Thin Solid Films

(1)

Critical review

Thin ﬁlm metallic glasses: Unique properties and potential applications

Jinn P. Chu

^a,

⁎ , J.S.C. Jang

^b

, J.C. Huang

^c

, H.S. Chou

^c

, Y. Yang

^d

, J.C. Ye

^d

, Y.C. Wang

^e

, J.W. Lee

^f

, F.X. Liu

^g

, P.K. Liaw

^g

, Y.C. Chen

^h

, C.M. Lee

^h

, C.L. Li

^h

, Cut Rullyani

^a

aDepartment of Materials Science and Engineering, National Taiwan University of Science and Technology, Taipei, 10607, Taiwan

bInstitute of Materials Science and Engineering, Department of Mechanical Engineering, National Central University, Chung-Li, 32001, Taiwan

cDepartment of Materials and Optoelectronic Science, National Sun Yat-Sen University, Kaohsiung, 80424, Taiwan

dDepartment of Mechanical Engineering, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong

eDepartment of Civil Engineering, National Cheng Kung University, Tainan, 70101, Taiwan

fDepartment of Materials Engineering, Ming Chi University of Technology, New Taipei City 24301, Taiwan

gDepartment of Materials Science and Engineering, University of Tennessee, Knoxville, TN 37996-2200, USA

hGraduate Institute of Applied Science and Technology, National Taiwan University of Science and Technology, Taipei, 10607, Taiwan

a b s t r a c t a r t i c l e i n f o

Article history:

Received 21 March 2012 Accepted 22 March 2012 Available online 9 April 2012

Keywords:

Metallic glass Amorphous

Solid-state amorphization Microcompression Adhesion Wear resistance

A new group of thinfilm metallic glasses (TFMGs) have been reported to exhibit properties different from conventional crystalline metalfilms, though their bulk forms are already well-known for high strength and toughness, large elastic limits, and excellent corrosion and wear resistance because of their amorphous structure. In recent decades, bulk metallic glasses have gained a great deal of interest due to substantial improvements in specimen sizes. In contrast, much less attention has been devoted to TFMGs, despite the fact that they have many properties and characteristics, which are not readily achievable with other types of metallic or oxidefilms. Nevertheless, TFMGs have been progressively used for engineering applications and, thus, deserve to be recognized in thefield of thin film coatings. This article will thus discuss both properties and applications of TFMGs including a review of solid-state amorphization upon annealing, the glass-forming ability improvement due to thinfilm deposition, and mechanical properties, including residual stress, hardness and microcompression, adhesion, and wear resistance. Potential applications and simulations will also be discussed.

Contents

1. Introduction . . . 5098

2. Early work on TFMGs. . . 5098

3. Unique properties . . . 5099

3.1. Annealing-induced amorphization of TFMGs. . . 5099

3.2. Smooth surface . . . 5100

3.3. Hardness and electrical resistivity . . . 5100

3.4. Magnetic properties . . . 5102

3.5. Glass-forming ability (GFA) . . . 5103

3.6. Mechanical properties . . . 5103

3.6.1. Residual stress . . . 5103

3.6.2. Nanoindentation . . . 5103

3.6.3. Microcompression . . . 5106

3.6.4. Multilayered TFMGs. . . 5108

3.7. Adhesion and tribological properties . . . 5109

3.7.1. Rockwell-C adhesion test . . . 5110

3.7.2. Wear resistance. . . 5111

⁎ Corresponding author.

E-mail address:jpchu@mail.ntust.edu.tw(J.P. Chu).

doi:10.1016/j.tsf.2012.03.092

Contents lists available atSciVerse ScienceDirect

Thin Solid Films

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / t s f

(2)

4. Potential applications . . . 5112

4.1. TFMGs for biomedical use . . . 5112

4.1.1. Antimicrobial . . . 5112

4.1.2. TFMGs for medical tools . . . 5112

4.2. Improved fatigue properties due to TFMGs. . . 5113

4.2.1. Governing mechanisms . . . 5114

4.2.2. Annealing effects . . . 5115

4.3. TFMGs for bendable bulk metallic glass . . . 5116

4.4. Micro and nano replication . . . 5116

4.5. Microelectronic and optoelectronic applications . . . 5118

5. Molecular-dynamics (MD) simulations . . . 5118

6. Conclusions and outlook . . . 5121

Acknowledgments . . . 5121

References . . . 5121

1. Introduction

Metallic glasses (MGs), or amorphous metallic alloys, are non- crystalline metals which lack long-range atomic periodicity because they are generally formed with fast quench rates for the retention of the glassy state from the melt. MGs are non-equilibrium materials;

important characteristics are the glass transition and crystallization temperatures when heated toward the liquid state. Due to the disor- dered atomic structure and the absence of grain boundaries, MGs have many superior properties including good soft-magnetic properties and excellent mechanical properties of high speciﬁc strength, large elastic limits (~2%), and high resistance to corrosion and wear.

Thus, they have been a subject of interest for scientific research and engineering application for decades. The pioneering work by Klement et al.[1], Chen and Turnbull[2], and Chen[3]in the 1960s and 1970s reported thefirst MG samples based upon binary Au\Si and ternary Pd\Si\X (X_Ag, Cu or Au), and Pd\Y\Si (Y_Ni, Co or Fe) systems, using rapid solidification methods such as splat quenching. MGs at that time were limited to relatively small sizes, typically of order μm, in order to achieve fast quench rates. The recent advent of MGs with relatively high thermal stability and low critical cooling rates, primarily through multi-component composition design, has led to significant improvements in characteristic specimen sizes (thickness and diameter). As a result, many MGs (e.g. Zr-, Cu-, Ti-, Fe-, Pd-, Pt- and Au-based systems) with sizes in excess of 1 mm (“bulk” metallic glasses, BMGs) have been reported[4]. These BMG systems generally have good glass-forming ability (GFA) such that rapid solidification is unnecessary and BMGs thus have become obtainable with conventional copper-mold casting methods.

Increasing interest in developing and understanding this new family of materials has also led to making thinﬁlm metallic glass (TFMG) processing possible, which was not readily achieved in the past when MGs were available only as powder or ribbons. To make use of unique properties, TFMGs with good GFA are evolving as alternativeﬁlm materials which are potentially useful for many applications such as micro- electro-mechanical system (MEMS) devices. While MGs and BMGs are considered as newcomers, there are already many succinct and thor- ough articles reviewing various topics from the science of glass forming to atomic structure to mechanical properties[4–9]. However, only a short review on TFMGs published in 2010[10]is available. The present review summarizes and discusses progress made in the area of TFMGs over the past decade. In addition, several challenging subjects will be proposed for future research.

