Content uploaded by Javier Hernández-Andrés
Author content
All content in this area was uploaded by Javier Hernández-Andrés on Oct 10, 2014
Content may be subject to copyright.
Virtual tunnels and green glass: The colors of common mirrors
Raymond L. Lee, Jr.a)
Mathematics and Science Division, United States Naval Academy, Annapolis, Maryland 21402
Javier Herna
´ndez-Andre
´s
Departamento de O
´ptica, Facultad de Ciencias, Universidad de Granada, Granada 18071, Spain
共Received 10 February 2003; accepted 25 July 2003兲
When a pair of common second-surface plane mirrors face each other, repeated mirror-to-mirror
reflections form a virtual optical tunnel with some unusual properties. One property readily analyzed
in a student experiment is that the color of objects becomes darker and greener the deeper we look
into the mirror tunnel. This simple observation is both visually compelling and physically
instructive: measuring and modeling a tunnel’s colors requires students to blend colorimetry and
spectrophotometry with a knowledge of how complex refractive indices and the Fresnel equations
predict reflectance spectra of composite materials. © 2004 American Association of Physics Teachers.
关DOI: 10.1119/1.1615524兴
I. INTRODUCTION
Across the centuries, mirrors have given us windows onto
unusual worlds. Some of these worlds are as familiar as they
are fundamentally odd: the one that lets us see our own face
or the one that shows us a transposed twin of the room in
which we stand. Yet the oddest mirror world may be the one
found between two plane mirrors that face each other. If
these mirrors are parallel 共or nearly so兲, then looking into
either of them brings us to the vertiginous edge of a visual
tunnel that seems to recede endlessly into the virtual dis-
tance.
We call this phenomenon of repeated mirror reflections a
mirror tunnel, and here we are interested in how one formed
by two common mirrors transforms the colors of reflected
objects. 共We define common mirrors as inexpensive second-
surface plane mirrors such as those found in homes.兲Under-
standing these transformations can teach students not only
about geometrical optics and spectral transfer functions, but
also how knowing the complex refractive indices of the
metal backing and glass in such mirrors lets us predict their
reflectance spectra via the Fresnel equations.1In a student
experiment, two especially instructive 共and surprising兲re-
sults are that common mirrors are not spectrally neutral re-
flectors, and their reflectance spectra depend on the absorb-
ing and reflecting properties of both the glass substrate and
its metal backing.
Because mirror tunnels are so visually compelling, we use
them as a pedagogical tool in this paper. In Sec. II we de-
scribe how to set up a mirror tunnel and observe its color
shifts, and in Sec. III we discuss the measured reflectance
spectra of some common mirrors and show how we use such
spectra to calculate the chromaticity trends seen in mirror
tunnels. In Sec. IV we use existing data on the complex
refractive indices of silver and soda–lime silica glasses in a
simple model that calculates the reflectance spectra and chro-
maticities for these mirrors. At a minimum, students should
already be familiar with the optical significance of the mate-
rials’ absorption, reflection, and transmission spectra. For
deeper insights, they also should understand how such purely
physical properties are translated into the psychophysical
metric of colorimetry. A very readable introduction to the
colorimetric system of the Commission Internationale de
l’Eclairage 共CIE兲is given in Ref. 2.
II. THE MIRROR TUNNEL DEMONSTRATION
One well-designed mirror tunnel is seen in Granada,
Spain’s Science Museum.3This virtual tunnel is produced
simply by placing two tall mirrors parallel to and facing each
other 共see Fig. 1兲. Viewers look at one mirror through two
small holes in the back of the other mirror; the holes are
spaced to match the average distance between human eyes.
This viewing geometry avoids the problem that can occur if
the viewer’s head is between the two mirrors, where it can
block the angularly smallest, most interesting regions of the
mirror tunnel.
On looking through the eyeholes, one sees a long progres-
sion of ever-smaller reflected images of the two mirrors that
plunges into the virtual distance, thus giving the illusion of
an infinite tunnel 共Fig. 2兲. If the mirrors are not exactly par-
allel 共as is true in Fig. 2兲, then the tunnel curves in the same
direction that the mirrors tilt toward each other.4After relish-
ing the vertiginous thrill of this virtual tunnel, we find an-
other, subtler feature: the farther we look into the tunnel, the
greener and darker its reflected objects appear. To track this
color and brightness shift in Fig. 2, we put a photographer’s
gray card at the base of one mirror. Although the card’s im-
age becomes progressively darker and more yellow–green
with successive reflections, no such shifts occur for distant
real objects seen by a single reflection in either mirror.
