Building upon the depth and breadth of Harold Winters's work, this paper pays tribute to his pioneering contribution in the field of plasma etching of metals, and how that knowledge base helps guide the fundamental research in these areas. The fundamental understanding of the plasma–surface interactions during metal etch is key to achieve desirable etch efficacy and selectivity at the atomic scale. This paper presents a generalized methodology, combining thermodynamic assessment and kinetic verification of surface reactions, using copper, magnetic metals, and noble metals as examples, in an effort to demonstrate the applicability of this strategy in tailoring plasma–surface interactions at the atomic scale for a wide range of materials.

The work of Harold Winters had profound impact and spanned over four decades, from his first publication in 1962 to his latest in 2007, as exemplified by the impressive number of publications and citations to his credit (Fig. 1). The sphere of Winters's influence encompassed multiple regimes of plasma science and technology, including the processing of silicon-based materials and metals by physical and chemical sputtering. This paper is not an attempt to review all of Harold Winters's work but a tribute to his pioneering accomplishment in plasma etching of metals, as well as a highlight of how his work continues to inspire advances in this area.

Fig. 1.

Summary of publications authored by Harold Winters, with the number of corresponding citations on the y-axis and the number of publications in the parentheses on the x-axis.

Fig. 1.

Summary of publications authored by Harold Winters, with the number of corresponding citations on the y-axis and the number of publications in the parentheses on the x-axis.

Close modal

Delineating the reaction mechanisms during plasma etching of metals is perhaps a less well-known but equally important area of Winters's work. He studied the energy transfer during collisions between noble gas ions and noble metals such as Pt,1–4 detailing the specific role of ionic bombardment on the etching process by quantifying the fraction of kinetic energy imparted as a function of the incident ion's kinetic energy [Fig. 2(a)].5 The use of noble ions with well-controlled energy distributions is of paramount importance in patterning metallic materials that do not form volatile compounds, and ion milling has become an industrial standard in patterning magnetic and noble metals.6–9 In the past decade, as some of these magnetic and noble metals find potential applications in magnetic data storage devices at nanoscaled device dimensions, findings from Winters's work at low ion energies (<50 eV) continue to be important in guiding the development of an effective patterning process to minimize surface damage.3 In addition to noble metals, Winters conducted some of the earliest work on chlorine etching of copper, where a copper chloride layer was formed at temperatures as low as 30 °C with its thickness increasing linearly with chlorine exposure up to 140 °C. Copper etching was observed at temperatures of 150–200 °C with and without the plasma.1,2 The primary chemical etch product during chlorine etching of copper was found to be trimers of CuCl, Cu3Cl3, as determined by mass spectrometry measurements. Figure 2(b) shows a typical mass spectrum during copper etching in chlorine gas at 590 °C, with (CuCl)3 and its cracking fragments.

Fig. 2.

(a) Fraction of the kinetic energy of Xe, Ar, and He atoms deposited onto Pt(111) surface for normal incidence and surface temperature of 800 K. Reprinted with permission from Winters et al., Phys. Rev. B 41, 6240 (1990). Copyright 1990 American Physical Society. (b) Mass spectrum of products from etching Cu(100) in Cl2 at 590 °C with dominant etch product Cu3Cl3. Adapted and reprinted with permission from Winters, J. Vac. Sci. Technol., B 3, 9 (1985). Copyright 1985 American Vacuum Society.

Fig. 2.

(a) Fraction of the kinetic energy of Xe, Ar, and He atoms deposited onto Pt(111) surface for normal incidence and surface temperature of 800 K. Reprinted with permission from Winters et al., Phys. Rev. B 41, 6240 (1990). Copyright 1990 American Physical Society. (b) Mass spectrum of products from etching Cu(100) in Cl2 at 590 °C with dominant etch product Cu3Cl3. Adapted and reprinted with permission from Winters, J. Vac. Sci. Technol., B 3, 9 (1985). Copyright 1985 American Vacuum Society.

Close modal

Despite Winters's work which demonstrates the feasibility of forming Cu3Cl3 trimers as the major reaction products in chlorine etching of Cu, Cu was deemed unetchable at low temperatures in the nineties. As a result, the microelectronic industry developed an alternate and complex damascene process10 to circumvent the perceived difficulty of etching copper. Two decades later, Cu has been shown to be etchable at lower temperatures, at a rate that is potentially commercially viable.11,12 In retrospect, this could have been realized earlier, building upon what Winters started.1,2 To ensure that the lessons learned in this case can help circumvent similar situations in the future, this paper presents a methodology, combining thermodynamic assessment and kinetic verification of surface reactions, in an effort to demonstrate the applicability of this generalized strategy to help tailor plasma–surface interactions at the atomic scale for a wide range of materials, including copper, magnetic metals, and noble metals.

In an effort to find a gas-phase chemistry capable of patterning copper for interconnect applications using Cl2 chemistry, Kulkarni and DeHoff pioneered the utilization of volatility diagrams as a means to assess the viable processing conditions.13 In the case of copper etching in chlorine [Cl2(g)], the volatility diagram in Fig. 3(a) was constructed by considering the individual chemical reaction and phase equilibria between the condensed phase Cu(c) and gaseous products, Cu(g), CuCl(g), CuCl2(g), and Cu3Cl3(g), as a function of the chlorine partial pressure at various temperatures. The trimeric Cu3Cl3(g) has the highest vapor pressure as an etch product, consistent with Winters's experimental findings.1 However, this vapor pressure at the equilibrium condition was deemed too low to be considered viable in etching copper in Cl2 at temperatures below 200 °C.13 To enhance the etching rate of copper, the utilization of atomic hydrogen was found to be most effective, through its reaction with CuCl2 [i.e., 3CuCl2(c) + 3H(g) ↔ Cu3Cl3(g) + 3HCl(g)]. In Fig. 3(a), the hydrogen isobaric lines have negative slopes due to the highly exothermic nature of the reduction reaction. The interception point between the hydrogen isobaric lines and the metastable extension for the reaction 3CuCl2(c) ↔ Cu3Cl3(g) + 3/2 Cl2(g) suggests that a much enhanced etch rate is attainable.

Fig. 3.

(Color online) (a) Complete volatility diagram for the Cu-Cl-H system at 50 °C, adapted from Kulkarni and DeHoff, J. Electrochem. Soc. 149, G620 (2002). The maximum equilibrium vapor pressure lines are solid, the stable portions of the equilibrium vapor pressure lines that are less than the maximum are dashed, and the metastable extensions are dotted. The gray horizontal line at 10−8 atm denotes the minimum partial pressure of the reaction products for the reaction to be considered effective in etching. Solid green isobaric lines for atomic hydrogen (in atm) are shown to intercept the metastable extensions. (b) Experimental verification of a low temperature (<20 °C), reactive plasma etch process for copper films in a two-step etch process: Cu was first exposed to a Cl2 plasma to preferentially form CuCl2, which was volatilized as Cu3Cl3 by reaction with a H2 plasma. The Cu film thickness is shown to decrease as a function of number of cycles of this two-step plasma etch process, and the surface chemical states are confirmed by XPS, in the inset. Reprinted with permission from Tamirisa et al., Microelectron. Eng. 84, 105 (2007). Copyright 2007 Elsevier.

Fig. 3.

