Você está na página 1de 6

Volume 17, number 1 OPTICS COMMUNICATIONS April 1976

A SIMPLE MODEL FOR DIFFUSE REFLECTION

J. David ZOOK
Honeywell Corporate Research Center,
Bloomington, Minnesota 55420, USA

Received 8 August 1975, first revised manuscript received 19 November 1975,


second revised manuscript received 8 January 1976

Diffuse reflection from a matte nonabsorbing inhomogeneous medium such as white paint or paper can be described
by a simple model in which light rays enter the volume of medium and then undergo a random walk until they reemerge
from the surface. Lambert's law of diffuse reflection is an immediate consequence of the random walk. Another conse-
quence of the volume interaction is that the light emerges from a different point than where it enters. This spreading of
the light was measured for BaSO4 white reflectance paint and for several kinds of paper. The random walk model implies
a diffusion equation which makes predictions that are in reasonable agreement with the experiments. The spreading is
proportional to an interaction length which, in this model, represents the range of distances that light rays penetrate be-
fore beginning their random walk.

1. Introduction Although no surface is known that strictly tbllows


Lambert's law, a large number of surfaces exist for
Lambert's law of diffuse reflection states that the which it holds approximately. These are "matte" or
luminance (or radiance) of a surface is constant, in- "flat" surfaces as opposed to "glossy" or "shiny" sur-
dependent of the viewing angle, and independent of faces which contain a significant component of spec-
the direction of the illumination. This law was first ular reflection. Becuase of its simplicity and conve-
stated by Lambert* in 1760 based on the observation nience, Lambert's law serves as the definition of a
that the moon and planets appear to be unitbrmly uniform diffuser and is widely used in photometry
bright even though the angle of illumination and tile and in radiation transfer. In spite of this, it has been
angle of viewing relative to tile normal obviously vary said that no simple model or theoretical basis for
widely over tile surface. Usually Lambert's law is Lambert's law exists [1,5]. The first explanation was
stated as offered by Bougier in 1762 [1,2] based on a micro-
facet model which supposes that a matte surface con-
l = lo cos O, (1)
sists of small mirrors, set at all possible inclinations.
where I is the luminous or radiant intensity (watts/ Recent developments in tile theory of scattering from
steradian or lumens/steradian) and 0 is the angle be- rough surfaces has been smnmarized by Bechmann
tween the surface normal and the viewing angle [ 1 - 3 ] . [5]. References to experimental work are given by
Tile cos 0 factor is the ratio of the apparent surface Sparrow and Cess [31 •
area to the actual surface area, so that eq. (1) is con- Eq. (1) also serves as the definition of diffuse emis-
sistent with constant luminance or radiance. The law sion, and is usually presented in standard texts as a
can be generalized to state that the reflected light is sort of axiom, the truth of which is ahnost self-evident
unpolarized, and is independent of the polarization [6]. It is indeed evident that thermal emission from
of the source [4]. a cavity in a body at uniform temperature (black body
radiation) should obey eq. (1) because for a given
* A survey and critique of the early literature is given in ref. energy, all normal modes of the radiation field are
[1]. equally populated. However, it is not so obvious that

