Patterning nickel for extreme ultraviolet lithography mask application I. Atomic layer etch processing

Ni is an element that could be found from dental fillings to space shuttles. It is traditionally valued for its corrosion resistance, i.e., as a common additive in stainless steel to promote mechanical properties in harsh environments. For the same reason, mints introduce Ni in coins to prevent depreciation in monetary values during circulation. Its chemical stability and economic affordability also make it a competitive catalyst for many hydrogenation reactions. It also exists in batteries in the form of nickel cadmium (NiCd) and nickel metal hydride (NiMH). Its optical, electrical, and magnetic properties are further explored and leveraged in recent times. Nickel silicide (Ni_xSi_y), for example, is a family of important intermetallic materials for reducing contact resistance at Ni-Si interfaces in microelectronics.¹ 

Among the various advantageous traits listed above, its potential application as a candidate material for next-generation EUV absorbing layer is the focus in this work. Patterns generated from reflection-based optics in vacuum require high contrast with minimum phase deformation. Due to its high refractive index (n) and high extinction coefficient (k) comparing to previous Cr- and Ta-based masks as listed in Table I, both experimental measurements and computational simulations suggest Ni to be a promising choice for the next-generation EUV mask absorber material.⁵ However, its chemical inertness mentioned above implies that Ni cannot be easily and selectively patterned, a prerequisite for its application in EUV.

There has been some reported work on etching Ni, as summarized in Table II, with many originated from the field of metallurgy. While the solution phase processes seem effective in etching Ni, the high etch rates and low selectivity to other materials render these processes incompatible with the requirement of patterning EUV absorbers. The predominant and successful gas-phase etching process relies heavily on either noble ion bombardment or halogenation. The former results in low selectivity and sidewall redeposition,¹⁰ while the latter is prone to potential corrosion of other materials present in the system.¹¹ 

To address these issues and realize Ni patterning capable of delivering patterns for sub−10 nm features, an Ar- and halogen-free gas-phase patterning technique with extreme precision and high selectivity needs to be developed.

Plasma atomic layer etching (ALE) of metals, or more generally on most of the materials with ALE reported, activates target surface via predominately halogenation, and the removal is accomplished via ion bombardment.¹² While a valid etching chemistry on a wide selection of metals, given clearly defined and easily targeted metal-metal and metal-metal halide binding energy differences, this approach could result in higher etch rate on silicon-based hard masks due to the high volatilities of silicon halides. On the other hand, thermal ALE functionalizes the metallic surface into a metal compound, which is removed upon organic ligands exchange.¹³ The viability of thermal ALE is subjected to the reactivity of activation radicals and leaves an isotropic etch profile, not suitable for applications that requires high directionality and vertical sidewall angles. The shortcomings of existing ALE approaches are listed in Table III. A recently proposed plasma-thermal ALE technique¹⁶ discussed the possibility of combining the advantages of both plasma ALE and thermal ALE to realize a highly selective and anisotropic etching chemistry for patterning metal. Using a directional modification step, anisotropic etch profile could be realized even with an isotropic removal chemistry.

To realize ALE, it is essential to have a chemically limited process that self-terminates. In the context of metal etching, a general strategy is to create a chemical contrast near the surface region whereby limiting the reaction when the chemically labile material is completely removed.

Ni has long been regarded as etch-resistant due to its chemical inertness and low gas-phase compound volatilities. This conventional image, however, is challenged by the development in the synthesis of Ni-bearing precursors for chemical vapor deposition (CVD) and atomic layer deposition (ALD) applications such as cyclopentadienylallylnickel¹⁷ and bis(cyclopentadienyl)nickel.¹⁸ The successful synthesis of stable gas-phase Ni-compounds suggests the possibility of Ni removal via ligand exchange processes in a “reverse ALD” manner. For this reaction route to be possible, the conversion of metallic state Ni to oxidized Ni is required since the deposition counterpart reduces the oxidation state of the divalent ion from plus two to zero.¹⁹ Reaction feasibilities between Ni in both metallic and oxidized states and various potential organic etchants were first assessed based on theoretical calculations of Gibbs free energy of reactions from HSC thermodynamic database. Beta-diketone organic etchants such as acetylacetone and hexafluoroacetylacetone were considered since Ni-beta-diketonates precursors are widely used.²⁰ However, the lack of thermochemical data on nickel acetylacetonate and nickel hexafluoro-acetylacetonate suggest the reaction products could be of low stability. It is also noted that metal-carboxylate complexes from carboxylic acids show considerable degree of stability in solution,²¹ thus the reaction favorability using formic acid and acetic acid were also included. The lower Gibbs free energy of reactions of NiO suggests potential chemical contrast when exposing metallic Ni and oxidized Ni to the same chemicals (Table IV).

