In addition to more conventional features such as the thickness, chemical composition, or phase constitution, the control of the morphology at the submicrometer scale and the associated porosity is accepted as an essential parameter, allowing the properties of thin films to be tailored. Yet, for many years, the design and fabrication of nano-sculpted thin films have been recognized as a new opportunity to improve the performance of the thin films in a wide variety of applications ubiquitous in our society in the fields of microelectronics, information processing, as well as energy generation and storage. [1
] The main interest toward nano-sculpted films mainly arises from their high surface-to-volume ratio, allowing for an important developed surface that can be used as examples to accommodate a large quantity of molecules (i.e., in dye-sensitized solar cells) or to strongly increase the active sites (i.e., in photocatalytic materials).
Specifically addressing the domain of solar energy harvesting, one of the most frequent nano-sculpted film architectures is based on fritted nanoparticles thin films [2
]. Such a structure, which is the one used nowadays in dye-sensitized solar cells (DSSCs) as an example, often presents a very high specific surface area and very good porosity, which is crucial to improving the dye absorption in the mentioned application. Nevertheless, they also present a dramatic limit in terms of the relatively low quality of charge carriers transport [3
]. This problem is often associated with the scattering of the charge carriers at the numerous grain boundaries between the fritted particles constituting the film [5
]. This has motivated the development of thin films based on one-dimensional (1D) nanostructures, expected to present a much better facility to transport charge through the film. In reality, such a thin film consists of a tri-dimensional (3D) arrangement constituted of nano-objects grown with a preferential direction from the substrate. This means that two of their dimensions are less than 100 nm while the other can reach several micrometers [6
]. Individually, such a 1D object would avoid the carriers’ transport limitations by creating direct pathways through the material without particle interconnections, while keeping a high specific surface area. As a consequence, the transport path for the charge collection is efficiently shortened and, in most cases, these materials have higher electron diffusion coefficients than nonordered nanostructures [7
], allowing a large increase in the thickness of the film. The price to pay for such an architecture is an often lower specific surface area, as well as porosity, in comparison with the conventional nanoparticles-based thin-film architecture. Indeed, for the latter, the conventional specific surface area and porosity are ~200 m2
/g and >75%, respectively, while the best reported values for 1D nanostructures-based films are ~140 m2
/g and 70%, respectively [8
]. This has motivated many efforts to improve such a structure in the last few years. Basically, researchers have been focused on the improvement in the design of these hierarchical nanostructures, aiming to reach an optimal equilibrium between the specific surface area, porosity, and charge transport efficiency. As a noteworthy example, Kuang et al. established a systematical strategy to grow TiO2
hierarchical nanostructures made of nanowires (NW) on which nanorods (NR) are branched, themselves being sources of nanorods (NR), labeled as NW/NR/NR. These hierarchical structures have been utilized in solar cell applications, allowing for a conversion efficiency of up to 9% due to a larger specific surface area, lower transport time, and longer electron lifetime than TiO2
]. From these results, they concluded that the design of 1D hierarchical nanostructure thin films is one of the keys to find the best agreement between a high carriers’ reservoir capability and good charge transport properties in DSSCs.
Numerous processes have been used to synthesize such nano-sculpted materials: Anodic oxidation [11
], electron beam evaporation [12
], atomic layer deposition [13
], sol–gel deposition in a template [15
], or hydrothermal methods, which has been intensively utilized by successive treatments to obtain branched nanowires [17
]. However, these methods are usually difficult to industrialize, often lead to the synthesis of amorphous materials, and make necessary the use of solvents and toxic chemicals. In this context, it is necessary to develop industrially viable alternative synthesis roads for such structures that would allow a crystallized material to be synthesized with a low environmental impact. Physical vapor deposition (PVD) techniques are well-established in various manufacturing areas such as microelectronics, automotive, and biomedical industries [19
]. In these fields, plasma-assisted processes are generally preferred to thermal evaporation in response to requirements of materials processing at reasonable temperatures. In PVD processes, the target material to be deposited as a thin film is transformed into the vapor phase by different means, generally involving plasma generation (except for thermal evaporation). The chemical composition of the deposited film can be tuned by the addition of various reactive gases (O2
, etc.) in order to form oxides, nitrides, or more complex compounds, which makes the technique versatile.
