optimization of coating uniformity in an ion beam

Optimization of coating uniformity in an ion beamsputtering system using a modified

planetary rotation method

Mark Gross, Svetlana Dligatch,* and Anatoli ChtanovCSIRO Division of Materials Science and Engineering, P.O. Box 218, Lindfield, NSW 2070, Australia

*Corresponding author: [email protected]

Received 30 July 2010; revised 13 November 2010; accepted 24 November 2010;posted 15 December 2010 (Doc. ID 132611); published 0 MONTH 0000

A modified planetary rotation system has been developed to obtain high uniformity optical coatings onlarge substrates in an ion beam sputter coater. The system allows the normally fixed sun gear to rotate,thus allowing an extra degree of freedom and permittingmore complexmotions to be used. Bymoving thesubstrate platen between two fixed positions around the sun axis, averaging of the distributions at thesetwo positions takes place and improved uniformity can be achieved. A peak-to-valley radial uniformity of∼0:15% (∼0:07% rms) on a single layer film on a 400mm diameter substrate has been achieved withoutthe aid of masking. © 2011 Optical Society of AmericaOCIS codes: 310.1860, 310.3840.

1. Introduction

The design of multilayer optical coatings that meetalmost any specification is now a routine process[1–3], but the actual manufacture of some designsis sometimes not so straightforward. The thicknessand refractive index tolerances on many optical mul-tilayer designs can be very demanding and requireprecise deposition control and monitoring [4,5].Apart from the common problems associated withprocess control and layer thickness monitoring, par-ticularly for coatings with small error tolerances,large substrates add another difficulty in that thenonuniformity of coating thickness and refractive in-dex may exceed the error tolerance of the design.

The usual methods for improving the thicknessuniformity in optical coating systems invariably in-clude some kind of simple or planetary substraterotation together with moving or stationary masksto modify the deposition plume profile. Such ap-proaches generally work well [6–9], even on curvedsurfaces, but with some limitations. The masks need

accurate fabrication, precise placement in the cham-ber, and can be sources of particulate contaminationif placed very close to the substrate surface. For pro-cesses such as ion beam sputtering, where the sput-ter plume distribution varies considerably withdifferent materials, several mask sets may need tobe exchanged between layers or, alternatively, com-pletely separate and optimized sources used for eachmaterial. The masks may also need to be frequentlymodified as the source is eroded and its plume distri-bution changes.

This work was undertaken to enable the fabrica-tion of multilayer optical coatings on large diameter,high specification optics in an ion beam sputter sys-tem. An example of the kind of coating envisaged forthe system is an antireflection (AR) coating with areflectance <0:002% at 1064nm over a diameter lar-ger than 250mm. To achieve this specification with asimple two-layer AR coating would require a peak-to-valley uniformity of optical thickness (product of re-fractive index and physical thickness) better than∼0:5%, and this is ignoring all other sources of error.In order to produce such coatings, a modified plane-tary substrate mount has been constructed and fittedto a commercial ion beam coating system specifically

1

2

0003-6935/11/0900C1-01$15.00/0© 2011 Optical Society of America

20 March 2011 / Vol. 50, No. 9 / APPLIED OPTICS C1

to handle single, large substrates. The principal ad-vantage of the system is that no masks are used andthe setup can readily be adjusted to account for dif-ferent or changing source plume distributions. Singlelayer radial uniformities of ∼0:15% peak-to-valley(p–v, ∼0:07% rms) have been achieved on 400mmdiameter substrates.

2. Planetary System Design

A conventional planetary system usually consists ofa central fixed “sun” gear about which are driven oneor more “planet” gears to which are attached the sub-strate mounts (Fig. 1).

Any given point on the substrate thus rotatesabout the central axis of the system in a cycloidalpath that exposes that point to many different partsof the source plume and results in some averaging ofthe film thickness. Careful placement of the sourceand choice of the sun-to-planet gear ratio to limitpath retracing can produce uniformities at the per-cent level, depending on the substrate size. In thepresent design (Fig. 2), the fixed sun gear is replacedwith one that can rotate on its axis and is driven by aservo motor separately from the planet drive.

