Optics and Geometry

with Applications to Photography

Tom Davis
[email protected]
http://www.geometer.org/mathcircles
November 15, 2004
1 Useful approximations
This paper can be classied as applied mathematics since every equation well discuss
has applications to useful physical objects, such as telescopes, cameras, prisms, eye-
glasses and so on. As is the case with almost any application of mathematics to the real
world, we will be making certain approximations to be able to solve the equations. For
that reason we realize that our solutions will ususally be slightly in error, but that the
errors can be made quite small. The main job of an engineer who is designing a real
optical system is to create a physical situation that reduces the size of these errors to an
acceptable level.
As an example, well talk about the thin lens approximation, where we will assume
that the lens is innitely thin. Obviously no real lens is innitely thin, but as long
as the thickness of the lens is small enough compared to the distances to the objects
being viewed and imaged, this approximation may be good enough. Well make further
assumptions about the nature of light, about the perfection of our lens material, and so
on. When its important, well point out places where these assumptions may cause
trouble in a real physical system.
OO
BB
AA XX

Figure 1: Trigonometric Approximations

We will make a couple of mathematical approximations from time to time as well. The
most important ones concern the calculations of angles and of the trigonometic func-
tions of those angles. We will always measure angles in radians rather than degrees.
1
Consider the angle

AOB in Figure 1. If the length OA = OB = 1, then the length
of the arc AB is the measure of the angle measured in radians. (If the length of
OA happens not to be 1, then the measure of in radians is the length of the arc AB
divided by the length of OA.) Similarly, if OA = OB = 1, then sin = BX/OB =
BX, since OB = 1. It should be obvious from the gure, and it can be proven
mathematically, that as the size of the angle gets smaller and smaller, the length BX
is closer and closer to the length of the arc AB. In fact, we can write this mathmatically
as:
lim
0
BX
AB
= lim
0
sin

