Low energy sputtering of nickel by normally incident xenon ions

https://doi.org/10.1016/j.nimb.2005.02.016Get rights and content

Abstract

New sputter deposition processes, such as biased target ion beam deposition, are beginning to be used to grow metallic superlattices. In these processes, sputtering of a target material at ion energies close to the threshold for the onset of sputtering can be used to create a low energy flux of metal atoms and reflected neutrals. Using embedded atom method potentials for fcc metals and a universal potential to describe metal interactions with the inert gas atoms used for sputtering, we have used molecular dynamics simulations to investigate the fundamental phenomena controlling the emitted vapor atom and reflected neutral fluxes in the low energy sputtering regime. Detailed simulations of low energy, normally incident Xe+ ion sputtering of low index nickel surfaces are reported. The sputtering yield, energy and angular distributions of sputtered atoms, together with the reflection probability, energy and angular distributions of reflected neutrals were deduced and compared with available experimental data. The average energy of sputtered metal atoms can be controllably reduced to 1–2 eV as the Xe+ ion energy is reduced to 50–100 eV. Normally incident Xe+ ion sputtering in this energy range results in reflected Xe energies that are narrowly distributed between 2 eV and 6 eV. These fluxes are ideally suited for the growth of metallic multilayers.

Introduction

Nanoscale multilayers have many important applications. For instance, Cr/Sc multilayers are critical X-ray optics materials [1] that can be used for microscopy [2], astronomy [3], lithography [4] and microanalysis [5]. Multilayers composed of ferromagnetic metal layers separated either by thin conductive metal or dielectric (tunneling barrier) layers are giant magnetoresistive (GMR) [6], [7], [8] and are being used for read head sensing [9], [10] and magnetic random access memory (MRAM) [11], [12], [13]. The performance of devices utilizing these metallic multilayers are optimized by minimization of both the interfacial roughness and interlayer mixing at each interface within the multilayer stack [1], [14], [15], [16], [17].

These metallic superlattices have been grown by a wide variety of vapor deposition techniques [1], [6], [17], [18]. Hyperthermal processes such as magnetron sputtering and ion beam deposition (IBD) that utilize metal atoms with average translational energies of several electron volts or more have been found to produce better multilayers than those made by molecular beam epitaxy (MBE) where the vapor atoms have energies of around 0.1 eV [19]. Numerous efforts have sought to experimentally optimize the IBD process for growth of GMR multilayers [18], [20], [21]. The best GMR films were obtained when the energy of the inert gas ions used for sputtering lay between 500 eV and 700 eV. In this ion energy range, the average kinetic energy of the atoms sputtered from the target is close to or greater than 10 eV [22].

Atomistic simulations of multilayer growth indicate that hyperthermal atom impacts activate surface flattening mechanisms resulting in smoother growth surfaces [23], [24], [25]. However, as the impact energy is increased, significant interlayer mixing by an impact induced atomic exchange (layer alloying) mechanism begins to occur [23], [24], [25]. Reflected (inert gas) neutrals and assisting ion fluxes have been shown to induce similar effects [26]. These simulations indicated that the best trade-off between interfacial roughness and interlayer mixing occurs at an atom impact energy of 2–3 eV [24], in agreement with the experiments [17].

Atomistic simulations also indicate that further improvements in film perfection are possible if the atoms are modulated from a low (<1 eV) to a higher (5–10 eV) energy during the growth of each layer [23], [24], [25]. The use of a low energy to deposit the first few monolayers of a new material layer reduces mixing at the interface, but at the expense of forming high roughness. Switching to a higher energy as the layer thickness increases enables the layer surface to be flattened without intermixing at the now buried interface.

