited simply by the limited number of different sieve sizes available for separating each sample into different grain size groups and this consequently influences the size of the confidence intervals. Each confidence interval represents the confidence bounds for the linear regression line calculated for the data set concerned. We use these extreme limits to estimate error margins in Polarisation Anomaly measured using the regression lines in question. For example, regarding silicon carbide, Fig. 4 shows how the mean value (Dpol = 0.288), upper value (Dmax = 0.303) and lower value (Dmin = 0.273) are measured using the mean, upper and lower confidence interval limits of the regression lines for silicon carbide and soda lime glass. This leads to the estimated Polarisation Anomaly value of Dpol = 0.288 ± 0.015 for silicon carbide relative to soda lime glass. The same procedure was also applied to the Umov plot for JSC-1a at the region of the plot associated with the largest grains. Fig. 5 shows these error margins whereas the estimated error in measuring the refractive index of JSC-1a (Bowles, 2014) was ±0.01 and is too small to be visible. Table 5 indicates the estimated error values in Polarisation Anomaly for silicon carbide and JSC-1a, and the resulting error interval in the corresponding value of refractive index when calculated by applying the measured Polarisation Anomaly to Eq. (7). 5. Observations of the Moon In this section we describe polarimetric observations of parts of the lunar surface. We show how we have identified Polarisation Anomalies within these observations and how we have converted these Polarisation Anomalies into maps of refractive index across the lunar surface. The method of polarimetric observation is the same, in its basic principle, as the method described above regarding our goniometer measurements. Of course, in our lunar observations we have replaced the goniometer grain samples by the lunar surface itself, and the imaging camera lens is replaced by a telescope as described below. 5.1. Equipment Schmidt-Cassegrain or Maksutov-Cassegrain telescopes were used, with mirror diameters of 20 cm or 25 cm. These axisymmetric instruments have no off-axis reflections that might affect the measured polarisation. Monochromatic CCD video cameras of two types were used to record data. They were, the DMK21AU04.AS camera containing a Sony ICX098BL CCD chip, and the Lumenera Infinity-2 camera containing a Sony ICX274 CCD chip. We found that each of the CCD cameras has a sensitivity to linearly polarised light depending on the orientation of the polarisation axis relative to the CCD chip. The physical reason for this sensitivity is not clear. To resolve this problem, we calibrated our cameras against an un-polarised light source and derived a correction factor which we applied not only to the telescopic observations, but also to the goniometer measurements described above. The polarising filters were optically-polished glass polarisers mounted to be rotated to calibrated angular positions relative to the telescope/camera. Video sequences were recorded in visible light through an infra-red rejection filter with a transmission range between 400 and 710 nm to cut out un-focussed near infra-red light. 4.3. Error analysis 5.2. Observational technique and data acquisition To estimate error margins we calculated 1r confidence intervals for each linear regression line shown in Fig. 4. Examples are We recorded 21 video sequences of a selected region of the Moon imaged telescopically through a linear polarising filter when 164 A. Fearnside et al. / Icarus 268 (2016) 156–171 Table 6 Refractive index values for terrestrial and lunar materials. Fused Plagioclase a c d e f g h j k l m o p q s t u v w x y b Crystalline c Mixed grains w 1.523 + 0.0227(An#) + 0.0264(An#)2 1.562–1.581 1.51–1.585d – – 1.57 1.58e 1.569–1.592f 1.535–1.60g (Diopside) (Pigeonite) (Augite) (Titanian Augite) (Hedenbergite) 1.61–1.68h 1.683 j 1.61–1.69l 1.591–1.749f – – – – – – – – – – – – – – – – 1.67c 1.681–1.693k 1.685–1.703m 1.65–1.71o 1.715–1.755p 1.675–1.692q 1.684–1.714q 1.685q 1.73q 1.753q – – – – – – – – Olivine (Forsterite) (Chrysolite) (Hyalesiderite) (Hortonolite) (Ferrohortonolite) (Fayalite) 1.758s – – – – – – – – – – – – – 1.78c 1.651t 1.680t 1.733t 1.786t 1.828t 1.869t – – – – – – Lunar ilmenite Lunar simulant Lunar pyroclast Lunar sample Integrated Moon – – – – – – – – – – – – – – – 2.58u 1.603v 1.569x 1.67y 1.78y Pyroxene b 1.485–1.580 – – – Thetomorphic a – – – w 1.768 (0.118 ⁄ (Mg#)) 1.726 (0.082 ⁄ (Mg#)) w w 1. 827 (0.192 ⁄ (Mg#)) (Stoffler and Hornemann, 1972). Range given corresponds to 0 wt.% Anorthite (lower) to 100 wt.% Anorthite (upper). Wavelength is 589 nm. (Engelhardt et al., 1970). Increasing from 0% An to 100% An. (Hiroi et al., 2009). (Stoffler and Hornemann, 1972). Average values for variation in anorthite content from 0% An to 100% An. Wavelength is 589 nm. Bytownite. From Egan and Hilgemann (1979). Wavelength is 550 nm. (Chao et al., 1971; Chao et al., 1970). (Chayes, 1952). Anorthite content varies from 0% to 100%. Assumed wavelength is 589 nm. (Segnit, 1953). For wt.% of Fe2O3 increasing from 0% to 20%. (Dorschner et al., 1995) (Mg0.5Fe0.5SiO3). Generally, the formula is MgxFe1xSiO3 in which X is varied from 1.0 to 0.4, from which refractive index is (approx.) n = 1.8–0.23X. See (h). For wt.% of TiO2 content increasing from 0% to 8%. See (h). For wt.% of TiO2 content increasing from 0% to 8%. See (h). For wt.% of Fe2O3 increasing from 0% to 8%. (Binns, 1970). For (Fe + Mn)/(Fe + Mn + Mg) atom % increasing from 0% to 30%. (Mason, 1974). Averaged over a, b and c indices. US Geological Bulletin 1627. (Dorschner et al., 1995) (MgFeSiO4). Generally, the formula is Mg2xFe22xSiO3 in which X is 0.5. US Geological Bulletin 1627, for series (Mg,Fe)2SiO4. Lunar sample 45.35.5 (Egan and Hilgemann, 1979). Lunar regolith simulant JSC-1a (Bowles, 2014). (Lucey, 1998; Warell and Davidsson, 2010; Carli et al., 2014). Mg# is the magnesium number Mg/(Mg + Fe), and An# is the anorthite number Ca/(Ca + Na). (Wilcox et al., 2006). Aristarchus Plateau pyroclastic deposits. (Hapke, 1994). Lunar Apollo sample 12070. the lunar phase angle was between 60° and 130°. Each video sequence was recorded at one of 21 different orientations of the transmission axis of the polarising filter. After a video sequence was recorded, the polarising filter was rotated (relative to the camera and Moon) to its next