As a widespread plasma technology, magnetron sputtering, which belongs to the PVD methods and consists of bombarding a target material with accelerated ions from the plasma, leading to the ejection of particles (mainly atoms and clusters), has been used to build the thin film. Oxides, nitrides, or carbides can also be grown by adding pure O2
, or C-based vapor sources inside the magnetron sputtering deposition chamber, the so-called reactive magnetron sputtering (RMS) regime [20
]. The magnetron sputtering process offers the opportunity for tuning the crystalline constitution depending on the energy brought to the growing film by adjusting the experimental parameters such as the applied power [22
]. In most of the aforementioned applications, magnetron-deposited films are meaningful because they are dense, homogeneous, and chemically pure [23
In this work, in order to generate a 1D structure-based thin film, the so-called nano-sculpted films, we have utilized magnetron sputtering in glancing angle geometry. Glancing angle deposition (GLAD) is a particular case of oblique deposition where the substrate position is manipulated during the film deposition [24
]. The technique takes advantage of the ballistic shadowing effect, which allows the formation of columnar microstructures as the film is growing. The basic operation principle is presented in Figure 1
and can be summarized as follows: While the substrate is tilted with an angle α compared to the target normal, the initial nuclei of the depositing film randomly roughen the surface. Subsequently, the depositing particles nucleate on the substrate, while the region behind the nucleus does not receive any vapor, because it falls in the shadow of the nucleus. Consequently, a larger number of particles will be deposited onto the nuclei than in the shadowed area. This inequality increases as growth continues. As only the tops of the nuclei receive the depositing material, the nuclei will develop into columns, tilted in the direction of the incident particles flux and forming an angle β
with the substrate normal (β
). The β
value depends on many experimental parameters as it will be discussed in this paper.
The key principle of GLAD consists of changing the vapor flux direction to operate ballistic shadowing and to provide control over the final thin-film morphology. Two degrees of freedom are obtained by tilting the substrate with respect to the source of particles (α), while rotating the substrate around its normal axis allows the substrate azimuthal angle ф to be controlled with a fixed substrate rotation speed (фs). Varying ф modifies the direction of the incident vapor flux and provides control over the shadowed regions of the substrate.
There are mainly four archetypal columnar microstructures, which illustrate how substrate motion affects the microstructure. Basically, inclined columns are obtained when working with a fixed tilt angle (α
) higher than 60°; zig-zags are generated when rotating the substrate by successive rotation of the substrate by a 180° angle in the latter configuration; and plots or helical structures can be grown continuously, rotating the tilted substrate during deposition [9
]. All of these structures are generated by modifying the substrate rotation as α
is unchanged for each one. The ability to sculpt the film and access the various morphologies is provided by the trajectory of the incident vapor flux relative to the substrate surface during deposition. Accordingly, each nanostructure is characterized by a given porosity (inter-columnar space) that mainly depends on the columnar tilt.
