Open Access
Add to BibSonomyAbstract
We report on a novel combination of measurement techniques for evaluating dimensional and compositional changes of selective-area-grown multiple-quantum-well laser diodes (SAG MQW LDs). This technique is based on C-V and I-V measurements of the fully fabricated LDs. Using this technique, the changes in the capacitance and voltage correspond to the layer thickness and bandgap energy. To verify the effectiveness of the proposed technique, we first fabricated an LD array containing ten different SAG MQW structures, and examined the effects of the dimensional and compositional changes on the wavelength shift both theoretically and experimentally. From our examination, we found that a wavelength shift of 83 nm is obtained for an SAG mask pattern with an opening width of 100 μm, and that a cross point between both dimensional and compositional changes exists for this mask pattern. As the following step, the fabricated LD array was tested using the proposed technique, and the growth rate enhancement and bandgap energy were extracted from the measured C-V and I-V results. The extracted data for each array channel were compared with the simulation results, which were well-fitted from the photo-luminescence (PL) measurements. They both show good agreement with the simulation results.
© 2014 Optical Society of America
1. Introduction
The selective-area growth (SAG) of compound semiconductors has attracted considerable interest as a useful technique for monolithically integrated optoelectronic and photonic devices [1]. SAG is based on metal-organic vapor phase epitaxy (MOVPE) on dielectric patterned substrates. Through this technique, the thickness and material bandgap of the epitaxial layer can be controlled simply by adjusting the opening width between the adjacent mask patterns under the specific growth temperature and pressure. As a result, a monolithic integration of various functions can be realized using a single growth process.
SAG has been widely studied by many research groups for different material systems (i.e., InGaAsP/InP, InGaAlAs/InP, InGaAs/GaAs, and InGaN/GaN), various growth conditions and pattern configurations (i.e., growth temperature, pressure, pattern geometry, and deployment), and diverse integration schemes (i.e., distributed feedback (DFB), distributed Bragg reflector (DBR) laser diodes (LDs)/electro-absorption modulators (EAM), semiconductor optical amplifiers (SOAs)/EAM, broadband super-luminescent diodes, and multi-channel laser diode arrays) [2–9].
During the SAG process, the epitaxial layer grown on an opened area (i.e., the SAG layer) will simultaneously undergo the dimensional changes caused by the enhanced growth rate of the group III precursors (i.e., In, Ga, and Al) and the compositional changes from the different diffusion characteristics of each precursor. For SAG layers employing a multiple quantum well (MQW) structure, it was reported that the photo-luminescence (PL) spectrum is red-shifted, which mainly originates from the lowering of the quantum level (caused by the enhanced growth rate) and the reduction in the bandgap energy (by the Indium enrichment). However, in analyzing and evaluating this results, because the growth rate and composition of all element layers (i.e., the well, barrier, and/or separate confinement hetero-structure (SCH)) are changed, it is difficult to discriminate the effect of each change on the wavelength shift separately without performing additional tests such as absorption spectra of individual transition energy level differences in the MQW structure or PL measurements for the individual bulk layers. In addition, a highly compressive-strained layer (used as a well material) is difficult to realize as a bulk layer owing to the critical thickness. Therefore, a new technique for effectively evaluating the dimensional and compositional changes of the MQW SAG structure is required.
On the other hand, changes in the growth rate and composition of a bulk layer have been verified experimentally through measurements of the layer thickness using scanning electron microscopy (SEM), transmission electron microscopy (TEM), optical interferometer microscopy (OIM), and a surface profiler (i.e., Dektak), and of the layer compositions using x-ray diffraction (XRD), Auger electron spectroscopy, and Raman spectroscopy. Although these measurements can offer direct and accurate results to a certain extent, they still have some limitations in their practical use owing to the need for test samples and expensive measurement equipment.