= 1.
If we multiply through by , then we can see that for very small values of , sin ,
and for that reason, when we do deal with very small angles, we will often replace
instances of sin by in the calculations.
It should also be obvious from the gure (and can, again, be rigorously proven math-
ematically) that the lengh of the line segment AB and the length of the arc AB are
also very similar. Thus if OA = OB = 1, we can approximate the angle as the
length of the line segment AB, or, if OA happens not to be 1, we can approximate
by AB/OB.
We will use the two approximations above repeatedly in what follows.
2 Images and Virtual Images
This paper will consider properties of lenses, where the word lens is used in a gen-
eral sense. We are going to examine how light is bent as it passes through different
materials. Obviously, the most common example is the bending of light through the
lenses of cameras, binoculars or microscopes that form images on lm or on the retina
of your eye.
The most important feature of lenses is that they are able to form images of objects
in the world. As an example, consider what happens when you use a camera to take
a photo. Lets consider one of the most boring possible photos: it is pitch-black in a
room except that there is a point-sized light source and the camera is pointed at a that
source. Your goal is to produce an image on the lm that is pitch-black except for a
single dot where the light from the point-source arrives on your lm.
The tiny light sends off photons in all directions, so rst imagine what would happen
if there were no lens, but rather that the lm were pointed at the light and exposed.
Photons would arrive more or less uniformly across the surface and the whole area of
the lm would be uniformly exposed. With the lens, however, some collection of the
photons will happen to hit the circular lens on your camera, and what we would like
the lens to do is to bend all of those photons back so that they hit a single point on the
lm, as shown in Figure 2. The point O is the location of the objectin this case,
the light source. The point I is the imagethe place where the light from the object
is focused. If the lm happens to lie at point I, the point source will produce a point
image, but if the lm happens to be a little in front of or behind point I, the light from
2
OO II
Figure 2: Object and Image
the source will produce a circle of light on the lmin other words, a blurred image.
If the lm is close to I, the blurring will be small, and if its far away, there will be a
lot of blur. When you focus a camera, you are basically changing the distance between
the lens and the lm so that the object you are interested in makes point sources of the
object focus to point images on the lm.
Figure 2 shows the effects of the lens on a single source of light; in reality, a perfect
lens will focus an entire plane of point sources onto the lm plane at the same time,
so that an entire plane in the real world is in perfect focus on the lm. For real lenses,
however, nothing is perfect, so the world-plane thats in perfect focus will not, in fact,
be a perfect plane, but will be some curved surface in 3-space. And the focus will never
be exact, either, but with expensive lenses, the errors can be made to be quite small.
2.1 Virtual Images
I
v
I
v
OO
Figure 3: Object and Virtual Image
The lens illustrated in Figure 2 bends the light from the source so that it forms a real
image where the photons diverging from the source meet again, hopefully on the lm
or on your retina. Other lenses, or locations of the object, like the one illustrated in
3
Figure 3, may cause the photons from the source to diverge instead of converge. In
this case, the diverging photons will diverge forever, and will never meet, but if such
a lens were perfect, those diverging lines would appear to be diverging from a single
point, labeled I
v
in the image. This is not a true, or real image, but the point is called a
virtual image.
A lens that forms a virtual image may at rst seem useless, but imagine that it is a part
of a series of lenses. The next lens along would see the light from the source at O not
as if it were coming from O, but as if it were diverging from the point I
v
. The dotted
lines in the gure simply indicate the direction of the diverging lines; the solid lines
represent the true paths of the photons.
2.2 Cones of Light
If you examine either of the gures 2 or 3 you see a pattern: light is emitted in all
directions fromthe object, but the part of interest is the collection of photons that strike
the circular lens. This light lls a cone-shaped volume of space.
When the light in the cone strikes a lens, the light coming out may have changed
direction somewhat but it now ts inside another cone-shaped piece of space. The
cone may point in the opposite direction, as in gure 2 or it may point in the same
direction but with a different angle, as in gure 3. Thus in a system of lenses, every
time the light arrives at the next lens, it is in the form of a cone, either getting larger or
smaller (or, if the beams happen to be parallel, its like an innite cone or cylinder).
What every perfect lens does is to take light coming from one side in a cone-shaped
volume and emit it on the other side in a different cone-shaped volume. If we know
how to calculate the new cone shape from the old one, we can trace the light through
lens after lens to nd the resultant cone shape after passing through any number of
lenses.
In fact, its not lenses that change the cone shapes; its lens surfaces, so to gure out
how a simple two-sided lens works, well merely need to gure out how the cone shape
is changed by each of the two surfaces.
Since this is a short paper, we will only do one sample calculation for each example, but
the formulas we derive will be very general in the following sense: Well assume that
the object is on one side of the lens and the image is on the other and well call those
distances to the lens o and i for the object and image distances. If the image happens
to lie on the same side as the object (as in the case of the virtual image in gure 3) the
same equations will work, but they will give a negative value for i. Similarly, when we
work out just how a lens ground with a certain radius r affects the light, well obtain
an equation, and if the lens surface happens to be concave instead of convex, the same
equation will work, but youll need to put in a negative value for r to obtain the correct
result.
4
2.3 Depth of Field
A camera with a perfect lens will focus every point on some focus plane in front of it
to a point on the lm plane that lies within the camera. Only points on the focus plane
will be focused perfectly. Points in the world in front of or behind the plane of perfect
focus will not form point images on the lm, but rather will form small circles.
The best way to visualize this is to think of the photons that come through the lens and
are bent back to a point as lying inside a cone of light. If the tip of the cone is at the
lm plane, we get perfect focus, but if the tip is a little in front of or behind the lm
plane, then photons from the object lie in a circle on the lm. The farther the tip of the
cone is from the lm plane, the larger the circle. In a camera, this circle is sometimes
called the circle of confusion.
Human eyes are not perfect either, and if the circle of confusion is small enough, we
will still consider the focus to be perfect, so objects near the focus plane also form
images that are almost in focus. Camera lens manufacturers dene a concept called
depth of eld which indicates the error allowable from the focal plane that still pro-
duces images that are good enough to seem perfect to humans.
Finally, notice that fat cones of light will allow less error than skinny cones, and the
cones of light can be made skinnier by restricting the amount of the lens that is used to
be a small part near the center. The diaphragm of the camera performs this function,
so by having a smaller aperture through which the light can travel, a photographer
can obtain a larger depth of eld, which may or may not be good, depending on the
goal of the photo.
Pinhole cameras, in fact, can be built that have no lens at all, but just use a tiny hole in
place of a lens. If the hole could be made arbitrarily small, a pinhole camera could form
arbitrarily good images. The problem is that light only approximately follows straight
lines and is bent slightly at the edges of the diaphragm or hole in a pinhole camera due
to a property called diffraction, so there is a limit to how good even a pinhole camera
can be.
3 Index of Refraction and Snells Law
Photons of light travelling through a vacuum move in a straight line at the speed of
light, which is about 300, 000 kilometers/second which physicists often denote by the
letter c. Each photon can be thought of as a wave packet with some wavelength.
Light travelling through any other material, like water, plastic or glass, does not move
with velocity c, but somewhat slower, and the ratio of c to the true speed of light in
the material is called the index of refraction of that material. Here are the indices of
refraction for a few common materials:
5
Material Refractive index
Vacuum 1.0
Air 1.0003
Water 1.33
Glass (zinc crown) 1.51
Glass (heaviest int) 1.89
Diamond 2.42
This means, for example, that light passing through a vacuum travels 1.33 times as fast
as it does through water. There are many different types of glass which allows lens
makers to construct lenses with very different properties. As we shall see shortly, the
index of refraction is closely related to how much light bends when it enters or leaves
a material, and the extremely high index of refraction for diamond offers some clue as
to why those precious stones are so interesting to look atit is very different from glass
and produces very different visual effects.
3.1 Snells Law
OO

