Large-scale nanopatterning of metal surfaces by target-ion induced plasma sputtering (TIPS)

Abstract

Target-ion induced plasma sputtering (TIPS) is a one-step self-organization nanofabrication method in which a conventional DC magnetron sputter with a negative substrate bias voltage is used. This process successfully leads to ion-induced sputtering on metal substrates, producing large-scale nanopatterns on various metal surfaces. We demonstrated that the obtained nanopatterns have size-tunability from nano- to micro-scales by modulating the negative substrate voltage. This large-scale, bottom-up nanofabrication technique will accelerate nanopattern applications in biomedical, chemical or magnetic devices through large surface-to-volume ratios and unique surface topography of induced ripple patterns on metals.

---

Top-down or bottom-up nanofabrication techniques have been actively developed in order to improve the accuracy, controllability, and scalability of nanopatterns, as well as the cost-efficiency of the nanofabrication process.^1,2 Such advances in nanofabrication techniques have accelerated the development of information technologies through the fabrication of silicon-based micro- and nanoelectronic devices.¹ On the other hand, the demand for the nanopatterning of metal surfaces is increasing in areas ranging from energy harvesting storage, fuel cells and electronics to biosensors and biomedical implants due to their large surface-to-volume ratios and unique surface topography.^3–5 These emerging applications tend to be more defect-tolerant and cost-sensitive compared to micro and nanoelectronic devices.^6,7 Therefore, new approaches to nanopatterned metal surfaces are required in a simpler and more cost-effective manner for defect-tolerant devices. In this context, bottom-up self-assembly approaches have significant potential due to their ability to create large-scale nanopatterns while maintaining the ease and cost-efficiency of processing.^8,9

The most feasible self-assembly approaches for large-scale nanopatterned metal surfaces include template self-assembly (TSA)¹⁰ and self-organization by ion beam erosion.¹¹ The TSA process is a nanofabrication method using a self-assembled thin coating layer as a template which guides the pattern formation of the underlying substrate surface in conjunction with various lithography techniques. The self-assembled template is often easily synthesized and processed with good scalability, chemical functionality, and size tunability of patterns. In particular, the pattern size of the template is easily adjusted by the selection of self-assembled materials, ranging from a few nanometers to a few hundreds of micrometers.¹⁰ However, nanopatterning through TSA inherently requires multiple processing steps due to the use of the template, which increases inconvenience and manufacturing cost. On the other hand, self-organization by ion beam erosion is a single-step nanofabrication method which can directly create self-assembled nanopatterns on a metal substrate through an incident ion beam.¹¹ The observed self-assembled nanopatterns are often either a regular array of dots or ripples. The topography and size of nanopatterns are also controlled by incident ion beam (e.g., ion species, ion energy, ion fluence, and incident angle), substrate conditions (e.g., substrate material, temperature) as well as existence of a target (e.g., target material, location).^11–13 However, this method has limitations on the fabrication of large-scale, homogeneous nanopatterns for various metal surfaces, particularly with complex geometry, due to the strong dependence of pattern formation on the incident angles of the ion beam.

Therefore, in this study, we report, for the first time, that large-scale, self-assembled nanopatterns can be directly fabricated on various metal surfaces from nanoscale to microscale via plasma sputtering, using a conventional direct current (DC) magnetron sputter. The DC magnetron sputter has been widely used for thin film coating of a substrate, where target atoms or ions sputtered by plasma are deposited on the substrate.¹⁴ On the other hand, we found that by modulating negative substrate bias voltage, the DC magnetron sputter can induce ion bombardment on the substrate in which incoming ions are accelerated to the substrate via negative substrate bias voltage, mimicking the ripple formation mechanism by ion beam erosion (Fig. S1 in ESI†).¹⁵ More importantly, the target ions entering the substrate lead to a compositional variation on the substrate, manipulating local sputtering yields. By controlling the differences in sputtering yields, the characteristic dimensions of the patterns can possibly be tuned from nanoscale to microscale, which is hardly achieved by normal ion beam erosion. Thus, this plasma sputtering process was denoted as target-ion induced plasma sputtering (TIPS). Herein, this TIPS process was directly applied to the nanopatterning of three metal alloy systems, cobalt-chromium (Co–Cr), nickel titanium (NiTi), and stainless steel (SUS 316L), which are commonly used in industry and medical prosthetics. As a target material, tantalum (Ta) was chosen because of its high corrosion resistance and good biocompatibility so that its incorporation during TIPS can potentially improve the corrosive and biological behavior of alloy systems.^16,17 The morphology of nanopatterns was investigated by varying the applied negative substrate bias voltage with corresponding compositional analysis through the depth of the substrate. Also, the nanopatterns were directly created on the surface of Co–Cr meshed tubes through TIPS in order to confirm the feasibility of TIPS for various metal applications.

