OPTICAL REVIEW Vol. 16, No. 2 (2009) 188–191

Wide-View-Angle =4 Plates for Diagnosing 193-nm Lithography Tools

Yohko FURUTONO and Hiroshi NOMURA Semiconductor Company, Toshiba Corporation, 8 Shinsugita-cho, Isogo-ku, Yokohama 235-8522, Japan (Received July 18, 2008; Accepted November 13, 2008)

We developed new =4 plates, which are insensitive to angle of incidence at 193 nm. Thanks to the development of the wide-view-angle =4 plates, a polarimeter reticle is able to measure the polarization of illumination of immersion lithography tools. The wide-view-angle =4 plates are composed of four crystal plates; two are made of crystalline quartz categorized as positive uniaxial crystals, and two are made of sapphire categorized as negative uniaxial crystals. Since the birefringent crystals differ in regard to which is the higher of the two refractive indices for the ordinary ray and for the extraordinary ray, their combination mitigates the retardation change at oblique incidence. We measure the retardation change of the new =4 plates by the Senarmont method. Results show that the change of retardation is mitigated within an angular range of incidence of 20. # 2009 The Optical Society of Japan Keywords: polarization, lithography, quarter-wave-plate, retardation, birefringent

1. Introduction Lithography is a key technology for the mass production of LSIs. For the progress of miniaturization of LSIs, it is necessary to control several properties of lithography tools. 0-λ plate ( n > n ) Polarization of illumination should also be one of these e o λ /4 plate properties in the near future.1) For the two-beam interference ( ne < no ) through an objective with an NA above 1, the p-component of polarization of an incident light in resist does not wholly Fig. 1. The wide-view-angle =4 plate composed of four plates. interfere with that of the other light, so that the contrast of A pair of crystalline quartz plates is a =4 plate, and the other pair fine images degrades. Techniques for measuring polarization of sapphire plates is a 0 plate. states of illumination should be constructed by a system on the reticle stage, such as a special reticle with and =4 plates. It has been difficult to realize such a system in visible and deep UV regions. The phase of a light wave because of space restrictions. shifts depending on a light path in the material, and the light In 2006, H. Nomura proposed a polarimeter reticle, which path changes according to angle of incidence; consequently, is composed of two new optical devices.2,3) One is a thin the retardation also changes according to angle of incidence. plate . Conventional polarizing such as To avoid this change of retardation, conventional =4 plates Nicol , Glan–Thompson prism, and Rochon prism require a collimating optics in front of them to penetrate all require a long space to split an input light into two orthogo- the rays to the face of the =4 plates at the normal incidence. nal linear polarizations. The thin plate polarizers made of It is hard to put the collimating optics on a reticle with a calcite (CaCO3), which has large birefringence at a wave- thickness of 6.35 mm. The wide-view-angle =4 plate makes length of 193 nm, can greatly economize on space. The other the polarimeter mask simple. device is the =4 plate with a wide range of view angles. This paper describes details of the wide-view-angle =4 3. Theory of Wide-View-Angle =4 Plates plates and measurement results of the retardation change The new =4 plates are composed of four crystal plates; with angle of incidence. two are made of crystalline quartz and two are made of sapphire. The two crystalline quartz plates compose a =4 2. The =4 Plates plate and the two sapphire plates compose a 0 plate as Prior to discussion on the wide-view-angle =4 plates, illustrated in Fig. 1. Index ellipsoids of crystalline quartz and conventional =4 plates are briefly described. Widespread sapphire are shown in Figs. 2(a) and 2(b), respectively. =4 plates at visible to deep UV regions are composed of Crystalline quartz is categorized as a positive uniaxial crystal two or more plates made of uniaxial birefringent materials, with refractive indices of no > ne, and sapphire is categorized and one linear component of polarization, which is parallel as a negative uniaxial crystal with refractive indices of no < to the slow axis (s-axis), retards the phase by 90 compared ne, where no and ne are refractive indices for the ordinary ray with the other orthogonal component, which is parallel to the (o-ray) and for the extraordinary ray (e-ray), respectively. fast axis (f-axis), through the plates. On the surface of a Whether the uniaxial crystal is ‘‘positive’’ or ‘‘negative’’ holder, there is a scar that indicates s-axis or f-axis of the determines which ray retards the other at oblique incidence. =4 plate. Crystalline quartz (-SiO2) and sellaite (MgF2) Therefore, combination between the positive and negative are widely used as materials of =2 and =4 plates employed uniaxial crystals may reduce the change of retardation with

