1. Introduction

Due to its non-destructive spectroscopic nature, photoluminescence (PL) measurement technique [1] has been a versatile and efficient method to analyze the optical characteristics of the semiconductor epitaxial layers. A spatially scanning PL system would give important information especially about the role of substrates on optical characteristics of epitaxial films grown over them. In order to cope with rapidly growing demand for PL morphology studies, confocal microscopy-based PL systems have been developed and applied to optical characterization of light emitting diode (LED) epi-wafers [2,3]. Due to compact and flexible light guiding capability, optical fibers have been recently adopted in PL systems. For instance, Jayakrishnan et al. [4] have used an optical fiber PL mapping system to evaluate the semiconductor deposition processes and the uniformity of the layers. Wang et al. [5] have demonstrated fluorescence spectroscopy using a double-clad fiber (DCF), where the core delivered the excitation light and the inner cladding collected the fluorescence light. In recent years, demands for PL or fluorescence research are being directed toward development of a compact and flexible system for in situ analysis. Conventional microscopy technologies have shown very good performances in terms of spatial resolution, but their large physical form factor is a fundamental obstacle for in situ PL analyses. Fiber optic PL systems have been rapidly developed to take advantage of its flexible and compact light guiding capability for micro PL applications [6,7]. However those prior fiber optic PL systems did not provide a sufficiently high spatial resolution required separate bulk lens systems, and couplers. Despite their high potential, fiber optic PL systems have not been fully adopted in recent LED processes and only a few PL studies have been reported on correlation between the substrate patterning and optical properties of epi wafers.

In recent years, GaN-based semiconductors have been intensively applied in optoelectronic devices such as LEDs and laser diodes (LDs), due to their direct, wide band-gap properties [8]. Conventional GaN epitaxial films have been routinely grown over sapphire substrates but threading dislocation (TD) has been unavoidable due to the large mismatch in the lattice constant and thermal expansion coefficient between the two materials [9]. Epitaxial lateral overgrowth (ELOG) [10] has been introduced to reduce TDs, but the technique suffers from a fundamental drawback of requiring repeated chemical processes. LEDs grown on a patterned sapphire substrate (PSS) have attracted much attention due to their process advantages such as the single growth without any dielectric mask layer, and significant improvement in the light extraction efficiency by scattering or redirecting the light in LEDs [11–13]. However, to the best knowledge of the authors, the fiber optic system has not been previously used for detailed PL comparison in terms of the luminescence uniformity of the epi-layers grown on patterned or un-patterned sapphire substrate (UPSS).

In this paper, we demonstrate an integrated all-fiber PL measurement system based on a hybrid fiber optic probe, which was applied to comparative analysis of optical characteristics of two types of blue LED epi-wafers: InGaN/GaN grown on UPSS and PSS. The salient features of the proposed fiber optic PL system are: 1) the concatenated fiber segments composed of DCF and single mode fiber (SMF) provided a high launching efficiency of the UV excitation light source with an inherent mode filtering capability, 2) a micro-lens formed on the coreless silica fiber (CSF), which was fusion spliced to the end of DCF-SMF hybrid fiber, procured an all-fiber integrated focusing capability, and 3) a low loss Y-type DCF cladding mode coupler enabled an efficient collection of the PL signal from the hybrid probe and a robust PL delivery to the photodetector. The proposed all-fiber device provides integrated optical paths for both the excitation laser and the PL signals in a single hybrid fiber unit, which significantly improves system integration capability. The spectral uniformity of the PL from the two types of GaN epi-wafers was systematically compared in terms of the peak intensity and the peak wavelength of the photoluminescence, for the first time.

