Application of the conservation of etendue theorem for

2-D subdomains of the phase space in nonimaging

concentrators

Juan C. Minano

The conservation of etendue for general 2-D bundles of rays (not necessarily coplanar) is examined (a 2-D
bundle of rays is that whose rays are distinguishable by giving each one two parameters). This is one of the
integral invariants of Poincar6 and it is directly related to the Lagrange invariant. The application of this
theorem to selected 2-D bundles of rays crossing an arbitrary cylindrical concentrator gives us a relationship
between the maximum geometrical concentration of a cylindrical concentrator and the angular field of view
which is more restrictive than the general one (i.e., the relationship is valid for an arbitrary concentrator)
when the collector is surrounded by a refractive medium.

1. Theoretical Background OL
a== _ _ _ _ _ _ _ _
* ~ (3)
In this paper we shall follow the Hamiltonian de- a.y (1 + 2 + y2)1/2
scription of geometrical optics, which has been very A look at Eqs. (2) and (3) reveals that p and q are, re-
useful for treating problems of nonimaging optics. In spectively, n times the cosine of the angle formed be-
particular, the Liouville theorem, which can be derived tween the ray and a parallel to the x axis and between
from this description, has given a simple answer to a the ray and a parallel to the y axis at point (x,y). Ob-
fundamental problem of nonimaging concentrators: to serve that a ray leaving the x-y plane through the sem-
calculate the smallest exit aperture area of a concen- ispace z positive is fully defined by the set of variables
1
trator that will collect a given beam of rays. This x,y,p,q (Fig. 1).
2 is
theorem is also known as the etendue invariant and the When a 4-D region of the phase space x,y,p,q
conservation of phase space volume; it is one of the in- transformed point to point into a 4-D region of other
4
tegral invariants of Poincar6.3 phase space x',y',p',q' by a canonical transformation,
Let x (z),y (z) specify a light ray in a set of Cartesian the Liouville theorem states 2 3 5 6
coordinates. In terms of these functions, the Fermat dxdydpdq = dx'dy'dp'dq'. (4)
principle can be expressed as
Let us consider an entry aperture x-y plane and an exit
2 2
6 fz n(x,y,z)(1 + x + 9 )dz = 0, (1) aperture plane of an optical system (Fig. 1). Each ray
Zl
leaving the first plane and reaching the last one estab-
where the dot denotes differentiation with respect to z, lishes a correspondence point to point between (xyp,q)
and n (xy,z) is the index of refraction corresponding to and (x',y',p',q'). This transformation is canonical and
point (x,y,z) (it will be simply noted by n). We can involves a special generating function called Hamilton's
identify the integrand as the Lagrangian L(x,y,z,i,5) characteristic function.7 2 Then, Eq. (4) holds for the
of the light ray. Then, the conjugate momenta corre- transformation of the bundle of rays entering the entry
sponding to x and y are aperture of the optical system into the bundle of rays
aL
OL- _ _
Xi_ _ _ _ _
(2)
leaving the exit aperture.
ax (1+x2+2)1/2( There is another canonical transformation that will
be in our interest: the transformation of the set of
variables (x,y,p,q) into (p,6,j,h) where p and 0 are the
polar coordinates of a point in the x-y plane and j,h are,
respectively, their conjugate momenta8
The author is with Universidad Politecnica de Madrid, ETSI
j = p cosO + q sinO, (5)
Telecomunicacion, Instituto de Energia Solar, Madrid 3, Spain.
Received 20 July 1983. h = p(q cosO - p sinO). (6)
0003-6935/85/122021-05$02.00/0.
© 1984 Optical Society of America. From these equations wecan verify that j is n times the

