Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Spherical aberration, coma, and the Abbe sine condition for physicists who don't

design lenses
Alex Small

Citation: American Journal of Physics 86, 487 (2018); doi: 10.1119/1.5036939


View online: https://doi.org/10.1119/1.5036939
View Table of Contents: http://aapt.scitation.org/toc/ajp/86/7
Published by the American Association of Physics Teachers
Spherical aberration, coma, and the Abbe sine condition for physicists
who don’t design lenses
Alex Smalla)
Department of Physics and Astronomy, California State Polytechnic University, Pomona,
California 91768
(Received 4 August 2016; accepted 9 April 2018)
Treatments of lens aberrations tend to be either short and descriptive surveys or detailed theoretical
treatments. In this article, we take an intermediate approach, showing that the mechanism of
spherical aberration can be easily understood on a qualitative level by comparing a sphere with the
ideal lens shape prescribed by the Principle of Least Time, and by using a similar argument to
understand coma aberration. We also use a simple argument from radiant flux conservation to
explain why aberration-free imaging of finite-sized flat objects (as opposed to discrete points or
curves) is impossible with just a single thin lens, and we provide a simple explanation of the Abbe
Sine Condition as a figure of merit in aberration correction. VC 2018 American Association of Physics Teachers.
https://doi.org/10.1119/1.5036939

I. INTRODUCTION far more elaborate diagrams. The second reason for limiting
our consideration to spherical aberration and coma is that
Ideal, stigmatic optical imaging (i.e., imaging such that all these aberrations are, respectively, zeroth and first order in
of the rays from one point on an object meet together at one object height h (i.e., the distance from a point object to the
point in the image) is difficult to achieve in practice and can optic axis) or off-axis angle h (the most appropriate measure
be almost as difficult to teach to undergraduates. The typical of off-axis deviation for telescopes imaging objects at infin-
optics textbook aimed at undergraduate physics courses ity), and are thus the most significant aberrations (though not
focuses its treatment of geometrical optics primarily (and the only significant ones) for many scientifically significant
largely appropriately) on paraxial rays and thin lenses, as imaging applications.
these limiting cases help beginners form intuition and help
practicing physicists in the initial design of optical experi- II. STIGMATIC IMAGING AND THE MECHANISM
ments. Aberrations are often given a survey treatment in a OF SPHERICAL ABERRATION
single chapter or section,1–4 with the goal of helping the stu-
dent recognize the tell-tale signatures of different aberrations One can design a curved surface that will stigmatically
and get a first exposure to a few concepts that will be useful image a chosen object point by invoking Fermat’s Principle,
in studying more advanced treatments of aberrations. To a which states that a ray of light will travel from point 1 to
large extent it is quite reasonable not to explore aberration point 2 via a path whose duration (travel time) is an extre-
theory in depth. Lens design is a well-established subject for mum relative to neighboring paths. A well-designed lens
specialists,5,6 and is mostly outside the scope of contempo- will make it so that all paths that light can take from point 1
rary physics research (although one does still encounter the to point 2 (via a passage through the lens) will take the same
occasional fundamental physics investigation centered on amount of time, irrespective of where the ray initially strikes
lens design7). Consequently, advanced textbook treatments of the lens. Consequently, any ray that hits the lens surface after
lens aberrations are mostly divided between in-depth theoreti- originating from point 1 will travel to point 2. Our task here
cal treatments8,9 and specialized engineering treatments.5,6 is to determine the shape of such a surface. For convenience,
Our goal in this article is to explore optical aberrations at we will begin our analysis by considering only a single
an intermediate level, for the benefit of people who desire curved surface. While a real lens consists of a pair of surfa-
some insight into why aberrations arise via an explanation ces, we can treat the image produced by one surface as a vir-
deeper than just “The lens has the wrong shape.” We will tual source for the second surface (as discussed below).
explore four key topics without the full mathematical It is particularly convenient to compare a path along the
machinery of aberration theory: The conditions under which optic axis with an off-axis path, as in Fig. 1. Since the speed
a lens can form a perfect stigmatic image of a point, the of light is c in vacuum and c/n in a medium with refractive
mechanism by which spherical aberration and coma arise index n, the time for light to pass from point 1 to point 2
through deviations from those conditions, the reason why a along the optic axis is (n1s1 þ n2s2)/c. Alternately, we could
single thin lens is insufficient for aberration-corrected imag- consider an off-axis path, which hits the lens a distance r
ing of a finite-sized flat object, and the relationship between from the axis and a distance z to the right of the center of the
the Abbe Sine Condition and aberrations.10,11 We will limit lens. Equating the time along the two different paths, we get
our consideration to spherical aberration and coma for two that
reasons. The first, a practical reason, is that it is possible to n1 s1 þ n2 s2 n1 l1 þ n2 l2
analyze the mechanism behind these aberrations while limit- ¼
ing one’s calculations to only those rays in a meridional c qcffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
n1 n2
plane, i.e., a plane that includes the optic axis and an off-axis ¼ r 2 þ ðs1 þ zÞ2 þ r 2 þ ðs2  zÞ2 :
object point. The other aberrations require more detailed cal- c c
culations for rays propagating outside of that plane, requiring (1)