Large-scale homogeneous nanopatterns on the Co–Cr surface were successfully created by TIPS with the negative substrate voltage of 800 V after 90 min of processing time as shown in Fig. 1. The detailed structural evolution of nanopatterns during the 90 min processing time was indicated in Fig. 2. At the beginning of the process, a nanoscale ripple structure with an average depth of ∼20 nm was observed on the surface, which was found to be identical to the ripple patterns induced by normal ion beam erosion in size and shape.¹⁰ After 5 min, the Co–Cr surface appeared to have prominent nanopatterns since the increased depth of the ripples (>100 nm) made the surface topography more visible. As the processing further proceeded, irregularly shaped clusters with an average diameter of ∼500 nm started being formed on the top of the ripple structure and finally covered the entire surface. The surface became nanoporous with a pore depth of ∼150 nm (Fig. 2A(c)). The nanopores were continuously lengthened by merging with neighboring pores, again forming a ripple structure (Fig. 2A(d–f)).

Fig. 1 Large-scale nanopatterned Co–Cr surface by TIPS with substrate voltage of −800 V after 90 min of processing time.

Fig. 2 (A) Top-surface and cross-sectional SEM images of Co–Cr obtained from different process time ((a) 1 min, (b) 5 min, (c) 10 min, (d) 30 min, (e) 60 min, and (f) 90 min) at the negative substrate voltage of 800 V. Inset of (a) corresponding to high-resolution magnification AFM phase image and insets of (c) to (f) corresponding to high-resolution cross-sectional SEM images. (B) Width and depth and (C) lateral length of ripple-like nanopatterns the Co–Cr over 90 min of processing time.

At steady state, the width and depth of nanopatterns reached the max values except for the continuous lengthening mechanism of ripples (Fig. 2B and C). According to the continuum theory of ion beam erosion,¹⁸ the dominant wavelength (or width) of ripples is dependent on ion energy, which is correlated with the applied substrate voltage in this study. Therefore, the steady-state width of nanopores was determined by the substrate voltage. The depth of nanopatterns from TIPS is closely related to the co-deposited Ta ions on the Co–Cr substrate during sputtering as reported in the surfactant-driven ion beam erosion.¹⁹ Equilibrium of etching rates between the top and bottom surfaces of ripples was induced by the concentration difference of Ta ions and may eventually lead to the steady-state depth of the substrate. On the other hand, the channel length of nanopores was continuously increased over time in the direction for minimizing surface energy. The topography evolution of the ripple structure by TIPS appeared to be similar to the ripple formation by ion beam erosion over time, but the depth of the nanopatterns was significantly amplified by a factor of 5–10.^19,20 We called the resultant nanoporous structure the ‘amplified ripple structure’.

For TIPS, the negative bias voltage applied to a metal substrate influences the amount of ion flux and ion energy simultaneously, because an induced negative electric field around the conductive metal substrate accelerates ions toward the substrate. The effect of negative substrate voltage on the nanopatterns of Co–Cr was closely examined at the steady-state conditions by varying the voltages from 0 V to −1600 V (Fig. 3). The threshold voltage (V_th) for nanopatterning by TIPS appeared to be −200 V. At V < V_th, a dense, smooth Ta film was formed on the Co–Cr substrate, since the Ta deposition rate exceeded the sputtering rate of the substrate (the deposition mode), as shown in Fig. 3A(a). On the other hand, at V > V_th, ripple-like nanopatterns were clearly observed on the substrate surface. With increased substrate voltage, the sputtering rate exceeded the deposition rate (the etching mode), as shown in Fig. 3A(b–f). Moreover, as the negative substrate voltage increased, the size of the ripple patterns increased while the pattern shape was still maintained. Two important structural parameters of nanopatterns, width (w) and depth (d), were also quantified in relation to the applied substrate voltage through image analysis (Fig. 3B). Both width and depth were found to significantly increase as substrate voltage increased, with the width increasing almost linearly with depth.