In this paper, we propose and demonstrate a novel combination of measurement techniques for evaluating the dimensional and compositional changes of SAG MQW structures. This technique is based on C-V and I-V measurements (which are widely used for most PN-junction laser diode tests). Compared to previous measurement methods, this technique is very simple and cost-effective, and because the measurement is performed on a fully fabricated device without any existing damage, there is no need to take additional test samples for evaluation. Through this technique, the changes in the capacitance shown in the C-V curve, and in the voltage shown in the I-V curve, mainly correspond to variations of the layer thickness (according to a simple parallel-plate junction capacitor model) and bandgap energy (according to a PN junction model), respectively. To confirm the effectiveness of the proposed technique, we conducted two different evaluations. First, we fabricated an LD array employing various SAG MQW structures and measured its PL spectra. The measured data were analyzed and compared with the simulation results for the SAG MQW structure. In this way, we were able to estimate and evaluate the effect of the dimensional and compositional changes on the wavelength shift. Next, we conducted C-V and I-V measurements for the SAG MQW LD array, which we then used to extract the growth rate enhancement and bandgap energy for each channel. Based on this evaluation, we could compare the extracted results with the well-fitted simulation results from the photo-luminescence (PL) measurements. They both showed good agreement with the simulation results.
The rest of this paper is organized as follows. Experimental and theoretical analysis for the SAG MQW structure is described in Section 2. Section 3 provides the experimental results obtained from the proposed technique, along with comparisons between the measured and calculated results. Finally, Section 4 summarizes the findings of this research.
2. Experimental and theoretical analyses for SAG MQW structure
2.1 SAG structure and its PL measurement results
The epitaxial layers were grown using a lateral-flow-type metal-organic chemical vapor deposition (MOCVD). The layer stack before the SAG process consists of an n-InP buffer and a lattice-matched 30-nm thick n-doped outer separate confinement hetero-structure (SCH) layer with a band-gap wavelength of 1.08 μm. Figure 1 shows (a) top and (b) cross-sectional schematic views of the SiNx mask patterns used for the SAG process. Each mask pattern was aligned along the <110> direction on the (100) InP substrate, and was composed of a straight structure with a length of Lst and a tapered structure at both sides. The patterns appear to be bilaterally symmetric with unit cell period P, as shown in Fig. 1(b). The spacing between adjacent mask stripes with a width of Wm is the same as the opening width Wo. For this pattern, we introduced an additional parameter, M, to describe the relationship between Wm and Wo (i.e., Wm + Wo = M/2). In this experiment, to generate a different SAG effect on each channel in an array, Wo was adjusted for the given pattern parameters, such as P, M, Ltot, and Lst.
After the mask-patterning, InGaAsP layers for SAG were grown at a temperature of 630 °C and a pressure of 100 mbar as follows: a 10-nm thick outer SCH layer, a 20-nm thick inner SCH layer with a band-gap wavelength of 1.24 μm, 7-pair QWs (i.e., 0.6% compressively strained 6-nm thick wells of 1.62 μm, and 0.45% tensile-strained 7.5-nm thick barriers of 1.3 μm), a 20-nm thick inner SCH layer, and a 10-nm thick outer SCH layer. After the SAG process, room-temperature PL measurements were performed at the centers of the opened areas. Figures 2(a) and 2(b) show the normalized PL spectra and their peak-wavelength shifts with the deviation for different values of Wo. The pattern parameters P, M, Ltot, and Lst were 500 μm, 400 μm, 3 mm, and 2.4 mm, respectively. Because of the intensity fluctuation near the peak (over the wavelength range of 5 ~8 nm), the PL-peak wavelength of each channel (cross) was estimated by performing Gaussian-fitting. The results indicate that, when Wo is decreased to 100 μm, the peak-wavelength is red-shifted to about 83 nm.
Fig. 2 (a) Normalized PL spectra and (b) PL peak-wavelength shifts for mask patterns with different values of Wo.
After the removal of the SAG mask, a 30-nm thick p-doped outer SCH layer, a 100-nm thick p-InP residual cladding layer, a 20-nm thick p-InGaAsP etch-stop layer, a 2-μm thick p-InP upper cladding layer, and a 0.2-μm thick p + InGaAs layer were grown in sequence. Reverse-mesa ridge waveguide with a ridge-neck width of about 2 μm were fabricated [9]. After the typical LD fabrication processes, a 300-μm long LD array was implemented in a chip-bar form.
