q 2006 International Society for Analytical Cytology Cytometry Part A 69A:712–734 (2006)

Reviews

Imaging Spectrometer Fundamentals for

Researchers in the Biosciences––A Tutorial
Jeremy M. Lerner*
LightForm Incorporated, Hillsborough, New Jersey
Received 30 August 2005; Revision Received 19 December 2005; Accepted 20 December 2005

Over the last 2 years there has been a dramatic increase in ple. We also show how to deduce the basic capabilities of
the number of bioscience laboratories using wavelength a spectral confocal system. Finally, we show how to deter-
dispersive spectroscopy to study in vivo, in situ fluores- mine the true spectral bandwidth of an object, the illumi-
cence. Transforming spectral information into an image nated area of a laser-excited object, and what is needed to
provides a graphic means of mapping localized ionic, mo- optimize light throughput. q 2006 International Society for Ana-
lecular, and protein–protein interactions. Spectroscopy lytical Cytology
also enables fluorophores with overlapping spectral fea-
tures to be delineation. In this study, we provide the tools
that a researcher needs to put into perspective instrumen- Key terms: spectral imaging; spectrometer; spectro-
tal contributions to a reported spectrum in order to gain graph; PARISS; hyperspectral imaging; confocal; spectrom-
greater understanding of the natural emission of the sam- eter design; wavelength calibration; spectral calibration

The use of spectroscopy has greatly simplified the task To provide background we also describe how readily
of characterizing and delineating autofluorescence, nat- available commercial, plane grating, concave holographic
ural fluorophores, and multiple man-made fluorophores, grating, and advanced prism-based spectrometers work,
many with overlapping spectral profiles, in the same sam- and discuss their inherent limitations and advantages.
ple. Consequently, spectroscopy is one of the fastest The goal is to help a researcher optimize light through-
growing techniques to be found in a bioscience laboratory put, check accuracy, and understand the real consequences
(1,2). It is also one of the least understood, especially of changing aperture sizes (such as a pinhole in a confocal
when both spectral and spatial information is required. In system) on spectroscopic performance. When comparing
this study, we focus on wavelength dispersive devices spectroscopic results with those of others, it is important to
rather than those that acquire spectra sequentially by understand that in some spectrometers spectral resolution
changing bandpass filters. The transformation of wave- degrades with an increase in the ratio of magnification to
length information into an image is often called hyper- numerical aperture (NA), pinhole (or slit) size, and in other
spectral or multispectral imaging, but these terms are so instruments spectral resolution is a constant. We also illus-
blurred that, given the current state of technology, using trate how to determine the undisclosed operating parame-
the simple term ‘‘spectral imaging’’ is appropriate. ters of a commercial spectral confocal microscope.
This study provides the researcher with the tools to There is a routine debate concerning the actual illumi-
understand how spectrometers work, and how the limits nated area of a sample in a laser confocal system. To help
of instrument performance can affect the accuracy, qual- put this issue into perspective there is a section on light
ity, validity, and interchangeability of acquired data. Spec- throughput as a function of variously scattering, laser-
trometers operate with multiple variables that have a excited samples.
significant influence on bandpass, wavelength disper-
sion, aberrations, and light throughput. To complicate
matters further not all spectrometers work well with *Correspondence to: Jeremy M. Lerner, LightForm Inc., 601 Route 206,
Suite 26-479, Hillsborough, NJ 08844, USA.
linear arrays or charge coupled devices (CCD) as a wave- E-mail: [email protected]
length detectors. We try to put all these factors into Published online in Wiley InterScience (www.interscience.wiley.com).
perspective. DOI: 10.1002/cyto.a.20242
 IMAGING SPECTROMETER FUNDAMENTALS FOR RESEARCHERS IN THE BIOSCIENCES 713

FIG. 1. Generic optical layout for both diffraction grating and prism spectrometers. (a) A monochromator with entrance and exit apertures in fixed loca-
tions where the WDE rotates to change wavelength; and (b) a spectrograph with a fixed WDE where the angle of diffraction (or refraction) varies with
wavelength. A multielement detector located at the focal detects all wavelengths simultaneously, where a, angle of incidence; b, angle of diffraction (or
refraction for a prism); k, order (prisms only refract in one order compared to diffraction gratings that present light in multiple orders); bmin, angle of dif-
fraction, or refraction, at the shortest wavelength on an array; bmax, angle of diffraction, or refraction, at the longest wavelength on an array; k, wavelength;
Dv, fixed angle at the center of the WDE; Normal: a reference line perpendicular to the optic. Angles are measured from the normal; La, distance from the
entrance aperture to the first active optic such as a collimating mirror; Lb, distance from the final active optic to the exit aperture or detector; Lh, Perpen-
dicular distance from the final focusing optic to the focal plane in a ‘‘field-flattened’’ spectrograph; bh, angle measured from the normal to the grating to Lh;
g, inclination of a ray Lb, at a specific wavelength, to the focal plane in a ‘‘flat-field’’ spectrograph. An alternative approach in a spectrograph configuration
is to translate an aperture/detector across the focal field.

WAVELENGTH DISPERSION THROUGH The theory behind diffraction gratings and prisms is
DIFFRACTION GRATINGS AND PRISMS well covered in the literature; so this section simply high-
Regardless of the nature of its wavelength dispersive lights the main issues that are of importance to the bio-
element (WDE), all spectrometers operate as a function of scientist (3,5–7).
the same geometric optics. Light strikes the WDE at an
angle of incidence, and depending on whether the WDE ‘‘Need to Know’’ Diffraction Grating Equations
is a diffraction grating or prism, it is either diffracted or
Diffraction gratings are available in three types: classi-
refracted at an angle of diffraction (refraction) that varies
cally ruled (CR), holographic surface relief (HSRG), and
with wavelength. By the very nature of imaging spectros-
volume holographic (VHG). CR and HSRG gratings work
copy, multiple wavelengths will be acquired to character-
in reflection on flat (plano), concave, or convex surfaces.
ize an object. In a monochromator such as that shown in
A VHG is typically used in transmission.
Figure 1a each wavelength is acquired sequentially and a
A classical diffraction grating is generated by mechani-
photomultiplier tube (PMT) measures the signal at each
cally ‘‘ruling’’ (actually burnishing) grooves into a coating
wavelength (3–6). of aluminum or gold on a glass blank. The first example of
In a spectrograph such as that shown in Figure 1b ei- a ruled grating occurred in 1817 when Fraunhofer con-
ther a one-dimensional linear array of detector elements, structed an engine to rule diffraction gratings. Since then,
or a matrix array such as a CCD acquire a series of wave- advanced ruling engines have been developed, which
lengths simultaneously. The spectrometer has to be have dramatically improved precision, accuracy, groove
designed to distribute the wavelength range of choice shape, spacing, and made it possible to rule very large,
over the fixed dimensions of the detector. In other words, high-groove-density (up to 5,000 g/mm) gratings, even on
first the detector has to be chosen, and then the optics of curved surfaces.
the spectrometer has to be designed around it. Holographic gratings are recorded at the intersection of
Although the same geometrical optic considerations two expanded laser beams to form a series of periodic
apply to both prism and diffraction grating based spectro- fringes in photoresist, which, after processing, form sinu-
meters, in this study we illustrate the concepts using the soidal grooves. A major breakthrough occurred in 1969
diffraction grating equations. Most WDE-based systems when Labeyrie, Cordelle, Flamand, and Pieuchard at Jobin
require collimating and/or focusing optics to bring light Yvon in France introduced concave aberration-corrected
from the entrance aperture and focus wavelength dis- holographic gratings (ACHG) that greatly reduced or elimi-
persed light onto the detector. For simplicity, collimating nated astigmatism and field curvature. A key advantage to
and focusing optics are not shown in Figure 1. these gratings is that they do not require any additional fo-

Cytometry Part A DOI 10.1002/cyto.a

714 LERNER

cusing or collimating optics. This simplicity makes them

very attractive for use as the WDE in compact monochro-
mators or spectrographs. Examples of the use of ACHG
gratings are described later in this paper (3,6,8–10).
The diffraction grating equation applies to both ruled
and holographic gratings:

sin a þ sin b ¼ knk ð1Þ

where the terms are based on those shown in the legend

of Figure 1.
The angles of incidence and diffraction, or refraction,
both vary as a function of wavelength in a monochroma-
tor, but the angle of incidence is fixed in a spectrograph,
and the angles of diffraction vary with wavelength (Fig. 1). FIG. 2. A classically ruled blazed diffraction grating. This figure illustrates
Transmission gratings modify Eq. (1) to accommodate the Littrow configuration, where x 5 the ‘‘Blaze’’ angle and x 5 a 5 b. It
is rare for this geometry to be used in practice, but catalog specifications
its refractive index and the thickness of its glass substrate. for efficiency and blaze are always quoted in Littrow.
The Deviation angle (Dv) is the fixed angle that forms
the center of the entrance slit, the center of the WDE, and Lb to the detector plane at the focal field g such that:
the center of the exit slit and is given by:

Dv ¼ b
a ð2Þ dk 106 cos b  cos g

¼
dx ðknLbÞ
In a spectrograph where the location of an array detector
is fixed, there will be a different value of Dv for the rays In summary, linear dispersion varies with wavelength as a
striking the first pixel on the array when compared with function of the angle of diffraction (b), the distance Lb at
that for rays striking the last pixel on the array, corre- each wavelength, and the inclination (g) of Lb to the focal
sponding to kmax and kmin. field.
The value of Dv is a constant of the system and does
not change if the groove density of the diffraction grating Blazing or Wavelength Optimized
is changed. However, the wavelengths appearing at kmax Diffraction Gratings
and kmin will change according to Eq. (1). To vary the Ruled diffraction gratings are optimized for maximum
spectral range on the array, the WDE will be rotated. efficiency at only one wavelength known as the ‘‘blaze’’
wavelength. Classically ruled gratings have a triangular
Angular Dispersion
shape with a blaze angle that is chosen to produce maxi-
The angular spread between two wavelengths is given by: mum diffraction efficiency at a particular wavelength in a
Littrow configuration (when the angle of incidence equals
db kn  10
6 the angle of diffraction, as shown in Figure 2.
¼ ðradiansÞ
dk cos b The diffraction grating Eq. (1) in Littrow simplifies to:

where n, groove density; k, order; db, the angle in radians; 2 sinðxÞ ¼ knk ð4Þ
dk, the separation between two wavelengths in nan-
ometers (nm). where x 5 a 5 b.
Catalogs typically specify diffraction gratings in terms of
Linear Dispersion
groove density (n), blaze wavelength, and blaze angle. A
Linear dispersion is measured in nanometers per milli- Littrow geometry is not commonly used, and in a spectro-
meter (nm/mm) and defines the extent to which a spec- graph is impossible, except at a single wavelength; how-
tral interval is physically spread out across a focal field in ever, deviations from Littrow rarely make a big difference
millimeters. A lower dispersion value increases the dis- to an efficiency profile when deviation angles are less than
tance between wavelengths and the potential for higher 40
. The use of the term ‘‘diffraction efficiency’’ always
spectral resolution. Linear dispersion is wavelength-speci- means the reflectivity relative to the metal coating the
fic and is measured perpendicular to the exit ray Lb at the grating at the wavelength in question. For example, if a
wavelength of interest. grating is coated with gold or aluminum, and is 50% reflec-
tive at a particular wavelength, then a diffraction grating
dk 106 cos b quoted with 50% diffraction efficiency will diffract 25% of
¼ ð3Þ
dx ðknLbÞ the incident light at that wavelength. The remainder will
appear in other orders including the zero order (a reflec-
Dispersion in a spectrograph varies with the length of Lb tion off the grating as if it were a mirror. See the following
as a function of wavelength and also by the inclination of section for detailed discussion on diffraction grating orders.)