(Color online) (a) Complete volatility diagram for the Cu-Cl-H system at 50 °C, adapted from Kulkarni and DeHoff, J. Electrochem. Soc. 149, G620 (2002). The maximum equilibrium vapor pressure lines are solid, the stable portions of the equilibrium vapor pressure lines that are less than the maximum are dashed, and the metastable extensions are dotted. The gray horizontal line at 10−8 atm denotes the minimum partial pressure of the reaction products for the reaction to be considered effective in etching. Solid green isobaric lines for atomic hydrogen (in atm) are shown to intercept the metastable extensions. (b) Experimental verification of a low temperature (<20 °C), reactive plasma etch process for copper films in a two-step etch process: Cu was first exposed to a Cl2 plasma to preferentially form CuCl2, which was volatilized as Cu3Cl3 by reaction with a H2 plasma. The Cu film thickness is shown to decrease as a function of number of cycles of this two-step plasma etch process, and the surface chemical states are confirmed by XPS, in the inset. Reprinted with permission from Tamirisa et al., Microelectron. Eng. 84, 105 (2007). Copyright 2007 Elsevier.

Close modal

The prediction from these thermodynamic calculations was experimentally verified using a cycle comprised of sequential steps of Cl2 (2 min) and H2 (5 min) plasma exposure [Fig. 3(b)].14,15 Hydrogen plasma was demonstrated to be effective at temperatures below 20 °C in removing the copper chloride layer generated upon Cl2 plasma exposure, as confirmed by x-ray photoelectron spectroscopy (XPS) analysis [Fig. 3(b) inset], resulting in etch rates up to 100 nm/cycle.15 Leveraging this demonstrated success, it was further shown that H2 plasma alone is effective for low temperature etching of copper at rates of 2 nm/min at a substrate temperature of −150 °C up to 9 nm/min at 10 °C.16,17 In addition to the attainable etch rate of copper at low temperatures, this hydrogen-based etch process exhibits substantial anisotropy, with sidewall slopes greater than 80° when used in conjunction with an SiO2 hard mask, suggesting the viability of low temperature patterning of Cu thin films.

Table I summarizes the literature reported reactive ion etch chemistries for patterning Cu, including etch rates, hard mask materials used, and etch selectivity to the mask (when available). While there are some reported works using CO and CH4 as the primary etchant, it is not clear that the etch products are well-known.18–20 

Table I.

Reactive ion etch chemistries reported for copper and the corresponding etch rates, temperatures, masks used, and selectivity (if reported).

Plasma chemistryEtch rate (nm/min)T ( °C)MaskSelectivityReferences
Cl2 120 >200 SiO2 2.0 21  
Cl2 (80%)/Ar 500 200 PR — 22  
H2 13 10 SiO2 — 16, 17, 23, and 24  
Cl2/H2 50 (nm/cycle) 15 — — 15  
CCl4 (50%)/Ar 500 225 PR — 25  
SiCl4 (20%)/Ar 25 220 Polyimide PR — 26  
SiCl4 (20%)/N2 70 
BCl3 (50%)/Ar 53 225 Spin-on glass — 27  
BCl3 (50%)/N2 
CO (12.5%)/NH3 20 20 Ti 8.0 28  
CH4, CH4 (50%)/H2 17, 20 10 SiO2, PR — 19 and 20  
Plasma chemistryEtch rate (nm/min)T ( °C)MaskSelectivityReferences
Cl2 120 >200 SiO2 2.0 21  
Cl2 (80%)/Ar 500 200 PR — 22  
H2 13 10 SiO2 — 16, 17, 23, and 24  
Cl2/H2 50 (nm/cycle) 15 — — 15  
CCl4 (50%)/Ar 500 225 PR — 25  
SiCl4 (20%)/Ar 25 220 Polyimide PR — 26  
SiCl4 (20%)/N2 70 
BCl3 (50%)/Ar 53 225 Spin-on glass — 27  
BCl3 (50%)/N2 
CO (12.5%)/NH3 20 20 Ti 8.0 28  
CH4, CH4 (50%)/H2 17, 20 10 SiO2, PR — 19 and 20  

While halogen based chemistries have been shown to be effective in etching copper, the associated corrosion issues are a concern to materials and device reliability. Utilizing organics in the gas-phase to pattern copper could potentially alleviate the corrosion issue related to the use of halogen chemistries. This paper addresses this concern by extending the thermodynamic assessment to evaluate viable organic chemistries that can be used to pattern copper. Leveraging knowledge from synthesizing organometallic precursors such as copper acetylacetonate [Cu(acac)2] and copper hexafluoroacetylacetonate [Cu(hfac)2] that have reasonable vapor pressure and have been used in chemical vapor deposition and atomic layer deposition processes, this work examined the changes in Gibbs free energy of reaction for Cu-acac and Cu-hfac systems. At 300 K and 1 atm, the changes in Gibbs free energy listed in Table II indicate that reactions with metallic copper are likely to occur in the liquid phase of these chemicals (with the slightly negative values) but not in the gas phase. While there are no available thermochemical data to support the calculations at higher temperatures, it should become feasible to etch Cu in both solution and gas phase as temperature increases and the processing pressure decreases.

Table II.

Gibbs free energies for reaction of metallic copper with liquid and gas phases of acetylacetone and hexafluoroacetylacetone at 300 K and 1 atm. Data only available at 300 K and 1 atm. Note that etch experiments conducted in the gas phase were carried out at elevated temperatures and reduced pressures.

ReactionΔGrxn (kJ/mol)
Cu(c) + 2 acac(l) ↔ Cu(acac)2(l) −39.2 
Cu(c) + 2 hfac(l) ↔ Cu(hfac)2(l) −57.6 
Cu(c) + 2 acac(g) ↔ Cu(acac)2(g) 508.4 
Cu(c) + 2 hfac(g) ↔ Cu(hfac)2(g) 946.6 
ReactionΔGrxn (kJ/mol)
Cu(c) + 2 acac(l) ↔ Cu(acac)2(l) −39.2 
Cu(c) + 2 hfac(l) ↔ Cu(hfac)2(l) −57.6 
Cu(c) + 2 acac(g) ↔ Cu(acac)2(g) 508.4 
Cu(c) + 2 hfac(g) ↔ Cu(hfac)2(g) 946.6 

Since patterning of copper should be achieved in the gas phase for device integration, an alternative route should be considered to enable the reaction as the thermodynamics suggest unfavorable reactions. Interestingly, copper oxides have been shown to dissolve through complexation in β-diketones and other organic acids in the solution phase.29,30 There are similar reports in vapor phase cleaning of copper where the hfac chemistry has been used to remove oxidized copper with a rate of up to 150 nm/min at 350 °C.31,32 Cupric oxide (CuO) is reportedly removed through the production of Cu(hfac)2. However, at elevated temperatures, the reduction of Cu(II) to Cu(I) directly competes with the chelation reaction.33 Nonetheless, the reaction of Cu2O with hfac remains important through disproportionation. Removal of Cu(I) and Cu(II) oxides was observed to terminate upon reaching the Cu0 surface, due to the reduced reactivity between metallic Cu and hfac,33 consistent with the thermodynamic analysis shown above. This implies a very high selectivity between etching copper and copper oxide and an effective way to achieve the removal of copper by tailoring its oxidation states.

The same thermodynamic assessment was applied to address the spontaneity of the CuO-acac and CuO-hfac systems by calculating the changes in Gibbs free energy in both liquid and gas phase reactions (Table III). The negative values reported for CuO reacting with acac and hfac in the liquid phase suggest that Cu(acac)2 and Cu(hfac)2 can be effectively generated. Though reactions between CuO and acac/hfac remained unfavorable in the gas phase, they are more likely than Cu reacting with acac and hfac. It has been shown that copper complexation with acac and hfac occurs with the organic molecule in the “standing up” mode, with C=O group pointing downward to the surface.32 Similar mechanisms have been reported for carboxylic acids,34 prompting the evaluation of additional organics in this study, namely, formic and acetic acid (no thermodynamic calculations are shown for these chemicals due to the lack of thermochemical data of the reaction products). Again, while there are no available thermochemical data to support the calculations at higher temperatures, it should become feasible to etch CuO in both solution and gas phase as temperature increases and the processing pressure decreases.