2.2 Theoretical analysis of SAG MQW structure
In principle, because no growth occurs on the dielectric mask patterns during the SAG process, the reactants (i.e., In and Ga) accumulated above the mask will have a higher concentration compared with those within the opened area. When growth rate enhancement R is defined as the growth rate in a selective growth area normalized by that in a planar area without the effect from the masks, the only parameter determining R is the diffusion coefficient, D/ks, which is the ratio of gas-phase diffusion coefficient D to the surface reaction rate ks. In and Ga have been reported to have D/ks values of 30–40 μm and 100–200 μm, respectively, and consequently will produce different R profiles along the lateral direction (i.e., the x-axis). For the given mask pattern parameters, the concentration profiles of In and Ga can be obtained by solving the well-known Laplace’s equation in 2D under the appropriate boundary conditions [4,5]. Figures 3(a) and 3(b) show the calculated R profiles for In and Ga along the x-axis (i.e., RIn and RGa) for a mask pattern of Wo = 100 μm, and their values at the center of the opened region (i.e., x = 250 μm) as a function of Wo, respectively. For this simulation, mask pattern parameters P and M were 500 and 400 μm, respectively, and the D/ks values for In and Ga were 35 and 190 μm. It appears that, as D/ks increases, the R profile becomes gradually flattened across the opened region with a reduced slope with respect to Wo.
Fig. 3 (a) RIn and RGa profiles along the x-axis at Wo = 100 μm and (b) their values at the center of the opened area (x = 250 μm) as a function of Wo.
For an In(x)Ga(1-x)As(y)P(1-y) material system, this difference between RIn and RGa will result in a change in the mole fraction x as well as an enhancement in the growth rate. Under the assumption of no changes in the group V precursors (i.e., constant mole fraction y) during the SAG process (this assumption is quite acceptable because the V/III ratio is very large under this growth condition), the growth rate enhancement and In mole fraction in an InGaAsP material, i.e., RInGaAsP and x′, can be expressed as
where x is the In mole fraction in a planar area without a SAG effect. For application to the SAG MQW structure, we calculated RInGaAsP and x′ for the well, barrier, and SCH materials independently with the assumption of the same (D/ks)In and (D/ks)Ga, and then obtained the first transition energy level difference in the QW (i.e., energy difference between the first quantized level of electrons in the conduction band and that of heavy holes in the valence band), i.e., Ee1,hh1, using the InGaAsP band parameters [10,11] and QW band-structure calculation [12]. For a comparison with the measured data shown in Fig. 2(b), Ee1,hh1 was converted into the bandgap wavelength, λbg,1,1, according to the energy-wavelength relation (i.e., λbg,1,1 [μm] = 1.2398/ Ee1,hh1[eV]), and its shift Δλbg,1,1 was then obtained by subtracting the initial value (i.e., bandgap wavelength without the SAG effect).Figures 4(a) and 4(b) show the measured PL peak-wavelength shifts (crosses with error bars) and calculated bandgap wavelength shifts Δλbg,1,1 (lines) as a function of Wo for different values of (D/ks)In and (D/ks)Ga, respectively. The calculated values are located within the error bars of measured data, and their results indicate that Δλbg,1,1 is susceptible to variations of (D/ks)In for a mask pattern with a relatively large Wo, and as Wo decreases, it is affected more and more by that of (D/ks)Ga. Under the condition of (D/ks)In = 35 μm and (D/ks)Ga = 190 μm, the calculated results agree well with the measured data. For this diffusion condition, we were able to obtain a growth rate enhancement for the SAG MQW structure, RMQW, by performing a weighted average of all SAG layers at x = 250 μm, as shown in Fig. 5. For this curve, RMQW was obtained as 1.672 at Wo = 100 μm.
Fig. 4 Calculated bandgap wavelength shifts (lines) as a function of Wo for different values of (a) (D/ks)In at (D/ks)Ga = 190 μm and (b) (D/ks)Ga at (D/ks)In = 35 μm. In both figures, the crosses with error bars denote the measured PL peak-wavelength shifts shown in Fig. 2(b).