Cytometry Part A DOI 10.1002/cyto.a

 IMAGING SPECTROMETER FUNDAMENTALS FOR RESEARCHERS IN THE BIOSCIENCES 715

FIG. 3. Virtually all diffraction gratings

diffract light into ‘‘orders,’’ with the ‘‘first’’
order used to present spectral data. How-
ever, all shorter wavelength multiples of
first order wavelengths will commingle
with the first order spectrum. For example,
400, 266.6, and 200 nm will be super-
imposed on 800 nm in first order; if they
are present in the light source. This can be
a significant spectral contamination pro-
blem when the quantum efficiency of the
detector is greater at shorter wavelengths
than longer wavelengths.

A good rule of thumb is to assume that relative diffrac- order possibilities, and increases overall first-order grating
tion efficiency will drop by 50% at 0.67 times the blaze efficiency.
wavelength and at 1.8 times the blaze wavelength. For Linear dispersion also varies linearly with order; in fact,
example, if the grating is blazed at 400 nm, the relative dif- Echelle grating spectrometers take advantage of increased
fraction efficiency level is above 50% over the range of dispersion to increase spectral resolution by working only
267–720 nm. in high orders. A prism is used following the grating for
‘‘order sorting’’ up to 80 diffraction orders. The net result
Diffraction Grating Orders is that Echelle spectrometers deliver very high-spectral re-
A diffraction grating acts like a mirror when the angle of solution (<0.1 nm, FWHM at 436 nm even though the
incidence (a) equals minus the angle of diffraction (a 5 groove density of the diffraction grating is only 52 g/mm),
2b). This is known as the ‘‘zero’’-order reflection. How- and can cover the wavelength range from 200 to 900 nm
ever, a diffraction grating can have an almost unlimited simultaneously (e.g., the ‘‘Mechelle,’’ Andor Corp, Belfast,
number of orders depending on the wavelength range and Northern Ireland).
wavelengths present in the light source. Figure 3 shows a Prisms work in a single order; therefore, their inherent
diffraction grating presenting 200–1,000 nm in first order efficiency profiles are significantly flatter and higher than
(k 5 1) in the focal field of a spectrograph. For given those of gratings and, of course, do not need order-sorting
values of a, b, and groove density, the grating Eq. (1) sim- filters. In addition, the wavelength transmission of most
plifies to: glasses is very high (>90%) over the majority of the visible
and near infrared (up to about 1,000 nm).
kk ¼ constant
Diffraction Gratings––Pros and Cons
From Eq. (4), we can see that a grating blazed when k 5 1 Classically ruled reflection gratings. These can be
is also blazed in all higher orders. For example, a grating very efficient at the blaze wavelength especially in Littrow.
blazed at 800 nm is also blazed at 400, 267, and 200 nm. Ion lasers rely on this property to produce high-energy,
All wavelengths are diffracted simultaneously; so all high-efficiency emission at target wavelengths. Wave-
orders, which can be present, will be present. Therefore, length blazing can be achieved anywhere in the spectrum
if 600 nm is diffracted into first order, then 300 nm will be from the X-ray to the far infrared, and is typically polariza-
present in second order, 200 nm in third order, and so on, tion dependant. Master ruled gratings can be replicated
and light from all orders will be commingled. Assuming with very high fidelity to make a very inexpensive optic.
that all wavelengths are present in the light source, the On the downside, high-groove-density gratings can pro-
only way to prevent higher orders from contaminating duce ‘‘ghosts’’ due to periodic ruling errors that vary as
first-order light is to use some form of blocking filter. If the square of the groove density (n) and order (k). Mod-
the wavelength range is from 200 to 399 nm, no order ern gratings have considerably reduced ghosting with the
sorting filters are needed, because wavelengths below use of interferometrically controlled ruling engines. Plane
200 nm are absorbed in the atmosphere. When wave- diffraction gratings must be used in conjunction with colli-
lengths appear in multiple orders, this causes reduced dif- mating and focusing optics. Stray light (scatter) can be up
fraction efficiency in first order. Increasing the groove to a factor of 10, greater than holographically produced
density and reducing the wavelength range reduces higher gratings.

Cytometry Part A DOI 10.1002/cyto.a

716 LERNER

FIG. 4. A simplified schematic of refraction through

a prism. Just as with diffraction gratings, there are
angles of incidence and refraction (rather than diffrac-
tion), and a need to collimate light onto the prism and
focus refracted wavelengths onto the wavelength dis-
persion plane.

Holographic surface relief diffraction gratings. when compared with that of a replicated surface relief dif-
HSRGs were developed in parallel, and essentially inde- fraction grating.
pendently, by Labeyrie and Flamand in France, and
Rudolph and Schmahl in Germany. HSRGs offer up to a Wavelength Dispersion Through a Prism
full-order of magnitude less stray light than classically
Sir Isaac Newton first described the properties of a
ruled gratings, absolutely no ghosts, and reduced polariza-
prism in 1670. Dr. Arnold Beckman produced the most
tion dependence. These advantages become significant at
successful UV spectrometer (arguably any spectrometer)
groove densities at or in excess of 600 g/mm, and/or for
ever built, the Beckman DU, by using a prism as the WDE.
use below 400 nm.
Indeed, Beckman Instruments owes much of its initial suc-
Basic HSRGs tend to offer less diffraction efficiency than
cess to this instrument.
a classically ruled grating, because the groove profile is si-
Figure 4 shows a simplified schematic of refraction
nusoidal. This has since been mitigated by ion-etching tri-
through a prism (4,12,13). Essentially, the operating pa-
angular grooves to ‘‘blaze’’ the groove profile (11). HSRGs
rameters (a,b,g, Lb, Lh) are identical to those in a diffrac-
are typically inefficient at wavelengths much above 2 lm
tion grating system. The same equations can be used to
where low groove densities are necessary, in which case
determine angular and linear dispersion, but cannot be
classically ruled gratings are superior.
used to calculate the values of the angles of incidence and
Both classically ruled and holographic surface relief
refraction (a,b). Most classical prism spectrometers, like
gratings are routinely replicated with very high fidelity
diffraction gratings, cannot be used without collimating
and make a very inexpensive optic, especially considering
and focusing optics, shown in Figure 4 as CO and FO.
the high level of technology and expertise that goes into
The angle of refraction (b) varies with the refractive
making them.
index of the prism material at each wavelength. For a flint
Volume holographic gratings. The diffraction effi-
glass prism, the linear wavelength dispersion at 436 nm is
ciency at blaze of a VHG can be high (>70%) over a parti-
about four times greater than that at 611 nm, delivering
cular wavelength range, but like plane reflection gratings,
approximately four times higher spectral resolution in the
they need collimating and focusing optics when used as
blue than in the red. Nonlinear wavelength dispersion
the WDE in a spectrometer. Mirrors make the best choice
makes it easy to ‘‘pack’’ an extended wavelength range
for collimating and focusing light because they are free of
onto a smaller detector array size. For example, the wave-
chromatic aberration; however, because a VHG usually
length range from 400 to 800 nm can be accommodated
works in transmission, there is a tendency for instrument
by a ½-inch array chip when compared with a diffraction
designers to use lenses as collimators and focusers to pro-
grating, which would require a 2/3-inch chip to provide
vide an approximately in-line system.
the same spectral range and competitive spectral resolu-
The lens approach is not without problems due to
tion.
reflections off lens edges, ghosting off lens surfaces, resid-
ual aberrations, including astigmatism at peripheral wave-
Pros and Cons of Prisms
lengths, degraded diffracted wavefront, and a restricted
wavelength range that can be limited by the degree of A prism offers a wide range of glass materials to ensure
chromatic aberration correction over the operating wave- very high-transmission efficiency from 400 to 1,000 nm
length range. For low-resolution systems, a VHG can be a (>90%), good transmission efficiency down to 360 nm
good solution, but the cost of a VHG can be very high (40%), and optimum refractive index characteristics over

Cytometry Part A DOI 10.1002/cyto.a

 IMAGING SPECTROMETER FUNDAMENTALS FOR RESEARCHERS IN THE BIOSCIENCES 717

FIG. 5. Generalized format of a wave-

length dispersive spectrometer. An object
in the FOV with an area ‘‘S’’ is imaged onto
the entrance aperture (slit or a pinhole) of
a spectrometer with an area S1. The aper-
ture must accommodate the full size of the
image of the object to prevent loss of light.
Light is collimated onto the WDE and then
focused onto the exit aperture. Light pas-
sing through the exit aperture can either
be measured or used to excite or illumi-
nate an object. All apertures must match
the size of the image of ‘‘S,’’ and all optics
must accommodate the incident numerical
aperture to prevent vignetting.