This essentially decouples the sun and planet mo-tions, which can now be independently controlled.When the sun drive is off, the system performs con-ventional planetary motion. With the planet driveoff, the planet can be driven by the second motor viathe rotating sun gear. This allows conventional sim-ple rotation with the additional feature that the cen-ter of rotation of the planet can be set by stopping theplanet drive at any angle.

3. Uniformity Simulations

The range of motions possible in such a system islarge. A software model was therefore developed toenable the coating uniformity obtained with variousmotion options to be more rapidly estimated. In de-veloping any such model, an accurate knowledge ofthe two-dimensional source plume distribution isneeded. The plume profile from an ion beam sputtertarget is complex and is influenced by many factors,including the incident ion beam profile, the targetmaterial, variations in target ion beam angle, the an-gle of the substrate to the target and the degree oftarget erosion. Rather than undertake a time con-suming discrete thickness mapping process over a

large area, an empirical description of the plume wasdeveloped so that an adjustable model of arbitraryresolution could be obtained. The two-dimensionalplume shape was approximated as a series of semi-concentric distorted ellipses of constant thickness(Fig. 3).

A skewed Gaussian line shape was then applied tothe horizontal axis of this distribution and the geo-metrical parameters were adjusted until the thick-nesses fitted a measured set of data. Figure 4 showsa photograph of the interference pattern formed by a∼1 μm maximum thickness tantala film deposited ona 600mm × 800mm stationary glass plate from atantalum sputter target in the current coating sys-tem. The agreement between the vertical and hori-zontal line thickness profiles measured on thispattern and the model fit is sufficient to allow rea-sonably accurate predictions to be made (Fig. 5).

With the plume profile determined and stored asa two-dimensional computer-based array, a simple al-gorithm can be developed to calculate the film thick-ness on the substrate for any position of the substratein the plume (Fig. 6). The substrate may then be in-crementally displacedaccording to thedesiredmotionand the thicknesses at each increment summed togive the total thickness. Figure 7 shows the experi-mental and modeled normalized uniformity of a

Fig. 1. Basic concept of a conventional planetary rotation mount.The substrate platen is attached to the planet gear.

Fig. 2. Modified planetary with driven sun gear. Servo motorsenable both gears to move independently.

Fig. 3. (Color online) The two-dimensional plume shape approx-imation: calculated rings of constant thickness superimposed on afalse color image of the source distribution. 4/CO

3

C2 APPLIED OPTICS / Vol. 50, No. 9 / 20 March 2011

tantala film deposited on a 200mm long silica glassstrip positioned along a radius of the rotating sub-strate planet. The planet center was located at var-ious fixed angles of the planet drive axis to thehorizontal axis. Here, the planetary system is verti-cally mounted and the horizontal axis is defined asthe line through the center of the sungearand the cen-ter of the plume distribution. The agreement betweenthe model and experimental data is quite good andcan be improved with finer adjustment of the modelparameters.

An important observation from these results is thatthere is an angle (in this case between 50° and 70°) atwhich the slope of the radial variation changes sign.For lower angles the deposition is thickest in themid-

dle of the substrate, while for higher angles it is thin-nest. This suggests the possibility of balancing thedeposition thickness over the entire substratebymov-ing the substrate between these two areas in somesuitable manner. While there may be several ap-proaches to do this, the obvious one is to find two an-gles where the radial thickness distributions are asclose to complementary as possible (given a constantscale factor) and then depositing the coatingwhile therotating substrate ismoved between these two anglesand allowed to dwell at each angle for a time deter-mined by themaximumdeposition rate at each angle.Themodel canbeused to estimate suitable angles anddwell times, as a starting point for subsequent experi-mental optimization. Figure 8 shows the simulatednormalized deposition uniformity predicted by coat-ing a rotating substrate first at 3° then at 77° tothe horizontal for fixed times in the ratio 1∶21. Thepredicted net uniformity in this case is <0:2% p–v.

