Predicting surface gure in diamond turned calcium uoride using in-process force measurement

Eric R. Marsh,a) Binu P. John, and Jeremiah A. Couey

Machine Dynamics Research Laboratory, The Pennsylvania State University, 21 Reber Building, University Park, Pennsylvania 16802

Jue Wang and Robert D. Grejda

Corning Tropel, 60 OConnor Road, Fairport, New York 14450

R. Ryan Vallance
Precision Systems Laboratory, The George Washington University, 738 Phillips Hall, 801 22nd St., N.W., Washington, DC 20052

(Received 7 July 2004; accepted 1 November 2004; published 5 January 2005) Single crystal calcium uoride CaF2 shows signicant variation in material properties as a function of crystallographic orientation. The surfaces generated by material removal processes such as diamond turning are inuenced by this anisotropy and consequently show periodic undulations aligned with the crystal structure. This article explores the relationship between surface gure and cutting forces measured during the diamond turning of single crystal calcium uoride. The cutting forces, when mapped to the physical geometry of CaF2 plano (at) optics, show good correlation with surface gure measured by interferometry. A model is presented to predict the surface gure error from the experimentally measured normal component of the cutting force. The model also shows how the surface gure obtained under various machining parameters may be extrapolated from force measurements made during a single diamond turning operation. 2005 American Vacuum Society. [DOI: 10.1116/1.1839897]

I. INTRODUCTION With the shift of excimer laser wavelengths to 193 and 157 nm, calcium uoride has emerged as an important optical material for lithography systems and high power light source systems due to its wide band-gap structure.1 Because of the intrinsic features of the uoride crystal materials, the high absorption of contamination, and scattering losses at these short wavelengths, well nished optical surfaces are preconditions for achieving high performance.24 Considerable effort is being devoted to develop economical processing techniques for CaF2 optics. Previous work demonstrated that CaF2 is diamond turnable and it is now possible to turn rough blanks into seminished optics in less time than traditional grinding and lapping.57 These seminished optics are immediately ready for nal guring and polishing with magnetorheological nishing, which is routinely used to generate calcium uoride surfaces of suitable quality. Despite promising results obtained in applications such as photolithography optics, calcium uoride is widely considered a challenging material to work with because of its high coefcient of thermal expansion 24 106 C1 and relatively low hardness (one-seventh that of silicon). However, calcium uoride is one of the few visible or ultraviolet wavelength optical materials that may be machined with single point diamond tools to obtain specular surfaces and minimal subsurface damage. Figure 1 shows an atomic force microscope (AFM) image of a typical diamond turned workpiece
a)

Author to whom correspondence should be addressed; electronic mail: emarsh@psu.edu J. Vac. Sci. Technol. B 23(1), Jan/Feb 2005

with 3.5 nm Ra and 18.4 nm peak-to-valley microroughness. The subsurface damage, as measured by the quasiBrewster angle measurement technique, ranges from 250 to 500 nm.8 Quarter-wave, diamond turned optics processed by a single iteration of magnetorheological nishing were improved to / 20 peak-to-valley gure error and 24 Ra microroughness (measured by AFM). These promising results justify the pursuit of solutions to the two issues associated with diamond turning crystalline materials such as CaF2; tool wear and spherical aberration aligned with the crystal structure. Diamond tool wear is a consideration when turning CaF2, but the wear rate is much slower than observed in silicon turning. We found that it is possible to achieve specular surfaces for 35 consecutive passes on CaF2 80 mm plano (at) optics with mineral oil coolant and monocrystalline synthetic diamond tools with chemically faceted cutting edges. The surface gure and surface nish obtained in diamond turning is strongly inuenced by the crystalline structure of calcium uoride. Yan reports that the same radial spokes of damage that are well known in silicon can arise in CaF2 with various combinations of typical diamond turning parameters.5 Yan also reports machining parameters that may be chosen to generate damage-free surfaces in single point diamond turning for all crystal orientations.7 This orientation-dependent behavior is further documented in the extensive literature on silicon and diamond, two other crystalline materials that show the same large variations in machinability with crystal orientation.913
2005 American Vacuum Society 84

84

0734-211X/2005/23(1)/84/6/$19.00

85

Marsh et al.: Predicting surface gure in diamond

85

FIG. 1. AFM image of 3.5

nm Ra diamond turned CaF2 (111).

