Depth Profiling of Ion-Implanted 4H–SiC Using Confocal Raman Spectroscopy

Abstract

For silicon carbide (SiC) processed by ion-implantation, dedicated test structure fabrication or destructive sample processing on test wafers are usually required to obtain depth profiles of electrical characteristics such as carrier concentration. In this study, a rapid and non-destructive approach for depth profiling is presented that uses confocal Raman microscopy. As an example, a 4H–SiC substrate with an epitaxial layer of several micrometers thick and top layer in nanoscale that was modified by ion-implantation was characterized. From the Raman depth profiling, longitudinal optical (LO) mode from the epitaxial layer and longitudinal optical phonon-plasmon coupled (LOPC) mode from the substrate layer can be sensitively distinguished at the interface. The position profile of the LOPC peak intensity in the depth direction was found to be effective in estimating the thickness of the epitaxial layer. For three kinds of epitaxial layer with thicknesses of 5.3 μm, 6 μm, and 7.5 μm, the average deviations of the Raman depth analysis were −1.7 μm, −1.2 μm, and −1.4 μm, respectively. Moreover, when moving the focal plane from the heavily doped sample (~10¹⁸ cm⁻³) to the epitaxial layer (~10¹⁶ cm⁻³), the LOPC peak showed a blue shift. The twice travel of the photon (excitation and collection) through the ion-implanted layer with doping concentrations higher than 1 × 10¹⁸ cm⁻³ led to a difference in the LOPC peak position for samples with the same epitaxial layer and substrate layer. Furthermore, the influences of the setup in terms of pinhole size and numerical aperture of objective lens on the depth profiling results were studied. Different from other research on Raman depth profiling, the 50× long working distance objective lens (50L× lens) was found more suitable than the 100× lens for the depth analysis 4H–SiC with a multi-layer structure.

1. Introduction

As a third generation semiconductor material, silicon carbide has many outstanding characteristics such as a wide forbidden bandgap, high thermal conductivity, high mobility, and high electrical breakdown strength [1]. Ion-implantation is generally accepted to be the only feasible method for selective doping in light of the low diffusion coefficient of impurities in silicon carbide [2]. Ion-implantation was followed by a high temperature annealing treatment to achieve the electrical activation of dopant impurities and recovery of defects. It is worth noting that the implanted regions typically have thicknesses in the sub-micrometer range. To evaluate the electrical properties of the material obtained after ion-implantation and annealing treatment, where the carrier concentration and carrier mobility are of particular interest, and many methods are being used to measure the electrical properties of silicon carbide such as the Hall effect [3], point contact current voltage technique (PCIV) [4], four-point probe [5], and mercury-probe capacitance voltage [6]. Currently, most characterization methods for the electrical properties above-mentioned need the preparation of special measurement structures or will destroy the sample in the process of obtaining the depth distribution of electrical parameters, while the operation procedures are laborious and time consuming: (a) the point contact current voltage (PCIV) [7] has several drawbacks including rather low sensitivity and the lack of required reference samples; (b) the electrochemical etching based depth profiling [8] was investigated but also exhibited insufficient accuracy and reliability. For doping profiles with lower maximum doping concentrations (<~5 × 10¹⁷ cm⁻³ or even lower), the standard in industrial process control is based on well-established mercury capacitance voltage measurements, whereas basically similar but contactless methods seem to have arisen [9]. In addition, other techniques have been investigated and developed such as scanning spreading resistance microscopy (SSRM) or scanning capacitance microscopy (SCM), which, however, are time consuming and need elaborate sample preparation. For SCM, higher doping concentrations might be problematic as indicated in earlier works for 6H–SiC and also need appropriate calibration samples [10]. The most relevant method for rather thin implanted 4H–SiC layers with high doping concentrations is secondary ion mass spectroscopy (SIMS) depth profiling. However, this gives only the “chemical” depth profile of the implanted species [11]. Unfortunately, though, especially in 4H–SiC (compared to e.g., Si), the electrically active doping concentration and, thus, the carrier concentration profile might differ significantly from the chemically implanted concentrations. This is much more pronounced for Al-implantation than for N-implantation. All according methods have not asserted themselves yet due to the above given reasons. Therefore, it is very important to find alternative and non-destructive methods to evaluate the electrical properties profiling of ion-implanted silicon carbide.