15 June 1984 / Vol. 23, No. 12 / APPLIED OPTICS 2021

 rays. Since the sum of the square of the three direction
cosines of a ray (with respect to three orthogonal axes)
is one, the field of variability of P/n(X,Y),Q/n(X,Y) [or
OPT ICA L P'/n(X',Y'),Q'/n(X',Y')] is restricted to the circle of
SYSTEM radius unity. We assume that the index of refraction
car-, nnH
(_ X.i'yYz
q outside the concentrator is 1 and the one of the medium
I ,z surrounding the collector is n,. Then we can express
y
y. E = Aed (9)
Fig. 1. Coordinate systems used for the rays at the entry aperture and affirm that
x,y,p,q and the exit aperture x',y',p',q' of an arbitrary optical system.
The two rays drawn have coordinates x = 0, y = 0 and x' = 0,
E Arn., (10)
Y = 0. where Ae is the area of 1eXAc is the area of 1c, and is
called "average acceptance area."'10
cosine of the angle formed between the ray and the Let us consider the region of the P-Q plane obtained
straight line of the x-y plane defined by 0 = constant, by the intersection of D and a domain of constant X, Y
and h is np times the cosine of the angle formed between (i.e., the collected rays that cross 2e at X,Y), a is the
the ray and the tangent to the circumference p = con- averaged area of all these regions corresponding to the
stant at point p,O;h is an invariant (skew invariant) in points of 1e,. Since P,Q is restricted to a circle of radius
a ray traversing an axisymmetrical system.2 The 1,i <•-r.
Liouville theorem applied to the two aforementioned Combining (9) and (10) we obtain
canonical transformations states that
Cg < 2nC = Cgm(d), (11)
dxdydpdq = dpdOdjdh = dx'dy'dp'dq' = dp'dO'dj'dh', (7) ela)

where p',0'j',h' are obtained from x',y',p',q' in the same where Cg is the geometrical concentration. Equation
way as p,Oj,h derive from x,y,p,q. (11) expresses the maximum Cg achievable with a con-
Equation (4) is valid even if the x-y and x'-y' planes centrator of given d. The quotient Cg/Cg () is called
are not parallel. Then we can derive an equation sim- the "degree of isotropy" g.11 For a concentrator of flat
ilar to (4) for the sets of variables (X,YP,Q) and (X', entry aperture which collects every ray forming, with
Y',P',Q'), where X and Y are the coordinates of a point the normal to e, an angle lower than q5,Eq. (11) takes
in a given curved entry aperture surface, and P,Q and its more usual form (observe that in this case p2 + Q2
n (X, Y) times the cosine of the angle formed between < sin2, ) 1:
the ray and the line Y = constant and between the ray
and the line X = constant, respectively. The lines X Cg n, (12)
= constant and Y = constant are assumed to be or-
thogonal. Note xlydXdYdPdQ is equal to a differ- 11. Conservation of Etendue for 2-D Subdomains of
ential dxdydpdq where the x axis is the straight line the Phase Space
tangent to the line Y = constant and the y axis is tan-
gent to the line X = constant; Ix and ly are two func- Another integral invariant of Poincare3 will interest
tions of X, Y that make lXdX and lyd Y be differentials us now. This integral invariant refers to a 2-D region
of length. Similar considerations can be done with re- of the phase space x,y,p,q and to the 2-D region of the
spect to X',Y',P',Q' and so lx,ly',dX'dY'dP'dQ = phase space x',y',p',q' obtained by a canonical trans-
dx'dy'dp'dq'. Then, using Eq. (4) we have that formation of the first one. It is not necessary that the
lxlydXdYdPdQ = lxly'dX'dY'dP'dQ'. 9 trajectories of the rays represented by these 2-D regions
Let us assume that X,Y are the surface coordinates are contained in a plane and, in general, this will not
of a point of the entry aperture 2e of an arbitrary con- occur. The differential form of this invariant is
centrator and X',Y' are the coordinates of a point of its dx dp + dy dq = dx'dp'+ dy' dq'. (13)
collector 1, The concentrator casts on its collector a
given bundle of rays arriving at its entry aperture. The A simple proof of the validity of (13) can be obtained
rays of this bundle occupy a region D of the X- Y-P-Q from the analogy between Hamiltonian optics and
space and a region D' of the X'-Y'-P'-Q' space. Ob- mechanics. 5 We can also see that (13) is only the La-
viously, if a ray (X,Y,P,Q) belongs to D, it is trans- grange invariant. Let us assume that u and v are two
formed by the concentrator into a point of the regionD'. parameters that determine each ray of the bundle de-
The conservation of etendue applied to the transfor- fined by either of these 2-D regions:
mation generated by each collected ray when linking a x = x(u,v), x' = x'(u,
set (X,Y,P,Q) to another set (X',Y',P',Q') states
that y = y(u,v), y' = y(u,v), (14)
p = p(u,v)
E= fD IxlydXdYdPdQ = f lx1ly'dX'dY'dP'dQ', (8)
q = q(u,v),
p' = P,(uv),
q' = q'(u,v).
where E is called the etendue of the bundle of collected What we should show is that the Jacobians fulfill