487 Am. J. Phys. 86 (7), July 2018 http://aapt.org/ajp C 2018 American Association of Physics Teachers
V 487
Fig. 2. A ray of light (originating in air on the left) intersects a spherical
glass surface with a larger angle of incidence than if it had intersected an
asphere (whose radius of curvature at the optic axis matches that of the
sphere). (Dotted aspheric surface is shown for comparison.) Consequently,
the aberrated rays undergo a larger change in direction than rays striking the
ideal aspheric surface.
Fig. 1. Imaging by a single curved surface separating media with refractive
indices n1 and n2. The shape of the surface is chosen so that light takes the consequence of spherical surfaces differing from aspheres in
same amount of time to traverse from object to image along both the axial a systematic manner, being more curved than aspheres across
and off-axis paths. the entire surface, and with the magnitude of the discrepancy
increasing as one moves farther away from the center. The
In order to obtain the desired shape of the surface it is nec- result is that rays that strike the lens farther from the optic
essary to solve for either r as a function of z or vice versa. axis suffer accordingly larger deflections from their ideal
The algebraic manipulation involved is tedious (first isolat- path. The fact that the discrepancy becomes larger farther
ing one square root and squaring, then isolating another away from the axis is made clear in Fig. 2. Near the axis
square root and squaring), and in general gives us a fourth- (where paraxial rays propagate) the spherical and aspherical
order curve known as a Cartesian oval.12,13 Two key excep- surfaces are nearly identical, but farther from the axis the
tions to the fourth-order nature of the curve are when the spherical surface systematically deviates from the prescribed
object or image distance approaches infinity, in which case aspherical shape. This also partly explains why spherical sur-
we get a hyperbolic surface,14 or when the image is virtual faces are of practical utility in optics, as spherical surfaces
with magnification n1/n2, in which case the lens shape is a are nearly indistinguishable from aspherical surfaces close to
sphere and the object is said to be at a aplanatic point of the the axis. (An additional justification for the utility of spheri-
sphere.15,16 The most important point for our purposes is that cal surfaces will be offered in Sec. IV.)
there do exist surfaces which can perform real and stigmatic To this point we have only considered imaging by a single
imaging of a single point. These surfaces, whether fourth- aspherical surface (imaging from one medium to another),
order or hyperbolic, are generally referred to as “aspheres,” and the attendant spherical aberration that occurs when it is
to distinguish them from spherical surfaces that are typically approximated with a spherical surface. In practice, however,
easier to fabricate (and hence are more common in optical lenses typically have two surfaces, and are usually designed
engineering). to image from air to air, or from some other medium to either
Because of spherical surfaces’ greater ease of fabrication, air or vacuum (the media in which camera sensors are typi-
in many cases the desired aspheric surface is approximated cally embedded in the image plane), not from air to glass.
with a spherical surface that has the same radius of curvature The second surface is designed to take rays that were
as the desired aspherical surface in the vicinity of the optic directed to a point on the right (treated as a virtual object
axis. In Fig. 2, we show the effect on rays that hit an air- here) and redirect them to a different point (often but not
glass interface far from the optic axis: The spherical surface always a closer point) in a medium different from the mate-
that they encounter is more curved than the ideal aspherical rial of the lens. In order to derive the appropriate shape for
surface for the chosen pair of object and image points, and the second surface, however, it may be conceptually simpler
so they are more strongly deflected than they would be if the to consider virtual imaging of real objects, instead of real
lens were aspherical. This is the origin of the phenomenon of
spherical aberration.
If we zoom out from the scenario in Fig. 2, and look at a
ray-tracing simulation (performed with the software package
OSLO17) of rays striking that spherical surface in Fig. 3, we
see that while rays striking the air-glass interface near the
optic axis nearly intersect each other in the image plane (at
precisely the point where the theory of paraxial optics would
predict for image formation), the rays that strike the interface Fig. 3. Rays of light (originating from an on-axis point object in air on the
farther off-axis experience larger deflections, and therefore left) intersect a spherical glass surface (refractive index n ¼ 1.5) and enter
the glass medium. Rays striking the interface near the optic axis are focused
intersect the axis in front of the paraxial image point. very close to the paraxial image point, while rays striking the interface far-
Spherical aberration is thus not just a matter of spherical ther off-axis undergo large deflections, the signature of spherical aberration.
surfaces differing from ideal aspherical surfaces (any surface Simulations were performed in OSLO and exported to graphics software for
other than the prescribed one is different), it is specifically a color editing and insertion of text.