Raman spectroscopy is a contactless optical detection method with good sensitivity and high detection efficiency. The confocal configuration formed by placing a micrometer sized pinhole in the back focal plane can greatly improve the spatial resolution of the obtained Raman spectrum and realize optical sectioning [12]. So far, only a small number of confocal depth analysis results have been published, but most of them have focused on the damage distribution after high energy (MeV range) implantation, which is not the typical case for silicon carbide doping in the SiC electrical device industry. Confocal Raman set-up was used to analyze the depth distribution of damage after high energy ion-implantation on the order of MeV before and after annealing treatment [13] and the information of the absorption coefficient could be extracted from a high dose Al-implanted layer in 6H–SiC [2,14]. Zuk destructively pretreated the sample and obtained the damage distribution of high-energy ion-implanted 6H–SiC by confocal Raman spectroscopy on the processed slope, which requires rigorous and complex sample pretreatment and the sample cannot be re-used or further processed after the test [15]. De Biasio demonstrated the use of Raman spectroscopy for the determination of carrier concentrations in epitaxial layers and substrates in n-type 4H–SiC obtained by doping during epitaxial growth without a detailed explanation of the overlap of the longitudinal optical (LO) mode and the longitudinal optical phonon-plasmon coupled (LOPC) mode [16]. Tao Liu et al. studied the phenomenon of photo-generated carriers excited by ultraviolet light (325 nm line) and calculated the carrier concentrations of the epitaxial layer [17]. However, the research did not focus on the epitaxial layer thickness, the variation of the LOPC peak position in the depth direction, and the influence of different doping concentrations on the position of the LOPC peak. Aside from this, there has been no integrated report on the depth profiling of p-type ion-implanted 4H–SiC by confocal Raman microscopy, which typically has multilayer structures along the depth direction.

In this study, confocal Raman spectroscopy was used to analyze the depth profiling of 4H–SiC processed by ion-implantation and annealing treatments for p-type ion-implanted 4H–SiC. Information on the heavily doped substrate and epitaxial layer were distinguished, the thickness of the epitaxial layer was evaluated, and the influence of the top ion modified layer on the Raman depth analysis was explained in detail.

2. Experimental Setups

2.1. Sample Description

The starting materials were 4H–SiC (0001) nitrogen doped (N-doped) n⁺ substrates obtained from Dow Corning and SiCrystal (Germany). N-doped epitaxial layers of 5.3 μm, 6 μm and 7.5 μm were grown on substrates with 4° ± 0.5° off-axis toward <11–20> by the chemical vapor deposition (CVD) process. Multiple aluminum ion-implantations were used to achieve the uniform p-type doping profile. The implantation depth varied between 200 and 800 nm (criterion: N_A = N_D) [18]. Then, the samples were annealed at different temperatures of 1700 °C, 1800 °C, and 1900 °C with a capping layer. The samples had a multilayer structure, as shown in Figure 1, which includes heavily doped n-type substrates of several hundred microns, n-type epitaxial layers of several microns, and p-type ion-implanted layers of several hundred nanometers (except for the un-implanted sample G and the substrate material). Details of the sample preparation parameters are shown in

Table 1. Details of the samples used in Raman characterization.