Fig. 4. (Color online) Photograph of the interference pattern of an∼1 μm thick tantala film deposited on a stationary glass plate.4/CO

Fig. 5. Horizontal and vertical line thickness measurements(symbols) through the film thickness maximum, together withmodel profiles determined by adjusting the parameters to fit thesepoints.

Fig. 6. Fixed angle positions of the rotating planet to thehorizontal axis of the plume center.

Fig. 7. (Color online) Measured and calculated uniformity over a200mm radius of a simple rotating substrate located at fixed an-gles to the horizontal axis of the source plume. The symbols repre-sent measured values, while the solid curves are the modelpredictions. 4/CO


4. Practical Application of the Method

The best experimental result achieved so far is pre-sented in Fig. 9, where a measured radial uniformityof ∼0:15% p–v has been obtained. Here, an ∼250nmthick tantala film was deposited on a silica stripmounted along a radius of the 400mm diameter sub-strate planet. Initially, the angles and times sug-gested by the model were used and then a trial anderror approach used to fine-tune these values to op-timize the result. In practice, this involved only fouriterations and the optimum angles and times dif-fered from the model by less than 5%. Spectrophoto-metric measurements of the transmission throughthe strip were made at points along the strip usinglong integration times and constant rechecking ofthe spectrophotometer baseline to minimize signaldrift and obtain the most precise measurements pos-sible from the instrument. The thickness of the filmwas extracted from the transmission spectra by op-tical modeling using a fixed refractive index. Themeasured thickness is thus more correctly the opticalthickness; however, this is the most important figurefor the design of optical coatings.

While the result in Fig. 9 is very encouraging,there are a number of factors that will limit the max-imum uniformity achievable in practice. One obviouslimit is the extent to which the optimum high andlow angle distributions for any particular materialcomplement each other. For example, in the case ofthe tantala coating above, potentially better unifor-mities (∼0:05% p–v) can be achieved, but over smal-ler diameters where the distributions are moreclosely matched. The uniformity obtained in practicewill also be limited by additional effects, includingthe accuracy of the stage positioning mechanism,the finite transition time between the upper and low-er angles, and the accuracy of the thickness monitor-ing system. Stage positioning and transit time effectscan, however, be modeled. As shown in Fig. 10, stagepositioning errors of�1°, which are larger than thoseactually achieved with the current mechanism, leadto modeled uniformities still within 0:3% p–v (thelack of smoothness in the traces can be attributedto the limited resolution of the model). The best mod-eled uniformity of 0:1% p–v in Fig. 10 assumes a 2 stransit time between the high and low angles, whichis limited only by the mass of the substrate and thepower of the motor drive. In fact, the transit timecan be eliminated entirely if the deposition is shut-tered during the movement between the two angles.As yet it has not been determined if a number ofoscillations between the high and low angles or a sin-gle deposition at each angle gives better results, par-ticularly for uniformity of refractive index, althoughthe latter case has advantages from the mechanicalperspective.

The results described here are limited by themeasurement method to radial nonuniformity.Azimuthal nonuniformity has not been measured di-rectly. The sources of azimuthal nonuniformity in-clude too slow passage of the substrate throughareas of high nonuniformity in the plume, as wellas wobble in the plane of rotation of the coated faceof the substrate. In the latter case, this can be mini-mized by careful design of the substrate jigging. Inthe former case, the nonuniformity can be minimizedif the substrate rotational velocity is high compared

Fig. 8. (Color online) Modeled thickness uniformity obtained byadding the deposition profiles obtained at 3° substrate angle tothat obtained at 77° in the ratio 1∶21 (by maximum thickness).4/CO

Fig. 9. Measured thickness uniformity of an ∼250nm thick tan-tala film on a silica substrate over a 200mm radius of a rotatingplanet oscillating between optimized fixed angles to the horizontalaxis of the source plume. The uniformity is ∼0:15% p–v.