FIG. 2. Single point diamond turning cutting geometry.

When diamond turning CaF2 optics, the cutting tool encounters different orientations of the crystal within each rotation of the workpiece. As with diamond and silicon, the orientation-dependent material properties of calcium uoride cause variation in the apparent material stiffness that leads to oscillatory cutting forces and undulations in the surface gure. During the course of this work, the surface gure of plano (111) CaF2 workpieces invariably showed three-lobed (trifoil) spherical aberration with amplitudes ranging from 25 to 250 nm peak to valley depending on the material removal rate and the condition of the diamond tool. Despite this lobing, single point diamond turning, followed by magnetorheological nishing as needed, is a good alternative to traditional grinding and polishing because the material removal rate is much higher, the subsurface damage propagated into the workpiece is low (submicrometer) and aspheric shapes can be generated and polished directly. An essential requirement in turning brittle materials such as CaF2 is the selection of cutting parameters such that the material removal mechanism is in the ductile mode regardless of the crystal orientation.14 This is achieved by keeping the effective depth of cut below the critical uncut chip thickness tcrit, which varies with tool geometry and crystal orientation. Yan studied this effect for (111) CaF2 turning and found that tcrit is 85 nm in the hard 112 direction with at diamond tools and 20 rake angle.5 In our work, we used round nosed tools and found the critical depth of cut to be somewhat larger, presumably as a result of the substantially different geometry of the chips. A microfracture model explaining the difference between the microfractures left by round and at tools is found in Blake and Scattergood.15 In summary, CaF2 presents an interesting opportunity in optical fabrication because the material offers both favorable optical properties and the possibility of reduced manufacturing costs by replacing the traditional grinding steps with a single diamond turning setup. The remainder of this article demonstrates the relation between crystal structure-induced spherical aberration and cutting forces that arises in diamond turning anisotropic CaF2. A model for predicting the magnitude of these aberrations from in-process force measurement is presented and validated. The model enables the prediction
JVST B - Microelectronics and Nanometer Structures

of aberrations at various conditions by extrapolation from force measurements made during a single diamond turning operation. II. PREDICTIVE MODEL FOR FIGURE ABERRATIONS This section presents a model to predict the trifoil spherical aberrations that result from machining an anisotropic material like (111) CaF2. In a single point material removal process such as diamond turning, the stiffness of the machine tool dictates the extent to which variations in cutting forces affect the deection of the cutting tool. A diamond turning machine, including the main structure, the spindle, the toolholder and tooling, and the CaF2 workpiece with xturing comprises a structural loop with some nite stiffness. This stiffness controls the deection of the tool, with respect to the workpiece, in the presence of cutting forces, and these deections result in spherical aberrations. The workpiece material also contributes a relatively small amount of additional stiffness when in contact with the cutting tool. This stiffening effect is well established because of its fundamental role in the prediction of regenerative chatter.16,17 Figure 2 shows a plano CaF2 workpiece that rotates as a diamond tool is slowly fed across the face. The axial depth of cut h (z direction) typically ranges from 1 to 100 m. The cross feed f (x direction) usually ranges from 1 to 10 m per revolution of the workpiece. A round-nosed diamond tool of radius R cuts along its leading edge as the tool slowly advances across the rapidly rotating workpiece. With a typical 2 mm radius round-nosed diamond tool, the arc-shaped contact length between tool and workpiece will be about 200 m if the axial depth of cut is 10 m (contact length 2hR). The stiffening effect of the tool-workpiece contact is somewhat complicated by the round geometry of optics diamond turning tools. Figure 3 shows the prole left by ve workpiece rotations with a round-nosed diamond tool of radius R. In practice, the feed per revolution f is very small with respect to the other dimensions, but is shown greatly