Sample	Epitaxy		Implantation		Annealing
Label	Carrier Concentration (cm−3)	Thickness (μm)	Element	Surface Concentration (cm−3)	Temperature (°C)	Time (min)
G	≈6.0 × 1015	≈5.3	-a	-	1700	30
Q1	≈6.0 × 1015	≈5.3	Al	5.0 × 1016	1700	30
Q2	≈6.0 × 1015	≈5.3	Al	5.0 × 1017	1700	30
Q3	≈6.0 × 1015	≈5.3	Al	1.0 × 1018	1700	30
A	≈1.4 × 1016	≈7.5	N	5.0 × 1019	1700	30
	≈1.4 × 1016	≈7.5	Al	5.0 × 1019	1700	30
B1	≈1.0 × 1016	≈6	Al	3.3 × 1018	1700	30
B2	≈1.0 × 1016	≈6	Al	5.0 × 1019	1700	30
C1	≈1.0 × 1016	≈6	Al	3.3 × 1018	1800	30
C2	≈1.0 × 1016	≈6	Al	5.0 × 1019	1800	30
D1	≈1.0 × 1016	≈6	Al	3.3 × 1018	1900	1
D2	≈1.0 × 1016	≈6	Al	5.0 × 1019	1900	1
Sub b	-	-	-	-	-	-

^a No ion-implantation was performed in this area. Thus, no surface concentration is given. ^b Substrate material without epitaxial growth from Xiamen Powerway Advanced Material Co. Ltd. (PAM-XIAMEN).

.

2.2. Measurement Setups

The Raman measurements were performed at room temperature in backscattering geometry using a Horiba XploRA PLUS (Minami-ku, Kyoto, Japan) equipped with a 250 mm focal length spectrometer. All spectra were excited by the 532 nm line of a solid-state Nd:YAG SHG laser with a power of 100% (≈6.69 mW). A grating of 1800 lines/mm, 2400 lines/mm, and a thermo-electric (TE) cooled charge-coupled detector (CCD) with 1024 pixels were used to obtain spectra with a resolution better than 1 cm⁻¹. An adjustable pinhole (100 μm, 300 μm, or 500 μm) was placed in the back image plane to block light from outside the focal plane to ensure a spatial resolution better than 1.5 μm in the lateral direction and better than 2 μm in the longitudinal direction (parallel to the optical axis). The light was focused and collected through a 50× long working distance objective lens (N.A. = 0.5), 50× (N.A. = 0.75), or 100× lens (N.A. = 0.9) using a typical acquisition time of 10 s and accumulation of two times. Furthermore, the peak position and intensity of the Raman modes obtained from different depths were fitted using LabSpec 6 software (Horiba Scientific), according to the Gauss–Lorentz function. The depth resolution of confocal Raman microscopy is mainly determined by both the objective and pinhole, as shown in

Table 2. The depth resolution of a confocal Raman microscope as a function of objective lens and pinhole diameter [12].

Objective	N.A. a	WD b (mm)	Half-Entrance Angle θ c (deg)	Depth of Focus ω (μm)	Pinhole Diameter (μm)
					500	300	100
					Depth Resolution d (μm)
50×	0.75	0.37	48.6	0.4	6.0	3.0	1.5
50L× e	0.5	10.6	30.0	1.6	14	8.0	3.0
100×	0.9	0.21	64.2	0.1	3.0	1.2	0.7

^a Numerical aperture. ^b Working distance. ^c Half-entrance angle θ = sin⁻¹(N.A.). ^d The values have been multiplied by two to obtain the total depth where Raman signal originates from (above and below the focal plane). ^e Long working distance objective.

[12].

As above-mentioned, the ion-implanted 4H-SiC possesses a multilayer structure. To obtain information from the different layers, we altered the position of the focal plane by adjusting the sample stage along the optical axis with a step of 1 μm and 0.5 μm for the 50L× lens and 0.2 μm/1 μm for the 100× lens while the objective lens remained stationary and recorded the spectrum for each step (Figure 2). The sample stage was driven by a mechanical motor with an accuracy of 0.1 μm.

