September 1952 SMALL PLANE GRATING MONOCHROMATOR 647

long were constructed which were not sufficiently cir- quality, and simplicity of construction and adjustment.
cular that their centers and ends were simultaneously It has a resolution in the visible and ultraviolet regions
positioned at the same wavelength. However, with of about 0.05 A.
these slits the resolving power recorded for any short
portion of the slit was the same as for any other short ACKNOWLEDGMENT
portion of the slit. A great many people have aided in the development
In the following paper in this issue0 further tests of this equipment. Mr. David Clements, Leeds and
and calculations are presented which give a quantitative
Northrup Research Design Section, is responsible for
picture of the potentialities and limitations of this form the mechanical design work. Messrs. Elmer Luft and
of monochromator. George Wells, Leeds and Northrup Production Machine
CONCLUSION Shop, did most of the mechanical construction. Messrs.
The new monochromator appears to be an improve- George C. Hill and N. E. Polster, of this laboratory,
ment in several respects over other types of spectro- aided in the development of the spectrometer and in
metric equipment, particularly from the standpoint testing it. Dr. H. M. Crosswhite, Physics Department,
of ruggedness, portability and compactness, optical The Johns Hopkins University, made many helpful
10 William G. Fastie, J. Opt. Soc. Am. 42, 647 (1952). suggestions.

JOURNAL OF THE OPTICAL SOCIETY OF AMERICA VOLUME 42, NUMBER 9 SEPTEMBER, 1952

Image Forming Properties of the Ebert Monochromator

WILLIAM G. FASTIE
The Johns Hopkins University,Baltimore,Maryland
(Received May 8, 1952)

The previous paper (Wm. G. Fastie J. Opt. Soc. Am. 42, 641 (1952)) described a plane grating monochro-
mator optical system which apparently has not been used since Ebert described it in 1889. The high resolu-
tion obtained has inspired the design of anf/6 spectrometer of that type for the infrared region. This paper
describes the results of preliminary tests and calculations which show that the Ebert system is capable of
near theoretical resolution at f/6 even when very long slits are used. Both slits must be curved in order to
remove the effect of astigmatism and to avoid wavelength errors at the exit slit which arise from curvature
of spectrum when a plane grating is employed. That the same curvature of slits simultaneously corrects
these two image errors is a fortunate coincidence which makes the Ebert system appear to be superior to
currently employed spectrometer optical systems. Comparison with other systems and discussion of the
limitations of the Ebert system are presented.

IN the preceding paper' experimental evidence was have been made with a 15-in. circular flat, in place of
presented which indicated that an f/10 monochro- the grating. The test chart described in the previous
mator of the type described by Ebert2 forms aberration- paper' was used to check the resolving power. The re-
free images of long slits if the slits are properly curved. solving power in the visible at f/6 was better than
An f/6 scanning infrared spectrometer of the Ebert 50 percent of theoretical. For infrared work, the defini-
type is under construction in this laboratory. The tion, therefore, would be substantially perfect. The
present paper describes some preliminary tests and image showed considerable astigmatism.
calculations which confirm the previous results and
which show that the Ebert system is superior in the DEFINITION OF I NUMBER
near infrared and at shorter wavelengths to other sys- For the purpose of comparison between various
tems when ultimate resolution is desired and when long systems, the term I number is introduced, it being
slits are used.
QUALITY OF THE IMAGE AT F/6 PLANEGRATNG CONCAVE MIRROR 1
The spectrometer under construction employs a
90-in. focal length mirror of 30-in. diameter and a 16-in.
X 10-in. grating as shown in Fig. 1. Preliminary tests
90 IN.
' Wm. G. Fastie, J. Opt. Soc. Am. 42, 641 (1952).
2 H. Ebert, Wied. Ann. 38, 489 (1889). FIG. 1. Optical diagram of f/6 spectrometer.
648 WILLIAM G. FASTIE Vol. 42

cally positioned with respect to the center line. The

length dL of the astigmatic images will depend only on
the value of dF and thef number. When the object and
image are at the edges of the optical flat of width W,
and since W is small compared to F,
2 2
dL=dF/f=0.4Wp =0.IW/f . (1)

The value of dL is about 0.1 cm for the system shown

in Fig. 1.