To assess the etching efficacy of these chemicals, two types of un-patterned Ni samples were prepared. Blanket 40 nm Ni thin film on Si wafer and 110 nm NiO thin film on Si wafer were deposited via sputtering. To assess the etching selectivity to a hard mask of SiO₂, a blanket SiO₂ thin film with an initial thickness of 100 nm was also prepared by plasma enhanced chemical vapor deposition using N₂O and SiH₄ chemistry. One patterned sample was prepared to evaluate the etching anisotropy and profile control: a 40 nm thick Ni film has a hard mask pattern formed by SiO₂ lines with an initial sidewall angle of 60–80° (SiO₂ thickness = 95 nm, line width = 125 nm, and line pitch = 1 μm).

To determine the etch rate, the thickness of Ni was imaged by cross-sectional scanning electronic microscope (SEM, FEI Nova 600, with 10 keV electron beam energy), before and after the etching process. The atomic composition and chemical states of nickel was obtained using x-ray photoelectron spectroscopy (XPS, Kratos XPS Axis Ultra DLD, with 20 eV passing energy). The thicknesses of SiO₂ were determined by spectroscopic ellipsometry (SE, J.A. Woollam LPS-400) using a common optical model for SiO_2.²² 

For solution phase etching, stock solutions of acetic acid, formic acid, and acetylacetone (99%, Sigma-Aldrich) were water bathed to maintain a constant temperature of 80 °C, while hexafluoroacetylacetone acetylacetone (99%, Sigma-Aldrich) was water bathed at a temperature of 60 °C due to its lower boiling temperature (71 °C). Nickel and nickel oxide samples were submerged in the stock solutions for a specified period of time and then dried by nitrogen (UHP, Airgas).

To analyze reaction products, the etched solution was diluted by 1 to 100 volume DI water then ionized in an electrospray-ionization mass spectrometer (ESI-MS, Waters LCT- Premier XE Time of Flight Instrument controlled by masslynx 4.1 software). The instrument was equipped with the multimode ionization source operated in the electrospray mode. An electric field of 2 kV is applied between the spray tip and the orifice. Each mass spectrum is acquired after averaging 10 consecutive scans to improve the signal to noise ratio and account for any flow variations. The measured spectra of the reaction products are compared to theoretically generated mass patterns based on the relative abundance of naturally occurring isotopes for each element.

For gas-phase etching, a cyclic oxidation-formic acid exposure process was developed. For oxidation, an oxygen (UHP, Airgas) plasma was generated at 30 mTorr in an etcher (Unaxis 770) with an RF power of 500 W. The samples were placed on a carrier wafer which received an applied bias ranging from 0 to 50 W, with no additional heating, for 1 to 2 min. To control the gas-phase formic acid exposure, formic acid (95%, Sigma-Aldrich) was stored in a glass ampoule with nitrogen (UHP, Airgas) as a carrier gas. The flow rate of nitrogen gas and formic acid solution were set at 100 sccm and 0.1 g/min, respectively, before entering a temperature-controlled vaporizer (HORIBA MI-1000) at 80 °C. The samples were placed in a custom-built hot-wall reactor where the chamber wall was held at 80 °C for all experiments, while the pressures were adjusted between 150 Torr to 550 Torr. The nominal formic acid exposure time is 60 min, unless otherwise specified. These two process steps (oxidation and formic acid exposure) can be repeated as a cycle, where there existed a 5-min vacuum break between oxidation and formic acid vapor exposure.

Efficient, repeatable, and non-destructive sidewall angle measurement on patterned samples was accomplished by using atomic force microscope (AFM, Bruker FastScan). Most commonly seen roughness-measuring probes are not capable of capturing the sidewall profile due to their relatively large sized tip. Since the patterned samples used in this work has periodic abrupt height transitions, a feature-targeting probe (Bruker VTESPA-300) that has a small radius of curvature (5 nm) and a slanted front angle (2° off the cantilever) was used. To quantify the sidewall profile, an average sidewall angle is reported by performing a linear fit to the sidewall contour in four equally partitioned sections in the bottom half of the nickel film, which translated to about every 5 nm thick sections. To validate the sidewall measurements by AFM, selected samples were also analyzed by cross-sectional high-resolution transmission electron microscopy (HRTEM). The etched pattern samples were treated by focused ion beam (Ga ions at 30 keV) and imaged using a JEOL JEM-ARM300CF system that has the energy dispersive spectroscopy (EDS) for elemental identification.