2022 APPLIED OPTICS/ Vol. 23, No. 12 / 15 June 1984

 Ill. 2-D Bundles in Cylindrical Concentrators

tconst. Let us assume that X,Y are the surface coordinates

of a point of the entry aperture of an arbitrary cylin-
drical concentrator and X',Y' are the coordinates of a
point of 1c, We can choose the coordinate systems so
that Ix = ly = Ix' = ly' = 1 and that the lines Y = const,
Y' = const are straight lines parallel to the axis of
translational symmetry, and the lines X = const, X' =
const are transverse to it. With this set of coordinates
we can say that P is an invariant of the ray in a cylin-
7aC CosI drical concentrator and so P = P' (Fig. 2). Let us con-
2
sider the 2-D bundle of rays which has, when crossing
Fig. 2. Coordinate system used for the rays at the entry aperture e
1, a constant X and a constant P. Since the concen-
and at the exit aperture Z, of an arbitrary cylindrical concentrator. trator is cylindrical, we know that the rays of this bundle
will have, in 1, a constant P' = P. Then, applying Eq.
(13) we have
D(x,p) + D(y,q) D(x',p') + D(y',q') dYdQ = dY' dQ'. (18)
= (15)
D(u,v) D(u,v) D(u,v) D(u,v) Observe that this equation holds not only for the rays
Equation (15) can also be written as (the dot denotes contained in a plane normal to the axis of translational
scalar product) symmetry (rays of P = 0) but also for any 2-D bundle of
(ax y\ ap aql - (ap aq x ay constant P (it is not necessary to have a constant X).
Let us call E2 -D(X,P) the etendue of the 2-D bundle of
\aluO uJav aid v
OF. dv duavid vav
collected rays crossing the entry aperture at a given X
and at a given P:
au TU) -v
ka TV) auaTO avavJ1
12
which is the Lagrange invariant. Note that we have E 2 D(X,P) = Sfls(xP) dYdQ = S.(X P,)dY'dQ', (19)
implicitly assumed that the 2-D bundle of rays is
crossing a plane z = constant (and z' = constant). The where S(XP) is the 2-D region of the X- Y-P-Q space
etendue of the 2-D bundle is obtained by adding (inte- of X = const and P = const and D 2-D(X,P') is the 2-D
grating) each differential contribution dxdp + dydq (or region obtained by transforming D fl S(X,P). The
dx'dp' + dy'dq') of the bundle. Considering that each region D2 D(X,P') has the properties that all its rays
of these contributions is positive, the integral will also have a constant P' = P and that there are not two rays
be positive. The value of this integral is sometimes with equal Y',P',Q' and different X'. Note for this that
called the geometrical energy flux.13 As was shown by X can be expressed (because it is a cylindrical system)
as A =_ ' + [ Y's's').
Luneburg, a 2-D bundle whose etendue is zero is a -. . - -11 no I-
(ZU)
normal system of rays, i.e., there is a family of surfaces
normal to the rays (the wave fronts). The converse Then we can say that
theorem is also true: a normal system of rays has zero E 2 -D(X,P) < 2Ae(1 - p2 )1 /2 (21)
etendue. From here it follows that the theorem of
Malus-Dupin is equivalent to the conservation of the [the index of refraction at Xe is assumed to be 1 and so
zero etendue of the normal system of rays. IQ I < (1 - P 2 ) 1 / 2 ] because of the first equality of (19),
The usual expression of the Lagrange invariant is a and that
line integral on a closed curve and it refers only to nor- E 2 D(X,P) < 2A,(n2 - p 2 )/2 (22)
mal systems of rays.1 4 This integral invariant relative
to closed curves can be transformed into the integral because of the second equality of (19). A and Ac are,
invariant obtained from Eq. (13) with the Stokes the- respectively, the areas per unit of length of 2e and 1,
orem. 3 n, is the index of refraction of the medium surrounding
The integral invariants of Poincar6 have also been X, so IQ' < (n 2 - P/ 2 )1/2. E 2 -D(X,P) is, in fact, only
used to describe the flow of phase space through non- a function of P, because the coordinate X of a ray at Ie
imaging concentrators by means of the geometrical does not care whether the ray is collected.
vector flux.1 5 The invariant was expressed there as the As a result of the first equality of (8), E can be ex-
conservation (zero divergence) of this vector. pressed as
By similar reasoning to that in sec. I we can obtain for X=1 P 1
the sets of variables X,Y,P,Q and X',Y',P',Q' that x=o P=1 E2-D(P)dX dP (23)
lxdXdP + IydYdQ = lx'dX'dP' + ly'dY'dQ'. Using By combining inequalities (21) and (22) with Eqs. (23)
the sets p,Oj,h and p',0',j',h' the equivalent of Eq. (13) and (10) and the definition of the degree of isotropy g,
is we can obtain that g for a cylindrical concentrator
dp dj + dOdh = dp' dj' + dO'dh'. (17) cannot be greater than the expression

