Ion polishing of copper: some observations

R. A. Hoffman, W. J. Lange, and W. J. Choyke

The change in absorptivity at 10.6 jtm of copper samples, at various stages of polishing including ion pol-ishing, is reported. The ion polishing apparatus is also described; it employs a low energy Xe ion beam in-cident at a glancing angle to maximize the removal rate while minimizing ion penetration and damage.The sample to be polished is rotated and the beam scanned across the surface to produce uniform removalof material. Starting with single crystal copper that had been mechanically polished and by using combi-nations of ion polishing and vacuum annealing, the absorptivity, determined by calorimetry, was reducedfrom 0.92% to 0.76%. Although optical figure and scatter were not monitored, SEM pictures show that thesurface features become smoother with ion polishing. A similar experiment was performed on a singlecrystal copper sample that had been electropolished, and the combination of ion polishing and annealinglowered the absorptivity from 0.97% to 0.89%.

1. Introduction

Mirrors for high power laser application must ofnecessity exhibit very low optical absorptivity at theappropriate wavelength(s). The mirrors must havethe proper optical figure, be smooth to minimize lightscatter, and have high thermal conductivity to pre-vent surface damage and undesirable mechanicalstresses. To obtain near-theoretical absorptivity thesurface must also be smooth and damage free.' Cer-tain metals that have the appropriate properties,such as copper and molybdenum, are candidates forlaser mirror substrates: However, conventional pol-ishing of metals usually damages the surface beingpolished; the depth of the damage depends upon thehardness of the material and the hardness and size ofthe polishing grit used. In addition, polishing gritmay be imbedded in or just below the surface. Theuse of electropolishing or etching frequently can beeffective except that a residue may be left on the sur-face from the chemical reaction. The combination ofdamage plus residue resulting from the polishingleads to surfaces with higher absorptivity, lower ther-mal conductivity, and lower damage threshold.

The work reported here was aimed at demonstrat-ing that ion polishing, if done properly, can be usedto remove material from surfaces in a controlledmanner, minimizing the damage to the surface andeliminating contamination. Ion polishing relies on

The authors are with Westinghouse Research Laboratories,Pittsburgh, Pennsylvania 15235.

Received 5 November 1974.

the sputtering of surface atoms by bombarding gasions; however, it is unlike conventional sputteringwhere the surface is bombarded by ions at near nor-mal incidence leaving the surface damaged androughened. With the ion polisher used here, ascanned and focused beam of low energy ions is inci-dent on the surface at a glancing angle. This maxi-mizes sputtering yield and should preferentially re-move high spots to yield a smoothened surface. Theion beam is programmed to move across the samplesurface, and the sample is simultaneously rotated inorder to avoid surface texturing and optimizesmoothing. The results to be presented give evi-dence for the premise that ion polishing of copper re-sults in surfaces that are smoother and have loweroptical absorptivity at 10.6 ,um.

II. Ion Polisher

The ion polishing equipment was built to provide alaboratory facility that could smooth and removecontrolled amounts of material from the surfaces offlat samples. Commercial ion polishers or millers donot provide the desired flexibility and control. Fig-ure 1 shows our apparatus in its present state of de-velopment. The principal design criteria were to re-move material by sputtering using an ion beam welldefined spatially, in energy, and angle of incidence soas to obtain reasonably high sputtering yields whileminimizing penetration and lattice damage producedby the incident ions. In addition, the arrangementpermits in situ evaporation of selected films on a pol-ished sample.

The ion source proper is of the hot cathode, arcdischarge type. A heavy tantalum filament (fourturns of 0.75-mm wire) is located near the center of a

August 1975 / Vol. 14, No. 8 / APPLIED OPTICS 1803

LN2 TrapDiffusion Pump -- l0 cm

Fig. 1. Ion polishing apparatus. The vacuum envelope, exceptfor the large glass cross on the right, is stainless steel at ground po-tential. The LN2-trapped, oil diffusion pump provides a basepressure measured by the ion gauge (BAG) of •5 X 10-3 Torr (7 X10-6 Pa). The system is inclined to the horizontal at 120, so thatthe sample is held on the rotatable holder by gravity alone. Elec-trons supplied by the filament F maintain a discharge in the anodeA into which gas G is admitted to maintain a pressure of -1 10-3 Torr, some 500 times that in the remainder of the system.Cylinders C1, C2, and C3 comprise the ion lens. Two pairs ofplates, D. and D, provide deflection of the beam in the x and y di-rections; for rastering, frequencies of 1000 Hz and 50 Hz, respec-tively, are used on Dx and Dy. Typical operating conditions usingXe as the sputtering gas are: filament at +3975 V; anode at +4000V; C at -2000 V; C2 at +2600 V; C3 at ground; target at ground;ion current 5 MA; no current through the solenoid, sample rotated

at 6 rpm. Removal rates for copper are 2.3 A/MA-min.