Table III.

Gibbs free energies for reaction of copper oxide with liquid and gas phases of acetylacetone and hexafluoroacetylacetone at 300 K and 1 atm. Data only available at 300 K and 1 atm. Note that etch experiments conducted in the gas phase were carried out at elevated temperatures and reduced pressures.

ReactionΔGrxn (kJ/mol)
CuO(c) + 2 acac(l) ↔ Cu(acac)2(l) + H2O(l) −119.1 
CuO(c) + 2 hfac(l) ↔ Cu(hfac)2(l) + H2O(l) −428.7 
CuO(c) + 2 acac(g) ↔ Cu(acac)2(g) + H2O(g) 409.2 
CuO(c) + 2 hfac(g) ↔ Cu(hfac)2(g) + H2O(g) 847.4 
ReactionΔGrxn (kJ/mol)
CuO(c) + 2 acac(l) ↔ Cu(acac)2(l) + H2O(l) −119.1 
CuO(c) + 2 hfac(l) ↔ Cu(hfac)2(l) + H2O(l) −428.7 
CuO(c) + 2 acac(g) ↔ Cu(acac)2(g) + H2O(g) 409.2 
CuO(c) + 2 hfac(g) ↔ Cu(hfac)2(g) + H2O(g) 847.4 

Etching of copper and copper oxide in both liquid and gas phases of selected organic chemistries were conducted in this study to corroborate the thermodynamic calculations. Cu films (50 nm) were deposited by electron-beam evaporation on a p-type silicon wafer. Thick CuO films were generated by rapid thermal annealing of copper thin films in an AccuThermo AW 610 RTP furnace with a flow of 3000 sccm of O2 at 200 °C. Organics, including acac, hfac, acetic acid, and formic acid were purchased from Sigma-Aldrich at 99% purity, with the exception of hfac (98% purity) and used in both solution and vapor phase etch studies. Etching efficacy was first tested in high purity solutions (no dilution) in a vial in a water bath at 80 °C. Changes in thickness of the films were measured by cleaving samples after processing and imaged by examining each cross section in an FEI Nova 600 scanning electron microscope (SEM) (1.4 nm resolution at 1 kV).

For etching elemental copper thin films, as shown in Fig. 4(a), both acac and hfac solutions were able to etch copper at a very small rate of 0.9 nm/min. Slightly higher etch rates of copper were observed in acetic acid (1.4 nm/min) and formic acid (2.2 nm/min) solution. These etch rates, while measurable, indicate that the corresponding etch rates in the gas phase would be negligible (due to the much lower concentration). Indeed, all four organic chemistries were used in the vaporizer system referenced.37 No measurable etch rates were found for these organic vapors reacting with metallic copper at 80 °C (data not shown). From literature, measurable etch rates of CuOx have been reported for acetic acid(g) and hfac(g), but at much higher temperatures 200–350 °C.32,38

Fig. 4.

(Color online) (a) Etch rates of Cu and CuOx in liquid phase acac, hfac, acetic acid, and formic acid are shown as measured at 80 °C. Shown in comparison are results from etching Cu and CuOx thin films in formic acid vapor at 150 °C and 10 Torr for 5 min. SEM cross-sections (b) before and (c) after etching CuOx thin films in formic acid vapor. The asterisk (*) denotes the zero etch rate as measured in this work.

Fig. 4.

(Color online) (a) Etch rates of Cu and CuOx in liquid phase acac, hfac, acetic acid, and formic acid are shown as measured at 80 °C. Shown in comparison are results from etching Cu and CuOx thin films in formic acid vapor at 150 °C and 10 Torr for 5 min. SEM cross-sections (b) before and (c) after etching CuOx thin films in formic acid vapor. The asterisk (*) denotes the zero etch rate as measured in this work.

Close modal

For etching CuOx thin films, much higher etch rates were observed at 80 °C, with etch rates of 110, 150, 540, and 338 nm/min in acac, hfac, acetic acid, and formic acid solutions, respectively [Fig. 4(a)]. The observed high etch rate of CuOx is consistent with literature reports using similar organics and other chemistries [aqua regia, CuCl2, FeCl3, and (NH4)S2O8].6,30,35,36 These observations suggest that an organic chemistry based vapor phase etching process may be feasible if copper is oxidized. To verify this point, formic acid was selected and used in the vaporizer to etch both Cu and CuOx films. No measurable etch rate was observed for etching Cu in the gas phase under all experimental conditions explored in this study. At reduced pressures and elevated temperatures of 10 Torr/150 °C/5 min and 540 Torr/80 °C/10 min, etch rates of CuOx at 19 and 5 nm/min were obtained, respectively. The etch rate as well as the cross-sectional SEM image of CuOx etched in formic acid vapor at 10 Torr/150 °C/5 min are shown in Figs. 4(a)–4(c).

The substantially higher etch rate of oxidized copper over metallic copper suggests that a viable dual step process could be developed to etch metallic Cu by first converting it to copper oxide then exposing it to the organic vapor. Thus in this work, an O2 plasma was used (500 W source power, −100 V bias) to oxidize the copper surface and generate a thin layer of copper oxide, a thin layer (∼3 nm) of copper oxide can be formed after the plasma oxidation for 2 min. Next, acac or formic acid vapor was delivered at flow rates up to 3 cm3/min to achieve pressures of 85 and 148 Torr, respectively. A cyclic processing alternating between oxidation and organic vapor exposure was used to etch the copper. Both chemicals achieved measurable etch rates of CuOx, as shown in Fig. 5. It was clear that these organic vapors are much more effective in removing CuOx than Cu. In addition, since this etch rate is reported in nm/cycle, the etching process can reach precise etch depth through the careful control of the oxide formation and the etching efficacy of the organic vapor.

Fig. 5.

(Color online) Measured etched thickness per cycle for copper as a function of a cyclic process alternating between O2 plasma and organic vapor exposure (the dotted lines are linear regressions with R2 values >0.98). An etch rate of 1.3 nm/cycle was determined for a cyclic process with acac vapor (85 Torr), while a rate of 3.7 nm/cycle was measured for a cyclic process with formic acid vapor (148 Torr).

Fig. 5.

(Color online) Measured etched thickness per cycle for copper as a function of a cyclic process alternating between O2 plasma and organic vapor exposure (the dotted lines are linear regressions with R2 values >0.98). An etch rate of 1.3 nm/cycle was determined for a cyclic process with acac vapor (85 Torr), while a rate of 3.7 nm/cycle was measured for a cyclic process with formic acid vapor (148 Torr).

Close modal

Based on the results shown in Fig. 5, formic acid was chosen for the subsequent experimental validation due to its substantial gas phase etch rate of copper oxide at a lower temperature and its high vapor pressure (50 Torr at 25 °C). Blanket Cu (35 nm) films were etched using alternating cycles of plasma oxidation and formic acid vapor exposure described above, using a reactor system detailed elsewhere.37,39 SEM cross-section measurement indicated that the 35 nm Cu film was completely removed after 15 cycles (Fig. 6). Energy dispersive x-ray spectroscopy (EDS) was used to verify Cu removal. Preprocessed samples show strong signatures of O Lα (0.52 keV), Cu Lα (0.93 keV) and Si Kα (1.73 keV) from the 35 nm Cu film and underlying Si substrate with a small peak for C Lα at 0.27 keV. Upon processing by 15 cycles of alternating O2 plasma oxidation and formic acid exposure with a terminal oxidation step, the Cu Lα signature is almost entirely removed, confirming removal of the Cu film [Fig. 6(b)].