2. Early work on TFMGs

Research on TFMGs in the 1980s and 90s mainly concern immiscible binary systems. These include Cu\Ta and Cu\W deposited by evaporation[11], as well as Cu\Zr[12]and Al\Fe, Bi\Fe and Bi\Ti[13]

List of symbols

β taper angle γ weighting factor ξ effective hopping integer σf residual stress in theﬁlm δ total displacement

δi initial depth of blade indentation ρ radius of curvature

ΔT supercooled liquid region, SCLR a effective contact radius

A, p,q system-speciﬁc parameters for molecular-dynamics simulations

dx increment of blade displacement D diameter of micropillar

e electron charge exp exponential function

Er reduced‘composite’ modulus of the ﬁlm/substrate system

E_si Young's modulus of Silicon

Ei Young's modulus of the diamond indenter Ef Young's modulus ofﬁlm

E_s Young's modulus of substrate f dimensionless function F applied force

h indentation depth H height of micropillar l length of cut surface Lf Firsov screening length JIc Mode I fracture toughness Ms substrate biaxial modulus

P applied load

r_i distance between two atoms, i and j r0 ﬁrst-neighbor distance

R0 radii of curvatures of the substrate beforeﬁlm deposition R radii of curvatures of the substrate after theﬁlm

deposition ts substrate thickness t_f ﬁlm thickness

Tg glass transition temperature Tx crystallization temperature ν Poisson's ratio

νi Poisson's ratio of the diamond indenter νf Poisson's ratio ofﬁlm

νs Poisson's ratio of substrate V_ij pair-wise Moliére potential Z1 ion gas atomic number Z2 neutral atom atomic number

5098 J.P. Chu et al. / Thin Solid Films 520 (2012) 5097–5122

(3)

by sputtering. Additional early studies on binary systems are those on the annealing-induced solid-state amorphization (SSA) of multilayerfilms[14–17]. In 1983, Schwarz and Johnson[14] fabricated thefirst La\Au TFMG by SSA during annealing of evaporated La/Au multilayerfilms. In 1986, Newcomb and Tu[16]and Cotts et al.[15]

used transmission electron microscopy (TEM) and differential scanning calorimetry (DSC), respectively, to characterize the SSA of crystalline Ni/Zr multilayer thinﬁlms prepared by sputtering. Since then, some elemental multilayerﬁlms have been fabricated by evaporation and sputtering for SSA studies. See reference[18]for reviews and summaries of SSA.

Apart from binary systems and SSA multilayerfilms, which are often limited to nanometer-scaled thicknesses and narrow composition ranges, multi-component TFMGs readily become amorphous in the as-deposited state withfilm thicknesses up to several, or even tens of, micrometers. TFMGs gradually have received attention in scientific research for potential applications since the late 1990s because many new multi-component BMGs with good GFA in Mg-, Ln-, Zr-, Fe-, Pd-, Cu-, Ti-, and Ni-based systems have been developed by Inoue and his group in the 1980s and 90s[19]. Thinfilms prepared by vapor-to-solid deposition are expected to be farther from equilibrium than those prepared by a liquid-to-solid melting/casting process. Thus, the GFA can be further improved and composition ranges for amorphization are wider when formed by thinfilm processing such as sputtering[20,21]. In fact, sputter deposition technology has been used for GFA determination of metallic glass systems, by varying thefilm composition and density when co-sputtered with Zr and Cu elemental targets[22].

In 1999 and 2000, Zr\Cu\Al and Pd\Cu\Si ternary TFMGs were sputter deposited for MEMS applications[23,24]. Because of excellent three-dimension forming ability, good corrosion resistance, and mechanical properties, compared to conventional crystallineﬁlm counterparts, Zr- and Pd-based TFMGs are appropriate choices for MEMS applications, such as conical spring linear micro-actuators. Zr\Al\Cu\Ni TFMGs prepared by sputter deposition and focused ion beam patterning are reported for nano-device applications[25].

3. Unique properties

3.1. Annealing-induced amorphization of TFMGs

Amorphization in solids may be achieved by mechanical alloying, solid-state reaction or solid state amorphization (SSA), high pressure, or shock loading techniques. As mentioned inSection 2, in some elemen- tally modulated crystallineﬁlms, SSA within the interfacial nanometer regions can occur through annealing-induced diffusion reactions[10].

Amorphization in sputtered TFMGs during annealing in the temperature range (the supercooled liquid region, SCLR orΔT) between the glass transition temperature (Tg) and the crystallization temperature (Tx) has been reported. TFMGs which exhibit annealing-induced amorphization withinΔT include Zr-[26], Fe-[27], and Cu-based[28,29]systems.

Yet, it is difficult to observe amorphization in TFMGs because the small scale of thinfilms and the presence of nanocrystalline phases have made characterization a challenging task. For instance, when Cu-based TFMG (Cu51Zr42Al4Ti3) with Tgand Txdetermined to be 440.2 °C and 493.7 °C, respectively, is annealed withinΔT for various length of times [29], XRD spectra inFig. 1 [29]from as-deposited and annealedfilms reveal samples are basically either amorphous or nanocrystalline, as indi- cated by typical broad diffraction humps with no detectable crystalline peaks. Based on the XRD results, annealing-induced amorphization can not be unambiguously determined because thefilms remain amorhpous or nanocrystalline with no obvious crystalline peaks up to 600 °C or after isothermal annealing at 470 °C for 3 min. Thus, the proper selection of analytical techniques to characterize the amorphous and crystalline structures is essential. In addition, it is important to carefully control the annealing process (such as the appropriate duration of annealing time under a protective atmosphere) for amorphization to take place.

The typical duration of time for annealing-induced amorphization in TFMGs is 1 min inΔT. Transmission electron microscopy (TEM) is a common and useful tool for microstructural and crystallographic analyses.

As an example of amorphization withinΔT,Fig. 2 [27]shows TEM bright-ﬁeld images and electron diffraction patterns of Fe-based TFMGs revealing that the minor nanocrystallites formed in the as-deposited state grow in size with annealing temperature up to 400 °C (Tg) and eventually transform into the amorphous phase withinΔT (i.e. at 500–

550 °C). Above Tx, crystallization and grain growth of FeNi takes place at 600–750 °C. Amorphization also yields smooth ﬁlm surfaces after annealing withinΔT, whereas the surface normally becomes rougher as the annealing temperature increases due to grain growth and crystallization. This can be clearly seen in the atomic force microscopy (AFM) topographic images inFig. 3that were acquired at the same temperatures as those inFig. 2for comparison[27]. The AFM images at 500 and 550 °C are featureless and consistent with those of the amorphous structure in TEM results (Fig. 2). In addition, the crystalline phases and their sizes in theﬁlm also affect surface morphology, i.e. the coarser crystalline grains at the higher temperatures inFig. 2have rougher surfaces in AFM results (Fig. 3).

To further clarify the above-described amorphization phenomenon, in-situ TEM observations are correlated with the ex-situ measurement results obtained using AFM and DSC.Figs. 4 and 5show TEM bright- ﬁeld images and corresponding diffraction patterns, respectively, of a Cu-based (Cu51Zr42Al4Ti3) TFMG heated to various temperatures[29].

Fig. 4(a) is an image obtained from the as-deposited sample, revealing small numbers of nanocrystallites in the amorphous matrix. Neverthe- less, a columnar structure, typical of sputter deposition at low temperature[30], is revealed. The diffraction pattern inFig. 5(a) indicates that the amorphous structure is the major phase with minor nanocrystallites. During heating at low temperatures [e.g. 400 °C,Fig. 4(b)], the spherical nanocrystallites increase in size, to 10–20 nm in diameter, and in number as well. The volume fraction of crystallites is estimated to be ~30%. The micrograph inFig. 4(b) and its corresponding diffraction pattern inFig. 5(b) indicate the nanocrystalline structures as minor phases within the dominant amorphous matrix. The nanocrystallites produce a spotty diffraction pattern, which has been identified as either tetragonal or cubic Cu10Zr7. At this temperature (400 °C), the large interfacial energy drives coarsening of the metastable crystalline phase through a process analogous to Ostwald ripening and the nanocrystallites thus increase in size.Fig. 4(c) shows a fully amorphous structure with no observable crystalline phase in the TEM image after heating at 438 °C within theΔT range. The typical broad diffuse diffraction ring further confirms the amorphous nature ofFig. 5(c). Neither obvious crystals nor splitting of the halo ring is observed in the bright-field TEM image or Fig. 1. XRD patterns from an as-deposited Cu-based TFMG and samples annealed within ΔT for various length of times[29].