Implementation of the deposition concepts identified by the simulations requires a process where the metal and the reflected neutral atom energy can be controlled in the 1–5 eV range. The average metal atom kinetic energy in conventional (700–2000 eV ion energy) IBD appears to be much higher than the ideal energy predicted by atomistic simulations. Reducing the ion energy is expected to decrease the average metal atom energy. However, the use of ion energies below 300 eV results in ion beam de-focusing (which causes overspill contamination by sputtering of the target holder and chamber walls) and a reduction in the deposition rate due to a reduction of sputtering yield (which results in increased contamination by residual gas species) [27]. A recently developed biased target ion beam deposition (BTIBD) technique has sought to extend the IBD process to a much lower ion energy range by resolving both the overspill contamination and deposition rate problems [28]. Preliminary studies of the growth of GMR multilayers using BTIBD at an ion energy of 300 eV has resulted in significantly improved GMR properties compared to identical multilayers grown using conventional IBD with an “optimized” ion energy of 600 eV [27].

The ion bombardment of a metal surface creates energetic reflected (inert gas) neutrals in addition to sputtered metal atoms [29], [30], [31], [32]. These neutrals can potentially reach the growth surface, altering the structure and hence the properties of the deposited films [33]. They can also reach other hardware, such as chamber walls, where they sputter off undesired materials that contribute to contamination. The metal and reflected neutral fluxes created by very low ion energy bombardment of a target are not as well understood as those of conventional sputtering, which has been widely studied since the 1950s [34], [35], [36], [37]. A complete characterization of the fluxes includes the sputtering yield, reflection (inert gas neutral) probability and the energy/emission angular distributions of both fluxes. For practical applications, such as X-ray optics [1] and GMR devices [9], many different materials (Sc, Cr, Ta, Ni, Fe, Cu, Co, etc.) are involved. In principle, a characterization of the sputtering fluxes of all of these materials is required for each ion species, ion incident energy and incident angle of practical interest. The experimental collection of this sputtering data is difficult and prohibitively time consuming. An alternative atomic scale simulation approach is used here to explore the sputtering of nickel by xenon ions. The wide availability of compatible interatomic potentials for the many metals [23], [38] and sputtering gases [39] makes the approach readily extendable to other systems of interest.

Our simulation approach extends considerably past theoretical efforts to understand and simulate the sputtering of metals by inert gas ion bombardment. In sputtering, the impinging ions transfer their energy and momentum to the target atoms upon collision. These primary knock-on target atoms then transfer their energy and momentum to other atoms via secondary recoils. This process repeats until a near surface atom receives a sufficiently high, outwardly-directed impulse that it overcomes its binding to the surface and is sputtered. A major advance in the theoretical analysis of sputtering was made by Sigmund in the 1960s [40]. His analytical theory assumed that sputtering proceeded by a linear collision cascade mechanism. High speed computing subsequently enabled linear-cascade sputtering processes to be simulated using Monte Carlo techniques and a binary collision approximation. These frequently utilized Monte Carlo methods include the codes MARLOWE [41] and TRIM [42], [43]. They have a well developed physical picture behind them and often give a good representation of experimental results [43]. However, these methods suffer from several drawbacks. First, they need a number of ad hoc input parameters that cannot be obtained fundamentally. For instance, the results of simulations are very sensitive to the surface binding energies [42], [43], which are poorly defined and often need to be modified to match experimental sputtering data [44]. Second, the simulations cannot address the phenomena outside the linear-cascade regime, such as cluster sputtering and the occurrence of high energy density (spike) zones. The binary collision approximation has also been found to fail at low incident energies [45], [46]. Finally, these Monte Carlo approaches are not easily extended to alloyed or compound targets or to cases where more complex sputtering molecules or particles are used.

The emergence of increasingly high fidelity interatomic potentials and computationally efficient molecular dynamics (MD) algorithms led to a widespread interest in the use of MD methods for investigations of sputtering [47], [48], [49]. In MD simulations, atom positions are deduced using Newton’s equation of motion where the interaction among all the atoms is treated simultaneously. MD therefore better captures the physics of sputtering. The applicability of MD approaches to sputtering has been limited in the past by high computational cost arising from (a) the large computational crystal that must be used to encompass the heat zone generated during a sputtering event, (b) the very short time step that must be used to correctly reflect both the lattice vibration and the energetic particle bombardment and (c) the duration of the sputtering processes, which can be significant compared to the time step. Modern desktop computers now enable MD simulations to handle 5000 or more atom crystals and allow real time simulation periods that exceed the time duration of low ion energy sputtering events (typically <2 × 10−12 s) [47].