This simple observation lies at the heart of our paper, and
it begs the question that students must answer: why do re-
peated reflections change the mirror images’ color and
brightness? Geometrical optics is silent on this point, be-
cause it predicts only the images’ location and size. Thus, to
explain the brightness and color changes, we must consider
the physical optics of common mirrors. Because color de-
pends on spectral variability, we start by examining the spec-
tral reflectances and transmissivities of the metal and glass
used to make these mirrors.
III. MEASURED REFLECTANCE SPECTRA
AND COLORS OF COMMON MIRRORS
Common household mirrors have long been made by de-
positing a thin film of crystalline silver on the rear surface of
float-process flat glass.5This silver film is optically thick
共minimum thickness ⬃10
m兲, and its rear surface is pro-
tected from oxidation and abrasion by successive films of
53 53Am. J. Phys. 72 共1兲, January 2004 http://aapt.org/ajp © 2004 American Association of Physics Teachers
copper and lacquer. The polished glass itself is usually clear
soda–lime silica whose constituents 共by weight兲are about
72% SiO2, 14% Na2O(⫹K2O), 9% CaO, 3% MgO, 1%
Al2O3, and variable, smaller amounts of Fe2O3共⬍1%兲.6,7
Other backing metals 共for example, aluminum兲and special
‘‘colorless’’ 共that is, lower iron oxide content兲or tinted glass
can be used to change a mirror’s reflectance spectrum appre-
ciably.
Nonetheless, a soda–lime silica glass substrate with silver
backing forms the optical core of most common mirrors.
Figure 3 shows the reflectance and transmission spectra of
these two materials in air at visible wavelengths (380⭐
⭐780 nm).8,9 Although the glass transmits most at 510 nm
共a green兲, its broad transmission maximum only imparts a
pastel greenish cast to white light, even after it has been
transmitted through several centimeters of the glass. This
greenish cast is best seen by looking obliquely through a
plate of common glass, bearing in mind that refractive dis-
persion can complicate the colors that students see. In other
words, students should learn to distinguish between the uni-
form green or blue–green that results from glass absorption
and the many prismatic colors that result from glass refrac-
tion. In contrast, although Fig. 3’s silver spectrum shows a
fairly steep decrease in spectral reflectance Rbelow 460
nm, at longer wavelengths the metal reflects much more
spectrally uniformly.
When we form a composite common mirror from these
two materials, what reflectance spectra result? Figure 4
shows some representative spectra both from Granada’s Sci-
ence Museum mirrors and from common mirrors in our own
laboratories and homes (400⭐⭐700 nm, measured in 10
or 5 nm steps兲. We measured these spectra at normal inci-
dence using either of two portable spectrophotometers: a Mi-
nolta CM-2022 or a HunterLab UltraScan.10 In Fig. 4, all
spectral reflectances except those labeled ‘‘R共lab 1兲’’ were
measured with the Minolta CM-2022. Given the instruments’
designs, these spectra necessarily include both external
specular reflections from the mirror glass and multiple reflec-
tions from within the mirror.
One of Fig. 4’s most striking features is how much the
overall reflectance Rvaries from mirror to mirror 共and, as
our measurements show, even within a given mirror兲; mean
Rvalues range from 80.5% for the ‘‘museum 1’’ mirror to
95.6% for the ‘‘home 2’’ mirror. Although some of this vari-
ability may be attributed to spectrophotometer errors, more
variability likely comes from differences in the mirrors’
original fabrication.
Nevertheless, Fig. 4’s spectra all have similar shapes: they
peak on average at 545 nm, a yellowish green. To restate
these spectral differences in colorimetric terms, after a single
reflection of a spectrally flat illuminant 共the equal-energy
spectrum兲,11 the most dissimilar mirrors in Fig. 4 共the mu-
seum 1 and home 2 mirrors兲slightly change that illuminant’s
color into two new colors. However, these two colors differ
by only 14% more than the local MacAdam just-noticeable
Fig. 1. Schematic diagram of the mirrors used to produce the virtual mirror
tunnel photographed in Fig. 2. The circles on the right mirror are eyeholes
through which observers view the mirror tunnel.
