3.1.

XPS Characterization

XPS characterization was performed on each sample using a Staib XPS system consisting of an ion pump, x-ray source, and energy analyzer. Samples were loaded into an antechamber with a base pressure of $6.7\times {10}^{-6}\mathrm{Pa}$ or better. The samples were then transferred into the analysis chamber which had a base pressure of $4.0\times {10}^{-8}\mathrm{Pa}$. XPS data were collected using a pass energy of 20 eV. All spectra were background corrected using a Shirley background subtraction. Component curves were fitted to each spectrum using a Levenberg−Marquardt algorithm that minimizes ${\chi }^2$, Fig. 2. Each component curve consists of a convolution of Gaussian and Lorenztian functions. Photoemission intensities were corrected to account for Scofield cross sections, kinetic energy, and analyzer transmission function. XPS data were collected with a standard Mg $K\alpha$ (1253.6 eV) excitation source. The energy scale was calibrated against Au $4f_{7/2}$ and Cu $2p_{3/2}$ at the binding energies of 84.0 and 932.6 eV, respectively. The XPS resolution was determined by the FWHM of the Au $4f_{7/2}$ line, which was found to be 1.2 eV.

Fig. 2

X-ray photoelectron spectroscopy (XPS) showing boron opposite of carbon for desired BN:C ratio. Both sets of stacked spectra consist of raw data (black), fitted components (blue), and sum curves (red).

Table 2 shows the binding energy ($E_B$) for boron and carbon (B $1s$ and C $1s$). The normalized concentration of each element is equal to the measured intensity divided by the instrument sensitivity factor for that element. The resulting percentage of the normalized concentration for each element is then the atomic percent. Table 2 shows the surface characterization measurements for the 50% C-50% BN sample. Oxygen contamination, O $1s$ peak, is shown as being some 28%; this is expected due to the fact that the samples have been exposed to air and were exposed to an ion beam for a brief time before sampling. The same measurement and analysis techniques were used on each of the four samples. Table 2 shows that the ratio of B $1s$ to C $1s$ is approximately slightly $<1.0$ and the ratio of B $1s$ to N $1s$ is just under 1.7.

Table 2

X-ray photoelectron spectroscopy surface characterization of 50% carbon–50% BN sample. Measurement error is ±0.1 atomic %.

Figure 2 shows plots of the binding energies for boron and carbon 1s peaks. The spectra show cross-contamination that occurred during growth, and are obvious for the first growth of 100% BN, i.e., a laser pulse of carbon would place small amounts of carbon onto the BN target, and pulse onto the BN target would contaminate the carbon target. This issue has been largely resolved with the addition of a shutter over the sample as well as a shield between the two PLD targets which provides for the ability to operate the ellipsometer even while the shutter is closed with PLD operating for the purposes of precleaning the targets, and has been shown to be required.

Figure 2 shows the composition summary of B, C, and N for each of the four films deposited and the corresponding XPS composition. The nitrogen content is lower than boron content for each sample, which can be explained by the fact that ionization and atomization of the BN target during the PLD process in vacuum results in a loss of nitrogen. Since the pressure was constant, HV, it was also expected that the same boron to nitrogen ratio will essentially be maintained regardless of BN:C ratio of growth, as shown in Fig. 3, and the $\mathrm{B}:\mathrm{N}$ ratio is $\sim 1.5$ for XPS characterized samples.

Fig. 3

XPS and EDX summary of $\mathrm{B}:\mathrm{N}$ ratio (atomic %) and $\mathrm{B}:\mathrm{C}$ ratio for each setpoint of carbon target pulses (%).

The desired growth setpoints that were also characterized by XPS were BN: 100%, BN: 80%, BN: 67%, and BN:50%, but as is shown in Fig. 3, the actual boron to carbon ratios ended up being 4.33, 2.96, 1.44, and 0.901, respectively. The ratio of boron to nitrogen ended up being 1.472, 1.493, 1.480, and 1.678, respectively. This represents ratios on the order of $\sim {\mathrm{B}}_4{\mathrm{CN}}_x$, $\sim {\mathrm{B}}_3{\mathrm{CN}}_x$, as well as two samples with a $\sim {\mathrm{BCN}}_x$ ratio near unity, i.e., one $\mathrm{B}:\mathrm{C}$ sample rich in boron and the other rich in carbon.