position, and another video sequence recorded. This process was repeated until all 21 video sequences were recorded within 15–20 min, to minimise the effects of the change of illumination angle and atmospheric transparency. Video sequences were recorded only when the sky was judged to be transparent and when the Moon was at an altitude of at least 30° to mitigate the effects of atmospheric turbulence and extinction nearer the horizon. Several hundred video frames were recorded in each video sequence, and individual frames were analysed to assess their quality. Those of poor quality were rejected. Using software, the remaining frames of a given video sequence were co-aligned and stacked to produce a single stacked image of increased signal-to-noise ratio. The stacking process also averaged out much of the effect of any poor seeing conditions. The lunar coordinates of the sub-solar and sub-Earth points at the time of observation, and the pixel coordinates, latitudes and longitudes of reference points/craters in each stacked image, were used to convert image brightness into equigonal brightness using the method described by Velikodsky et al. (2011). Equigonal brightness has the effect of suppressing variations in surface brightness arising solely due to solar illumination angle above the local horizon. 5.3. Data analysis 5.3.1. Data reduction All of the 21 stacked images were assigned a common image coordinate system. For each pixel in this common coordinate system the variation in pixel value through each of the 21 stacked images (i.e. the 21 polarising filter angular positions) was fitted to Malus’ Law as described above in relation to our goniometer experiments. The degree of linear polarisation (P) associated with each stacked image pixel was individually calculated as P = A/(A + 2B) using that law. An equigonal brightness map was converted into a polarisation map by treating each image pixel in this way. 165 A. Fearnside et al. / Icarus 268 (2016) 156–171 5.3.2. Error analysis The accuracy of our measurements of the degree of linear polarisation rests upon the confidence with which we can say that the variation of image pixel brightness (I) follows Malus’ Law: I = Acos2(h + u) + B, as the polarising filter angular position (h) changes. A least-squares fit of Malus’ Law to observational data (I, h), requires optimisation of the fitting parameters A, B and u. This is achieved by solving the matrix equation: I = M [h] for the optimal least-squares value of the vector M given the input vector of equigonal brightness values: I = [I1, I2, . . .In] for a given image pixel location across the full sequence of n different polarising filter angular positions (h). Here, n = 21. The matrix [h] contains the position angle data: 2 cosð2h1 Þ cosð2hn Þ 6 ½h ¼ 4 sinð2h1 Þ 1 3 7 sinð2hn Þ 5 1 The vector M contains all of the variable fitting parameters A, B and u and is given by: M ¼ ½M1 ; M 2 ; M 3 ¼ A A A ; ; þB 2 cosð2uÞ 2 sinð2uÞ 2 Solving for the vector M to obtain the optimal least-squares values for each of its vector components gives fitting parameter values as: 1 2 u ¼ arctan M2 2M 1 A ; A¼ ; B ¼ M3 2 M1 cosð2uÞ Upper and lower bounding values are calculated for each of these fitting parameters which correspond with a 1r confidence level. We then identify the combination of these parameters required to give the maximum deviation of resulting polarisation value P = A/(A + 2B), relative to the optimal value. This range in polarisation deviation is used as our 1r confidence interval for polarisation measurements. Fig. 6(a) shows an example of a data set of 21 (n = 21) brightness values corresponding to an image pixel position corresponding to a location in Oceanus Procellarum west of the Aristarchus Plateau. The optimal least-squares fit to Malus’ Law is shown together with upper and lower bounding values according to a 1r confidence interval. The mean error as a fraction of amplitude of the fitted curve is 0.015. That curve and mean error correspond to an (a) estimated polarisation of 0.165 ± 0.0034 which ultimately results in a measured refractive index of 1.75 ± 0.01. To make a final estimate of error margins for measured refractive index values we take account of the effect of the confidence interval we have calculated for the theoretical Umov plots we have generated using our 3D grain model (see Fig. 3(b)). This particular confidence interval quantifies the level of confidence we can have in asserting that the data points within an Umov plot do indeed correlate to a given linear correlation trend line. In Section 5.4 of this paper we describe how a linear correlation trend line is applied to observational data in an Umov plot in order to measure refractive index and generate refractive index maps for the lunar surface. In doing this we assume that the confidence intervals calculated for our 3D grain model apply equally to observational data. When this is combined with error estimates for the polarisation values calculated using Malus’ Law, as described above, we find that error margins in measured refractive index values for the lunar surface range from about ±0.01 to about ±0.02, with the typical error margin being ±0.015. Fig. 6(b) illustrates the distribution in the size (dn) of the error margin (±½dn) for all points in the refractive index map of Fig. 8(a) covering northern Oceanus Procellarum including the Aristarchus Plateau. This general distribution of errors in the refractive index measurements is typical of each of the refractive index maps shown in this paper. As a general rule, slightly higher error margins of about ±0.02 in refractive index occur in the brightest regions of the lunar surface within the mapped region, where polarisation values are lowest. The lowest error margins, of about ±0.01 in refractive index occur at the darkest regions where polarisation is highest. In the next section of this paper, we present polarimetric data obtained in this way for regions of the Moon found by Shkuratov to produce a ‘polarisation deficit’, as well as regions showing a ‘polarisation excess’. We apply our interpretation of Polarisation Anomaly in Umov’s Law to generate maps of the refractive index of the regolith of these regions. We discuss how these refractive index maps are consistent with independent estimates of the mineralogy of these regions. 