IMAGE ERRORS RESULTING FROM ASTIGMATISM

WITH STRAIGHT SLITS
In the evaluation of the effect of aberration error,
the width of the imperfect image must be compared
FIG. 2. Geometrical relationship between exit slit and line image with the half-width of the diffraction pattern. The
of a point at the end of the entrance slit.
half-width of the diffraction image (Wd) is given by the
formula for a rectangular aperture:
defined as the ratio of the length of the slit to the focal
length of the mirror, (I=L/F). The number has the Wd= FX/W=fX. (2)
significance that, for a given grating, it is the only
factor which determines the amount of energy delivered Figure 2 shows the position of the line image dL of
through the exit slit per unit spectral band. The energy a point a at the end of the entrance slits with relation
E in a spectral band in the image of the slit is propor- to the exit slit. The slits have a length L and are
tional to the area of the slit, the area A of the grating, separated by a distance of approximately W. Since W
and is inversely proportional to the square of the focal is much larger than L, the projected width Wa of the
length, F, thus: astigmatic image dL in the direction perpendicular to
E-LdA/F, the slits is Wa=LdL/W. Accepting a value of wa equal
to the half-width Wd of the diffraction image as an
where d= slit width, L= slit length, and A = projected allowed upper limit,
area of grating.
For a fixed spectral band d-F, then wa LdL/W=fX= Wd. (3)
E-LAIF-AI. Substituting from Eq. (1) and solving for L,
Thus, the I number is the critical number in evalua- L= 10Xf3 .
ting the energy efficiency of a spectrometer employing
a dispersing element of a given projected area. It will Atf/6 and X =3 the allowed value of L is 0.65 cm,
be shown later that the Ebert monochromator can be and the allowed value of I is 0.003. Even if the resolving
used with large I numbers without loss of resolution power desired were only one-third of theoretical,
but that it cannot form perfect spectral images when the straight slits no longer than 2 cm may be used with this
f number is small. The above discussion is introduced 3
monochromator for the A region.
here to emphasize the fact that a large f number does
not constitute an energy limitation.
MEASUREMENT OF ASTIGMATISM
It was necessary to know the astigmatism of the
system in order to calculate the limitations of curved
and straight slits. The astigmatism was measured by
placing a vertical and horizontal crosshair in the posi-
tion of the entrance slit Si (Fig. 1) and observing it
in the image plane at S2, with a microscope. The dif-
ference in position dF between the horizontal and verti-
cal foci was found to fit the empirical formula

dF=0.4Fso2 ,
where F is the focal length of the concave mirror and p
is the half-angle subtended at the center of the concave FIG. 3. Geometrical relationship between line image dL of
mirror by the object and image, which are symmetri- point a on circular exit slit.
September 1952 EBERT MONOCHROMATOR 649
CORRECTION OF ASTIGMATISM ERROR WITH N
CURVED SLITS
The above results show that the slit length is seriously
limited in an Ebert system if straight slits are used. If
circular slits are used with their center of curvature at
the center C of the system (Fig. 3), any point on the
entrance slit will be formed into a line image on the D
exit slit perpendicular to and with its center on the line
Ca. The maximum deviation of the line dL from the
exit slit will be half the sagitta of the chord dL. Thus,
for curved slits:
wa,=dL2 /8W.
For the system of Fig. 1, wa=0.3 microns which is
negligible compared to the value of Wd= 18 microns
obtained from Eq. (2) when =3 and f=6. Thus, if
the slits are properly curved, astigmatism in the Ebert
system does not affect the quality of the image no
matter how long the slits are. FIG. 4. Skew ray diagram to describe interference relationship
Except at the center of the curved slit, the direction between adjacent facets.
of the spectrum is not perpendicular to the slit, re-
quiring that the slit opening be crescent-shaped in order therefore, the perpendicular distance between the
that the spectral width be the same at all points on the facets will be t= a sin 0. Consider oly those diffracted
slit. This condition can be met if the inner and outer rays D and D' which are specularly reflected from the
jaws of the slit have the same curvature and if the slit facets and which, therefore, make the angle sowith the
width is adjusted by having the chord connecting the facet normal. Under these restricted conditions the
ends of the jaw move in the direction of the spectrum. grating formula becomes identical to the thin plate,
The area of the opening of such a slit is the same as that air spaced, interference formula:
of a straight slit having the length of the chord and the
same spectral width. Therefore, the energy which nX= 2t cosso.
crescent-shaped slit openings can pass is the same as Applying the relationship between a and t,
that passed by straight slits.
It will be noted that the use of long curved slits - n= 2a sinG cos y. (4)
does not require a larger concave mirror to avoid vignet-
Since the angle which the facets make with the grat-
ting, the additional mirror area which is required being
ing normal affects only the distribution of energy in
provided in the top and bottom sections of the mirror.
the various orders and does not affect the angular
In the preceding paper it was indicated that machine
shop limitations made construction of curved slits for spectral dispersing properties of a grating, the line N
a small ultraviolet monochromator very difficult. can be redefined as the bisector of the incident and
diffracted ray, and as being perpendicular to the lines
However, for near infrared work, and for the larger
of the grating. With this broader definition of the line
instrument considered in the present paper, the dimen-
N, Eq. (4) is applicable to all positions of a grating
sional tolerances present no problem. Accordingly,
and is not restricted to the center of the blaze.
an I number of as large as 1=0.05 (4.5 inch slits) is
For a given position of the grating and for a given
being considered. The problem of illuminating a long
incident ray there are a multiplicity of diffracted rays.
entrance slit and of concentrating the energy from a
long exit slit onto a small detector has recently been con- and a multiplicity of bisector lines N. However, in a
monochromator only one such bisector must be con-
sidered by Benesch and Strong,3 who describe an image
transformer for the purpose. sidered, and it applies to rays from all points on the
entrance slit and for all positions of the grating; that
CURVATURE OF SPECTRAL LINES is, there is a fixed line, passing through the center of
the grating which besects the central ray of all incident
It is well known that plane gratings produce curved and diffracted beams which reach the exit slit. There-
spectral lines when straight slits are used. This can be fore, Eq. (4) exactly represents the spectral image
understood by the following argument. Consider two forming properties of a plane grating monochromator.
adjacent facets F1 and F2 of an echellette grating with This general statement can be confirmed with respect
spacing a (Fig. 4) and consider any skew incident ray to the Ebert system by reference to the end view shown
I making the angle with the normal N to the facets. in Fig. 2. Consider a ray from the center of the entrance
The line N makes an angle 0 with the grating normal P; slit which travels perpendicular to the paper. This ray,
3 W. Benesch and J. Strong, J. Opt. Soc. Am. 41, 252 (1951). after reflection from the mirror, will strike the center
650 WILLIAM G. FASTIE Vol. 42

