Chapter 1

Introduction
Thomas A. Germer*, Joanne C. Zwinkels{ and Benjamin K. Tsai*
*National Institute of Standards and Technology, Gaithersburg, Maryland, USA
{
National Research Council Canada, NRC, Measurement Science and Standards, Ottawa,
Ontario, Canada

Chapter Outline
1.1 Opening Remarks 1 1.3 Overview 8
1.2 Uncertainties 4 References 9

Exper. 6. And as these Colours were not changeable by Refractions, so neither were they
by Reflexions. For all white, grey, red, yellow, green, blue, violet Bodies, as Paper, Ashes,
red Lead, Orpiment, Indico Bise, Gold, Silver, Copper, Grass, blue Flowers, Violets,
Bubbles of Water tinged with various Colours, Peacock’s Feathers, the Tincture of
Lignum Nephriticum, and such-like, in red homogeneal Light appeared totally red, in blue
Light totally blue, in green Light totally green, and so of other Colours. In the homogeneal
Light of any Colour they all appeared totally of that same Colour, with this only Difference,
that some of them reflected that Light more strongly, others more faintly. I never yet
found any Body, which by reflecting homogeneal Light could sensibly change its Colour.
Sir Isaac Newton [1]

1.1 OPENING REMARKS

Spectrophotometry is the quantitative measurement of the interaction of ultra-
violet (UV), visible, and infrared (IR) radiation with a material and has an
impact on a wide field of science and technology. The nature of this interaction
depends upon the physical properties of the material, for example, transparent
or opaque, smooth or rough, pure or contaminated, and thin or thick. Thus,
spectrophotometric measurements can be used to quantify, in turn, these impor-
tant physical properties of the material. The choices of spectrophotometric
measurements include spectral reflectance, transmittance, absorptance, emit-
tance, scattering, and fluorescence and can be classified as phenomenological
optical properties of the material. Spectrophotometric measurements can also

Experimental Methods in the Physical Sciences, Vol. 46. http://dx.doi.org/10.1016/B978-0-12-386022-4.00001-7

© 2014 Elsevier Inc. All rights reserved. 1
2 Spectrophotometry

be used to probe the intrinsic or internal physical nature of the material, such as
its refractive index and extinction coefficient.
The design and performance of optical instruments, ranging from low-cost
cell-phone cameras to high-cost microlithography projection tools and satel-
lite telescopes, require knowledge of the optical properties of the components,
such as their refractive index, roughness, subsurface scatter, and contamina-
tion. The pharmaceutical and chemical industries use optical absorption and
fluorescence measurements to quantify concentration, required for accurate
dosing and elimination of contaminants. Global climate change simulations
require accurate knowledge of the optical properties of materials, gases, and
aerosols to calculate the net energy balance of our planet. The properties of
thin films, even when they are not intended for optical applications, are often
related to their optical reflection, transmission, and scattering properties.
Commercial products are often selected by consumers based upon appearance,
a complex attribute that encompasses more specific terms, such as color,
gloss, and texture. Renewed interest in solar energy has driven the need to
maximize the light capture efficiency of solar collectors.
When we are asked to inspect a piece of material, it is our natural inclina-
tion to view it by holding it up to a light. The interaction of the light with the
material gives us an overall impression of its quality. Our vision is also inher-
ently multispectral, by providing color discrimination on a relatively high spa-
tial resolution. Binocular vision, by allowing us to view the object from
multiple directions simultaneously, gives us an ability to perform rudimentary
tomography. The spectral, spatial, and directional properties permit us to iden-
tify materials, characterize topography, and observe defects, without ever
coming into contact with the object. It is not surprising, then, that we seek
to make measurements of optical properties of materials in order to better
quantify what our own eyes sense qualitatively.
While certain aspects of optics, such as the laws governing refraction of
light and the ray nature of light, were well established by the mid-1600s, it
was Isaac Newton who discovered that white light was a mixture of colors
that could be separated into its components using a prism. It could be argued
that Newton performed the first spectrophotometric measurements of this
light interaction with a prismatic material. This chapter’s epigraph [1] gives
an account of his discovery that, in the absence of fluorescence, rays of one
color cannot be changed into rays of another, but that different materials sim-
ply reflect the colors in different amounts. Newton noted that the color purple
was not in the rainbow, but could be created by mixing violet and red rays. He
then proposed the basic structure of the color circle and noted that mixtures of
any two opposing colors yield a neutral gray.
Newton, of course, used his eyes as the detector. While he could be quite
quantitative in measuring angles of refraction, he had more difficulty in esti-
mating intensity or quantifying color. Furthermore, because of his reluctance
to accept the wave nature of light, he would never correlate the colors that he
Chapter 1 Introduction 3