As was previously mentioned, the four samples grown without mechanical shielding between targets and characterized by XPS show approximately a $4:1$ ratio of boron to carbon for the first sample which should not have had any significant amount of carbon. This error can be from either XPS measurement technique due to air exposure, or due to contamination occurring during the deposition process. Contamination can occur by the fact that if the targets are not physically separated using mechanical shielding, then each laser pulse will have material both deposited on the substrate, but also onto the other laser target and thereby skewing the desired deposition parameters and potentially creating particulates. To test this, six additional samples were grown to approximately 1 μm in thickness with mechanical shielding added between PLD targets to prevent cross-contamination. Energy dispersive x-ray characterization was performed with a Hitachi 3000 tabletop SEM with energy-dispersive X-ray spectroscopy (EDX) attachment. As shown in Fig. 3, the results were virtually identical within the error in terms of composition. However, the first XPS sample shown with a $\mathrm{C}:\mathrm{B}$ ratio of $\sim 0.25$ demonstrates the first sample was contaminated as the EDX sample shows a carbon content of zero as expected, Fig. 3. PLD results in a direct stoichiometric transfer of material from target to substrate, but in vacuum also in all directions, hence the need to have a shield between targets to prevent cross-contamination. Adding a mechanical shield prevented cross-contamination issues.

When depositing from the BN target, there is a loss of nitrogen (Fig. 3, $\mathrm{N}:\mathrm{B}$ ratio below 1) due to the fact that the deposition is being performed in vacuum, which is typical for multielement targets containing nitrogen or oxygen. One of many techniques to raise the resulting thin film nitrogen content is to perform depositions at a pressure of a few Pa of nitrogen. Figure 3 demonstrates that the composition, $\mathrm{C}:\mathrm{B}$ ratio, can be adjusted with precision using PLD, but optimization process with regard to ${\mathrm{BC}}_x{\mathrm{N}}_y$ would need to be performed to obtain the desired nitrogen content as well. For single element targets, it is clear that subnanometer mixing can be accomplished using sequential laser pulses on each target.

3.2.

Raman Characterization

Figure 4 shows four Raman spectra corresponding to each of the four samples and a carbon sample for the reference. A Renishaw inVia Raman microscope having a 514 nm laser excitation and $50\times$ microscope objective was used for the measurements in a confocal mode providing a measurement depth of approximately 1 μm. The peaks shown in each of the spectra below $1000{\mathrm{cm}}^{-1}$ are from the Si substrate and corresponding native oxide layer of ${\mathrm{SiO}}_2$ ($\sim 2.25\text{-}\mathrm{nm}$ thick). The carbon sample was deposited with the same conditions as the other four, but for a much shorter period of time yielding a film thickness of $\sim 30\mathrm{nm}$, and as such the Si and ${\mathrm{SiO}}_2$ peaks are much larger than those observed in the other samples.

Fig. 4

Raman spectra of each of the five samples as deposited.

Raman spectra of carbon are expected to have a D peak at $1360{\mathrm{cm}}^{-1}$ and a G peak at $1580{\mathrm{cm}}^{-1}$, and boron nitride is expected to have Raman peaks at $1055{\mathrm{cm}}^{-1}$ for cubic BN (c-BN) and $1364{\mathrm{cm}}^{-1}$ for hexagonal BN (h-BN).¹² As is shown in Fig. 4, the spectra above $1000{\mathrm{cm}}^{-1}$ are as could be expected, essentially a D peak component, a much larger G peak component, and second harmonic of the D peak. By peak fitting the D and G components using two Gaussian functions, for pure carbon, the algorithm returns that there is dominantly graphitic carbon having a G peak centered at $1557{\mathrm{cm}}^{-1}$ and a much smaller D peak centered at $1401{\mathrm{cm}}^{-1}$, and for the pure BN sample, there is only h-BN with a broad peak centered around $1300{\mathrm{cm}}^{-1}$. The D peak and h-BN peak overlap and the peak position of their summation shifts lower in wavenumber with the increasing BN content. Since the depositions were performed without a heated substrate c-BN was not expected to be formed, which is confirmed by the absence of a $1055{\mathrm{cm}}^{-1}$ peak. The spectra in Fig. 4 for 80% BN-20% C deposited by PLD are equivalent to the Raman spectra presented in Ref. 13 from sputtering of ${\mathrm{B}}_4\mathrm{C}$ in 10% ${\mathrm{N}}_2$ to obtain a film composition of 48% B, 12% C, and 40% N.¹³