almost any wavelength range of choice. Light throughput is If the spectrometer is designed to operate as a mono-
enhanced because wavelengths are only refracted into a chromator, then each wavelength is selected sequentially
single order. Scattered light characteristics are exception- by rotating the WDE, and an increment of the spectrum,
ally good; exceeding most, if not all diffraction gratings. with a given bandpass, passes through an exit aperture in
It is relatively easy to exchange one diffraction grating a fixed location Figure 1a. After the aperture, the mono-
with a particular groove density with another of different chromatic light can either be used to illuminate or excite
groove density, whereas this is impractical with a prism. a sample or it can be passed to a measurement device,
In general, diffraction gratings provide significantly higher such as a PMT or diode. If the spectrometer is configured
spectral resolution than prisms in monochromator mode. as a spectrograph, then an entire spectral range can be
In spectrograph mode, it is the width of the entrance aper- imaged simultaneously onto an array detector. In special
ture and the length of the detector that determines band- cases, a spectrograph can be designed to correct severe
pass and wavelength range; consequently, in this mode, aberrations, such as astigmatism, spherical aberration,
the spectral resolution of prisms and diffraction gratings coma, and correct field curvature, for use with a matrix
are competitive. array detector, such as a CCD. Aberration-corrected instru-
ments of this type can be used for spectral topographical
mapping to create images with considerable data content
GEOMETRIC OPTICAL CONSIDERATIONS
(4,5,14).
OF WAVELENGTH DISPERSIVE
Aperture matching ensures lossless energy trans-
SPECTROMETERS
fer. Figure 5 depicts an object in the FOV with an area S
Wavelength Dispersive Spectrometer Designs
that is imaged onto an entrance aperture by an optic L1.
Image transfer basics. Wavelength dispersive spectro- The areas of S, S1, and S2 will never be equal in a WDE-
meters come in many optical configurations; however, based spectrometer; consequently, a failure to match aper-
they all follow a generalized format consistent with Figure ture sizes will either result in lost photons at the detector,
5. Regardless of the WDE in the spectrometer, light has to or, if the apertures are too large, will suffer from increased
be collected from an object in the field of view (FOV). In stray light.
an optimized system, optic L1 can be the microscope First let us consider the effects of magnification, or
objective or a telescope, and projects an image of an demagnification, on photon flux density taking the en-
object with area, S, onto the entrance aperture of the trance optic L1 as an example:
spectrometer at a NA consistent with the NA of the optics
of the spectrometer. To prevent vignetting, the entrance Magnification ¼ SQRTðS1 =SÞ ¼ q=p ¼ sin X1 = sin X
aperture must be matched in size to the image of the
¼ NAin =NAout ¼ F numberout =F numberin
object with area, S1.
As shown, an optic, CO, collimates light passing ð5Þ
through the entrance aperture onto the WDE. The WDE
diffracts or refracts collimated, wavelength-dispersed light The same relationships apply following the entrance slit
that is focused by an optic, FO, onto an exit aperture or through the spectrometer, where SQRT is the square root;
array detector. The exit aperture should be matched to S is the location and area of an object in the FOV; S1, S2
the size of the wavelength dispersed image of the en- are the areas of projected images of S; p is the distance of
trance aperture (S2). A photomultiplier tube, linear array, the object with area S from optic L1; q is the distance from
or CCD can then measure photon flux at each wave- L1 to the entrance aperture; La is the distance from center
length. The geometry illustrated in Figure 5 also applies to of the entrance aperture to the center of the first active
an ACHG and other active optic configurations in which (often collimating) optic, or the center of a concave grat-
case CO and/or FO can be ignored. The basic principle, ing; Lb is the distance from the center of the focusing
however, remains intact. optic to the exit aperture or from the center of a concave

Cytometry Part A DOI 10.1002/cyto.a

718 LERNER

grating to the exit aperture; X is the half angle subtended tells us that the optics must accommodate the cone angles
by L1; X1 is the half angle subtended by focusing optic FO. of light passing through the system.
NAin is NA into L1 (given by sin X); NAout is NA leaving L1 In confocal microscopy, the Etendue equation can be
(given by sin X1); and F number is F# 5 projected width used to test and determine how well a system is opti-
of an optic/image distance (e.g., F#in of L1 5 diameter/p; mized, and the functional area of a laser-excited sample.
F#out 5 L1/q). We show an example of this later in the paper.
NA ¼ l sin X ð6Þ Real Life Bandpass and Resolution Characteristics
of a Spectrometer
where l is the angle of refraction. For most spectro-
meters, this is unity because spectrometers almost always Theoretical resolution of a diffraction grating. The
operate in air. ‘‘resolving power,’’ R, of a diffraction grating is at best an
impractical, theoretical concept and is given by:
F number ðF#Þ ¼ 1=ð2NAÞ ð7Þ
R ¼ k=dk
(Note: F number should be calculated from the NA not
vice versa. This is due to the difficulty in identifying the where dk is the difference in wavelength between two
position of the principle plane that defines the ‘‘effective spectral lines of equal intensity. Resolution is the ultimate
diameter,’’ or projected width, of a lens or mirror used in ability of an instrument to separate two spectral lines. By
an off-axis configuration. When tan a (tan a 5 width the Raleigh criterion, two peaks are considered resolved
WDE/Lb) is approximately equal to sin a, Eq. (7) is reversi- when the maximum of one falls on the first minimum of
ble; consequently, when NA >0.16, we can use Eq. (7) to the other. It can be shown that:
convert F number to NA and back again. NA is an absolute
value; F number is a relative value.) R ¼ k=dk ¼ knWg ¼ kN
As we progress through the remaining optics in the sys-
tem, the image of S will undergo magnification and/or where k, the central wavelength to be resolved; Wg, the
demagnification and photon density will be given by: illuminated width of the grating; and N, the total number
of grooves on the illuminated width of the grating.
Photon density ¼ S1 =S ¼ ðq=pÞ2 ¼ ðNAint =NAout Þ2 Actual spectral resolution and bandpass depends on the
ð8Þ width of the entrance aperture and the focal length of the
¼ ðF#out =F#in Þ2 system; so numerical resolution, R, should not be con-
fused with observed resolution or bandpass of an instru-
The F# and NA relationships in Eq. (8) will be familiar to ment system. Hence, the original comment that resolving
photographers and fluorescence imagers, because they power is an impractical concept. It is only included for
directly govern relative exposure time at constant magnifi- completeness.
cation. Observed instrumental spectral resolution. Band-
pass and resolution can be easily determined in any instru-
Calculating Light Throughput
ment by using a light source that emits a spectrum with a
Nothing is more important than ensuring that all avail- pure monochromatic line, k0. Figure 6a shows the natural
able light is transferred through the system. Geometric light ‘‘real’’ line width. Figure 6b shows how it would be char-
throughput or Etendue defines the ability of an optical sys- acterized by a perfect spectrometer; so Figure 6b should
tem to accept and transfer light. Etendue, also known as the be identical to Figure 6a.
‘‘geometric extent,’’ is a constant of an optical system and Spectrometers are not perfect and record a line spectrum
represents the ‘‘bottleneck’’ when considering the transfer with finite width. This is known as the ‘‘instrumental line
of light. A failure to preserve the nominal value of the Eten- profile’’ and can be determined by characterizing the spec-
due will result in a loss of real signal. Using Figure 5: trum of a single mode laser or with a low-pressure Hg1/Ar1
emission lamp with the entrance and exit slits at minimum
Etendue ¼ G ¼ Sðsin XÞ2 ¼ S1 ðsin X1 Þ2 width. The bandpass is the full width at half maximum
¼ S2 ðsin X2 Þ2 ¼ ::::: (FWHM) of the recorded spectrum Figure 6c.
In its simplest case, the bandpass of a spectral feature
We can replace (sin X) with NA, and the Etendue equation presented by the FOV is influenced by its natural line
width, the influence of the slits, and the resolution of the
simplifies to:
instrument. For a monochromatic emission and a
G ¼ S ðNAÞ2 ð9Þ high-resolution spectrometer, the instrumental bandpass
(BP) is given by:
The Etendue equation enables us to optimize the light
throughput of any series of optics to ensure that the maxi- BPsw ¼ Disp  Wexap ð10Þ
mum available real signal is either passed to the detector
or illuminates a sample. It tells us that the areas of aper- where Disp is the linear wavelength dispersion (in nm/
tures must be matched to collect all available light. It also mm) at a particular wavelength; Wexap is the width of the

Cytometry Part A DOI 10.1002/cyto.a

 IMAGING SPECTROMETER FUNDAMENTALS FOR RESEARCHERS IN THE BIOSCIENCES 719

FIG. 6. The natural spectrum of a pure monochromatic light source (a); the same light source imaged through a theoretically ‘‘perfect’’ spectrometer (b);
the pure monochromatic light imaged through a real-life spectrometer (c). The finite bandwidth (FWHM) is an instrumental function that is imposed on the
natural bandwidth of the monochromatic light.

exit aperture or width of the image of the entrance aper- ble to calculate the linear dispersion of an instrument
ture, whichever is greater. when the bandpass and distance one bandpass occupies
The total recorded bandpass, BPnet, for an emission are known. For example, if the light source is a monochro-
with finite spectral bandwidth, such as a fluorescence matic emission line from a low-pressure Hg lamp and the
emission, assuming an approximately Gaussian distribu- FWHM in Figure 6c is 1 nm, and occupies three 9 lm pix-
tion, is given by the ‘‘generalized bandpass equation’’ for a els on a CCD, then we know that the linear dispersion at
real emission: 436 nm is 37 nm/mm (1/3*0.009) corresponding to BPnet
in Eq. (11). Therefore, by knowing the functional operat-
BPnet ¼ SQRTðBP2nat þ BP2slit þ BP2res Þ ð11Þ ing characteristics of a spectrometer and rearranging
Eq. (11), it is possible to reveal the natural spectral band-
where BPnet is the net bandpass after accommodating the width of an emitting source.
finite emission bandwidth of the light source and instru- If a light source emits a continuum, then the resolution
mental factors; SQRT is the square root; BPnat is the nat- is the smallest spectral increment that can be isolated, and
ural spectral bandwidth of the emitting source; BPslit is the bandpass is a user selected spectral increment.
the bandpass determined by the bandpass Eq. (10); BPres
is the limiting resolution of the instrument (ultimate band-
Magnification and System Anamorphism
pass with a line emission source).
In real world acquisitions, the FWHM of a typical fluo- We know from Eq. (10) that bandpass is determined by
rescence emission is significantly greater than the limiting the product of linear dispersion, and either the image of
spectral resolution of the instrument; therefore, the the entrance aperture or the exit aperture, whichever is
reported resolution will be dominated by the bandpass greater. However, the width of the image of the entrance
determined by the slit width and the natural spectral slit aperture, shown as W * in Figure 7, for either a mono-
bandwidth of the natural emission spectrum. chromator or a spectrograph varies with wavelength.
In essence, real life bandpass and resolution indicate Note that W * 5 Wexap when the width of the image of
the limits of a finite instrument’s ability to separate real ad- the entrance aperture is the same as the exit aperture. In a
jacent spectral features. Bandpass is set by the user, and confocal system or a spectrometer with a fiber optic feed,
resolution is limited by the functional limits of the instru- the aperture could be a pinhole or a circle, rather than a slit.
ment. The smallest possible bandpass is the resolution, The projected height can be determined by considering
and is determined when the FWHM of a monochromatic the magnification or demagnification through the system.
emission line is not reduced even when the slit width con- Sometimes, it is more convenient to use F number, rather
tinues to be narrowed. than NA. We can calculate the projected width of the en-
By rearranging Eq. (3), Disp 5 BP/Wexap; where Wexap is trance and exit slits by W cos a, and W * cos b as perceived
the FWHM of a monochromatic emission line, it is possi- by the WDE, and divide these values by the entrance

