Communication—Electrochemical Atomic Layer Etching of Copper

A novel process for the electrochemical atomic layer etching (e–ALE) of copper (Cu) is presented. In this process, Cu ﬁrst undergoes surface-limited sulﬁdization to form a monolayer of copper sulﬁde (Cu 2 S). The Cu 2 S layer is then selectively etched in hydrochloric acid without etching the underlying Cu. The steps of surface-limited sulﬁdization of Cu and selective etching of the resulting Cu 2 S are repeated sequentially to achieve a net etch rate of close to one Cu monolayer etched per e–ALE cycle. Surface-limited etching is shown to minimize roughness ampliﬁcation thereby preserving the near-atomic ﬂatness of the original Cu electrode.

Atomic layer etching (ALE) processes are critically important for the precise tailoring of materials and structures in nano-electronics. 1 For atomically precise etching of metals, plasma-based approaches are available which generate nonvolatile etch products thereby contaminating the metal surface. Hess and co-workers have developed a two-step process that etches copper (Cu) films with chlorine and hydrogen plasmas at low temperature (below 20 • C). This process generates a volatile etch product that minimizes surface contamination. 2 In most plasma-based approaches, the metal etching rate is higher than 1 nm per etch cycle. Such high etch rates do not provide the requisite atomic-scale control over etching required in ALE. While plasma-assisted ALE processes for oxides are mature, 1 ALE of metals is still in its infancy and numerous development efforts are currently underway. 3 In this communication, we report on an electrochemical approach for the layer-by-layer etching of Cu with atom-scale control over the etching rate. The two-step approach consists of surface-limited electrochemical sulfidization of Cu followed by selective etching of the resulting copper sulfide (Cu 2 S) monolayer. Surface-limited sulfidization has been used previously for fabricating semiconductors. 4 Feasibility of the electrochemical ALE of Cu is demonstrated and process performance parameters (etch rate, surface roughness) are characterized.

Experimental
Cyclic voltammetry and chronoamperometry.-Surface-limited sulfidization of Cu was studied using cyclic voltammetry (CV) performed using a three-electrode cell consisting of a sputter-deposited Cu substrate as the working electrode, a platinum (Pt) wire as the counter electrode, and a saturated Ag/AgCl reference electrode (Fisher Scientific). The electrolyte contained 0.1 M potassium hydroxide (KOH, Fisher Chemical) and 0.5 mM sodium sulfide (Na 2 S, Sigma-Aldrich) and was prepared using de-aerated DI water. A VersaSTAT 3 potentiostat was used for all electroanalytical measurements. All potentials reported below are referenced to the standard hydrogen electrode (SHE). The Cu substrate was rinsed first with ethanol and then with DI water before drying with N 2 . The cleaned substrate was immersed in 2 M sulfuric acid (H 2 SO 4 , Fisher Scientific) for 1 min to remove surface Cu oxides. A potential of -1.2 V vs. SHE was applied for 100 s to further reduce Cu oxides before the working electrode potential was scanned (at 20 mV/s) from -1.2 V to -0.6 V vs. SHE and back. Chronoamperometry was performed at an applied potential of -0.75 V vs. SHE. * Electrochemical Society Student Member.
* * Electrochemical Society Member. z E-mail: rna3@case.edu Electrochemical atomic layer etching.-Electrochemical ALE (e-ALE) of Cu was performed on a Cu substrate fabricated using electrochemical atomic layer deposition (e-ALD) detailed elsewhere. 5 Cu e-ALD for 10 cycles was performed on a sputter-deposited Ru substrate to form a ∼2 nm Cu film with RMS surface roughness of ∼0.2 nm. Surface-limited sulfidization of Cu was performed at -0.75 V vs. SHE in an alkaline Na 2 S-containing electrolyte (composition reported above). The substrate was then removed from the sulfidization electrolyte, rinsed with de-aerated DI water, dried under N 2 , and transferred immediately to the etching electrolyte. The Cu x S layer formed during sulfidization was selectively etched by immersion in de-aerated 2 M HCl for 30 s. This process did not etch the underlying Cu. After selective etching, the Cu substrate was again rinsed with de-aerated DI water and dried under N 2 before it was transferred back to the sulfidization electrolyte. The sulfidization and selective etching steps were repeated to achieve layer-by-layer removal of Cu with atomic precision. The surface-limiting signature of the sulfidization step was confirmed by performing e-ALE on sputter-deposited Cu substrates.
Anodic stripping coulometry.-After e-ALE was performed, the Cu remaining on the electrode was electrochemically stripped in 50 mM H 2 SO 4 by first scanning the electrode potential from OCP (∼0.4 V) to 0.6 V vs. SHE at 20 mV/s and then holding the potential at 0.6 V until the stripping current dropped to zero. Since hydrogen evolution was thermodynamically prohibited at 0.6 V, the stripping current measured was solely due to Cu dissolution. The integrated stripping charge density provided via Faraday's law the mass of Cu remaining after various number of e-ALE cycles. Substrate RMS roughness before and after e-ALE was measured using Dimension 3100 (Veeco Digital Instruments) AFM.

