Home

Search

Collections

Journals

About

Contact us

My IOPscience

Sub-nanometre double shearing heterodyne interferometry for profiling large scale planar surfaces

This article has been downloaded from IOPscience. Please scroll down to see the full text article. 2004 Meas. Sci. Technol. 15 2435 (http://iopscience.iop.org/0957-0233/15/12/013) View the table of contents for this issue, or go to the journal homepage for more

Download details: IP Address: 122.174.20.192 The article was downloaded on 20/08/2011 at 07:34

Please note that terms and conditions apply.

INSTITUTE OF PHYSICS PUBLISHING Meas. Sci. Technol. 15 (2004) 24352443

MEASUREMENT SCIENCE AND TECHNOLOGY PII: S0957-0233(04)82957-6

Sub-nanometre double shearing heterodyne interferometry for proling large scale planar surfaces
T Yokoyama1, S Yokoyama1, K Yoshimori2 and T Araki1
Department of Mechanical Science and Bioengineering, Graduate School of Engineering Science, Osaka University, Japan 2 Department of Computer and Information Science, Faculty of Engineering, Iwate University, Japan
1

Received 4 July 2004, in nal form 14 September 2004 Published 19 November 2004 Online at stacks.iop.org/MST/15/2435
doi:10.1088/0957-0233/15/12/013

Abstract A novel interferometric method, called double shearing interferometry, is presented for proling large-scale quasi-planar surfaces, such as semiconductor wafers, optical ats and x-ray mirrors. The surface prole is measured even if there is an inclination in the scanning stage. By adopting the common-path optical conguration and heterodyne detection, the proposed method achieves excellent resolution at the subnanometre scale and allows a robust measurement in the presence of unwanted disturbances in the measurement environment. A height resolution of 0.1 nm was achieved experimentally. The standard deviation of the analytical surface proles was found to be 1.3 nm, even when using a conventional screw-lead scanning stage. The measured results show that the surface prole and stage inclination are determined separately. We conrmed that the resultant prole was quite consistent with that measured with a Fizeau interferometer.
Keywords: heterodyne interferometer, shearing interferometer, proler,

nanometre measurement

1. Introduction
There are great demands for precision measurement of the surface topography of large-scale industrial products, such as silicon wafers, optical ats, and grazing incidence mirrors for synchrotron radiation. The size of silicon wafers is currently up to 300 mm in diameter, and a grazing incidence mirror can be more than 1 m long. An optical system commonly used to measure the topography of surfaces is the Fizeau interferometer. The Fizeau interferometer provides information on the difference in prole between the surface under test (SUT) and a reference optical at (ROF) by analysing the interference fringe pattern. The Fizeau interferometer can measure the whole SUT simultaneously, but the size of the ROF limits the size of SUT that can be measured. Although various methods of fringe pattern analysis, such as the phase shifting method (sometimes called the fringe scanning method) [1, 2]
0957-0233/04/122435+09$30.00

and the Fourier transform method [3], are employed to increase the measurement resolution, the uncertainty in measurement is mainly restricted by that of the ROF prole. The Fizeau interferometer requires a rigid and stable construction to avoid unwanted effects from mechanical vibrations because the SUT and ROF are placed independently, even if their optical paths are aligned on a common axis. With the aim of measuring large specimens with high precision, an instrument has been developed that is equipped with a ROF whose diameter is 350 mm [4]. The measurement uncertainty was investigated and reported as /77, where is 633 nm [5]. For measurements of larger materials such as the grazing incidence mirror, a long trace proler (LTP) has been developed [610]. This instrument scans a single beam that illuminates the SUT and measures a slope prole based on an angle-to-position conversion feature of the Fourier transform lens. It was reported recently that a 500 mm long mirror was 2435