188 OPTICAL REVIEW Vol. 16, No. 2 (2009) Y. FURUTONO and H. NOMURA 189

(a) Quartz (b) Sapphire Table 1. Azimuthal angles (deg) of each optical element. n z nz (a) (b) (c) (d) (e) n =4 plate =4 plate Sample Wollaston prism y n no y no 45 0 þ 45 0 45 45 45 þ 90 45 0 nx n nx þ ne e 45 90 45 90 45

90° Fig. 2. Index ellipsoids of crystalline quartz and sapphire. Crys- talline quartz is categorized as a positive uniaxial material and 45° therefore, the refractive index for the ordinary ray is higher than that for the extraordinary ray. On the other hand, sapphire is categorized as a negative uniaxial material and therefore, the refractive index for 0° the extraordinary ray is higher than that for the ordinary ray.

Optic axis (a) (b) (c) (d) (e) λ /4 plate Wollaston λ /4 plate sample Wollaston prism λ /4 plate prism (f) Fig. 4. Measurements were made in three cases of azimuthal FIX ° θ FIX FIX FIX 0 - detector Ray source angles, namely, 0, 45, and 90 , between the f-axis of the sample and ArF the incident plane. 193nm

under N2 environment

Fig. 3. The measurement configuration. The Senarmont method second Wollaston prism (e) for the measurement at three was adopted to measure the retardation change with angle of azimuthal angles of sample (d) are listed in Table 1. incidence. Figure 4 shows the azimuthal angles of sample (d). This measurement is done for these three cases. For example, the measurement of the retardation change angle of incidence. When the ray is incident to the crys- with angle of incidence when the incident plane lies at 45 talline quartz =4 plate at the normal, the retardation angle to the f-axis of the sample =4 plate (d) is explained. =4 of the o-ray to the e-ray is designed to be just 90. However, Plate (a), =4 plate (c), and Wollaston prism (e) were set at for an incidence other than the normal, the retardation azimuthal angles of 45, 90, and 0, respectively. The gradually exceeds 90 as the angle of incidence increases. intensities were measured by a power sensor (f), rotating the On the other hand, when the ray is incident to the normal Wollaston prism (b) in increments of 5. The retardation to the sapphire 0 plate, the retardation must be nothing. error of the wide-view-angle =4 plate (d) from 90 is The retardation gradually arises as the angle of incidence connected by twice the angle of the first Wollaston prism increases, and the sign of the retardation is contrary to the (b) as =4 plate. Therefore, retardation change at the oblique ¼ 2; ð1Þ incidence can be mitigated by the 0 plate. Interestingly, the retardation change depends on angles where is defined as the angle of Wollaston prism (b) to the between incidence plane and the s-axis of the =4 plate. sample (d) for the minimum intensity. Other cases such that When the incidence is in the plane that involves the f-axis or the incident planes are parallel and perpendicular to the the s-axis, the behavior of the retardation change is not f-axis of the sample (d) were also checked. complicated. When the incidence plane is parallel to the s- The measurement chamber was connected with argon axis, the o-ray retards the e-ray, that is, the retardation rises fluoride (ArF) excimer laser equipment. The wavelength of according to the angle of incidence. On the other hand, when the light is 193 nm. All the optical paths are in the nitrogen the incidence plane is parallel to the f-axis, the o-ray (N2) environment. The ArF laser emits a high-power light advances to e-ray, that is, the retardation declines. The linearly polarized in the horizontal direction. For fear that former is symmetric in curve to the latter. too high power of a light could cause errors from thermal expansion, the light was diminished by mirrors before being 4. Experiment led to the measurement chamber. Prior to the measurement, We adopted the Senarmont method in the measurement of we defined purity of polarization as retardation change with angle of incidence. Figure 3 shows I I the measurement configuration. By exchanging =4 plate (c) I ¼ max min ; ð2Þ and sample =4 plate (d) for comparison with the original Imax þ Imin Senarmont configuration, we measured the retardation of and confirmed that the purity of the ArF laser is 0.998 at sample (d). The three combinations among azimuthal angles least. The first =4 plate (a) with an azimuthal angle of 45 of first Wollaston prism (b), second =4 plate (c), and and the first Wollaston prism (b) can produce high-purity 190 OPTICAL REVIEW Vol. 16, No. 2 (2009) Y. FURUTONO and H. NOMURA

1 -50 deg. 180 Measured 0 deg. Sample azimuth 45 deg. -30 deg. Measured 45 deg. -15 deg Measured 90 deg. 0 deg. Recalculated 0 deg. 150 15 deg. Recalculated 45 deg. 30 deg. Recalculated 90 deg. 0.1 120

90

Intensity [a.u.] 0.01

60

Guide for eye Retardation [deg.]