Fig. 2. Mirror tunnel photographed at the Science Museum in Granada,
Spain. The large white circle framing the mirror tunnel is one of Fig. 1’s
eyeholes, which also appear as small dark circles within the tunnel itself.
Fig. 3. Normal-incidence spectral reflectances Rfor optically thick silver
and spectral transmissivities Tfor a 5 mm thick layer of nominally clear
soda–lime silica (SiO2) glass. The optical properties of each material are
considered separately here 共that is, they are not combined as a mirror兲,and
the incident medium is air.
54 54Am. J. Phys., Vol. 72, No. 1, January 2004 R. L. Lee, Jr. and J. Herna
´ndez-Andre
´s
difference.12 In other words, after a single reflection, al-
though many observers might be able to distinguish between
colors reflected by Fig. 4’s most dissimilar mirrors when
viewed side by side, this is by no means certain. Thus, the
color shifts caused by Fig. 4’s very different-looking reflec-
tance spectra will appear nearly identical.
Yet each mirror’s color shift is by itself only marginally
perceptible. Using the same equal-energy illuminant as be-
fore, we find that after one reflection, even the most spec-
trally selective mirror in Fig. 4 共the museum 1 mirror兲in-
creases that light’s colorimetric purity by ⬍3% or ⬃2.7 just-
noticeable differences. More significantly for the mirror
tunnel, the dominant wavelength of this low-purity reflected
light is 560 nm, a yellow–green. In the mirror tunnel, of
course, light undergoes not one but many reflections between
the two mirrors before reaching our eyes. Energy conserva-
tion requires the mirror tunnel’s effective spectral reflectance
R,Nafter Nreflections to be
R,N⫽共R,1兲N,共1兲
where R,1 is as given above, the spectral reflectance after
one reflection by a mirror.
Figure 5 shows the pronounced spectral effects of 50 re-
flections by the Science Museum mirrors. The 50th reflection
has a much weaker and spectrally narrower reflectance spec-
trum than the first reflection. Obviously, this makes colors in
the mirror tunnel’s most ‘‘distant’’ images darker and more
saturated. In particular, 50 reflections by the museum 1 mir-
ror give a white object a dominant wavelength ⬃552 nm and
colorimetric purity ⬃71%,13 compared with the 3% purity
resulting from one reflection. Equally dramatic is the fact
that the white object’s reflected luminance is reduced by a
factor of 5780 after 50 reflections.14
Between these extremes, the mirror tunnel’s luminances
and chromaticities follow easily predicted, although dis-
tinctly nonlinear, paths. To provide a context for these chro-
maticities, Fig. 6 shows the entire CIE 1931 x, y chromaticity
diagram,2to which we have added a dashed chromaticity
locus for blackbody 共or Planckian兲radiators. The dimension-
less CIE 1931 chromaticity coordinates x, y are derived from
psychophysical experiments in which subjects matched the
colors of test lights by mixing red, green, and blue reference
lights in varying intensities. To a first approximation, the x
coordinate represents the relative amounts of green and red
in a color, and the ycoordinate indicates the relative amounts
of green and blue.
The chromaticity diagram’s curved border in Fig. 6 con-
sists of monochromatic spectrum colors that increase clock-
Fig. 4. Measured spectral reflectances Rfor several common mirrors. All
measurements were made at normal incidence and include external specular
reflections from the mirror glass and multiple reflections from within the
mirror.
Fig. 5. Effective spectral reflectances for the museum 1 mirror in Fig. 4 after
one (R,1) and 50 (R,50) mirror-to-mirror reflections. To emphasize the
R,50 spectrum’s greatly increased spectral selectivity, we normalize it to
have the same maximum as the R,1 spectrum. In reality, after 50 mirror-to-
mirror reflections, the mean reflectance and reflected luminance decrease by
factors of 11 620 and 5780, respectively.
Fig. 6. CIE 1931 x, y chromaticity diagram, to which is added the curved
chromaticity locus of blackbody 共or Planckian兲radiators. The rectangular
box that straddles the Planckian locus is the region shown in Fig. 7; its upper
right corner just touches the horseshoe-shaped spectrum locus.