Fig. 6.

(Color online) (First column) SEM cross-section and EDS spectrum of preprocess Cu (35 nm) on 35 nm of low-k material and 200 nm of SiO2. (a) SEM cross-section of 35 nm Cu etched under 15 cycles consisting of alternating plasma oxidation (500 W, −100 V, and 2 min) and formic acid vapor etch (50 s). (b) Energy dispersive x-ray spectrum (EDS) of sample surface after exposure to 15 cycles. The interface between the mask and copper is outlined with a thin dashed line to help guide the eyes for comparison.

Fig. 6.

(Color online) (First column) SEM cross-section and EDS spectrum of preprocess Cu (35 nm) on 35 nm of low-k material and 200 nm of SiO2. (a) SEM cross-section of 35 nm Cu etched under 15 cycles consisting of alternating plasma oxidation (500 W, −100 V, and 2 min) and formic acid vapor etch (50 s). (b) Energy dispersive x-ray spectrum (EDS) of sample surface after exposure to 15 cycles. The interface between the mask and copper is outlined with a thin dashed line to help guide the eyes for comparison.

Close modal

While etch rate measurements with blanket films are useful to confirm the disparity in etching metallic and oxidized copper, a direct comparison between solution- and gas-phase formic acid etching processes was made in this work utilizing prepatterned copper films. Specifically, a 35 nm Cu thin film was patterned by a 40 nm TiN hard mask to form 70 nm × 10 μm lines and 1 × 1 μm pads, as shown in the first column in Fig. 7 with both cross-sectional and birds-eye views.

Fig. 7.

(Color online) (First column) As-patterned 1 × 1 μm pads and 70 nm × 10 μm lines of 40 nm TiN hard mask (HM) patterned 35 nm Cu films, in cross-sectional and birds-eye views. The interface between the mask and copper is outlined with a thin dashed line to help guide the eyes for comparison. (a) and (b) SEM images of patterned copper etched in formic acid (500 s) in solution (no oxidation). (c) and (d) SEM images of patterned copper etched by 10 cycles of alternating O2 plasma (500 W, −100 V bias, and 2 min) and formic acid vapor (50 s), terminating with a final exposure to O2 plasma.

Fig. 7.

(Color online) (First column) As-patterned 1 × 1 μm pads and 70 nm × 10 μm lines of 40 nm TiN hard mask (HM) patterned 35 nm Cu films, in cross-sectional and birds-eye views. The interface between the mask and copper is outlined with a thin dashed line to help guide the eyes for comparison. (a) and (b) SEM images of patterned copper etched in formic acid (500 s) in solution (no oxidation). (c) and (d) SEM images of patterned copper etched by 10 cycles of alternating O2 plasma (500 W, −100 V bias, and 2 min) and formic acid vapor (50 s), terminating with a final exposure to O2 plasma.

Close modal

In formic acid solution, a 500 s exposure at 80 °C resulted in removal of a substantial amount of metallic copper at a rate consistent with the reported value in the earlier section and as shown in Figs. 7(a) and 7(b): severe undercutting (more than 200 nm in the lateral direction) was observed under the 1 × 1 μm hard mask and a complete suspension of the 70 nm lines. In the gas phase, formic acid was not able to etch metallic copper (SEM not shown). However, the cyclic process involving oxidation of copper followed by formic acid vapor exposure (148 Torr, 50 s, 80 °C) achieved an anisotropic etch profile after 10 such cycles with a terminal oxidation step, for both 1 × 1 μm pads [Fig. 7(c)] and 70 nm × 10 μm lines [Fig. 7(d)].

Although the delivery of the formic acid vapor to the surface and its subsequent reaction is isotropic, the removal of a directionally oxidized CuOx under a patterned feature results in high selectivity to metallic copper and an anisotropic etch profile because the oxidation step is a directional one, involving the use of an oxygen plasma and a substrate bias.

The alternating oxidation and formic acid vapor etch process provides a number of benefits for furthering the patterning of copper films. First, the high selectivity with which formic acid vapor removes copper oxide over metallic copper indicates that this cyclic approach can be utilized for pursuing etch at the atomic scale by controlling the etched thickness through the modified layer thickness. Second, directionality of etch can be induced by applying a bias voltage to the sample during plasma oxidation. Finally, this dual step approach is functional at temperatures substantially lower than those used for halogen-based plasma etch.

As shown by Winters in his earlier work, ion milling has been successfully used to pattern a wide variety of magnetic materials, such as those in a magnetoresistive random access memory (MRAM) stack. However, ion milling's inherently low selectivity between the hard mask and target material can cause sidewall deposition and shorting across the tunnel barrier. Having established the efficacy of applying thermodynamic assessment to enable the identification of viable plasma chemistries and operating conditions for copper etching, this approach is further validated by applying it to address the patterning of magnetic thin films in an MRAM stack. As shown in Fig. 8, the material stack contains TiN, Ru, CoFeB, MgO, CoFe, and PtMn, with many layers at a thickness of 0.8–3 nm. Due to the wide variety of materials present within a magnetic tunnel junction (MTJ) and the presence of nanometer or subnanometer thicknesses, controlling selectivity at an atomic level is key in patterning each layer. Furthermore, the reaction chemistries must be capable of removing materials from within high aspect features in order to meet the criteria of increasing device density for storage capacity.

Fig. 8.

Schematic and a transmission electron micrograph of a reactive ion etched MTJ stack. Adapted from and reprinted with permission from Garay et al., Electrochem. Solid State 4, P77 (2015). Copyright 2015 The Electrochemical Society.

Fig. 8.

Schematic and a transmission electron micrograph of a reactive ion etched MTJ stack. Adapted from and reprinted with permission from Garay et al., Electrochem. Solid State 4, P77 (2015). Copyright 2015 The Electrochemical Society.

Close modal

Using Co and Fe as model materials, halogen chemistry was first applied to assess the etching efficacy. Thermodynamic calculations were performed for Co and Fe films to achieve insights into patterning CoFe and CoFeB where thermodynamic data are not readily available.40 The thermodynamics predicts that the halogen reacts with Co and Fe to form metal chlorides, whose volatility is not high enough at temperatures below 200 °C for the process to be considered viable. The feasibility for Co and Fe etching was then expanded to consider reactions between the meta-stable state of MX2 (M = Co, Fe and X = F, Cl, Br) and a secondary chemistry. Among all secondary chemistries examined by thermodynamics, atomic hydrogen was shown to be the most effective in shifting the equilibrium to generate more volatile etch products (metal hydrides).40 Experimental verification of these predictions is shown in Fig. 9. The etch rates of Co and Fe in continuous Cl2 were measured at 3.9 and 8.0 nm/cycle, respectively. An alternating Cl2/H2 plasma etch process resulted in enhanced etch rates up to 4.6 and 8.5 nm/cycle for Co and Fe [Figs. 9(a) and 9(b)], and the XPS confirmed the formation and removal of metal chlorides [Figs. 9(c) and 9(d)].40 

Fig. 9.

Etch rate measurement of (a) Co and (b) Fe exposed to alternating Cl2 and H2 plasma cycles. The corresponding XPS of the etched surfaces for (c) Co and (d) Fe. Adapted and reprinted with permission from Kim et al., J. Vac. Sci. Technol., A 32, 041305 (2014). Copyright 2014 American Vacuum Society.

Fig. 9.