488 Am. J. Phys., Vol. 86, No. 7, July 2018 Alex Small 488
imaging of virtual objects. Such a scenario is illustrated in the same argument as above applies if one approximates the
Fig. 4. right surface with a spherical surface: The spherical surface
Rays of light originating from the object point on the right will be more curved than a Cartesian oval, and will hence
travel on the left as though they are all diverging from the bend rays more strongly toward the optic axis, resulting in
same image point; this is the definition of a virtual image. If spherical aberration.
we look at a spherical surface centered on the image point Interestingly, while both surfaces will be Cartesian ovals
(i.e., a wavefront) then all of the rays will have taken the in a lens designed for stigmatic imaging, we have some free-
same amount of time to reach that surface, just as rays dom in designing those surfaces: The left surface must be
diverging from a point would have taken the same amount designed to image from the left object point to some other
of time to reach that surface. The time taken may not be the point, but we have latitude in choosing that second point.
actual time that would be required to get from the image Whatever second point we pick, the right surface must take
point on the left to the wavefront on the right if the light rays that were aimed at that point and bring them to the
were traveling in a uniform medium with refractive index desired image point on the right. This degree of freedom will
n2, but it will at least be the same for all paths. We can become important in Sec. IV, when considering designs that
therefore write minimize aberrations.
n1 l1 þ n2 DO n1 s1 þ n2 DA
¼ : (2) III. THE MECHANISM OF COMA ABERRATION
c c IS RELATED TO SPHERICAL ABERRATION
As it stands, Eq. (2) does not specify a lens shape because in The key signature of the spherical aberration discussed
addition to the object-lens surface distances l1 it also contains above is that it can occur even for rays originating from an
the off-axis and axial distances DO and DA. However, because object on the optic axis. If the object is not on the optic axis,
the wavefront is diverging spherically from the image point it however, there are many more things that can “go wrong,”
follows that the time to get from the image point to the wave- i.e., many more ways in which rays can fail to intersect at a
front surface would also be path-independent (if the medium single object point. A full understanding of the multitude of
were uniform), giving us the following relationship: possible aberrations would require consideration of rays
propagating in many different planes. However, the basic
n2 l2 þ n2 DO n2 s2 þ n2 DA
¼ : (3) mechanism of one important aberration (called “coma,” for
c c reasons that will soon become apparent) can be qualitatively
Equation (3) also does not specify a lens shape, as it also con- understood by considering rays in the plane containing the
tains the off-axis and axial distances DO and DA. Fortunately, off-axis object and the axis through the center of the lens. It
if we subtract Eq. (3) from Eq. (2) we eliminate DO and DA, turns out that the mechanism of coma aberration is closely
leaving related to the mechanism of spherical aberration discussed
above.
n1 l1  n2 l2 n1 s1  n2 l2 In Fig. 5, we show four rays from an off-axis object: two
¼ : (4) extreme rays hitting the interface far above or below the axis,
c c
and two paraxial rays hitting the interface close to the axis.
This equation is indeed sufficient to specify the shape of the The paraxial rays, of course, combine to form an image with
right lens surface, and it is indeed a Cartesian oval, just like few noticeable aberrations. The extreme rays, however, will
the left lens surface. We can thus obtain stigmatic imaging of exhibit noticeable aberrations. The curved interface does not
a single point (i.e., imaging without spherical aberration) if “know” (at the risk of anthropomorphizing a lens) that these
we use a lens whose surfaces are Cartesian ovals. Moreover, two rays come from off-axis points. The rays are propagating
as though they came from on-axis points (denoted A and B for
the top and bottom rays, respectively). Consequently, they
experience spherical aberration, and get deflected toward the
axis more steeply than they would ideally (in the absence

Fig. 4. Formation of a virtual image by a single curved surface separating


media with refractive indices n1 and n2. The shape of the surface is chosen Fig. 5. Paraxial and extremal rays from an off-axis object pass through a
so that rays traversing both the on-axis and off-axis paths take the same lens, and because they either pass through the axis or can be extrapolated to
amount time to reach the wavefront (a circular surface centered on the image the axis they experience spherical aberration, being deflected away from
point). their ideal paths and toward the axis.

