A NanoSys 500 deposition system, built by Mantis Deposition Ltd (Thame, Oxon, UK), was used to produce size-selected TiN nanoclusters 19 . Figure 1 shows a schematic diagram of the experimental setup. The fabrication of noble metal nanoclusters using Nanosys 500 has been described elsewhere by other authors 18 . The difference in producing TiN clusters is the use of a reactive gas (N₂) aggregation process.

Inside the cluster generation zone, the TiN clusters were produced by sputtering a Ti target in an inert gas atmosphere of Ar⁺ and by adding directly over the sputter target a small flux of nitrogen gas (20 sccm). When the Ti target is sputtered, the clusters are formed by multiple collisions between the high-density particles (sputtered particles and also argon and nitrogen gasses). During cluster formation, the sputter chamber is kept at a low temperature by a coolant mixture (2°C), and the pressure was set at 1 × 10^-4 Torr. The residence time within the aggregation zone can be diversified by varying the length of the aggregation region with the linear motion drive. The cluster size can be controlled by varying the principal parameters: the sputter power, flow of gasses, and the aggregate zone length (variable using a linear drive). The clusters flow together with the argon gas through a variable orifice towards a mass filter. The mass distribution is monitored in situ by a MesoQ mass filter (Mantis Deposition LTD, Thame, Oxon, UK) before the deposition. This mass filter has been specifically added for the purpose of high-resolution measurement and filtering of nanoclusters between 1 × 10³ and 3 × 10⁷ amu. A typical mass distribution spectrum of the obtained TiN⁺ clusters is presented in Figure 2. The mean cluster diameter was estimated from the cluster masses by using the specific density and the molar volume of the TiN.

Once the size clusters are selected by the MesoQ mass filter, they are accelerated by applying a bias voltage (V _b) to a substrate in a high vacuum with a base pressure of 10^-8 Torr. The nanoparticles were deposited onto silicon wafer. The substrates were cleaned in successive ultrasonic baths of acetone and isopropyl alcohol. The depositions were performed at room temperature without any heating and were applied different bias voltages (3 and 6 kV). The nanoparticle size was controlled by regulating the magnetron power, gas flow (Ar and N₂), and aggregation zone length. These parameters were varied to produce particles of different sizes onto the substrate.

The size distribution and morphological characterization were performed by AFM analysis, using a (Veeco Instruments Inc., Plainview, NY, USA) multimode scanning probe microscope in hard tapping mode. The sample surface was scanned at 1 Hz and within 1 μm². The Image Processing and Data Analysis software (Version 2.1.15) by TM Microscope (Camarillo, CA, USA) was used for image analysis.

For TEM characterizations, TiN nanoparticles were grown on copper grids. Samples were characterized by HRTEM using a 300-kV FEI Titan 80-300 STEM/TEM microscope (FEI Co., Hillsboro, OR, USA). The HRTEM images were used to study the TiN crystalline structure by micrograph FFT analysis. A detailed statistical analysis of TiN nanoparticles after deposition was performed by measuring several hundreds of nanoparticles using the HRTEM micrographs. The size distribution, the nearest neighbor distance (d _NN), and the covered area on the surface were extracted from these calculations. The procedure was carried out by manually outlining the particles from several dozens of low- and high-resolution TEM images. Once digitized and saved in the proper format, the image was processed using the Gatan Digital Micrograph and Mathematical software (Gatan, Inc., Pleasanton, CA, USA).

PL studies were carried out at room temperature in a conventional PL system. An He-Cd laser (λ = 325 nm at 16 mW) was employed as the excitation source. The outgoing radiation from the sample was focused on the entrance slit of a 50-cm Acton monochromator (Princeton Instruments, Trenton, NJ, USA). The detection was carried out using a Princeton Instrument photomultiplier tube to a photon counter. All the spectra were corrected for the spectral response of the system.

Figure 3 displays the typical size distribution recorded by the mass filter for three samples containing nanoparticles with different diameters. The mean nanocluster diameter was obtained after fitting a Gaussian function to the experimental data (Figure 3, solid lines). The figure displays the nanoclusters having a narrow size distribution of 4.1, 5.3, and 7.1 nm with a full width at half maximum of 1.2, 1.4, and 2.1 nm, respectively. Moreover, a good separation of peaks is clearly observed demonstrating that the mass filter can resolve the nanoparticle diameter with a difference at approximately 1 nm. Hence, the filtering process allows the selection of particles with a high resolution in size higher than 1 nm, which could be used in either scientific research or technological development. Varying the critical parameters on the system (gas flow, partial gas pressure, magnetron power, aggregation zone length), we were able to produce small nanoparticles at different sizes: 4.1 ± 0.2, 5.3 ± 0.4, 6.2 ± 0.2, 7.1 ± 0.3, 14.9 ± 0.7, and 20.3 ± 0.8 nm.