86

Marsh et al.: Predicting surface gure in diamond

86

The product of this linearized form of the instantaneous chip area and the cutting force coefcient Kt of the calcium uoride crystal predicts the normal component of the cutting force Fnormal = KtA.
FIG. 3. Instantaneous chip area from consecutive workpiece rotations with feed f and nominal depth of cut h.

exaggerated to reveal the underlying geometry. In the presence of some disturbance, if the tool penetrates the workpiece by some amount beyond the equilibrium axial depth of cut h, the force on the tool will increase. In fact, the normal force may be assumed to be proportional to the instantaneous chip area, which is shaded for the current pass in Fig. 3.16 Under steady-state conditions = 0 , the instantaneous chip area of a round-nosed tool remains constant at approximately f h. The additional instantaneous chip area generated by an increase in depth of cut may be determined by integrating to nd the shaded area under the curve in Fig. 3. The resulting area may be expressed as a function of the tool radius R, the nominal depth of cut h, the feed f, and the displacement of the tool from its equilibrium position .18 As a result of the round-nosed (circular) tool geometry, the function is nominally nonlinear 2 2R h 3

The force within the structural loop of the machine tool must balance this cutting force. The dynamic terms are neglected since the diamond turning process is stable and the rotation rate of the workpiece is low with respect to the resonant frequencies of a precision lathe. In that case, the equilibrium equation is given in Eq. (4). The structural loop stiffness is kloop and x is the total deection within the structural loop from the tool to the workpiece kloopx = Fnormal . 4

A=h f+

+hh h+

+h+

In diamond turning, the feed rate is so small with respect to the tool radius R and the nominal depth of cut h that the instantaneous chip area may be accurately linearized by a simple Taylor series expansion A h f+ 2hR . 2

These equations provide the necessary information to (1) compute the surface gure in an anisotropic workpiece given cutting force data and (2) estimate the surface gure that would be obtained for arbitrary cutting parameters given either the structural loop stiffness of the machine kloop or the surface gure of a trial workpiece (to obtain Kt). The surface gure is estimated by dividing the measured (orientation dependent) cutting force by the loop stiffness kloop. In our tests, the structural loop stiffness of our machine with toolholders and workpiece xturing is 10 N / m. This value was obtained by applying known loads with a strain gauge load cell (Interface SM-100 with Omega DP25-S meter) and measuring the deection across the structural loop with a capacitance probe (Lion C1-C with DMT 12 driver). A typical test cut with 2 N of variation in the normal component of the measured force will be expected to have a 2 / 18.4= 110 nm peak-to-valley trifoil surface gure. Similarly, the cutting force coefcient of this material may be calculated given the nominal machining parameters of feed per revolution f and depth of cut h. Since Fnormal varies with the orientation of the CaF2 crystal, the cutting force coefcient Kt has a range of values

FIG. 4. Hydrostatic way diamond turning lathe with air bearing spindle. J. Vac. Sci. Technol. B, Vol. 23, No. 1, Jan/Feb 2005

87

Marsh et al.: Predicting surface gure in diamond

87

FIG. 5. Calcium uoride cutting forces after ltering to remove three-lobe uctuations.

Kt =

0.5 to 2.5 N kN Fnormal = = 15 . A 5 m 100 m mm2

Degradation of the cutting edge also leads to increases in Kt. Therefore, monitoring changes in cutting force can provide valuable feedback on the tool edge wear. III. EXPERIMENTAL SETUP In order to validate the cutting force model presented in the previous section, calcium uoride turning tests were performed on a hydrostatic way, diamond turning lathe with a 10 nm programmable resolution (Moore Nanotechnology Systems AG150). An illustration of the machine and its vertical air bearing spindle (Professional Instruments TwinMount) is shown in Fig. 4. A high resolution three-axis dynamometer (Kistler MiniDyn 9256A2) measures the three components of force on the tool. A capacitance probe (Lion Precision C1-C with DMT 10 driver) targets the rotating workholder (not shown) in order to synchronize the cutting force data with the rotation of the workpiece. This index pulse allows accurate mapping of the cutting forces to the trifoil structure of the (111) CaF2 crystal in the presence of spindle speed variations.