55 55Am. J. Phys., Vol. 72, No. 1, January 2004 R. L. Lee, Jr. and J. Herna
´ndez-Andre
´s
wise from short wavelengths at the diagram’s left vertex
共near x⫽0.175, y⫽0.0) to long wavelengths at its right ver-
tex 共near x⫽0.735, y⫽0.265). This curved part of the bor-
der is called the spectrum locus. A straight line connects
these two spectral extremes, and mixtures of monochromatic
red and blue along it generate purples. Thus, purples are not
spectrum colors proper 共they are not monochromatic兲, but
they are the purest possible colors that link the ends of the
visible spectrum. One point in Fig. 6’s interior is a white or
achromatic stimulus, and for the equal-energy illuminant its
chromaticity coordinates are x⫽0.33333, y⫽0.333 33. We
can form any other color by additively mixing a specific
amount of this achromatic stimulus with a spectrum color 共or
a purple兲. An arbitrary color’s dominant wavelength is found
by extending a straight line from the achromatic stimulus
through the arbitrary color and then on to intersect the spec-
trum locus. The monochromatic spectrum color at this inter-
section defines the arbitrary color’s dominant wavelength.
Furthermore, the arbitrary color’s fractional distance be-
tween white and its spectrum color defines its colorimetric
purity.
Figure 7 expands the boxed area in Fig. 6, and in it we plot
chromaticity curves for a white object seen after Nmirror-
to-mirror reflections by the mirrors in Fig. 4. Each curve in
Fig. 7 is constructed by repeatedly drawing from the chro-
maticity coordinates xN,yNfor the RNreflection to xN⫹1,
yN⫹1for the RN⫹1reflection (1⭐N⭐50). We calculate each
xN,yNby multiplying the equal-energy illuminant’s power
spectrum by a white object’s constant Rand the mirrors’
changing R,N. To show the maximum color range for mir-
ror tunnels resulting from our measured spectra, we substi-
tute in Fig. 7 a slightly more greenish common mirror
共‘‘home 4’’兲for the ‘‘lab 1’’ mirror in Fig. 4. Note that all
colorimetric calculations given here can be done easily in an
electronic spreadsheet that contains mirror spectra and the
human color-matching functions.15
As expected, Fig. 7’s chromaticities grow steadily purer
with increasing N. However, for the largest N, the most dis-
tant colors in the tunnel are difficult 共or impossible兲to see
because they are so dark. In fact, Fig. 8 shows that the re-
flected luminance L
decreases nearly logarithmically with
N, as one would expect from Eq. 共1兲.16 Figure 8 does not
include the ‘‘home 4’’ L
shown in Fig. 7 because they
would overlie the L
of museum 1. Note that after 50 reflec-
tions the latter mirror not only yields the darkest colors 共Fig.
8兲, but also the purest ones 共Fig. 7兲. This decreased bright-
ness is no coincidence, because increased spectral purity nec-
essarily comes at the price of reduced luminance for nonfluo-
rescing reflectors 共see Fig. 4兲.
IV. MODELING MIRROR-TUNNEL COLORS
One of the most instructive and satisfying aspects of the
mirror–tunnel exercise for students is that they can simulate
a fairly complex optical system with only modest mathemati-
cal effort. We start with the facts that a second-surface mir-
ror’s spectral reflectance Rdepends on the mirror surface’s
preparation, the illuminant’s incidence angle and polarization
state, the glass thickness and composition, and the optical
properties of the metal itself. We then base our model on the
following assumptions.
共1兲The mirror is a polished flat substrate of soda–lime
silica (SiO2) glass that is 5 mm thick and is backed by an
optically thick silver 共Ag兲film. This glass matches the thick-
ness of the mirror glass shown in Fig. 2.
共2兲We assume normal incidence for all reflections. Strictly
speaking, the eyeholes in an actual mirror tunnel 共see Figs. 1
and 2兲keep us from seeing normal-incidence reflections be-
tween the two mirrors, but this fact only negligibly affects
the relevance of the color and luminance trends simulated
here.