15 June 1984 / Vol. 23, No. 12 / APPLIED OPTICS 2023

 .78 upper bound of Cgvs a for cylindrical concentrators is
lower than the general one Cgm(6):
I Cg • gM(Cgnc)Cgm(i). (27)
.76 The equality of Eq. (27) can be solved by iterations to
obtain the upper bound of Cg for cylindrical concen-
3 5 7 9 trators as a function of 6 (see Ref. 16). Figure 4 shows
GemetricalCon tmttlon It this upper bound compared with the general one, when
Fig. 3. Upper bound of degree of isotropy for cylindrical concen- n = 1.5. This upper bound is achieved by the cylin-
trators whose collector is surrounded by a medium of index of re- drical CPCs for any n, The combination of Eqs. (24)
fraction 1.5. and (26) is analyzed in Ref. 16. We have been able to
use Eq. (13) for studying the degree of isotropy of cy-
lindrical concentrators because the conservation of P
10 has given us information about the region transformed
I
of D n S(X,P). In the concentrators of axial symmetry
a the rays conserve their value of h (skew invariant).
a This suggests that we can obtain similar information in
an axisymmetrical concentrator. Let us consider one
i. 5
2
of these concentrators of flat 1e and 1, The coordi-
nates of a point of 2e are p and 0 and those of a point of
I
co
I 1, are p' and 0'. The study of the etendue of the 2-D
1 2 3 bundle of rays crossing z2 e at a constant 0 and at a con-
stant h does not give us more information than that
Fig. 4. Upper bound of geometrical concentration vs average ac- derived from Eq. (11). This is because the upper bound
ceptance area a for cylindrical concentrators whose collector is sur- of this E2 -D(h)obtained from the conditions at 1e [i.e.,
rounded by a medium of index of refraction n = 1.5. The dashed the equivalent inequality to (21)] and the upper bound
line shows the general upper bound for arbitrary concentrators of nc obtained from the conditions at 1c [the equivalent in-
= 1.5. equality to (22)] cannot be applied together to a single
concentrator, i.e., if Ae > ncAc,the conditions at c are
2(n,2-1 1/2 more restrictive than the conditions at 1e for any value
g gM(Cgnc) = 2Cg sin' C 1)
of h, and if A < n2Ac, the conditions at ye are more
rn2in'1
(C,2 - n 1/21l 24 restrictive than the conditions at 1, These inequali-
+ n2 sin 1 |-(I, n) ]i' (24) ties are
ncg
where Cgis the geometrical concentration. For details
of these calculations see Ref. 16. In Fig. 3, gM(Cg,1.5) E 2 -D(h) p=R j=[-(h/p)2]1/2 (28)
is plotted. The proof of Eq. (24) can be done without Jp=h j=[l-(h/p)21/2
referring explicitly to Eq. (13) (as is done in Ref. 16).
The degree of isotropy is an important parameter of r2Dh
p-r rj dp=n2-d/p2]
the concentrators which, as we have seen in Sec. I, gives E2Dh)SJPp=h/n J j=-[n2-(h/p')21/2 d j (29)
the ratio of the Cgof a concentrator to the maximum Cg
corresponding to its average acceptance area . This
parameter also appears when studying the concentra- R is the radius of 2e and r is the radius of I. Note that
tors illuminated by a given source. Then Cgand a be- the field of variability of j (and j') is given by the rela-
come parameters less useful than the coupled optical tion between the direction cosines j2 + (h/p)2 < 1 [and
j]2 + (h/p) 2 < n2] and that p cannot be smaller than h
concentration C, intercept factor I (COis the ratio of
the power collected to the maximum power on the col- (and p' cannot be smaller than h/nc); this would require
lector when this is without concentrator and I is the having a direction cosine >1. If we change the variable
p' in (29) to a new variable p = p'nc, we obtain the in-
ratio of the power collected to the maximum power
reaching the entry aperture surface). It has been shown equality (28) with the only difference being that the first
elsewhere1 0 that the source illuminating the concen- integral of (29) is extended from p = h to p = mc.
trator, which in Ref. 10 is assumed to have a flat entry Then, if R < rn, (29) is a more restrictive condition for
aperture, sets an upper bound of the C achievable by any value of h, and if R < rnc, (28) is the most restrictive
the concentrators that have a given I,COM(I). The ratio condition. None of these inequalities gives us a new
of the C achieved by a concentrator to the upper bound information by itself.
corresponding to its I is gy, where -y is called the ad- IV. Conclusions
aptation factor. Summarizing, we have that
We have presented an application to nonimaging
Cg = gCgm(), (25) concentrators of one of Poincar's integral invariants
referring to a two-parameter bundle of rays (2-D bun-
C. = gyCM(I). (26)
dles). This integral invariant is directly related to the
Then, using (24) and (25), we have that (if n > 1) the Lagrange invariant. It permits us to assign a value