2004 IOP Publishing Ltd Printed in the UK

T Yokoyama et al

measured with a standard deviation of 4.6 nm in height when the temperature was controlled to within 0.1 K [8]. Shearing interferometry is one method for resolving the issues inherent in the Fizeau interferometer described above, because it does not require a ROF. The shearing interferometer uses two laterally shifted probe beams, and measures the differential of the surface prole from the interference signal of those beams. This type of interferometer can employ the heterodyne detection technique, with which sub-nanometre resolution is easily obtained. Sub-nanometre prolers using heterodyne interferometry have been reported [11, 12], but their span of measurement was less than a few millimetres. To obtain whole surface prole with a shearing interferometer, mechanical scanning of the SUT is necessary and this causes a problem in that any inclination of the scanning stage can hide small variations in the surface prole. In this report, we will present a novel interferometric measurement method named double shearing heterodyne interferometry. The proposed system is composed of two heterodyne shearing interferometers. Here, the SUT is illuminated at three points located equidistantly in a straight line with a pair of shearing interferometers to measure two height differences between adjacent two points. Because any inclination of the scanning stage is regarded as contributing equally to the measured two height differences, this contribution is cancelled out by subtracting one height difference from the other. Then, one can obtain the secondorder differential and retrieve the prole of the SUT. In each shearing interferometer, two probe beams propagate along a nearly common optical path by introducing an anisotropic uniaxial crystal (AUC) as a beam splitter. Because of the common-path conguration, the effects of the mechanical vibration and uctuation in the refractive index caused by air turbulence are effectively reduced. Furthermore, temperature drifts of the two shearing interferometers are also reduced by an appropriate subtraction. Based on such a common-path optical conguration and heterodyne detection, a height resolution of 0.1 nm was achieved experimentally in the usual environment. The height of the prole was measured with a standard deviation of 1.3 nm using the screw-lead scanning stage. The system is conrmed to be robust and capable of highly sensitive measurement without requiring extremely stability.

Figure 1. Schematic of the double shearing interferometer. HWP: half-wave plate, PL: polarizer, BS: beam splitter, PD: PIN photodiode, APD: avalanche photodiode, AUC: anisotropic uniaxial crystal, N: the optical axis of AUC, o and e: the ordinary and extraordinary rays, SUT: surface under test, PM: phase meter.

2. The principle of double shearing heterodyne interferometry

2.1. Optical system A schematic of the optical system is shown in gure 1. The light source of the system is a frequency-stabilized transverse Zeeman HeNe laser ( = 633 nm) [13]. The output power is 2 mW and the beam diameter (e2 of peak intensity) is 0.8 mm. The output light beams are linearly polarized and their polarization directions are orthogonal due to the Zeeman effect [14]. The frequency difference between two orthogonal beams caused by the cavity mirror anisotropy [15] is approximately 190 kHz. After passing through a half-wave plate (HWP), half of each laser beam is reected at a beam splitter (BS) and the 2436

orthogonally polarized elds are combined with a polarizer (PL) and detected with a PIN photodiode (PD). Then, the reference beat signal, whose frequency is 190 kHz, is generated at the PD due to the square law of the detector. The light beams transmitted through the BS are split at the parallel plate into two parallel beams whose spatial separation is equal to the sampling pitch of the measurement, X, and these parallel beams are transmitted through another BS. Each of them is further split with an AUC, into the ordinary ray and the extraordinary ray, which are depicted as o and e, respectively, in gure 1. The dimension of the AUC is designed so that the walk-off distance between the ordinary ray and the extraordinary ray is equal to X. The optical axis of the AUC, N, is set in the plane including two incident parallel beams, so that two pairs of ordinary and extraordinary rays lie in the same plane. The ordinary ray from one input beam and the extraordinary ray from the other input beam are superimposed at the output end of the AUC. Therefore, the output of the AUC consists of parallel three beams, which illuminate three spots on the SUT, Xn1, Xn and Xn+1, located equidistantly in a straight line. The HWP is rotated so that the direction of one of two orthogonal polarizations lies in the plane including N and the incident beams. This realizes the condition that one polarization becomes the ordinary ray and the other polarization becomes the extraordinary ray. Thus, each pair of ordinary and extraordinary rays forms a heterodyne probe of the shearing interferometer whose frequency difference is 190 kHz. The output beams of the AUC are reected at the SUT and each pair of ordinary and extraordinary rays is superimposed again at the front end of the AUC. Then, each superimposed beam contains two frequency components that bear the interferometric information of two adjacent spots on the SUT. In this manner, the optical conguration forms two heterodyne shearing interferometers, which measure