2. A hybrid fiber optic probe with an integrated fiber lens

Most of DCFs used in PL applications have the inner cladding composed of pure silica and the outer cladding of a low refractive index polymer. The refractive index of the core is raised by adding GeO₂ in the silica. DCF provides an efficient multi-mode guidance for the PL signals through the inner cladding by the large index difference between the silica cladding and the lower refractive index outer polymer cladding. DCF can guide excitation light through the core in conventional manner by the index difference between the core and inner cladding [14]. However, most of commercially available DCFs have been designed to operate in the IR region for fiber laser and amplifier applications, which do not provide the single mode guidance for the UV-violet excitation photon in PL for GaN LEDs. In order to adopt the DCF for a blue LED PL system, it is necessary to deliver the UV excitation light through the core in the fundamental mode in order to avoid random modal couplings and secure a well-defined minimum beam spot diameter. In this study, we transformed a commercially available SMF in the visible range (Corning RGB400) to an in-house DCF that can guide the excitation laser at λ = 405 nm in the LP₀₁ mode along the core, and the PL signal centered near 430 nm along the inner cladding. Firstly the high refractive index coating of the commercial SMF was chemically removed in an organic solvent bath. Then the bare SMF was mounted on a fiber drawing tower (SG-Controls) and continuously re-coated by UV-curing a low refractive index polymer (SSCP EFiRON® PC-375, refractive index is 1.382 at 850nm) layer as shown in Fig. 1(a) . The DCF structure is shown in Fig. 1(b). The diameters of the DCF core, inner cladding and outer polymer cladding were 4, 125 and 180 μm, respectively.

 

Fig. 1 Dual cladding fiber (DCF) fabrication. (a) Recoating a single mode fiber with a low refractive index polymer in a continuous manner using an optical fiber drawing tower. (b) Structure of the DCF with a 4 μm core, 125 μm inner cladding, and 180 μm outer cladding.

Download Full Size | PDF

The numerical aperture (NA) between the core and the inner cladding was 0.12 and that between the inner cladding and the outer polymer cladding was 0.44.

When a UV-visible laser with an output beam diameter of a few mm is launched directly to the DCF, most of the light is guided along the inner cladding to result in a very undesirable output beam spot: a large modal area fluctuating with speckles generated by the modal interference. Selectively launching of the UV-violet laser through the core using a microscopic objective (MO) is not an optimal solution either, because the tight focusing via MO inevitably demands a high NA, which results in not only significant light-coupling to the inner cladding, but also power fluctuation in the core-coupled light that is very sensitive to external vibrations. This makes it even more problematic for consistent PL mapping experiments.

In order to solve this coupling issue between DCF and the UV-violet excitation laser, we proposed a hybrid fiber optic probe that can provide both a mechanical robustness in the launching and an efficient core-guidance for a stable and small spot output beam. The hybrid fiber is composed of concatenated segments of ‘input DCF’-‘SMF’-‘output DCF’ using an arc fusion splicer as shown in Fig. 2(a) . Key feature in the proposed hybrid fiber, input DCF-SMF-output DCF, is efficient removal of higher order modes of the “input DCF” by the radiation loss along the high refractive index polymer jacket of SMF [15,16]. The novelty of the proposed cascade fiber was to keep the glass optical waveguide structure identical in the axial direction but the polymer jacket indices were locally modified to selectively induce the radiation losses for higher order modes. We also utilized an tandem fiber lens [17,18] to further reduce the output beam diameter, which is directly related with the spatial resolution of the PL system.

 

Fig. 2 Proposed hybrid fiber optic probe. (a) Schematic diagram for concatenated fiber structure and transformation of the guided modes in the axial direction. (b) Fabrication of the fiber-lens using a coreless silica fiber (CSF) spliced to the end of ‘output DCF. (c) Intensity profile of the excitation laser (λ = 405 nm) output at the focal length.

Download Full Size | PDF

The excitation laser at λ = 405nm is butt-coupled to ‘input DCF’ and guided along both the core and inner cladding modes. We took advantage of high NA and large inner cladding diameter of the ‘input DCF’. The robust coupling between the laser and the ‘input DCF’ was achieved using a commercially available SMA (sub-miniature version A) type fiber connector via butt-coupling without any focusing optics.

