Effects of surface tension and viscosity on gold and silver sputtered onto liquid substrates Free

Metal nanoparticles and films can be produced via sputtering onto low vapor pressure liquids such as oils and ionic liquids for a range of applications in optics,¹ electronics,^2,3 and catalysis.⁴ The sputtering process parameters⁵ and the physical and chemical properties of the liquid can affect the morphology of the deposited metal.^6,7 While some studies acknowledged the effects of viscosity or surface tension on nanoparticle size and film structure, other studies showed no effect or neglected the effects of the liquid properties.⁶ In order to resolve this conflicting information, we have systematically studied DC magnetron sputtering of gold and silver onto a wide range of liquids in order to deconvolute the effects of surface tension and viscosity. For each system, we pipetted 0.5 ml of liquid onto a 2.5 × 2.5 cm glass slide, allowed the liquid to spread over the entirety of the glass slide, and placed the sample 50 mm underneath the sputtering target. DC magnetron sputtering is advantageous because of its ability to confine the target atoms to the substrate and allow for efficient deposition without degrading the substrate.⁸ We kept the sputtering conditions constant and chose a series of low vapor pressure liquids to isolate the physical properties from chemical interactions such as electrostatic interactions.^9,10 We chose a series of silicone oils with viscosities of 48, 96, 339, and 485 cP because their similar chemical makeup and surface tension allows us to isolate the effects of viscosity. In order to study the effects of surface tension, squalene, poly(ethylene glycol) (400 MW PEG), and glycerol were chosen because of their lack of charged chemical moieties. Our previous studies have shown that the deposition of polymers onto liquid surfaces is governed by the spreading coefficient $S=γLV1+cos θ−2γPV$⁠, where $γLV$ is the liquid-vapor surface tension, $θ$ is the advancing contact angle of the liquid on the polymer, and $γPV$ is the polymer-vapor surface tension.^11,12 Particles form when the spreading coefficient is negative, whereas films form when the spreading coefficient is positive. The key difference between the deposition of polymers and metals is that the surface energy of metals is much higher than polymers, and therefore, the spreading coefficient will always favor particle formation.¹³ 

Our hypothesis is that viscosity and surface tension impact the rate at which the liquid wets the deposited metal which determines whether the sputtered nanoparticles deposited onto the liquid surface will diffuse and aggregate, submerge into the bulk liquid, and diffuse and aggregate within the bulk. Nakamura et al. showed that the rate of wetting for liquid argon on platinum at the nanoscale was proportional to the liquid surface tension and inversely proportional to the liquid viscosity.¹⁴ Therefore, we expect the high surface tension liquids to have a faster rate of wetting than the low surface tension liquids and high viscosity liquids to have a slower rate of wetting than the low viscosity liquids. Additionally, the liquid viscosity is an important factor because it directly impacts the nanoparticle diffusion velocity and aggregation behavior. The nanoparticles in our study are approximately 4 nm in radius, which is similar to the size measured by others in comparable sputtering systems,¹⁵ and therefore, the energy required for these particles to penetrate the liquid surface into the bulk varies from ∼20 to 60 eV. Previous studies have found that sputtered metals have energies in the range of 5–15 eV irrespective of the applied voltage,¹⁶ and therefore, the sputtered metal nanoparticles likely do not penetrate the surface. These nanoparticles impinge on the surface with a stagnation pressure (∼10 MPa) less than their hardness values (∼3 GPa),¹⁷ and therefore, the nanoparticles should not deform. The nanoparticles at the vapor-liquid interface may submerge into the bulk liquid leading to aggregation behavior within the bulk.¹⁸ The minimum energy required for a particle to submerge into the liquid is given by the equation ΔG = πr²γ (1 − cos θ_e)² for 0 ≤ θ_e ≤ 90°, where r is the particle radius, γ is the liquid-vapor surface tension, and θ_e is the equilibrium three phase contact angle.¹⁹ Our results indicate that varying the viscosity and surface tension can lead to nanoparticles within the bulk liquid, aggregated nanoparticles within the bulk, or films comprised of branch-like (ramified) aggregates or dense (clustered) aggregates.