5.4. Observations and interpretation Fig. 7 shows the results of lunar polarimetry in terms of Umov plots for data associated with the following three regions of the Moon: Aristarchus Plateau, Harbinger Mountains and adjacent (b) Fig. 6. Curve fitting of Malus’ Law to observational data. (a) Observational data (data points) of equigonal brightness of a location on the lunar surface for 21 successive polaroid angular positions, and the optimal least-squares fitted curve (solid line) together with 1r upper and lower confidence intervals (dashed lines). (b) Distribution in the size (dn) of the error margin (±½dn) for all points in the refractive index map of Fig. 8(a) covering northern Oceanus Procellarum including the Aristarchus Plateau. 166 A. Fearnside et al. / Icarus 268 (2016) 156–171 1.6 -0.7 (a) (d) 1.4 log10 (P) -0.8 1.2 1 -0.9 0.8 0.6 -1 0.4 -1.1 0.2 4.4 4.5 4.6 4.7 0 4.8 log10 (B) (b) -0.7 (e) 2 -0.8 log10 (P) 1.5 -0.9 -1 1 -1.1 0.5 -1.2 4.4 4.5 4.6 4.7 4.8 4.9 5 0 log10 (B) 2.5 (c) -0.8 (f) 2 log10 (P) -0.9 1.5 -1 -1.1 1 -1.2 0.5 -1.3 4.4 4.6 4.8 5 0 log10 (B) Fig. 7. Regions of interest (left images) and corresponding Umov plots (right images). The regions of interest are: (a) Aristarchus Plateau, Harbinger Mountains and adjacent mare. (b) The Marius Hills and surrounding mare. (c) Mare Crisium and surrounding terrain. The colour side bar shows data density on a log scale and each Umov plot ^ 2 ¼ 1:68 ((d) and (e)) or n ^ 2 ¼ 1:60 (f). (For interpretation of the references to colour in includes a reference line assumed to be associated with material of refractive index n this figure legend, the reader is referred to the web version of this article.) mare (Fig. 7(a) and (d)); the Marius Hills and surrounding mare (Fig. 7(b) and (e)); Mare Crisium and surrounding terrain (Fig. 7 (c) and (f)). In each of Fig. 7(a)–(c) a specific region, or regions, of interest (ROI) is indicated by a white box. As other workers have noted, Umov’s Law is obeyed well not only by the peak value of polarisation (Pmax) but also by polarisation values at lunar phases close to the phase angle at which Pmax occurs. This is the case for the results shown in Fig. 7. The value of polarisation and the corresponding equigonal brightness, are calculated for each image pixel within a given ROI, and this information is plotted in the adjacent Umov plot of Fig. 7(d)–(f), respectively. The Umov plot of Fig. 7(d) corresponding to Aristarchus Plateau, the Harbinger Mountains and mare south of the Plateau shows two well-defined parallel linear data correlations. Compare and contrast this basic structure to the prediction shown in Fig. 3(b). Each correlation takes the form log10(P) = c1 0.97log10(B) as indicated by a reference line in the Umov plot where the parameters P and B are the degree of polarisation and the equigonal brightness, respectively. The term c1 is a constant which differs as between the upper and lower data correlations. The upper correlation, along which the indicated reference line passes, is associated with the mare regions of both ROIs, whereas the lower correlation is associated with Aristarchus Plateau and the Harbinger Mountains. We now derive a refractive index map using this Umov plot as follows. ^ 2 Þ) the measured common graWe apply to Eq. (7) (i.e. n2 ðDpol ; c2 ; n dient (c2 = 0.97), the assumed value of reference refractive index ^ 2 ¼ 1:68Þ for the Umov data falling upon the reference line, and ðn the measured Polarisation Anomaly of each data point in the Umov plot relative to the reference line. From this we calculate the refrac^ 2 Þ for the region as a whole as shown in Fig. 8 tive index n2 ðDpol ; c2 ; n (a). One can see that the refractive index of Aristarchus Plateau and the Harbinger Mountains is significantly lower than that associated with the embaying mare regions and takes values typically in the range 1:55 6 n2 6 1:6. This is consistent with independent estimates (Wilcox et al., 2006) which suggest that n2 = 1.569 for 167 A. Fearnside et al. / Icarus 268 (2016) 156–171 (a) (b) 1.75 1.7 1.65 1.6 1.55 1.5 1.45 1.4 Fig. 8. Refractive index map of (a) Aristarchus Plateau, Harbinger Mountains and surrounding mare; (b) Marius Hills, and surrounding mare and western highlands. Map coverage corresponds directly to Fig. 7(a) and (b). The colour side bar shows refractive index on a linear scale. Refractive index is derived from the Polarisation Anomaly ^ 2 ¼ 1:68. (For interpretation of the references to within the Umov plot of Fig. 7(d) or (e), relative to a reference line assumed to be associated with material of refractive index n colour in this figure legend, the reader is referred to the web version of this article.) pyroclastic glasses assumed to blanket these regions. Also, we note how highland terrains on the western limb are of low refractive index (1:50 6 n2 6 1:55) consistent with plagioclase, whereas the mare regions have higher refractive index (1:65 6 n2 6 1:80) consistent with pyroxenes and olivines (Shkuratov et al., 2005). Fig. 7(b) shows an ROI encompassing Marius crater, the Marius Hills, the surrounding mare, and the highland terrains extending to the western limb of the Moon. The Umov plot derived from this ROI is shown in Fig. 7(e). It shows two well-defined linear correlations each parallel to the reference line: log10(P) = c1 1.18log10(B). The data defining the upper correlation is associated with the embaying mare regions of this area, whereas the data defining the lower correlation is associated with the Marius Hills and the highland terrains. As before, we apply to Eq. (7) the measured common gradient (c2 = 1.18), the assumed value of reference refractive index ^ 2 ¼ 1:68Þ for the data falling upon the reference line, and the ðn measured Polarisation Anomaly of each data point in the Umov plot relative to the reference line. The refractive index map for the region as a whole as shown in Fig. 8(b). One can see that the refractive index of the Marius Hills is significantly lower than that associated with the embaying mare regions and the floor of Marius crater. This region comprises materials of refractive index typically in the range 1:5 6 n2 6 1:6. Fig. 9 shows an enhanced version of the refractive index map of Fig. 8(b) projected onto a surface relief map of the terrain of the Marius Hills. Local surface slope (illumination angle) variations are calculated using lunar surface topographic data made publicly available from NASA’s Lunar Orbiter Laser Altimeter (LOLA). A more accurate equigonal brightness of the Marius Hills was calculated accordingly. This was used in generating