Cytometry Part A DOI 10.1002/cyto.a

720 LERNER

FIG. 7. System anamorphism results

in an image of the entrance aperture ap-
pearing with a different width at each
wavelength. The height of the slit is a
constant in a monochromator, when Lb
is fixed, (h 5 h*) and a variable in a
spectrograph when there is a different
Lb for each wavelength (h „ h*).

and exit arm lengths to obtain the input and output F creases with wavelength; however, in a prism system,
numbers. the angle of refraction varies inversely with wavelength.
W  cos b The height of the image of the entrance slit in the exit
F#out ¼ plane is determined by the ratio of the arm lengths
Lb
alone:
W cos a
F#in ¼
La h Lb
h ¼
La
Therefore:
W cos a Lb
W ¼ ð12Þ In the case of a spectrograph, the arm length Lb varies
cos b La with wavelength; consequently, the vertical magnification
also varies with wavelength. In a monochromator, the
We can now substitute Eqs. (3) and (12) into Eq. (10) to length Lb will be fixed unless the exit slit is translated
obtain the relationship: across the spectrum to select wavelength, such as in a
Leica SP series spectral confocal system.
W cos a Lb 106 cos b
BP ¼
cos b La knLb
3ðprojected slit width multiplied by linear dispersionÞ How to Estimate the Operating Conditions
of a Spectral Confocal Microscope
106 W cos a The equations listed in the previous sections are those
BP ¼ ð13Þ that are needed to estimate the groove densities of diffrac-
knLa
tion gratings, focal length, changes in wavelength disper-
From Eqs. (13) and (3), we note that bandpass is a func- sion, theoretical bandpass, magnification of the pinhole,
tion of the angle of incidence, and linear dispersion a func- and the optical geometry (Dv angles). All we need to
tion of the angle of diffraction, or refraction. know in advance is the observed wavelength-range over a
Using Eq. (12), we can match the width of the exit known width of an array detector.
slit to W * or assign the correct number of detector ele- Nikon C1-Si spectral imaging system with an IPMT
ments in a linear or matrix array so as to maximize real linear array detector. Let us take as an example the
signal throughput. This is of key importance for a diffrac- Nikon C1-Si spectral imaging system. From published liter-
tion grating system with a wide wavelength range and a ature, we are informed that a user can sample a spectrum
long kmax. The reverse is true for a prism-based system. in 2.5, 5, or 10 nm increments, in what Nikon refers to
With a diffraction grating, the angle of diffraction in- ‘‘wavelength resolution,’’ by exchanging three diffraction

Cytometry Part A DOI 10.1002/cyto.a

 IMAGING SPECTROMETER FUNDAMENTALS FOR RESEARCHERS IN THE BIOSCIENCES 721
Table 1
Estimated Geometric Parameters for a WSI of 2.5 nm
Observed (F )
wavelength Eq. (1) Eq. (2) Eq. (12) Eq. (3)
range (nm) (C ) b (
) (C ) Dv (
) (C ) Pinhole (mm) (C ) Disp (nm/mm)
531 11.04 215.46 0.091 2.46
569 13.70 212.80 0.092 2.44
607 16.39 210.11 0.093 2.41
WSI 5 2.5 nm; total length of the IMPT array detector 5 32 mm; diffraction grating 5 1,200 g/mm; angle of incidence a 5 26.5
; focal
length 332 mm; pinhole diameter 5 100 mm. (C ) 5 computed, (F ) 5 actual observed values. The value of a was iterated; and Lb was cal-
culated using the wavelength dispersion at the center of the chip of 2.44 nm/mm. This simple approach provides a good approximation.

gratings. It is more accurate to refer to these settings as cise value of Lb using 2.44 nm/mm dispersion at the cen-
the ‘‘wavelength sampling increment’’ (WSI). The use of ter of the chip at 569 nm (the wavelength that falls on the
three gratings implies that the wavelength dispersion will center detector element), using Eq. (3).
increase by a factor of two and four, based on the 10 nm This process will display the Dv values at the extremes
condition. From Eq. (3), we know that to change disper- and the center of the wavelength range. The Dv values are
sion we must either change the focal length (Lb) or the then fixed and will not change when we select alternative
groove density (n). Consequently, we know that it is the groove densities to obtain the 5 and 10 nm WSI values.
groove density that will change because Lb is fixed. For this We find by iteration that a 1,200 g/mm grating provides
exercise, we took the pinhole to be 100 lm in diameter. the reasonable solution shown in Table 1. We also note
Observed wavelength scans acquired with a 2.5 nm that the image of the pinhole appears to demagnify in the
WSI places the wavelength range of 531–607 nm across a dispersion plane; however, in these calculations, we
32 element, imaging photomultiplier tube (IPMT), linear assume that the entrance and exit arm lengths are the
array detector (Hamamatsu Corp, Bridgewater, NJ) in a same, and that the pinhole will change in size only as a
single shot. The elements are on 1 mm centers with an function of the cosines of the angles of incidence and dif-
active width of 800 lm, 7 mm in height, with 200 lm fraction. In an actual instrument, unequal arm lengths are
dead-space between elements, and each spectrum will be possible and are certainly different across the array. Chan-
characterized by up to 32 wavelength data points (WDP). ging values of Lb with wavelength will contribute to the
(Note: As described the term ‘‘wavelength resolution’’ is pinhole magnification.
really the WSI. The actual resolution, calculated by mea- To summarize the results:
suring the FWHM of a monochromatic emission line, will
be between 2.5, 5, or 10 nm only when an image of the Lb (focal length) 5 332 mm.
entrance pinhole strikes the center of an IPMT detector a (the angle of incidence) 5 of 26.5
for the wave-
element.) length range from 531 to 607 nm.
Therefore, the wavelength dispersion is 75.6 nm (607– Diffraction grating 5 1,200 g/mm
531.4) spread over 31 mm (allowing 2 3 0.5 mm from
center-to-center) for an average wavelength dispersion of To select an alternate wavelength range, the grating would
2.44 nm/mm (76/31). From diffraction grating catalogs, be rotated to change the angle of incidence while keeping
we note that off-the-shelf gratings are available in 150, Lb and the Dv angles constant.
300, 600, 1,200, 1,800, and 2,400 g/mm, blazed at a vari- To determine a for a 10 nm WSI, the groove density of
ety of wavelengths. (Horiba/Jobin Yvon Edison NJ, New- the diffraction grating must be four times less than the
port Corp-Richardson Gratings, Rochester NY). 1,200 g/mm grating used earlier. By using a 300 g/mm
The easiest way to estimate the geometric parameters is grating, and keeping Lb and the Dv angles constant, we
to construct Table 1 in an Excel spreadsheet. Then, select vary a until we reach the desired wavelength range
a catalog diffraction grating, vary a, and calculate the pre- shown in Table 2. Here, the wavelength range from 406 to

Table 2
Estimated Geometric Parameters for a WSI of 10 nm

Eq. (1) (D)

wavelength Eq. (2) (F ) Absolute Dv Eq. (12) (C ) Eq. (3)
range (nm) (C ) b (
) [From Table 1] Pinhole (mm) Disp (nm/mm)
406 24.21 215.46 0.098 10.01
560 21.55 212.80 0.098 10.04
717 1.14 210.11 0.098 10.04
WSI 5 10 nm; total length of the IMPT array detector 5 32 mm; diffraction grating 5 300 g/mm; angle of incidence 5 11.25
; focal
length 332 mm; pinhole diameter 5 100 mm. (C ) 5 computed; (D) 5 derived by varying a. Lb and Dv values were fixed in Table 1.

Cytometry Part A DOI 10.1002/cyto.a

722 LERNER

consequently, the actual instrument may have an alterna-

tive geometry.

DIFFRACTION GRATING BASED

SPECTROMETER SYSTEMS
Spectrometers characterize the spectral characteristics
of an object by measuring wavelength intensity as a func-
tion of wavelength. Spectrometers work in reflection,
transmission, absorption, fluorescence, Raman, and all
forms of luminescence and phosphorescence whether
chemically or electrically induced. If wavelengths are
acquired sequentially by rotating the WDE, then system is
a monochromator. If all wavelengths in a specific range
are acquired simultaneously, then it is a spectrograph. The
spectrometer systems that follow can all be made to oper-
FIG. 8. Nikon C1-Si: simultaneously acquisition of the LightForm MIDL ate as a monochromator; however, we are going to focus
lamp spectrum with a 10 nm WSI. The wavelength accuracy and contrast on spectrograph systems due to their increasing impor-
(peak to valley ratio measured at PV1) are good and well within expecta- tance to the life science community.
tions. It is evident that the image of the pinhole at the Hg 545 nm line
strikes a single detector element and the 611 nm line is split between two The literature describes a wide variety of spectrometer
elements. Values in parentheses are the true wavelength values. designs, but we will concentrate on the most common
and arguably the most successful. These include the
Czerny turner (CT), concave diffraction grating (CDG),
717 nm (311 nm) is acquired in a single shot in 10 nm and aberration corrected holographic diffraction gratings
increments with a 5 11.25
. The results appear to be con- (ACHG), and prism systems.
sistent with observations. As is implied by the above-mentioned list, all spectro-
To keep astigmatism and coma to a minimum, the NA of graphs do not perform equally well. Without implanting
the system should be kept to a minimum. Astigmatism var- specific corrections, spectrometers are subject to field
ies with the square of the NA and coma with the cube of curvature, astigmatism and a variety of off-axis aberrations
the NA. The projected NA of a 633 objective (NA 5 1.32) and anamorphism. Astigmatism does not have a dramatic
is 0.02 (1.32/63), and a 103 (0.3 NA) projects at 0.03. If effect on spectral resolution, but it significantly degrades
we used a 30-mm wide diffraction grating, the NA of the spatial resolution. Field curvature will not support a linear
spectrometer would be 0.05 (F/10) and would accommo- array, such as an IPMT or a CCD. Of the two, field-curva-
date the NA of both the high- and low-magnification ture is the easiest to correct (15–17).
microscope objectives. In relative terms, an NA of 0.05 is
low enough to keep aberrations to a minimum, and the Field-Flattened Czerny Turner Spectrograph
large size of the detector elements will mask whatever re-
Most spectrometer systems, such as those found in an-
sidual aberrations remain.
alytical laboratories, are incapable of determining the
Figure 8 shows an acquisition of the MIDL wavelength
location of a spectrum presented by an object in the FOV,
calibration lamp with a 10-nm WSI taken in a single shot
and are designed to characterize samples in a cuvette or
(see section titled ‘‘Wavelength Calibration of Spectral Sys-
capillary.
tems’’ for details). The emission maxima match the abso-
Figure 9 illustrates a CT design that significantly flattens
lute values very well considering that each acquisition
the focal plane to accommodate array detectors, including
was 10 nm wide. The scan also illustrates that, when an
the previously mentioned Hamamatsu IPMT. A field-flat-
image of the pinhole strikes a single detector element, the
tened CT (FFCT) design is asymmetric, and the focal field
bandpass measured at the FWHM of the 545 nm peak is
is tilted in the wavelength dispersion plane. In this geome-
indeed 10 nm. However, the 611 nm line clearly falls
try, u and u1 are unequal, and distances La and Lb may also
between two detector elements reducing the apparent
be unequal. Although corrected for focus, spherical aber-
FWHM and peak intensity by a factor of two. This is alias-
ration, and coma, astigmatism remains a problem; conse-
ing and is most noticeable when the WSI is large.
quently, a FFCT is very well suited to an IPMT with large
Although expected, this effect can cause unreliable inten-
rectangular detector elements.
sity ratios between wavelengths. Changing the initial
wavelength of the scan can push a particular wavelength
Using a Spectrograph with a Linear Array or CCD
from one detector to another and change intensity ratios
as a Wavelength Detector
and FWHM values. This is a hazard or an opportunity
depending in whether the instrument operator is aware of We know that bandpass is a function of wavelength dis-
the consequences of changing a scan starting wavelength. persion and the greater of the image of the entrance aper-
Note that although we derived a plausible solution ture or the exit aperture Eq. (10). Therefore, the detector
there are many variables. For example, Nikon may have element size plays a critical role in determining both spec-
used custom diffraction gratings and/or focusing optics; tral and spatial resolution. A linear array, such as the pre-