共3兲We include the effects on Rof 5 reflections within
each mirror 共see Fig. 9兲; higher-order internal reflections
contribute almost nothing to R. However, we exclude in-
terference among the internally reflected rays because the
mirror glass is sufficiently thick 共⬃8620 times the mean vis-
Fig. 7. CIE 1931 x, y chromaticity curves for a white object seen after N
mirror-to-mirror reflections in a mirror tunnel. The mirrors are those whose
measured Rare plotted in Fig. 4, plus one additional mirror 共the home 4
curve兲. In order to show as much detail as possible, the x, y scaling is not
isotropic 共unlike Fig. 6兲.
Fig. 8. The reflected mirror-tunnel luminance L
共arbitrary maximum value兲
decreases nearly logarithmically with an increasing number of mirror-to-
mirror reflections N. The mirrors are those whose Rare plotted in Fig. 4.
56 56Am. J. Phys., Vol. 72, No. 1, January 2004 R. L. Lee, Jr. and J. Herna
´ndez-Andre
´s
ible 兲 so that its optical pathlengths are far greater than the
illuminant’s coherence length.17 Thus we add the intensities,
rather than the amplitudes, of successive internal reflections.
共4兲As was true in Sec. III, we assume an unpolarized,
equal-energy illuminant.
During the light’s traversal of the mirror, it variously un-
dergoes absorption within a medium or reflection and trans-
mission at an interface between two media. Note that all the
following equations are implicit functions of wavelength. At
the air–glass interface, external specular reflection r0is
given by the normal-incidence Fresnel equation:1
r0⫽rair/glass⫽
共nglass⫺nair兲2⫹kglass
2
共nglass⫹nair兲2⫹kglass
2,共2兲
where nglass and kglass are the real and imaginary components
of the refractive index of soda–lime silica glass,7,8 and nair
⫽1 at visible wavelengths.
Transmittance through the glass Tglass is modeled by the
Lambert–Bouguer–Beer exponential absorption law:
Tglass⫽exp共⫺4
dkglass /兲,共3兲
where dis the mirror glass’ thickness. Normal-incidence re-
flection at the glass/Ag interface is given by
rglass/Ag⫽共nglass⫺nAg兲2⫹共kglass⫺kAg兲2
共nglass⫹nAg兲2⫹共kglass⫹kAg兲2,共4兲
where nAg and kAg are the real and imaginary components of
the refractive index of silver.9Here rdenotes component
reflectances from within the mirror, in contrast to their sum
R, the reflectance that we can measure with a spectrophotom-
eter.
Using Eqs. 共2兲–共4兲, we add the external specular and mul-
tiple within-mirror reflections as follows 共see Fig. 9兲:
R⫽r0⫹兺
j⫽1
5
rj,共5兲
where rj⫽(1⫺rair/glass)2rair/glass
j⫺1Tglass
2jrglass/Ag
jis the jth inter-
nally reflected contribution to Rby light that is 共a兲initially
transmitted from air to glass and, ultimately, from glass to air
共the factor of (1⫺rair/glass)2), 共b兲reflected internally at the
air/glass interface if j⬎1共the factor of rair/glass
j⫺1), 共c兲trans-
mitted twice through the glass en route to and from the silver
backing 共the factor of Tglass
2j), and 共d兲reflected at the silver
backing 共the rglass/Ag
jfactor兲. Note that for all j,rjⰇrj⫹1,
which means that each successive internal reflection makes a
much smaller contribution to the observed R.
At non-normal incidence and reflection angles, the exter-
nal reflection r0and internal reflections rjfrom slightly dif-
ferent directions actually contribute to the observed R, but
the small relative size of r0共⬃0.04 at visible wavelengths兲
and the rapid decrease in rjmeans that r1dominates R. Thus,
common mirrors function much as desired—most of their
reflected luminance comes from a single reflection by the
mirror’s metal backing. We include the higher-order reflec-
tions here for completeness and because they do contribute
to a second-surface mirror’s images, even if only by slightly
degrading their quality.
Figure 10 shows Rpredicted by our model for three com-
posite mirrors, each using the same silver backing but a dif-
ferent glass. In addition to the clear and low-iron soda–lime
silica glasses,8Fig. 10 also plots the simulated Rfor a mir-
ror whose glass is pure silica.18,19 Not surprisingly, the
impurity-free silica mirror is the least absorbing of the three,
and so its reflectance spectrum most closely resembles that
of the silver backing 共Fig. 3兲. With increasing amounts of
iron oxide in the glass substrate, the mirrors not only have
lower mean R, but also become more spectrally selective.