Fig. 5 Calculated growth rate enhancement of SAG MQW structure RMQW at x = 250 μm under the conditions of (D/ks)In = 35 μm and (D/ks)Ga = 190 μm.
On the other hand, to examine the effect of dimensional and compositional changes on the bandgap wavelength shift in SAG MQW structures, we performed a simulation on the SAG MQW structure when considering only a single effect (i.e., either a thickness change without a compositional change, or a compositional change without a thickness change) independently. The sum of both dimensional wavelength shift Δλdim and compositional wavelength shift Δλcom were identical to the bandgap wavelength shift Δλbg,1,1. For the shift in energy level difference, ΔEe1,hh1, the same result was obtained. Figure 6 and Fig. 7 show the dimensional and compositional wavelength shifts (i.e., Δλdim (solid) and Δλcom (dashed lines)) and energy level difference shifts (i.e., ΔEdim (solid) and ΔEcom (dashed lines)) as a function of Wo for different values of (D/ks)In and (D/ks)Ga, respectively. From both figures, it is found that the changes of diffusion coefficients have a larger influence on the compositional shifts (i.e., Δλcom and ΔEcom) in this mask pattern. It appears that, as Wo decreases, Δλdim (or ΔEdim) is changed monotonically while Δλcom (or ΔEcom) is changed quadratically. This monotonic change in Δλdim (or ΔEdim) is quite reasonable because in this structure the well thickness is changed quadratically with Wo (see Fig. 5), and the energy level is nearly proportional to the square of the well thickness. The important thing to note here is that Δλdim has a dominant effect on the wavelength shift at a relatively large Wo, and as Wo decreases, the effect of Δλcom increases quickly; as a result, a cross point in both curves exists. For this SAG pattern, the cross point appears at a Wo of about 100 μm (or a normalized opening width No (2Wo/M) of 0.5, i.e., M = 400 μm), and at this point a total wavelength shift, Δλtot ( = Δλdim + ΔEcom), of about 83 nm was obtained.
Fig. 6 Dimensional and compositional wavelength shifts (Δλdim (solid) and Δλcom (dashed lines)) as functions of Wo for different values of (a) (D/ks)In at (D/ks)Ga = 190 μm and (b) (D/ks)Ga at (D/ks)In = 35 μm.
Fig. 7 Dimensional and compositional energy level difference shifts (i.e., ΔEdim (solid) and ΔEcom (dashed lines)) as a function of Wo for different values of (a) (D/ks)In at (D/ks)Ga = 190 μm and (b) (D/ks)Ga at (D/ks)In = 35 μm.
On the other hand, No and Δλtot at a cross point can be very important in examining the maximum available wavelength shift and channel wavelength controllability for the given SAG pattern parameters, and for designing SAG layers with various functions such as thickness and strain controls. In this simulation, we found that both No and Δλtot at a cross point depend highly on pattern parameter M at a given P. When M is decreased, No is increased and Δλtot is decreased. For example, for an M of 500 μm, No = 0.4 and Δλtot = 107.3 nm were obtained at a cross point, and for an M of 200 μm, No = 0.88 and Δλtot = 26 nm were obtained.
3. Experimental results from the proposed technique
Since the SAG MQW structure is used as an intrinsic layer in a P-i-N junction, a dimensional change of this intrinsic layer can be examined by measuring the C-V properties of the LD. To confirm this, a fabricated SAG MQW LD array was mounted on a copper-tungsten (CuW) metal optical bench, and C-V measurements were taken using a CV meter (Keithley 590 C-V analyzer). Figure 8 shows the measured C-V curves for the ten-channel SAG MQW LD array used. As the channel number was increased (i.e., Wo was decreased), the C-V curve was reduced. To extract the growth rate enhancement from this result, we introduced a simple junction capacitance model. For this model, each channel capacitance Ci can be expressed through εiA/di, where εi is the averaged dielectric constant of channel i, A is the area of the intrinsic region, and di is the total thickness of the channel intrinsic layer. From this model, the growth rate enhancement of channel i to reference channel r (i.e., r = 0, or no SAG), i.e., Ri,r (di/dr), can be expressed through εiCr/εrCi, where εi/εr can be obtained from calculating the refractive indices of the MQW SAG structure using the modified effective single-oscillator method (MESO) [10].