489 Am. J. Phys., Vol. 86, No. 7, July 2018 Alex Small 489
IV. ABERRATION CORRECTION
While some amount of aberration is almost inevitable in
lenses (especially spherical lenses), the magnitude of the
problem posed by aberrations is often (at least partially) cor-
rectable. In this section, we will discuss what it means to cor-
rect an aberration. We will find that standard correction
formulas for coma and spherical aberration give rise to very
similar lens shapes, consistent with our point about the com-
mon mechanism of these two aberrations.
If we consider the common case of a glass lens whose sur-
faces are spherical and whose object and image planes are
both in air, the designer still has (at least) three parameters
indicated in Fig. 7(a): The radii of curvature (R1 and R2) of
the left and right surfaces, and the lens thickness d (measured
from the left surface to the right surface, along the central
axis of the lens). There are many combinations of these
parameters that can give the same focal length f, as the focal
length is given by the lens maker’s equation
 
1 1 1 ðn  1Þd
ð
¼ n1 Þ  þ ; (5)
f R1 R2 nR1 R2

where n is the refractive index of the lens (the refractive


index of air is assumed to be very close to 1) and the radius
is positive if the surface bulges out to the left and negative if
the surface bulges to the right. Given that we have flexibility
in choosing R1 and R2, our goal is to choose them so as to
minimize the spot size.
Fig. 6. (a) Ray of light (originating from axial and off-axis point objects in
air on the left) intersect a spherical glass surface (refractive index n ¼ 1.5)
In order to minimize the spot size, we need a quantitative
and enter the glass medium. The image plane was chosen to minimize the model to predict a (possibly aberrated) ray’s arrival position
spot size in the image of the axial point object. The image of the axial point h2 in the image plane in terms of the object’s location in the
is symmetric, consistent with spherical aberration, while the image of the off- object plane (h1), the lens’ shape (i.e., radii of curvature R1
axis point is asymmetric in the manner of coma aberration. (b) A spot dia- and R2, thickness d), the imaging conditions (i.e., distances
gram (pattern of rays in the image plane) corresponding to the off-axis
s1 and s2 from object point to lens and from object to axis),
(bottom right in part (a)) image. The vertical asymmetry in the distribution of
rays, with somewhat more rays in the upper portion of the image, is a signa- and where the ray hits the lens (describable in terms of a dis-
ture of coma aberration. tance r and an angle / in the plane of the lens, as shown in
Fig. 7(b)). Treatises on aberration theory outline elaborate
formulas to describe these quantitative relationships. What
of spherical aberration). This results in an asymmetry in the matters here is that these formulas can be broken down into
distribution of arriving rays in the image plane. The same series expansions, and each term in the series has a different
asymmetry is present in quantitatively accurate ray-tracing cal- quantitative and qualitative form, corresponding to different
culations in Fig. 6(a). aberrations and hence to different shapes in the image plane.
This asymmetry is also apparent in the propagation of rays The first few terms in the series for h2 are
outside of the plane containing the object point and the optic
axis, as shown in Fig. 6(b). The extreme top and bottom rays h2 ¼ mh1 þ C30 r3 þ C21 r 2 h1 ð2 þ cos 2/Þ þ    : (6)
in Fig. 5 define the top and bottom of a pattern of rays strik-
ing the edge of the interface, and because they are not posi- The first term corresponds to the predictions of paraxial
tioned symmetrically with respect to the paraxial image optics, telling us that a ray will arrive in the image plane a
point the result is a comet-shaped image (hence the name distance mh1 from the axis, where m is the magnification of
“coma”). the image.

Fig. 7. (a) Schematic of a lens, showing the radii of curvature and the thickness. Note that by convention the right surface has a negative radius of curvature if
it is convex to the left. (b) Polar coordinate system used to describe the point where a ray intersects the left surface of a lens.