3. Results and Discussion

3.1. Depth Profiling Set-Ups

The selection of measurement parameters has a significant impact on the data fitting and analysis. For depth profiling of a multilayer structure, both spatial resolution and signal intensity need to be considered. As is well known, the lateral resolution of a Raman microscope is directly related to the laser spot size, which is mostly governed by the laser wavelength and the N.A. of the objective lens. Except for these factors that determine the lateral resolution, the size of the confocal pinhole influences the depth resolution significantly. The micro scale pinhole placed in the back image plane filters most of the scattered light from outside the focal plane to obtain an optimized depth resolution. Furthermore, the sample characteristics, especially the refractive index and absorption coefficient, also have a remarkable impact on the depth resolution of the system [12,19].

To evaluate the depth resolution of the measurement system, depth profiling with different pinholes and different lenses was studied using sample Q1 processed by Al ion-implantation with a surface concentration of 5 × 10¹⁶ cm⁻³ followed by annealing at 1700 °C for 30 minutes (Table 1). After calibrating the spectrometer using a silicon chip under the condition of 532 nm laser excitation, single spectra of Q1 with different pinhole sizes and lenses were collected focusing on the surface of the sample, as shown in Figure 3a,b. In the backscattering geometric configuration, A₁ modes (longitudinal acoustic mode (LA mode) at 610 cm⁻¹, LO mode at 964 cm⁻¹, and LOPC mode at ~980 cm⁻¹), E₂ mode at 776 cm⁻¹, E₁ mode at 797 cm⁻¹, and the second order Raman peaks can be observed. The spectra showed two overlapping peaks in the range of 950–1000 cm⁻¹ (framed with red dashed lines in Figure 3a,b, which were the longitudinal optical mode (LO mode) located at 964 cm⁻¹ and the LOPC peak located at around 980 cm⁻¹ [20], respectively. As the diameter of the confocal pinhole becomes smaller or the numerical aperture of the objective lens becomes larger, the LOPC peak information is filtered out more, so the LOPC information is assumed to be derived from the substrate of the sample. Pinholes with diameters of 300 μm and 500 μm exhibit a similar intensity ratio because the true confocal mode is only available when the pinhole diameter is less than 200 μm. The objective lens with large N.A. corresponds to a smaller entrance pupil size, which filters more information from the substrate, thus shows a much weaker LOPC mode. To further confirm the origin of the LOPC mode, a Raman spectrum of the heavily doped n-type 4H–SiC substrate without epitaxial growth was measured and is shown in Figure 3c. For the substrate, only the LOPC peak can be observed at around 978 cm⁻¹, which is estimated to correspond to a carrier concentration of 2.7 × 10¹⁸ cm⁻³ [21].

The distribution of the Raman signal in the depth direction (along the z-axis) is shown in Figure 4, which corresponded well to the multilayer structure of the sample. When focusing on the substrate layer (negative defocus distance), no LO peak appeared, whereas an obvious LOPC peak was observed. When the focal plane moved toward the surface of the sample (which corresponded to 0 for the defocus distance), the LO peak gradually appeared and the LOPC peak decreased. Thus, we classified the LO peak as the longitudinal optical mode (LO mode) of the epitaxial layer and the LOPC peak as the longitudinal optical phonon-plasmon coupled mode (LOPC mode) of the heavily doped substrate, respectively. Druy et al. arrived at similar conclusions [16]. Figure 4a–c indicate that the distribution of the LO peak along the depth direction had a narrower width for smaller pinholes, which implies a stronger ability to distinguish multilayer structures for smaller pinholes. Figure 4d–f show that the distribution of the LO peak intensity along the depth direction was very steep for the 100× lens due to the higher spatial resolution of the objective lens with a large numerical aperture.

