Interhalogen plasma chemistries for dry etch patterning of Ni, Fe, NiFe and NiFeCo thin films

Transcription

1 Ž. Applied Surface Science Interhalogen plasma chemistries for dry etch patterning of Ni, Fe, NiFe and NiFeCo thin films H. Cho a,), K.B. Jung a, D.C. Hays a, Y.B. Hahn a,1, E.S. Lambers a, T. Feng a, Y.D. Park a, J.R. Childress b, S.J. Pearton a a Department of Materials Science and Engineering, UniÕersity of Florida, 132 Rhines Hall, PO Box , GainesÕille, FL 32611, USA b IBM Almaden Research Center, San Jose, CA 95120, USA Received 28 August 1998; accepted 1 October 1998 Abstract IClrAr and IBrrAr plasmas operated in an inductively coupled plasma Ž ICP. source have been examined for dry etching of Ni, Fe, NiFe and NiFeCo. The removal of the Fe etch products limits the etch rates under most conditions, but rates of y1 ; 500 A min are obtained for both NiFe and NiFeCo in both chemistries. The etched surfaces are smooth Žatomic force microscopy root-mean-square roughness - 1 nm. over a broad range of plasma conditions, with small residual halogen concentrations Ž F2 at.%.. q 1999 Elsevier Science B.V. All rights reserved. Keywords: Interhalogen; Dry etch; Thin film 1. Introduction The dry etching of magnetic multilayer structures represents a challenge because of the relative inw1 8 x. For this reason, volatility of the etch products most of the patterning for magnetic sensors, nonvolatile memory elements, and readrwrite heads are performed using ion beam milling to physically sputw2 7 x. High ion energies Ž ;1 kv. during this process have been found to lower the ter the material coercivity of magnetic elements by up to a factor of eight, probably due to creation of magnetic dead- ) Corresponding author. Fax: q Present address. Department of Chemical Engineering and Technology, Chonbuk National University, Chonju , South Korea. layers on the exposed sidewalls wx 9. Etch processes with a chemical component, in addition to purely physical sputtering, should have a number of advantages, including higher etch rates, better selectivity to mask materials, lower ion energies and reduced redew10 x. One method of position on feature sidewalls enhancing the etch product volatility in plasma etching is to heat the sample during the process Žgener- ally C would be required. wx 1, but in giant magneto-resistive Ž GMR. multilayers, the component layers may only be A thick and there is only a very limited thermal budget available before interdifwx 2. fusion occurs Another method for removing the etch products is by providing a high ion flux incident simultaneously with the reactive neutral flux. This provides impetus for ion-assisted desorption of the etch products. Ex r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved. Ž. PII: S

2 216 ( ) H. Cho et al.rapplied Surface Science perimentally, we have found that ion-neutral ratios G0.02 are necessary for achieving high etch rates of NiFe and NiFeCo in Cl 2rAr plasma chemistries w11 13 x. These flux ratios are only available in high density plasma sources, such as inductively coupled plasmas Ž ICP. or electron cyclotron resonance Ž ECR. microwave plasmas. In conventional reactive ion etch tools, the ion-neutral flux ratio is typically in the 10 y5 10 y6 range, and the absence of a strong ionassisted desorption contribution leads to the build-up of a selvedge or reaction layer of chlorinated etch products on the sample surface. While the Cl rar plasma chemistry operated un- 2 der high density conditions produces effective etching of magnetic materials, there are other mixtures of interest. In particular, the interhalogens, ICl and IBr, have been found to dissociate readily in high density Fig. 2. Etch rates of Ni, Fe, NiFe and NiFeCo in 250 W rf chuck power, 5 mtorr discharges of 2 IClr13 Ar Ž top. or 2 IBrr13 Ar Ž bottom., as a function of source power. plasma sources, producing high concentrations of reactive species w14 x. In this paper, we report a parametric investigation of the effect of plasma conditions on etch rates, surface morphology and surface composition of Ni, Fe, Ni 0.8Fe0.2and Ni 0.8Fe0.13- Co in ICP discharges of IClrAr and IBrrAr Experimental Fig. 1. Etch rates of Ni, Fe, NiFe and NiFeCo in 750 W source power, 250 W rf chuck power, 5 mtorr discharges of IClrAr Ž top. or IBrrAr Ž bottom., as a function of plasma composition. Direct current Ž DC. magnetron sputtering was used to deposit A thick layers of each of the materials on Si substrates. For etch rate experiment, samples were masked with Apiezon wax, which was removed after the plasma exposure and the etch step measured by profilometry. Etching was performed in a Plasma Therm 790 system, involving an ICP source with a three-turn coil antenna operat-