an associated Umov plot (see Fig. 7(e)). The Polarisation Anomaly and resulting refractive indices across the Marius Hills are consequently more accurate. Two independent studies based upon reflection spectra of this region have both indicated that it incorporates elevated levels of anorthositic materials (e.g. plagioclase) (Lehman et al., 2013) or pyroxenes of high calcium content (Besse et al., 2011). Either material can be expected to possess a reduced refractive index as compared to embaying mafic mare basalts in this region. The refractive index map of Fig. 9 is consistent with these studies in that regard. We note that a clear boundary is visible at the northern and eastern edges to the Marius Hills complex at which the refractive index of the regolith falls from relatively high values of about 1.65–1.80 in the embaying mare, down to relatively low values predominantly in the range 1.55–1.60 on the complex itself. This boundary coincides well with the geographical limits of the plateau 1.75 1.7 1.65 Marius 1.6 1.55 1.5 1.45 1.4 Fig. 9. The refractive index map of the Marius Hills and immediately surrounding mare, corresponding to Fig. 8(b). The local terrain relief is shown (with vertical scale exaggerated) and has been taken into account when calculating equigonal brightness for the Umov plot used to measure refractive index from Polarisation Anomaly. The colour side bar shows the refractive index linear scale. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) 168 A. Fearnside et al. / Icarus 268 (2016) 156–171 containing the Marius Hills complex as defined in Fig. 5 of Besse (Besse et al., 2011) who have studied the compositional variability of this region in detail using data collected by the Moon Mineralogy Mapper (M3). Fig. 6 of Besse presents a map of multi-spectral reflectance for the region which indicates that the materials of the plateau possess a stronger 950 nm absorption band than do the embaying basalt flows of Oceanus Procellarum. This was interpreted as possible evidence of high-calcium pyroxene and/or variation in plagioclase content within the plateau. Referring to Fig. 9 of the present paper, we note that refractive index values of 1.65–1.80 outside the plateau coincide with those one may expect of pyroxenes and olivines, whereas refractive indices within the plateau as low as 1.55 are possibly too low to be associated with pyroxenes, even those of high calcium content such as diopside (see Table 6). One might associate such low refractive index values with plagioclase. It is interesting to note that the lower refractive index values seen on the Marius Hills complex are generally associated with the locations of volcanic domes, especially those to the south-west and north of Marius crater, which may be the source of felsic basalts according to some current petrogenic theories. It is worth comparing the refractive index maps of Fig. 8 with the map of ‘Polarimetric Anomaly Parameter’ presented in Fig. 6 of Shkuratov et al. (2007). It is straight-forward to show that Eq. (7) above may be re-cast in terms of variation in the ‘Polarimetric Anomaly Parameter’ (Pmax)aA = d in place of the Polarisation Anomaly, Dpol, to permit a refractive index map to be recovered from a map of ‘Polarimetric Anomaly Parameter’ as follows. b1 b2 þ Seð1Þ R1=a ^ 1Þ ðn dð2Þ ^2 Þ ¼ ; Seð1Þ ¼ b1 2 n2 ðR; a; n þ b2 ; R ¼ ð1Þ ð1Þ 1=a ^ ðn2 þ 1Þ d b1 þ b2 Se R ð8Þ Here, d(1) is a reference ‘Polarimetric Anomaly Parameter’ associated ^ 2 , and with regolith of an assumed reference refractive index value n d(2) is a value of ‘Polarimetric Anomaly Parameter’ that deviates from the reference value d(1) and is associated with a regolith of refractive index value n(2). A similar pattern is seen in the Umov plot for Mare Crisium. Fig. 7(c) and (f) shows the ROI of Mare Crisium and its associated Umov plot. The data correlation is split into two separate regions extending parallel to an indicated reference line defined by log10(P) = c1 0.79(B). The grouping of data lying on the trend line, to the upper left of the plot, is associated with the surface (a) of Mare Crisium itself. The grouping of data points below the trend line, to the lower right of the plot, is associated with the terrains surrounding the mare. The refractive index map of Fig. 10(a) was obtained by applying Eq. (7) to calculate the refrac^ 2 Þ for the region as a whole. Note that a tive index n2 ðDpol ; c2 ; n ^ 2 ¼ 1:60 was assumed for the average refractive index value of n of the surface of Mare Crisium, and this lower reference value was used due to the relatively high Al/Si concentration ratios across Mare Crisium (0.35–0.45) as compared to other maria. This is indicative of a less mafic surface composition in Mare Crisium (Taylor, 1975). However, the composition of the basalts of Mare Crisium is more mafic than the materials of the surrounding terrains which would tend to contain relatively more plagioclase of lower refractive index (see Table 4). This is reflected in the refractive index maps of Fig. 10. We also see a slightly elevated refractive index across Palus Somni, immediately west of Mare Crisium, as compared to highland terrains east and south of Mare Crisium. This is consistent with previous works which suggest that the plains north of crater Taruntius, and in Palus Somni, may be volcanic units emplaced as fluid flows with a surface composition intermediate between mare basalts and highland materials. Schonfeld (1981) has used Mg/Al ratio maps to show that this region has a Mg/Al ratio similar to that found in mare regions (Hawke et al., 1985). A higher Mg/Al ratio should be expected to be associated with a higher magnesium number and, therefore, a higher refractive index. In Fig. 10(a) the floors of Picard and Peirce craters show notably lower refractive index. Conversely, a region of elevated refractive index appears between Picard crater and Curtis crater to the east of Picard. Earlier studies (Andre et al., 1978) showed that this region comprises basalts of elevated Mg/Si ratio presumed to come from more mafic material underlying the surface of Crisium, and exhumed by the impact that formed Picard crater. Curtis crater itself defines a striking region of unusually high refractive index (Lat. = 14.57°, Long. = 56.79°) in Fig. 10(a). The circularly symmetrical shape of the refractive index enhancement at Curtis crater suggests that it may be associated with an impact ejecta pattern. We suggest that Curtis crater may be located in a stratigraphic layer of Mare Crisium which overlies a layer of material with a higher refractive index that was exhumed by the impact that produced Curtis. We have found a number of similar examples of elevated refractive index centred upon the location of small impact craters within other mare regions of the Moon. We believe these are examples of the ‘polarisation excess’ referred to by Shkuratov. (b) 1.75 Peirce 1.7 1.65 1.6 Proclus Picard 1.55 1.5 1.45 1.4 Fig. 10. A refractive index map for Mare Crisium. (a) Mare Crisium and surrounding highlands. (b) Western Mare Crisium including craters Proclus, Picard and Peirce. (b) local terrain relief is shown and has been taken into account when calculating equigonal brightness for an Umov plot, and thence refractive index from Polarisation Anomaly. The colour side bar shows the refractive index linear scale. