Professional Documents
Culture Documents
should be taken into account in a separate calibra- performed in ;7 s, when performed on a Pentium PC
tion. with sufficient memory.
The grating molds are made from a stack of SiO2 on Equation ~1! is not well suited to establish the
Si covered by a resist layer and exposed to an electron shear ratio s because it is difficult to determine R with
beam. On development, the resist is removed from satisfying accuracy. A simpler and more accurate
the unexposed areas and subsequently the grating is way is to examine the interference pattern, as re-
dry etched in the SiO2 layer. The grating depth corded by the CCD camera, such as the one shown in
achieved in this way is found to be within 30 nm of Fig. 3, and determine the radius r of each of the two
the target. beams in pixels and the displacement q in pixels.
A piezoelectric translator performs the translation The shear ratio is then given by
of one of the two gratings mounted on a leaf-spring
construction. This piezostack can expand 20 mm,
which is sufficient for the 12.5-mm translation re- q
s5 . (2)
quired for a phase step of 2p. Through calibration, 2r
the nonlinear response of the translator to the ap-
plied voltage is compensated for in software such that In this way, the shear ratio in Fig. 3 was found to be
deviations from linearity remain below 0.05 mm over 0.113 6 0.002.
the required 12.5-mm range. The stationary grating To test the absolute accuracy and repeatability of
can be aligned with respect to the movable grating by the phase-stepped LS interferometer, the system was
adjustment of the small rotation table. The distance compared with a phase-stepped T-G interferometer.
between the gratings can be adjusted between 5 and To this end, the modified T-G arrangement of Fig. 4
35 mm by the use of spacers of variable length. In was built. The reference mirror is piezodriven for
this way the shear ratio can be varied. The free phase stepping. A beam splitter in the measure-
aperture is 10 mm. This whole assembly is fitted ment arm directs part of the beam under consider-
into a cylinder approximately 100 mm long and 50 ation into the LS interferometer. The influence of
mm in diameter. This in turn is mounted on a ro- the beam splitter was checked by the measurement of
tator that can switch between two perpendicular po- one wave front both after reflection and after trans-
sitions for horizontal and vertical shear. The mission by the beam splitter. It was thus found that
inaccuracy of the rotation angle is of the order of 1°. the beam splitter does not contributes measurably to
A computer-controlled rotation over 90° takes ;3 s in the wave-front aberrations. If the shutter is open,
our prototype. the T-G interference patterns can be recorded by the
The LS interference pattern is imaged on a CCD CCD camera in the lower left of the figure. If the
camera. A PC controls the setup and performs the shutter is closed, the LS interference patterns are
data analysis of the interferometer. It comprises an recorded on the CCD camera at the top of the figure.
analog-to-digitalydigital-to-analog card controlling This interferometric setup is built virtually vibration
the rotator and the calibrated high-voltage supply for free on an air-pressure-stabilized optical table. This
the piezotranslator and a framegrabber card to col- ensures optimal conditions for T-G interferometry.
lect the phase-stepped CCD images. A complete The reference front in the T-G interferometer was
measurement plus reconstruction of the wave front is measured separately with the LS interferometer. It
Fig. 4. Experimental setup designed to measure a test sample in The fractional errors Dr and Dq are both 0.5. Typ-
both a T-G interferometer and a shearing interferometer. If the
ical values are r 5 150 pixels and q 5 75 pixels,
shutter is opened, the arrangement functions as a T-G interferom-
eter, and if closed, the beam wave front can be analyzed with the
giving a relative inaccuracy of less than 1%. This
LS interferometer. gives an absolute error of approximately Ds 5 0.002
at a shear ratio of 0.25. Ds 5 0.002 results in a
relative wave-front error of ;0.5% at shear ratios
was found to be almost perfectly flat, i.e., all aberra- between 0.15 and 0.5, as is evident from Fig. 6. In
tion terms were less than 3 ml rms ~1 ml 5 ly1000!. first order this relationship is linear, i.e., an increase
In Subsection 5.B the displayed T-G results have in the shear ratio error results in a similar increase of
been corrected for the measured reference-front data. the wave-front error.
1. Estimation of s
The accuracy in the determination of shear ratio s
@see Eq. ~2!# is in practice limited only by the error
that is due to the use of a limited amount of pixels.