MD has been used to simulate the sputtering during hyperthermal (5–400 eV) Ne, Ar and Xe atom impacts on crystalline Cu surfaces [50]. Here we use an MD approach to simulate the sputtering of low index ({1 1 1}, {1 1 0} and {1 0 0}) nickel (target) surfaces by normally incident, low energy xenon ions. The sputtering yield, the energy and angular distributions of the sputtered nickel atoms, together with the reflection probability, the energy and angular distributions of the reflected xenon atoms are all quantified for xenon ion incident energies between 50 eV and 1000 eV. Time-resolved simulations are also used to show mechanisms of low ion energy sputtering.

Section snippets

Interatomic potentials

Realistic MD simulations require high fidelity interatomic potentials to calculate interatomic forces. For metal systems, especially the fcc transition metals such as nickel, the embedded atom method (EAM) potential originally developed by Daw and Baskes has been widely used [51]. In addition to pair potentials, EAM potentials include an embedding energy term, which effectively incorporates the local environment dependence of atomic interactions and allows for a realistic description of

Nickel sputtering yield

Sputtering of the three Ni surfaces were simulated at Xe+ ion incident energies of 50–1000 eV, with most simulations carried out in the low energy end (50–300 eV). The sputtering yield was determined as a function of incident Xe+ ion energy. Similar data have also been experimentally measured [55], [56], [57], [58]. The simulated and the experimental data are compared in Fig. 2. It can be seen that the incident ion energy dependence of the sputtering yield obtained from simulations is close for

Collision mechanisms

The low energy sputtering has been well established to proceed through a single-knock-on mechanism [34]. MD simulations enable the detailed sputtering mechanisms to be visualized. Here we examine several representative low energy (100 eV) impacts and their collision sequence resulting in low energy sputtering phenomena.

During ion impacts at energies above the sputtering threshold, significant displacement of atoms occurs. To analyze the collective motion of many atoms is not a simple task. The

Conclusions

MD simulations have been used to study the low energy (50–1000 eV) normal incident Xe+ ion sputtering on low index Ni surfaces. The study focused on the low energy end (50–300 eV) exploited by new deposition technologies, such as biased target ion beam deposition. The sputtering yield, average energy, energy and angular distributions of sputtered atoms, as well as reflection probability, average energy, energy and angular distributions of reflected neutrals, were all quantified as general

Acknowledgments

We are grateful to the Defense Advanced Research Projects Agency and Office of Naval Research (C. Schwartz and J. Christodoulou, Program Managers) for support of this work through grant N00014-03-C-0288. We also thank S.A. Wolf for numerous helpful discussions.

References (71)

  • M. Julliere

    Phys. Lett. A

    (1975)
  • J.Q. Wang et al.

    Mater. Sci. Engr. B

    (2000)
  • J.P. Nozieres et al.

    J. Magn. Magn. Mater.

    (1993)
  • W.H. Butler et al.

    J. Magn. Magn. Mater.

    (1995)
  • S. Schmeusser et al.

    J. Magn. Magn. Mater.

    (1997)
  • X.W. Zhou et al.

    Acta Mater.

    (2001)
  • G. Betz et al.

    Int. J. Mass Spectrom. Ion Processes

    (1994)
  • M. Hou et al.

    Nucl. Instr. and Meth. B

    (1994)
  • H.M. Urbassek

    Nucl. Instr. and Meth. B

    (1997)
  • C. Kim et al.

    Nucl. Instr. and Meth. B

    (1995)
  • J. Bohdansky

    J. Nucl. Mater.

    (1980)
  • N. Matsunami et al.

    At. Data Nucl. Data Tables

    (1984)
  • Y. Yamamura et al.