(4)

diffraction pattern. At this point, the metastable nanocrystalline phases may have dissolved in the amorphous matrix. As the temperature rises to 525 °C, well above the crystallization transition temperature, nanocrystallites reappear in the amorphous matrix [Fig. 4(d)]. Appearance of crystalline spots in the diffraction pattern ofFig. 5(d) indicates that the amorphous phase starts to transform into the crystal state as the temperature increases. The sizes of nanocrystalline phases at 525 °C are in the range of 20 to 35 nm.

3.2. Smooth surface

In BMGs, surface defects on the scale of a few micrometers tend to disappear without crystallization as the glassy surface undergoes“self- healing” when heated within ΔT due to a sufﬁciently low viscosity[31].

In the case of TFMGs, both amorphization and low viscosity are thought to cause the recovery of minor scratches, such as a nanoindentation mark. AFM images inFig. 6reveal a decrease in the size of a nanoindentation after annealing withinΔT. The indentation, measured using AFM, becomes ~13.8% shallower in depth, decreasing from 64.63± 0.82 nm to 55.97 ± 0.53 nm, after annealing inΔT for 1 min. The figure shows that thefilm surface is also smoother and the pile-up around the indentation decreases as a result of annealing. According to AFM and TEM results, annealing-induced amorphization withinΔT provides a unique and ad- vantageous property for TFMGs over other types of thinfilms because smooth surfaces can be readily obtained through a proper annealing process.

3.3. Hardness and electrical resistivity

Annealing-induced amorphization is generally associated with significant alterations in film properties. These property changes, in addition to the smooth surface shown inFig. 3, include a decrease in hardness and an increase in electrical resistivity[26–28]. The hardness drop (or softening) withinΔT is mainly due to the single phase of amorphous structures, compared to composite structures consisting of crystalline phases in the amorphous matrix annealed at temperatures other thanΔT[27,28]. Overall, thefilm hardness tends to increase with the temperature upon annealing up to Tg, as reported for several TFMG systems[26,27,29,32]. The hardness increases are thought to be attributable to the combined effects of composite structure and free volume annihila- tion due to structure relaxation.

For electrical property changes due to amorphization withinΔT,Fig. 7 shows an example of the dramatic increase in the electrical resistivity of Cu-based (Cu51Zr42Al4Ti3) TFMGs[28]. According to differential scanning calorimetric (DSC) scan data inFig. 7(a), Tg, Tx, andΔT are determined to be 452 °C, 502 °C, and 50 °C, respectively. InFig. 7(b), annealing within theΔT region is shown to cause a distinct increase in the film resistivity to a maximum of ~4628μΩ-cm at 502 °C. When the temperature reaches 527 °C, the resistivity then decreases to ~177μΩ-cm. The random atomic structure in the amorphous phase is believed to cause the electrical resistivity increase withinΔT, while the presence of crystalline phases in the films annealed at temperatures other than ΔT result in relatively low resistivity. The extent of resistivity change withinΔT is larger in Cu-based TFMGs than those reported for Zr- and Fe-based TFMGs[26,27]. The Fig. 2. TEM bright-field images and diffraction patterns from Fe-based TFMGs revealing phase evolution during one-min anneals: amorphous matrix with γ-fcc FeNi nanocrystallites (as-deposited)→growth of nanocrystallites (up to 400 °C)→amorphous (500 °C and 550 °C)→formation of cubic Fe(Ni) crystallites (600 °C) with minor cubic FeNi phase (650 °C to 700 °C)→formation and grain growth of FeNi phase (750 °C)[27].

(5)

Fig. 3. AFM topographic images (5 × 5μm², height scale = 50 nm) of Fe-based TFMGs in as-deposited and annealed (one min) conditions.

Fig. 4. In-situ TEM video image frames of a Cu-based TFMG: (a) as-deposited and after one-min anneals at (b) 400 °C, (c) 438 °C and (d) 525 °C. Approximately same locations in the images are denoted by white reference arrows. Black arrows indicate the presence of crystalline phases[29].

(6)

signiﬁcant increase (~25×) of resistivity in this case may have a potential for applications such as safety breakers and switches where rapid electrical resistivity change with temperature is required.

3.4. Magnetic properties

Thermal annealing affects not only microstructural, surface morphology, hardness, and resistivity, but also the magnetic properties of Fe-based TFMGs[27]. Magnetic force microscopy (MFM) images in Fig. 8show variations in the magnetic domain structure upon annealing. Until the temperature reaches 600 °C, there is no obvious color contrast in the images, suggesting the magnetization direction lies in the plane of theﬁlm. Since the ﬁlm thickness in this case is about 500 nm,

which is relatively thin, the demagnetization energy in theﬁlm is neg- ligible due to the high aspect ratio ofﬁlm geometry and thus a single domain structure is formed. Furthermore, the domain structure formation is not affected by amorphization at 500 and 550 °C.

When the temperature reaches 650–700 °C, the color-contrast MFM images show the distinctive features of a“dark” and “bright”

stripe magnetic domain structure. Since the MFM tip is magnetized downward, the bright stripe domains are those with magnetization upward, whereas the dark domains are magnetized downward. The

“dark/bright” stripe images indicate the presence of strong perpendicular magnetic anisotropy with the magnetic easy axis normal to the plane of thefilm. As the magnetization is along the easy axis and normal to thefilm plane, the film demagnetizing fields in the film increase. To Fig. 5. In-situ TEM diffraction patterns of a Cu-based TFMG: (a) as-deposited and after one-min anneals at (b) 400 °C, (c) 438 °C and (d) 525 °C. Major diffraction spots in (b) are indexed to be the Cu10Zr7phase, while those in (d) are from Cu10Zr7and AlZr2[29].

Fig. 6. AFM images showing the recovery of a nanoindentation in a Zr-based TFMG after annealing at 460 °C for 1 min.

(7)

minimize the demagnetizingfields, the bright and dark stripe domain structures are thus formed. The stripe domains are found to increase in width with annealing temperature between 650 and 700 °C, which increases the effective magnetic anisotropyfield in the film [27]. The annealing-induced perpendicular magnetic anisotropy originates from the positive magnetostriction and the compressive stress due to the large FeNi lattice phase present in the smaller Fe(Ni) lattice matrix [27]. When thefilm is annealed at 750 °C, the magnetic anisotropy van- ishes because the stress diminishes due to the absence of minor phases, resulting in no apparent color contrast in the magnetic force image. This

“stress-induced perpendicular magnetic anisotropy” is commonly reported in multilayeredfilms, but is rare in single layer thin films [33,34]. This again exemplifies TFMGs as fascinating thin film materials exhibiting unique and extraordinary properties, many of which have not been explored or well studied.