cylindrical anode (31.2-mm diam by 53 mm long).The filament leads enter the anode through boron ni-tride insulators whose tolerances are held to providereasonably tight seals. The gas inlet line for thesputtering gas is electrically isolated from the vacu-um envelope and enters the anode via a gas-tightseal. Thus the only exit for gas entering the source isthe 2-mm diam hole through which the ions are ex-tracted. The pressure in the source is estimated tobe 1 X 10-3 Torr (1.3 X 10-1 Pa) for this particulargeometry and associated vacuum system. The effectof an axial magnetic field on the source, provided bythe solenoid shown in Fig. 1, was found to be margin-al. Typical operating conditions are: filament heat-ing current, 26 A; filament to anode current 1 A; fila-ment to anode voltage, 25 V; no magnetic field.

Ions are extracted by a 6-kV potential drop acrossa 1.25-mm gap into the first element of a three-mem-ber, 2.54-cm diam, cylindrical, decelerating electro-static lens whose resultant focal length is 45 cm.Aperture stops, 6.25-mm diam, located in the firstlens cylinder and on the exit end of the third assurean ion beam of low divergence. Two orthogonal setsof deflection plates permit the ion beam to be posi-tioned and rastered on the sample by application ofsuitable dc and ac potentials (100 V). A phosphorscreen, mounted on a movable probe, permits conve-nient alignment and focusing of the beam. Thus theion spot and raster can be visually positioned on thesample and any spillover observed. The ion spot is

approximately 3 mm in diameter for a 5-gA beam.Attempts to obtain higher currents without increas-ing the spot size by using electrons to cancel spacecharge in the ion beam were unsuccessful.

The ion beam is incident, typically with 4 keV, onthe sample, which is inclined at an angle of 120 to thebeam. Using Xe ions under these conditions thepenetration depth of ions into the sample is estimat-ed from LSS theory2 to be -10 A. Other useful fea-tures incorporated in the apparatus include provisionfor thermal evaporation and thickness monitoring toenable in situ deposition of films on samples and anelectron source to provide charge neutrality whenworking with insulating samples.

Ill. Experimental ProcedureFor this study we chose to work with convenient,

laboratory-size samples, typically 1.27-cm diamdisks, in order to understand the effects of ion pol-ishing on surfaces rather than to produce operationalmirrors. Once the technique and effects are estab-lished, scaling up to work with the size of samplesthat are needed for laser mirrors should not be diffi-cult. Single-crystal,3 copper disks were selected forsamples because copper is used as a laser mirror ma-terial and yet is relatively difficult to polish by con-ventional means. The choice for using single crystalswas made to eliminate other effects and enable easierinterpretation of the results. Optically flat samplesurfaces were prepared using a combination of vari-ous techniques including mechanical polishing, elec-tropolishing, and ion polishing. At each stage in thesurface preparation, the optical absorptivity at 10.6Am was measured and the surface was examinedusing scanning electron microscopy. (It should benoted that were we attempting to prepare actual lasermirrors other measurements, including scatter andoptical figure, would be required. However, thethrust of the present work was to investigate the ef-fect of ion polishing on absorptivity at 10.6 gm undervery well defined and controllable conditions.)

Since only small changes in the sample reflectivitywith various methods of surface preparation are to beexpected, it was clear that measurement of absorptiv-ity should be done using optical calorimetry. Thecalorimeter is based on a thermal equilibrium tech-nique with substitutional electrical calibration and isfully described elsewhere.4 This apparatus is capa-ble of measuring reflecting surfaces having an ab-sorptivity of 0.01 at 10.6 Atm with a precision of <1.4%and a systematic error or accuracy of <1.6%.

IV. Results

A. Ion Polishing of Mechanically Polished CopperTable I gives the value of the optical absorptivity

at 10.6 ,m for a single crystal copper sample at eachstage of surface preparation starting with a surfacethat had been mechanically polished and followed bycombinations of ion polishing and vacuum annealing.Figure 2 shows scanning electron micrographs of the

1804 APPLIED OPTICS / Vol. 14, No. 8 / August 1975

Table I. Absorptivity of an Initially Mechanically PolishedSingle Crystal Copper Sample

Surface preparation (sequential) Absorptivity

Mechanical polish + light etch . 0.915%Vacuum anneal, 450 C, 1 h 0.812%Ion polish, 3600 A removed 0.959%Vacuum anneal, 4500C, 1 h 0.772%Ion polish, 12,000 A removed 0.879%Vacuum anneal, 4500 C, 1 h 0.762%