Fig. 3 (A) Top-surface SEM images of Co–Cr obtained from different substrate bias voltage after 60 min of TIPS ((a) −200 V, (b) −400 V, (c) −600 V, (d) −800 V, (e) −1200 V, and (f) −1600 V). Insets of (b) to (d) corresponding to high-resolution cross-sectional SEM images. (B) Depth and width of amplified ripples on the Co–Cr as a function of substrate bias voltage from −400 to −800 V.

As TIPS proceeded, the nanopatterns on Co–Cr became homogeneous, showing the amplified ripple structure. The cross-section STEM image of the amplified ripple structure at the steady state is shown in Fig. 4A with the corresponding EDX elemental analysis through the ripple from the surface (Fig. S2 in ESI†). The STEM image of the ripple clearly exhibits two regions with different contrasts. EDX analysis of those two regions indicates that the brighter region has ∼50 atom% Ta in addition to Co and Cr, whereas the darker region is pure Co–Cr without Ta. Within the Ta-rich region, the concentration of Ta was increased towards the surface because the penetration efficiency of Ta ions is decreased with the increased penetration depth during ion implantation induced by TIPS.^19,21,22 However, the maximum concentration of Ta did not exceed ∼60 atom%. The difference between the local sputtering yields of the substrate due to incorporated target ions was modeled as follows:

ΔY ∼ Δ(j_D/j_ionY^bulk_S)Y^bulk_substrate = δΔC_dY^bulk_substrate

where j_D and j_ion are the target-ion deposition flux and the target-ion flux, respectively, and Y^bulk_S and Y^bulk_substrate are the target-ion sputtering yield and pure substrate sputtering yield, respectively.¹⁸ The difference between target-ion deposition flux (j_D) and target-ion sputtering flux (j_ionY^bulk_S) is related to the amount of deposited target ions at each region. Thus, the ratio of target-ion deposition flux and target-ion sputtering flux can be expressed with C_d and δ, which are the local concentration of deposited target-ions on the substrate, and the ion-flux related coefficient, respectively. The larger difference in local target-ion concentrations induces a larger difference in sputtering yield, and results in a higher pattern depth (d) (ΔY ∝ Δd). The different sputtering yield was attributed to the evolution of amplified ripples over time. The Ta-rich region started appearing after 5 min of processing time, at which time irregularly shaped clusters coexisted with ripples on the surface (Fig. S3 in ESI†). EDX spot analysis of the surface layer (p1) and the substrate (p2) for a ripple shows that the thin surface layer had ∼20 atom% Ta, while the underlying substrate was Co–Cr only, without Ta. However, in the case of the clusters, the Ta-incorporated layer (p3) was observed on the wavy Co–Cr substrate (p4) with a larger Ta amount (∼50 atom% Ta) compared to the ripples. The significant difference of Ta concentration between the surface layers of ripples and clusters indicates inhomogeneous Ta deposition at this stage. Thus, cluster regions have a lower sputtering yield than ripples, leading to the formation of clusters with higher amplitude.

Fig. 4 (A) STEM image and on Co–Cr after 60 min of TIPS with compositional profile through the distance from the surface of amplified ripples, (B) corresponding electron diffraction patterns (q1 to q3) of amplified ripples of three spots (q1 to q3) indicated in the STEM image of (A), and (C) steady-state depth of the ripples as a function of voltage-dependent sputtering yield coefficient, S(V_s).

The diffraction patterns of three chosen spots (q1–q3) within a ripple (Fig. 4A) are shown in Fig. 4B. The q3 region was crystalline while the q1 region was amorphous. The q2 region was mixed with crystalline and amorphous phases. Ta incorporation into the ripples changed the crystal phase of Co–Cr into an amorphous phase of Co–Cr–Ta. The incorporated Ta with high ion energy onto the surface of the substrate caused crystallographic damage, completely amorphizing the surface of the substrate.²³ This amorphization allows large-scale homogeneous nanopatterned metal surfaces without any clear crystallographic dependence (e.g., surface orientation, grain boundaries). The substrate temperature (T_s) is known to influence the amorphization of the metal substrate, associated with the diffusion rates of substrate atoms or target ions.²⁴ The inter-layer mass transfer becomes more active above an activation barrier which is related to the substrate temperature. For this study, the diffusion of surface atoms did not act as a limiting factor of the pattern formation at T_s > 300 K because the pattern morphology and size showed no temperature dependence when T_s was varied from 300 to 500 K.