3.5. Glass-forming ability (GFA)

For TFMG deposition, amorphous sputtering targets are difficult to prepare because of the large sizes. However, a recent work by Chen et al. reported that both amorphous and crystalline targets yield no apparent differences infilm microstructure and crystal structure[35]. In addition, it is well known that the composition window for achieving amorphous thinfilms is much wider than that for achieving BMGs by rapid casting because the state produced by the vapor-to-solid deposition is farther from equilibrium than the state produced by the liquid-to-solid casting process. For instance, fully amorphous thin films can be easily prepared by co-sputtering of Zr\Cu or Zr\Cu\Ti systems[20,21]. Partially amorphous thinfilms are achieved by co- sputtering of Mg\Cu based alloys[36]. Some elements having positive heat of mixing with the parent element can still be retained in

the amorphous matrix of TFMG prepared by sputtering. As an example, Zr-based TFMGs with immiscible Ta are successfully fabricated via co-sputtering[21,37]. The Ta content in Zr-based TFMGs ranges from 0 to ~75 atomic percent (at.%); thus, the GFA is distinctly larger than those of BMGs. When the Ta content is over 75 at.%, the amorphous TFMG structure gradually transforms to include nanocrystalline β-Ta, as shown in the TEM results in Fig. 9 [20,36]. The mechanical properties are strongly inﬂuenced by the Ta content.

Theﬁlm's elastic modulus remains nearly constant with 0–50 at%

Ta, but increases rapidly at 50–100% Ta. The key factor appears to be the population of strong Ta\Ta bonds. The highest elastic modulus and hardness, ~ 140 GPa and 10 GPa, respectively, of Ta-rich TFMGs, much higher than for Zr-, Cu and Ti-based amorphous counterparts, might be useful for applications in the area of hard coatings or interlayers for diffusion barriers.

3.6. Mechanical properties

3.6.1. Residual stress

Residual stress, either in compression or tension, in magnetron sputtered films is associated with the film/substrate lattice misfit, energetic-particle bombardment, and deposition conditions such as working pressure and substrate bias. The presence of residual stress plays a vital role in altering mechanical properties offilm and substrate, includingfilm hardness, adhesion, and substrate fatigue. In a previous study on a Zr-based TFMG (Zr47Cu31Al13Ni9) [38], a 200 nm-thickfilm is reported to exhibit a compressive residual stress of 88.2 MPa on a Si(001) wafer, as determined by substrate curvature measurements. According to Stoney's equation,

σf¼ M_s 6 1ð −νsÞ

t_s²

tf

! 1

R− 1 R0

; ð1Þ

whereσfis the residual stress in thefilm, tfthefilm thickness, tsthe substrate thickness, Ms the biaxial modulus of the substrate (180.3 GPa), andνsis the Poisson's ratio of the substrate. R0and R are the radii of curvature of the substrate before and after thefilm deposition, respectively. Small compressive stress in thinfilms is generally favorable for increasing the resistance to crack initiation and inhibition of crack propagation [38]. Improvement of fatigue- properties on steel substrates due to the presence of Zr-based TFMGs will be discussed in a subsequent section of this article.

It is important to further evaluate the influence of deposition conditions (such as substrate bias) on thefilm residual stress. The compressive stress in Zr-based TFMGs increases from 88.2 MPa without bias to 764.3 MPa with 125 V bias, an increase of ~8.6×. The residual stress drops noticeably to 441.7 MPa, a ~42% decrease, when an ~10- nm-thick Ti buffer layer is deposited between the bias-grownfilm and the Si substrate.

The results inFig. 10show the compressive stresses present in Zr- based TFMGs and the increasing trend of stress withﬁlm thickness when a−125 V bias is applied to the Si substrate during deposition.

Bias-deposited Zr-based TFMGs have a compressive stress of 7.8 GPa, when theﬁlm thickness is 800 nm.

The effect of bias on theﬁlm microstructure is also pronounced, as shown in the TEM micrographs ofFig. 11. It is seen that theﬁlm in this case becomes much denser, the column size decreases to

~11 nm, and nanocrystallites are not apparent, as compared with that deposited without bias inFig. 4. The signiﬁcant increase in the residual stress andﬁlm density is attributed to the ion bombardments when the bias is applied.

3.6.2. Nanoindentation

To characterize the mechanical properties of TFMGs, nanoindentation is a convenient approach since theﬁlm hardness and Young's Fig. 7. (a) DSC and (b) electrical resistivity results for Cu-based TFMGs in as-deposited

and annealed (for one min) conditions. Dashes lines indicate approximateΔT window [28].

(8)

modulus can be obtained experimentally[10,39,40]. However, as with otherfilm materials, care should be taken when interpreting the experimental results derived directly from indentation load-displacement curves using the well established Oliver and Pharr (OP) method for bulk materials[41]. In principle, when the OP method is utilized in data processing for a TFMG, the indentation hardness and modulus obtained reflect not only the mechanical attributes of the film, but also those of the underlying substrate. To circumvent this difficulty, a sequential-nanoindentation approach can be employed to obtain the

‘intrinsic’ properties of the ﬁlm through data ﬁtting.Fig. 12(a) displays a typical load-time curve consisting of multiple loading and unloading cycles for sequential nanoindentation. As an analog of a conventional nanoindentation test, each cycle is composed of loading, holding, and unloading segments, from which the hardness and Young's modulus data can be derived using the OP method as a function of the indentation depth as shown inFig. 12(b).

As a demonstration, Zr-based (Zr53Cu29Al12Ni6) TFMGs with thicknesses of 200, 400, 600, 800, and 1000 nm on Si(001) substrates were used for the nanoindentation evaluation. Nanoindentation tests were carried out with resolutions of 0.1μN in load and 1 nm in displacement.

Young's modulus and hardness of a bare Si substrate werefirst determined to be 187±8 GPa and 12.5±0.4 GPa, respectively, in the range of indent depths from 300 to 1500 nm. These values are consistent with those reported for Si(001)[42]. Sequential nanoindentation measurements were then performed onfilms with different thicknesses. Young's modulus and hardness values display strong indentation-depth dependence, particularly for normalized contact depth h/tfexceeding ~10%, as shown inFig. 13(a) and (b), where h and tfdenote indentation-depth andfilm thickness, respectively.

To extract Young's modulus of theﬁlm, a modiﬁed King's model is used[43]:

1 E_r¼1−v²i

E_i þ1−v²f

E_f 1−e⁻^{γ tf −h}ð Þ

a

! þ1−v²S

E_S e⁻

γ tf −hð Þ

a ; ð2Þ

where Eris the reduced‘composite’ modulus of the film/substrate system obtained via the original OP method[41]. Ei,Ef, Esandνi,νf, νsdenote the Young's modulus and the Poisson's ratio of the diamond indenter,film, and Si(001) substrate, respectively, a the effective contact radius approximated as a≈2.8 h for the Berkovich nanoindenter, andγ is a weighting factor that accounts for the continuously changing contribution of thefilm and substrate to the overall stiffness[43].

According to this equation, it can be readily seen that the reduced modulus of thefilm/substrate system is dominated by the properties of thefilm only when h/tfapproaches zero. Also from this equation, the trend of the reduced composite modulus vs. h/t_fcan befit as shown in Fig. 13(a). Here, the values of Eiandνiare taken to be 1140 GPa and 0.07, respectively, for the diamond indenter, whileνfandνsare assumed as 0.36 and 0.28 for Zr-based TFMG and the Si substrate[44,45]. Tofit the experimental data, Ef, Esandγ are treated as fitting parameters. Through nonlinear datafitting, the values of the three fitting parameters were obtained, as tabulated inTable 1.

Similar to the trend for the reduced composite modulus, the hardness of theﬁlms also levels off as h/tfapproaches zero, implying a dimin- ishing substrate effect when the indent depth decreases. According to the result inFig. 13(b), the depth-insensitive hardness for the current ﬁlm/substrate system corresponds to h/tf~ 0.1. However, it is worth Fig. 8. MFM images (5 × 5μm²) of Fe-based TFMGs as-deposited and annealed for 1 min at several temperatures[27].