77
Volume 17, number 1 OPTICS COMMUNICATIONS April 1976

therlnal emission from dielectric or nretallic bodies elnission fromn a cavity (blackbody radiation), which
with smooth surfaces, is described by eq. (1). In this is characterized by isotropic intensity and by Lambert's
case, electromagnetic theory predicts variations from law [2,3,7].
eq. (1), but these become significant only at large val- One purpose of the present paper is to point out
ues of 0 13] • Certainly in the case of reflection from that, because of the volume interaction, light emerges
matte surfaces, eq. ( i ) is even less obvious, because from a different point than where it enters. The
tile optical properties of the surface, the randomness spreading of light is readily naeasurable and the ran-
of surface height and surface slope and the wavelength dom walk model implies a diffusion equation which
of light relative to the scale of roughness are factors makes predictions that are in reasonable agreement
to be considered. with experinrent, in spite of the use of the simplified
The present discussion applies to the r e f e c t i o n model.
from matte surfaces where tile interaction does not The main feature of a random walk is that the
occur on the surface. Rather, light enters into the mean square displacement is proportional to the num-
volume of the material and is multiply scattered with ber of steps (collisions) N,
little or no attenuation until it reemerges at the sur-
r 2 = NI 2, (2)
face. This type of interaction should occur in white
diffuse r e f e c t o r s such as paint, paper, perhaps even where l is tire step length (mean free path). The nun>
snow, where the material consists of small transparent bet of steps, in turn, is proportional to t, the time of
particles having a higher index of refraction than the the walk. A relationship of this form holds even if the
media in which they are imbedded. The theory of steps are not completely random, such as in tile theo-
scattering from rough surfaces mentioned above is ry of polymer structure, where successive bonds (ran-
not directly applicable to a volume interaction. d a m steps) may be at a fixed bond angle, but are able
to rotate freely about tile bond (see, for example, ref.
[8] ). If we consider n(r, t) to be the number of pho-
2. The model tons per unit volume, or equivalently, the energy den-
sity, then eq. (2) implies the diffusion equation:
The simplest possible model of such a process is
an/~t = D g2n (3)
a ray picture in which the reflected rays undergo a
random walk starting at some point within the vol- as was shown by Einstein 4:. If we assume no absorp-
ume until they emerge at the surface. In the case of tion, the photon or energy density must also obey
a very porous material the distance a ray enters be- the continuity equation
fore being scattered is related to the surface rough-
an~at = - v . J , (4)
ness which in turn is related to tile porosity of the
material. An alternative viewpoint is that each scat- where J is the flux density or exitance. Thus (3) and
tering center divides the rays into an isotropically (4) imply that
scattered beam and an unscattered beam. Thus each
d = D gn, (5)
scatterer in the incident beam acts as the starting
point for a random walk. Lambert's law is an innne- and we are interested only in the steady-state condi-
diate consequence of this model because if tile walk tion, an~at = O.
is truly random the scattered photons within any The simplest case is that in which the light begins
volume element will have m o m e n t u m vectors uni- tile random walk at a point a distance b below the
formly distributed on a sphere in m o m e n t u m space, surface. In terms of diffusion this case is represented
giving equal radiance in all directions. After a suffi- by a point diffusion source and we need to solve La-
cient number of randomizing scattering events any place's equation for a point source. The boundary
correlation between the initial and final plane of po-
* See ref. [8], p. 350, for a statement of Einstein's deriva-
larization would be lost so that the uniform distribu- tion. Credit for developing the relationship between random
tion applies to b o t h polarizations. Thus, with this walks and diffusion is due to L. Bachelier, according to ref.
model diffuse r e f e c t i o n becomes very similar to [9]. Chapters 3 and 14 of the latter discuss random walks.

78
Volume 17, number 1 OPTICS COMMUNICATIONS April 1976

Ii
IRIS , IE '

DIFFUSE
I REFLECTOR~

[ LASER BEAM
REDUCER BEAM
SPLITTER

Fig. 1. Arrangement for measuring the spatial distribution of reflected light. The width of the incident beam from the laser is
exaggerated to show the action of the beam reducing lens, which focusses the beam to a waist (plane wave) at the objective lens.
The objective lens projects a magnified image of diffuse reflecting object onto the photo diode.

condition is that the surface must represent an infi- portional to the integral of eq. (6):
nite sink. In the analogous case of heat flow which is
y
also described by the diffusion equation, an infinite
sink implies a constant temperature surface and the
Io(r)=f 4r)rdr= 1 1
0 Vql + (r/b) 2 ' (7)
component of heat flow tangential to the surface is
zero. The situation is also analogous to the electro- where the constant c has been chosen to give I0(oo) = 1
static problem of a point charge a distance b from a Within the depth of field tile illuminating beam
conducting plane; it can be solved in a trivial way by can be considered to be a parallel beam, larger than
the method of images. The result is that the light in- typical particle size in the object. All the scattering
tensity at the surface (analogous to the normal com- will not occur at a single point, but rather will occur
ponent of electric field) is given by along a certain interaction length. Thus the reflected
¢
radiation is a sum of contributions from scatterers at
different distances b:
J(r) = ( r2 + b2)3/2 , (6)
Ylo(r/b ) dN(b)
where e is a constant and r is the cylindrical coordi- I(r) - y dN(b) ' (8)
nate, measured from the point source.
Experimentally it is easier to measure the inte- where dN is strength of the point sources within the
grated flux passing through a circle of radius r than range b and b + db.
to measure J(r) in eq. (6). This can readily be accom- For example, if each scattering event removes a
plished by the use of a beam splitter and an objective certain fraction of energy from the beam, and the
lens which images the surface onto a photodetector number of scatterers per unit volume is constant, we
with a variable aperture, as shown in fig. 1. The re- would have an exponential law:
flected light is collected from a circle whose radius
dN/db = e x p ( - b / a 1 )"
is equal to that of the variable aperture (labelled
"iris" in fig. 1) divided by the magnification of the In this case eq. (8) gives
objective. Thus the amount of light collected is pro-
79
Volume 17, number 1 OPTICS COMMUNICATIONS April 1976