The size of patterns was analyzed in terms of sputtering yields based on applied negative substrate voltage (V_s). Based on the EDX analysis, the difference in Ta concentrations (ΔC_d) between the top and bottom surfaces remained almost constant with the increment of V_s since the Ta concentrations of the top and bottom surfaces was ∼50 atom% and 0 atom%, respectively under various substrate voltage conditions. The depth of ripple patterns is proportional to the difference in sputtering yield (ΔY ∝ Δd).¹⁹ Sputtering yield is also proportional to ion energy, E(V_s) = qV_s and is often expressed as follows:

Y^bulk_substrate(V_s) = C(1 − √(V_th,s/V_s))^a = CS(V_s)

where V_th,s and V_s are the threshold substrate bias voltage for sputtering and substrate bias voltage, respectively, C is the constant, and a = 2.8 for Rutherford scattering.²⁵ Therefore, setting the term, (1 − √(V_th,s/V_s)), as the voltage-dependent sputtering yield coefficient, S(V_s), the pattern depths were correlated with S(V_s) as shown in Fig. 4C, assuming a = 2.8. The curve fitting clearly indicates that pattern depths were linearly proportional to S(V_s). The theoretical prediction shows good agreement with the experimental results. Therefore, the length scale of ripple patterns can be tuned by substrate voltage.

Interestingly, optical property of TIPS-treated Co–Cr surfaces was found to be varied with the pattern size of induced ripples (Fig. S4 in ESI†). Dramatic increase of the light absorption was observed along with the increased pattern size of ripples, which eventually led to a surface blackening effect.^26,27 In particular, the remarkable pore cavity of ripple structures with larger pore gap than 400–1400 nm efficiently trap and absorb the visible and near-infrared regions of the incident light spectrum during the iterative reflection of the trapped light, which resulted in the prominent blackening effect on the patterned surfaces. Moreover, other morphological features (e.g. alignment, curvature) of the ripples might also cause the blackening effect.

The TIPS process was applied to (1) different metal substrates (NiTi, SUS 316L), (2) the Co–Cr substrate with complex geometry and (3) selective nanopatterning. For (1), nanopatterns on both NiTi and SUS 316L were successfully formed via TIPS with the substrate voltage of −800 V after 60 min of processing time (Fig. 5A). The evolution of ripple structures varied depending on the substrate materials, apparently because of the different sputtering yields.¹² The ripples on NiTi already achieved a width of ∼50 nm and depth of ∼400 nm after 30 min of processing time, and continued to increase in length over the remaining processing time (Fig. 5A(a), S5 and S6 in ESI†). However, the evolution of the ripples on SUS 316L was slower than that on both Co–Cr and NiTi, resulting in a continuous increase in width, depth, and length for 60 min (Fig. 5A(b), S5 and S6 in ESI†). For (2), we confirmed that TIPS can induce the formation of ripples on a metal substrate with various geometries by the introduction of a proper jig design (Fig. S7 in ESI†). A ripple structure was successfully formed on the complex-shaped Co–Cr substrate via TIPS with the substrate voltage of −800 V (Fig. 5B). The Co–Cr meshed tube was held with a conductive jig, which allows rotation of the tube for a uniform-patterned surface. At the early stage, inhomogeneous ripple patterns were randomly formed on the tube. After a significant processing time, the ripples became uniform with the saturated width and depth, consistent with the observation on the flat Co–Cr surface. Lastly, for (3), selective nanopatterning on the surface by TIPS could be achieved by covering the surface with a patterned mask. The Co–Cr surface beneath the mask remained intact because the mask fundamentally blocked the formation of a ripple structure, while the uncovered surface was exposed for the TIPS process with the substrate voltage of −800 V, where both etching and nanopatterning occurred at the same time (Fig. 5C). Therefore, through this process, nanopatterns could be only formed on the selective region of the surface.

Fig. 5 Formation of nanopatterns on various metal substrates: (A) top-view SEM images of nanopatterns on NiTi (a) and SUS 316L (b) after 60 min of TIPS. (B) Three dimensional image of a nanopatterned Co–Cr tube with complex geometry and corresponding low and high magnification SEM images of nanopatterned Co–Cr surfaces after 60 min of TIPS. (C) Selective nanopatterning of a metal substrate with a patterned mask and corresponding low and high magnification SEM images of nano- and micropatterned Co–Cr surfaces after 60 min of TIPS.