The launched excitation laser propagates along the ‘input DCF’ in both the core mode and the inner cladding modes. The red and the blue lines in Fig. 2(a) represent the fundamental mode in the core and the inner cladding modes, respectively. The ‘input DCF’ segment was then fusion spliced to SMF as in Fig. 2(a). Note that the bare DCF and the bare SMF are identical in their waveguide structure and a seamless power transfer of guided modes was successfully achieved. When the inner cladding modes of the ‘input DCF’ propagate along the ‘SMF’ sections, they experience strong leaky radiation loss due to the high refractive index in the SMF outer cladding. See the schematic refractive index profiles in the bottom of Fig. 2(a). After the suppression of the cladding modes in SMF, we further spliced the SMF to the ‘output DCF’ to guide the core mode to the output. The proposed leakage mechanism along the hybrid fiber did not completely remove the fundamental inner cladding mode of the DCF. The output beam was a combination of the fundamental core mode and the inner cladding mode, which resulted in a mode field diameter larger than that of Corning RGB 400 fiber.

In order to further reduce the output beam spot size, we added micro fiber-lens in a tandem structure [17,18] at the end of the hybrid DCF-SMF-DCF. We spliced a segment of coreless silica fiber (CSF) to the cleaved end of ‘output DCF’, which was subsequently melted by applying an electric arc discharge from a fusion splicer to form a spherical lens tip as shown in Fig. 2(b). The curvature of the fiber lens was controlled by changing the arc current and duration time. The focal length and beam waist of the fiber lens depends on the CSF length and lens curvature. Using the CSF with the diameter of 125 μm and the length of 500 μm, we formed a fiber lens with the radius of curvature of 69 μm, and a focal length of ~125 μm. The excitation laser at λ = 405 nm had a beam diameter of ~1mm and optical power of 100 mW. Using the proposed hybrid fiber composed of 6cm of “input DCF”-6cm of SMF-6cm of ‘output DCF’, we were able to obtain a near Gaussian output beam whose beam diameter was ~13.6 μm at the focal length of 125 μm. See Fig. 2(c). The output power was about ~8 mW, which is comparable to the case when a bulk microscopic lens is used to focus the laser beam to the core.

3. Integration of ‘DCF to DCF cladding mode coupler’ for PL delivery

Recently DCF couplers have been widely adopted in various fiber optic medical imaging techniques [5,19,20]. Most of prior DCF couplers have been based on fusion and tapering technique which required strict alignment of two fibers in a parallel manner and sophisticated process optimization [21]. In this study, two strands of bare DCF were inserted in a custom-made silica V-groove chip and then a UV curable low refractive index polymer was applied to provide a consistent parallel contact between two DCFs. Mode coupling between two parallel fibers have been well-established theoretically [22,23], which laid a fundamental basis for various fiber couplers. When two cores are touching each other as in our case for DCFs, the coupling constant in non-vanishing even for high V-numbers and its values are large enough to result in good coupling within a relatively short distance [23]. A schematic structure of the proposed coupler integrated to the hybrid fiber optic probe is shown in Fig. 3(a) . The V-groove provided a well-defined contact length between the two parallel DCFs, which (subsequently) resulted in an efficient coupling for the inner-cladding modes without affecting the core mode. In experiments, the coupling region length was measured to be ~2.5 mm with the power splitting ratio of ~18:82 for the PL signal near λ = 430 nm. The cross-talk between the two cores of DCFs for the excitation light at λ = 405 nm was measured to be negligibly small below −30 dB, which is ideal for isolation of the excitation laser from PL signals at the photodetector. The loss for the core-guided excitation light was less than 5%, comparable to prior reports [5,19,20]. The proposed DCF coupler was further integrated to the hybrid ‘DCF-SMF-DCF/ CSF-fiber lens’ assembly for an all-fiber solution. Again, coupling between the ‘output DCF’ of the hybrid probe and the ‘receiving arm DCF’ was made using the proposed V-groove technique.

 

Fig. 3 Fiber lens integrated over hybrid fiber optic probe. (a) Structure of the proposed Y-type DCF coupler integrated on the hybrid fiber optic probe, (b) Schematic diagram for collecting PL signals using the proposed all-fiber probe. Here we used TiO₂ film on the end of ‘receiving arm DCF’ to collect only the PL directly excited by the focused beam on the sample wafer.