2
BB
AA
CC
A A
B B
C C
Figure 4: Snells Law
Simply by knowing the relative speeds of light through the two materials and knowing
that light is a wave phenomenon, we can calculate the amount by which it will be bent
as it moves across a surface from one material to another. Consider a straight line
interface between two different materials as illustrated in gure 4.
The line CC

is the interface between the materials. The two lines with arrows on
them indicate that light is coming in from the left and leaving on the right at a slightly
6
different angle. The line AA

is is called the surface normal and is perpendicular to the

surface-surface interface and the regularly-spaced lines perpendicular to the lines OB
and OB

indicate the crests of the waves of the moving light. The length
1
= OB, for
example, is the wavelength of the light arriving from the left, and the length
2
= OB

is the wavelength of the light leaving the interface to the right.

The angles of incidence and refraction,
1
=

BOA and
2
=

B

OA

, respectively,
are measured from the surface normal AA

.
It is easy to see that

OCB =
1
and that

OC

=
2
. Since the waves are
uniformly spaced, OC = OC

= d, and simple trigonometry gives us: sin

1
=
1
/d
and sin
2
=
2
/d. Combining these last two yields:
sin
1

1
=
sin
2

2
. (1)
Since it requires the same amount of time for each wave crest to pass in either material,

1
n
1
=
2
n
2
. We can use this in combination with equation 1 to obtain Snells Law:
n
1
sin
1
= n
2
sin
2
. (2)
3.2 Total Internal Reection
Imagine that we have the same situation as in the rst example in the section 3 where
two materials meet along a straight line. If the light is coming from the side with a
larger index of refraction, we recall that as it enters the less dense medium, the light
speeds up and is bent away from the normal vector by a certain amount, depending on
the relative sizes of the indices of refraction. For some incidence angle, this angle of
refraction will equal 90

which means that the refracted light does not even enter the
material and if light strikes at this angle or greater, no light enters the other material.
This angle is called the critical angle.
Energy doesnt just disappearit has to go somewhereand what happens is that instead
of passing through the surface, all of the light is simply reected from the surface. This
doesnt happen all of a sudden, however. What really occurs is that as the incidence
angle gets shallower and shallower, more and more light is reected.
If you shined a laser beam at a surface, for example, and started having it point at 90