In order to systematically study the effect of the viscosity of the liquid substrate on the sputtered gold, we performed depositions onto 48, 96, 339, and 485 cP silicone oils at constant sputtering parameters of 0.8 mbar and 20 mA for 30 s. In order to visualize the deposited gold at various length scales, the samples were imaged using a camera, an optical microscope, and a transmission electron microscope immediately after deposition (Fig. 1). The camera images of the 48 and 96 cP silicone oil systems show gold nanoparticles suspended in solution [Fig. 1(a)]. We confirmed the presence of nanoparticles dispersed throughout the z-direction using optical microscopy [Fig. 1(b)]. Since silicone oil has a very low contact angle of approximately 4° ± 2° on gold, it is energetically favorable for the particles to submerge into the bulk liquid rather than aggregate on the surface of the liquid to form a film. No film was formed on the surface of the liquid even when the deposition time was increased to 5 min because the gold nanoparticles deposited at the surface continuously submerge into the bulk silicone oil. The optical microscope image of the 48 cP silicone oil system shows aggregates on the order of tens of microns within the bulk liquid. These aggregates are bigger than in the 96 cP silicone oil because the diffusion of the nanoparticles is inversely proportional to viscosity [Fig. 1(b)]. The aggregation behavior in the 48 cP silicone oil system is consistent with work by Cai et al.²⁰ that showed that the diffusion velocity of non-sticky nanoparticles in polymeric liquids is mainly dependent on the solvent viscosity. While the camera and microscope images were taken of the nanoparticles as deposited, the TEM images are from grids that were gently dipped into the bulk liquid immediately after sputtering, rinsed with hexane, and allowed to dry for 1 day. The TEM images show that nanoparticles aggregated in the 48 cP silicone oil system, whereas the nanoparticles are more dispersed in the 96 cP silicone oil system, which is consistent with the microscope images [Fig. 1(c)]. Image J analysis of 100 nanoparticles and nanoparticle aggregates from representative TEM images show that both systems consistently have sub-20 nm diameter individual nanoparticles and aggregates, with the 48 cP silicone oil resulting in more nanoparticle and aggregate sizes greater than 9 nm, indicating a greater degree of aggregation [Fig. 1(c)]. By increasing the sputtering time to 15 min to increase the gold concentration within the bulk, we were able to use UV-Vis spectroscopy to determine the relative size and relative concentration of gold nanoparticles.²¹ Since the TEM images show that the nanoparticle diameters are less than 20 nm, the larger ratio of the absorbance at the surface plasmon resonance peak (A_spr) to the absorbance at 450 nm wavelength (A₄₅₀) in the 48 cP silicone oil system (1.14) compared to the 96 cP silicone oil system (1.06) confirms a higher degree of aggregation in the 48 cP silicone oil system. After 48 h, the gold nanoparticles form larger aggregates so the surface plasmon resonance peaks can be directly compared to give relative sizes. A shift to 638 nm for the 48 cP silicone oil system indicated a greater degree of aggregation when compared to the shift to 551 nm for the 96 cP silicone oil system.

Thermodynamically, sputtered gold always favors nanoparticle submersion into the bulk silicone oil because the ΔG for submerging is close to zero. However, the camera and optical images of the 339 cP and 485 cP oils show that a film forms on the surface of the liquid [Figs. 2(a) and 2(b)]. Our hypothesis is that the film forms because the increase in viscosity results in a lower rate of wetting, and therefore, nanoparticles at the vapor-liquid interface accumulate due to adparticle deposition. In the microscope images, we observe wrinkling of the films along the edges in both the 339 cP silicone oil system and the 485 cP silicone oil system likely because the sputter deposition process slightly heats the surface causing the liquid surface to expand during deposition [Fig. 2(b)]. The wrinkles form as the surface cools along the stress-relieving edges similar to the wrinkles Yang observed when using thermal evaporation to deposit gold onto 175 cSt silicone oil.²² TEM images show that the surface structures of the 339 cP and 485 cP silicone oil systems have significantly different morphologies compared to their bulk counterparts. We observe films composed of dense nanoparticle aggregates with crack-like features [Fig. 2(c)]. There is more empty space between the structures in the 339 cP silicone oil system than in the 485 cP silicone oil system as evidenced by the light grey areas in the TEM images [Fig. 2(c)]. These crack-like features are likely due to a combination of wetting of the gold by the silicone oil and surface diffusion of gold aggregates. The TEM images of the 339 cP silicone oil bulk and the 485 cP silicone oil bulk shows a lower concentration of gold and less aggregation than the lower viscosity systems [Fig. 2(d)]. Image J analysis shows that the bulk contains mostly sub-8 nm gold nanoparticles [Fig. 2(d)]. The absence of a peak in the UV-Vis spectra of 339 cP and 485 cP silicone oil systems after 15 min of deposition further confirms that the concentration of gold in the bulk is insignificant because most of the deposited gold is likely incorporated into the films.