The TIPS technique we have developed in this study is a simple, one-step process for the fabrication of large-scale nanopatterns on a metal surface at low cost. However, there are two major challenges to be resolved for the further improvement of TIPS. Intrinsic pattern irregularity and defects of self-assembly methods are often attributed to low accuracy of nanopatterns, thus the nanopatterns by TIPS appear more feasible for defect-tolerant applications such as biomedical, chemical or energy devices.^28–30 Moreover, unlike other nanopatterning approaches, TIPS is only applicable to conductive materials, mainly metals and alloys due to its unique mechanism.¹⁵ For the TIPS technique, surface nanopatterns are developed by energy transfer of high energetic ions under the extremely high negative substrate bias of the substrate. Thus, electrical conductivity of substrate is essential to attract ions and to introduce an energetic ion bombardment phenomenon on the substrate surface.

The large surface area and pore volume capacity of ripple structures on metal surfaces through TIPS could be used as “carriers” for trapping biomolecules, drugs, organic dyes or nanoparticles for the development of functional biomedical devices, biosensors, optoelectronics and magnetic media for high-density storage.^28,29 The normalized surface area (actual surface area/projected surface area) and pore volume of ripple-patterned surfaces were predicted, varying the pattern width (λ) with the fixed aspect ratio of the pattern depth and width as ∼2.5, approximated from Fig. 3B (Fig. S8 and S9 in ESI†). Regardless of the increased pattern depth, the surface area of ripples appeared >5.5 times larger than their projected surface area, whereas the pore volume of a ripple pattern was increased linearly along with the increment of the pattern width, indicating that the loading amount of trapping agents could be controlled by the pattern size of ripples. Moreover, the unique surface topography of ripple patterns on metals could facilitate the alignment of biomolecules (e.g., DNA) or metal oxide nanowires (e.g., Fe₂O₃).^31–33

Conclusions

We demonstrate that TIPS via conventional DC magnetron sputtering with a Ta target and a negative substrate bias voltage can produce large-scale, self-assembled nanopatterns on various metal substrates in one step. This Ta-incorporated sputtering process induced composition-dependent sputtering on the metal substrates, enabling the size of nanopatterns to be tuned from nanoscale to sub-microscale levels with a negative substrate bias voltage (V > V_th) where etching became dominant. Our sputtering method was applicable to the nanopatterning of various metal substrates with complex geometries, proving its significant potential as a new approach for fabricating defect-tolerant nanopatterned devices in various industrial fields.

Acknowledgements

This work was supported by the Technology Innovation Program (Contract grant No. 10047860, Selective Plasma Etching (SPE) surface treatment for vascular stent and the development of low temperature complex coating technology for drug), and Industrial Strategic Technology Development Program (Contract grant No. 10045329, Development of customized implant with porous structure for bone replacement), funded by the Ministry of Trade, Industry & Energy (MI, Korea).