to the interface, but then gradually tilted it, you would nd that the more it was tilted,
the closer the refracted beam would be to being parallel to the surface interface, and
at the same time, the more light would be reected rather than passing through. When
the angle nally reached the critical angle, all of the laser light would be reected.
At what angle does this occur? To nd out, all we need to do is to set
2
to 90

in
Snells Law (equation 2):
sin
1
= n
2
/n
1
.
Since material 2 has a lower index of refraction, n
2
< n
1
, so this has a solution. If the
light is coming from the side of the interface with a higher index of refraction there can
be no solution, since the sine function can never be larger than 1.
7
3.3 Prisms and Spectra
With real light, of course, the situation is a bit more complicted. It turns out that
the speed of light in a material depends on the wavelength (which corresponds to the
color) of the light. Higher energy photons travel faster, so the index of refraction for
blue light is less than the index of refraction for red light, for example. (Blue light has
higher energy photons than red light.)
Because of this variation of index of refraction with wavelength, if a beam of white
light (which is a mixture of all wavelengths) hits the surface of a prism, each different
color of light is bent by a different amount, and the result is a spectrum.
4 Focus through a Surface
oo ii
OO
AA
VV II

2
rr
Figure 5: Focus distances
Let us consider the situation illustrated in Figure 5 where light is emitted from the
point O (the object) and then passes through a spherical lens surface of radius r and is
focused at the point I (the image). The length o is the distance from the object to the
surface of the glass at point V and the length i is the distance from the surface of the
glass to the point I where the image is focused.
In this example, we assume that everything to the left of the lens surface is one material
(say air) and everything to the right is another (say glass), so in this situation, we
assume that the image is actually formed inside a giant lens. Later well use the results
here to show how normal lenses work where the focus is in the air on the other side
of the lens. But the situation above is general, and all that we really require is that the
index of refraction is the same for all points on the left of the glass surface and is the
same on all points to the right. Well let n
1
be the index of refraction on the left and n
2
be the index of refraction on the right.
Consider any ray of light emitted from point O that strikes the lens surface at point A.
Due to the different indices of refraction on both sides of the lens surface, the light will
be bent and may focus at a point I on the other side of the surface, forming the various
angles and lengths indicated in Figure 5.
Simple Euclidean geometry gives the rst two equations below and Snells Law yields
8
the third:

1
= +
=
2
+
n
1
sin
1
= n
2
sin
2
Since were going to assume that point A is quite close to point V so sin
1

1
and
sin
2

2
, we can replace the equation due to Snells Law with:
n
1

1
n
2

2
.
With a little algebra, we can eliminate
1
and
2
from the above equations, and obtain:
(n
2
n
1
) n
1
+ n
2
. (3)
Finally, we can denote by h the length of the segment AV and substitute the small-
angle approximations: h/o, h/r and h/i into equation 3 where all the
instances of h cancel to obtain:
n
2
n
1
r

n
1
o
+
n
2
i
. (4)
Equation 4 relates the distance fromthe object to the image given the relative indices of
refraction in the two media on both sides of the surface. Although we have not worked
out all the cases, this equation is true for concave or convex lenses (where a concave
lens would be indicated by a negative value of r), and where the distance of the image
from the lens surface may be negative (on the same side of the surface as the object,
meaning that it is a virtual image).
5 Lens Formulas
In section 2 we noted that as light passes through perfect lenses, the emitted photons
always appear to be diverging toward a point or converging to a point. Thus if we
consider the result of passing light through a series of lens surfaces, at each surface
we can simply calculate how the diverging or converging light will be refocused by the
new surface. Thus the calculation above is sufcient, for perfect lenses, to work out the
charactertistics of any optical system composed of spherical lenses simply by tracing
the light through one surface at a time.
In this section we will do exactly thatwe will consider a single lens where light enters
from air (or a vacuum), passes through a lens, both of whose surfaces are spherical,
and enters air (or a vacuum) on the other side. We will do so by using the result from
the previous section as we trace the path of light from a source, through a surface into
glass, through the glass, and then out though the opposite lens surface.
In gure 6 we will assume that the object is at the point O and that the light passes
through a lens having two surfaces of radius r
1
and r
2
centered at C
1
and C
2
, respec-
tively, and that the surfaces are a distance l apart.
9
o
1
o
1
r
1
r
1
r
2
r
2
i
2
i
2
ll i
1
i
1
C
1
C
1
v
1
v
1
C
2
C
2
v
2
v
2
I = O I = O OO II
Figure 6: Fat Lens with Two Surfaces
Light from point O passes through the rst surface which is a distance o
1
away and the
beams are bent to form an image (or in the case of this gure, a virtual image) at the
point I