Thus, the two mirror tunnels with the most dissimilar colors
should be those produced by the least and most absorbing
mirror glass 共pure silica and clear glass, respectively兲.
We see this trend in the chromaticity diagram of mirror
tunnels predicted for our simulated mirrors 共see Fig. 11兲.To
aid the comparison with mirror tunnels produced by actual
common mirrors, we repeat the chromaticity curves from
Fig. 7 in Fig. 11. In addition, we include the mirror tunnel for
Fig. 9. The specular external reflection, r0, and multiple internal reflections
r1¯r5that contribute to the measurable reflectance Rof a composite glass/
silver second-surface mirror. For clarity, these reflections are shown at non-
normal incidence. Fig. 10. Simulated spectral reflectances Rfor several types of second-
surface glass mirrors, modeled using Eqs. 共2兲–共5兲. Each simulation assumes
normal incidence and includes external specular reflections from the mirror
glass, multiple reflections from within the mirror, anda5mmthick glass
substrate.
57 57Am. J. Phys., Vol. 72, No. 1, January 2004 R. L. Lee, Jr. and J. Herna
´ndez-Andre
´s
a fourth simulated mirror, this one made using the greenish
soda–lime silica glass from Ref. 8. Note that pure silica pro-
duces a mirror tunnel whose chromaticities quickly curve
toward the yellow as Nincreases. This curvature occurs be-
cause pure silica absorbs very little, and thus the repeated
reflections merely serve to increase the purity of silver’s
slightly yellowish reflectance spectrum. However, the mirror
tunnels predicted for our other simulated glasses in Fig. 11
have chromaticity curves that point toward shorter dominant
wavelengths 共that is, toward the greens兲as the mirror glass’
iron oxide content increases 共shifting from pure silica to the
greenish glass兲. Table I quantifies these colorimetric trends
for N⫽25 reflections.
Students can glean several insights from Fig. 11. First,
mirror–tunnel colors depend on the properties of both the
backing metal and the glass substrate, as seen in the tunnels’
chromaticity curves when we change either mirror compo-
nent. To make this subtle point, we suggest that students first
examine samples of polished silver, aluminum, and low-iron
and ordinary clear window glass. They will see that the met-
als look similar, as do the two glasses. Next, have them cal-
culate how much changing the mirror’s backing metal
changes the mirror–tunnel colors. The result is startling: Fig.
12 shows that if aluminum20 is used rather than silver, the
simulated mirror tunnels in Fig. 11 shift markedly toward
shorter wavelengths. Thus, compared to their silver counter-
parts, aluminum-backed mirrors yield chromaticity curves
that are rotated counterclockwise around the point marked
N⫽1. Furthermore, the amount of rotation varies inversely
with mirror-glass absorption 共that is, weakly absorbing pure
silica yields a larger shift than does clear glass兲. This shift
toward shorter dominant wavelengths occurs because alumi-
num has a slightly bluish reflectance spectrum.
Second, if we reduce the mirror glass thickness from 5 to
2mm共the smaller value is found in some home mirrors兲, this
reduction rotates a mirror’s chromaticity curve clockwise in
Fig. 11, and the amount of rotation increases with kglass .
Thus, a chromaticity curve’s degree of rotation increases as
intrinsic absorption by the mirror glass increases, and the
rotation caused by reduced glass thickness is toward longer
wavelengths 共that is, toward the chromaticity curve that
would result from the silver backing alone兲. Figure 11 shows
the rotation that results from replacing the 5 mm substrate of
clear glass with one that is 2 mm thick. However, no percep-
tible shift occurs if we reduce the thickness of the pure silica
mirror glass to 2 mm, becausea5mmlayer absorbs 共and
colors兲so little of the incident light. When students calculate
and ponder such dramatic colorimetric shifts, it will reinforce
for them just how much the appearance of composite mate-
rials depends on each material’s optical properties.