It is important that the incident vapor remains highly directional to avoid merging of the shadowed regions. Indeed, deposition with a poorly collimated incident flux allows vapor to directly access the shadowed area. This implies that the mean free path of the particles should be greater than the distance to the substrate. It is thus evident that PVD techniques such as thermal and electron beam evaporation are the most prevalent in GLAD research because they allow high target-to-substrate distances and low operating pressures with small vapor sources. Therefore, GLAD has mainly been utilized in combination with an evaporation source to grow various nano-sculpted materials [26
]. Nevertheless, using evaporation, the energy brought to the growing film does not allow us to control the crystalline structure of the deposited films in most of the cases. By contrast, as mentioned, magnetron sputtering is a recognizable technique to do this even without intentional heating of the growing material. Indeed, in this case, the crystallization is promoted by the bombardment of energetic particles, as well as infrared radiation emitted from the target during the sputtering process [27
]. Although promising, the utilization of GLAD geometry in combination with magnetron sputtering, i.e. magnetron sputtering in glancing angle geometry (MS-GLAD), is surprisingly quite recent [28
], and has attracted considerable interests for 10 years [29
In this paper, we aim to overview the work that has been developed in our group during the past few years, utilizing this original synthesis process to design nano-sculpted thin films that are ultimately utilized in energy-related applications. Due to the unusual character of the MS-GLAD, we first had to answer many questions related to the growth mechanism of the nano-sculpted films by this approach in order to be able to control the synthesized films features (morphology, chemistry, crystalline constitution, etc.). Indeed, if magnetron-sputtered thin films can be crystallized by a higher supply of energy, the price to pay is the generation of dense films, inhibiting the effect of the grazing mode configuration [30
]. In order to obtain a full picture of the growth mechanism, experimental as well as simulation works have been developed. Finally, we will summarize the benefits that can be associated with the utilization of these nano-sculpted films in energy-related applications, specifically as the photo-anode in dye-sensitized solar cells (DSSCs), although other applications of our films have been investigated in our group [31
2. Materials and Methods
2.1. Experimental Setup
The nano-sculpted films were synthesized in a cylindrical stainless-steel magnetron sputtering chamber (height: 60 cm, diameter: 42 cm), schematically presented in Figure 2
. The chamber was evacuated down to a residual pressure of 10−4
Pa by a turbo molecular pump (Edwards nEXT400D 160W, Edwards, Irvine, California, CA, USA), backed by a dry primary pump (Edwards nXDS10i, Edwards, Irvine, California, CA, USA).
An unbalanced magnetron cathode was installed in front of the substrate, at the top of the chamber on which a 2 in. (5.08 cm) diameter and 0.25 in. (0.635 cm) thick target was connected. The target/substrate distance was fixed at 7 cm. Pure Ti and Mg (both with 99.99% purity) were used as target materials in this study. In order to modify the phase constitution of the deposited films, the target was sputtered either in direct current (DC) mode or in the high power impulse magnetron sputtering (HiPIMS) regime. For the DC mode, an Advanced Energy MDK 1.5 K (Advanced Energy, Denver, Colorado, CO, USA) power supply was used. The power (P) was fixed at 150 W, corresponding to a power density on the target surface of 7.5 W·cm−2
, which was calculated by taking into account the target surface exposed to the plasma (~20 cm2
). In the HiPIMS regime, a lab-made power supply based on a one-quadrant chopper topology was used, allowing the generation of short high-power pulses at the cathode [33
]. We have to mention that the discharge voltage is measured at the output of the lab-made power supply, not directly at the cathode. More information can be found elsewhere [34
Argon, which is the sputtering gas with or without O2 as the reactive gas (both with 99.999% purity), was introduced in the chamber using two distinct mass flow meters in order to grow oxide or metallic compound, respectively. Note that the gases were mixed prior to being injected in the vacuum chamber. All thin films were deposited at constant total gas flux. It was fixed at 15 sccm (standard cubic centimeter per minute) to allow low working pressure (0.13 Pa) according to the pumping rate.
The substrate was installed on a 2-axis manipulator, allowing two rotation motions: Along the α angle to tilt the substrate from α = 0° to α = 90° with regard to the cathode axis and/or along the ф angle to rotate the substrate step by step or in continuous mode with a given angular speed фs. The α and ф angles were varied in order to generate various morphologies: Discrete rotations (ф = +180° or −180°) allow zigzag structures to be grown, while continuous rotations (фs = 0.1, 1.0 or 10°/s) lead to vertical pillars and helicoidal structures.
Silicon single crystals with an (100) orientation and whose resistivity is 5·10−3 Ω.cm, or fluorine-doped tin oxide (FTO)-coated glasses, were utilized as substrates depending on the subsequent type of characterization. The substrates were cleaned with detergent solution, rinsed with ultra-pure water, and placed at the ground potential and at ambient temperature prior to deposition.