Etch rate measurement of (a) Co and (b) Fe exposed to alternating Cl2 and H2 plasma cycles. The corresponding XPS of the etched surfaces for (c) Co and (d) Fe. Adapted and reprinted with permission from Kim et al., J. Vac. Sci. Technol., A 32, 041305 (2014). Copyright 2014 American Vacuum Society.

Close modal

Building upon the etching efficacy of Co and Fe, the same plasma chemistries were shown to be successful in patterning CoFe.41 Specifically, patterned samples of TiN(300 nm)/CoFe(45 nm)/Ti(10 nm) on Si wafer were etched for 2 min in a Cl2 plasma (500 W source power, 100 W bias power, and 20 mTorr) without or with a subsequent 4 min H2 plasma etching (500 W source power, 50 W bias power, and 20 mTorr). SEM was used to observe the etch profile and patterned surface after Cl2 plasma etch and subsequent H2 plasma exposure [Figs. 10(a) and 10(b)]. EDS was used to spatially resolve elements present on the surface after each processing step [Figs. 10(c) and 10(d)]. Exposure to Cl2 plasma resulted in etching of both the CoFe and TiN hard mask [Fig. 10(a)]; however, substantial redeposition of nonvolatile metal chlorides was observed immediately adjacent to the corner of the patterned mask on CoFe, evidenced by the emergence of a Cl Kα peak at 2.67 keV [Fig. 10(c)]. Subsequent exposure of the chlorinated surface to hydrogen plasma for 4 min visibly removed sidewall redeposition [Fig. 10(b)], corroborated by EDS analysis showing the disappearance of Cl Kα [Fig. 10(d)].41 The increased etch rates, as well as removal of metal chlorides, validated the thermodynamic assessment that led to the selection of viable plasma etch chemistries for patterning ferromagnetic alloys such as CoFe.

Fig. 10.

(Color online) SEM images of patterned samples with 300 nm TiN-mask and 45 nm CoFe films, etched in (a) a 20 mTorr Cl2 plasma with a source power of 500 W, a bias power of 100 W, for 2 min and (b) identical Cl2 plasma exposure with a subsequent H2 plasma exposure at 500 W source power, 50 W bias power, and 20 mTorr for 4 min. The corresponding energy dispersive x-ray spectra (EDS) (c,d) are shown under each SEM image, taken near the TiN mask border and on the exposed CoFe film, and indicate that metal chloride deposits were removed upon exposure to H2 plasma. The spectrum shown in (d) was taken at the border region circled in (b). Adapted and reprinted with permission from Kim et al., J. Vac. Sci. Technol., A 33, 021308 (2015). Copyright 2015 American Vacuum Society.

Fig. 10.

(Color online) SEM images of patterned samples with 300 nm TiN-mask and 45 nm CoFe films, etched in (a) a 20 mTorr Cl2 plasma with a source power of 500 W, a bias power of 100 W, for 2 min and (b) identical Cl2 plasma exposure with a subsequent H2 plasma exposure at 500 W source power, 50 W bias power, and 20 mTorr for 4 min. The corresponding energy dispersive x-ray spectra (EDS) (c,d) are shown under each SEM image, taken near the TiN mask border and on the exposed CoFe film, and indicate that metal chloride deposits were removed upon exposure to H2 plasma. The spectrum shown in (d) was taken at the border region circled in (b). Adapted and reprinted with permission from Kim et al., J. Vac. Sci. Technol., A 33, 021308 (2015). Copyright 2015 American Vacuum Society.

Close modal

Other advantages of a terminal hydrogen plasma exposure step include the elimination of corrosion concerns associated with halogen and the restoration of the desired magnetic properties crucial to the device operation in a MTJ stack, namely, the coercivity, the required applied magnetic field to switch the magnetization. Superconducting quantum interference device magnetometry was utilized to measure the magnetization of a 45 nm CoFe thin film before and after exposure to optimized processing steps consisting of 30 s exposure to Cl2 plasma as well as a subsequent 30 s exposure to H2 plasma. The coercivity for preprocessed CoFe was measured to be ∼21 Oe. Upon exposure to Cl2 plasma, the coercivity increased substantially to ∼64 Oe, consistent with the presence of magnetically hard metal chlorides on the surface. This is detrimental to the magnetic device performance. Upon subsequent exposure to H2 plasma, coercivity was decreased to ∼23 Oe, nearly the same value measured from the as-prepared sample. This is consistent with the removal of metal chlorides and confirms the efficacy of a hydrogen plasma to restore the desirable magnetic property (Fig. 11).41 

Fig. 11.

(Color online) Magnetic hysteresis shown in normalized magnetization as a function of applied magnetic field, for pre-etched CoFe, and CoFe films etching in Cl2 plasma and by a Cl2/H2 plasma process. Adapted and reprinted with permission from Kim et al., J. Vac. Sci. Technol., A 33, 021308 (2015). Copyright 2015 American Vacuum Society.

Fig. 11.

(Color online) Magnetic hysteresis shown in normalized magnetization as a function of applied magnetic field, for pre-etched CoFe, and CoFe films etching in Cl2 plasma and by a Cl2/H2 plasma process. Adapted and reprinted with permission from Kim et al., J. Vac. Sci. Technol., A 33, 021308 (2015). Copyright 2015 American Vacuum Society.

Close modal

The demonstrated success in patterning CoFe is recently extended to etch CoFeB as summarized below. The incorporation of boron in CoFe has been shown to improve the performance of magnetic materials through better matching of the interface with MgO.42,43 In fact, CoFeB/MgO is at the heart of advanced MTJs owing to the coexistence of excellent tunneling magnetoresistance and strong perpendicular magnetic anisotropy.44 Table IV summarizes the literature reported plasma chemistries used for reactive ion etch of CoFeB. Halogen chemistries were used in conjunction with Ar or the aforementioned H2, while incorporation of numerous organics, such as CH4, CH3OH, and CH3COOH, were also employed but suffered from undesirable carbon deposition or oxidation caused from methanol and acetic acid fragmentation.45–48 None of the chemistries reported exhibits etch rates higher than 20–30 nm/min, with moderate selectivity to Ti or TiN masks.

Table IV.

Literature reported results on reactive ion etch of CoFeB.

Plasma chemistryEtch rate (nm/min)T (°C)MaskSelectivityReferences
Cl2 (10%)/Ar 15 90 — — 49  
H2O (20%)/Ar 15 TiN 3.3 46  
CO (25%)/NH3 12 50 Ti 2.0 50  
CH4 (20%)/Ar 30 15 Ti 2.0 45  
CH3OH (15%)/Ar 10 12 Ti 4.0 51  
CH3OH (25%)/H215 TiN 5.5 47  
CH3COOH (25%)/Ar 15 15 W/TiN 0.35 52  
Plasma chemistryEtch rate (nm/min)T (°C)MaskSelectivityReferences
Cl2 (10%)/Ar 15 90 — — 49  
H2O (20%)/Ar 15 TiN 3.3 46  
CO (25%)/NH3 12 50 Ti 2.0 50  
CH4 (20%)/Ar 30 15 Ti 2.0 45  
CH3OH (15%)/Ar 10 12 Ti 4.0 51  
CH3OH (25%)/H215 TiN 5.5 47  
CH3COOH (25%)/Ar 15 15 W/TiN 0.35 52  