3 ( ) H. Cho et al.rapplied Surface Science Results and discussion The influence of plasma composition on Ni, Fe, NiFe and NiFeCo etch rates in ICP IClrAr Ž top. and IBrrAr Ž bottom. discharges at fixed source power Ž 750 W., rf chuck power Ž 250 W. and pressure Ž5 mtorr. is shown in Fig. 1. For the IClrAr plasma chemistry, the rates for NiFe, NiFeCo and Ni initially increase as ICl is added, but then decrease beyond particular discharge compositions. This is consistent with a mechanism in which the adsorbed reactive neutral flux must be balanced with the ionw11 13 x. Beyond the optimum discharge compositions, assisted removal of the resultant etch products we believe there is blocking of the surface to ion bombardment by the high chlorine and iodine concentrations. The rate-limiting step appears to be re- Fig. 3. Etch rates of Ni, Fe, NiFe and NiFeCo in 750 W source power, 5 mtorr discharges of 2 IClr13 Ar Ž top. or 2 IBrr13 Ar Ž bottom., as a function of rf chuck power. ing at 2 MHz and powers up to 1000 W. The samples are thermally bonded to a radio frequency Ž rf. -powered Ž MHz, W., He backsidecooled chuck. Process pressure was varied from 5 to 20 mtorr, with a gas load of 15 standard cubic centimeters per minute Ž sccm.. ICl and IBr are crystalline solids with melting temperatures of 27 and 458C, respectively, and were contained in a stainless steel vacuum vessel heated to ;508C to enhance the vapor pressure w14 x. The resulting gases were injected directly into the ICP source through electronic mass controllers. Electronic grade Ar was always added to provide a strong physical component to the etching. The etched surfaces were examined by atomic force microscopy Ž AFM. and Auger electron spectroscopy Ž AES. to look at morphology and near-surface composition, respectively. Ž. Ž. Fig. 4. Etch rates top and etch yields bottom of Ni, Fe, NiFe and NiFeCo in 2 IClr13 Ar, 750 W source power, 250 W rf chuck power discharges, as a function of process pressure.

4 218 ( ) H. Cho et al.rapplied Surface Science moval of the FeCl x and FeI x etch products, based on the low etch rate of the Fe layers. Indeed, for very high ICl percentages, there is net deposition on the Fe due to the inability to effectively remove the chloride etch products. Note also that as the ICl percentage in the discharge increases, the chuck Fig. 5. The AFM scans of NiFe after etching in 750 W source power, 250 W rf chuck power, 5 mtorr discharges, as a function of plasma composition.

5 ( ) H. Cho et al.rapplied Surface Science self-bias also increases. This indicates that the posiw15 x, as tive ion density in the plasma is decreasing expected since both chlorine and iodine are electronegative gases. The results for IBrrAr discharges are shown at the bottom of Fig. 1. The etch rate behavior for Ni is similar to that with IClrAr, but the Fe etches much more rapidly in IBrrAr discharges, especially at Fig. 6. The AFM scans of NiFeCo after etching in 750 W source power, 250 W rf chuck power, 5 mtorr discharges, as a function of plasma composition.