giving an illuminating spot size* of 5(4X/n) = 4 ~m in


I (a;)
r = t - ~ a lr 2rrrIHl(~--l)
a7 1I/ (~1)1' (9)
diameter. For an objective lens focal length of 3 ram,
where H 1 is the Struve function and 111 is a Bessel tile incident beam diameter at the principal plane o f
function [10]. the object should be 3 mm/5 = 0.6 ram, and should
Alternatively, the material can be viewed as being have a plane wavefront. The beam reducing lens in
rough and porous, so that different rays enter differ- fig. t focusses the laser beam to a waist at the objec-
ent distances before interacting. If tile surface rough- tive. In the above example, the beam would have an
ness is describable by a gaussian function, then F-number of 0.6 mm/(4X/n) ~ 750 and a focal length
of 750 trim if the beam diameter at the beam reducer
dN/db = exp (- b2/a~), is 1 ram.
and eq. (8) becomes:
I r r2 r
3. Comparison with experiment
\a2/ 2
where erfc is the complimentary error function. A The results of measurements taken with a 1 mW
plot of eqs. (9) and (10) with a 2 = 1.56 a t shows that HeNe laser and a lens with NA = 0.85 on white re-
the curves are almost superimposed, so that it would flectance paint [12] (BaSO4) and two types of paper
be quite difficult to differentiate experimentally are shown in fig. 2, normalized with respect to an
among these cases and other cases SUCh as a linear aluminum front surface mirror. The white reflectance
distribution with a cutoff. paint has been proposed as a secopdary standard for
The amount of light collected depends on the col- reflectance measurements [ 1 2 - 1 4 ] . In the case of
lection angle of the objective. If we assume that eq. the papers several thicknesses were used because of
(1) applies, then the fraction of the total ligllt col- tile non-zero transmission of a single sheet. The meas-
lected by the objective is ured value of R(r) depends on the focussing of the
objective lens, and the values vary somewhat from
f cos0 dgZ 0m
collection angle 2n fo cos 0 sin 0 dO point to point. If the focussing is not optimum, lower
values of R(r) will be obtained at small values of r,
2n f~)/2cos 0 sin 0 dO and quadratic rather than linear behavior is observed.
cos 0 df2
Tile results shown were obtained by focussing with
hemisphere
crossed polarizers to avoid focussing on the front sur-
sin 2 011.1 = NA 2 , ill) face, especially in the case of the glossy surface. A
complicating factor in the experiment is that available
where 0 m is the maximum collection angle subtended
beam splitters act as fairly good polarizers so that
by the objective, and, by definition, sin 0m is the nu-
measurements had to be made of both components
merical aperture, NA, of the lens. In view of eqs. (7) of the reflected light. The results shown in fig. 2 refer
and (8) the ratio of the light collected from a circle
to the total reflected light (square root of the sum
of radius r to the incident ligl~t is given by
of the squares of the components).
R(r) = NA 2 R I(r), (12) The transmission of objective lens undoubtedly
depends on angle and this was not corrected for in
where R ~ is the hemispherical reflectance or albedo
the experiment. However, the correction is not likely
of the diffuse reflector and I is given by eq. (8).
to be large. The reflected light is approximately lam-
One degree of freedom in the experiment is the
bertian for all points in the object, and therefore all
choice of the illuminating spot size. As we show be-
the points in the image plane are nearly equally af-
low, the value of b is the range of 20 p m , so that the
fected by the unequal transmission of the lens as a
focussed beam should have a depth of field on the
function of angle. Evidence for this is that a good
order of 20 ~m. 'Fhe depth of field is hf2/w 2 where w
microscope objective such as the one used will show
is the width of the beam at the objective and f i s the
focal length of the objective. Thus the F-number,
* Ref. [11 ] gives a brief review of laser beam optics.
f/w, of the illuminating beam should be about 5,
80
Volume 17, number 1 OPTICS COMMUNICATIONS April 1976

.7 I I t I I I I I I i t

MINIMUM RADIUS OF /
iNCIDENT BEAM J
.6 j.u

rr
.5
Z
_o
I-
t.) .4
rt"
h
r~ .5
UJ
I.-
IJ
.2 'dr ~'~ 13 GLOSSY WHITE PHOTO
_J y : 2::S :::R E PHOTO
O
L)
.I

1 I I 1 I I i I I I
0 20 40 60 80 IO0 120 140 160 180 200/.~
RADIUS r

Fig. 2. Spatial distribution of reflected light. The radius r is the radius of the iris divided by the magnification of the objective.
The points are experimental averages of several measurements. The lines are plots of eq. (9) combined with eq. (12), having pa-
rameters al = 18, 26, and 40 #m and NA2R~ = 0.64, 0.6, and 16 from top to bottom.