Etching of nickel in organic solutions was first performed to screen viable chemistries, establish the selectivity, and characterize the reaction products. Thickness for as-deposited Ni after 5, 10, 15 and 30 min of stock solution etch using aforementioned chemicals are shown in Fig. 1(a). Both acetic acid and formic acid showed effective etching of Ni, while acetylacetone and hexafluoroacetylacetone seemed ineffective in etching Ni. A more linear etch rate was observed in formic acid etching Ni (∼1.5 nm/min), as opposed to a more non-linear etch rate of Ni in acetic acid—the etch rate is >3 nm/min initially but decreased to <0.5 nm/min after 10 min. A possible reason is the small volume of the etchant used, which became saturated with the reaction products. Nonetheless, the more linear etch rate obtained with formic acid makes it a more suitable candidate for subsequent analysis. Given the more negative Gibbs free energies of reactions between these chemicals with NiO, the same experiments were carried out with the NiO thin films using formic acid. Due to the thinness of NiO and the faster etch rates of NiO in formic acid, experiments were carried out at 80 and 100 °C for 0.5, 1, 1.5, 2, 2.5, and 5 min. Figure 1(b) shows a rather non-linear dependence on the etching rate of NiO in formic acid solution, possibly due to the lack of a surface cleaning before the etching experiment. At 80 °C, an initial etch rate of ∼20 nm/min was observed in the first 1.5 min, then a much higher etch rate of ∼85 nm/min was observed. Increasing the formic acid temperature to 100 °C, a lower etch rate (both initially and overall) was measured, likely due to the boiling temperature of formic acid (100.8 °C) being close to 100 °C and some etchant evaporated. Comparing the etch rates of NiO to that of Ni, a selectivity of 30–100 in formic acid solution at 80 °C can be determined, depending on if the overall average etch rate or the higher bulk etch rate was used.

The top panel in Fig. 2 shows the theoretically calculated mass pattern for nickel formate [Ni(COOH)₂] with peaks at m/z values of 146.92 (69.1%), 148.92 (27.0%), and 150.92 (3.9%). Since hydrogen, carbon, and oxygen all have a dominant isotope (>99%), the calculated mass pattern is primarily due to the presence of three dominant Ni isotopes (⁵⁸Ni = 68%,⁶⁰Ni = 26%, and ⁶²Ni = 4%). The bottom panel in Fig. 2 shows the actual mass spectrum of the analysis. A much more complex mass pattern is observed, compared to the theoretically calculated one. The additional peaks could be due to deprotonation during the ionization process. In fact, the experimentally measured mass pattern can be matched quite closely to the theoretical prediction if the following two species are considered: NiHC₂O₄ and Ni(COO)₂. These are derived from nickel formate which loses one or two hydrogen atoms. Each of these three compounds is color coded with the appropriate proportionality, and it is clear that the sum of the mass pattern matches the experimentally acquired spectrum well. Little work has been reported on ESI-MS of nickel formate, and a considerable amount of complexity has been noted for work reported on other nickel-containing complexes, and partial oxidation states has been discovered on Ni possibly due to “sharing of the extra electron among the metal ion and the ligands”.²³ 

In reviewing available literature concerning Ni-complexes regarding their potential applications in soldering,¹⁹ catalysis,²⁴ and as CVD precursors,²⁴ the physical properties of nickel acetate have been detailed,²⁵ indicating a stable complex in condensed form. The solubility of nickel acetylacetate was studied, with anhydrous Ni(acac)₂ forming trimers to achieve an octahedral coordination around each nickel atom. When water is present, each Ni(acac)₂ prefers to form an octahedral complex with two water molecules.²⁵ Volatility of Ni(hfac)₂ was cited to be higher than that of Ni(acac)₂ due to the weaker intermolecular hydrogen bonding and van der Waals force.¹⁹ Regardless of the type of ligands bonded to Ni atom, multiple literature reports emphasized on the role of oxidizing agents, either as designed reactant¹⁹ or as source of adventitious reactions.²⁶ The formation of nickel oxide, either from oxidizing nickel in metallic state or from decomposition of nickel-complex implies that the potential reaction path could be explored with controlled surface oxidation. This is consistent with the observed high selectivity between etching NiO and Ni in formic acid, making the creation of a chemical contrast in nickel by oxidation a viable approach to achieve ALE.