(9)

noting that there is a limit for continuously reducing the indent depth in a nanoindentation test, below which uncertainties arise, due to imper- fections in the nanoindenter tip. For this example, the minimum indent depth is approximately 30 nm, which is ~15% of the 200 nmfilm thickness. As a consequence, the hardness of the 200-nmfilm seems not yet saturated and still remains above the value devoid of the substrate effect [Fig. 13(b)]. For the sake of comparison, the apparent hardnesses of thefilms obtained at h/tf~ 0.15, which is closest to the truefilm property, are listed inTable 1.

From the tabulated results, a thickness effect on theﬁlm modulus is observed. As the thickness decreases from 1000 nm to 200 nm, the corresponding modulus increases by nearly 20%, from 107 GPa to 128 GPa.

However, considering the 10% data scattering in the modulus measurement [the error bar inFig. 13(a)], such an increase in the Young's modulus may not be significant. However, at a relatively large h/tf, there is a tendency to overestimate the Young's modulus of the TFMG due to the indentation pile-up effect[46]. Thus, thefitted Efvalues may be system- atically biased if the majority of the data points used for datafitting are affected by indentation pile-up, which could result in afitted Efhigher than its actual value, particularly for the relatively thin TFMGs, as shown in the AFM analyses of nanoindentations inFig. 14. In thisfigure, the Fig. 9. Bright-field TEM images of Zr-based TFMGs with various Ta contents: (a) Zr45Cu27Ti15Ta13, (b) Zr41Cu23Ti12Ta24, (c) Zr31Cu15Ti10Ta44(d) Zr10Cu5Ti3Ta57(e) Zr19Cu6Ti7Ta68, (f) Zr14Cu7Ti5Ta74, (g) Zr10Cu5Ti3Ta82, and (h) Zr4Cu3Ti1Ta92. Selected area diffraction patterns are shown in insets[21].

Fig. 10. Relationship between the residual compressive stress andﬁlm thickness of Zr-based TFMGs sputter deposited with a−125 V bias on Si(001).

Fig. 11. Cross-sectional TEM bright-ﬁeld (BF) image, with a selected-are diffraction pattern in the inset, and a Z-contrast high-angle annular dark-ﬁeld (HAADF) micrograph of a Zr-based TFMG sputter deposited with a−125 V bias on Si(001). The Pt cap is deposited as a protective layer for FIB sample preparation.

(10)

pile-up is more apparent when thefilm thickness decreases. Considering all these facts, the modulus of the MGfilms is taken as the average of all fitted results, which is 117±9 GPa, in order to avoid over-estimation.

This gives a Si substrate modulus of 183±8 GPa, which is in excellent agreement with the result obtained from nanoindentation of the bare Si substrate.

Compared to theﬁlm modulus, the ﬁlm hardness measured at h/

tf~ 0.15 oscillates around ~5.5 GPa, displaying no clear sign of a thickness effect except for the 200-nmﬁlm exhibiting a hardness of ~6.5 GPa, ~20%

higher. While the hardness does not appear to increase withﬁlm thickness, other thickness effects may be playing in role. These include residual stress and underlying substrate and indentation pile-up[47].

3.6.3. Microcompression

In the MG literature, microcompression measurements on micropillars, fabricated using the focused ion beam (FIB) technique, have been used to study the size effect on shear-banding mediated plasticity [48–52]. This is because the macroscopic deformability of MG at room temperature is a shear-band dominated phenomenon and shear band formation becomes important to study at the small scale. The emphasis here is placed on the application of microcompression methodology, emphasizing its functionality as a tool for mechanical characterization of TFMGs. Unlike nanoindentation, microcompression enables nearly uniaxial-loading on micropillars, which makes it attractive as an analog of the conventional compression test, thus facilitating the interpretation of experimentalﬁndings. However, issues like the geometrical im- perfection of the FIB-milled micropillars (pillar tapering)[46], the substrate compliance[46], and FIB-induced surface radiation damage[48],

complicate the translation of experimental data into the intrinsic material properties; this can lead to serious concern as to whether the mechanical properties obtained from microcompression are comparable with those obtained from bulk testing[53]. For TFMGs, further difﬁculty stems from the underlying supporting substrate, which could result in a biased estimation of theﬁlm mechanical properties if not accounted for appropriately. Despite these pitfalls, microcompression methodology has been widely utilized for mechanical testing of a variety of materials [54].

Similar to nanoindentation, control experiments mustﬁrst be performed for characterizing the mechanical properties of the Si(001) substrate. To this end, micropillars are fabricated on the surface of the wafer Fig. 12. Typical sequential nanoindentation results from a 600-nm-thick Zr-based

TFMG on Si(001) substrate: (a) load-time and (b) load-displacement (depth) curves. Fig. 13. Variations of the (a) reduced modulus and (b) hardness as a function of the normalized contact depth h/tffor Zr-based TFMGs on Si(001) substrates. The error bars correspond to the standard deviation of the experimental data obtained at different indentation sites. The solid curves in (a) are plotted based on a modiﬁed King's model[43]. The shaded area in (b) indicates the range in h/tfwhere the indentation hardness becomes insensitive to the indent depth.

Table 1

Physical properties extracted from sequential nanoindentation analyses of Zr53Cu29Al12Ni6

TFMGs with different thicknesses (tf). Esis Young's modulus of the Si(001) substrate and Efis Young's modulus of theﬁlm. Note that the hardness H corresponds to that at normalized contact depth h/tf~ 0.15 where h is the contact depth.

tf(nm) Es(GPa) Ef(GPa) H (GPa)

200 184 128 6.5 ± 0.1

400 191 125 5.6 ± 0.2

600 189 117 5.7 ± 0.2

800 178 110 5.2 ± 0.1

1000 172 107 5.4 ± 0.1

(11)

using FIB. In the microcompression literature, there are two approaches to FIB machining, i.e., the ion-beam lathe[55,56]and the sequential ion-beam milling approach[57]. The former approach enables the machining of materials to form non-tapered cylindrical pillar shapes, but it is time consuming and limited to micro-sized pillars. In contrast, the latter approach is time efﬁcient for obtaining pillars at the submicron size, but generally results in a tapered pillar shape[46,54]. Thus, it is generally a compromise decision to select the appropriate approach which ﬁts a particular research object. Here, the sequential ion-beam milling method is adopted to prepare micropillars for microcompression of TFMGs. The details of FIB milling have been thoroughly described else- where[46,57].

Fig. 15(a) shows the typical shape of a tapered micropillar milled into an uncoated Si(001) wafer. By replacing the Berkovich nanoindenter with aﬂat-end diamond nanoindenter, 20 μm in diameter, the same nanoindentation system can be modiﬁed for microcompression under either load or displacement control. After microcompression, the deformed or fractured micropillars are examined by electron microscopy.

As shown inFig. 15(b), a Si micropillar with a diameter of 1μm and an aspect ratio of ~2 shattered into many pieces after microcompression, a signature of the intrinsic brittleness of Si, even at the micro-scale. Like- wise, micropillars coated with TFMGs can be fabricated using FIB, as shown inFig. 15(c) to (f) for Zr53Cu29Al12Ni6TFMGs. In this way, one can vary the shape and size of the micropillars for a systematic investi- gation of the substrate effect, which is analogous to changing the indent depth in nanoindentation.