an image of a uniformly illuminated diffusely scatter- In our experiment, the beam size of 4 # m was less
ing object that is just as bright at the edges as it is at (after magnification by the objective lens) than the
the center. Thus the uneven transmission affects all m i n i m u m iris size.
image points in essentially the same way. It was found that the light reflected at small values
Although it is predicted in the theory discussed of r was fairly well polarized parallel to the incident
above, the linear behavior at small values o f r was un- polarization. Such polarization behavior is to be ex-
expected when first observed. According to eq. (6) pected since light that has undergone only one scat-
the intensity is constant for r ~ b. The integrated in- tering event would be included. The linear behavior
tensity R(r) is thus proportional to area and would at small values o f r is a basic feature that was retained
be expected to vary as r 2. However, the inclusion in even when measurements were taken with crossed
eq. (8) of diffusion sources having non-zero values of polarizers, a technique used to exclude specular re-
dN/db at b = 0 (surface contributions) results in lin- flections. We note that Billmeyer et al. [13] have
ear behavior at small r. Effectively then, eqs. (9) and found that when crossed polarizers were used in bidi-
(10) include surface as well as volume contributions rectional reflectance measurements on pressed BaSO4,
to the reflected light. lambertian behavior is observed for all angles of il-
The prediction of linear behavior is actually an lumination up to at least 85 ° (5 ° from grazing).
approximation based on illumination from a line
source, whereas the incident beam has a finite width
which causes quadratic behavior for values of r less 4. Conclusions
than the beam width. The approximation can be
quite good if the m i n i m u m beam size compatible with In summary, reflection from diffuse reflectors in-
the desired depth of field is chosen as discussed above. volves a volume interaction as evidenced from spatial

81
Volume 17, number 1 OPTICS COMMUNICATIONS April 1976

spreading of reflected light. A very simple model of References


the volume interaction is that the light rays follow a
random walk after entering a certain distance into [ 1] V.G.W. Harrison, Definition and Measurement of Gloss
the medium. Lambert's law follows naturally from (The Printing and Allied Trades Research Association,
London, 1945).
this model which also predicts the spatial distribu-
[2] J.W.T. Walsh, Photometry (Constable and Co., London,
tion of light as it emerges at the surface. Experiments Sec. Ed., 1953).
with focussed light at normal incidence give a meas- [3] E.M. Sparrow and R.O. Cess, Radiation Heat Transfer
ure of the surface roughness, or more precisely, an (Brooks/Cole Publishing Co., Belmont, California, 1970).
interaction length, a, which represents the range of [4] D.C. Cramer and M.E. Blair, Appl. Opt. 8 (1969) 1597
distances light rays penetrate before beginning their
1605.
[5 ] P. Beckman, in: Progress in Optics, ed. E. Wolf, Vol. 6
random walk. (North-Holland Publishing Co., Amsterdam, 1967) p. 57.
Other experiments, for example, measurements of [6] M. Born and E. Wolf, Principles of Optics, 3rd Ed.
the transmission of light by thin diffuse reflectors (Pergamon Press, Oxford, 1965) p. 182;
such as paper, would provide additional tests of the M.V. Klein, Optics (Wiley & Sons, New York, 1970)
p. 124.
simple theory. More elaborate experiments and theo-
[7] M.A. Melvin, Am. J. Phys. 23 (1955) 508-510, and
ries could provide a more detailed link between the references therein.
particle size and spacing and the parameter a. In any [8] C. Tanford, Physical Chemistry of Macromolecules
case, it is hoped that the theory and experiments (Wiley & Sons, New York, 1961) 152-161.
presented here will give additional insight into the [9] W. Feller, An Introduction to Probability Theory and
Its Applications, Vol. I, 2nd Ed. (John Wiley & Sons,
nature of diffuse reflection and radiative transfer
New York) p. 323.
that will prove helpful in the design of reflectors such [10] M. Abramowitz and I.A. Stegun, eds., Handbook of
as paper, paints and coatings for solar energy conver- Mathematical Functions, Nat. Bur. of Stand., Appl.
sion. Mathem. Ser. 55 (1964).
[11] J.D. Zook and T.C. Lee, Appl. Optc. 11 (1972) 2140.
[12] F. Grum and G. W1. Luckey, Appl. Opt. 7 (1968) 2289-
2294.
Acknowledgements [13] F.W. Billmeyer, Appl. Opt. 8 (1969) 737-750.
[ 14] F.W. Billmeyr Jr., D.L. Lewis and J.G. Davidson, Color
The author gratefully acknowledges helpful dis- Engineering 9 (1971) 31-36.
cussions with F.W. Billmeyer, P. Beckmann, C.W.
Erickson, and R. Henisch, who also provided the sam-
ple of white reflectance paint, and the able assistance
of R. Hegel with the experiments.

82

Você também pode gostar