Fig. 10. (Color online) Optimized Hi and Lo angles for best mod-eled thickness uniformity (including effect of expected transitiontime. The uniformity is ∼0:1% p–v. 4/CO


to the azimuthal variation in deposition rate. In thepresent system, the substrates essentially only ro-tate about their own axis and have no significantmovement, if any, azimuthally through the deposi-tion plume. The rotational velocity of the substrateis not limited by a gearing ratio or the mass or bal-ance of the whole planetary mechanism. In the cur-rent coating system, the substrate planet can executemore than 30 revolutions per nanometer of filmthickness, leading to very high averaging of any azi-muthal variations. In the result reported in Fig. 9,although the measurement has been made along aradius of the substrate planet, several measure-ments of the same process using random starting po-sitions has not resulted in any variation outside themeasurement error bars, suggesting the uniformityover the entire substrate is predominantly radialin character.

As a final point, the nontrivial question ofmeasure-ment of uniformities at this level must be considered.Typical methods, such as the spectrophotometrictransmissionmeasurements used here, often have in-strumental errors at the ∼0:1% level. In addition, thethickness of the layer(s) must then be extracted frommodels that can have a number of fitting parameters,some of which may have little or no sensitivity at thislevel, leading to limited precision in determination ofthe thickness.Ultimately, highuniformitiesmayhaveto be inferred by, for example,measurement of the op-tical performance of thickness-sensitive multilayercoatings.

5. Conclusions

A modified planetary rotation system has been con-structed and fitted to a commercial ion beam sputter-ing system. The system allows the normally fixed sungear to rotate, thus allowing an extra degree of free-dom and permitting more complex motions to be

used. By moving the substrate planet between twofixed angles to the horizontal axis, averaging ofthe distributions at these two angles takes placeand improved uniformity can be achieved. At pre-sent, a single layer radial uniformity of ∼0:15% p–v (0:07% rms) on a 400mm diameter substrate hasbeen achieved without the aid of masking. Althoughresults presented here are for tantala films, the sameprocess can be applied in principle to any material.Multilayer coatings with high uniformity on alllayers can thus, in principle, be achieved.

References1. H. A. Macleod, “The Essential Macleod,” software by Thin

Film Center Inc..2. TFCalc by Software Spectra, Inc., 14025 N. W. Harvest Lane,

Portland, Ore. 97229, USA.3. A. V. Tikhonravov and M. K. Trubetskov, “OptiLayer Thin

Film” (OptiLayer Consulting Ltd., software copyright 2010).4. B. T. Sullivan, G. A. Clarke, T. Akiyama, N. Osborne, M.

Ranger, J. A. Dobrowolski, and L. Howe, “High-rate auto-mated deposition system for the manufacture of complexmultilayer coatings,” Appl. Opt. 39, 157–167 (2000).

5. B. Badoil, F. Lemarchand, M. Cathelinaud, and M. Lequime,“Interest of broadband optical monitoring for thin film filtermanufacturing,” Appl. Opt. 46 (20), 4294–4303 (2007).

6. J. B. Oliver and D. Talbot, “Optimization of deposition unifor-mity for large-aperture National Ignition Facility substratesin a planetary rotation system,” Appl. Opt. 45, 3097–3105(2006).

7. B. Sassolas, R. Flaminio, J. Franc, C. Michel, J.-L. Montorio, N.Morgado, and L. Pinard, “Masking technique for coating thick-ness control on large and strongly curved aspherical optics,”Appl. Opt. 48, 3760–3765 (2009).

8. G. I. Abzalova, R. S. Sabirov, and A. V. Mikhailov, “Depositinguniform-thickness coatings on large surfaces by means of elec-tron-beam evaporation in vacuum,” Opt. Zh. 72, 76–78 (2005).

9. F. Villa, A. Martínez, and L. E. Regalado, “Correction masksfor thickness uniformity in large-area thin films,” Appl. Opt.39, 1602–1610 (2000).

4


Queries

1. Is there a definition of CSIRO that we can provide?2. Is antireflection correct for AR?3. Is p–v correct for peak-to-valley (p–v) was introduced later without definition)?4. This query was generated by an automatic reference checking system. Reference [8] could not be located in

the databases used by the system. While the reference may be correct, we ask that you check it so we canprovide as many links to the referenced articles as possible.


optimization of coating uniformity in an ion beam

Documents