Calculations based on available thermodynamic data indicate spontaneous reactions between gas-phase formic acid with both metallic nickel and nickel oxide. However, given the much reduced reactant concentration in a vacuum environment, very little etching was observed in the gas phase with metallic nickel: a thickness reduction of about 1 nm, which is within the detection limit of the SEM measurement, may have occurred at a chamber pressure of 350 Torr and a chamber temperature of 80 C for 3 h [Fig. 3(a)]. Shown in the same figure, it is noted that the same chemistry has no effect on etching SiO₂. The work was then focused on the use of an oxygen plasma to promote surface oxidation with a substrate bias of 50 W, creating an oxide layer, which was then exposed to formic acid vapor at 350 Torr and 80 °C. The data shown in Fig. 3(a) are the etched thickness of nickel, as measured by SEM, as a function of the cumulative formic acid exposure time in this cyclic process (since the nominal formic acid exposure time per cycle is 60 min, this corresponded to three cycles). A very linear etch rate was observed with this cyclic process, with an etch rate of 0.1 nm/min. The lower etch rate compared to that in the solution is attributable to the reduced reactant concentration. The oxidation step is not self-limiting; the self-limiting reaction step is where the organic vapor removes the oxides, as very high etching selectivity were recorded on NiO to Ni (∼100) and NiO to hard masks (essentially infinite) at 80 °C and 350 Torr. The effectiveness of this cyclic etching process is also confirmed by the XPS analysis of the etched Ni samples. As shown in Fig. 3(b), there are many features in the measured XPS spectra, namely, the Ni 2p core level emission in various chemical states (colored solid lines) as well as the satellite peaks (black dash lines). To differentiate these chemical states, a detailed XPS analysis was done with reference spectra taken on clean metallic Ni and fully oxidized Ni thin films in the same XPS system.²⁷ By fitting the Ni 2p_3/2 and 2p_1/2 regions simultaneously with the fixed spin orbital split and ratio, these features were assigned to nickel in metallic (852.6 eV), oxide (853.7 eV), and hydroxide (856.0 eV) forms. The formation of Ni(OH)₂ has been widely observed with the presence of moisture in the air.²⁸ Given the purity of formic acid used in the experiment and the base pressure of the reactor, nickel hydroxide formation was expected. The formation of higher oxidation state oxide [NI(III)] was not considered, as “attempts to prepare Ni₂O₃ by heating the hydroxide, basic carbonate or nitrate of nickel in air or oxygen resulted in the formation of NiO only.”²⁹ It has been suggested in the literature that the differentiation between Ni(II) and Ni (III) could be realized by measuring electrical conductivity;³⁰ however, this was not viable with such thin films. Finally, the possibility of incomplete removal of the reaction production and redeposition of the reaction products was also considered in analyzing the XPS data. The redeposition of etch product is possible; however, the EDS map from TEM analysis detailed later did not show Ni over the hard mask, which suggests that this is unlikely. The incomplete removal of the reaction products is also possible, but the XPS resolution did not allow such differentiation.

Figure 3(b) clearly shows that the signal intensities of both metallic Ni and NiO decreased significantly, confirming removal of nickel after cyclic treatments. The gas-phase etch product identification and characterization by the in situ mass spectrometry analysis were attempted but not successful due to the very low etch rates as a result of very low concentrations of the reaction products. Though the melting temperature of nickel formate is reported to be ∼140 °C at STP condition,³¹ as etching in a reduced pressure leads to a gaseous product, the reaction would be thermodynamically more favorable due to the entropy gain.

The effects of substrate bias during oxidation and the chamber pressure during formic acid exposure were examined in this work. Since a higher substrate bias leads to higher energy ions, which could physically sputter the hard mask, the effect of lower energy ions was also evaluated. The lowest attainable ion energy in this work is equivalent to the plasma self-bias, when the applied bias was set to zero. Figure 4 shows compositional analysis of the etched nickel surface where the oxide was formed with an oxygen plasma with zero substrate bias, while the formic acid exposure was carried out at various formic acid pressures. The ex situ XPS analysis included Ni 2p, C 1 s, and O 1 s spectra that are deconvoluted to show the chemical bonding configurations on the etched surface. Qualitatively, the spectra obtained after the cyclic etch process at 550 Torr were distinct from the other two spectra. The amount of surface oxide remained could be quantitatively represented by the signal intensity ratio of NiO over that of Ni, which is 0.80, 0.82, and 0.45 at formic acid exposure pressures of 150 Torr, 350 Torr, and 550 Torr, respectively. Based on the attenuation length of electrons in Ni and NiO,²⁷ the oxide thicknesses remained on the surface were 5.3 nm, 5.2 nm, and 0 nm (due to absent Ni-O peak in O 1 s) at 150 Torr, 350 Torr, and 550 Torr, respectively, based on references to clean metallic nickel and thick nickel oxide references.²⁸ Since all the oxidation half-cycles were performed under the same conditions in this set of experiments, the calculated oxide thickness suggested incomplete removal of the oxidized nickel at lower pressure conditions but complete removal of oxidized nickel at the highest pressure used.