Sub-nanometre double shearing heterodyne interferometry for proling large scale planar surfaces

the height differences between Xn1 and Xn, and between Xn and Xn+1. After reection at the BS, two frequency components of each beam are combined with the polarizer and detected with an avalanche photodiode (APD), producing the electrical beat signal of frequency 190 kHz. The phase differences of these beat signals relative to the reference beat signal are measured with two phasemeters (PMA and PMB) and registered to a computer at every X movement of the scanning stage. 2.2. Prole retrieval The heights of the sample at the positions Xn1, Xn and Xn+1 are denoted as Zn1, Zn and Zn+1, respectively, as shown in gure 1. The height differences Zn1 and Zn are dened as, Zn1 = Zn Zn1 , Zn = Zn+1 Zn . (1a) (1b)

The measured height differences between Xn1 and Xn and between Xn and Xn+1 are denoted as ZA(n) and ZB(n), respectively. When the inclination of the scanning stage is Tn against X, the measured height differences include Tn in addition to the height differences, Zn1 and Zn. Then, ZA (n) and ZB (n) are expressed as ZA (n) = ZB (n) = Zn1 + Tn + Tn , Zer , (2a) (2b) Zn +

where Zer in (2b) represents the relative error between two independent interferometers. This error is introduced in the calibration procedure of the initial phases. From equations (2a) and (2b), a recursive formula is obtained: Zn then
n

Zn1 =

ZB (n)

ZA (n)

Zer ,

(3)

In equations (4)(7), Z0 is the initial height at the point X0 and Z0 is the initial height difference between X0 and X1. We note that, for retrieving the prole, Z0 and Z0 can be chosen as any values without loss of generality, because they are merely origins in height and inclination. We also note that the relative error Zer gives the second-order polynomial bias to the prole Zn (see equation (6)). It also gives the rst-order polynomial bias to the stage inclination Tn (see equation (7)). To determine the true value of Zer, the output signals of the two independent interferometers must be compared using a completely at plane. However, it is not practical to provide such a reference at. Therefore, instead of the true value of Zer, the presumed value of Zer, denoted Zerp, is introduced to calculate the analytical prole. We determine Zerp under the assumption that the average surface of the prole is at, by applying the standard linear regression method to equation (6), that is, Zerp is determined so that the second-order polynomial component of the prole vanishes. Then, this presumed value Zerp is used to calculate equations (6) and (7). Consequently, the obtained prole, Zn, represents the true prole that involves third- and higher-order polynomial components, at the expense of losing its original parabolic component. On the other hand, the obtained stage inclination, Tn, represents the true inclination that involves second- and higher-order polynomial components, at the expense of losing its original linear component. We now summarize the actual procedure to retrieve the sample prole: (1) the initial values of Z0 and Z0 are set to zero, then Zerp is determined so that the second-order polynomial component of the prole vanishes. (2) The calculations of equations (6) and (7) are performed using Zerp instead of Zer. (3) For the convenience of showing the prole, the linear component of the obtained prole, Zn, is extracted and subtracted from Zn. The resultant Zn prole is shown in the gures in the following sections.