6 220 ( ) H. Cho et al.rapplied Surface Science high halogen concentrations. This is due to the higher volatility of the FeBrx etch products relative to FeCl x w12 x, and is not a strong function of bias. This effect leads to a small increase in NiFe and NiFeCo etch rate at IBr percentages beyond ; 60%. The effect of ICP source power on the material etch rates is shown for IClrAr Ž top. and IBrrAr Ž bottom. in Fig. 2. For the IClrAr, the etch rates for Ni, NiFeCo and NiFe increase monotonically with increasing ion flux, even though the self-bias decreases because of the larger conductivity of the plasma. For Fe, there is essentially no etching until source powers ) 600 W, which illustrates the point that balancing the ion and reactive neutral fluxes can lead to a positive etch rate w11 x. The behavior of NiFe and NiFeCo in IBrrAr discharges is basically similar to that with IClrAr. The Ni and Fe etch rates go in different directions at high flux, due to the ion energy falling below that needed to efficiently desorb NiBrx and NiI x. The low rates may also be attributed to the chemical kinetics of the reaction. Under high flux conditions, the reactive species may sputter off the surface prior to reaction. The dependence of material etch rates on rf chuck power is shown in Fig. 3 Ž top. for 2 IClr13 Ar and Fig. 3 Ž bottom. for 2 IBrr13 Ar discharges at fixed source power Ž 750 W. and pressure Ž 5 mtorr.. For both chemistries, the etch rates Ž except those for Ni. are basically linearly dependent on chuck power, indicative of a desorption-limited process. For Ni in both chemistries and Fe in IBrrAr, the rates initially increase as the rf chuck power Žand hence DC self-bias is increased., but then decrease beyond particular maxima. This is often observed in highdensity plasma etching of materials, and is usually ascribed to desorption of the adsorbed chlorine neutrals before they can react with the surface of the metal w10 x. The reaction rate is presumably different on the alloys, where this trend is not observed up to our maximum of chuck power. Fig. 4 shows the pressure dependence of material etch rates in 2 IClr13 Ar discharges Ž750 W source power, 250 W rf chuck power.. We were not able to produce stable IBrrAr discharges at pressures above 5 mtorr. Even though DC self-bias increases with pressure, the etch rates of all of the materials decrease with increasing pressure. We suspect that the ionrneutral ratio falls below that necessary for effective balance of the product formation and desorption. Once again, the rate-limiting step is removal of the Fe. The etch yields and ion fluxes calculated from the etch rate and DC self-bias on the chuck are shown at the bottom of the figure w16 x. The low etch yields show why high density plasma conditions are needed to produce practical etch rates for the magnetic materials. The surfaces of the NiFe and NiFeCo were smooth over a broad range of plasma conditions. Fig. 5 shows AFM scans from NiFe samples after etching of ;2000 A of material in IBrrAr discharges Ž750 Fig. 7. The AES surface scans of NiFe after etching in either IClrAr Ž top. or IBrrAr Ž center and bottom. discharges Ž750 W source power, 250 W rf chuck power, 5 mtorr., as a function of plasma composition.