observed after dispersion through a prism with the corresponding wavelengths

of light. Through the years, however, acceptance of the wave properties took
hold, first through the double slit experiment of Young [2] and then through
the progressive works of Augustin Fresnel, Michael Faraday, James Clerk
Maxwell, and others. By the 1800s, the world was ready for precision mea-
surements of wavelength, and the birth of quantitative spectroscopy occurred.
In the early 1920s, it was recognized that it was important that the results
from spectrophotometric measurements not only provide qualitative informa-
tion but are also reliable and meaningful quantitatively. The Optical Society
of America convened a progress committee on spectrophotometry, which
issued a report in 1925 [3]. This report gives an amazing account of the status
of spectrophotometry at that time, and except for the obvious lack of automa-
tion, many of the issues that were covered in this report are still relevant today
to obtaining meaningful spectrophotometric results. These include establish-
ing a common terminology, spectral calibration, stray light exclusion, polari-
zation, differentiation between diffuse and specular components, and
precision and accuracy.
The first automated, recording spectrophotometer was developed between
1926 and 1928 by Hardy and his colleagues at the Massachusetts Institute of
Technology [4]. Before this time, spectrophotometers were extremely tedious
to use. In a retrospective written a decade later [5], Hardy pointed out that the
first months of operation of this instrument were very exciting, that they
measured everything within sight, and that it took less time to make a mea-
surement than it took to decide whether the measurement would be signifi-
cant. Their results brought hundreds of visitors to their laboratory, and they
soon realized that virtually every industry was in need of such measurements.
Soon, Hardy made arrangements with the General Electric Company [6] to
commercialize the instrument. In the intervening years, there have been sig-
nificant advances in the design of spectrophotometers including the emer-
gence of faster multi-wavelength designs in the 1970s and the introduction
of a commercially available diode array spectrophotometer in 1979. The vari-
ety of different types of spectrophotometers has also increased dramatically
over the years, including many specialized features for measuring any type
of sample and every type of optical property.
In its simplest form, a spectrophotometer contains three parts: a source, a
sample holder, and a detector. The source usually contains some sort of spec-
trometer so that the optical radiation is monochromatic, covering a range of
interest. The wavelength range for spectrophotometric measurements depends
upon the application and can cover the ultraviolet, visible, or various ranges in
the infrared. For example, for characterizing materials for use in solar energy
applications, spectrophotometric measurements extending from about 200 to
2500 nm (i.e., the region of the solar spectrum) are important. The detector
is designed to be sensitive over the range of interest, and an instrument might
employ multiple detectors so that it can cover a broader wavelength range
4 Spectrophotometry

than that covered by a single detector. Operated in this fashion, a spectrophoto-

metric curve for the sample can be obtained, by comparing at each wavelength
the signal collected after interaction of the monochromatic source with the
sample to the signal recorded without the sample in the measurement beam.
More complex measurements can be achieved on commercial instruments
through the use of specialized accessories that modify the beam path, move or
substitute the detector, or manipulate the sample orientation. In this manner,
specular, diffuse, or angle-resolved reflectance or transmittance measurements
can be performed over a wide range of wavelengths, making the commercial
spectrophotometer a very versatile tool. Accurate measurements of optical
properties of materials using spectrophotometric instrumentation remain a
challenge, and the wide variety of modern instruments and applications has
heightened this need for improved standardization and traceability. This book
aims to address this need by providing both the novice and the experienced
user of spectrophotometry an authoritative reference document with compre-
hensive terminology, guiding principles, and best measurement practices,
including examples of important applications. In this text, we do not limit our-
selves to measurements made on commercial instruments. In many cases, the
commercial instrument is designed to rely upon a reference standard, with
which a relative measurement is made. The reference standard, on the other
hand, often has its reference values certified using an absolute method that
does not rely upon a physical standard, but rather, upon methods used to real-
ize the definition of the quantity.