Download Full Size | PDF

The location of the coupling zone was carefully optimized to maximize the PL signals at the receiving arm, which extended about ~100 μm from the interface between ‘output DCF’ and CSF, as shown in Fig. 3(a). The ‘receiving arm DCF’ was cleaved near the end of the coupling zone to make a Y-coupler. Its cleaved end was then coated with a TiO₂ powder paste as shown in Fig. 3(b) and this served to block the direct collection of PL signals bypassing the fiber lens probe, providing a consistent PL morphology capability in the proposed PL system. The fabricated DCF coupler integrated with the hybrid fiber probe procured both a small beam diameter of 13.6 μm for the excitation laser at λ = 405 nm and an efficient collection/delivery over 15% for PL signal λ = 430 nm to the photodetector and spectrum analyzer. Due to these salient features, the proposed fiber optic PL system provided a well-defined and consistent spatial resolution less than 14 μm along with a high signal to noise ratio in PL spectra.

4. Configuration of the fiber-based PL system

We set up a scanning PL measurement system using the proposed all-fiber probe, as schematically shown in Fig. 4 . A laser at λ = 405 nm (405 Blue Laser, RGBLase Co.) with the maximum power of 100 mW was used as an excitation light source. A spectrometer (SM240, CVI Spectral Products) with the spectral range from 300 to 1050 nm was used to analyze the PL signals. A x-y nano-positioning stage (Em4sys Co.) was used as a movable sample mount, whose displacement was electronically controlled by a computer with a stable spatial accuracy of 100 nm. The LED epi-wafer was mounted on the XY nano-positioning stage and it formed a right angle in reference to the fiber optic probe with an error of ± 0.5þ. This alignment range did not influence either the excitation power or the PL detection in the system due to the unique alignment tolerance in the CSF-fiber lens [18].

 

Fig. 4 Schematic diagram of the proposed fiber optic PL measurement system, consisting of a laser source, spectrometer, XY-nano positioning stage, and Y-type DCF coupler including the integrated fiber probe. The violet line represents the excitation laser beam, and the green arrows represent the PL signal route from the epi-wafer.

Download Full Size | PDF

The prior fiber optic micro PL systems [6,7] required two optical paths: one for the excitation laser delivery and the other for the PL signal collection. In reference [6], the excitation laser at λ = 442 nm was delivered through a tapered fiber optic scanning near-field optical microscope (SNOM) probe, but the PL signal was collected by an additional bulk microscopic objective lens system. In reference [7], two separate fiber optic probes were used for the excitation laser and the PL detection. In contrast, our proposed all-fiber device can provide optical paths in a single fiber unit for both the excitation laser and the PL signals, which significantly improves system integration capability and a wider scanning range.

The PL morphology measurement processes, including PL signal detection, saving its spectra, and moving the x-y stage, were all controlled by a LABVIEW program. The measurement procedure is schematically shown in the inset of Fig. 4. The excitation laser was focused on the LED wafer, and the PL signals were collected by the fiber lens and guided through the ‘output DCF’, from which it was transferred to the ‘receiving arm’ of the Y-type DCF coupler for further spectral and optical power analysis. This cycle was repeated over a desired area of the wafer by using the x-y stage to analyze the morphological characteristics.

5. PL of the InGaN/GaN LED epi-wafers and its morphology

Two types of InGaN/GaN LED epi wafers grown on UPSS and PSS were used to evaluate the performance of the proposed fiber-based PL system. The epi-wafers grown on two inch c-plane sapphire substrate by metal-organic chemical vapor deposition (MOCVD) and they were provided by EpiValley Co. Ltd.. The epi-structures consisted of a GaN buffer layer, a n-type GaN layer, InGaN/GaN multi-quantum well active layers, Mg doped AlGaN layer cladding layer, and p-type GaN layer, which was the same as the conventional blue LED structure [24,25]. Figure 5 shows the cross-sectional scanning electron microscope (SEM) images of the two epi-layers on (a) UPSS and (b) PSS. The thicknesses of the LED layers on UPSS and PSS were ~5 and ~7 μm, respectively. To fabricate the PSS, the well-known thermal photo resist (PR) reflow and dry etching method were employed [26].