In principle, errors in the aspect ratio of the imaging
system ~apparent deviation from squareness of the
CCD pixels! also affect the accuracy, but we find that
this ratio can be determined with sufficient precision,
such that the estimation of the shear ratio is not
affected. The relative error in the shear ratio s
therefore is
Ds
s
5 FS D S D G
Dr
r
2
1
Dq
q
2 1y2
. (3)
Fig. 6. Error propagation in the matrix-inversion method: rel-
ative errors in the calculated Zernike coefficients anm as a results
of a shear ratio error Ds 5 0.002.
ff
v5 , (6)
MR
where f and M are the focal length and the magnifi-
cation of the camera lens, respectively. The appar-
ent shear direction is tilted over g 5 arctan~vys! '
vys, and the evaluated amount of a1, 11 is
aberration term. Except for the third sample, the is estimated with less accuracy; e.g., an error of 0.02
agreement is very good: the deviation is never more in the determination of s results in a 5% error in the
than a few milliwaves. Note, furthermore, that the wave-front reconstruction, as discussed in Section 4,
s values are generally lower in the LS measurements which explains the deviations observed in Table 2.
than in the T-G measurements. s values of only a The accuracy of LS interferometry follows from the
few tenths of a milliwave indicate that the method is comparison with the results of the T-G measure-
stable, as is expected from its common-path nature. ments. Here the T-G interferometer serves as a ref-
The third sample has been included to illustrate erence that is assumed to be fully proven and reliable.
that a good determination of the shear ratio is crucial In practice, the limit of absolute accuracy of T-G in-
to obtain high accuracy. After a severe heat treat- terferometers in analyzing wave fronts is of the order
ment of this plastic lens, the pupil has lost its per- of 3–5 ml rms. This is of the order of the discrep-
fectly round shape, and consequently the shear ratio ancies found in Table 2. We therefore conclude that
the accuracy of the LS interferometer is at least of
comparable magnitude.
In another experiment, the LS interferometer was
set up on an ordinary tabletop, with no stabilization
or dampening facilities. We did not find any effect
on the standard deviations over 10 measurements.
Even if moderate vibration sources, such as a venti-
lator or a transportable radio playing at loud volume,
were operated on the same tabletop, the measured
values were not significantly affected and the s val-
ues remained below 1 ml per aberration and below 2
ml for the overall wave front. This evidences the
robustness and stability against vibrations of this
wave-front analysis method.
aberrated wave front with the LS interferometer by of Fig. 6 have little significance here. The measured
using different shear ratios. In each series the mea- effects for these three aberration terms are in good
surement was repeated 10 times to determine the agreement with the calculations of Fig. 5.
variation s. The sample was also measured in the
T-G interferometer. C. Alignment Errors
Figure 13 shows the values of a number of aberra- Grating misalignment leads to erroneous results,
tions terms present in the wave front as measured most notably in the a4, 24, a6, 24, . . . , terms, as dis-
with the LS interferometer ~data points! compared cussed in Section 4.
with the T-G results ~dashed lines: 3s limits!. The In a series of measurements a nearly flat wave
error bars on the data points indicate 3s ranges. front originating from a laser diode was analyzed
Note that these limits do not refer to absolute accu- with the LS interferometer, with a shear ratio of s 5
racies. 0.35, while the angle a was varied between the grat-
The measured values are in good agreement with ing orientations between 0° and 1° ~17.5 mrad!. The
the T-G result for 0.2 , s , 0.4, and we conclude that alignment angle was set with an accuracy of ;29.
the best shear ratios for performing LS interferome- Figure 14 summarizes the measured results. The
try are in this range. appearance of large an, 24 terms after misalignment
We compare the experimental results of Fig. 13 is obvious. Near the 0-mrad position the relation
with the error propagation properties of the matrix- between the angle and the magnitude of these terms
inversion method as summarized in Figs. 5 and 6. is linear, as expected from Section 4. We also see the
The standard deviation in the measured a22 is largest expected parabolic deviation of the a20 term. Fur-
at the low end of the range of shear ratios, which is in thermore, we note that for larger misalignment an-
agreement with Figs. 5 and 6. The observed devia- gles the response of the an, 24 also deviates from
tion of measured coma a31 from the T-G result at both linearity. For a 5 1°, the apparent shear ratio is s 5
the low and the high ends of the range is also in 0.38 ~see Fig. 7!. From Table 1 we see that this
agreement with the expectations from the error prop- increase in shear ratio results in the calculation of
agation properties. The measured values a40 and larger an, 24 terms. This effect is additional to the
a60 deviate only at the high end of the range, whereas linear response and explains the observed relation-
the agreement for a62 between the two interferom- ship. Finally, we observe that other aberration
eters holds over the whole range. Because these ab- terms are not strongly affected by misalignment.
erration terms are all quite small, the relative errors Figure 14 demonstrates that for a , 3 mrad, all