1.2 UNCERTAINTIES
The accuracy of spectrophotometric measurements, and the derived optical
properties of the material under test, depends upon the design and calibration
of the spectrophotometer, the choice of reference standard, and the interaction
of the sample with the measuring instrument. A discussion of high-accuracy
measurements is not complete without a discussion of measurement uncertain-
ties. That is, one cannot speak of “high accuracy” without asking the question,
“Just how accurate is the measurement?” Many of the chapters of this text
illustrate how such an estimate can be made for each measurement. In this
introduction, we present the basic principles of how such an uncertainty anal-
ysis is performed in accordance with methods described in the ISO Guide to
the Expression of Uncertainty in Measurements [7]. For a complete descrip-
tion, consult this reference or a suitably abridged version [8]. In the following,
we attempt to reduce a few hundred pages into just a few.
Accuracy and precision are two concepts that are important to understand
and differentiate in any field of metrology. When we perform any measurement,
we never know what is the actual value of the measurand. Precision is a term
that is used to loosely describe how close different measurements on the same
instrument are to one another and is relatively simple to assess. Accuracy, on
Chapter 1 Introduction 5

the other hand, is the closeness that a measurement result is to its true value.
Unfortunately, accuracy is much harder to assess because it requires having a
much higher degree of understanding of all the possible sources of error and
their impact on the measurement. For absolute measurements, that is, those that
do not rely on a reference material for their scale, the determination of the
uncertainty can be quite complicated. For relative measurements, where the ref-
erence material comes with a stated uncertainty, the determination of the uncer-
tainty is usually quite a bit simpler.
In short, each measurement has a measurement equation, which can be
written as a functional relationship. For the general case, this is expressed as
y ¼ f ðx1 , x2 , ... , xN Þ, (1.1)
where y is the output quantity of interest, and the xi are all the input quantities
that are combined through a functional relationship f to yield y. The function f
is referred to as the measurement function. The quantities xi include every-
thing that affects the measurement result in any way. Often, we assume that
the measurement function is of the form

f ðx1 , x2 , ... , xN Þ ¼ ef ðx1 , x2 , ... , xM ÞC1 ðxM + 1 ÞC2 ðxM + 2 Þ.. .CNM ðxN Þ, (1.2)

where ef ðx1 , x2 ,.. ., xM Þ is the essential measurement function, which is that

part used to calculate the output value, and Cj(xM+j) are correction factors,
which we assume are supposed to be unity, but which account for most of
the nonidealities in the measurement, including nonlinearity and stray light.
These correction factors are often not explicitly shown in the measurement
equation, but nonetheless, their contributions to the combined uncertainty will
appear in a measurement’s uncertainty budget. It is recommended, however,
that the measurement equation be written out in full, whenever possible.
For each of the identified input quantities xi, we need an estimate of its
standard uncertainty u(xi), which represents an estimate of the standard devi-
ation of xi. Furthermore, we need an estimate of the covariance, u(xi, xj),
between each xi and xj. The combined standard uncertainty, uc(y), is then
determined, using the law of propagation of uncertainty from
N 
X  N 1 X
X
@f 2 N
@f @f  
u2c ðyÞ ¼ u2 ðxi Þ + 2 u xi , xj : (1.3)
i¼1
@xi i¼1 j¼i + 1
@xi @xj