When machining brittle materials such as calcium uoride, a negative rake angle is benecial due to its mitigating effects on fracture.19 The synthetic, monocrystalline diamond tools used in these turning experiments achieve the negative rake angle with a chamfer, or radial rake. Most tests were run with 0 rake angle or 30 radial rake angle with a 10 clearance angle, and 1.5 mm nose radius (Edge Technologies, Inc.). More negative radial rake angle tools (45 and 60) were also used with less favorable results. The tool edge radius, which may be measured in a low voltage scanning electron microscope or by AFM, is less than 50 nm.20 The tools are chemically faceted on the rake and nose facets to remove surface and subsurface damage from the tool edge. Mineral oil coolant was used during all cutting tests. The three components of the diamond turning cutting forces contain useful information about the state of the material removal process. Changes in the forces reect changes in the geometry of the tool (wear) and the surface nish of the machined surface. Figure 5 shows the three components of the cutting force (with the undulations due to anisotropy removed by digital ltering) as the tool progresses from the outside to the inside of a calcium uoride workpiece. The nominal force levels remain roughly constant, despite the

FIG. 6. Two revolutions of cutting force data showing the three-lobe uctuations on a (111) CaF2 crystal. JVST B - Microelectronics and Nanometer Structures

88

Marsh et al.: Predicting surface gure in diamond

88

FIG. 7. Cutting force components mapped to the surface of a 80 mm plano (111) CaF2 workpiece.

fact that the cutting velocity approaches zero at the center of the workpiece. In these tests, a small hole was ground into the center of the workpieces prior to cutting so the cut ends 1mm from the axis of rotation. Although not shown here, the cutting force results from subsequent passes over the workpiece are very similar to the data shown in Fig. 5; however, after many facing cuts, the forces eventually increase as the tool edge begins to wear. At this point, the surface of the workpiece shows visible damage in the hard machining directions of the CaF2 crystal and the diamond tool must be resharpened. Figure 6 shows a close-up view of the unltered cutting forces with uctuations over two revolutions of the workpiece that are due to anisotropy. The three-lobe uctuations result from interaction of the cutting tool and the crystalline structure of the (111) calcium uoride workpiece. The uctuations in forces are highly repeatable and additional similarity is seen between the subsequent revolutions (rst three peaks versus the second three peaks). The three peaks in each revolution are not identical because the crystal is not perfectly aligned with the cutting plane. As a result, the diamond tool engages the crystal at a small angular offset that leads to the slight variation between the three lobes. This effect was originally documented in diamond polishing.13 Figure 7 shows the three components of cutting force mapped to the physical geometry of the round, at workpieces. The three-lobe (trifoil) uctuations in force resulting from the (111) symmetry of the crystal are clearly evident.

FIG. 9. Normal cutting force and measured surface gure of a diamond turned CaF2 workpiece at 1200 rpm and 4 m / rev feed rate.