490 Am. J. Phys., Vol. 86, No. 7, July 2018 Alex Small 490
The next term, proportional to r3, is independent of h1,
which tells us that it will affect imaging of any object, regard-
less of its location with respect to the axis. The value of C30
depends on the magnification m, the radii of curvature R1 and
R2, the lens thickness d, and the refractive indices of the left
medium, lens material, and right medium. Qualitatively, this
term describes the effects of spherical aberration, but the
more formal name for it is “third order spherical aberration,”
because it is a cubic term (r3). If one were to do more detailed
calculations one would get additional terms proportional to
r5, r7, etc. that are required if one wishes to more accurately
predict the full effects of spherical aberration.
The third term’s dependence on h1 indicates that it only Fig. 8. Coddington shape factor q þ 1=q  1 divided by scale factor s2  s1 =
applies to off-axis objects. The angular dependence gives s2 þ s1 for correction of coma and spherical aberration, plotted as a function
of lens refractive index n for values typical of optical glasses.
rise to asymmetry in the image, just as in the case of coma
aberration, and the coefficient C21 describes the magnitude
of coma effects. However, just as the r3 term formally aberration is, for typical values of the refractive indices of
describes only third order spherical aberration, the h1r2 term lens materials, very close to that required for correcting third
formally describes a phenomenon referred to as “third order order coma. This is in keeping with our observation above
coma” (as it involves three powers of distances, r2 and h1). that the mechanism of coma aberration is directly related to
As with spherical aberration, one could compute higher- spherical aberration.
order terms (proportional to r4h1, r6h1, etc.) that give rise to Finally, the fact that radii of curvature can be chosen in
a similar pattern in the image plane and provide additional such a way as to at least partially mitigate aberrations is an
quantitative corrections. additional reason (aside from ease of fabrication) why spher-
The task of a lens designer is to choose R1, R2, d, and the ical lens surfaces are so frequently used. Not only do spheri-
lens material in such a way as to minimize at least some of cal surfaces provide satisfactory performance for paraxial
the coefficients in these series, for given object and image rays, but also for non-paraxial rays a series of spherical sur-
distances. We will not derive the forms of such coefficients, faces can be designed to mitigate aberrations. Indeed, profes-
but will simply quote other authors. Jenkins and White18 sionally designed lens systems typically consist of several
showed that (for a lens thin enough that a paraxial ray’s dis- lenses in series, with each lens’ pair of surfaces providing
tance r from the axis does not change appreciably while tra- degrees of freedom for a designer seeking to mitigate aberra-
versing the lens) minimizing the effects of spherical tion effects. There is no simple analytical theory to predict
aberration, i.e., minimizing the value of C30, requires the final arrival point of a ray in the image plane from the
   radii of curvature of the many surfaces used in a multi-lens
R2 þ R1 2ðn2  1Þ s2  s1 design, but with professional ray-tracing software the arrival
¼ : (7)
R2  R1 nþ2 s2 þ s1 point can be efficiently computed, and optimization algo-
rithms can be applied to search parameter space for a high-
The minus sign on the right side of Eq. (7) is directly quality design. The design process may be challenging, but
related to the common guideline19 that when a lens is used to the fabrication of the spherical surfaces is a manufacturing
produce a magnified image (i.e., s2 – s1 > 0) the flatter side task that has been refined over the years.
of the lens should face the object while the more curved side
faces the image (so that R1 > 0 and jR1 j > jR2 j). One intui- V. THIN LENSES, RADIANT FLUX
tive way of explaining this guideline is that this lens orienta- CONSERVATION, AND OFF-AXIS ABERRATIONS
tion minimizes the angle between rays from the object and
the normal to the lens surface, and thus reduces the effects of In Secs. I–IV, we established the mechanisms of spherical
aberrations. aberration and coma by computing the ideal lens shape for
Jenkins and White also showed20 that for a thin lens third stigmatic imaging between a pair of points and examining
order coma can be eliminated, i.e., C21 can be set to zero, if the effect of deviations from that ideal shape. That argument
 2   was grounded in the Principle of Least Time. We will now
R2 þ R1 2n  n  1 s2  s1 use the principle that the total power collected from an object
¼ : (8) by a lens should equal the power focused to an image by a
R2  R1 nþ1 s 2 þ s1
lens (more formally known as “radiant flux conservation”) to
In both of these formulas, s2  s1 =s2 þ s1 is called the “scale argue that simple thin lenses cannot produce sigmatic and
factor,”21 and R2 þ R1 =R2  R1 is called the Coddington real images of a flat object of non-zero size. This argument is
shape factor. We can thus compare the prescribed lens not a rigorous proof, and is not sufficient to exclude the pos-
shapes (q values) for the two different aberration corrections sibility of multi-element lenses without off-axis aberrations,
at a given magnification if we compute the ratio ðR2 þ R1 = but it does serve to motivate two important points in optics:
R2  R1 Þðs2 þ s1 =s2  s1 Þ (shape factor divided by scale fac- First, in practice, high-quality imaging is accomplished with
tor). (We choose to examine this ratio so that on the right a series of lenses, i.e., even a lens for which the thin lens
side we are comparing factors that only depend on the type approximation is accurate may not be the optimal lens for a
of aberration under consideration but not the specifics of given task in imaging. Second, as a general theoretical result,
the imaging situation.) In Fig. 8, we show that the ratio of when stigmatic imaging of a finite-size region is possible it
shape factor to scale factor required for correcting spherical is only possible for a curved surface, not a flat one.22