Cytometry Part A DOI 10.1002/cyto.a

 IMAGING SPECTROMETER FUNDAMENTALS FOR RESEARCHERS IN THE BIOSCIENCES 723

FIG. 9. Asymmetric Czerny Turner designed to flat-

ten a typically curved focal to accommodate an array
detector. Residual astigmatism is irrelevant because
the height of the detector array captures all incident
light and there is no spatial resolution requirement.

viously identified IMPT, has large detector elements and can vary with wavelength in a monochromator and is a
these determine the limits of spectral resolution because constant in a spectrograph. The height of the optic is not
the size of the image of the entrance aperture is unlikely a variable, and rays distributed vertically along the WDE
to exceed the size of the detector elements. are brought to a sagittal focus. When the sagittal and tan-
In comparison, a CCD matrix array, such as the QICAM gential foci fail to coincide, the system is ‘‘astigmatic,’’ and
(Q-Imaging, Burnaby, Canada), is made up of 1,392 3 a point object at the entrance slit will be imaged as a line
1,040 lm2, 4.65 3 4.65 lm2 detector elements (pixels), in the exit plane (Fig. 10a).
each of which provides an individual measure of the Astigmatism causes two adjacent points to merge verti-
amount of light incident upon it. In this example, the cally, and the intersection of sagittal edge rays at the tan-
pixel size will almost always be smaller than the image of gential focus will determine the height of the astigmatism,
the entrance aperture; consequently, pixel size will not as shown in Figure 10a. In this example, each optic has
limit spectral bandpass or resolution. the same radius of curvature (ROC) in both the horizontal
Each spectrum incident on a CCD in an imaging spec- and vertical directions, where: R 5 Rt 5 Rs.
trometer will occupy a row (x-axis) of pixels and will not Astigmatism is evident perpendicular to the dispersion
produce an ‘‘image’’ that could be likened to a digital axis; consequently, the spectrometer illustrated in Figure
photograph. The y-axis correlates a position in the FOV 9 can be designed to deliver optimum spectral resolution,
with a position on the entrance slit with a spectrum in a but will not deliver good spatial resolution. This is a com-
row of pixels. If the entrance aperture is a pinhole, then mon design used with a wide variety of linear arrays.
only a few rows of pixels will be illuminated, and the de- Astigmatism is not a problem in a laser confocal spectral
tector should be a linear array, rather than a CCD, because system, because the laser submits single ‘‘points’’ from the
there will be no spatial component. FOV sequentially, not multiple points simultaneously. If ei-
ther an exit slit or detector elements are large enough to
Astigmatism and Spatial Resolution catch all available photons, irrespective of how much they
Poor spatial resolution is dominantly degraded by astig- are vertically spread or otherwise aberrated, then all
matism-an off-axis aberration that increases as the square energy from each point in the FOV will be captured and
of both the off-axis angle and NA. Figure 10 illustrates the quantified. As a result, laser-scanned spectral confocal sys-
origins and consequences of astigmatism. Light strikes the tems do not have to have an astigmatism-free spectrome-
WDE in a spectrometer along its width and its height. Rays ter. Similarly, classic spectrophotometers used to charac-
striking along the width of the object are brought to a tan- terize a solution in a cuvette use either an exit slit or
gential focus. The width of the optic is a variable because detector, which is tall enough to collect all vertically dis-
it depends on the cosine of the angle of incidence, which tributed light.

Cytometry Part A DOI 10.1002/cyto.a

724 LERNER

FIG. 10. (a) All wavelength dispersive spectrometers operate off-axis resulting in astigmatism that presents a point on the entrance slit as a line at the exit plane. (b) Astigmatism can be corrected with
The most obvious solution to the astigmatism problem
is to bring the two foci together by making mirrors with a
sagittal ROC (Rs) curve that will bend the sagittal focus
onto the tangential focus and Rs „ Rt. Figure 10b shows
an optic with different radii of curvature along the vertical
and horizontal axes. The two curves are mechanically
ground into the blank of collimating or focusing mirrors,
or recorded into the hologram of a holographic diffraction
grating.

Astigmatism Corrected CT Spectrometer

Figure 11 shows an example of a nearly stigmatic CT
where either the collimating or focusing mirror is a toroid.
This is a successful solution that is stigmatic at one wave-
length and much improved over a specific wavelength
range. It does not work as well when the entrance slit is
particularly high, or for a highly extended wavelength
range.
A stigmatic spectrometer images points along the en-
trance slit as spectra with about the same height as a point
on the entrance slit. There will be geometric magnifica-
tion due to differences in the angles of incidence and dif-
fraction (or refraction), Eq. (12) as well as differences in
path length along a linear array. Pure 1:1 imaging is impos-
sible at all wavelengths. If the source is monochromatic,
then a point will be imaged as a point.
We can assume that a stigmatic spectrometer will use a
matrix array, such as a CCD, as the wavelength detector.
The actual width of each ‘‘point’’ will be constrained by
toroidal optics with the result that a point on the entrance slit is imaged as a point in the exit plane.

the height of a row of pixels on the CCD, plus any other

residual aberrations that may be present. The greater the
number of points on the entrance slit that can be differen-
tiated, the better the spatial resolution and imaging capa-
city will be.
If a spectrometer is to be used as a source of excitation
energy, then astigmatism is a significant problem because
the photon density is reduced by any increase in area at
the exit of the spectrometer. For illumination or excitation
purposes and spectral topographical mapping, the spec-
trometer should be stigmatic.

Classical Concave Grating Options

Astigmatism is also observed with a classical concave
diffraction grating operating on the Rowland circle (RC).
The diameter of the RC is the radius of curvature of the
grating, as shown in Figure 12. If the entrance slit is
located on the RC, then all diffracted wavelengths will
also be focused on the RC. The benefit of this configura-
tion is the simplicity of the design, which does not need
any collimating or focusing optics and accommodates a
very wide spectral range. However, it is not capable of
point-to-point imaging because of astigmatism. The con-
siderable field curvature makes the use of linear arrays or
CCD chips impractical. This design is widely used in many
high-resolution ‘‘direct-reading’’ spectrometers found in
analytical laboratories (18).

Cytometry Part A DOI 10.1002/cyto.a

 IMAGING SPECTROMETER FUNDAMENTALS FOR RESEARCHERS IN THE BIOSCIENCES 725

FIG. 11. A toroidal focusing or collimating

mirror bends the sagittal focus onto the tangen-
tial focus.

Holographic Recording of Aberration-Corrected The theory behind correcting the aberrations of concave
Concave Gratings gratings is described in the literature. It is enough to say
that a classical concave grating has grooves that are equidis-
Aberration-corrected holographic gratings (ACHGs) can
tant, whereas ACHG are recorded with asymmetrically dis-
be recorded in such a way that a spectrum falls on a flat-
tributed grooves, and no longer operate on the RC (Fig.
field with greatly reduced or eliminated astigmatism. No
13). The asymmetry of the grooves is somewhat analogous
focusing or collimating optics are required and it can be
to using a toroidal optic with asymmetric radii of curvature.
well corrected over a significant wavelength range. For additional correction, the usual spherical grating blank
of an ACHG can also be both toroidal and be recorded with
asymmetrically distributed grooves (3,6,9, 10,19,20).
Balaban et al. at the National Institute of Health ele-
gantly demonstrated the use of an ACHG flat-field spectro-
graph grating (UFS 200, Horiba Jobin-Yvon) in 1985, by
generating fluorescence spectral acquisitions across the
nucleus of a living trophoblast cell, using a SIT matrix
array camera as the wavelength detector (21).
By 1990, Benedetti and Evangelista at the Instituto di
Biofisica in Italy used a more advanced ACHG flat-field
spectrograph grating (Horiba Jobin Yvon ref model
52300070) coupled to a CCD detector to perform spectro-
scopic, slit-confocal microscopy. A conventional confocal
system takes a point excitation at the sample and reimages
it onto a circular aperture, whereas in this case, the sam-
ple was excited by a ‘‘line’’ of light, rather than a point.
The areas of the sample that fluoresced under the influ-
ence of the line excitation were projected onto the slit of
the spectrometer to make a slit-confocal spectral micro-
scope, thereby, reducing scatter and enhancing contrast.
The full spectral data from each point along the slit was
captured simultaneously (as illustrated in Fig. 13), and the
FIG. 12. Concave diffraction grating operating on the Rowland circle sample was translated sequentially until the entire FOV
(RC). Light entering on the RC will be diffracted and focused on the RC.
Spectral resolution is high and it is not capable of spatial resolution due to was acquired. Software then reconstructed the spectral in-
astigmatism. formation into a spectral topographical map (22).

Cytometry Part A DOI 10.1002/cyto.a

726 LERNER

FIG. 13. Holographically aberration-corrected concave grating spectrometer. This solution does not require any ancillary collimating or focusing optics.