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) A. Fearnside et al. / Icarus 268 (2016) 156–171 Fig. 10(b) shows an enhanced version of the refractive index map of Fig. 10(a) projected onto a surface relief map of a swath of terrain. By taking account of local surface slope (illumination angle) variations, using lunar surface topographic data made publicly available from NASA’s Lunar Orbiter Laser Altimeter (LOLA), a more accurate equigonal brightness of the swath of terrain from Proclus crater to (and beyond) Peirce and Picard craters was calculated. This was used in generating the Umov plot (Fig. 7(f)). The Polarisation Anomaly and resulting refractive indices across the swath are consequently more accurate. Over-bright Sun-facing slopes cause some underestimation of refractive index, if not corrected in this way. The converse is true of slopes facing away from the Sun. Note that the dark blue crescent upon the northern inner wall of Proclus crater (Fig. 10(b)) is a consequence of CCD camera saturation due to very high surface brightness of that Sun-facing slope, and should be ignored. One can see immediately from Fig. 10(b), that the refractive index for the floor regions of Picard crater (Lat. = 14.6°, Long. = 54.7°) and of Peirce crater (Lat. = 18.3°, Long. = 53.4°) are each markedly low against the higher refractive index of the surrounding mare surface. This indicates that the floors of these craters each contain highland-type materials of relatively low refractive index. Spectral studies (Head et al., 1978) and X-ray data indicate that the floors of these craters are composed of a material of elevated Al/Si ratio which underlies Mare Crisium, implying that both of these craters have exhumed underlying mantle material. Elevated Al levels in minerals and glasses are generally accompanied by a reduced refractive index value. Interestingly, this map clearly shows a north/south asymmetry in the composition of the ejecta material of Proclus crater. Specifically, the northern portion of the ejecta pattern is rather high in refractive index, somewhat like that of Mare Crisium, whereas the southern portion of the ejecta is somewhat lower and comparable to that of Palus Somni. This asymmetry is faintly indicated (though not acknowledged) in the refractive index map derived by Henry et al. (1976) in the ultra-violet optical range. There, it can be seen that only the northern ejecta pattern of Proclus crater displays a refractive index comparable to that of Mare Crisium. More recently, ChangE2 orbital data has been used to generate maps of potassium abundance in and around Mare Crisium (Zhu et al., 2013) which clearly show elevated abundances in isolation both immediately to the north of Proclus (coincident with its northern rays) at levels comparable to mare abundances, and immediately surrounding Picard crater. As has been suggested by others, Proclus crater appears to have exposed an anomalous soil type (Bielefeld et al., 1978). 5.5. Factors influencing regolith refractive index The value of refractive index of lunar minerals is dependent upon the variation of their composition (Carli et al., 2014), as discussed by Lucey (1998) for mafic minerals and by Warell and Davidsson (2010) for plagioclases. This is summarised in Table 7. Furthermore, the refractive indices of lunar impact glasses, which comprise a significant component of the regolith, have been found to vary in value from n2 = 1.58 to n2 = 1.75 (Chao et al., 1970, 1972) in direct relation to the proportion of FeO and TiO2 present within the glass. Table 7 Mg# is the magnesium number defined as the molar ratio Mg# = Mg/(Mg + Fe), and An# is the anorthite number defined as the molar ratio An# = Ca/(Ca + Na). Mineral Refractive index Orthopyroxene Clinopyroxene Olivine Plagioclase n2 = 1.768 (0.118 ⁄ (Mg#)) n2 = 1.726 (0.082 ⁄ (Mg#)) n2 = 1. 827 (0.192 ⁄ (Mg#)) n2 = 1.523 + 0.0227(An#) + 0.0264(An#)2 169 5.5.1. Stratigraphy We see in our refractive index maps that some small craters are associated with patches of elevated refractive index. Examples of this include Curtis crater in Mare Crisium and Reiner K in Oceanus Procellarum. These craters may have exhumed material from a buried basalt layer of higher FeO content than the surface mare basalts. If so, the ejecta blanket surrounding these craters would have a higher FeO content than the local mare surface. Given that the magnesium number (Mg# = Mg/(Mg + Fe)) of this excavated basalt material would be lower than that of the local surface basalts then one might expect the refractive index of any pyroxene and/or olivine grains and impact-induced glass within the ejecta blanket to be elevated. 