491 Am. J. Phys., Vol. 86, No. 7, July 2018 Alex Small 491
Suppose that on the left side of a lens we have a circular n1 h1 sin h1;max ¼ n2 h2 sin h2;max ; (11)
object of radius h1 radiating light isotropically in all direc-
tions, as shown in Fig. 9. Conservation of radiant flux is or
related to the conservation of a quantity called etendue, which
is proportional to the product of surface area and solid angle h2 n1 sin h1;max
¼ : (12)
for light propagating in an optical system. In an optical system h1 n2 sin h2;max
without scattering (which increases the number of paths light
can follow, and thus increases etendue) or absorption (which Equation (11) is an example of the Abbe Sine Condition (dis-
decreases the amount of light in a system), etenude is con- cussed in Sec. VI).
served.23,24 The qualitative interpretation of etendue conserva- Unfortunately, this is not a requirement that a thin lens
tion is that if we compute the optical power passing through a can satisfy, as we will now show. The fact that it is thin
surface (by multiplying the area of the surface and the range means that a ray enters and exits the lens at the same distance
of angles from which light is coming), and then see what hap- r from the optic axis, so we have
pens to the rays of light that hit that surface as they continue
traveling, expanding the area covered by the light rays means r ¼ s1 tan h1;max ¼ s2 tan h2;max : (13)
that each element of surface area receives light from a smaller
range of angles. Conversely, focusing the light to a smaller Because the magnification of a thin lens is proportional to
area means that each element of the surface gets light from a h2/h1 ¼ n1s2/n2s1, we can rewrite Eq. (13) as
wider range of angles.
In the limit of an object small enough that the maximum h2 n1 tan h1;max
¼ : (14)
collection angle is approximately the same for all points on h1 n2 tan h2;max
the object, the etendue can be computed as
ð Equations (12) and (14) cannot be simultaneously satisfied
unless either n1/n2 ¼ h2/h1 (a special case) or tan h1;max
G1 ¼ ph21 cos hn21 dX
 sin h1;max and tan h2;max  sin h2;max (i.e., the paraxial
ð h1;max case). Thus, aside from the special case of magnification
¼ 2p2 h21 n21 cos h sin hdh ¼ p2 h21 n21 sin2 h1;max : equal to the index ratio a single thin lens cannot stigmatically
0 image a flat region of non-zero area in the non-paraxial
(9) regime, because the area over which the lens would spread
the collected light does not correspond to the magnification
The factor of ph21 cos h is the apparent surface area of the that one would calculate in the paraxial regime from object
object to a section of the lens that is viewing the object at an and image distances.
angle h. Even if every point on the object is emitting iso-
tropically, a surface element on the lens is “seeing” the VI. THE ABBE SINE CONDITION
object from an angle, and the cross-section presented to the
lens (which determines the integrated flux of energy through As mentioned above, Eqs. (11) and (12) are examples of a
the surface) will be ph21 cos h. The factor of n21 in front of the more general condition called the Abbe Sine Condition
differential element of solid angle reflects the fact that the (ASC), sometimes also called the Optical Sine Theorem,2
solid angle spanned by a bundle of rays can change due to which should be satisfied in a well-designed imaging system.
refraction. If this condition is to hold irrespective of the lens aperture
We will assume stigmatic imaging, i.e., all of the rays from (i.e., we should be able to place an adjustable pupil in our
each point on the object on the left side will be focused to system to vary the range of rays that the lens receives) then
corresponding image points on the right side. Consequently, we can write it slightly more generally as
the etendue on the right can be computed in the same manner
as on the left n1 h1 sin h1 ¼ n2 h2 sin h2 ; (15)

G2 ¼ p2 h22 n22 sin2 h2;max : (10) or

Equating etendue on the left and right gives (after some sim- h2 n1 sin h1
¼ : (16)
ple algebraic manipulation) h1 n2 sin h2