2.2. Characterization Techniques
Field emission gun scanning electron microscopy (FEG-SEM Hitachi SU8020, Hitachi, Tokyo, Japan) was used to observe the microstructure of the nanostructured films, while the nanostructure was investigated by transmission electron microscopy (TEM Philips CM200, Philips, Amsterdam, Netherlands). The cross-sectional lamellae of the untreated nanostructured films were prepared by mechanical polishing and ion milling. Individual columns of the single crystalline thin film were scratched to observe each column separately.
Grazing incidence X-ray diffraction (GIXRD) analysis (Panalytical Empyrean, Malvern Panalytical, Malvern, UK) was used to determine the phase constitution of the samples. The Cu Kα1 source (1.5406 Å) was used and the X-ray source voltage was fixed at 45 kV and a current at 40 mA.
The experimental procedure used to design the DSSCs is described in depth elsewhere [35
]. Briefly, the dye grafting of TiO2
-based nano-sculpted electrodes (0.25 cm2
) was performed by immersion overnight in a solution of acetonitrile and tertbutyl alcohol (volume ratio: 1/1) containing dye sensitizer (0.3 mmol) and (3R),(7R)-dihydroxy-5-cholic acid (Sigma-Aldrich, Saint-Louis, Missouri, MO, USA) (2 mmol) to avoid aggregation of the dye. Then, the sensitized electrode was assembled with a platinized FTO electrode, both separated by 25 μm-thick Surlyn®
(Dow Chemical, Midland, Michigan, MI, USA) to prevent the electrolyte from leaking. The internal space was filled with a liquid electrolyte by using a vacuum backfilling system. The photovoltaic performances of the cells were then measured under a simulated air mass (AM 1.5) Global spectrum and 1000 W/m2
2.3. Simulation Protocol
The simulation of nano-sculpted film growth is possible with kinetic Monte Carlo (kMC) algorithms [36
]. This approach is useful for the modeling of various surface processes such as the nucleation, growth, structural modifications, or dynamic evolution of obliquely deposited structures [38
]. These features are implemented into the freely distributed NASCAM (nanoscale modeling) code (version 4.6.2 rev. 6; University of Namur: Namur, 2018) [40
], which is particularly suitable to simulate GLAD processes as it takes into account the motion of the substrate during deposition (translation, rotation, and oscillation).
In this code, the incoming vapor flux is represented by hard spheres and their mobility is simulated according to the ballistic deposition approximation for minimizing the computation time. Taking into account the energy and the angular distribution of the vapor source, those atoms travel toward the substrate along linear trajectories. Then, the deposited particles become part of the growing film.
As input, the code uses the kinetic energy and angular distribution of the sputtered atoms calculated by SRIM (stopping and range of ions in matter, SRIM-2013; Chester, Maryland, MD, 21619, USA) [42
] and SIMTRA (simulation of the metal transport, version 2.2; University of Ghent: Ghent, 2018) codes, respectively. Indeed, the energy and the direction of the particles that are sputtered from the target material are first calculated by SRIM. Then, SIMTRA simulates the transport of these species toward the substrate, taking into account all collisions happening in the gas phase. At the end, the PoreSTAT plugin can be used to evaluate the porosity of the simulated films from the NASCAM output files [43
]. This simulation strategy is presented in Figure 3
The mobility of the atoms that reach the substrate is severely dependent on the energy of incoming deposited atoms. However, the morphology of the films depends on the different deposition parameters [44
]. Consequently, to simulate the growth of thin films synthesized at high temperature or with high-kinetic-energy incoming atoms, the approximation of ballistic deposition has to be completed by the diffusion phenomenon [22
]. Diffusion and evaporation events can take place between two atom depositions at an equal time interval determined by the deposition rate. For each step of the simulation, a list of atoms that can diffuse at the surface or evaporate is created for each possible physical event. The evolution of the system is, thus, determined by the probabilities of the events that may occur during the simulation. This probability can be implemented into NASCAM code via their activation energies (Ea
), which can either be found in the literature or calculated by molecular dynamics or potential models.