5.5.2. Impact-induced glasses Impact processes may also contribute to the observed enhancements of refractive index due to impact-induced glass components of the regolith. It has been observed that around 50% of the material in returned lunar soil samples consists of glasses (Ryder, 1981). Diverse melts tend to be homogenised and the resulting glasses represent homogenised mixtures of the original target lithologies (Masaytis et al., 1975; Grieve, 1981). Thus, whole rock impact melts can be expected to be simple mixtures of pre-existing target lithologies in terms of composition. Model calculations have shown that extensive melting occurs in a typical lunar impact (O’Keefe and Ahrens, 1975), and melt masses of about 100 times the impactor mass may be produced. However, studies have shown (Schaal and Horz, 1977) that different lithologies respond to high impact shock pressures in different ways. A prominent feature of shock pressures is the formation of shock-metamorphic glasses (also known as diaplectic glass, or thetomorphic glass) produced by in situ conversion of crystal into amorphous glass of the same composition. No melting is involved. Shock-metamorphic glasses are only known to occur in quartz and feldspars, such as plagioclase. They possess densities and refractive indices intermediate between those of the crystalline form of the target rock and the impact glasses produced by melting the target rock. 5.5.3. Grain composition Studies of lunar soil have shown that the composition of the regolith changes with grain size. An important finding is that, when compared to coarser grains, the smallest fraction of grains within mare soils are consistently richer in plagioclase, and are depleted in elements (e.g. Fe, Mg) associated with ferromagnesian minerals (Evenson et al., 1974; Korotev, 1976; Laul et al., 1987; Papike et al., 1981; Taylor et al., 2003). A hypothesis that may account for this is that a greater proportion of plagioclase from disaggregated mare basalt is concentrated in the finer grain size fraction (Korotev, 1976; Papike et al., 1982; Laul et al., 1987). Laboratory studies of impacts into basalt rocks have shown that a preferential enrichment of plagioclase occurs in the finest grain fraction of the disaggregated basalt particles produced by impact shock. Plagioclase preferentially enriches the smallest grain fraction because this mineral tends to shatter into a finer grain size than do other minerals common in basalt rocks, such as pyroxene in particular (Horz et al., 1984). An examination of the compositions of the impact-induced grain size fractions of lunar soil samples has revealed that the abundance of plagioclase remains relatively constant or slightly increasing with decreasing grain size. However, the abundance of pyroxenes and olivine decreases with decreasing grain size. Thus, a relative enrichment of finestgrained plagioclase occurs as against a relative decrease of fine-grained pyroxenes (Taylor et al., 2003). This differential comminution of plagioclase during regolith formation would result in a gradual evolution of the bulk refractive index of the regolith. When it is an immature soil, such as fresh crater ejecta, the regolith 170 A. Fearnside et al. / Icarus 268 (2016) 156–171 should possess a larger average grain size and relatively little plagioclase grain enrichment. As time passes those large grains are slowly disaggregated by impact processes into ever finer grains which increasingly enrich the plagioclase component of the ejecta at the expense of the pyroxene grain content. The net effect, as young crater ejecta matures, would be to reduce its bulk refractive index from an initially relatively higher value approximating that associated with pyroxenes, to a lower value closer to that associated with plagioclase. The most optically relevant portion of the regolith, for the purposes of remote sensing, is the finest grain fraction in the uppermost surface of the regolith which tends to coat the larger grains and presents the greater optical cross section. Thus, the evolution of the composition of this finest grain fraction will determine the evolution of the bulk optical refractive index observed remotely. The mechanism of differential comminution would result in a local relative refractive index increase in the fresh, large-grained and immature soils of young impact craters, albeit temporarily. This would also be consistent with the prior interpretations of the ‘Polarimetric Anomaly Parameter’ (Shkuratov, 1981; Dollfus, 1998; Shkuratov and Opanasenko, 1992; Geake and Dollfus, 1986; Novikov et al., 1982) and the ‘polarisation excess’ being found in fresh bright craters and caused by anomalously large median grain sizes. We suggest that the ‘polarisation excess’ observed in those cases was not a consequence of median grain sizes being anomalously large. Instead, it may have been due to the fact that the anomalously large grains in question possess a greater fraction of pyroxene which has a relatively higher refractive index, and a lesser fraction of fine-grained plagioclase with a relatively lower refractive index. The effect would be an anomalously higher bulk refractive index. 5.5.4. Summary One may conclude that a key factor controlling the value of the bulk refractive index of the uppermost surface of the lunar regolith is the relative contribution of: crystalline pyroxene grains; crystalline plagioclase grains; plagioclase shock-metamorphic glass; plagioclase quenched glass; and, quenched whole-rock melt glass. Immature regoliths are expected to be relatively abundant in the larger-grained first factor, producing an enhancement in refractive index. Increasing maturity enriches the regolith with each of the other smaller-grained factors, which each reduce the refractive index. Thus, we expect refractive index enrichment in the ejecta blankets of young mare craters to fade over time due to this impact mechanism which should also reduce median grain size. 6. Conclusion The linear correlation between a peak value in the polarisation of moonlight reflected from a lunar surface region and its associated albedo has been reproduced by each of two different mathematical models which represent the salient features of lunar regolith grains. Both models predict that variations in grain properties such as grain diameter, bulk opacity and extent of surface space weathering each serve to generate the linearity of the correlation enshrined by Umov’s Law. However, both models also predict that variation in the value of the real component of the refractive index of grains has the effect of inducing a transverse shift in the linear correlation. We refer to this as Polarisation Anomaly. This means that Polarisation Anomaly varies on the basis of variations in grain refractive index, but is independent of the effects of variation in grain size, opacity and space weathering. This could provide a powerful new tool for investigating the chemistry of the lunar regolith which mitigates the obscuring effects of space weathering (Noble et al., 2007; Taylor et al., 2001a,b) and grain size variations. Polarimetric studies of different selected terrestrial materials and lunar simulant JSC-1a have revealed good agreement with the theoretical predictions of Polarisation Anomaly. Reinforced by the agreement between theoretical predictions and experimental observations, we