7 ( ) H. Cho et al.rapplied Surface Science W source power, 250 W rf chuck power, 5 mtorr. at different gas compositions. The as-grown rootmean-square Ž RMS. roughness is ; 0.55 nm. At low IBr compositions, the surface is significantly rougher Ž 1.8 nm RMS., but as the chemical component of the etching increases, the surfaces are as good or slightly better than the control value. A similar trend was observed with NiFeCo layers, as shown in Fig. 6. The main difference is that even for the low IBr concentration, the RMS roughness is still as good as the control value. These data show that there is a wide process window for maintaining high quality surfaces with the interhalogen plasma chemistries. The AES data showed that the samples retained their initial stoichiometry under these conditions. The surfaces were also relatively clean after etching. Fig. 7 shows AES surface scans of NiFe after either IClrAr Ž top. or IBrrAr etching Žcenter and bottom. at different plasma compositions. We observe adventitious carbon and a native oxide originating from exposure to ambient during transfer from etch chamber to analysis chamber. There is only a slightly amount of residual chlorine detected on the ICl etched material Ž F 1 at.%., which is consistent with the mechanism involving efficient desorption of the etch products by the attendant ion flux. There was no Br Žmain Auger transition at 1396 ev. detected on any of the samples, while any I signal would be swamped by that due to oxygen Ž iodine transition at 511 ev.. Similar data were obtained for etched NiFeCo samples. Once again, the surfaces were relatively clean, with residual bromine concentrations below the detection limit of AES w11 13 x. We have previously reported that use of in-situ H 2 plasma cleaning is effective in volatilizing halogen residues, producw13 x. More ing clean surfaces on the etched field work needs to be done to establish the chemical state of the sidewalls of etched features, since this is what will determine the extent of long-term corrosion on patterned magnetic multilayers. 4. Summary and conclusions The interhalogen compounds, ICl and IBr, are effective dry etchants for Ni, Fe, NiFe and NiFeCo under high ion density conditions. The maximum etch rates are similar to those we have achieved with pure Cl 2 under the same conditions in the same reactor, but the surfaces are smoother over a broader range of conditions than with Cl Ž 2 which typically produced RMS values a factor of 2 3 higher.. This appears to be related to the lower amount of residual halogen on the etched surfaces Žour past results with Cl typically show 1 3 at.% chlorine residues. 2. The etch rates are strongly dependent on plasma composition, source power, rf chuck power and pressure. All of these trends with plasma parameters are consistent with the etching being limited by the removal of the halogenated reaction products, and the need to balance the formation and removal of these species. Acknowledgements The work at UF is partially supported by a DOD MURI monitored by AFOSR Ž H.C. DeLong., contract F , and by ONR contract N C-2114 through the Honeywell MRAM program. The work of H.C. is partially supported by KOSEF. YBH gratefully acknowledges the support of Korea Research Foundation for Faculty Research Abroad. References wx 1 K. Kinoshita, K. Yamada, H. Matutera, IEEE Trans. Magn. 27 Ž wx 2 G.A. Prinz, in: B. Heinrich, J.A.C. Bland Ž Eds.., Ultra-Thin Magnetic Structures, Vol. II, Springer, Berlin Ž wx 3 M.J. Vasile, C.J. Mogab, J. Vac. Sci. Technol. A 4 Ž wx 4 M. Balooch, D.S. Fischl, D.R. Olander, W.J. Siekhaus, J. Electrochem. Soc. 135 Ž wx 5 D.W. Hess, Plasma Chem. Plasma Proc. 2 Ž wx 6 D.W. Danner, M. Dalvie, D.W. Hess, J. Electrochem. Soc. 134 Ž wx 7 F.C.M.J. van Delft, J. Magn. Mag. Mater Ž wx 8 B. Khamsehpour, C.D.W. Wilkinson, J.N. Chapman, Appl. Phys. Lett. 67 Ž wx 9 W. Vavra, private communication. w10x R.J. Shul, M.C. Lovejoy, D.L. Hetherington, D.J. Rieger, J.F. Klem, M.R. Melloch, J. Vac. Sci. Technol. B 13 Ž

8 222 ( ) H. Cho et al.rapplied Surface Science w11x K.B. Jung, E.S. Lambers, J.R. Childress, S.J. Pearton, M. Jenson, A.T. Hurst Jr., J. Vac. Sci. Technol. A 16 Ž w12x K.B. Jung, J. Hong, H. Cho, J.R. Childress, S.J. Pearton, M. Jenson, A.T. Hurst Jr., J. Electron. Mater. 27 Ž w13x K.B. Jung, E.S. Lambers, J.R. Childress, S.J. Pearton, M. Jenson, A.T. Hurst Jr., Appl. Phys. Lett. 71 Ž w14x J.W. Lee, J. Hong, E.S. Lambers, S.J. Pearton, J. Vac. Sci. Technol. B 15 Ž w15x O.A. Popov Ž Ed.., High Density Plasma Sources, Noyes Data, Park Ridge, NJ Ž w16x Y.B. Hahn, to be published.