Fig. 16(a) displays typical load-displacement curves obtained from a set of micropillars with dimensions 1μm×3 μm. The experimental results include uncoated Si and Si coated with Zr₅₃Cu₂₉Al₁₂Ni₆TFMGs, from which it can be seen that the uncoated Si micropillars fractured at much higher loads than those causing yielding in the TFMGs. Since the uncoated Si micropillars behave elastically before fracture, the mechanical response of the TFMGs can therefore be extracted by subtracting the displacement of the Si underlayer from the total displacement measured, as shown inFig. 16(b). For a monolithic Si micropillar, the load-displacement response in microcompression can be expressed as:[46]

P¼ ESif H D; β; v;ρ

D

πD²

H δSi; ð3Þ

where P is the applied load; E_Si, H, D,β, ρ and ν denote the Young's modulus, height, diameter, taper angle, radius of curvature at the base, and Poisson's ratio of the Si micropillar, respectively, as shown inFig. 16(b);

δSiis the displacement of the Si micropillar. Here, f is the dimensionless function deﬁned in Reference[46]. Because of the complexity of this function, for the sake of brevity, readers are suggested to refer to the original paper[46]for further details.

The displacementδfoccurring in the TFMG is (δ−δSi), whereδ is the total displacement measured during microcompression. It should be noted that ESidiffers from that measured from nanoindentation owing to the elastic anisotropy of Si. E_Sican be extracted directly from microcompression of uncoated Si micropillars. Due to the uniaxial stressfield applied for microcompression, ESishould be close to the out-of-plane modulus along theb100> direction, while the indentation modulus re- flects a value averaged over those sampled by the complex indentation stressfield. Based on experimental microcompression data , ESiis found to be ~130 GPa, which is consistent with the theoretical value[45].

Once ESiis known, the TFMG load-displacement curves can be extracted, as shown inFig. 16(c). From these curves, thefilm's Young's modulus and apparent yielding strength can be estimated by assuming a uniform film cross section and using the data from films on micropillars with a taper angle less than 2°. In doing so, the Young's modulus of the TFMGs was found to be 108 ±15 GPa, irrespective offilm thickness. This is very close to the nanoindentation results.

Unlike the Young's moduli of TFMGs, their apparent yielding strengths exhibit strong shape dependence, as shown inFig. 17. Regard- less of theﬁlm thickness, the apparent yield strengths remain approximately constant at ~2.6 GPa for aﬁlm thickness to pillar diameter ratio, tf/D, greater than ~0.5, while they increase dramatically for tf/Db0.5.

This phenomenon of shape dependence has a simple physical explana- tion. According to Packard et al.[58], global yielding occurs in MGs when material on a potential shear plane reaches the yielding point.

For TFMGs with a high aspect ratio tf/D, the stress state is closer to uniaxial and it is easy to form a potential shear plane without the hindrance of the underlying substrate. However, for TFMGs with a low aspect ratio, the stress state is closer to triaxial, particularly in regions near theﬁlm/substrate interface. In this case, global yielding does not occur until the applied stress is increased to a higher level to cause local yielding of material at the interface where a strong compressive stress exists.

That is to say, the substrate effect arises once a potential shear plane is impeded by thefilm/substrate interface. Assuming a 45° shear angle, one can then estimate the critical aspect ratio tf/D triggering such a shape effect to be approximately ~0.5, at which point the shear plane from the film center just intersects the interface (see the insets in Fig. 17). Based on these experimentalfindings, it is concluded that the compressive yield strength of Zr53Cu29Al12Ni6TFMGs, with minimized substrate effects and therefore comparable with that of their bulk counterparts, should be ~2.6 GPa.

It is worthwhile to point out that, even though BMGs are quite brittle under uniaxial loading without noticeable plasticity at room temperature, TFMGs appear very ductile as can be seen from load- Fig. 14. Section analyses (left) and AFM height images (right) of nanoindentations in

Zr53Cu29Al12Ni6TFMGs of various thicknesses, showing the thickness effect on indentation pile-up. All the indents were made at the same peak load of 8 mN.

(12)

displacement curves inFig. 16(c). In general, this size-effect behavior can be attributed to stable shear banding in MGs at a small size scale (typically less than 1 mm), which is then related to a size-affected elastic energy release upon yielding[48,49,51,52,59–61]. However, the physical/structural origin for such a size effect is not well understood at the present time, and still remains an important ongoing research topic. In summary, it has been demonstrated that TFMGs possess high mechanical strength because of their amorphous structure, and good ductility owing to their small thickness. Such a unique integration of mechanical properties renders TFMGs promising materials for a variety of engineering applications[10,38,62,63]in recently emerging MG- based micro- and nano-technology that have already attracted research interest[60,62,63].

3.6.4. Multilayered TFMGs

In addition to monolithicfilms, multilayered TFMGs have attracted attention in recent years. For mechanical property improvement, multilayered systems consisting of thin layers of nanocrystalline metal and metallic glass are prepared[64–68]. The metal layer should be suffi- ciently strong in modulus and strength while being deposited with the appropriatefilm orientation. Face centered cubic Cu(111) films appear to be too soft, whereas body-centered cubic Mo(110)films are too brittle.

Hexagonal close-packed Zr(001) matches the above requirements [65,68]. During deformation of multilayered TFMGs, the shear bands

initiated in the TFMG layer can be absorbed and accommodated by the nanocrystalline Zr(001) layer via nano-twinning. Zr-based-TFMG/Zr multilayers can be highly ductile with semi-uniform plastic deformation of 55%. They are even more ductile than many pure metals, as shown inFig. 18 [68]. The amorphous-crystalline interface exhibits good strain compatibility after appreciable plastic deformation.

Another example is Zr-based-TFMG/crystalline-Ta multilayers. FIB- machined multilayer micro-pillars have been evaluated by microcompression tests[37], as shown inFig. 19. Compared with monolithic Zr-based-TFMG pillars, the TFMG/Ta multilayer pillars exhibit consider- ably higher yield stress as well as improved deformability. The en- hanced mechanical properties are attributed to positive effects from high-stiffness Ta layers, which are not involved in formation of shear bands, but can retard shear band propagation by forming a“plastic zone” in the constrained Zr-based-TFMG layers.

The tensile behavior of multilayer TFMGs has been also reported.

Monolithic Zr-based TFMGs and Zr-based-TFMG/Cu multilayer coatings on pure Cu foils have been examined[69]. The extracted tensile modulus and strength of 1-μm-thick multilayers are in good agreement with rule-of-mixtures predictions. The results reveal that the multilayer coatings exhibit much better tensile performance than monolithic TFMGs, despite the brittle deformation.

In addition to nanocrystalline metal interlayers, TFMG interlayers of different composition have also been used to form TFMG/TFMG Fig. 15. SEM micrographs of (a) as-fabricated and (b) fractured Si micropillars on an uncoated Si(100) substrate; (c) to (f) as-fabricated micropillars, different sizes and shapes, coated with Zr-based TFMGs.

(13)

multilayered structures[70,71], as shown inFig. 20. An example is the Zr-based-TFMG/Pd-based-TFMG multilayered system, whose uniaxial microcompression and nanoindentation responses have been investigated[70]. It is found that the apparent deformation mechanism transforms from a highly inhomogeneous mode in monolithic amorphous alloys to a more homogeneous mode in micropillars of TFMG/TFMG

multilayers. Similar phenomena were observed under nanoindentation.

The presence of sharp amorphous/amorphous interfaces, which could hinder the propagation of shear bands, is a possible reason for the observed transition in the deformation mode.