 

Fig. 5 Cross-sectional Scanning Electron Microscope (SEM) images of GaN LED epi-layers on (a) UPSS and (b) PSS

Download Full Size | PDF

A typical PL spectrum measured in this study is shown in Fig. 6 . The PL signal had a peak wavelength near 442.2 nm and its full width at half maximum (FWHM) was ~14.7 nm, which is comparable to prior reports [27,28]. The excitation laser at λ = 405 nm was also identified in the PL spectrum due to the reflection at the epi-wafer surface. To further quantify PL morphology of the two types of epi-wafers, the fiber probe was scanned over the square area of 200 × 200 μm² and the PL signals were collected with the spatial spacing of 10 μm, maintaining the focal length along the z-axis. At each local point, total of 121 spectra were obtained in a detection time of 1 second. The PL signals were Gaussian fitted to identify the peak wavelength, its full width at half maximum (FWHM), and its peak intensity. We found out that the average PL intensity from the PSS wave was more intense by a factor of 2.68 than that from UPSS. In order to level the PL intensity, we kept the launched power of the excitation laser at 75 mW and 35 mW for the UPSS, and PSS samples, respectively.

 

Fig. 6 A typical PL spectrum of the InGaN/GaN LED epi-wafer measured by the proposed PL system. Here the excitation laser was at λ = 405 nm.

Download Full Size | PDF

Figure 7 shows the maps of (a) the peak intensity and (b) the peak wavelength, and (c) the FWHM of PL signals from the UPSS (left) and PSS (right). The averaged values of these parameters and their standard deviations are summarized in Table 1 . PSS had a higher peak intensity and its deviation was also larger, which shows an effect of PSS consistent to prior reports. However this relative comparison would not give detailed physical information on the quality of crystalline structure simply because the thickness of epi-layers was different as shown in Fig. 5. Further systematic comparison is being pursued by the authors. In terms of the average peak wavelength, the LED on UPSS had 458.8 nm with a deviation of 4.1 nm, while PSS sample had average wavelength of 440.7 nm with a deviation of 2.4 nm. This relative comparison in the average wavelength deviation might indicate that the multi-quantum layer on the PSS would be more uniform. FWHM consistently showed a larger value of 19.6 nm in UPSS with the deviation of 8.6 nm in comparison to 12.1 nm with the deviation of 7.8 nm in PSS.

 

Fig. 7 Contour maps of (a) Peak Intensity, (b) Peak Wavelength (c) FWHM of the LED epi-wafers on the UPSS (left column) and PSS (right column).

Download Full Size | PDF

Table 1. Average PL peak properties for two types of LED epi-wafers.

View Table

The proposed fiber-optic PL system successfully provided PL morphology data that can be further utilized to characterize the impacts of epi-layer structure and the sapphire substrate patterns over the LED optical properties. Further refinement of beam spot size is being pursued by the authors and subsequently enhancement of the spatial resolution is expected.

6. Summary

We fabricated an integrated all-fiber PL probe consisting of a DCF-SMF-DCF hybrid fiber, a tandem fiber lens, and a Y-type DCF coupler, which were used in a compact scanning PL mapping system for LED epi-wafer characterization. The DCF-SMF-DCF hybrid fiber provided a mechanically robust butt-coupling to an excitation laser along with an efficient core mode conversion. The tandem fiber lens focused the excitation laser at λ = 405 nm to a beam diameter of 13.6 μm. The Y-type DCF coupler based in a V-groove made an efficient coupling with more than 15% of the launched PL power. PL scanning system enabled fast and consistent morphology study over a 200 × 200 μm² area with a 10 μm moving step size for InGaN/GaN LED epi-wafers grown on two types of substrates, PSS and UPSS. Morphology of the PL peak intensity, the peak wavelength, and its FWHM were successfully obtained and analyzed. The proposed all-fiber scanning PL system could make a compact and versatile alternative to prior bulk PL systems. We expect that further optimization of focused beam diameter could improve the spatial resolution of the system and provide an ample potential to epi-layer optical characterization.