Zn is expressed by ( Zi
i=1 n

Zn = =
i=1

Zi1 ) +

Z0 Z0 . (4)

3. Experiments
In the experimental system, the sampling pitch X was 2.7 mm. A calcite beam displacer with a walk-off distance of 2.7 mm (Melles-Griot 03 PPD001) was used as the AUC. The distance between AUC and SUT was about 20 mm. The SUT was scanned with stepping-motor-driven screw-lead stages of 85 mm moving span (SGSP20-85, Sigma Koki Co.) or 1200 mm moving span (SGSP65-1200, Sigma Koki Co.). The phases of the signals, ZA(n) and ZB(n) were measured with two lock-in-ampliers (5610B, NF Co.). The SUT was moved with a speed of 1 mm s1 and stopped at every X of 2.7 mm when the phase data of the lock-in ampliers were obtained with a time constant of 10 ms and registered to a computer. The optical system was set up on a vibration-proof table and the optical path was covered with plastic sheets to prevent large air turbulence. There were heat sources such as the laser, the lock-in ampliers and the computer in the experimental room, but the room temperature was not specially controlled. Before the sample prole measurement, the system noise was investigated to evaluate the measurement resolution. The phases of ZA(n) and ZB(n) were recorded in 10 min. The scanning stage was not moving during this measurement. The result is shown in gure 2. The phase uctuation 2437

( ZB (i)

ZA (i)) n Zer +

The height of the sample, Zn, is given from equation (1a) as Zn = Zn1 +
n1

Zn1 Z0 . (5)

=
i=1

Zi + Z0 +

Then, Zn is obtained from equations (4) and (5). The nal expressions for Zn is given by
n1

Zn =
i=1

(n i)( ZB (i)

ZA (i))

(n 1)n Zer + n Z0 + Z0 . (6) 2 The stage inclination Tn is obtained from equations (2a) and (4) as
n1

Tn =

ZA (n) +
i=1

( ZA (i) Z0 .

ZB (i)) (7)

+ (n 1) Zer

T Yokoyama et al

Figure 2. The chart record of phase signals,

(a )

ZA(n) and

ZB(n).

(b )

Figure 3. Measured proles of the aluminium-coated glass plate (70 mm 10 mm (D)). (a) Measured with the double shearing interferometer 17 times; the standard deviation is 1.32 nm. (b) Measured with the Fizeau interferometer.

within a short period was smaller than 0.1 , which corresponds to a height variation of 0.1 nm. In spite of appreciable air turbulence and residual mechanical vibration in the experimental environment, such excellent height resolution was achieved because of the nearly common-path conguration in each shearing interferometer. The slow phase drift was also found in gure 2. The temperature change in the experimental room is considered to cause this drift. As seen in gure 2, the phase drifts of ZA(n) and ZB(n) arise in the same way, both in direction and amount. This is because the optical conguration of the two interferometers is the same and their optical path differences are approximately the same. Consequently, the drifts of the two phases are successfully canceled through the subtraction procedure in the evaluation of Zn and Tn (see equations (6) and (7)). 2438

The measured samples were (1) an aluminium-coated glass plate (70 mm 10 mm (D)), (2) a silicon plate (100 mm (L) 50 mm (W) 10 mm (D)) and (3) a silicon block for the x-ray mirror (400 mm (L) 50 mm (W) 30 mm (D)). Sample (1) was set on an 85 mm long stage, while samples (2) and (3) were set on a 1200 mm long stage. The scanned lengths of samples (1)(3) were 60 mm, 57 mm and 380 mm, respectively. The measurement time for one scan was 4 min for samples (1) and (2) and 20 min for sample (3). The obtained proles of samples (1)(3) are shown in gures 3(a), 4 and 5, respectively. In gure 5, a rising part is seen on the right side of the retrieved curve, but it is not a cumulative error of measurement, because Zerp is given so that the retrieved prole does not have a second-order polynomial component. To evaluate the system

Sub-nanometre double shearing heterodyne interferometry for proling large scale planar surfaces

Figure 4. The prole of the silicon plate (100 mm (L) 50 mm (W) 10 mm (D)). It was measured four times; the standard deviation is 1.30 nm.