After the deposition of TiN nanoparticles on Si (111) wafer, the morphology and size distribution were analyzed using the AFM images. Figure 4 shows a typical AFM image of TiN nanoparticles. From the figure, it is possible to observe the aggregation of nanoparticles, which could be explained due to clusters landing on top of each other before the layer is completed. For longer deposition times, it is expected that even more clusters get aggregated. The magnification in Figure 4a, b was used to measure the height profiles of four nanoparticles (Figure 4c). The height shows a roughly Gaussian profile, which allows to consider the nanoparticle as quasi-spherical.

AFM height histograms for TiN nanoparticles at different filtered sizes are shown in Figure 5. These histograms show narrow height distributions with averages of 15.5 nm (Figure 5a) and 23.0 nm (Figure 5b). Usually in an AFM measurement, the distance in the z-axis is directly related to the nanoparticle size. However, in the xy plane, the distance is enlarged regarding the real spatial dimensions because the width of the AFM tip is distorted by the combination of the nanoparticle shape and tip geometry. In this case, the widths of height histograms do not refer to the real sizes of nanoparticles. Therefore, the results of size distribution for both mass filter and AFM height profiles support the effective mass filter to control the TiN nanoparticle size.

The nanoparticle size was directly measured by HRTEM micrograph. Figure 6 shows TiN nanoparticles with sizes of 7.1 (Figure 6a), 6.2 (Figure 6b), and 4.1 nm (Figure 6c), where the two-dimensional atom arrangement and circular shape can be observed. The nanoparticle size was directly measured from the HRTEM micrographs using suitable tools of the image processing software. The crystalline structure of the TiN nanoparticles was further confirmed by FFT analysis (insert in the figures). The FFT shows a diffraction pattern corresponding to planes of the TiN face-centered cubic [FCC] structure with a lattice parameter of 4.2417 Ǻ 20 : (111) [d ₁₁₁ = 2.529 Å], (200) [d ₂₀₀ = 2.134 Ǻ], and (220) [d ₂₂₀ = 1.561 Å]. According to the AFM height profiles and plan-view TEM image, we can strongly consider the shape of nanoparticles as quasi-spherical.

The size distribution, cover surface, and nearest neighbor distance were also statistically analyzed as a function of the bias voltage. TiN nanoparticles produced at 3 and 6 kV bias voltage on TEM grids are shown in Figures 7a and 8a, respectively. The covered surface by the nanoparticles at 3 kV is around 8.2%, while at 6 kV, they cover only 2.1% of the surface.

Also, the figures display the corresponding size distributions (Figures 7b and 8b) and the nearest neighbor distances (Figures 7c and 8c). In Figure 7a, b, it can be seen that the nanoparticles have an average diameter of 5.7 nm and a distance between particles of 6.7 nm. When the bias voltage is increased at 6 kV, the average diameter and the d _NN are enlarged to 6.7 nm and 22.9 nm, respectively. The increase of the nanoparticle size and the nearest neighbor distance when the bias voltage is raised can be explained by the coalescence of two or more nanoparticles, as shown in Figure 9. The bias polarization increases the cluster kinetic energy towards the substrate and enhances the nanoparticles' mobility on the surface, allowing that one nanoparticle could reach other nanoparticles.

Figure 10 displays the spectral PL of TiN nanoparticles at different sizes: 20, 15.5, and 5.4 nm. In the figure, the increase of the PL intensity when the diameter decreases from 20 to 15.5 nm can be observed. However, when the diameter decreases down to 5.4 nm, a strong visible PL is clearly seen. The origin of this strong PL is only due to the nanocluster size confinement because in transition metals, the light emission is normally very weak due to ultrafast nonradiative decay and the absence of a bandgap 21 . The PL has been previously reported in another transition metal nanoparticle with a size below 10 nm 22 . However, this is the first report in TiN nanoparticles.

It can also be observed in Figure 10 that the strong PL is deconvoluted by three Gaussians, which correspond to the emission at different wavelengths. According to the bulk band structure studies, these emissions could be related to the transition between L1 and Gamma1. This is because the emission energies are close to the difference of energy band (2.3 ± 0.3 eV) 23 . In this case, we considered that the observation of three emissions is due to the size distribution on the surface (observed in AFM and HRTEM). A first attempt to explain this phenomenon is shown in Figure 11.

A fitting equation (Figure 11, inset) and the HRTEM diameter distribution of TiN nanoparticles at 3 kV were used to describe the PL intensity. It can be noticed that energy distribution, as a function of nanoparticle size, is in an agreement with PL peak position. The slight difference (2.5 eV) of both the theoretical energy band and the PL peaks are due to the quantum confinement effect, where the energy band is enlarged when the nanoparticle size becomes smaller (for spherical particles) 24 25 . Thus, in TiN, the PL peaks at a higher energy displacement with the nanoparticle size. For larger sizes, the energy displacement is less, (peaks or shoulders at approximately 2.6 eV) while for smaller sizes, the energy displacement is greater (peaks at approximately 3.0 eV).