2024 APPLIED OPTICS/ Vol. 23, No. 12 / 15 June 1984

(called etendue) to the 2-D bundle. This etendue can
be calculated on any surface intercepting the whole
bundle and it is zero if the bundle is a normal system of
rays, i.e., if there is a family of surfaces (wave fronts)
normal to all the rays of the bundle.
The application of this invariant to selected 2-D
bundles of rays crossing an arbitrary cylindrical con-
centrator has given us two useful inequalities with which
we have been able to obtain the particular upper bound
of geometrical concentration vs average acceptance area
for cylindrical concentrators.
References
1. R. Winston, "Light Collection Within the Framework of Geo-
metrical Optics," J. Opt. Soc. Am. 60, 245 (1970).
2. W. T. Welford and R. Winston, The Optics of Nonimaging
Concentrators (Academic, New York, 1978).
3. H. Poincar6, Les methodes nouvelles de la mecanique celeste
(Dover, New York, 1957), Vol. 3.
4. R. Weinstock, Calculus of Variations (McGraw-Hill, New York,
1952).
5. L. D. Landau and E. M. Lifshitz, Spanish translation: Mecdnica
(Revert6, Barcelona, 1965).
6. W. T. Welford, Aberrations of the Symmetrical Optical System
(Academic, London, 1974).
7. R. K. Luneburg, Mathematical Theory of Optics (U. California
Press, Berkeley, 1964).
8. A generating function of this transformation can be G(p,q,p,O)
= -(pp cosO + qp sinO).
9. The Hamiltonian characteristic function can also be defined for
the cases where the terminal points of the rays lie on a curved
surface. See H. A. Buchdall, An Introduction to Hamiltonian
Optics (Cambridge U.P., London, 1970); G. W. Forbes, "New
Class of Characteristic Functions in Hamiltonian Optics," J. Opt.
Soc. Am. 72, 1698 (1982).
10. J. C. Mifiano and A. Luque, "Limit of concentration under
Nonhomogeneous Extended Light Sources," Appl. Opt. 22, 2751
(1983).
11. For the sake of simplicity we have assumed in this work that the
collector is monofacial (only collects rays coming from one face
of the collector surface). If it is assumed that the collector is
bifacial, as is done in Ref. 10, the right-hand side of Eq. (10)
should be multiplied by 2 and the definitions of Cgm(a) and g
suffer the corresponding changes. Note that, with the treatment
used here, a concentrator with a monofacial collector can have
g = 1, while in the treatment of Ref. 10 g cannot, in this case, be
>1/2. From an optical point of view, a bifacial collector can be
considered as having two monofacial faces. Richard Wyatt
12. See, for example, M. Herzberger, "Mathematics and Geometrical
Optics," Supplementary Note III in Ref. 7. British Telecom Research Labs.
13. M. Herzberger, Modern GeometricalOptics (Wiley Interscience, Photo: F. S. Harris, Jr.
New York, 1958).
14. M. Born and E. Wolf, Principles of Optics (Pergamon, New York,
1970).
15. R. Winston and W. T. Welford, "Geometrical Vector Flux and
Some New Nonimaging Concentrators," J. Opt. Soc. Am. 69, 532
(1979).
16. J. C. Mifiano and A. Luque, "Limit of Concentration for Cylin-
drical Concentrators Under Extended Light Sources," Appl. Opt.
22, 2437 (1983).

15 June 1984 / Vol. 23, No. 12 / APPLIED OPTICS 2025