In order to study the effects of increasing the surface tension of the liquid, we used constant sputtering parameters and chose squalene (12 cP, 31.2 mN/m), 400 MW PEG (112 cP, 43.2 mN/m), and glycerol (946 cP, 64.3 mN/m). These liquids were selected instead of ionic liquids to minimize electrostatic interactions with the deposited metals. While increasing viscosity reduces diffusion and aggregation of the metal nanoparticles on the surface and within the bulk, increasing the liquid surface tension has the opposite effect since the liquid molecules prefer to minimize the surface contact area with the gold in order to minimize the free energy of the system, thereby causing aggregation. Nakamura et al. showed that the rate of wetting is proportional to the liquid surface tension.¹⁴ However, since the energy required for submerging (ΔG) is proportional to surface tension and the liquids have a nonzero contact angle on gold (14° ± 3°, 12° ± 2°, and 11° ± 3° for squalene, 400 MW PEG, and glycerol, respectively), nanoparticle aggregates are generated on both the surface and within the bulk. The camera image of gold deposited onto squalene shows ramified aggregates along the edges of the liquid and a continuous film-like structure in the center on the liquid surface [Fig. 3(a)]. The center structure likely forms due to fast surface diffusion due to a combination of low viscosity and high surface tension [Fig. 3(b)]. The TEM of this surface structure shows that the film is comprised of a network of ramified aggregates [Fig. 3(c)]. Dispersed gold nanoparticles are observed within the bulk of squalene likely because energy may be supplied to the system through heating of the liquid surface to overcome the ΔG requirement to submerge. Sputtering gold onto the 400 MW PEG results in large ramified aggregates on the surface of the liquid and smaller ramified aggregates within the bulk of the liquid [Figs. 3(a) and 3(b)]. In comparison to squalene, 400 MW PEG has a higher surface tension, resulting in denser structures both on the surface and within the bulk of the liquid [Figs. 3(c) and 3(d)].

Glycerol has the highest surface tension (∼63.4 mN/m) and viscosity (∼946 cP) of all the liquids tested. A film is formed on the surface since the high viscosity hinders diffusion and aggregation [Fig. 3(a)]. The film is unstable because the system wants to reduce the interfacial area between the gold and the glycerol [Figs. 3(a) and 3(b)]. The microscope images show fragments of the gold film which have aggregated [Fig. 3(b)]. Since the energy to break gold bonds²³ is much higher than the energy gained from aggregating the gold,^24,25 we may conclude that the aggregation occurs during deposition. The TEM image of the surface shows dense aggregate structures [Fig. 3(c)]. The TEM image of the bulk liquid shows ramified aggregates on the order of 100 nm despite the high viscosity [Fig. 3(d)].

Our data demonstrates that surface tension has an effect on the growth and aggregation behavior because dispersed nanoparticles readily form in the 96 cP silicone oil [Fig. 1(c)] whereas ramified aggregates form on the surface and within the bulk of the 400 MW PEG at a similar viscosity but nearly doubled surface tension [Figs. 3(c) and 3(d)]. While PEG may have a small electrostatic interaction with the gold, the effect is minimal and would promote nanoparticle stability rather than aggregation, allowing us to conclude that the ramified aggregates are a result of the high surface tension. The surface structures on each of the liquids (339 cP, 485 cP, squalene, 400 MW PEG, and glycerol) can be described by the diffusion-limited regime or reaction-limited regime. In the diffusion-limited regime, nanoparticles are strongly attracted to one another and tend to form open ramified aggregate systems.²⁶ In the reaction-limited regime, clustered or dense aggregates form more favorably since nanoparticle attraction is weak and surface diffusion is slow.²⁷ The surface tension data show that the diffusion-limited regime at low viscosity (squalene, 400 MW PEG) forms open ramified aggregates [Fig. 3(c)]. The system switches to a reaction-limited regime as the viscosity increases (glycerol), evidenced by the dense aggregates seen as black spots in the TEM images [Figs. 3(c) and 3(d)]. High viscosity silicone oils (339 cP and 485 cP) also form compact clusters because of the high viscosity and low surface tension driving force of the systems [Fig. 2(c)].

In order to understand whether our observed trends could be generalized to high surface energy materials, we repeated our set of experiments using silver. The camera images for both the 48 and 96 cP silicone oil systems show dispersed nanoparticles [Fig. 4(a)]. The TEM images show greater aggregation in the 48 cP silicone oil, which is in agreement with the gold results [Fig. 4(c)]. Similar to the gold system, film formation occurs on the higher viscosity systems (350 and 485 cP) [Figs. 4(a) and 4(b)]. TEM images of the 339 cP and 485 cP bulk show only sub-8 nm nanoparticles [Fig. 4(c)]. Since the 339 cP silicone oil system allows for a faster rate of wetting than in the 485 cP silicone oil system, more empty space is observed. Figure 5 shows a summary of the results for sputtering silver onto liquids of varying surface tension. Similar to gold, sputtered silver diffuses and aggregates according to the properties of the liquid. Squalene causes the silver to aggregate in the diffusion limited regime, creating ramified aggregates whereas the glycerol causes the silver to aggregate in the reaction-limited regime, creating densely packed aggregates. 400 MW PEG has an intermediate effect on the aggregation behavior of the silver. A morphological phase diagram to summarize our results is given in Fig. 6.

In conclusion, we elucidated the influence of viscosity and surface tension on the resulting morphologies of sputtered gold and silver onto low vapor pressure liquids. The resulting morphologies were nanoparticles within the liquid bulk, aggregated nanoparticles within the bulk, and films comprised of sparsely or densely aggregated nanoparticles on the liquid surface. These results allow for tuning the liquid properties to fabricate structures for applications in optics, catalysis, and electronics.

This work was supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences under Award No. DE-SC0012407. M.D.L. was supported by a National Science Foundation Graduate Research Fellowship under the Grant No. DGE-1418060.