.03 correction. This solution is further undesirable in an

Ebert system because the astigmatism correction is not
accomplished.
WAVELENGTH ERROR RESULTING FROM
.02 X
Il STRAIGHT SLITS

-J By use of Eq. (4) and geometrical relationships of

the optical- system (Fig. 2) the wavelength error due
.01 x
25)1
to straight slits can be determined. Equation 4 can be
written (when o is small)
x x
.41
p.611
sinO1-
nX=2a ,
2xIO
W
4 l

4 6
w

8
l

10
l l

12
ff

14 8F2
RESOLVINGPOWER

FIG. 5. Limitation on I value imposed by curvature of the and the wavelength error from the center to the end of
spectrum when straight slits are used. Crosses indicate best the slit is
resolving power obtained in various wavelength regions by five dX= aL2 sinO/4nF 2 ,
different spectroscopists using straight slits.
but a/n sinO X; then
of the grating and will come to the center of the exit dX= XL 2/8F2= X12/8.
slit traveling perpendicular to the paper. Thus, the
center line C of the system coincides with the line N Equate the foregoing to the theoretical value of dX
for these rays. It will be seen that a ray from one end of = X/R, and the largest allowed I value becomes
entrance slit which goes to the center of the grating and I= (8/R)k. (5)
finally to the other end of the exit slit is also bisected,
at the grating, by the center line C. Figure 5 is a plot of Eq. (5) and shows that the I value
It will also be noted by reference to Fig. 2 that the is limited to less than 0.01 in a monochromator with
angle so will vary for different points along straight straight slits if a resolving power of 105 is required.
slits. Therefore, for a given position of the grating This limitation is completely removed if properly
(0 constant) the wavelength emergent from the mono- curved slits are employed.
chromator varies along the exit slit.
If curved slits are used, with the center of curvature WAVELENGTH ERROR RESULTING FROM
CURVATURE OF WAVE FRONT
at the point C (Fig. 3), the angle p is the same for all
points on the slits and there is no wavelength error Although the Ebert system corrects at the second
along the exit slit. There is, thus, a remarkable and reflection for the curvature in the wave front introduced
fortunate coincidence in the simultaneous correction at the first reflection, the grating is between and does
of wavelength error and astigmatism error when curved not receive parallel rays. The angle sovaries therefore
slits are used in an Ebert monochromator. across the grating face, which results in a wavelength
The principle of carving both slits to correct for error at the exit slit. It will be shown that this effect
wavelength error can be applied to any plane grating only limits thef value and does not affect the allowed
monochromator with fixed slits, with the exception of I value.
the Littrow type. If the bisector line N is carried Czerny and Turner4 showed that the wave front,
through the system, the distance from the center of the after the first reflection, is S shaped, the extreme rays
slits to the line N, in the plane of the slits determines the being parallel, the central ray being most deviated from
radius of curvature of the slits. If the focal lengths of
R.
the entrance and exit systems are not the same, the
curvature of the entrance and exit slits will be different,
and will be in the same ratio as the focal lengths. It can
be seen that the radius of curvature of the slits would
be very small in the Littrow type spectrometer because
the slits are very close to the line N. Since the total
length of the slit is limited to a semicircle, the I value
for a Littrow spectrograph would be seriously limited
if both slits were curved. FIG. 6. Geometrical relationship between extreme and central
rays. The extreme rays R and R are parallel after reflecting
Of course, it is possible to correct for the wavelength from the concave mirror. The central ray R2 is not parallel to
error by using a straight entrance slit and a properly R1 and R 3 after reflecting from the concave mirror. The point A
curved exit slit. However, the curvature of the slit must is the center of curvature of the concave mirror.
then be variable with wavelength to provide complete. 4 M. Czerny and A. F. Turner, Zeits. Physik 61, 792 (1930).
September 1952 EBERT MONOCHROMATOR 651