reproducibility, samples (1) and (2) were measured 17 and 4 times, respectively, and the resulting standard deviations of Zn were 1.32 nm and 1.30 nm, respectively. Sample (3) was measured twice and the average difference of the obtained Zn values was 11.4 nm. It is considered that the manner of holding sample (3) was not appropriate, and thus the sample might have slipped slightly during the stage movement. We are going to optimize the holding mechanism for large samples. Now, we demonstrate the ability of our system to separate the prole from the stage inclination. For this purpose, the data of sample (3) before the calculation of equation (6) are shown in gure 6. This gure shows A: ZA(n); B: ZB(n) Zerp, which is the compensated ZB(n) with the presumed error value, Zerp; and B A: ZB(n) ZA(n) Zerp,. The plots of A and B show rugged variations, for example periodic variations with periods of about 10 mm and 160 mm. These variations are caused mainly by the stage tilt, because they appear in both A and B in the same manner. A small difference between plots A and B (see B A in this gure) is the second derivative of the prole as indicated in equation (3).

Thus, by integrating this value twice according to equations (4) and (5), the prole of the SUT is retrieved as shown in gure 5. The variation of the stage tilt, Tn, is also calculated using equation (7), separately from the prole of the SUT, which is shown in gure 7. From the plots shown in gures 5 and 6, one can nd how the system retrieves the prole using a small signal extracted from the measured signals, which are mainly noise due to the stage inclination. Next, the proposed method was compared with Fizeau interferometry. Sample (1) was measured with a Fizeau interferometer whose uncertainty is reported as /77 [4, 5]. The second-order polynomial component was subtracted from the obtained prole for comparison. The result is shown in gure 3(b). In this measurement, the difference in reectivity between the sample and the reference plane was large, resulting in a possible decrease of the signal-to-noise ratio and the system performance. However, the prole of the sample was obtained with sufcient reliability. Comparing gures 3(a) and (b), the prole obtained with the double shearing interferometer is seen to be quite consistent with that obtained with the Fizeau interferometer, although there remain quantitative differences in the absolute values. In the proposed system, one of the main causes of the absolute error is an optical misalignment of probes. For example, misalignment of the parallel plate leads to situations such that two probe light beams illuminating the centre spot of the three spots on the SUT are not superimposed completely, and further, the three spots are no longer in a straight line. These possible systematic errors should be investigated in the future and the uncertainty should be further examined. As a result of the experiments, the good reproducibility of the proposed system and the good agreement with the Fizeau interferometer are obtained, and these indicate that double shearing interferometry is capable of subnanometre proling of large samples.

4. Interpolation of height within the probe interval

The spatial frequency that can be measured with the proposed method is limited by the interval of the probes, X.

Figure 5. The prole of the silicon block (400 mm (L) 50 mm (W) 30 mm (D)).

2439

T Yokoyama et al

Figure 6. The measured height differences before the retrieval procedure. The sample is the same silicon block as in gure 5.

Figure 7. The variation of the stage inclination,

Tn, calculated from the data of gure 6.

The highest spatial frequency of the experimental system corresponds to twice X of 2.7 mm. Although a narrower X can be obtained by selecting an appropriate size for the crystal and the parallel plate, it is hard to detect two interference light beams separately when X is less than the beam diameter of 0.8 mm. To achieve the narrower X, an additional improvement of the optical conguration should be made. Instead of this, we show here an improved algorithm of measurement that expands the measurable spatial frequency without changing the optical conguration. In the improved algorithm, the intermediate points between the original sampling points are additionally measured and the height information of the intermediate points is interpolated. As shown in gure 8(a), the original sampling points, which are X apart, are represented by the rst subscript i of Xi,0 (i = 0, 1, 2, . . . , N), and the intermediate points between Xi, 0 and Xi+1,0 are represented by the second subscript j of Xi,j ( j = 1, 2, . . . , M). Here, the probe measures the height difference between Xi1, j and Xi,j , ZA(i, j), and 2440