PRISM-BASED SPECTROMETER SYSTEMS standards, and also able to correct astigmatism present in
Classical Prism Spectrometer Designs the off-axis refracted light, can be a challenge. Conse-
Figure 14 shows a generic wavelength dispersive prism quently, it is not unusual to find a variety of residual aber-
system with a typically curved focal field. Optic CO colli- rations, including astigmatism, coma, and spherical aberra-
mates light from the FOV onto the prism at an angle a. Af- tion in the focal field.
ter refraction, each wavelength exits at an angle b. The sim- This or a similar prism geometry is probably used in the
ilarity to a diffractive system is self-evident with the same Leica SP series spectral confocal microscope, using the
influences that contribute to aberrations. However, prisms confocal pinhole as the entrance aperture of the spec-
only refract a single order; consequently, light transmission trometer. The system does not need elaborate aberration
efficiency can be very high and only depends on the wave- correction, because the image of the pinhole is imaged
length and the material that is used to make the prism. not onto another circular aperture, but a slit. Any astigma-
The most effective means to collimate and focus light in tism is accommodated by making the exit slit high enough
any wavelength dispersive system is with front surface to capture any vertical image enlargement. This works
concave mirrors; however, in order to produce a more because the laser excites the sample point-by-point
compact system, lenses are often used at CO and FO. Find- sequentially; so that there are never two objects being
ing lenses that are transmissive over all wavelengths, field imaged at the same time that could interfere with each
flattened and chromatically corrected to spectroscopic other. To acquire a complete spectrum, the exit slit is

FIG. 14. Generic wavelength-dispersive prism

system.

Cytometry Part A DOI 10.1002/cyto.a

 IMAGING SPECTROMETER FUNDAMENTALS FOR RESEARCHERS IN THE BIOSCIENCES 727

FIG. 15. PARISS aberration-corrected prism-

based spectrometer.

translated across the focal field and a single PMT measures spectrum from 365 to 950 nm acquired on the PARISS sys-
the signal at each wavelength. tem, with a single 10 ms exposure, using a CCD detector
with a 1-inch CCD chip (Retiga 2000R with 2 3 2 binning,
An Aberration-Corrected Prism-Based Q-Imaging Corp, Burnaby, BC, Canada) (see section titled
Imaging Spectrometer ‘‘Wavelength Calibration of Spectral Systems’’ for details).
The measured bandwidth of the 436 nm line is 1 nm.
In 1989, Warren and Hackwell, at the Aerospace Cor- Although it is not obvious, this spectrum is remarkable
poration in El Segundo, California, developed a highly ab- because Argon lines above 650 nm fade to extinction
erration-corrected, prism-based spectrograph. Originally within seconds of turning on the lamp. The fact that the
called SEBASS, and designed for use in the wavelength Hg 365 nm line and the Ar 912 nm line are both captured
range 2.9–13.5 lm, it used a telescope for light collection simultaneously highlights the power of a prism to rapidly
and a matrix array as the detector. After LightForm capture a very wide wavelength range with no contami-
acquired the license to the patent, it was redesigned for nating higher orders.
use in the wavelength range from 360 to 950 nm, and
named ‘‘PARISS’’ (prism and reflector imaging spectros- Spectral Resolution and CCD Pixel Density
copy system). PARISS was optimized for use in fluores-
The number of pixels in a CCD chip, when used as a
cence, absorption, and reflection microscopy with a digi-
wavelength detector in the focal plane of a spectrograph,
tal CCD as the detector (Fig. 15). PARISS acquires all wave-
has a very limited bearing on spectral resolution. The
lengths in the wavelength range from 400 to 800 nm
spectral resolution for a slit-based instrument is governed
simultaneously over a ½-inch CCD chip and from 365 to
by the generalized bandpass Eq. (11). Consequently, the
800 nm over a conventional 2/3rd inch chip (23,24).
spectral resolution in the dispersion plane (x-direction) is
This novel design uses a front surface concave mirror at
a function of the width of the slit and in the spatial (y-
finite conjugates (to create an image of an object) with a
direction) is governed by the height of a row of pixels on
small off-axis angle, and a highly optimized, computer
the CCD. In both cases, residual aberrations, magnifica-
designed prism, with concave and convex sides. The com-
tion, and the natural energy distribution (atypically Gaus-
bination of curves and distances from optic to optic pro-
sian) will each contribute to image enlargement.
duces an almost aberration-free flat focal plane that
For example, let us take the previously mentioned
accommodates a digital CCD camera. The curved sides of
the ‘‘prism’’ makes it closer to being a lens with a wedge. QICAM with 1,392 3 1,040 pixels each of which is 4.65 3
The design is made highly resistant to reflections and 4.65 lm2 in size. We know that an image of the entrance
ghosting by tilting the optical components so that no aperture is imaged onto the CCD at all wavelengths present
optic ‘‘looks’’ directly at any other. In this way, any reflec- in the emitting source simultaneously. As a practical exam-
tions pass harmlessly out of the optical path, and never ple, the PARISS system uses a 25 lm entrance slit width,
reach the detector. The light transmission efficiency over 5 mm in height and its operating geometry (La < Lb) magni-
the design range is >90% in the visible, and is capable of fies the width and height of the entrance slit by 10%
true point-to-point imaging over a wide FOV. Figure 16 resulting in an 27 lm wide image of the entrance slit.
shows the spectrum of a low-pressure Hg calibration lamp This corresponds to an observed bandpass of 1 nm, FWHM,

Cytometry Part A DOI 10.1002/cyto.a

728 LERNER

FIG. 16. Wavelength calibration spec-

trum of a low pressure Hg lamp. All wave-
lengths were acquired simultaneously
with the PARISS system using a Retiga
2000R with a 2/3rd inch chip. The emis-
sion lines above 650 nm are Ar1 lines that
have a very brief lifetime after the lamp is
turned on.

at the Hg 436 nm line. Therefore, the linear dispersion of chip; so it is to be expected that the spatial resolution will
the PARISS system at 436 nm is 37 nm/mm. be moderated at least by the Gaussian spread of the image
For a monochromatic emission wavelength, we know that at the entrance slit.
bandpass is the linear dispersion multiplied by either the At the CCD chip (where the image of the entrance slit is
width of the image of the entrance slit or the exit aperture, located), the spatial resolution along the slit height (y-axis
whichever is larger. However, because we are using a CCD, perpendicular to dispersion) is again limited by the height
we can select the number of pixels that will accommodate of a single row of pixels and the Gaussian energy in a point
the image of the entrance slit. Given that each pixel is 4.65 object. After binning to match the width of the entrance slit,
lm in width, the image of the entrance slit will occupy 6 the vertical resolution also approximates ~0.6 lm with a
pixels (the integer value of 27/4.65). However, by the Ray- 403 microscope objective. Therefore, we will observe up
leigh criterion, we only need three data points to define the to 240 spectra along the slit, corresponding to 240 objects,
FWHM, so that there will be no loss in spectral resolution by 0.6 3 0.6 lm2 in size, distributed along a linear slice of
binning pixels 2 3 2 and we will benefit by an increase in the FOV. If objects in the FOV are true point sources, such
linear dynamic range, and signal-to-noise ratio. It is also de- as nanoparticles or single molecules, then if only one such
monstrable that there was no measurable image enlarge- emitter is imaged through the slit, the slit acts as a field aper-
ment apart from the expected geometric magnification. ture. The image of the point source effectively defines the
entrance aperture and will fill one (or partially two) rows of
Spatial Resolution and CCD Pixel Density pixels. The accuracy of the location of the point object in
the FOV will continue to be 0.6 3 0.6 lm2.
The spatial resolution in the dispersion plane, at the
Spatial resolution is only limited when the object is
CCD, is given by the entrance slit width divided by the
extended in size. If the image of an object is smaller than
magnification of the light collection optic imaging the
the slit width, then it is the image of that object that acts
sample, such as the microscope objective. For example, if
as the entrance aperture.
we image an object in the FOV with a 403 objective onto
a slit 5 mm in height by 25 lm in width, then the spatial
Generating Spectral Images with a Wavelength
resolution at the sample will be 25 lm divided by 403 5
Dispersive Imaging Spectrometer
0.6 lm at an object’s FWHM. In theory, a 1003 lens
should give us a spatial resolution of 0.25 lm; however, When we take a digital photograph with a CCD camera,
depending on the wavelength of light, diffraction effects, we acquire a fixed FOV that is characterized pixel-by-pixel
and residual aberrations (from the microscope objective, in the array. If we take an image of the same FOV through a
the spectrometer, and wavefront errors through relay series of WDP through wavelength bandpass filters, we
lenses and filters), the practical FWHM spatial resolution would generate a stack of as many images as there are
will most likely plateau around 0.4 lm. It is also unlikely WDPs. Each pixel will contain a ‘‘spectrum’’ consistent with
that the image of any object will strike only 1 pixel on the the number of WDP in the series. Motion of objects in the

Cytometry Part A DOI 10.1002/cyto.a

 IMAGING SPECTROMETER FUNDAMENTALS FOR RESEARCHERS IN THE BIOSCIENCES 729

FIG. 17. Classification and spectral mapping of the fluorescence spectra of pollen grains (pollen was chosen for the wide range of fluorescence signatures
present). (a) A rectangular section of the FOV is allowed to pass through the entrance slit. (b) Spectra from points in the FOV incident on the entrance slit.
(c) The spectra in the FOV are sorted into classes and inserted into a library. There are three spectra in Set 1, pseudocolor coded blue, yellow, and red; Set 2
has five spectra. (d) Correlated spectra painted onto a grayscale image of the FOV. Note how the three spectra in Set 1 differentiate in the spectral image of
the FOV. Areas in gray do not correlate with any library spectrum.

FOV or a chemical reaction during the time it takes to ac- In a wavelength dispersive system, using a CCD camera
quire the full WDP series will compromise the integrity of as a detector, the process is reversed. A spectrograph sub-
the spectra. A filter-based approach takes a fixed FOV simul- mits all wavelengths present in an object simultaneously
taneously, and wavelength acquisitions sequentially.

FIG. 19. Spectral topographical map of Alexa 555, 569, 594 in a tissue
FIG. 18. Spectral characteristics of Alexa 555, 568, and 594 acquired on section acquired on the PARISS spectral imaging system. A pseudocolor
the PARISS spectral imaging system as observed within a tissue section was assigned to each of the dyes. Correlated spectra were ‘‘painted’’ onto
shown in Figure 19. a grayscale image of the tissue section with pixel perfect accuracy.