which is at a distance of i
i
from the original surface.
Now we just assume that the light is emitted from I

as if it were a new source, and

see how that light will be affected as it passes through the other surface. The light will
nally converge at a point I, and we want to nd a relationship between the various
distances and the radii of the two different lens surfaces.
To simplify things slightly, well assume that the lens is glass and has an index of
refraction n and that the material outside the lens is a vacuum with index of refraction
1. (Air has an index of refraction of about 1.0003 so if the lens is used in air instead of
in a vacuum, the errors will be very slight.)
Using the formula from section 4, and under the assumption that the rst image is
virtual (and therefore that I

is to the left of v
1
) we obtain
1
:
1
o
1

n
i
i
=
n 1
r
1
. (5)
If we analyze light assumed to be emanating from I

and passing through the second

surface (so its distance to that surface is i
1
+ l) using the same method, we obtain a
second equation:
n
i
1
+ l
+
1
i
2
=
1 n
r
2
. (6)
Combining equations 5 and 6, and letting l = 0 (this is the thin lens approximation),
1
It works out ne if the rst surface forms a real image to the right of the lens surface, since that will ip
the sign on the second fraction below, but will measure i
1
in the opposite direction, effectively negating it.
10
we obtain:
1
o
1
+
1
i
2
= (n 1)

1
r
1

1
r
2

. (7)
The focal length of a lens is dened to be the distance of the focus from the lens
assuming that the incoming light is parallel, which is equivalent to stating that the
source is innitely far away. If we set o
1
to in equation 7 we obtain what is known
as the lens makers equation, where f is the focal length of the lens:
1
f
= (n 1)

1
r
1

1
r
2

. (8)
Thus if you know the two radii of the lens surfaces and the index of refraction of the
glass youre using, you can calculate to fairly good accuracy the focal length of the
lens produced. In the equation above, we make the assumption that the radii r
1
and r
2
are measured from centers that lie to the right of the surface. If the center lies to the
left, the radius will be negative. For the double-convex lens in gure 6, for example, r
1
is positive and r
2
is negative (so in this case both fractions to the right of the equality
signs will be positive).
If you are not a lens maker, but just purchase a lens with xed surface radii and glass
with a xed index of refraction then the term on the right of both equations 7 and 8 is a
constant, and equation 8 tells us that the constant is 1 divided by the focal length of the
lens. Keeping this in mind, we have derived what is known as the thin lens equation
where i and o are the distances of object and image on opposite sides of an innitely
thin lens:
1
o
+
1
i
=
1
f
.
6 Consequences of the Thin Lens Equation
Equation 5 contains a huge amount of information. Light emitted from a focus point
will leave the opposite side of the lens in parallel. Light arriving in parallel will focus
at a distance f from the lens on the other side. Light emitted from a point f/2 from
one side of the lens will focus at the same distance on the other side. As the object gets
closer to the lens, the image moves away from the lens until the object is closer than
the focal length in which case the lens fails to converge the light and a virtual image is
formed on the same side of the lens as the object (and hence has a negative distance to
the lens).
An interesting feature of lenses is their magnication. The magnication is simply the
ratio of the size of the image to the size of the original object. If the image is twice as
big as the object, we say that the magnication is 2. Given a lens with focal length f,
how can we calculate the magnication of an object placed at a distance o from that
lens?
See gure 7. From the thin lens equation, we know that the only way to obtain a real
image (as opposed to a virtual image) is to have the object (and hence image) farther
11
yy
y y
x x xx
OO II ff ff
Figure 7: Magnication
away from the lens than the focal length. Suppose the object has height y and is x
beyond the focal length. This will form an image of height y

a distance x

beyond the
focal length on the other side of the lens.
To measure the magnication of this setup, y