that between Xi,j and Xi+1,j , ZB(i, j). This measurement is conducted by scanning the SUT stepwise by the distance X between Xi,j and Xi, j +1. The calculation of retrieving the prole is performed for ZA(i,j ) and ZB(i,j ) which have the same j. The resultant prole is specied as Pj, which consists of a sequence of heights, Zi,j (i = 0, 1, . . . , N), corresponding to Xi,j (i = 0, 1, . . . , N). In this way, the original prole P0 and M additional proles P1, P2, . . . , PM are retrieved. The M retrieved proles are connected to P0 by modifying the parameters Zerp, Z0 and Z0 of each Pj. Since the measurement is conducted with the same probe that has the unique relative phase difference, Zerp of P0 is employed for other proles. The parameters Z0 and Z0 of Pj, Z0,j and Z0,j are determined with the following steps: (1) the height (0) at Xi,j of P0, Zi,j , is calculated as the linear interpolation of the height between Zi,0 and Zi+1,0, as
(0) Zi,j =

j j Zi+1,0 + 1 Zi,0 M +1 M +1

(8)

Sub-nanometre double shearing heterodyne interferometry for proling large scale planar surfaces

(a )

(b )

where Z(x) is the assumed prole and a is the radius of the probe beam at e2 of maximum power. In the experimental system, a is given as 0.4 mm. Curves (B) and (C) in gure 8(b) show the simulation results without and with the interpolation, respectively. The parabolic component has been eliminated from each prole. The high frequency rugged portions are sensed successfully by the interpolation (curve (C)), while they are lost without interpolation (curve (B)). The sensed rugged portions in curve (C) appear broader than those of the assumed prole. This is because the beam diameter of the probe is larger than X. The simulation results indicate that the algorithm of the interpolation within the probe interval expands the measurable spatial frequency without changing the optical system. We consider that our improved algorithm enhances the capability of the double shearing interferometry.

5. Discussion
The height of the sample, Zn, obtained by the proposed system is slightly different to its true height, i.e. the height perpendicular to the stage surface. When the inclination angle of the stage, n, is given by Tn n = tan1 , (13) X the height difference of the sample, Zn, is expressed in terms of the true height difference, Hn, as Hn . (14) Zn = cos n By substituting equation (14) into equations (2a) and (2b), true height Hn is expressed in the same way as Zn, as
n1

Figure 8. Simulation results of interpolation within the probe interval. (a) Schematic of the sampling points for interpolation. (b) Simulation results: (A) assumed prole, (B) simulated prole without interpolation and (C) simulated prole with interpolation.

(2) The modied Zi, j , Zi,j , is given as Zi,j = Zi,j + i Z0,j + Z0,j . (9) Zi,j has been calculated assuming both Z0,j and Z0,j are zero. (3) Z0,j and Z0,j are determined so that the square sum of the (0) difference between Zi,j and Zi,j becomes minimum:
N (0) Zi,j i=0

Hn =
i=1

(n i) ( ZB (i)
n1

ZA (i)) cos i

Zi,j

min.

(10)

i=1

(n i) cos i Zer + (n 1) H0 + H0 .

(15)

This algorithm for the interpolation within the probe interval was examined using computer simulation. The line (A) in gure 8(b) shows the assumed surface prole for the simulation, which is a low frequency sinusoidal prole with rugged portions whose width is narrower than the probe interval of X. Both the amplitude of the low frequency sinusoidal prole and the height of rugged portions are 10 nm. The scan length is 50 X and the number of the intermediate points for the interpolation within X is nine (M = 9), that is, X is 0.27 mm. The simulation was performed with the assumption that the probe beam has a Gaussian distribution of intensity. Then, the measured height difference, Zmeas, with a single shearing interferometer illuminating the points X and X + X, is given by Zmeas (X) = eimeas (X) = meas (X) , 4 2 (x X)2 21 exp a a2 X) Z(x)) dx, (12) (11)

When n is small, equation (15) reduces to

n1

Hn
i=1

(n i) ( ZB (i)
n1

ZA (i)) 1

i2 2

i=1

(n i) 1

i2 2

Zer + (n 1) H0 + H0 . (16)