Based upon the above reported thermodynamic analysis and subsequent experimental verification of Co, Fe and CoFe etched in Cl2 and H2,40,41 30 nm CoFeB films were etched in Cl2 (500 W source power, 50 W bias power, 5 mTorr, and 30 s), H2 (800 W, 50 W, 5 mTorr, and 30 s) and an alternating process comprised of both Cl2 and H2 plasmas. As shown in Fig. 12, in Cl2 plasma, CoFeB was etched at a rate of 6.5 nm/min. In H2 plasma, CoFeB showed no measurable etch rate, except for the observed removal of the surface oxide. Using an alternating Cl2/H2 plasma etch process resulted in an etch rate of 10.9 nm/cycle. One cycle was equivalent to one minute, resulting in an etch rate of 10.9 nm/min, an increase of 68% over Cl2 plasma alone. A 5 s delay was necessary between each processing step to allow time for gas switching; however, this was not incorporated into the etch rate calculation. While this etch rate is less than that reported for a number of other chemistries in Table IV, the cyclical process reported here can be further optimized by tuning each step in the cycle. These findings further confirm the validity of thermodynamic analysis utilized before, which suggested that processing under a reductive environment was beneficial to not only increase the etch rate of ferromagnetic materials through volatilization of metallic chlorides but also mitigate corrosion. These etched surfaces were examined by XPS and the analysis of Co 2p, Fe 2p, and Cl 2p spectra confirmed the surface chlorination by chlorine plasma and removal of metal chlorides by hydrogen plasma, and magnetic properties of CoFeB were restored after the hydrogen plasma treatment (not shown here).53 

To circumvent the corrosion issue that can deteriorate the static magnetic properties, organic chemistries seem a viable alternative in patterning magnetic metals, given the success in synthesizing volatile metalorganic precursors for many magnetic elements. Based on the findings earlier, the utility of organic chemistry in patterning magnetic and noble metals is expanded in Sec. IV to enable atomic layer etch (ALE).

In recent years, ALE has gained much attention due to its role as a critical processing technique the fabrication of devices <35 nm.54 Since first formalized for Si etch using Cl2/Ar chemistry,8,9 ALE has been shown to be viable in patterning SiO2,7 Si3N4,55–58 as well as other semiconductors and metal oxides.59,60 While all these reported results show some level of efficacy in etching targeted materials, there is currently limited literature demonstrating ALE of metals. This is partly because metallic bonding, caused by the sharing of free electrons across the metal lattice, makes it challenging to direct the bond formation between metals and the gas-phase chemistries to realize self-limiting reactions that are essential to ALE.

Building upon the success of tailoring the copper etch by its oxidation states, exposing a metallic thin film to highly electronegative plasma species such as oxygen could provide the chemical contrast between metal and metal oxide to allow select organic chemistries to selectively remove one but not the other. The relative reactivity of both metal and metal oxide can be assessed through the use of thermodynamics by comparing change in free energy upon reaction with a chosen chemistry. Using Co and CoO as an example in their reaction with acetylacetone (liquid), the calculated ΔGrxn for formation of Co(acac)2 from Co and acac is approximately −410 kJ/mol, whereas for CoO reacting with acac is calculated to be about −710 kJ/mol. Similar effectiveness in organic chemistries was found in removing iron oxide through the reaction between Fe2O3 and hfac(g).61 Even for noble metals, such as Pt, aqueous nitrate (NO3) ions act as strong oxidizing agents, promoting the conversion of metallic Pt0 to Pt4+, which subsequently reacts with chloride ions (Cl) to generate soluble PtCl4 and PtCl62−. This substantial difference in the change in free energy of reaction between the metallic and oxidized condensed phase can be exploited as the basis to achieve a highly selective removal process. Imparting directionality to the modification chemistry is possible through the use of ionized reactive species that have a narrow angular distribution—the directional surface modification thus leads to an anisotropic profile once the modified layer is removed.

In this work, O2 plasmas were used to chemically modify transition and noble metals such as Co, Fe, Cu, Pt, and Pd. Figure 13(a) presents the oxidation of metals as a function of O2 plasma oxidation time based on the XPS overlayer-substrate model. The thickness of metal oxides increased immediately in the first few seconds, and the growth rate saturated at five minutes for most metals. A measured self-bias of ∼5 eV, when compared to the sputtering threshold energies of the metal elements in oxygen, indicated that the oxygen source was mainly atomic oxygen, and the ion sputtering effect could be neglected.39 

Fig. 12.

(Color online) Etched thickness as a function of time for 30 nm CoFeB films etched under continuous Cl2 plasma (500 W source power, 50 W bias power), continuous H2 plasma (800 W, 50 W), and alternating Cl2 and H2 plasmas (same conditions).

Fig. 12.

(Color online) Etched thickness as a function of time for 30 nm CoFeB films etched under continuous Cl2 plasma (500 W source power, 50 W bias power), continuous H2 plasma (800 W, 50 W), and alternating Cl2 and H2 plasmas (same conditions).

Close modal

Various oxidation mechanisms at low temperature have been proposed empirically, and the inverse logarithmic rate law presents the best fit for Fe, Cu, and Co, suggesting that the rate-determining step is the incorporation of metal cations into interstitial sites. The parabolic rate law exhibits the best fit for Pd and Pt oxidation, suggesting that beyond the adsorption of the first oxygen monolayer, oxidation is rate-controlled by the diffusion of reactive oxygen species to the metal.39 Formic acid was chosen to test the efficacy of the cyclic oxidation/etch process detailed earlier (O2 plasma oxidation at 0 V, 500 W, 5 min and formic acid vapor at 148 Torr and 50 s). Figure 13(b) confirmed that the thickness removed in each etching step is essentially the oxide thickness formed in the proceeding step, with the linear relationship and a slope of near unity. This result also demonstrates that the formic acid vapor can achieve a high etch selectivity between the metal and the metal oxide. This highly selective organic vapor etch step can be qualified as a self-limiting reaction, which is an important requirement in realizing atomic layer etching. As such, the etch rate of metals can be controlled at the atomic scale through the control of the thickness of oxide grown.

Fig. 13.

(Color online) (a) Best model for the oxidation of Fe, Cu, and Co is the inverse logarithm rate law, with the reciprocal thickness (nm−1) of oxides formed on Fe, Cu, and Co shown as a function of the natural logarithm of the plasma oxidation time. The best model for the oxidation of Pd and Pt is parabolic rate law, with the inset showing the thickness of oxides (nm) formed on Pd and Pt as a function of O2 plasma oxidation time. (b) Metals, including Pt, Pd, Co, Cu, and Fe, are removed by alternating cycles of O2 plasma (oxidation) followed by formic acid vapor exposure (etch). Reprinted with permission from Chen et al., J. Vac. Sci. Technol., A 35, 05C304 (2017). Copyright 2017 American Vacuum Society.

Fig. 13.

(Color online) (a) Best model for the oxidation of Fe, Cu, and Co is the inverse logarithm rate law, with the reciprocal thickness (nm−1) of oxides formed on Fe, Cu, and Co shown as a function of the natural logarithm of the plasma oxidation time. The best model for the oxidation of Pd and Pt is parabolic rate law, with the inset showing the thickness of oxides (nm) formed on Pd and Pt as a function of O2 plasma oxidation time. (b) Metals, including Pt, Pd, Co, Cu, and Fe, are removed by alternating cycles of O2 plasma (oxidation) followed by formic acid vapor exposure (etch). Reprinted with permission from Chen et al., J. Vac. Sci. Technol., A 35, 05C304 (2017). Copyright 2017 American Vacuum Society.