Third, Fig. 11 indicates the power of even a simple model
to explain the complicated world of everyday appearances—
the measured colors of a range of common mirrors closely
match those simulated for similar mirrors made of silver and
clear soda–lime silica glass. Finally, students can learn that
spectrally selective absorption from multiple reflections in
glass and mirrors has many practical implications in research
Fig. 11. Chromaticity curves for a white object seen after Nmirror-to-mirror
reflections in a mirror tunnel produced by our measured 共Fig. 4兲and simu-
lated 共Fig. 10兲common mirrors, plus curves for simulated mirrors made
either with slightly greenish glass 共green glass curve兲or clear glass that is 2
mm thick 关clear glass 共2mm兲curve兴. See Fig. 7 for a closer view of the
measured mirror tunnels’ chromaticities.
Fig. 12. The colorimetric consequences of changing the backing metal in
several of the simulated mirrors shown in Fig. 11. Replacing silver with
aluminum rotates each mirror–tunnel chromaticity curve counterclockwise
共that is, toward shorter wavelengths兲, and the amount of rotation increases
with decreasing absorption by the mirror glass.
Table I. Dominant wavelengths, colorimetric purities, and approximate CIE
color names for a white object seen in the simulated mirror tunnels in Fig.
11 after N⫽25 mirror-to-mirror reflections. Each mirror is backed with sil-
ver, and the light source is the equal-energy illuminant. Unless noted other-
wise, the mirror glass thickness is 5 mm.
Dominant
wavelength 共nm兲
Purity
共%兲CIE color name
Pure silica 580 59.7 Yellow
Low-iron glass 572 52.7 Greenish yellow
Clear glass 共2mm兲566 48.0 Yellow green
Clear glass 540 35.1 Yellowish green
Greenish glass 501 57.9 Green
58 58Am. J. Phys., Vol. 72, No. 1, January 2004 R. L. Lee, Jr. and J. Herna
´ndez-Andre
´s
areas ranging from signal loss in fiber optic cables19 to spec-
troscopic analysis of gases within optically resonant
cavities.21
V. SUMMARY
Mirror tunnels have the twin advantages of being visually
and intellectually intriguing. Thus, an experiment that lets
students explore the relationship between the exotic visual
world of the mirror tunnel and the measured and modeled
optical properties of common mirrors makes for an instruc-
tive, appealing bridge between geometrical and physical op-
tics. The path that we have taken here has the virtue of using
readily observed phenomena 共the mirror tunnel’s color and
brightness trends兲to develop and test a simple, but fairly
rigorous, optical model that explains them. Along the way,
students will find that seemingly abstruse ideas such as colo-
rimetry, the Fresnel equations, and complex refractive indi-
ces can be combined to explain everyday visual experience.
ACKNOWLEDGMENTS
We are grateful to the Science Museum of Granada, Spain
for permitting us to photograph its mirror tunnel demonstra-
tion. Several agencies generously funded this research. Lee
was supported by United States National Science Foundation
Grant No. ATM-0207516 and by the United States Naval
Academy’s Departments of Physics and Mathematics.
Herna
´ndez-Andre
´s was supported by Spain’s Comisio
´n Inter-
ministerial de Ciencia y Tecnologı
´a共CICYT兲under research
grant No. BFM 2000-1473.
a兲Electronic mail: raylee@usna.edu
1M. V. Klein and T. E. Furtak, Optics, 2nd ed. 共Wiley, New York, 1986兲, pp.
263–307.
2R. S. Berns, Billmeyer and Saltzman’s Principles of Color Technology, 3rd
ed. 共Wiley, New York, 2000兲, pp. 44–66.
3Parque de las Ciencias, Avda. del Mediterra
´neo s/n, 18006 Granada, Spain,
具http://www.parqueciencias.com典.
4Thus in Fig. 2, the ascending ‘‘staircase’’ of floor images arises because
the mirrors’ tops are closer than their bases. The mirror tunnel’s curvature
is useful in itself for discussing the geometrical optics of plane mirrors and
the behavior of their virtual images.
5B. Schweig, Mirrors: A Guide to the Manufacture of Mirrors and Reflect-
ing Surfaces 共Pelham Books, London, 1973兲, pp. 3–4, 23–77.
6H. G. Pfaender and H. Schroeder, Schott Guide to Glass 共Van Nostrand
Reinhold, New York, 1983兲, pp. 45, 66.
7T. Uchino, K. Nakaguchi, Y. Nagashima, and T. Kondo, ‘‘Prediction of
optical properties of commercial soda–lime–silicate glasses containing
iron,’’ J. Non-Cryst. Solids 261, 72–78 共2000兲.