To assess the profile control of the developed ALE process, pattern samples were used. Processing pressures ranging from 150 to 550 Torr were evaluated, while the chamber temperature was maintained at 80 C. In performing the AFM measurement, due to the slanted tip used, the sidewall angle analysis focuses on the side of the feature where the tip points toward. While repeating measurement on the exact locations of the pattern was difficult, efforts were made to keep the measurement position as consistent as possible. It is also noted that the initial sidewall angles varied between 60 and 80°, which is another parameter to consider in analyzing the results. Finally, since the calibration experiments confirmed that formic acid vapor does not etch SiO₂ hard mask [as shown in Fig. 3(a)], the AFM measured height change is fully attributed to the etching of NiO (which is also corroborated by HRTEM analysis).

Figure 5 shows the AFM contour plots for patterned samples etched by the plasma-thermal cyclic process at three different pressures and after various cycles (only results after two and six cycles are shown for clarity). At a low pressure (150 Torr), some changes in the sidewall profile may have been resulted but there is no significant reduction in the Ni film thickness in the open area. This is likely due to the low concentration of the reactant at this pressure and is consistent with the XPS results shown in Fig. 4. As the pressure increased to 350 Torr, a measurable etch rate of 3 nm/cycle was recorded, with some erosion on the corner of the hard mask. As the pressure increased further to 550 Torr, a much more substantial etching was observed, with an overall higher etch rate of 6 nm/cycle, consistent with the compositional analysis by XPS on etched blanked samples. More importantly, the average sidewall angle after six cycle of treatment was measured to be 87°, making it possible to categorize the etching as anisotropic.

Figure 6 summarizes the results by evaluating the remaining Ni layer thickness and sidewall angle as a function of the number of cycles of the plasma-thermal process. The shaded gray region at the top represents the targeted sidewall angle (87°−90°). It is observed that processing at the lower formic acid pressure did not reach the etching end point, leaving a less anisotropic sidewall profile. The etch rate as well as the sidewall angle increased as the formic acid pressure increased. The repetition of cycles at higher pressures allow more thorough removal of the oxidized region, which is anisotropic due to the directionality of the oxygen ions when traversing through the plasma sheath, thereby reaching the etching end point after six cycles.

To confirm whether etching is anisotropic with higher precision, as well as to inspect the elemental distribution of the final feature, HRTEM and EDS were performed on a sample treated by 6 cycles of ALE with an oxidation step at zero bias and a pressure of 550 Torr during formic acid exposure, as shown in Fig. 7. HRTEM analysis confirmed the vertical Ni sidewall (87°), as determined by AFM measurements. The EDS showed no Ni over the hard mask, suggesting that redeposition of the etch products is unlikely. The EDS also showed that the exposed sidewalls are slightly oxidized, which is expected due to ambient exposure before the HRTEM analysis. Comparing the oxygen and nickel EDS mapping, the oxidized sidewall thickness is on the order of 10 nm, which is about the same as the normally observed native oxide layer thickness due to ambient exposure (∼5 nm). It is concluded that this ALE process is capable of etching patterned Ni thin film with high selectivity and anisotropy.

A highly selective and directional removal of Ni is demonstrated via controlled oxidation and organic acid exposure. An etching selectivity of greater than 100 was recorded for etching of nickel oxide over metallic nickel using gas-phase formic acid at 80 °C. Infinite selectivity is measured for etching of nickel oxide and silicon dioxide hard mask. Directional surface oxidation was employed to leverage such selectivity and establish chemical contrast between nickel and its oxide at defined openings. The demonstrated method is proven to be uniform over a micrometer range and is capable of realizing complete Ni removal with 87° final sidewall angle.

The authors acknowledge the financial support from the Semiconductor Research Corporation (SRC, No. 2802.001), National Science Foundation (No. 1805112), Lam Research, and Center of Design-Enabled Nanofabrication (CDEN). The authors thank Dr. Changju Choi and Dr. Tristan Tronic at Intel as well as Dr. Nathan Marchack at IBM for providing patterned samples. The authors thank Dr. Gregory Khitrov at UCLA for providing guidance on ESI-MS data analysis, Dr. Adam Steig at UCLA for providing guidance on AFM measurements, and Dr. Mingjie Xu at UCI for helping with TEM imaging and EDS.