(a) a =0.915% (d) a = 0. 762%lpm

(b) a =0.8127% (c) a =0.772%

Fig. 2. Scanning electron micrographs of a single crystal coppersurface (a) after mechanical polishing, (b) after vacuum annealing,(c) after ion polishing and annealing, 3600 A removed, (d) after ion

polishing and annealing, 12,000 A removed.

surface along with values of the absorptivity. Initial-ly the copper sample, after mechanical polishing to aflatness <X/2 at 5876 A and light etching in dilute ni-tric acid (3%) to remove the oxide layer on the sur-face, had a value of 0.915% for the absorptivity,which is significantly lower than reported values formechanically polished copper.5 The scanning elec-tron micrograph of the surface [Fig. 2(a)] shows it tobe quite rough with pits and scratches. The samplewas next annealed in vacuum at 4500C for h andthe absorptivity decreased to 0.812%. Examinationof the surface [Fig. 2(b)] indicates some smoothing ofthe surface after annealing. Note the annealing tem-perature is considerably lower than the standard an-nealing temperature for copper, about 8000C; how-ever, the lower temperature appears to give the sur-face atoms enough mobility to cause significant sur-

face diffusion. Next, the surface of the sample wasion polished, removing about 3600 A, and surprising-ly the absorptivity increased to 0.959%. The samplewas then subjected to a vacuum anneal at 4500C, andthe absorptivity decreased to 0.772% [see Fig. 2(c)], avalue that is significantly lower than that before ionpolishing. No discernible differences were found be-tween the scanning micrographs of the surface afterion polishing and after the second annealing; only apicture of the surface after both ion polishing and an-nealing is included. Examination of Fig. 2(c) showsthat the surface is smoother with more rounded fea-tures. Also seen, are white spots on the surface thatwere not present in the previous photos. We believethe white spots are dielectric particles of polishinggrit, charging up and appearing white in the SEM,which were ibedded in the surface and are now un-covered by the ion polishing. The polishing grit hasa lower sputtering yield than copper and would re-main as the surrounding copper was sputtered away,and, in fact, it looks as if some of the white spots aresitting on the tops of mountains of copper. We re-peated the cycle of ion polishing, removing 12,000 A,and vacuum annealing and found that the absorptivi-ty after annealing had decreased slightly to a value of0.762%. The scanning micrograph of the surfaceafter the second ion polishing and annealing, [Fig.2(d)] indicates that the surface is now much smooth-er than before with no white spots visible. Appar-ently, enough copper was removed to get below theregion of trapped polishing grit. The presence ofpolishing grit in the surface of a laser mirror or win-dow could lower the laser damage threshold consider-ably,6 and we believe that it is essential that it be re-moved if durable laser components are to be pro-duced.

Referring again to Table I, it is seen that afterstarting with a mechanically polished sample havingan absorption of 0.915%, which is already lower thanpreviously reported values, the sequential operationsof ion polishing and vacuum annealing further de-creased the absorptivity to 0.762%. Thus, we con-clude that the combination of ion polishing and vacu-um annealing can remove damaged surface layersalong with imbedded polishing grit and result in sur-faces that are smoother and have lower absorptivity.

B. Ion Polishing of Electropolished Copper

Values for the absorptivity, at 10.6 m, of a singlecrystal copper sample with various surface prepara-tions are shown in Table II. The sample that was

Table II. Absorptivity of a Mechanically Polished andSubsequently Electropolished Single Crystal of Copper

Surface preparation (sequential) AbsorptivityMechanical polish + electropolish, 0.970%

50 m removedVacuum anneal, 450 C, 1 h 0.992%Ion polish, 4500 A removed 1.026%Vacuum anneal, 4500 C, 1 h 0.893%

August 1975 / Vol. 14, No. 8 / APPLIED OPTICS 1805

(a) a = 0. 970% 2 ' (bI a = 0. 893%20 i'm

Fig. 3. Scanning electron micrographs of a single crystal coppersurface (a) after electropolishing, (b) after ion polishing and vacu-

um annealing, 4500 A removed.

mechanically polished and then electropolished inphosphoric acid,7 removing about 50 m, exhibitedan absorptivity of 0.970%. The surface appeared vi-sually to be bright but wavy. Subjecting the sampleto vacuum annealing, at 450 C for 1 h, increased itsabsorptivity to 0.992%. The sample was next ionpolished, removing about 4500 A, and the absorptivi-ty further increased to 1.026%. The absorptivitythen decreased to 0.893%, significantly below that be-fore ion polishing, when the sample was once againannealed. Figure 3 shows scanning electron micro-graphs of the sample surface at two stages in its prep-aration. Examination of Fig. 3(a), taken after elec-tropolishing, reveals the presence of a residue fromthe electropolishing, and moving the sample in themicroscope showed the residue to cover a large frac-tion of the surface. The appearance of the surface asviewed in the SEM did not change with annealing.Figure 3(b) shows the final surface, after ion pol-ishing and annealing. From this picture one can seethat the residue from the electropolishing is com-pletely removed by the ion polishing. In fact, no res-idue was found when the sample was moved in theSEM to examine all areas of the surface. Scanningelectron micrographs at higher magnification showedportions of the electropolished surface to be verysmooth, and the ion polishing, in which 4500 A wereremoved, did not roughen this surface. We concludethat the combination of ion polishing and vacuumannealing is capable of removing contamination fromsurfaces and significantly lowering the optical ab-sorptivity.