Cytometry Part A DOI 10.1002/cyto.a

730 LERNER

camera (sometimes after binning). Figure 17c shows clas-

sified spectra from the FOV that have been inserted into a
spectral library, each with a user assigned, unique color-
code. In Figure 17d, all newly acquired spectra that corre-
late to library spectra, at a user defined threshold, are
painted onto a grayscale image of the FOV. The colors are
the same as those that the user assigned in the spectral
library. Because the FOV is mechanically translated, there
is no limit to the extent of the field that can be spectrally
mapped.
Figure 18 shows the spectral characteristics of Alexa
555, 568, and 594 that was used to fluorescently label a
tissue sample. The dyes were pseudocolored red in areas
corresponding to Alexa 555, yellow to Alexa 568, and
green to Alexa 594. Figure 19 shows the spectral topo-
graphical distribution of the Alexa fluorophores acquired
with the PARISS system. All wavelengths were acquired
simultaneously over the wavelength range from 540 to
720 nm. Whenever an acquired spectrum correlated with
that of one of the Alexa dyes, its assigned pseudocolor
was painted with pixel perfect accuracy onto a gray scale
image of the tissue section.

WAVELENGTH CALIBRATION OF
SPECTROMETER SYSTEMS
Wavelength accuracy is best determined with a low-
pressure Hg1/Ar1 discharge lamp that covers the wave-
length range from 365 to 850 nm. The emission spectrum
FIG. 20. Spectra of two calibration sources: (a) pure Hg/Ar low-pres- of this lamp from 400 to 842 nm is shown in Figure 20a.
sure discharge lamp; (b) LightForm multi-ion discharge lamp (MIDL). Both
spectra are as presented and digitized by the PARISS spectrometer.
(‘‘Spectroline,’’ Spectronics Corp., Westbury NY, and Oriel
Corp., Stratford CT supply a wide variety of wavelength
calibration lamps.) The main drawback of pure Hg1/Ar1
to the detector. However, the entrance slit in this case acts lamps is that they emit deep UV light that is dangerous to
as a field stop by limiting the area of the FOV passing exposed skin and can cause blindness to unprotected
through the spectrometer. To generate an image, the FOV eyes. Consequently, we use an eye-safe, multi-ion dis-
must be translated sequentially. Remote earth-sensing ima- charge lamp (MIDL) distributed by LightForm, Inc. (Hills-
ging spectrometers mounted in a satellite or aircraft create borough, NJ). The MIDL presents monochromatic emis-
spectral topographical maps by ‘‘flying over’’ an unlimited sion features emitted by Hg1, Ar1 as well as narrow, but
FOV. not monochromatic, inorganic fluorophores to cover the
It is the same principle in a microscope system. Here, spectrum from 400 to 840 nm, as shown in Figure 20b.
the sample is translated in the x-direction in increments The lamp actually emits down to 365 nm, but the 365 nm
consistent with the slit dimensions. If the slit is 25 lm
wide and 5 mm in height, and objects are viewed through Table 3
a 403 objective, then the area of the FOV submitted to Peak Maxima for Principal Spectral Features in the
MIDL Calibration Light Source
the detector is 0.625 3 125 lm2 (25/40, 5,000/40) and
the sample would be translated in 0.625 lm increments. Wavelength Emission
To create an image, spectra collected from the FOV are A 404.7 Hg
categorized into ‘‘classes’’ where each class correlates B 435.8 Hg
with an object, condition, or interaction. A library of the C 546.0 Hg
classes of spectra stores pseudocolor codes for each class. D 577/579 Hg
E 696.5 Ar
Each time a spectrum from the FOV correlates with a par- F 763.5 Ar
ticular class of spectrum, to a user defined threshold, it is G 811.5 Ar
‘‘painted’’ onto the FOV in its associated pseudocolor. Fig- H 842.0 Ar
ure 17 is a general schematic of the process. Figure 17a F1-1 485.0 Fluorophore
shows that only a slice of the FOV is acquired in a single F1-2 544.0 Fluorophore
F1-3 586.0 Fluorophore
acquisition. The FOV is mechanically translated across the F1-4 611.5 Fluorophore
projection of the entrance slit in the FOV. In Figure 17b, V1 605.0 Valley
each acquisition records all wavelengths simultaneously BK1 525 Background
with as many spectra as there are rows of pixels on the BK2 642 Background

Cytometry Part A DOI 10.1002/cyto.a

 IMAGING SPECTROMETER FUNDAMENTALS FOR RESEARCHERS IN THE BIOSCIENCES 731
Hg line is weak in this lamp (for eye safety), and is not
shown here.
The MIDL lamp is battery operated and emits light from
a 6-mm diameter ‘‘pencil’’ tube that replaces the sample
on the microscope stage. Table 3 shows selected Hg1,
Ar1, and fluorophore emission features corresponding to
the letters in Figures 20a and 20b.
The benefit of an MIDL spectrum is that it emits a spec-
tral fingerprint that can be used to calibrate the perform-
ance of any spectroscopic system. (A full listing of wave-
length emission features can be obtained from the
National Institute of Standards and Technology (NIST)
(25).) Most importantly, however, an MIDL can be used to
compare the performance of any instrument with any
other and validate the accuracy of wavelength data (26).

PRACTICAL APPLICATIONS OF THE SPECTROMETER

MATH TO PREDICT THE AREA OF A LASER ‘‘SPOT’’ FIG. 21. Intensity versus pinhole diameter for various samples. Note
AT THE SAMPLE that the MIDL curve closely follows the curve for the area of the pinhole,
Spot Size and Pinhole Optimization as predicted by Eq. (9). The plots for the mirror form a plateau after twice
the size of the Airy disc using the 10 and 633 objectives, indicating that
Probably one of the most discussed issues in confocal the Airy disc determines the illuminated area of the pinhole. In the case
of magic marker and thick fluorescent plastic, as the pinhole is opened
microscopy are the questions, ‘‘what is the true spot size signal continues to increase well past the size of the Airy disc and never
at the sample?’’ and, ‘‘what can I expect when the pinhole forms a plateau. This proves that, with these two sample types, the size of
the Airy disc does not correspond to the illuminated area of the sample. It
is opened?’’ The literature informs us in great detail how also indicates that the bandpass will vary with the size of the pinhole.
to calculate the size of an Airy disc produced by the laser Leica 103, 0.4 NA objective was used unless otherwise indicated.
at the sample, and transmit it through the confocal optics
until it passes through the pinhole. All good theory, but it
pinhole. A spectrum is acquired in 5-nm steps or wave-
is possible to test what happens in reality for any given
length sampling increments (WSI).
sample.
Figure 21 shows a series of plots in which the curve
Looking at the problem from a geometrical optics point
with diamonds is simply a plot of the normalized area of
of view, the amount of light passing to the detector must
the pinhole. If we have a perfectly homogenous extended
obey the Etendue Eq. (9). If the pinhole is perfectly
light source evenly illuminating the pinhole, then we
matched in size to the projected image of the Airy disc at
would expect that transmission curve to exactly follow
the sample, then all available light should pass through
the area of the pinhole as predicted by the Etendue
the system to the detector. If there is no scattering, signal
Eq. (9). If the sample is truly a small ‘‘point,’’ then we
will increase for as long as the image of the Airy disc is
would expect the curve to rise until the pinhole is larger
greater than the size of the pinhole, and then form a pla-
than the image of the spot and then plateau.
teau when the pinhole is larger than the image of the Airy
We tested the principle first with an example where
disc.
there is no Airy disc and the sample can be considered to
If the spot size at the sample is large due to scatter, or
be an ‘‘infinitely’’ scattering extended source. We chose
other less obvious reasons, then we can expect an
the emitting area of an MIDL fluorescent calibration lamp
increase in the size of the projected image of the spot and
for this purpose because it was certain to fully and homo-
it will overfill the pinhole. If the pinhole is homogenously
genously illuminate a pinhole opened to its maximum
overfilled, then we can expect an increase in signal that
extent.
will vary as the square of the diameter of the pinhole, until
All the following acquisitions were made using a 103
finally the pinhole is larger than the spot size, at which
objective. Using a fluorescent MIDL calibration lamp
point it should again form a plateau.
(monitored at one wavelength) as an extended source, we
We tested this experimentally on a Leica SP1 spectral
indeed observe that as the pinhole is opened the curve
confocal microscope (LCSI).
closely follows the area of the pinhole as expected and as
shown in red circles.
Estimation of Illuminated Area of
To compare this with a rigorously nonscattering sam-
a Laser-Excited Sample
ple, we imaged the laser on a front surface mirror and
The LCSI system uses a prism as the WDE and the con- incrementally opened the pinhole. Again we observed
focal pinhole forms the entrance aperture to the spec- that, as expected, the slope of the light throughput curve
trometer. A spectrum is acquired wavelength-by-wave- rose rapidly and then formed a plateau. We tried this with
length by sequentially translating an exit slit assembly both a 103 (red squares with an ‘‘X’’ within the box) and
along the focal field. It seems evident from the results that a 633 objective (blue circles) with essentially the same
the width of the exit slit tracks the nominal width of the result.

Cytometry Part A DOI 10.1002/cyto.a

732 LERNER

We then selected samples that would be more likely to Table 4

scatter light. The first of these samples was a produced by Pinhole Diameter Versus Airy Pinhole Number (Leica SPI) for
a 103, Plan Apo, NA 5 0.4
drawing a very thin line of magic marker on a glass slide
and exciting it with a 488 nm laser, focused to maximize Pinhole diameter (lm) Pinhole number
signal with the smallest pinhole size. We incrementally 80 1
increased the pinhole diameter and observed that the 156 2
throughput curve, shown with triangles, did not follow ei- 241 3
ther the pure reflector model, suggested by the front sur- 319 4
397 5
face mirror, or the extended source model suggested by 475 6
the calibration lamp. However, it was much closer to the 558 7
extended source model than we expected.
We repeated the experiment with a thick fluorescent
plastic slide and found that the curves for both the magic To determine the agreement between the theoretical
marker (triangles) and the plastic slide (squares) were the predictions of the generalized bandpass Eq. (11) and the
same up to a pinhole size of 150 lm (corresponding to observed spectral bandpass as the pinhole diameter
two Airy discs) and then as the pinhole increases in size, increases, we ran a wavelength scan using the MIDL as a
the slopes of the two curves begin to diverge, with the light source between 520 and 580 nm to capture the 544/
thick plastic slide tending more toward the curve plotting 546 Eu/Hg lines on the Leica SP1 system. We varied the
the area of the pinhole than that of the magic marker. pinhole diameters as shown in Table 4. Figure 22 shows
Neither the plastic slide nor the magic marker followed an overlay of the spectral scans for pinholes 1, 4, and 7
the intensity curve as a function of the area of the pinhole, corresponding to pinhole diameters of 80, 319, and
indicating that the pinhole is nonuniformly filled and the 558 lm.
area of excitation at the sample is neither a ‘‘point’’ nor a
homogenously extended illuminated area (such as an Comparison of Observed Versus Theoretical
MIDL). FWHM with Increasing Pinhole Size
The fact that it appears that even with a supposedly
nonscattering sample, such as the magic marker, the For a real life experiment with an emitting source with
apparent illumination of the pinhole is greater than the finite bandwidth we can use the generalized bandpass Eq.
pinhole size. Consequently, increasing the size of the pin- (11) to determine the theoretical bandpass BPnet. First,
hole will result in an increase in light intensity according however, we need values for the five terms Wexap, BPnat,
to Eq. (9) and a decrease in spectral bandpass according BPres, BPslit, and wavelength dispersion (Disp).
to Eq. (10). Wexap: We take the actual size of the pinhole with the
initial assumption that the image of the entrance pinhole
Degradation of Bandpass with Increasing is matched to the width of the exit slit. If the width of the
Magnification and NA slit is not matched to the width of the image of the pin-
hole, this will become apparent when we correlate theory
In the case of a laser confocal system, from Eq. (9), we
with observation.
know that the size of a pinhole should be matched to the
size of an Airy disc, and the size of the Airy disc is a func-
tion of the magnification of the lens and the NA. It may be
counter-intuitive, but while the spatial resolution may
increase as a function of magnification and NA, spectral re-
solution decreases with the size of the Airy disc (27–30).