Beginning with the pure carbon sample Raman spectra, increasing the amount of BN causes the G peak to be reduced relative to the h-BN peak as the boron content is increased, and hence the h-BN peak of $1364{\mathrm{cm}}^{-1}$ is increased in intensity. The broad peak around $2700{\mathrm{cm}}^{-1}$ is the second harmonic of the $\sim 1360{\mathrm{cm}}^{-1}$ peak. The second harmonic of bulk h-BN (h-BN′′) is represented by a broad peak at $2765{\mathrm{cm}}^{-1}$, while the BN thin film second harmonic is $\sim 2500{\mathrm{cm}}^{-1}$, Fig. 4.¹⁴

One noticeable detail is that with each of the four films having approximately the same thickness, the pure BN film has a Raman intensity at $\sim 1300{\mathrm{cm}}^{-1}$, which is much smaller than the substrate related to peak intensity at $\sim 970{\mathrm{cm}}^{-1}$. This lack of Raman response for pure BN would imply that either the film is not nearly as thick or that BN is perhaps much more transparent to 514 nm laser excitation than carbon. The film thickness was verified with ellipsometer modeling, AFM, and a KLA Tencor profilometer, so the lack of response must be from the actual material properties such as the absorption coefficient at 514 nm.

3.3.

AFM Characterization

An MFP3D AFM (Asylum Research) was used to measure film thickness and surface root-mean-squared (RMS) roughness for each sample composition. The data shown in Fig. 5 were acquired in ac (tapping) mode by an Olympus ac160 silicon probe with a nominal tip radius of $9\pm 2\mathrm{nm}$. Sample thickness was measured relative to the silicon substrate (disregarding an uneven edge formed by the removal of a mask adjacent to the deposited film). A measured thickness of 103.7 nm is indicated by the cross section in Fig. 5(a) for a BN-50:C-50 scan. An image analysis function in the MFP3D (Igor Pro-based) software was used to calculate the RMS roughness of unmodified topography data according to $\sqrt{(1/n)\sum z_i^2}$, where $n$ is the total number of $z$ height values. Figure 5(b) shows a representative 10 μm scan of BN-80: C-20, which yielded a full image RMS value of 7.87 nm. RMS values were on the order of 7 to 11 nm for randomly sampled 10 μm areas. The horizontal lines in Fig. 5(b) were artifacts from the instrument of the tallest features in the scan line. The particulates captured in Fig. 5(b), displayed as white spots, were observed for each sample composition and could be attributed to the deposition of material onto the opposing PLD target with the ablation pulse. Each focused laser pulse was approximately $1.2\mathrm{mm}\times 0.4\mathrm{mm}$, with each target material having some 45-mm diameter active area of ablation, allowing buildup of particulates from many pulses onto each opposite target before an actual deposition pulse. In this manner, some degree of particulates are expected, but could be mitigated with the addition of a shield between the laser targets.¹⁵

Fig. 5

Atomic force microscopy (AFM): (a) height profile across the edge of a BN-50:C-50 sample indicates approximate 100 nm film thickness relative to the silicon substrate. The uneven edge morphology was formed by the removal of a mask after deposition. (b) Representative 10 μm BN-80:C-20 surface scan yields 7.9 nm RMS roughness. RMS values were $\sim 7$ to 11 nm over randomly sampled 10 μm areas, and the surface morphology is typical for the four compositions evaluated. Particulate heights range between $<10$ and $>100\mathrm{nm}$; horizontal lines across the tallest features are AFM scanning artifacts.

3.4.