    At. Data Nucl. Data Tables

    (1996)
  • M.W. Thompson

    Phys. Rep.

    (1981)
  • M.W. Thompson

    Vacuum

    (2002)
  • T. Mousel et al.

    Nucl. Instr. and Meth. B

    (1999)
  • M. Urbassek

    Nucl. Instr. and Meth. B

    (1984)
  • V.I. Shulga

    Nucl. Instr. and Meth. B

    (2001)
  • X.W. Zhou et al.

    Surf. Sci.

    (1999)
  • J. Birch et al.

    Vacuum

    (2003)
  • M. Berglund et al.

    J. Microsc.

    (2000)
  • A.B.C. Walker et al.

    Science

    (1988)
  • D.G. Stearns et al.

    Appl. Opt.

    (1993)
  • A. Grudsky, A. Rudnev, Optics and microanalysis 1992, Proceedings, 13th International Congress, Bristol, IOP, UK, 1993,...
  • M.N. Baibich et al.

    Phys. Rev. Lett.

    (1988)
  • G. Binasch et al.

    Phys. Rev. B

    (1989)
  • C. Tsang et al.

    IEEE Trans. Magn.

    (1994)
  • G.A. Prinz

    Science

    (1998)
  • R.S. Beech et al.

    IEEE Trans. Magn.

    (1996)
  • X. Zhu et al.

    IEEE Trans. Magn.

    (2003)
  • D.M.C. Nicholson et al.

    J. Appl. Phys.

    (1994)
  • J.C.S. Kools

    J. Appl. Phys.

    (1995)
  • X.W. Zhou, W. Zou, R.A. Johnson, H.N.G. Wadley, Proceedings, Non-volatile Memory Technology Symposium, San Diego, CA,...
  • Y. Saito et al.

    Phys. Rev. B

    (1995)
  • H. Hegde et al.

    J. Appl. Phys.

    (1999)
  • Cited by (16)

    • Incident angle dependence of reflected particles in low-energy xenon-ion impacts on metal surfaces

      2021, Computational Materials Science
      Citation Excerpt :

      Winters et al. discussed how the energy reflected from a metal target may influence thin films deposited in various sputtering systems [8]. Zhou et al. conducted simulations of low-energy, normally incident Xe + ion sputtering of low-index nickel surfaces [11]. They reported that normally incident Xe + ion sputtering in this energy range results in reflected Xe energies that are narrowly distributed between 2 eV and 6 eV.

    • How do energetic ions damage metallic surfaces?

      2015, Current Opinion in Solid State and Materials Science
      Citation Excerpt :

      Other studies have focused on the mechanisms of sputtering. For example, results of irradiation by low (few hundred eV) and moderate energy (few keV) ions were discussed in [10–18]. Sputtering by atoms or ions with a few keV of kinetic energy was modeled in many materials and significant progress was achieved in understanding and development of predictive models [19,20].

    • Sputtering of cubic metal crystals by low-energy xenon-ions

      2015, Computational Materials Science
      Citation Excerpt :

      The time between subsequent impacts was 1 ps, which was observed to be sufficiently long for most sputtering events to have occurred. The number of projectile atoms that were impacted in each simulation was chosen to provide a flux of 0.04 atoms/Å2, matching simulations conducted by Zhou et al. [7]. Following the simulation, the final coordinates and velocities of the atoms are used to compute a range of statistics.

    • Structural and electrochemical characterization of Ni nanostructure films on steels with brush plating and sputter deposition

      2007, Current Applied Physics
      Citation Excerpt :

      The electroless nickel–phosphorous plating with differing phosphorous contents was employed as an interlayer to modify the properties of CrN hard coating on mild steel [14]. Detailed simulations of low energy, normally incident Xe+ ion sputtering of low index nickel surfaces are reported [15]. Milkani et al. [16], have reported the effect of DC bias voltage on the deposition rate for Ni thin films by RF-DC coupled unbalanced-magnetron sputtering.

    View all citing articles on Scopus
    View full text