Close modal

To confirm that the directional surface modification can lead to an anisotropic profile with high selectivity, a vapor phase etch of patterned Co films was demonstrated through sequential exposure to O2 plasma oxidation (500 W, 5 min, with and without substrate bias) and formic acid vapor etch (148 Torr, 50 s). The as-patterned Co film stack was comprised of 40 nm TiN hard mask/10 nm SiN/30 nm Co/10 nm barrier layer/Si substrate patterned into lines 70 nm in width and 10 μm in length. To study the directionality of the vapor etch, the etch profile of the patterned Co films was characterized by cross-sectional SEM. For comparison, as-patterned Co is shown in the first image column of Fig. 14. The cross-section of the as-prepared patterned Co shows a line width of 70 nm, a slight undercut into the SiN layer and a 30 nm Co film. Figure 14(a) shows the profile evolution after the sample was subjected to 6 alternating cycles of O2 plasma (0 V, 500 W, and 5 min) and formic acid vapor (50 s), after which, a significant undercut below the hard mask into Co is observed. This undercut is attributed to isotropic oxidation and subsequent formic acid vapor removing all generated oxides. The hard mask remained intact, further confirming the high selectivity of Co to TiN in formic acid vapor.37 Directional modification of the Co was explored using an applied substrate bias of −200 V during the O2 plasma exposure. An optimized cyclical process comprised of an O2 plasma (−200 V, 500 W, and 5 min) followed by formic acid vapor (50 s) was developed for anisotropic etch. Figure 14(b) shows that an anisotropic etch profile was observed with the optimized process, with a 5 nm recess in the TiN hard mask, demonstrating the viability of a cyclic oxidation and organic chemical vapor etch process for achieving anisotropic etch profiles.37 

Fig. 14.

(Color online) Cross-sectional schematics and corresponding SEM images of 40 nm TiN hard mask (HM) and 10 nm SiN, patterned on 30 nm Co, with a 10 nm barrier layer (BL) on Si before and after processing. The boundary between each layer is outlined with a thin dashed line to help guide the eyes for comparison. The chemical processes are exposure to an O2 plasma (500 W, 5 min) and formic acid vapor (50 s). Etched profiles of patterned Co after (a) six alternating cycles of oxidation (0 V bias) and formic acid vapor exposure, and (b) four alternating cycles of oxidation (−200 V bias) and formic acid vapor exposure. Reprinted with permission from Chen et al., J. Vac. Sci. Technol., A 35, 05C305 (2017). Copyright 2017 American Vacuum Society.

Fig. 14.

(Color online) Cross-sectional schematics and corresponding SEM images of 40 nm TiN hard mask (HM) and 10 nm SiN, patterned on 30 nm Co, with a 10 nm barrier layer (BL) on Si before and after processing. The boundary between each layer is outlined with a thin dashed line to help guide the eyes for comparison. The chemical processes are exposure to an O2 plasma (500 W, 5 min) and formic acid vapor (50 s). Etched profiles of patterned Co after (a) six alternating cycles of oxidation (0 V bias) and formic acid vapor exposure, and (b) four alternating cycles of oxidation (−200 V bias) and formic acid vapor exposure. Reprinted with permission from Chen et al., J. Vac. Sci. Technol., A 35, 05C305 (2017). Copyright 2017 American Vacuum Society.

Close modal

This review began with an overview of the accomplishments of Harold Winters in plasma etching of metals, especially Cu. Winters's identification of the Cu3Cl3 trimer as the volatile etch product for chlorine-based etch of Cu thin films laid the foundation for future thermodynamic analysis of plasma etch chemistries, such as Cl2, which were later experimentally verified alongside the use of reductive environments to mitigate detrimental halogen-induced corrosion of not only Cu but also ferromagnetic Co, Fe, CoFe, and CoFeB and restore the crucial magnetic properties upon which next generation magnetic memory technologies rely. The same volatility analysis was extended to identify organic chemistries capable of etching metallic thin films in both solution and gas phase. To attain higher selectivity, chemical surface modification was employed, creating contrast between metal oxide and metal and enabling atomic scale control over material removal by controlling generated oxide thickness. Expert knowledge from Harold Winters's publications continues to drive the mechanistic understanding of the surface reaction pathways, thereby enabling the control of surface chemistry to realize high specificity and selectivity in patterning various materials at the atomic scale.

J.P.C. had the honor of receiving the Coburn and Winters Award from AVS and the privilege of meeting Harold Winters, whose professional and humble manner is vividly remembered, and his words of wisdom and encouragement are always inspiring. Harold's impact continues as he constantly inspired many to continue the pursuit in the understanding of the complex plasma, from the viewpoints of both chemistry and physics. The authors acknowledge the financial support from the Lam Research Corporation and Integrated Modeling, Process, and Computation for Technology Plus (IMPACT+) Program. A portion of this work was supported by the Semiconductor Research Corporation (Award No. 2001-MC-425.046) through the SRC Engineering Research Center for Environmentally Benign Semiconductor Manufacturing (ERC). The authors thank Taeseung Kim and Ernest Chen at UCLA and Meihua Shen, Thorsten Lill, and Shuogang Huang at Lam Research for their contributions to this work and fruitful discussions.