Then, the difference between Zn and Hn, denoted Zn, is obtained as Zn = Zn Hn

n1

=
i=1

(n i)( ZB (i)

ZA (i)

Zer )

i2 , 2

(17)

4 + i (Z(x +

where Z0, Z0, H0 and H0 are assumed to be zero. The magnitude of Zn depends on the prole of the SUT, but Zn is rather small when each i is small. In the present experiment, the overall change of Tnwas approximately 300 nm as seen from gure 7, and thus the overall change of i is estimated as 1.1 104 radian. In this case, the error between Zn and Hn is considered to be negligible. 2441

T Yokoyama et al

In the current system, two phases are measured discontinuously at every X interval. When the phase changes by more than between two consecutive measurements, the actual phase change may not be determined uniquely from the measured phase change. Thus, the change in the inclination of the sample must be less than /4 X. This limitation is expanded by monitoring the phase change continuously during the scanning; however, there remains another limitation which arises from the limited range of light detection. Two photodetectors of this system are aligned initially so that each interference beam illuminates the centre of the corresponding detector. The diameter of the detector is 2 mm and the path length from the SUT to the detectors is approximately 100 mm, thus the overall stage inclination during the measurement should be less than 24 mrad. In the proposed retrieval procedure, the second-order polynomial component is lost from the recovered prole, while the higher-order polynomial components are obtained correctly. This arises from the presumption of the relative error value, Zer, between the phases of two independent shearing interferometers. However, in the eld of atness measurement, low-order polynomial components of the prole are often insignicant, and thus these components are removed by the ltering when measuring local proles or roughness. In the eld of silicon wafers, especially for those more than 300 mm in diameter, there is a demand for measurement of nanometre height variations with a spatial wavelength of 0.220 mm. The proposed method meets such a demand because it is capable of measuring a long span with great sensitivity in height. For large silicon mirrors or optical ats, such requirements to measure height variations of specic spatial frequencies are expected in the manufacturing process. The measurement time of the experimental system is quite long, particularly for a large sample, because the stage stops to measure at every sampling pitch. Such a measurement may produce erroneous results because of the effects of environmental change during the measurement, although the proposed system is robust for temperature variation owing to the conguration of two common-path interferometers. In addition to this, rapid measurement is very important for industrial use. To decrease the measurement time, it is necessary to scan the SUT continuously. In continuous measurement, the effects of the stage travelling during phase observation must be considered and simultaneous acquisition of two phases is essential. Since the time constant of the lockin amplier and the speed of stage scanning are 10 ms and 1 mm s1 respectively, the travelling distance of the stage during the time constant is calculated as 10 m. In this case, the effects of the stage travelling during observation are considered to be small enough in comparison with the probe interval (2.7 mm) and diameter (0.8 mm). Currently, each phase datum is acquired at an independent trigger clock of each lock-in-amplier and the acquisition time differs by 0.1 s in the worst case. Thus, synchronization of two clocks will enable continuous measurement. The objective of the proposed double shearing interferometry is to measure long samples in the range of hundreds of millimetres or more with subnanometre height resolution. We have conrmed that our method is capable of this through the experiment as described in section 3. 2442

We are going to improve the moving stage and the holding mechanism to measure longer samples and investigate the uncertainty of the system. On the other hand, there is a further requirement for the measurement of large samples with higher spatial frequency, for example in the case of silicon wafers as described above. In section 4, we proposed an improved algorithm of the measurement to increase the measurable spatial frequency without changing the optical conguration. Higher spatial-frequency components of the prole can be measured if we achieve a smaller spot size and a narrower spot separation. The authors are now implementing an improved optical conguration of the double shearing interferometry. In the new optical system, each probe beam is focused on the SUT with a spot size of several tens of microns and the spot separation is narrowed to a few hundreds of microns. We plan to report on this system in the future.