There are three basic questions to ask about the ASC:


What is its physical origin, what is its physical meaning, and
why is it a result worthy of attention?
In regard to the source of the ASC, it can be derived via sev-
eral distinct lines of reasoning, including Fermat’s Principle in
geometrical optics,1,9 wave optics,4,11 and even the resolution
of a celebrated thermodynamic paradox involving focused
light.24,25 The fact that the same conclusion can be reached by
reasoning from at least three different sets of premises indi-
cates that the result is a deep one worthy of further investiga-
tion. For full, careful derivations it is best to examine the
Fig. 9. Object and image sizes and ray angles for imaging of a finite-sized aforementioned references; here we will briefly mention the
object. basic ideas of the geometrical and wave derivations.

492 Am. J. Phys., Vol. 86, No. 7, July 2018 Alex Small 492
In the context of geometrical optics, it can be derived from problem affects images formed by any object point, whether
an assumption that the lens has been designed for stigmatic on-axis or off-axis, and the coma aberration affecting images
imaging of an axial object point. One considers how the travel of off-axis objects thus arises from the same mechanism as
time changes when the object and image points move off-axis the spherical aberration affecting images of on-axis objects.
and the ray’s path is varied, and requires that the travel time Fortunately, while a spherical lens does not have the proper
be the same for all paths. When derived from wave optics, the shape for stigmatic imaging, a spherical lens does have two
ASC amounts to a relationship between magnification and degrees of freedom, and a system of several lenses has pro-
spatial frequency (since the spatial frequency of a light wave portionately more. Therefore, even if the focal length of the
is proportional to n sin h). If we produce a magnified image system is held constant there are still unconstrained degrees
(h2/h1 > 1) then the length scales in the image must be propor- of freedom, so it is possible to either minimize spherical aber-
tionately larger and thus the spatial frequencies in the image ration or eliminate coma while holding the focal length and
(proportional to n2 sin h2 ) must be proportionally smaller than magnification constant. One means for evaluating the quality
the spatial frequencies in the object (n1 sin h1 ). of an optical design is the Abbe Sine Condition, which relates
Regardless of the approach that one uses to derive the the angle of a marginal ray (as well as any other ray) to the
ASC, its basic optical content is a requirement that for stig- system’s magnification. The ASC is a rather deep concept,
matic imaging of off-axis points the magnification ratio h2/h1 being derivable from several distinct lines of reasoning, but it
should be the same for all rays, irrespective of the angle h1 at can also be motivated by a simple calculation that illustrates
which they were traveling in medium 1. It follows that the a shortcoming of a single thin lens, namely, that it cannot
ratio n1 sin h1 =n2 sin h2 should also be the same for all rays. conserve radiant flux while imaging stigmatically. This pre-
The ASC thus relates an intuitive requirement that the mag- sentation is not intended as a substitute for intensive training
nification of an image should be the same regardless of in optical engineering, but it is hoped that it might serve as a
where the rays hit the lens surface to a less intuitive state- starting point for those who wish to pursue geometrical optics
ment about the sines of angles made by rays. beyond the coverage typically given in physics classes.
While the requirement on the sines of angles may not be
intuitive, it has a very practical application. Aberrations gen- ACKNOWLEDGMENTS
erally have their worst effects for the ray striking the edge of
the lens at the largest angle, commonly known as the I thank my students for being patient as I have tried to
“marginal ray.” In Fig. 5, the marginal ray is the bottom ray, teach aberrations and ray tracing to the best of my ability
passing through point B. The ASC thus gives the optical engi- (hopefully improving each year). I thank my many colleagues
neer a simple and quantitative target: Optimize a lens combi- in the Optical Society of Southern California for showing me
nation so that for the marginal ray the ratio n1 sin h1 =n2 sin h2 just how rich of a subject geometrical optics can be.
is as close as possible to the magnification h2/h1. Some work- a)