3.7. Adhesion and tribological properties

Among the mechanical properties of TFMGs, there are few reports of adhesion and tribological properties[24,72,73]. In general thinﬁlm technology, scratch adhesion tests are frequently used for coating evaluation. Qualitative and quantitative information about the adhesion of a coating to a substrate can be obtained from nano/micro/macro scratch experiments. A scratch test provides not only the critical load to failure, but also the type of failure mode[74–76].

For BMGs, the scratch resistance of Mg-based-BMGs has been evaluated using a nanoindenter with a Berkovich probe[77]. The wear mode changed from rubbing and/or ploughing to cutting as the normal load was increased from 5 to 200 mN. The tribological properties of Zr- based BMGs have been examined using ramping-load nanoscratch and multiple-sliding nanowear techniques[78]. The best wear performance is obtained for structures consisting of an amorphous matrix with dispersed nanocrystalline particles. A study of the tribological be- haviors of Ce-, Ti- and Fe-based BMGs by ramping-load nanoscratch tests reported that higher hardness correlated with better scratch resistance[79].

The tribological behavior of Zr-based TFMG (Zr47Cu31Al13Ni9) deposited on 316 L stainless steel, probed using the micro-scratch technique, was reported by Liaw et al.[73]. For this TFMG, a penetration depth profile, the coefficient of friction (COF) during the scratch test, and an SEM image of the scratch track after ramp-loading of 25–250 mN are shown inFig. 21 [73]. The critical load is determined to be 110 mN and is marked with a vertical arrow. The average COF value was 0.23 at a 100 mN normal load. When the normal load increased to 300 mN, the COF value increased to ~0.62. The scratch tracks near thefilm failure area are shown in the SEM micrographs ofFig. 22 [73]. It is clear that a high density of shear bands is formed along the sliding direction. However, no obvious debonding is observed. The smooth scratch groove and the small shear band spacings indicate a ductile deformation mode with goodfilm plasticity.

For sputtered Zr-based-TFMGs on stainless steel substrates, Jang et al.

found that adhesion increased with sputtering power to a saturated critical scratch-test load of 70 N[72]. Conversely, poor adhesion properties of sputter-deposited Pd-based TFMGs on Si(001) substrates have also Fig. 16. (a) Load-displacement curves obtained from TFMG-coated Si(001) micropillars

with pillar diameters ~ 1μm, height ~3 μm, and the taper angle ~2°, (b) a schematic illustration demonstrating the method for extracting the mechanical properties of the TFMG by subtracting the mechanical response of the Si substrate from the total response; and (c) extracted load-displacement curves for the Zr53Cu29Al12Ni6TFMGs.

Fig. 17. The shape dependence of the yield strengths obtained from Zr53Cu29Al12Ni6

TFMGs of different aspect ratios (the insets on the top illustrate the aspect-ratio effect on shear band formation). tfis theﬁlm thickness and D is the substrate diameter.

(14)

been reported[24]. Hence, the adhesion properties of TFMGs show wide variations with composition, deposition technique, and substrate.

In recent work, a 210-nm-thick Zr-based TFMG (Zr58Cu25Al11Ni6) deposited on Si(001) by pulsed DC magnetron sputtering[80], with a synchronous pulsing bias of−50 V applied to the substrate, has been evaluated for adhesion and tribological properties. The adhesion behavior was examined using a scratch tester with the normal load from 0.1 to 50 N and scratch lengths of 5 mm.Fig. 23illustrates the coefﬁcient of friction (COF) and friction force versus scratch length, together with SEM images of surface morphologies of scratch tracks. It is clear that the mean COF value was around 0.3–0.4. This result is similar to that of Zr47Cu31Al13Ni9metallic glassﬁlms on Si(001) reported by Liaw et al.[73]. No apparent crack or delamination was found adjacent to the scratch track of the coating, suggesting good adhesion.

3.7.1. Rockwell-C adhesion test

The Daimler–Benz Rockwell-C (HRC-DB) adhesion test[81]is an easy and commercially adopted method to evaluatefilm/substrate adhesion quality. A load of 1471 N is applied to causefilm damage, and the degree of the damage is classified into six grades, HF1–HF6. In general, the adhesion strength quality HF1 to HF4 defines sufficient adhesion, whereas HF5 and HF6 represent poor adhesion in the HRC-DB test. The adhesion of Zr55Cu27Al12Ni6 TFMGs deposited on tool steel (SKD 61) substrates by pulsed-DC magnetron sputtering has been evaluated by using Rockwell-C adhesion tests and compared to results for CrN[82]and TiNfilms[83]deposited on SKD 61 and high speed steel

substrates, respectively, by the same method. Thefilm thickness, hardness, and elastic modulus of TFMG, CrN and TiNfilms are listed in Table 2. As expected, the TFMG, due to its metallic nature, has lower hardness and modulus than those of hard ceramic CrN and TiNfilms.

The HF values obtained from the HRC-DB test listed in this table reveal excellent adhesion; HF1 for allfilms examined. Different surface morphologies are observed in SEM micrographs (Fig. 24) obtained from the indentation craters after the adhesion tests. For the TFMG, there is negli- gible radial cracking andfilm delamination; however, brittle fracture features are observed for both the CrN and TiNfilms. The brittle features include many circular bulge wrinkles adjacent to the crater in CrN and a number of brittle cracks emanating radially from the crater in TiN.

This suggests that the Zr-based TFMGs exhibit good adhesion to steel Fig. 18. (a) SEM micrograph showing a TFMG/Zr (100/100 nm) multilayer on Si(001)

pillar compressed to a predetermined displacement of 500 nm (∼25% strain), and (b) the recorded engineering stress/strain curve.

Fig. 19. SEM micrographs (a) monolithic Zr-based TFMG (b) Zr-based TFMG/Ta (50/

5 nm) multilayer, and (c) Zr-based TFMG/Ta (50/50 nm) multilayer on Si(001) pillars deformed by micro-compression[37]. The pillars, 900 nm in diameter by 1500 nm high, were formed using an FIB.

(15)

with better ductility than ceramic hard coatingﬁlms. A similar result has been obtained when TFMGs and the hard coating (such as TiN) are deposited on BMG surfaces for a comparison study, in which TFMGs are found to exhibit ductile feature whereas TiN is brittle[84].

3.7.2. Wear resistance

Wear resistance plays an important role in many applications. For BMGs, much effort has been devoted to examining tribological

properties. Ni-based BMG micro-gears have longer lifetimes than micro- gears made of steel[85]. The wear resistance and hardness of Zr-based BMGs increase with increasing Ag content over the range from 2 to 8 at.%[86]. A study of the tribological behavior of the B₄C-reinforced Fe-based BMG plasma-sprayed coatings reveals that the observed reduc- tion in wear loss, hardness increase of the wear surface, and inhibited plasticﬂow propagation are largely due to the incorporated hard material, B4C, in the amorphous coating[87]. The tribological behavior of the Fig. 20. Electron micrographs showing MG/MG multilayers on Si(001) pillars compressed to a displacement of 300 nm. SEM: (a) 5/50 nm Zr-based MG/Pd-based MG system, and (c) 50/50 nm Pd-based MG/Zr-based MG system. TEM: (b) 5/50 Zr-based MG/Pd-based MG, (d) 50/50 Pd-based MG/Zr-based MG. Higher resolution views of sample (d): (e) upper left corner, (f) upper center, and (g) layer deformation due to shear band propagation[70].

Fig. 21. Penetration depth and COF vs. the scratch distance along a Zr-based TFMG on 316 L stainless steel for ramp loading of 25–250 mN[73]. SEM micrograph taken from the test region is shown above the graph.