Results and Discussion
Surface-limited sulfidization of copper.-CV of a Cu substrate in an electrolyte containing 0.5 mM Na 2 S and 0.1 M KOH (pH = 13.2) is shown in Fig. 1a (blue). The background CV without Na 2 S in the electrolyte is also shown (red). The positive limit of the potential scan was set at -0.6 V to prevent surface Cu x O formation. 6 Hydrogen evolution was observed at potentials negative with respect to -1.0 V. In the anodic scan direction, an oxidation peak was observed near -0.68 V corresponding to bulk Cu x S film formation. A relatively small oxidation peak located at -0.98 V corresponding to surfacelimited sulfidization of Cu was also observed consistent with prior reports of sulfur adlayer formation on Cu at such potentials. 6,7 In the cathodic (reverse) scan direction, a peak appeared at -0.93 V indicating reduction of bulk Cu x S.
To demonstrate surface-limiting characteristics of the proposed sulfidization scheme, we performed chronoamperometry on Cu at -0.75 V. This potential is anodic with respect to the small oxidation peak at -0.98 V but cathodic with respect to bulk Cu x S formation at around -0.68 V (Fig. 1a). As seen in Fig. 1b, after an initial unsteady state behavior which can be attributed to transient diffusion and electro-nucleation processes, 8 the current eventually (after ∼35 s) decays to nearly zero. This indicates self-terminating behavior once the electrode surface is covered with a monolayer of Cu x S. The integrated charge density associated with this self-limiting surface sulfidization was Q sat ≈ 300 μC/cm 2 and in agreement with previous reports. 7 Assuming surface molar density of Cu(111) to be about N = 2.96 nmol/cm 2 , the measured charge density Q sat corresponds to a Cu oxidation state of n = Q sat /NF = 1.05 where F is the Faraday's constant. This suggests the formation of a monolayer of Cu 2 S (x ≈ 2) on Cu. 7 Fig.  2. In this sequence, surface-limited sulfidization of Cu and selective etching of the formed Cu 2 S by HCl are repeated sequentially:

Electrochemical atomic layer etching (e-ALE) of copper.-A schematic representation of the Cu e-ALE sequence is provided in
(i) Surface-limited sulfidization of Cu: In Na 2 S-containing electrolytes, solution-phase equilibrium favors formation of hydrosulfide species (HSaq ). 7 At suitable electrode potentials, i.e., -0.75 V vs. SHE, the Cu sulfidization reaction is facilitated: 6,9 2Cu When the surface Cu atoms are oxidized completely to Cu 2 S, Reaction 1 self-terminates as observed in Fig. 1b The chloro-complex formed in Reaction 2 may also exist as CuCl 2 as described elsewhere. 11 Note that, in the absence of dissolved oxygen, Cu 0 cannot be oxidized. Thus, as Cu 2 S is etched via Reaction 2, the exposed underlying Cu remains protected. When the surface Cu 2 S is entirely removed, Reaction 2 is also terminated and the subsequent e-ALE cycle begins.
The e-ALE sequence described above was characterized by electrochemical techniques. First, the surface-limited sulfidization signature, i.e., the decay of the oxidation current to zero in ∼35 s, is confirmed via chronoamperometry during step (i) of the 1 st -5 th cycles of e-ALE (Fig. 3). During each e-ALE cycle, the integrated sulfidization charge density (Fig. 3 inset) remained fairly constant  (except for the 1 st cycle) indicative of quite repeatable sulfidization and etching steps. Anodic stripping coulometry was also performed to determine the amount of Cu remaining after various e-ALE cycles. The integrated Cu stripping charge density (Q) after various e-ALE cycles normalized to the charge density of the original un-etched Cu electrode prepared by e-ALD (Q 0 ) is shown in Fig. 4. It is observed that Q/Q 0 decreases linearly with increasing e-ALE cycles. From the slope of this curve, an equivalent etch rate was computed to be 0.103 (10.3%) per etch cycle. Since the original un-etched electrode was prepared by depositing 10 atomic layers of Cu via e-ALD, 5 one expects ∼10 cycles of e-ALE to be required to etch this film completely which is consistent with the measured etch rate of ∼10% per e-ALE cycle. This suggests an etch rate of ∼1 Cu monolayer per e-ALE cycle. Table I shows surface RMS roughness (from AFM) after various e-ALE cycles. The RMS roughness of the original un-etched Cu surface was ∼0.2 nm. 5 After 4 and 6 cycles of e-ALE, the measured RMS roughness of the electrode remained in the ∼0.2 nm range and thus relatively unchanged compared to the un-etched Cu surface. In control experiments, Cu samples subjected to step (ii) but not to step (i) exhibited significant roughness amplification. These observations confirm that e-ALE facilitates atomic layer-by-layer etching of Cu with negligible surface roughness amplification.

Conclusions
Feasibility of the electrochemical atomic layer etching of Cu is demonstrated. The e-ALE process consists of two steps: (i) surfacelimited sulfidization of Cu to form Cu 2 S followed by (ii) selective etching of Cu 2 S in HCl. The e-ALE process provides an etch rate of close to 1 Cu monolayer per e-ALE cycle. Furthermore, substrate RMS roughness is not amplified during e-ALE suggesting a layerby-layer etching mode. The e-ALE approach presented herein opens new avenues for atomically precise tailoring of surfaces in advanced nano-electronics.