Notes and references

1. J. Chai, D. Wang, X. Fan and J. M. Buriak, Nat. Nanotechnol., 2007, 2, 500–506.
2. W. Cheng, N. Park, M. T. Walter, M. R. Hartman and D. Luo, Nat. Nanotechnol., 2008, 3, 682–690.
3. Y. Ding, M. W. Chen and J. Erlebacher, J. Am. Chem. Soc., 2004, 126, 6876–6877.
4. Surface design: applications in bioscience and nanotechnology, ed. R. Förch, H. Schönherr and A. T. A. Jenkins, Wiley-VCH Verlag GmbH & Co. KGaA, 2009.
5. G. M. Whitesides, Nat. Biotechnol., 2003, 21, 1161–1165.
6. K. Zhang, M. Uhrmacher, H. Hofsass and J. Krauser, J. Appl. Phys., 2008, 103.
7. L. Richert, F. Vetrone, J. H. Yi, S. F. Zalzal, J. D. Wuest, F. Rosei and A. Nanci, Adv. Mater., 2008, 20, 1488–1492.
8. O. Azzaroni, M. H. Fonticelli, G. Benitez, P. L. Schilardi, R. Gago, I. Caretti, L. Vazquez and R. C. Salvarezza, Adv. Mater., 2004, 16, 405–409.
9. Q. He, T. Yuan, S. Wei, N. Haldolaarachchige, Z. Luo, D. P. Young, A. Khasanov and Z. Guo, Angew. Chem., 2012, 51, 8842–8845.
10. J. Y. Cheng, C. A. Ross, H. I. Smith and E. L. Thomas, Adv. Mater., 2006, 18, 2505–2521.
11. S. Facsko, T. Dekorsy, C. Koerdt, C. Trappe, H. Kurz, A. Vogt and H. L. Hartnagel, Science, 1999, 285, 1551–1553.
12. J. Muñoz-García, L. Vázquez, R. Cuerno, J. A. Sánchez-García, M. Castro and R. Gago, in Toward Functional Nanomaterials, Lecture Notes in Nanoscale Science and Technology, ed. Z. M. Wang, Springer Science+Business Media, LLC, 2009, pp. 323–398.
13. S. Rusponi, C. Boragno and U. Valbusa, Phys. Rev. Lett., 1997, 78, 2795–2798.
14. W. D. Sproul, Science, 1996, 273, 889–892.
15. V. H. Pham, T. S. Jang, H. D. Jung, H. E. Kim and Y. H. Koh, J. Mater. Chem., 2012, 22, 24798–24804.
16. H. El-Sayed, S. Singh, M. T. Greiner and P. Kruse, Nano Lett., 2006, 6, 2995–2999.
17. T. Lu, J. Wen, S. Qian, H. L. Cao, C. Q. Ning, X. X. Pan, X. Q. Jiang, X. Y. Liu and P. K. Chu, Biomaterials, 2015, 51, 173–183.
18. H. Hofsäss and K. Zhang, Appl. Phys. A, 2008, 92, 517–524.
19. K. Zhang, M. Brötzmann and H. Hofsäss, New J. Phys., 2011, 13, 013033.
20. K. Zhang, O. Bobes and H. Hofsass, Nanotechnology, 2014, 25, 085301.
21. A. Meldrum, R. F. Haglund, L. A. Boatner and C. W. White, Adv. Mater., 2001, 13, 1431–1444.
22. S. Rosset, M. Niklaus, P. Dubois and H. R. Shea, Adv. Funct. Mater., 2009, 19, 470–478.
23. M. Engler, F. Frost, S. Muller, S. Macko, M. Will, R. Feder, D. Spemann, R. Hubner, S. Facsko and T. Michely, Nanotechnology, 2014, 25, 115303.
24. E. Chason, T. M. Mayer, B. K. Kellerman, D. T. Mcilroy and A. J. Howard, Phys. Rev. Lett., 1994, 72, 3040–3043.
25. Y. Yamamura and H. Tawara, At. Data Nucl. Data Tables, 1996, 62, 149–253.
26. H. Huang, L. M. Yang, S. Bai and J. Liu, Appl. Opt., 2015, 54, 324–333.
27. P. Gecys, A. Vinciunas, M. Gedvilas, A. Kasparaitis, R. Lazdinas and G. Raciukaitis, J. Laser Micro/Nanoeng., 2015, 10, 129–133.
28. T. Kumeria, A. Santos, M. M. Rahman, J. Ferre-Borrull, L. F. Marsal and D. Losic, ACS Photonics, 2014, 1, 1298–1306.
29. K. F. Huo, B. Gao, J. J. Fu, L. Z. Zhao and P. K. Chu, RSC Adv., 2014, 4, 17300–17324.
30. X. Guo, L. Liu, Z. Zhuang, X. Chen, M. Ni, Y. Li, Y. Cui, P. Zhan, C. Yuan, H. Ge, Z. Wang and Y. Chen, Sci. Rep., 2015, 5, 15947.
31. Q. L. He, T. T. Yuan, X. R. Yan, Z. P. Luo, N. Haldolaarachchige, D. P. Young, S. Y. Wei and Z. H. Guo, Chem. Commun., 2014, 50, 201–203.
32. Q. He, T. Yuan, Y. Wang, A. Guleria, S. Wei, G. Zhang, L. Sun, J. Liu, J. Yu, D. P. Young, H. Lin, A. Khasanov and Z. Guo, Nanoscale, 2016, 8, 1915–1920.
33. B. Teshome, S. Facsko and A. Keller, Nanoscale, 2014, 6, 1790–1796.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra00443a

---