In Situ Ellipsometer Characterization

In situ ellipsometer data were collected for each of the four depositions, with the 50% BN–50% C and the 80% BN–20% C deposition dynamic data being shown in Fig. 6. Each deposition lasted approximately 15 min, with the deposition conditions being $300\mathrm{mJ}/\text{laser}$ pulse, and a 16 Hz pulse rate from the laser. During the deposition, the growth rate of the thin film thickness was linear as shown in Figs. 6(c) and 6(d). The value of $\mathrm{\Psi }$ changed significantly during growth with each associated wavelength. The effect of different composition of materials and the difference in optical property of the material was significant, with the 80% BN–20% C having the largest change in the magnitude of $\mathrm{\Psi }$. The change in $\mathrm{\Psi }$ is most pronounced at 371.3-nm wavelength, Figs. 6(a) and 6(b).

Fig. 6

Dynamic ellipsometer spectra collected during deposition: (a) BN-50:C-50 and (b) BN-80:C-20. In situ thickness modeled using ellipsometry (c) for BN-50:C-50 and (d) for BN-80:C-20.

The ellipsometer measures psi and del across the wavelength range of 370 to 1690 nm. CompleteEase software from J.A. Woollam provides for definition of optical layers, and in this case, a modeled layer for the Si wafer substrate, the ${\mathrm{SiO}}_2$ native oxide layer, and a Cauchy model for each thin film of interest. Each optical layer is parameterized by the thickness, $n$ and $k$, and an appropriate model such as Tauc-Lorentz, Sellmeier, or in the simplest form Cauchy for transparent materials. The CompleEase software iteratively fits all measurements to all of the defined model layers to determine optimal model parameters. The Cauchy model, Eqs. (1) and (2), is an approximate function of the Sellmeier model in which a series expansion is used to compute $n(\lambda )$ and $k(\lambda )$:

In Eq. (1), the constant $A$ represents the approximate index of refraction and together with $B$ and $C$ define the index dispersion. Band edge is a parameter in nanometers that is set manually, whereas the $k1$ and $k2$ paramters determine the shape of the extinction coefficient dispersion. Once the best fit of the parameters $A$, $B$, $C$, $k1$, and $k2$ are obtained by CompleteEASE, where $n$ and $k$ and thickness are calculated.

From a Cauchy model fit of the psi and del in situ data of two deposition conditions, $n$ and $k$ were calculated at five representative wavelengths for the 50% BN–50% C and 80% BN–20% C cases, Fig. 7. The 50% BN case shows that particular film is very absorptive, especially at 514 nm where strong Raman spectra were collected, and the 80% BN case was not nearly as absorptive, Fig. 7. The extinction coefficient for pure BN, Fig. 7, was closest to zero at 514 nm as compared to the others, and was correspondingly the weakest Raman scattering thin film at $1364{\mathrm{cm}}^{-1}$, Fig. 4.

Fig. 7

$n$ (a) and $k$ (b) with the wavelength for pure BN and samples having tuned stoichiometry $\sim 100\mathrm{nm}$ in thickness and C $\sim 30\text{-}\mathrm{nm}$ thick.

Additionally, all samples $n$ and $k$ values are shown in Fig. 7: pure BN, 80% BN–20% C, 67% BN–33% C, 50% BN–50% C, and pure carbon. BN has the lowest value of index of refraction of the films grown at 2.0, and carbon the highest around 2.4.¹⁵ As demonstrated in Fig. 7, changing the stoichiometry by adding amounts of carbon will cause a shift in the actual optical properties of index of refraction and absorption coefficient, which can be desired for filters, optical lenses, coatings, metamaterials, or other applications. Increasing carbon concentrations, Table 3, show that when compared at a specific wavelength such as 632 nm, the $n$ and $k$ values increase in an increasing fashion. As the carbon concentration is linearly increased from 0 to some 41 atomic %, the index of refraction increases from 2.03 to 2.36 at 632 nm and the value of $k$ increases from $\sim 0.006$ to $\sim 0.304$. Typical values of index of refraction for h-BN are 1.8 and 3.3 for ${\mathrm{B}}_4\mathrm{C}$.

Table 3

Changes of n and k according to carbon concentration at 632 nm.