Ropt ¼ A kM=NA ð14Þ

where: Ropt is the pinhole diameter; A, a constant; k,

wavelength; M 5 magnification; NA, numerical aperture
Bandpass reminder: BP 5 linear dispersion 3 the exit
slit width (or the image of the entrance aperture, which-
ever is greater).
To compare the diameter of the Airy disc at 103, NA 5
0.4 lens with a 633, NA 5 1.32 lens we only have to take
the ratio of magnification (M) to NA in Eq. (14). The nom-
inal ratio for the 103 lens is 25, and 48 for the 633 lens
keeping A and k constant. The increase in pinhole diame-
ter will be 48/25 5 1.9. Consequently, from Eq. (10), we
can expect spectral resolution and bandpass to degrade FIG. 22. Pinhole diameter versus bandpass. Using the MIDL lamp as a
light source, the FWHM of the 545 nm spectral feature increases nonli-
by approximately by a factor of two when using the 633 nearly, but predictably, as the pinhole diameter increases. This data were
lens compared to the 103 lens. acquired on a Leica SP1 spectral confocal microscope.

Cytometry Part A DOI 10.1002/cyto.a

 IMAGING SPECTROMETER FUNDAMENTALS FOR RESEARCHERS IN THE BIOSCIENCES 733

FIG. 23. Observed and theoretical plots

of bandpass versus pinhole size at 545 nm
(the midway point between the 544 nm
Eu fluorescent maximum and the 546 Hg
line). Note that the agreement is excellent.

BPnat: The emission band centered at 545 nm is a com- It is evident that there is very good agreement. This con-
posite of the Hg 546 nm line (which is essentially mono- firms that expectations predicted by the bandpass and
chromatic) and an inorganic fluorophore peaking at Etendue equations are powerful tools in predicting instru-
545 nm; the composite FWHM is 4.75 nm (measured from mental performance.
a high-resolution spectrum of the MIDL lamp on the pre-
viously mentioned Mechelle, spectrograph).
BPres: Using basic optical principles, for a flint glass CONCLUSIONS AND SUMMARY
prism spectrometer, and focusing optics consistent with We have seen that the design and implementation of a
the size of the Leica spectrometer box, we expect a limit- spectrometer has a profound effect on its ability to deter-
ing resolution between 4 and 7 nm. We iterated the terms mine either the spectral characteristics or the location of
BPres and BPslit in the generalized bandpass equation to an object in a FOV. Best imaging is achieved when an
determine that the closest fit for BPres was 4.7 nm. object in the FOV is illuminated or excited by a point
Disp: Leica does not supply this information; therefore, source. When accurate spectral imaging is required over a
the dispersion value was estimated by iterating until the large area of the FOV, the spectrometer must be capable of
theoretical BPnet corresponded to the observed BPnet. The meeting the nontrivial challenge of point-to-point imaging.
calculated value that produced the closest fit was 25 nm/ Understanding the equations that govern the bandpass
mm at 545 nm. and wavelength dispersion of light enables a researcher to
BPslit: This is calculated from Eq. (10) by multiplying optimize light throughput, and understand the true emis-
the deduced dispersion by the pinhole size (Wexap). By sion characteristics of an object. We also illustrated how
the operating parameters of a spectral confocal system can
using 4.7 nm for BPres and 25 nm/mm for the wavelength
be estimated starting with some simple observations, and
dispersion at 545 nm, we obtain an almost perfect fit with
how to determine the true illuminated area of an object in
the observed FWHM values, as shown in Figure 23. Band-
the FOV. With a monochromatic line emission source such
pass is seen to degrade more or less linearly with pinhole
as an Hg or MIDL lamp, we can measure wavelength accu-
diameter. The excellence of the fit with the theoretical
racy and residual instrumental aberrations, as well as the
FWHM also indicates that either the width of the exit slit
relationship between aperture size and spectral resolution.
is always matched to the size of the pinhole or that the
image of the pinhole is always greater than the effective
width of the slit. We can expect to see the same degrada- ACKNOWLEDGMENTS
tion in bandpass whenever the pinhole size is increased Thanks to Robert Zucker of the USEPA for running the
on this instrument. Any system using detector elements Leica SP1 experiments. The copyright to the spectra
that are always larger then the image of the pinhole will shown in Figure 18 are owned by Molecular Probes Inc.,
show no change in bandpass as a function of pinhole di- probes.introgen.com. Thanks to Dr. Michael Donovan at
ameter. This is because bandpass is determined by the Aureon Corporation for permission to reproduce the
larger of either the image of the pinhole of the exit width Alexa stained tissue section in Figure 19, and David Cook
aperture width. When using an IPMT, the exit aperture is of Spectrum Scientific Inc., for his suggestions concerning
the width of a single detector element [Eq. (10)]. volume transmission diffraction gratings.

Cytometry Part A DOI 10.1002/cyto.a

734 LERNER

LITERATURE CITED 17. Reader J. Optimizing Czerny-Turner spectrographs: A comparison between

analytical theory and ray tracing. J Opt Soc Am 1969; 59: 1189–1196.
1. Dickinson ME, Bearman G, Tille S, Lansford R, Fraser SE. Multi-spectral
18. Beutler HG. The theory of the concave grating. J Opt Soc Am 1945;
imaging and linear unmixing add a whole new dimension to laser scan-
35:311–350.
ning fluorescence microscopy. BioTechniques 2001;31:1272–1278.
19. Chrisp M. Aberration corrected holographic gratings and their mount-
2. Garini Y, Katzir N, Cabib D, Buckwald RA, Soenksen DG, Malik Z.
ings. In: Wyant J, Shannon RR, editors. Applied Optics and Optical En-
Spectral bio-imaging. In: Wang XF , Herman B, editors. Fluorescence
gineering, Vol. X. SPIE; 1987. p 391–454.
Imaging Spectroscopy and Microscopy. New York: Wiley; 1996. p 87–
20. Pieuchard G, Flamand J, Passereau G. Monochromator with concave
124.
grating. US Pat. 3,930,728, 1976.
3. Hutley MC. Diffraction Gratings. London: Academic Press; 1982.
21. Balaban RS, Kurtz I, Cascio HE, Smith PD. Microscopic spectral ima-
4. James JF, Sternberg RS. The Design of Optical Spectrometers. London:
ging using a video camera. J Microsc 1986;141:31–34.
Chapman & Hall; 1969.
22. Benedetti PA, Evangelista V, Guidarini D, Vestri S. Confocal-line mi-
5. Lerner JM, Thevenon A. The Optics of Spectroscopy: A Tutorial. Metu-
croscopy. J Microsc 1992;165:119–129.
chen NJ: Jobin Yvon/JY Optical Systems; 1998.
23. Warren DW, Hackwell JA. Compact prism spectrograph suitable for
6. Loewen EG, Popov E. Diffraction Gratings and Applications. New
broadband spectral surveys with array detectors. US Pat. 5,127,728,
York: Marcel Dekker; 1997.
1992.
7. Stroke GW. Diffraction gratings (in English). In: Flugge S, editor. Hand-
24. Warren DW, Hackwell JA, Gutierrez DJ. Compact prism spectrographs
buch der Physik XXIX. Berlin: Springer; 1967. based on aplanatic principles. Opt Eng 1997;36:1174–1182.
8. Lerner JM. Ruled and holographic gratings: A historical perspective. 25. National Institute of Standards and Technology (NIST). Atomic Spec-
Proc SPIE 1995:2532;2–14. tra Database: Lines Data. NIST, 2000.
9. Lerner JM, Chambers RJ, Passereau G. Flat-field imaging spectroscopy 26. Lerner JM, Zucker RM. Calibration and validation of confocal spectral
using aberration corrected holographic gratings. Proc SPIE; 1981:268; imaging systems. Cytometry A 2004;62:8–34.
122–128. 27. Centonze V, Pawley J. Tutorial on practical confocal microscopy and
10. Flamand J, Labeyrie A, Pieuchard G. Diffraction gratings. US Pat. use of the confocal test specimen. In: Pawley J, editor. Handbook of Bi-
3,628,849, 1971. ological Confocal Microscopy, 2nd edition. New York: Plenum; 1995.
11. Flamand J, Bonnemason F, Thevenon A, Lerner JM. The blazing of p 559–567.
holographic gratings using ion-etching. In: Proc SPIE, Adar F, Griffiths 28. Centonze V, Pawley J. Practical laser scanning confocal light micros-
JE, Lerner JM, editors. 1989;1055. copy: Obtaining optimum performance from your instrument. In: Celis
12. Fery C. Comptes Rondue 1910;150:216. J, editor. Cell Biology, 2nd edition. New York: Academic Press; 1998.
13. Hecht E. Optics. Reading, Massachusetts: Addison-Wesley; 1979. Vol. 3, p 149–169.
14. Loewen EG, Popov E. Instrumental Systems. Diffraction Gratings and 29. Pawley J. Fundamental limits in confocal microscopy. In: Pawley J, edi-
Applications. New York: Marcel Dekker; 1997. p 437–480. tor. Handbook of Biological Confocal Microscopy. New York: Plenum;
15. Shapiro H. Practical Flow Cytometry. New York: Wiley-Liss; 1995. 1995. p 19–36.
p 106–135. 30. Wilson T. The role of the pinhole in confocal imaging systems. In:
16. Murty MVRK. Compensation for coma and anamorphic effect in dou- Pawley J, editor. Handbook of Biological Confocal Microscopy, 2nd
ble monochromators. Appl Opt 1972;11:1637–1638. edition. New York: Plenum; 1995. p 167–182.

Cytometry Part A DOI 10.1002/cyto.a