The combined standard uncertainty represents the estimated standard devi-

ation of the final measurement result y. It is often assumed that there is little
or no correlation between the input quantities, that is, the covariance terms are
negligible and the combined standard uncertainty can be calculated as the
simple quadrature sum of the individual uncertainty components. However,
for designing a measurement for optimum accuracy or for careful uncertainty
analysis of any measurement result, it is important to have an eye out for
6 Spectrophotometry

covariance between terms, since, left unnoticed, they can significantly impact
the uncertainty. While the factors in the first term in Eq. (1.3) always increase
the uncertainty, those in the second term do not necessarily do so and in some
cases can reduce the combined standard uncertainty.
It is common practice to identify estimates of uncertainty, u(xi), as either
Type A or Type B. In a Type A evaluation, the input quantity is obtained using
a statistical analysis. In a Type B evaluation, the uncertainty is estimated from
other nonstatistical sources. Examples of Type A uncertainty estimates are
those obtained from the standard deviation of the mean of a series of indepen-
dent measurements, the estimated standard deviations of parameters obtained
using the method of least squares to fit a curve of data, or those obtained from
an analysis of variance (ANOVA). Examples of Type B uncertainty estimates
are those based on experience, previous measurement results, uncertainties
provided in calibration reports, and manufacturers’ specifications.
The partial derivative in the first term on the right of Eq. (1.3) is often
referred to as the sensitivity coefficient for the ith input quantity. This quan-
tity is determined analytically, numerically, or experimentally (i.e., by mea-
suring the change in y for a small intentional change in the value of xi
while keeping the other input quantities constant). In the absence of covari-
ance terms, we often express the uncertainty budget with a table, listing the
various contributions, their sensitivity coefficients ci ¼ @f/@xi, their respective
uncertainties u(xi), and their contributions ciu(xi) to the combined standard
uncertainty. The combined standard uncertainty is then determined by adding
all the contributions in quadrature. As a result, one does not necessarily see
Eq. (1.3) written out explicitly.
When the measurement equation contains a moderate to large number
of independent inputs that each contribute to the uncertainty about equally
and the degrees of freedom in the measurement are large, the result y will
be normally distributed, and we will have 68% confidence that the true value
will lie between y  uc(y) and y + uc(y). Since this is a fairly low level of
confidence, it is standard practice to multiply the combined standard
uncertainty by a coverage factor k to obtain an expanded uncertainty U(y) ¼
kuc(y) so that in an interval between y  U(y) and y + U(y), we have a signifi-
cantly higher level of confidence. For example, under the conditions given
above, k ¼ 2 corresponds to a 95% confidence, and k ¼ 3 corresponds to a
99.7% confidence. For a small to moderate number of degrees of freedom,
the interval of the Student’s t-distribution that encompasses the fraction p of
the distribution should be considered. Table 1.1 gives appropriate coverage
factors for some representative degrees of freedom and some common confi-
dence levels.
To estimate the effective degrees of freedom in the measurement, when
the covariance terms can be ignored, one should use the Welch–Satterthwaite
formula
Chapter 1 Introduction 7

TABLE 1.1 Appropriate Values of Coverage Factor k as a Function of the

Number of Degrees of Freedom, n, and the Confidence Level p

n 68.27% 95.45% 99.73%

1 1.84 14 236
2 1.32 4.5 19
3 1.20 3.31 9.22

4 1.14 2.87 6.62

5 1.11 2.65 5.51
10 1.05 2.28 3.96
25 1.02 2.11 3.33
100 1.01 2.03 3.08
1 1 2 3

u c 4 ðyÞ
neff ¼ X , (1.4)
N ci u ðxi Þ
4 4

i¼1 ni
where ni is the number of degrees of freedom of u(xi). For Type
A uncertainties, the number of degrees of freedom ni should be chosen by
the appropriate statistical method. For example, if the value is taken from
the average of n samples, and the standard uncertainty is taken as the standard
deviation of the mean, then ni ¼ n  1. For Type B uncertainties, and in the
absence of any other guidance (e.g., that given on a calibration certificate),
it is common practice that one takes ni ! 1, which eliminates its contribution
to the denominator of Eq. (1.4).
In many cases, the values are expected to be normally distributed and sta-
tistical analysis can estimate the standard deviation. For many Type
B estimates, we may only have a tolerance such that we know that xi lies in
an interval between xi  a and xi + a. In this case, we assume a probability dis-
tribution that quantifies the different possible values of the input value. Then,
the standard deviation of that probability distribution is taken to be the stan-
dard uncertainty of that input. For example, for a rectangular probability dis-
tribution, we find that the appropriate standard uncertainty is
a
uðxi Þ ¼ pffiffiffi : (1.5)
3
8 Spectrophotometry