C. Effects of Annealing

The relatively large increase in absorptivity afterion polishing and its subsequent dramatic decreasewith annealing, as seen in the data of Table I, are

puzzling. Since the effects of ion polishing should, atmost, extend 10-20 A below the surface, it is difficultto imagine how imbedded Xe ions or damage causedby them can have such a significant influence on theabsorptivity. One must consider that this damagedepth is only a small fraction of the optical penetra-tion depth, and the surface scattering of the conduc-tion electrons is presumed to be totally diffuse.",7

In order to gain further insight into what happensduring the annealing process, an ion polished samplewas heated in a vacuum system equipped with a massspectrometer. Significant amounts of Xe were foundto be released when the sample was heated to only2000C. Further heating to higher temperature re-sulted in negligibly more Xe release. This is taken asan indication that the Xe was trapped very close tothe surface in agreement with theoretical estimates.The observation that the trapped Xe is released byheating to only 2000C allows us to distinguish be-tween the effect of trapped Xe and the effect of theXe induced damage as the cause for the increase inabsorptivity with ion polishing. Consequently, thefollowing experiment was performed: an ion pol-ished sample was annealed at low temperature,2000C for 1 h, to effect the release of the trapped Xe.The value of the absorptivity was measured andfound not to change as a result of the low tempera-ture annealing. Since the absorptivity did changewith 4500C annealing, we conclude that the damagecaused by the bombarding Xe ions and not the im-bedded Xe atoms themselves is responsible for theincrease in the absorptivity after ion polishing. Sig-nificantly, this damage can be annealed out by heat-ing to 4500C for 1 h, as evidenced by the measureddecrease in absorptivity.

V. Conclusions

It is shown that ion polishing of single crystal cop-per samples is capable of removing damage layers ofmaterial and contamination from surfaces that hadbeen prepared by other polishing methods, and thatthe resulting surfaces are smoother than those pro-duced by the initial polishing. Furthermore, thecombination of ion polishing and 4500C vacuum an-nealing yields values of absorptivity that are signifi-cantly lower than those before ion polishing.

We believe that the combination of lower valuesfor the absorptivity, the elimination of polishing gritimbedded in the surface, and the further removal ofthe damage layer will yield laser mirrors with signifi-cantly increased damage thresholds. Although thepresent work was confined to laboratory size samplesof a single crystallographic orientation, there is noapparent difficulty in scaling for ion polishing of

1806 APPLIED OPTICS / Vol. 14, No. 8 / August 1975

large laser mirrors and windows. Automatic elec-tronic control of the ion polishing should permit han-dling of more complicated optical shapes. The com-bination of grazing incidence, low ion energy, smalland rastered ion beam, and target rotation makes thepresent approach different than the usual ion pol-ishing and hopefully can even be effective for poly-crystalline material. Finally, the combination of ionpolishing and deposition of overcoating films in situshould produce a very clean film-substrate interfacewith improved bonding. This could be of particularsignificance for coated alkali-halide laser windowsthat presently exhibit film adhesion problems.

References1. H. E. Bennett, J. M. Bennett, E. J. Ashley, and R. J. Motyka,

Phys. Rev. 165, 755 (1968).2. J. Lindhard, M. Scharff, and H. E. Schiott, Mat. Fys., Medd.

Dan. Vid. Selsk. 33 (1963).3. Materials Research Corporation, Orangeburg, N.Y. The sam-

ple surfaces had an orientation such that the normal to the sur-face was 15° from the [111] toward the [3211.

4. R. A. Hoffman, Appl. Opt. 13, 1405 (1974).5. S. Holmes, A. Klugman, and P. Kraatz, Appl. Opt. 12, 1743

(1973).6. N. Bloembergen, Appl. Opt. 12, 661 (1973).7. J. A. McKay, doctoral dissertation, Carnegie-Mellon University

(1974) and Phys. Lett. 47A, 385 (1974).

Boris P. StoicheffUniversity of TorontoOSA President Elect

at1975 OSA Anaheim Meetings

Photo A. L. Schawlow

August 1975 / Vol. 14, No. 8 / APPLIED OPTICS 1807

��o