A series of calcium uoride optics were diamond turned with 30 radial rake angle diamond tools of 1.5mm radius with mineral oil coolant. Trial cuts were made at a depth of cut of 25 m feed rate of 4 m / rev and two spindle speeds (600 and 1200 rpm). The three force components were measured during all tests. The nished workpiece surfaces were measured with a grazing incidence interferometer (Tropel FlatMaster 200) to quantify the height of the trifoil aberration. The predictive model presented in the previous section is validated by the results shown in Figs. 8 and 9. A 1.2 N PV normal force was measured when the CaF2 workpiece was turned with a spindle speed of 600 rpm and 4 m / rev feed rate. A 120 nm peak to valley (PV) gure aberration is predicted by Eq. (4) with the measured loop stiffness of 10 N / m. The actual aberration was measured to be 130 nm. The results in Fig. 9 for a second workpiece turned at 1200 rpm show a predicted trifoil aberration of 80 nm PV and the actual 90 nm PV aberration. IV. CONCLUSIONS This article demonstrates that trifoil aberration is predictable using in-process force measurements taken while diamond turning (111) calcium uoride. The predictions agree with actual measurements within less than 13% error. This level of agreement allows a deterministic approach to process planning and optimization with predictable results. The model also shows how the trifoil aberration can be minimized using a higher stiffness machine tool. The predictable nature of the trifoil aberration makes it possible to design a cutting process capable of generating rough-turned optics of sufcient quality for nal polishing. ACKNOWLEDGMENTS The authors wish to thank George Kim and James A. Simonson of Edge Technologies for providing the diamond tools used in this testing and for suggesting the use of force measurement as a means of quantifying the trifoil aberration in the surface gure of (111) calcium uoride optics. The authors also thank Corning Tropel and QED for their assistance throughout the project.

FIG. 8. Normal cutting force and measured surface gure of a diamond turned CaF2 workpiece at 600 rpm and 4 m / rev feed rate. J. Vac. Sci. Technol. B, Vol. 23, No. 1, Jan/Feb 2005

89
1

Marsh et al.: Predicting surface gure in diamond

11

89 P. Blake and R. Scattergood, J. Am. Ceram. Soc. 73, 949 (1990). K. Blaedel, J. Taylor and C. Evans, S. Jahanmuir, M. Ramulu, and, P. Koshy editors: Machining of Ceramics and Composites, edited by S. Jahanmuir and M. Ramulu (Dekker, New York, 1999), pp. 139176. 13 J. Wilks and E. Wilks, Properties and Applications of Diamond (Butterworth-Heinemann, Oxford, 1991). 14 T. Bifano, T. Dow, and R. Scattergood, in Ref. 10. 15 P. Blake and R. Scattergood J. Am. Ceram. Soc. 73, 949 (1990). 16 J. Tlusty, Machine Dynamics, Handbook of High Speed Machining Technology, edited by R. I. King (Chapman and Hall, New York, 1995). 17 E. R. Marsh, D. S. Yantek, M. A. Davies, and D. E. Gilsinn, ASME J. Manuf. Sci. Eng. 120, 230 (1998). 18 E. R. Marsh and A. J. Schaut, ASPE J. Precis. Eng. 22, 252 (1998). 19 J. A. Patten and W. Gao, Precis. Eng. 25, 165 (2000). 20 W. Preus and O. Riemer, Single Point Diamond Machining, ASPE Annual Meeting Tutorial, 2003.
12

V. Liberman, T. M. Bloomstein, M. Rothschild, J. H. C. Sedlacek, R. S. Uttaro, A. K. Bates, C. V. Peski, and K. Orvek, J. Vac. Sci. Technol. B 17, 3273 (1999) 2 Handbook of Optical Materials, CRC, edited by M. J. Weber (Boca Raton, FL, 2003), Chap. 1. 3 T. M. Bloomstein, International SEMATECH Technology Transfer Report No. 01104175A-ENG, 2001. 4 S. Gliech, J. Steinert, and A. Duparr, Appl. Opt. 41 3224 (2002) 5 J. Yan, K. Syoji, and J. Tamaki, J. Vac. Sci. Technol. B 22, 46 (2004). 6 J. Yan, K. Syoji, T. Kuriyagawa, K. Tanaka, and H. Suzuki, ASPE Proc. 22, 66 (2000). 7 J. Yan, J. Tamaki, K. Syoji, and T. Kuriyagawa, Key Eng. Mater. 257258, 95 (2004). 8 J. Wang, R. L. Maier, and J. H. Bruning, Proc. SPIE 5188, 106 (2003). 9 C. Syn, J. Taylor, R. Donaldson, and S. Shimada, Lawrence Livermore National Laboratory Report No. 98100, 1988. 10 T. Bifano, T. Dow, and R. Scattergood, Proc. SPIE 966, 108 (1988).

JVST B - Microelectronics and Nanometer Structures