6. Conclusions
Double shearing heterodyne interferometry has been presented for proling large-scale quasi-planar surfaces with subnanometre resolution. In this method, the prole of the SUT can be measured separately from the variation of the inclination of the scanning stage. Since the system is composed of a pair of common-path shearing interferometers, the effects of mechanical vibration and temperature change are minimized. Based on this optical conguration and the heterodyne detection, the proposed method achieves a sensitivity of subnanometre order and a robust measurement against disturbances in the measurement environment. We have demonstrated the performance of the proposed system. A height resolution of 0.1 nm was achieved in a normal laboratory environment. The standard deviation of a surface prole with a scan length of 60 mm was 1.3 nm, even with the use of the common screw-lead scanning stage. The measured results show that the surface proles are determined separately from the stage inclinations. The obtained prole was conrmed to be in good agreement with the prole measured with the Fizeau interferometer. With these results, we may conclude that the proposed method is promising for the measurement of subnanometre surface proles.

Acknowledgments
The comparison measurement in this work was performed with the Fizeau interferometer of the National Institute of Advanced Industrial Science and Technology (AIST), with the help of Dr T Takatsuji and Dr H Matsumoto of AIST. The silicon block measurement was supported by Professor S Shimada of Osaka Electro Communication University, Dr A Komura and Mr T Inoue of Hitachi Zosen Co. and Mr M Chikamoto of the Department of Precision Science & Technology, Osaka University. The authors are grateful to all of them for encouraging results.

References
[1] Yatagai T and Kanou T 1984 Aspheric surface testing with shearing interferometer using fringe scanning detection method Opt. Eng. 23 35760

Sub-nanometre double shearing heterodyne interferometry for proling large scale planar surfaces

[2] Bruning J H, Herriott D R, Gallagher J E, Rosenfeld D P, White A D and Brangaccio D J 1974 Digital wavefront measuring interferometer for testing optical surfaces and lenses Appl. Opt. 13 2693 [3] Takeda M, Ina H and Kobayashi S 1982 Fourier-transform method of fringe-pattern analysis for computer-based topography and interferometry J. Opt. Soc. Am. 72 156 [4] Takatsuji T, Ueki N, Hibino K, Osawa S and Kurosawa T 2001 Japanese Ultimate Flatness Interferometer (FUJI) and its preliminary experiment Proc. SPIE 4401 83 [5] Takatsuji T, Osawa S, Kuriyama Y and Kurosawa T 2003 Stability of the reference at used in Fizeau interferometer and its contribution to measurement uncertainty Proc. SPIE 5190 431 [6] Bieren K V 1982 Pencil beam interferometer for aspherical optical surfaces Proc. SPIE 343 101 [7] Qian S, Jark W and Takacs P Z 1995 The Penta-Prism LTP: a long-trace-proler with stationary optical head and moving penta-prism Rev. Sci. Instrum. 66 2562

[8] Takacs P Z, Church E L, Bresloff C J and Assoud L 1999 Improvements in the accuracy and the repeatability of long trace proler measurements Appl. Opt. 38 5468 [9] Zhao Y, Li Z, Li D, Tiqiao X and Shaojian X 2001 Principle of -phase plate long trace proler for synchrotron radiation optics Opt. Commun. 200 23 [10] Chhatterjee S and Kumar Y P 2002 Long-trace proler with cyclic optical conguration Appl. Opt. 41 5857 [11] Sommargren G E 1981 Optical heterodyne prolometry Appl. Opt. 20 610 [12] Pantzer D, Politch J and Ek L 1986 Heterodyne proling instrument for the angstrom region Appl. Opt. 25 4168 [13] Yasui T, Araki T and Suzuki N 1996 An intermittent frequency offset lock of a transverse Zeeman laser to an iodine stabilized HeNe laser Opt. Rev. 3 197201 [14] Wood R W 1961 Physical Optics 3rd edn (New York: Dover) p 668 [15] Umeda N, Tsukiji M and Takasaki H 1980 Stabilized transverse Zeeman laser as a new light source for optical measurement Appl. Opt. 19 442

2443