the extreme rays. Figure 6 shows the relationship

between the various rays. The entrance slit SI and the I0
exit slit S2 are separated by the distance W, the width
of the plane mirror. (The argument below is equally
valid if a grating is considered, but the geometry is
less obvious.) The radial ray RI makes the angle ,
5
=tan-'(W/aF)=tan-'(1/2f) with the center line AC
throughout its path to the exit slit. The other extreme
ray R3 also makes the angle spewith the center line
AC throughout its path to the exit slit. The angle e
can, therefore, be used in Eq. (4) to determine X for
t | * e
t 4
the extreme rays. 5x 10 10 5 20 25 30
RESOLVING POWER
The central ray is more complicated, but it can be
seen that whatever path it follows to the point B will be FIG. 7. Relationship between f number of Ebert spectrometer
and resolving power. When the f number is small the resolution
identical to its path from B to the exit slit. The angle it is limited because of curvature of the wave front incident on the
makes with the line AC when incident at B is therefore grating.
the angle to be used in Eq. (4) to determine X. The
parallel ray R 2 does not go quite to the point B but it (solid line) is most closely approached by infrared spec-
is substantially parallel to the central ray and makes the trographs, which is to be expected, because aberration
angle V = 2 sinl-(1/4f) with the center line AC. Making and figure errors become less important when the half-
the substitutions in Eq. (4) for the center and extreme width of the diffraction pattern is large.
rays similar to those made in the last section, Systems using parabolas can only form perfect images
1/R= cos tan- (1/2f) - cos2sin-'(1/4f). (6) of a point, the ends of the slit being off axis and not
being imaged perfectly. The magnitude of the error
The above analysis is only qualitative since the thus introduced by long slits is not known but probably
wavelength error does not vary linearly with the f is the contributing factor to the low values of I which
number and only a small area at the edge of the grating are generally obtained in the visible and ultraviolet
contributes to the extreme wavelength error. However, region with parabolic systems. The ability of the Ebert
Fig. 7, which is a plot of Eq. (6), shows that even the system to produce almost perfect images for very
pessimistic analysis predicts a resolution of 60,000 at long slits is one of its most attractive features.
f/6. For visible or ultraviolet applications, where resolu-
tions of several hundred thousand may be desired, CONCLUSION
large f numbers should be used. It was shown earlier
that the f number does not determine the amount of The important sources of image error in an Ebert
energy delivered to the exit slit. type monochromator have been shown to be negligible
over a wide range of practical conditions. The analysis
COMPARISON WITH OTHER SYSTEMS has been used as a guide in the design of anf/6 infrared
In comparing the Ebert system with other plane spectrometer. The analysis qualitatively confirms the
grating monochromator systems it is necessary to con- experimental results obtained with the f/10 mono-
sider two factors: limitations caused by curvature of chromator described in the preceding paper. It is an-
spectra, and limitations caused by aberrations. A private ticipated that quantitative confirmation with the new
survey of five different spectroscopists covering the f/6 system will be obtained in the near future.
spectral range from 0.4tt to 10u and obtaining resolving
ACKNOWLEDGMENT
powers from 10,000 to 80,000 indicates that no instru-
ment with straight slits employs an I value larger than I wish to thank Professor John Strong of this labor-
the value given by Eq. (5), which is the limitation im- atory for his many helpful suggestions. This work was
posed by curvature of spectral lines. financed by research contract Nonr 248 (01) between
The various values obtained by these investigators the Office of Naval Research and the Institute for
are plotted as crosses in Fig. 4. The theoretical I value Cooperative Research, The Johns Hopkins University.