0.001 30 -45 0 45 90 135 180 225 270 315 Azimuthal angle of Wollaston prism (b) [deg.] 0 -50 -40 -30 -20 -10 0 10 20 30 40 50 Fig. 5. An example of measurement results at an azimuthal angle Angle of incidence [deg.] of 45. We plot the data each angle of incidence. The angle at the minimum indicates the retardation for the azimuthal angle. This Fig. 7. Measured retardation data agree well with recalculated retardation shifts from 90 as the incident angle becomes higher. results, taking the ray tracing method for sixteen rays split from an incident ray inside the wide-view-angle =4 plate into account.

180 Measured 0 deg. Measured 45 deg. The retardation at the normal incidence was controlled to Measured 90 deg. 90 with an offset less than 10. An incident angle of 20 at 150 Initially designed 0 deg. Initially designed 45 deg. the object plane is equivalent to an NA of 1.35, which is the Initially designed 90 deg. maximum for water-based immersion lithography. The 120 measured retardations vary in the range between more than

° 90 about 85 and about 100 for a range of incident angle from 15 20 to 20. This means that the retardation errors are less 60 than 7.5 for the range of incident angles. When the wide- Retardation [deg.] 40° view-angle =4 plate is used on the reticle stage, we have

30 already confirmed that a very small level of measurement errors occurs in the measurement of polarization states of

0 illumination. Without the offset, the retardation at the -50-40 -30 -20 -10020304050 10 normal incidence should be 90. The measured retardations Angle of incidence [deg.] at oblique angles of incidence were different from our Fig. 6. Initially calculated data and measurement data for a wide- expectations obtained by the initial calculation. Although the view-angle =4 plate with a total thickness of 0.8 mm. Solid lines calculation described that the retardation at the 45 azimuth and dashed lines indicate experimental and calculated values, remains constant at oblique incidence, measured data do not respectively. agree with the calculation but lie on a curve. We considered there could be two reasons for this difference. One is that uncertainness of the optical indices linear polarization in any direction. The power sensor of quartz and sapphire at a wavelength of 193 nm might measures intensities for an angular range of incidence from result in the difference. Unfortunately, for most materials, 50 to +50 in 5 increments. Figure 5 is an example of optical indices at the wavelength are still unknown to us. data at an azimuthal angle of 45. The angle for the And the other reason considered was that the calculation minimum intensity shifts away from 90 according to angle method was no good. of incidence. Retardation shift of the sample depends on We recalculated the retardation by a strict method. incident angles. The curves are obtained by least-squares Figure 7 shows measured data of retardation and recalcu- fitting of the measured data. lated results. This calculation includes ray tracing where the incident ray to this wide-view-angle =4 plate is split into 5. Results and Discussion sixteen rays. These graphs agree well with our experimental Figure 6 shows the initially calculated and measurement results. This suggests that we should take the effects of beam data. Conventional two-plate type wave plates have a split into account for the detailed calculation. retardation shift of much more than 360 even if the incident angle is less than 10. According to our initial design, the 6. Conclusions retardation shift was mitigated to be less than 25 at an We designed and fabricated a wide-view-angle =4 plate, incident angle of 10, as shown in Fig. 6. However, which is composed of two sapphire plates and two measurement results actually show that a combination crystalline quartz plates, and confirmed that it has stable between crystalline quartz and sapphire can mitigate much retardation in the range from 85 to 100 for incident angles more than the initial calculation. between 20 at a wavelength of 193 nm. The combination OPTICAL REVIEW Vol. 16, No. 2 (2009) Y. FURUTONO and H. NOMURA 191 of crystalline quartz and sapphire successfully mitigates the References effect of oblique incidence on retardation shift. 1) T. Higashiki: Hikari Risogurafi Gijutu (Optical Lithography Acknowledgements Technology) (ED Research, Tokyo, 2002) p. 17 [in Japanese]. 2) H. Nomura: Dr. Thesis, School of Engineering, Tohoku The authors would like to thank Yuji Hirose and Shin-ichirou University, Sendai (2007). Toda of Kogaku Giken Co., Ltd. for development of optical 3) H. Nomura and Y. Furutono: Microelectron. Eng. 85 (2007) devices. They are grateful to Dr. Osamu Wakabayashi of 1671. Gigaphoton Inc. for experimental discussions and assistance.