For each working condition, the energy and the angular distribution of the species can be adapted by the introduction of the experimental parameters such as the working pressure, the power applied to the target, the racetrack size, and the target-to-substrate distance. In order to compare simulated and experimental thin films having the same thickness, the number of deposited atoms (N) and the substrate size (XYZ) can be tuned in the NASCAM input file. X and Y correspond to the length and width of the substrate, respectively, while Z accounts for the height of the deposited film in atom units.
2D NASCAM simulations were performed for direct comparison with the cross-sectional film morphology, while 3D simulations were performed for the porosity evaluation. The Ti and Mg deposition rates were fixed at 0.5 monolayers per second (0.16 and 0.30 nm/s, respectively), which is of the same order of magnitude in comparison with their experimental values (0.17 and 0.32 nm/s, respectively).
This work summarizes our recent research related to the development of nano-sculpted thin films by magnetron-sputtering-related technologies and to their use in energy-related applications. We first describe our understanding of the growth mechanism associated with the novel utilization of the glancing-angle geometry in magnetron sputtering processes, the MS-GLAD process. The synthesis of model nano-sculpted Ti and Mg films was investigated by using a joint experimental–modeling approach based on kMC simulations implemented in the NASCAM code.
Based on the different morphological properties of Ti and Mg coatings grown for similar experimental conditions, it appears that the homologous temperature as defined in structural zone models, Ts/Tm, is one of the key parameters to finely control the growth of nano-sculpted coatings by MS-GLAD. When comparing the two considered metals, this parameter is different enough to allow for different growth regimes, from Zone 1 for Ti to Zone 2 for Mg. Basically, this parameter mainly defines the importance of the adatom diffusion processes that is, at room temperature, negligible in the Ti case (the ballistic deposition approximation is sufficient), while events such as hops up and hops down from one atomic layer to another should be taken into account to accurately describe the growth of Mg nano-sculpted thin films. This rationalization based on the homologous temperature of the deposited material can even been extended when considering a different chemistry of the system, i.e., an oxidation of the deposited material. Indeed, in such a situation, it is shown that, as expected by the values of the homologous temperature, Ti and TiO2 behave almost similarly while a strong impact is observed for Mg.
The other key parameters are related to the collimated nature of the depositing flux and on its impact on the shadowing effect, which is the basic effect when GLAD geometry is considered. Therefore, it has been demonstrated that the deposition pressure, which strongly affects the collimated character of the depositing flux through the mean free path of the particles, has to be low enough (<0.26 Pa) to trigger the formation of the different nano-sculpted structures.
In comparison with the conventional combination of GLAD with evaporation, it is shown that the utilization of magnetron sputtering in given conditions allows for a good crystallization of the deposition material, which is important in many applications. In particular, for TiO2 thin films, we demonstrate that anatase monocrystalline-like nanocolumns-based thin films can be synthesized.
From an application point of view, nano-sculpted TiO2 coatings were integrated into the photo-anode of dye-sensitized solar cells (DSSCs). First, it appears that the devices based on nano-sculpted thin films outperform nanoparticles-based DSSCs both in terms of charge harvesting and charge recombination. However, the photo-anode thickness drastically affects the cell performances, indicating that the critical parameter is the adsorbed dye density. This problem has been addressed by combing the nano-sculpted TiO2 films with a spin-coated TiO2 nanoparticles solution. This hybrid system demonstrates a synergetic effect between the columnar thin film and the absorbed nanoparticles, which significantly improved the efficiency of the DSSCs by simultaneously enhancing the charge transport and the quantity of adsorbed dye molecules.
We believe that the development of novel magnetron sputtering approaches to design materials presenting a well-defined morphology at the nanoscale consists of an important opportunity for this well-established technology.