1.
H. F.
Winters
,
J. Vac. Sci. Technol., A
3
,
786
(
1985
).
2.
H. F.
Winters
,
J. Vac. Sci. Technol., B
3
,
9
(
1985
).
3.
H. F.
Winters
,
H.
Coufal
,
C. T.
Rettner
, and
D. S.
Bethune
,
Phys. Rev. B
41
,
6240
(
1990
).
4.
H.
Coufal
,
H. F.
Winters
,
H. L.
Bay
, and
W.
Eckstein
,
Phys. Rev. B
44
,
4747
(
1991
).
5.
J. W.
Coburn
and
H. F.
Winters
,
Nucl. Instrum. Methods, B
27
,
243
(
1987
).
6.
P.
Walker
and
W. H.
Tarn
,
Handbook of Metal Etchants
(
CRC
,
Boca Rotan, FL
,
1991
).
7.
A.
Agarwal
and
M. J.
Kushner
,
J. Vac. Sci. Technol., A
27
,
37
(
2009
).
8.
S. D.
Athavale
,
J. Vac. Sci. Technol., B
14
,
3702
(
1996
).
9.
S. D.
Athavale
and
D. J.
Economou
,
J. Vac. Sci. Technol., A
13
,
966
(
1995
).
10.
P. C.
Andriacos
,
C.
Uzoh
,
J. O.
Dukovic
,
J.
Horkans
, and
H.
Deligianni
,
IBM J. Res. Dev.
42
,
567
(
1998
).
11.
Y.
Kuo
and
S.
Lee
,
Vacuum
74
,
473
(
2004
).
12.
Y.
Kuo
and
S.
Lee
,
Appl. Phys. Lett.
78
,
1002
(
2001
).
13.
N. S.
Kulkarni
and
R. T.
DeHoff
,
J. Electrochem. Soc.
149
,
G620
(
2002
).
14.
F.
Wu
,
G.
Levitin
, and
D. W.
Hess
,
J. Electrochem. Soc.
157
,
H474
(
2010
).
15.
P. A.
Tamirisa
,
G.
Levitin
,
N. S.
Kulkarni
, and
D. W.
Hess
,
Microelectron. Eng.
84
,
105
(
2007
).
16.
F.
Wu
,
G.
Levitin
, and
D. W.
Hess
,
ACS Appl. Mater. Interfaces
2
,
2175
(
2010
).
17.
F.
Wu
,
G.
Levitin
, and
D. W.
Hess
,
J. Vac. Sci. Technol., B
29
,
011013
(
2011
).
18.
T.
Abe
,
Y. G.
Hong
, and
M.
Esashi
,
IEEE International Conference on Microelectronics
(
2003
), p.
574
.
19.
T. S.
Choi
and
D. W.
Hess
,
ECS J. Solid State Sci. Technol.
4
,
N3084
(
2014
).
20.
T. S.
Choi
,
G.
Levitin
, and
D. W.
Hess
,
ECS J. Solid State Sci. Technol.
2
,
P506
(
2013
).
21.
H.
Miyazaki
,
J. Vac. Sci. Technol., B
15
,
237
(
1997
).
22.
J. W.
Lee
,
Y. D.
Park
,
J. A.
Childress
,
S. J.
Pearton
,
F.
Sharifi
, and
F.
Ren
,
J. Electrochem. Soc.
145
,
2585
(
1998
).
23.
F.
Wu
,
G.
Levitin
, and
D. W.
Hess
,
ECS Trans.
33
,
157
(
2010
).
24.
F.
Wu
,
G.
Levitin
, and
D. W.
Hess
,
J. Electrochem. Soc.
159
,
H121
(
2012
).
25.
G. C.
Schwartz
and
P. M.
Schaible
,
J. Electrochem. Soc.
130
,
1777
(
1983
).
26.
B. J.
Howard
and
C.
Steinbrüchel
,
Appl. Phys. Lett.
59
,
914
(
1991
).
27.
B. J.
Howard
,
J. Vac. Sci. Technol., A
12
,
1259
(
1994
).
28.
T.
Abe
,
Y. G.
Hong
, and
M.
Esashi
,
J. Vac. Sci. Technol., B
21
,
2159
(
2003
).
29.
F.
Rousseau
,
A.
Jain
,
L.
Perry
,
J.
Farkas
,
T. T.
Kodas
,
M. J.
Hampden-Smith
,
M.
Paffett
, and
R.
Muenchausen
,
Mater. Res. Soc. Symp. Proc
268
,
57
(
1992
).
30.
K. L.
Chavez
and
D. W.
Hess
,
J. Electrochem. Soc.
148
,
G640
(
2001
).
31.
H. L.
Nigg
and
R. I.
Masel
,
Surf. Sci.
409
,
428
(
1998
).
32.
R.
Steger
and
R. I.
Masel
,
Thin Solid Films
342
,
221
(
1999
).
33.
M. A.
George
and
D. W.
Hess
,
J. Electrochem. Soc.
142
,
961
(
1995
).
34.
D.
Panias
,
M.
Taxiarchou
,
I.
Paspaliaris
, and
A.
Kontopoulos
,
Hydrometallurgy
42
,
257
(
1996
).
35.
O.
Çakır
,
H.
Temel
, and
M.
Kiyak
,
J. Mater. Process. Technol.
162–163
,
275
(
2005
).
36.
O.
Cakir
,
J. Mater. Process. Technol.
175
,
63
(
2006
).
37.
J. K.-C.
Chen
,
N. D.
Altieri
,
T.
Kim
,
E.
Chen
,
T.
Lill
,
M.
Shen
, and
J. P.
Chang
,
J. Vac. Sci. Technol., A
35
,
05C305
(
2017
).
38.
T.
Suda
,
N.
Toyoda
,
K.
Hara
, and
I.
Yamada
,
Jpn. J. Appl. Phys., Part 1
51
,
08HA02
(
2012
).
39.
J. K.-C.
Chen
,
N. D.
Altieri
,
T.
Kim
,
T.
Lill
,
M.
Shen
, and
J. P.
Chang
,
J. Vac. Sci. Technol., A
35
,
05C304
(
2017
).
40.
T.
Kim
,
J. K. C.
Chen
, and
J. P.
Chang
,
J. Vac. Sci. Technol., A
32
,
041305
(
2014
).
41.
T.
Kim
,
Y.
Kim
,
J. K. C.
Chen
, and
J. P.
Chang
,
J. Vac. Sci. Technol., A
33
,
021308
(
2015
).
42.
J. D.
Burton
,
S. S.
Jaswal
,
E. Y.
Tsymbal
,
O. N.
Mryasov
, and
O. G.
Heinonen
,
Appl. Phys. Lett.
89
,
142507
(
2006
).
43.
J. Y.
Bae
,
W. C.
Lim
,
H. J.
Kim
,
T. D.
Lee
,
K. W.
Kim
, and
T. W.
Kim
,
J. Appl. Phys.
99
,
08T316
(
2006
).
44.
C. Y.
Yang
,
S. J.
Chang
,
M. H.
Lee
,
K. H.
Shen
,
S. Y.
Yang
,
H. J.
Lin
, and
Y. C.
Tseng
,
Sci. Rep.
5
,
17169
(
2015
).
45.
E. H.
Kim
,
T. Y.
Lee
,
B. C.
Min
, and
C. W.
Chung
,
Thin Solid Films
521
,
216
(
2012
).
46.
I. H.
Lee
,
T. Y.
Lee
, and
C. W.
Chung
,
Vacuum
97
,
49
(
2013
).
47.
S. M.
Hwang
,
A. A.
Garay
,
I. H.
Lee
, and
C. W.
Chung
,
Korean J. Chem. Eng.
31
,
2274
(
2014
).
48.
A. A.
Garay
,
S. M.
Hwang
,
J. H.
Choi
,
B. C.
Min
, and
C. W.
Chung
,
Vacuum
119
,
151
(
2015
).
49.
K.
Kinoshita
,
H.
Utsumi
,
K.
Suemitsu
,
H.
Hada
, and
T.
Sugibayashi
,
Jpn. J. Appl. Phys.
49
,
08JB02
(
2010
).
50.
J.-Y.
Park
,
S.-K.
Kang
,
M.-H.
Jeon
,
M. S.
Jhon
, and
G.-Y.
Yeom
,
J. Electrochem. Soc.
158
,
H1
(
2011
).
51.
Y. B.
Xiao
,
E. H.
Kim
,
S. M.
Kong
, and
C. W.
Chung
,
Thin Solid Films
519
,
6673
(
2011
).
52.
A. A.
Garay
,
J. H.
Choi
,
S. M.
Hwang
, and
C. W.
Chung
,
ECS Solid State Lett.
4
,
P77
(
2015
).
53.
N. D.
Altieri
,
J. K. C.
Chen
, and
J. P.
Chang
, “
Development of a modified halogen-based reactive ion etch process for CoFeB
,”
J. Vac. Sci. Technol., A
35
(unpublished).
54.
Lam Research Introduces Dielectric Atomic Layer Etching Capability for Advanced Logic (Fremont, CA, 2016), available at http://investor.lamresearch.com/releasedetail.cfm?releaseid=987888.
55.
D.
Metzler
,
C.
Li
,
S.
Engelmann
,
R. L.
Bruce
,
E. A.
Joseph
, and
G. S.
Oehrlein
,
J. Chem. Phys.
146
,
052801
(
2017
).
56.
D.
Metzler
,
C.
Li
,
S.
Engelmann
,
R. L.
Bruce
,
E. A.
Joseph
, and
G. S.
Oehrlein
,
J. Vac. Sci. Technol., A
34
,
01B101
(
2016
).
57.
D.
Metzler
,
R. L.
Bruce
,
S.
Engelmann
,
E. A.
Joseph
, and
G. S.
Oehrlein
,
J. Vac. Sci. Technol., A
32
,
020603
(
2014
).
58.
G. S.
Oehrlein
,
D.
Metzler
, and
C.
Li
,
ECS J. Solid State Sci. Technol.
4
,
N5041
(
2015
).
59.
S. M.
George
and
Y.
Lee
,
ACS Nano
10
,
4889
(
2016
).
60.
Y.
Lee
,
J. W.
DuMont
, and
S. M.
George
,
ECS J. Solid State Sci. Technol.
4
,
N5013
(
2015
).
61.
M. A.
George
,
D. W.
Hess
,
S. E.
Beck
,
K.
Young
,
D. A.
Bohling
,
G.
Voloshin
, and
A. P.
Lane
,
J. Electrochem. Soc.
143
,
3257
(
1996
).