have applied this new interpretation of Polarisation Anomaly to new telescopic observations of selected sites and regions on the Moon in an attempt to determine the location and extent of variations in the bulk refractive index of the grains of the lunar regolith there. The selected sites were chosen in view of their known mineralogical variety that, in turn, can be expected to display a corresponding variety in refractive index. We have indeed found strong evidence that this is the case and that Polarisation Anomaly provides a means for measuring the variation in the refractive index of regolith grains. We propose that the existing interpretation of the ‘Polarimetric Anomaly Parameter’, as being due to the presence of anomalously large regolith grains, may require a subtle revision. This revision is that Polarisation Anomaly is not caused by an anomalous size of regolith grains per se, but is instead caused by those grains having an anomalous mineralogy associated with a higher (or in some cases, lower) refractive index relative to their surroundings. Minerals of higher refractive index may break into fragments that tend to be larger than those formed by the breakage of minerals of lower refractive index. This mechanism may explain why the ‘Polarimetric Anomaly Parameter’ has previously been associated with grain size variation at fresh lunar craters. Acknowledgments We wish to acknowledge the kind support of Dr. K. Joy of the University of Manchester in providing a sample of lunar simulant JSC-1a, and of Dr. N. Bowles of the University of Oxford for measuring the refractive index of grains of this material. Our thanks go to Chris Dudman for his help with the preparation of samples used in the experiments described above. The JSC-1a simulant was supplied by Orbitec in support of the Chandrayaan-1 X-ray Spectrometer lunar mission activities. We also thank Kevin Kilburn FRAS and the Manchester Astronomical Society for their steadfast encouragement and their kind permission to use the Godlee Observatory in support of this work. References Andre, C.G., Wolfe, R.W., Adler, I., 1978. Evidence for high-magnesium subsurface basalt in Mare Crisium from orbital X-ray fluorescence data. In: Merrill, R., Papike, J.J. (Eds.), Mare Crisium: The View from Luna 24 (Conference Proceedings). Pergamon Press, New York, pp. 1–12. Besse, S. et al., 2011. Compositional variability of the Marius Hills volcanic complex from the Moon Mineralogy Mapper. J. Geophys. Res. 116, E00G13. http://dx.doi. org/10.1029/2010JE003725. Bielefeld, M.J., Wildey, R.L., Tombka, J.I., 1978. Correlation of chemistry with normal albedo in the Crisium region. In: Merrill, R., Papike, J.J. (Eds.), Mare Crisium: The View from Luna 24 (Conference Proceedings). Pergamon Press, New York, pp. 33–42. Binns, R.A., 1970. Pyroxenes from non-carbonaceous chondrite meteorites. Mineral. Mag. 37 (290), 649–669. Bowles, N., 2014. University of Oxford. Private communication. Carli, C. et al., 2014. Spectral variability of plagioclase–mafic mixtures (2): Investigation of the optical constant and retrieved mineral abundance dependence on particle size distribution. Icarus 235, 207–219. http://dx.doi. org/10.1016/j.icarus.2014.03.022. Chao, E.C.T. et al., 1970. Lunar glasses of impact origin: Physical and chemical characteristics and geologic implications. J. Geophys. Res. 75 (35), 7445–7479. http://dx.doi.org/10.1029/JB075i035p07445. Chao, E.C.T. et al., 1971. Petrology of unshocked crystalline rocks and evidence of impact metamorphism in Apollo 11 return samples. Proc. Lunar Sci. Conf. 1, 287–314 (Apollo 11). Chao, E.C.T., Best, J.B., Minken, J.A., 1972. Apollo 14 glasses of impact origin and their parent rock types. Proc. Lunar Sci. Conf. 1, 907–925. Characterisation summary of JSC-1A bulk lunar mare regolith simulant, 2007. <www.orbitec.com/store/JSC-1A_Bulk_Data_Characterization.pdf>. A. Fearnside et al. / Icarus 268 (2016) 156–171 Chayes, F., 1952. Relations between composition and indices of refraction in natural plagioclase. Am. J. Sci. Bowen volume, 85–106, n.d. <http://earth.geology.yale. edu/~ajs/1952A/85.pdf>. Dollfus, A., 1998. Lunar surface imaging polarimetry: 1. Roughness and grain size. Icarus 136, 69–103. http://dx.doi.org/10.1006/icar.1998.6008. Dollfus, A., Bowell, E., 1971. Polarimetric properties of the lunar surface and interpretation. I. Telescopic observations. Astron. Astrophys. 10, 29–53. Dorschner, J. et al., 1995. Steps towards interstellar silicate mineralogy II. Study of Mg–Fe-silicate glasses of variable composition. Astron. Astrophys. 300, 503– 520. Egan, W.G., Hilgemann, T.W., 1979. Optical Properties of Inhomogeneous Materials. Academic Press (ISBN 0-12-232650-4). Engelhardt, W. et al., 1970. Shock metamorphism of lunar rocks and origin of the regolith at the Apollo 11 landing site. Proc. Lunar Sci. Conf. 1, 363–384 (Apollo 11). Evenson, N.M., Murthy, V.R., Croscio Jr., M.R., 1974. Provenance of KREEP and the exotic component: Elemental and isotopic studies of grain size fractions in lunar soils. Proc. Lunar Sci. Conf. 5, 1401–1417. Geake, J.E., Dollfus, A., 1986. Planetary surface texture and albedo from parameter plots of optical polarization data. Mon. Not. R. Astron. Soc. 218, 75–91. http:// dx.doi.org/10.1093/mnras/218.1.75. Grieve, R.A.F., 1981. Impact cratering. Nature 291, 16. http://dx.doi.org/10.1038/ 291016a0. Hapke, 1994. The refractive index of the regolith of mercury. Lunar Planet. Sci. 25, 507 (held in Houston, TX, 14–18 March 1994). Hapke, B., 2005. Theory of Reflectance and Emittance Spectroscopy, first ed. Cambridge University Press (paperback, ISBN 0 521 61927 0). Hawke, B.R., Spudis, P.D., Clark, P.E., 1985. The origin of selected lunar geochemical anomalies: Implications for early volcanism and the formation of the light plains. Earth Moon Planet. 