/y, we need to gure out how big y

is,
compared to y. If we can trace any two rays from the top of the object to the image, we
can determine the height y

of the image. But we know that rays arriving parallel to the

axis of the lens must pass through the focus, and rays passing through the focus of the
lens must leave parallel to the axis, and those two rays are drawn in gure 7. Simple
geometry of similar triangles yields the equations: y

/x

= y/f and y/x = y

/f. Thus
the magnication y

/y is given by:
y

y
=
f
x
=
x

f
.
We can cross-multiply the nal equality above to obtain xx

= f
2
.
7 Real Lenses
Real lenses are not innitely thin, and have plenty of other characteristics that make
them not quite obey the equations we have derived in this article. In any real lens or
even optical system composed of a series of lenses, there are some deviations from per-
fect images, and the general term used to describe this sort of error is lens abberation.
Lens abberations come in various types; for example, chromatic abberations, spher-
ical abberations, astigmatism, eld curvature, and so on. Lens makers usually try
to reduce all of these at the same time by means of clever design.
The rst sort of error encountered is that real lenses generally have spherical surfaces
which is not the perfect shape, but is a shape that is incredibly simple to produce very
accurately. It is possible today to purchase non-spherical lenses for specic applica-
tons, but they tend to be quite expensive. The reason that spherical lenses are so easy
12
to produce is that if you rub two materials together until they slide smoothly in any
direction, the surface between them has to be either a plane or a portion of a sphere.
Lenses can be ground by placing an abrasive material between two pieces of glass and
rubbing the glass pieces together in every direction.
Figure 8: Optical System
What you will nd in any real optical system where it is important to reduce the abber-
rations is that instead of using a single lens, a series of lenses is used, where the series
is arranged in such a way that the abberations in one lens tend to be cancelled out by
another. The glass in the different lenses can be made of different materials having
different indices of refraction, the lenses may be of different shapes, and so on. You
have probably seen pictures of such lens systems like the one illustrated in gure 8 in
advertisements for camera lenses, et cetera.
The rst thing to notice for such real systems is that even if you assumed that the lenses
were innitely thin, the combination certainly is not. Thus your formulas for the thin
lens approximation will not work well if you have to work with optical systems instead
of with single lenses.
The following approximation is approximately true, however. There are two planes
in any such system perpendicular to the axis of the lenses called the principal planes,
and if you imagine the space between them as being squashed down to zero width, the
resulting system would behave like a single thin lens. This is true for any system of
lenses aligned along their axis.
What this means is that if youre not too worried about distortion, you can attach to-
gether any sequence of lenses, and it will behave like a single lens. Photographers take
advantage of this in many ways: you can screw a so-called diopter lens on the front
of your normal lens to turn it into a closeup lens. You can push the lens away from
the camera body with a hollow tube (called an extension tube) that will allow it to
focus much closer. You can put in a tele-extender between the camera body and a tele-
photo lens to multiply its focal length by some given amount (and thus turn a telephoto
lens into a super-telephoto lens, for example). You can reverse a lens and shoot photos
through it backwards, which typically increases its ability to take closeup shots. And
really, you are only limited by your imagination. Of course, only experiment will tell
whether the amount of distortion introduced is acceptable. The lens company goes to
a great deal of trouble to make sure the lens has low distortion in normal use, but they
13
probably dont do the calculations assuming that the lens is turned around, or sits three
inches farther from the camera than the lens mount would indicate.
7.1 Designing Lens Systems
Rather than do sophisticated mathematical calculations that work out the details of the
abberations of different lenses, modern lens design is more like trial and error. A set of
lenses is used as a starting point, and then a computer simply traces rays of light through
the lenses to evaluate the nal image. If its good enough, thats the design. If its not
good enough, a tiny change is made and the whole ray tracing exercise is repeated to
see if the new system is better than the old. If not, make a different correction and if
so, start from the new arrangement and try to improve it.
14