Key contributions to the Type A uncertainty are the measurement repeatability

and reproducibility. Repeatability indicates the variation in the measurement
result if the measurement is performed multiple times under the same conditions.
This usually entails performing the measurement on the same instrument and by
the same operator. Generally, an assessment of the repeatability involves removal
of the sample under test and replacement, as if the measurement were being done
each time independently. Sometimes, one distinguishes between short-term
repeatability and long-term repeatability. These two cases are differentiated by
the time interval between the successive measurements. Short-term repeatability
tests may involve the operator removing the sample from the instrument and then
immediately repeating the procedure from the start, while long-term repeatability
tests may involve occasionally performing the same measurement over days,
weeks, or months. However, strictly speaking, this requires that no other condi-
tions that impact this sample measurement change over the course of the time
interval of the repeat run, for example, the environmental conditions, or this would
constitute an assessment of reproducibility due to this influence parameter.
Reproducibility indicates the variation in the measurement result when the
conditions of the measurement have changed significantly. This usually entails
different instruments, possibly even using different measurement methods or
principles, or in different laboratories. When reporting reproducibility, it is
important to describe the conditions that changed between the measurements.

1.3 OVERVIEW
This textbook is organized as follows. Chapter 2 describes a number of theoreti-
cal concepts important for the field of spectrophotometry, including defining
radiometric quantities, expressing their relationships to the electromagnetic
fields, defining the spectrophotometric quantities, and providing a mathematical
foundation for describing reflection, transmission, absorption, emission, and
scattering from materials. Chapters 3 and 4 are dedicated to describing the
two primary methods, dispersive and interferometric, by which spectroscopic
measurements are performed. Chapters 5–8 then discuss the measurements of
reflection, transmission, fluorescence, and scattering, which are common to
spectrophotometry over its wide range of application. Chapters 9 and 10 discuss
two spectrophotometric measurements that are specific to the IR and visible
ranges. These are emissivity and color, respectively, which, while being related
to those discussed in Chapters 5–8, have their own specific measurement issues.
Each of these core chapters provides detailed information on the considerations
and approaches needed for precise and accurate measurements of optical prop-
erties using their particular spectrophotometric method. Finally, we devote three
chapters, Chapters 11–13, to overviews of three specific industries or important
users that illustrate the diverse applications of spectrophotometric measure-
ments. We hope that this book will become the handbook of choice for those
intending to make accurate spectrophotometric measurements.
Chapter 1 Introduction 9

REFERENCES
[1] I. Newton, Opticks: or, a Treatise of the Reflexions, Refractions, Inflexions and Colours of
Light, Royal Society, London, 1704, p. 89.
[2] T. Young, Bakerian Lecture: experiments and calculations relative to physical optics, Phil.
Trans. R. Soc. Lond. 94 (1804) 1–16.
[3] K.S. Gibson, et al., Spectrophotometry: report of O.S.A. Progress Committee for 1922–3,
J. Opt. Soc. Am. 10 (1925) 169–241.
[4] A.C. Hardy, A recording photoelectric color analyser, J. Opt. Soc. Am. 18 (1929) 96–117.
[5] A.C. Hardy, History of the design of the recording spectrophotometer, J. Opt. Soc. Am.
28 (1938) 360–364.
[6] Identification of commercial names is not intended to imply recommendation or endorsement
by the National Institute of Standards and Technology or the National Research Council of
Canada.
[7] ISO Guide to the Expression of Uncertainty in Measurement, ISO, Geneva, Switzerland,
1995.
[8] B.N. Taylor, C.E. Kuyatt, Guidelines for Evaluating and Expressing the Uncertainty of NIST
Measurement Results, National Institute of Standards and Technology, Gaithersburg, MD,
1994.