To compare the depth profiling results using different objective lenses, the authors compared the results of the 50L× and 100× objective lenses and 100 μm pinhole with the z-scanning step of 1 μm. Figure 5 shows the LO mode and LOPC mode distributions in the depth direction after normalization with the intensity maximum. As shown in Figure 5a, the signal intensity of the 100× objective lens changed steeply, showing its high spatial resolution. When comparing the peak position of LO and LOPC mode in Figure 5b,c, the results of the 50L× lens had a better coherence and showed a more stable regular change. Spectra when focusing on the sample surface in Figure 5d,e showed that the luminous flux of the 100× objective lens was smaller and the signal-to-noise ratio of the signal was worse than that for the 50L× lens. Therefore, for the 100× lens, the fitting error was larger than that for the spectra measured with the 50L× lens, which was not conducive to the analysis of the subsequent laws of LO and LOPC mode position for the depth profiling.

3.2. Determination of Epitaxial Thicknesses by Depth Profiling

Depth profiling can distinguish the LOPC and LO signal from the substrate and the epitaxial layer. Considering the high depth resolution of the 100× lens (Table 2), depth profiling experiments were first performed using the 100× lens, as shown in Figure 6a. The LOPC peak began to decrease at around −2 to −3 μm, which is shallower than the actual position of the interface between the substrate and epitaxial layer (Table 1), with deviations of around 3 to 4.5 μm. This phenomenon conforms to the law of refraction. Silicon carbide has a large refractive index of 2.68 at the laser wavelength. As shown in Figure 6b,c for the 100× lens, which has a large half-entrance angle θ of 64.2°, the actual focus point (Z’) caused by refraction would be much deeper than the theoretical focus point Z (controlled by the positioning of Z-axis sample stage) without considering the influence of refraction. Therefore, although the 100× lens has the best depth resolution in principle, because of its large angle of incidence, the large depth deviation of the actual focal point caused by refraction will degrade its depth characterization performance to some extent.

As shown in Figure 7, for samples with different epitaxial thicknesses, the inflection point at which the LOPC intensity began to decrease was close to the interface of the substrate and the epitaxial layer using the 50L× lens. The thicknesses of the epitaxial layers are shown in Table 1, which were measured by FTIR (Fourier-transform infrared spectroscopy) measurements [22,23,24,25]. The inflection point was determined using the depth at which the maximum intensity of the LOPC mode appeared (negative defocus distance). Multiple measurements (at least five times) with a depth step of 0.5 μm were performed on the same epitaxial layer. The mean position of the maximum of the LOPC peak intensity x and the standard deviation of the mean σ_x were calculated. The deviation (Δ = x – Th _epi) between the average position of the LOPC peak maximum (x) and the real interface position measured by FTIR (Th _epi) are listed in

Table 3. The interface position measured by FTIR measurements, average position of the maximum of LOPC peak intensity, and the deviations of the Raman depth profiling for samples with three kinds of epitaxial thicknesses. The step of depth profiling is 0.5 μm.

Label	Interface Position Measured by FTIR (Thepi/μm)	Average Position of LOPC Intensity maximum (x/μm)	Standard Deviation of the Average Position (δx/μm)	Deviation of Raman Depth Profiling (Δ/μm)
Q1	–5.3	–7.1	0.4	–1.8
Q2	–5.3	–6.3	0.4	–1.0
G	–5.3	–7.6	0.1	–2.3
B1	–6.0	–6.7	0.2	–0.7
B2	–6.0	–7.9	0.1	–1.9
C1	–6.0	–7.0	0.5	–1.0
An	–7.5	–8.6	0.4	–1.1
Ap	–7.5	–9.1	0.2	–1.6

. The deviations were distributed in the range of −1.2~−1.7 μm. For the three kinds of thicknesses for the epitaxial layer (5.3 μm, 6 μm, and 7.5 μm), the average deviations of the Raman depth analyses were −1.7 μm, −1.2 μm, and −1.4 μm, respectively. The standard deviation of the average position was between 0.1 μm and 0.5 μm. Furthermore, it should be noted that, in the actual sample, there is a buffer layer between the substrate layer and the epitaxial layer with a thickness of ~0.5 μm. The presence of the buffer layer may be a cause of the deviation in the results.