32 (3), 257–273. Head, J.W. et al., 1978. Regional stratigraphy and history of Mare Crisium. In: Merrill, R., Papike, J.J. (Eds.), Mare Crisium: The View from Luna 24 (Conference Proceedings). Pergamon Press, New York, pp. 43–74. Henry, R.C. et al., 1976. Moon 15, 51–65. http://dx.doi.org/10.1007/BF00562471. Hines, D.C. et al., 2008. Probing the lunar surface using spatially resolved optical and near-IR imaging polarimetry with AST/ACS and HST/NICMOS. In: ‘HST Optical and Near-Infrared Imaging Polarimetry of the Moon’, a white paper in response to lunar science with HST initiative solicited by STScI. Hiroi, T., Pieters, C.M., 1994. Estimation of grain sizes and mixing ratios of fine powder mixtures of common geologic minerals. J. Geophys. Res. 99 (E5), 10867–10879. http://dx.doi.org/10.1029/94JE00841. Hiroi, T. et al., 2009. Reproducing visible and near infra-red reflectance spectra of lunar rocks directly from their end-member spectra: Importance of ilmenite in estimating the lunar surface composition. Lunar Planet. Sci. 40, 1723.pdf. Horz, F. et al., 1984. Grain size evolution and fractionation trends in an experimental regolith. J. Geophys. Res. 89 (S01), C183–C196. http://dx.doi.org/ 10.1029/JB089iS01p0C183. Johnson, P., Christy, R., 1974. Optical constants of transition metals: Ti, V, Cr, Mn, Fe, Co, Ni, and Pd. Phys. Rev. B 9, 5056–5070. http://dx.doi.org/10.1103/ PhysRevB.9.5056. Joy, K., 2014. University of Manchester. Private Communication. Jung, M. et al., 2014. Preliminary results from multi-band polarimetric observations of the lunar surface. Lunar Planet. Sci. 45, 1315.pdf. Korotev, R.L., 1976. Geochemistry of grain size fractions of soils from Taurus– Littrow valley floor. Proc. Lunar Sci. Conf. 7, 695–726. Laul, J.C. et al., 1987. The lunar regolith: Chemistry and petrology of lunar 24 grain size fractions. Geochim. Cosmochim. Acta 51 (3), 661–673. http://dx.doi.org/ 10.1016/0016-7037(87)90077-9. Lehman, K.M. et al., 2013. Composition analysis of the Marius Hills volcanic complex using Diviner lunar radiometer experiment and Moon Mineralogy Mapper. Lunar Planet. Sci. 44, 1225 (held March 18–22, 2013 in The Woodlands, Texas. LPI Contribution No. 1719). Lucey, P.G., 1998. Model near-infrared optical constants of olivine and pyroxene as a function of iron content. J. Geophys. Res. 103 (E1), 1703–1713. http://dx.doi. org/10.1029/97JE03145. Masaytis, V.L., Mikhailov, M.V., Selivanovskaya, T.V., 1975. The Popigay Meteorite Crater. Nauka Press, Moscow, pp. 1–124. Mason, B., 1974. Aluminium–titanium-rich pyroxenes, with special reference to the Allende meteorite. Am. Mineral. 59, 1198–1202. 171 Muinonen, K. et al., 1996. Light scattering by Gaussian random particles a ray optics approximation. J. Quant. Spectrosc. Radiat. Transfer 55 (5), 577–601. http://dx. doi.org/10.1016/0022-4073(96)00003-9. Nimura, T., Hiroi, T., Pieters, C.M., 2008. An improved scheme for modeling the reflectance spectra of space-weathered regoliths. Earth Planets Space 60, 271– 275. http://dx.doi.org/10.1186/BF03352791. Noble, S., Pieters, C.M., Keller, L.P., 2007. An experimental approach to understanding the optical effects of space weathering. Icarus 192 (2), 629– 642. http://dx.doi.org/10.1016/j.icarus.2007.07.021. Novikov, V.V. et al., 1982. Correlation between Albedo and polarization properties of the Moon (heterogeneity of the relative porosity of the surface of the western part of the visible hemisphere). Sov. Astron. 26, 79–83. O’Keefe, J.D., Ahrens, T.J., 1975. Shock effects from a large impact on the Moon. Proc. Lunar Sci. Conf. 6, 2831–2844. Papike, J.J. et al., 1981. The relationship of the lunar regolith <10 micron fraction and agglutinates. Part 1: A model for agglutinate formation and some supportive evidence. Proc. Lunar Sci. Conf. 12b, 409–420. Papike, J.J., Simon, S.B., Laul, S.C., 1982. The lunar regolith: Chemistry, mineralogy, and petrology. Rev. Geophys. 20 (4), 761–826. http://dx.doi.org/10.1029/ RG020i004p00761. Rubin, M., 1985. Optical constants and bulk optical properties of soda lime silica glasses for windows. Solar Energy Mater. 12, 275–288. Ryder, G., 1981. Distribution of rocks at the Apollo 16 landing site. Workshop on Apollo 16, LPI Technical Report 81-01. Lunar and Planetary Institute, Houston, pp. 112–119. Schaal, R.B., Horz, F., 1977. Shock metamorphism of lunar and terrestrial basalts. Proc. Lunar Sci. Conf. 8, 1697–1729. Schonfeld, E., 1981. High spatial resolution Mg/Al maps of western Crisium and Sulpicius Gallus regions. Lunar Planet. Sci. 12B, 809–816. Segnit, E.R., 1953. Some data on synthetic aluminous and other pyroxenes. Mineral. Mag. 30 (223), 218–226. Shaffer, P.T.B., 1971. Refractive index, dispersion, and birefringence of silicon carbide polytypes. Appl. Opt. 10, 1034–1036. http://dx.doi.org/10.1364/ AO.10.001034. Shkuratov, Yu.G., 1981. Connection between the albedo and polarization properties of the Moon – Fresnel component of reflected light. Sov. Astron. 25, 490–493. Shkuratov, Yu.G., Opanasenko, N.V., 1992. Polarimetric and photometric study of the Moon: Telescope observation and laboratory simulation: 2. The positive polarisation. Icarus 99 (2), 468–484. http://dx.doi.org/10.1016/0019-1035(92) 90161-Y. Shkuratov, Yu.G., Kaydash, V.G., Pieters, C., 2005. Lunar clinopyroxene and plagioclase: Surface distribution and composition. Sol. Syst. Res. 39 (4), 255– 266. http://dx.doi.org/10.1007/s11208-005-0041-2. Shkuratov, Yu.G. et al., 2007. Multispectral polarimetry as a tool to investigate texture and chemistry of lunar regolith particles. Icarus 187, 406–416. http:// dx.doi.org/10.1016/j.icarus.2006.10.012. Stoffler, D., Hornemann, U., 1972. Quartz and feldspar glasses produced by natural and experimental shock. Meteoritics 7 (3), 371–394. Taylor, S.R., 1975. Lunar Science: A Post Apollo View. Pergamon Press (ISBN 0-08018274-7). Taylor, L.A. et al., 2001a. Lunar mare soils: Space weathering and the major effects of surface correlated nanophase Fe. J. Geophys. Res. (Planets) 106 (E11), 27985– 27999. http://dx.doi.org/10.1029/2000JE001402. Taylor, L.A. et al., 2001b. The effects of space weathering on Apollo 17 mare soils: Petrographic and chemical characterisation. Meteorit. Planet. Sci. 36 (2), 285– 299. http://dx.doi.org/10.1111/j.1945-5100.2001.tb01871.x. Taylor, L.A. et al., 2003. Mineralogical characterisation of lunar highland soils. Lunar Planet. Sci. 34. Abstract 1774 (CD-ROM). Velikodsky, Y.I. et al., 2011. New Earth-based absolute photometry of the Moon. Icarus 214 (1), 30–45. http://dx.doi.org/10.1016/j.icarus.2011.04.021. Warell, J., Davidsson, B.J.R., 2010. A Hapke model implementation for compositional analysis of VNIR spectra of Mercury. Icarus 209 (1), 164–178. http://dx.doi.org/ 10.1016/j.icarus.2009.11.037. Wilcox, B.B., Lucey, P.G., Hawke, B.R., 2006. Radiative transfer modelling of compositions of lunar pyroclastic deposits. J. Geophys. Res. 111, E09001. http://dx.doi.org/10.1029/2006JE002686. Zhu, Meng-Hua et al., 2013. Potassium map from Chang’E-2 constraints the impact of Crisium and Orientale basin on the Moon. Sci. Rep. 3. http://dx.doi.org/ 10.1038/srep01611 1611.