8M. Rubin, ‘‘Optical properties of soda lime silica glasses,’’ Sol. Energy
Mater. 12, 275–288 共1985兲.
9Handbook of Optical Constants of Solids, edited by E. D. Palik 共Aca-
demic, Orlando, FL, 1985兲, Table IX, p. 356.
10CM-2022 spectrophotometer from Minolta Corporation, 101 Williams
Drive, Ramsey, NJ 07446; HunterLab UltraScan spectrophotometer from
HunterLab Associates, Inc., 11491 Sunset Hills Road, Reston, VA 20190-
5280. Both instruments were calibrated shortly before we measured the
spectra in Fig. 4.
11Although the equal-energy illuminant is rarely encountered, we use its
spectral simplicity to help us isolate the colorimetric effects of Rin
creating the mirror tunnel’s colors. For more realistic illuminants 共for ex-
ample, incandescent lighting兲, the mirror tunnel’s chromaticities depend on
the wavelength-by-wavelength multiplication of Rand a more compli-
cated power spectrum.
12G. Wyszecki and W. S. Stiles, Color Science: Concepts and Methods,
Quantitative Data and Formulae, 2nd ed. 共Wiley, New York, 1982兲,pp.
306–310. Here we follow convention and set the just-noticeable difference
equal to the semimajor axis length of the MacAdam color-matching ellipse
at the relevant chromaticity. For the equal-energy illuminant, this chroma-
ticity is CIE 1931 x⫽0.333 33, y⫽0.33333. We define chromaticity co-
ordinates and other basic CIE metrics elsewhere in the main text.
13Dominant wavelengths and purities are measured with respect to the
equal-energy illuminant’s chromaticity. Here we define a white object as a
diffuse reflector with constant Rat visible wavelengths, say R⫽0.9; a
photographer’s white card approximates this behavior.
14Because each of the ⬃12 reflected gray-card images visible in Fig. 2
occurs after one pair of reflections between the two mirrors, N⫽24 total
reflections contribute to the color and brightness of its smallest gray card
image. To measure the relative luminance 共or radiance兲of the N⫽24 mir-
ror reflection, we restrict the spectroradiometer’s field of view to the 12th
smallest image of some easily recognized object in the mirror tunnel 共for
example, the gray card in Fig. 2兲. Then we divide this reading by the
luminance or radiance of the object seen directly 共the N⫽0 image兲. Given
the large luminance reductions involved, N⫽50 reflections is a generously
large upper limit on the visibility of mirror tunnels formed by common
mirrors.
15On request, we will gladly provide readers with as many of our measured
mirror spectra as desired, plus instructions on how to use this data in
colorimetric calculations. We also can give readers smoothly interpolated
complex refractive indices of glasses and metals in a form that is suitable
for simulating mirrors’ chromaticities 共see the model of Sec. IV兲.
16The relationship between L
and Nis not quite log-linear because lumi-
nance is calculated by a spectral integral that includes the changing shape
of the mirrors’ R,N. If a mirror’s original spectral reflectances R,1 were
completely uniform, then both the reflected luminance and radiance would
decrease exactly logarithmically with increasing N.
17J. Herna
´ndez-Andre
´s, E. M. Valero, J. L. Nieves, and J. Romero, ‘‘Fizeau
fringes at home,’’Am. J. Phys. 70, 684– 688 共2002兲.
18For the complex refractive indices of pure silica, see Ref. 9, Table X, pp.
759–760.
19We use data on pure silica’s very weak visible-wavelength absorption
found in B. Keck, R. D. Maurer, and P. C. Schultz, ‘‘On the ultimate lower
limit of attenuation in glass optical waveguides,’’ Appl. Phys. Lett. 22,
307–309 共1973兲.
20For the complex refractive indices of metallic aluminum, see Ref. 9, Table
XII, pp. 398–399.
21Cavity-Ringdown Spectroscopy: An Ultratrace-Absorption Measurement
Technique, edited by K. W. Busch and M. A. Busch 共American Chemical
Society, Washington, DC, 1999兲.
59 59Am. J. Phys., Vol. 72, No. 1, January 2004 R. L. Lee, Jr. and J. Herna
´ndez-Andre
´s