Thus, the thickness of the epitaxial layer can be determined by the inflection point at which the LOPC peak intensity reaches the maximum with deviations of –1.7 μm, –1.2 μm, and –1.4 μm for epitaxial layer thicknesses of 5.3 μm, 6 μm, and 7.5 μm. Therefore, depth profiling with a 50L× lens is an effective approach to determine the thickness of the epitaxial layer with deviations of –1.2~–1.7 μm.

Compared with the depth profiling results, the thickness of the epitaxial layer could be better characterized by the 50L× lens with smaller deviations. Two key parameters of the objective lens can well explain the above results. On one hand, the smaller angle of incidence θ for the 50L× lens (Table 2) reduced the refraction influence. The length of the laser beam waist, which is also called the depth of focus ω [11], is shown in Table 2 and Figure 6b. ω is usually defined as Equation (1).

$$\omega =6.4\times \frac{\lambda }{2\pi }\times {\left(\frac{1}{tan\theta }\right)}^2\approx \lambda \times {\left(\frac{1}{tan\theta }\right)}^2$$

(1)

where λ is the wavelength of the incident laser and θ is the half-entrance angle of the objective lens. On the other hand, the 50L× lens has a larger depth of focus (Table 2), which can compromise the refraction influence on depth profiling to some extent.

3.3. Depth Profiling of Net Doping Concentration

The above analysis identified the overlap of the LO and LOPC mode related peaks in the range of 950–1000 cm⁻¹. Furthermore, we aimed to extract the information of the nanoscale surface layer being modified by ion-implantation and annealing from the Raman depth analysis. Depth profiling of the samples in Table 1 with different carrier concentrations and different annealing treatments were studied. The variation of the LO and LOPC peak position with nominal focus position (also called the defocus distance in this paper) in Figure 8 were excited by 532 nm laser using 100 μm pinhole and the 50L× lens, where the gray blocks represent the substrates and the different colored blocks represent the epitaxial layers of different samples (according to the data in Table 1).

The results in Figure 8a,b, which have the same coordinate scale showed that the LO peak position in all regions was relatively stable, fluctuating within 0.5 cm⁻¹, while the LOPC peak showed a blue shift of around 1.5 cm⁻¹ in the depth direction between −30 µm and the 0 µm defocus distance. Similarly, in Figure 8c,d, the LO peak position showed a line shift no more than 0.5 cm⁻¹ and the LOPC peak showed a blue shift of 1.0–1.5 cm⁻¹ from the substrate to the sample surface. The Raman shift of the LOPC peak was relatively stable when focusing on the substrate and above the surface of the sample. In contrast, a manifest blue shift was observed when the focal plane moved from the interface of the substrate layer and the epitaxial layer to the surface of the sample (between −7.5/−6/−5.3 µm and 0 µm defocus distance). Many studies have shown the sensitivity of the LOPC mode on carrier concentration of the n-type 4H–SiC [26,27,28,29,30]. The heavily doped n-type substrates had relatively high carrier concentrations in the orders of 10¹⁸ cm⁻³, while the n-type epitaxial layers had rather low carrier concentrations of ≈1.0 × 10¹⁶ cm⁻³ comparatively. When focusing on the substrate, most of the signal was from the substrate layer, so the LOPC peak position was relatively stable. When the focusing point moved from the substrate to the epitaxial layer, the apparent difference in carrier concentration between the substrate (~10¹⁸ cm⁻³) and the epitaxial layer (~10¹⁶ cm⁻³) led to a blue shift in the LOPC peak. As the sampling volume was not infinitely small, the change in LOPC peak position was not as steep as the change in carrier concentration. Furthermore, the buffer layer, which had a carrier concentration lower than the substrate and higher than the epitaxial layer, also led to a smoother transition of the LOPC peak position from the substrate to the epitaxial layer. Thus, the blue shift in the depth direction of the LOPC mode can be attributed to the carrier concentration gradient from the substrate to the epitaxial layer. Additionally, the different annealing process parameters did not result in a significant difference in the peak positions of the LOPC mode. In addition, when comparing the frequency shift of the LOPC peak position among samples with different thicknesses of the epitaxial layer such as sample A and Q1/Q2/Q3 in Figure 8d, a greater blue shift around 1 cm⁻¹ in the depth direction was found for sample A, which had a thicker epitaxial layer.