(16)

porous and WC particle-containing Fe-based TFMG coating layers has been explored, revealing excellent tribological and wear characteristics, attributable to the amorphous matrix[88].

However, a fundamental understanding of tribological properties of TFMGs is still lacking. Recent research on the wear resistance of Zr-based TFMGs deposited on SKD 61 steel using pulsed DC magnetron sputtering has been carried out. A pin-on-disk wear method was used to investigate the wear resistance. A SiC ball, 6 mm in diameter, was adopted as the stationary pin with a normal load of 1 N. The sliding speed was 9.8 mm/s with a wear track diameter of 6 mm and a wear length of 20 m.Fig. 25shows the COF versus wear length for both the TFMGs and the bare steel substrate. It was found that thefilm was al- most worn through at a wear length ~6.8 meters as the COF values of film and substrate merged. The wear scar morphology of the film after the pin-on-disk wear test for a wear length of ~228μm is shown in the SEM micrograph inFig. 26. No obvious debonding is found, but plastic deformation can be observed in many areas. Plastic deformation of the wear track might have caused a higher COF value for the TFMG in this case.

4. Potential applications

4.1. TFMGs for biomedical use

4.1.1. Antimicrobial

Nosocomial infection frequently occurs in hospitals. The infection is found to be caused by transmission of pathogens from the hospital environment (such as instruments or equipment) and by direct hand- to-hand contact between health-care providers and patients[89]. The most common causative pathogens have been reported as Staphylococcus aureus, Escherichia coli, Pseudomonas, and mixed-microbial infections [90,91]. In addition, Acinetobacter baumannii is another important path- ogen in recent years[92,93]. Generally, stainless steel is used for many

medical environmental surfaces such as door handles and push plates, and also in many different devices such as surgical instruments[91,94].

Michels et al.[95]reported that copper alloys or copper-containing materials exhibit high efficacy levels as antimicrobial materials at temperature and humidity levels typical of indoor environments, such as hospitals. Therefore, surface conditions of stainless steel located at possible hand-contacting regions (door handle, push plate) can be modi- fied by adding a thin coat of hard, smooth Cu-containing film such as Zr-based TFMGs (Zr61Cul7.5Ni10Al7.5Si4), which improve antimicrobial properties[72]. This TFMG has good hydrophobic properties, with a wetting angle of 92° compared to 46° for 304 stainless steel. Additionally, thefilms have extremely smooth surface profiles (roughness ~1 nm) and a homogenous composition distribution while exhibiting strong antimicrobial effects for different microbes including E. coli, S. aureus, Pseudomonas aeruginosa, A. baumannii, and Candida albicans for at least 24 h as shown inFig. 27 [72]. More recently, Jang et al. have reported a Ag-containing Zr-based TFMG [(Zr42Cu42Al8Ag8)99.5Si0.5], also deposited by sputtering on stainless steel surfaces[96]. The 500-nm-thickfilms have physical properties similar to Zr-based TFMGs: surface roughness (b1 nm), hardness (~6 GPa), adhesion (~70 N load), and amorphous structure as characterized by XRD and TEM (seeFig. 28). However, the Ag-containing TFMGs exhibit much better antimicrobial capabilities, as shown inFig. 29. No microbe growth of E. coli, C. albicans and S. aureus is found on surfaces coated with Ag-containing TFMGs after 76 h.

4.1.2. TFMGs for medical tools

Potential applications of Zr-based TFMGs can be extended to medical tools such as surgical blades and micro-surgery scissors.

Commercial blades made of martensitic stainless steel always present micron-scale roughness on the edge surface and edge tip[97]. This rough blade edge directly inﬂuences the blade sharpness as well as the surgical quality, and eventually the performance and life of the tool. The sharpness of a blade is a key parameter in cutting soft solids, Fig. 22. SEM micrographs of scratch tracks along a Zr-based TFMG on 316 L stainless steel under a ramp load of 25–200 mN at (left) lower and (right) higher magniﬁcations[73].

Fig. 23. (Upper) SEM micrographs in secondary electron (SEI) and backscattered electron (BEI) modes and (lower) friction force and COF vs. the scratch distance along a Zr-based TFMG on Si(001).

(17)

such as biological tissues or elastomeric materials, and it is related to the energy required to cut, the quality of the cut surface, and the life of the cutting instrument. In addition, the depth of blade indentation neces- sary to initiate a cut or crack in the target material is a function of sharpness of the blade's cutting edge, characterized by the“blade sharpness index” (BSI) [98]. BSI is deﬁned as ∫_δ

iFdx=δitJ_Jc and illustrated in Fig. 30, where F is the applied force, dx the increment of blade displacement,δithe initial depth of blade indentation before test material fracture, l the length of cut surface, JIcthe Mode I fracture toughness (crack opening directly to two sides), and t the thickness of test material. The index is zero for an inﬁnitely sharp blade and increases with bluntness in a quadratic manner.

The blade sharpness index BSI of a commercial stainless steel blade (Surgical Blade No.11, FEATHER Safety Razor Co.) with and without sputter-deposited Zr-based TFMGs has been evaluated by cutting a rubber material (styrene-butadiene rubber, SBR) in a specially-designed test rig (seeFig. 31). The setup is mounted on a 50 kN universal testing ma- chine for measuring the applied force. The SEM results inFig. 32 [96]

clearly show that the surface roughness of a commercial scalpel (with

~1.2μm average roughness) can be signiﬁcantly decreased to ~0.4 μm by coating it with a 200-nm-thick Zr53CuNiAlSi TFMG. The coating also decreases the BSI value from ~0.36 to ~0.25, a 26.5% improvement in sharpness.

4.2. Improved fatigue properties due to TFMGs

Beneﬁcial effects of TFMGs on fatigue-properties of crystalline metallic substrates have been reported and some important results are summarized in this section. The TFMGs investigated to date are primarily sputter-deposited Zr47Cu31Al13Ni9, Cu51Zr24Hf18Ti7, and Fe65Ti13Co8Ni7B6Nb on substrates including stainless steels[38,99], Ni-[100], Al-[96], and Ti-[101]based alloys. Substrates are polished withﬁne grits, or electropolished, prior to TFMG deposition. The coated substrates are then subjected to tension during four-point bending fatigue tests at room temperature. The experiments are conducted with an R ratio of 0.1, where R =σmin/σmax, in whichσminandσmaxare the applied minimum and maximum stresses, respectively, at a frequency of 10 Hz.

Theﬁrst known study reporting fatigue property improvement was for 200-nm-thick Zr-based TFMG coated 316 L stainless steel substrates[38]. The stress versus number of fatigue cycles to failure Table 2

Film thickness, hardness, elastic modulus and adhesion of Zr55Cu27Al12Ni6TFMG, CrN [79], and TiN[80]layers.

Films Thickness

(nm)

Hardness (GPa)

Elastic modulus (GPa)

Adhesion strength quality (HF value)

Zr-based TFMG 500 5.9 ± 0.1 109.2 ± 2 1

CrN 1500 22.2 ± 3.8 239.7 ± 26.3 1

TiN 3000 32 ± 4 470 ± 50 1

Fig. 24. SEM micrographs of indent craters in Zr-based TFMG/SKD61, CrN/SKD61[82], and TiN/HSS samples[83]following Rockwell-C adhesion tests.

Fig. 25. The coefﬁcient of friction of a 500 nm-thick Zr-based TFMG/steel sample and the bare substrate against a SiC ball as a function of wear length.