ers even refer to the discrepancy between the magnification Electronic mail: arsmall@cpp.edu
1
and the ratio of sines n1 sin h1 =n2 sin h2 as the “Offense F. A. Jenkins and H. E. White, Fundamentals of Optics (McGraw-Hill,
New York, 1976).
Against Sine Condition” (OSC) and use it as a design metric. 2
E. Hecht, Optics (Addison-Wesley, Reading, MA, 2002).
In fact, one can relate the OSC to the magnitude of coma 3
F. L. Pedrotti, L. S. Pedrotti, and L. M. Pedrotti, Introduction to Optics
aberration, and derive Eq. (8) from the requirement that the (Pearson Prentice Hall, Upper Saddle River, NJ, 2007).
OSC be zero.6 4
Ariel Lipson, Stephen G. Lipson, and Henry Lipson, Optical Physics
Additionally, the ASC gives a method for quantifying the (Cambridge U.P., Cambridge, 2010).
5
order of the aberration in a system. For any ray (not just the W. Smith, Modern Optical Engineering, 4th ed. (McGraw-Hill Education,
New York, 2007).
marginal ray) we can write the angle h2 on the right as a func- 6
R. Kingslake and R. B. Johnson, Lens Design Fundamentals (Elsevier
tion of h1 or vice-versa, and then express n1 sin h1 =n2 sin h2 Science, Amsterdam, 2009).
as a power series in one of those angles. The zeroth order 7
J. B. Pendry, “Negative refraction makes a perfect lens,” Phys. Rev. Lett.
term should be equal to the magnification, and the goal for 85, 3966–3969 (2000).
8
the coefficients of higher order terms is to either set them to H. A. Buchdahl, An Introduction to Hamiltonian Optics (Dover
zero or else minimize them. Though we referred to “third Publications, New York, 1993).
9
Max Born, Emil Wolf, and A. B. Bhatia, Principles of Optics:
order coma” in Sec. IV, the cancellation of a factor in the Electromagnetic Theory of Propagation, Interference and Diffraction
ratio of odd functions (sines) means that the second order of Light, 7th ed. (Cambridge U.P., New York, 1999), pp. xxxiii,
term in n1 sin h1 =n2 sin h2 arises from the third order term in p. 952.
10
the expansion of the sine, and hence is related to third order J. Higbie, “Microscope resolution,” Am. J. Phys. 49, 40–42 (1981).
11
coma. If that coefficient is zero then third order coma has Jack Higbie, “Abbe’s sine theorem from a thermodynamic and Fourier
been eliminated. transform argument,” Am. J. Phys. 49, 788–789 (1981).
12
Edward Harrington Lockwood, A Book of Curves (Cambridge U.P.,
Cambridge, 1961).
VII. CONCLUSIONS 13
Ref. 2, p. 234.
14
Ref. 2, p. 151.
The chief purpose of this article is to meet the needs of an 15
Ref. 4, p. 83.
instructor who is not a specialist in geometrical optics but 16
Ref. 1, p. 166.
17
nonetheless wishes to explain aberrations to students on A free version is available for educational use <http://www.lambdares.
some level more sophisticated than merely noting that most com/oslo>.
18
lenses have the wrong shape. The key concept is that the Ref. 1, p. 159.
19
For instance, Ref. 3 p. 446. This guideline is often found in manuals pro-
shape of a spherical lens is not just incorrect but systemati- duced by optics manufacturers (e.g., https://www.edmundoptics.com/
cally incorrect, so that rays striking the surface farther off- resources/application-notes/optics/understanding-optical-lens-geometries/,
axis are systematically bent more sharply away from the accessed March 27, 2018) and even on Wikipedia, which states that “a
path that would be prescribed for stigmatic imaging. This plano-convex lens, which is used to focus a collimated beam, produces a

493 Am. J. Phys., Vol. 86, No. 7, July 2018 Alex Small 493
22
sharper focal spot when used with the convex side towards the beam A. Walther, “Irreducible aberrations of a lens used for a range of magnifi-
source.” (https://en.wikipedia.org/wiki/Lens_(optics), accessed March 27, cations,” J. Opt. Soc. Am. A 6, 415–422 (1989).
23
2018). Roland Winston, Juan C Mi~ nano, Pablo G Benitez, et al., Nonimaging
20
Ref. 1, p. 165. Optics (Academic Press, Burlington, MA, 2005).
21 24
For finite object and image distances, it is straightforward to compute the T. J. Yoder and G. S. Adkins, “Resolution of the ellipsoid paradox in
scale factor. If either the object or image is at infinity it is necessary to thermodynamics,” Am. J. Phys. 79, 811–818 (2011).
25
take a limit. In the case of an object at infinity, s2  s1 =s2 þ s1 ! 1, and W. T. Welford and R. Winston, “The ellipsoid paradox in
for an image at infinity s2  s1 =s2 þ s1 ! 1. thermodynamics,” J. Stat. Phys. 28, 603–606 (1982).

494 Am. J. Phys., Vol. 86, No. 7, July 2018 Alex Small 494

You might also like