For samples Q1, Q2, and Q3, the LOPC mode relationships in Figure 8d corresponded well to our previous single Raman spectrum results focusing on the sample surface [20]: when the surface Al carrier concentration was equal or larger than 10¹⁸ cm⁻³ (i.e., Q3), the LOPC mode showed a blue shift whereas no apparent difference was found in the LOPC peak position between Q1 and Q2.

Furthermore, samples Q1, Q2, and Q3 had exactly the same substrates and epitaxial layers and only differed in the implantation dose, thus the same LOPC mode position of the substrate was expected to be seen for Q1, Q2, and Q3. However, the LOPC peaks position were different when focusing on both the substrates and the epitaxial layers. From the comparison of the processing of Q1, Q2, and Q3, the only reason for the difference in the LOPC peak position was the different ion-implantation treatments. After the exciting photons reach the substrate layer, they carry the information of the substrate and scatter to the sample surface and reach the detector. In this process, photons from substrate layers travel twice through the surface modified layer (ion-implanted layer), as shown in Figure 9. Since the LOPC mode is sensitive to the carrier concentration, the twice travel of the photons through the p-type (hole-conducting) surface modified layer will bring inevitable influence from the p-type ion-implanted layer. The differences were evident between the LOPC positions of the Q1 and Q2 peaks (which were very similar) and the Q3 peaks. We believe that the second coupling of those photons (from the substrate layers) and the hole carriers in the surface Al ion-implanted layers led to the difference in the LOPC peak position of Q1, Q2, and Q3.

4. Conclusions

The confocal Raman depth profiling method is a fast and non-destructive method to obtain the electrical properties profiling of ion-implanted silicon carbide in the depth direction. The results of the depth analysis corresponded well to the surface single-point spectra we reported earlier and provides plentiful information on the following. (1) Raman spectra distribution in the depth direction identified the LO mode as the longitudinal optical mode (LO mode) as representative for the epitaxial layer and the longitudinal optical phonon-plasmon coupled mode (LOPC mode) as representative for the heavily doped substrate, respectively. (2) The epitaxial layer thickness could be evaluated according to the intensity variation of the LOPC peak in the depth direction from the position of the maximum of the LOPC peak intensity. For three kinds of thicknesses for the epitaxial layer (5.3 μm, 6 μm, and 7.5 μm), the average deviations of the according Raman depth analysis were −1.7 μm, −1.2 μm, and −1.4 μm, respectively. (3) For ion-implanted 4H–SiC samples, the apparent difference in carrier concentration between the substrate (~10¹⁸ cm⁻³) and the epitaxial layer (~10¹⁶ cm⁻³) led to a blue shift in the LOPC peak along the depth direction (from the substrate to the epitaxial layer). A greater blue shift was found for the thicker epitaxial layer of sample A. (4) The twice travel (excitation and collection) of photons from the substrate resulted in a second coupling with the surface ion-implanted layer, which led to the difference in the LOPC peak position of Q1, Q2, and Q3.

Furthermore, the depth profiling results with different sized pinholes and lenses showed that the 100 μm sized pinhole and 50L× lens were more suitable to undertake the depth analysis of the 4H–SiC substrate with an epitaxial layer of several micrometers in thickness and a top layer in the nanoscale, considering both the depth resolution, signal to noise ratio, and refraction effect.