In this work, each channel capacitance was selected based on its maximum value to remove the effect of the space-charge region. The maximum capacitance of the LD without SAG C0_max was measured as 46.5 ± 2 pF, and the averaged refractive index of MQW n0 was calculated as 3.413. From the MESO calculation, because ε10/ε0 was obtained as 1.0053, we assumed εi/εr ≅ 1. On the other hand, as can be inferred from Fig. 3(a), since the growth rate enhancement of SAG MQW structure has a profile along the x-axis, it is effective to use the value averaged across an interval of x. Figure 9 shows the growth rate enhancement extracted from the C-V curve (crosses, Ri,0 = C0_max/Ci_max) as a function of Wo and the calculated results RMQW for different values of x interval (lines). As Wo decreases (i.e., the channel number increases), Ri,0 increases. These extracted results agree well with the calculated results as a whole. As a result, it was confirmed that this technique is very useful in evaluating the dimensional change for the SAG MQW structure. The increase of the extracted Ri,0 near a Wo of 130 μm can be explained through the reduction and broadening of the peak values in C, as shown in Fig. 8, which can be related to the change of built-in voltage and the non-uniform potential profile in the ridge waveguide [13]. At a relatively large Wo, the discrepancies between experimental points and calculated curves are shown. Such discrepancies can be controlled and reduced in the calculation when we use different diffusion coefficients for each SAG layer.
Fig. 9 Ri,0 as a function of Wo for SAG MQW LD array (crosses with error bars) and the calculated RMQW for different values of x interval under the conditions of (D/ks)In = 35 μm and (D/ks)Ga = 190 μm (lines).
On the other hand, the bandgap of the SAG MQW structure can be examined by measuring the I-V properties of the LD because the built-in voltage of diode Vbi is closely related to the bandgap of the intrinsic layers, Eg (i.e., Vbi = Eg/q - ϕ, where q is the electron charge and ϕ is the work function). To confirm this, I-V measurements were taken using an IV meter (HP 4156A precision semiconductor parameter analyzer). Figure 10 shows the measured I-V curves for the ten-channel SAG MQW LD array. It can be clearly seen that, as the channel number is increased, the turn-on voltage is decreased. To compare these results with the calculated results in Fig. 7(a) (red dashed line), the channel voltage was selected for the same current. In addition, a material bandgap change for the SAG MQW structure, ΔEg, MQW, was calculated by performing a weighted average of all SAG layers. Especially, this change includes the contribution of SCH layers, unlike ΔEcom. Figure 11 shows the channel voltage measured at a current of 0.2 ± 0.01 mA (crosses with error bars) as a function of Wo, ΔEcom calculated in Fig. 7(a), and ΔEg, MQW calculated at the condition of (D/ks)In = 35 μm and (D/ks)Ga = 190 μm. Both of calculation results have nearly identical shapes (i.e., in this mask pattern, all the SAG layers change in a similar proportion), and the measured results agree well with the calculated results. As a result, it was confirmed that this method is very useful in evaluating the compositional change for a SAG MQW structure.
Fig. 11 Measured channel voltage at a current of 0.2 mA (crosses) as a function of Wo and the calculated results ΔEcom (red dashed) and ΔEg, MQW (blue solid line).
4. Summary
We analyzed and examined the effects of the dimensional and compositional changes for a SAG MQW structure both theoretically and experimentally. To perform this work in a systemic way, we conducted two evaluations. First, a ten-channel SAG MQW LD array was fabricated and its PL spectra were measured. The measured PL-peak wavelength shifts were then compared with the simulation results for the SAG MQW structure. Under the condition of (D/ks)In = 35 μm and (D/ks)Ga = 190 μm, the calculated results agreed well with the measured data. For this diffusion condition, we were able to obtain a growth rate enhancement of 1.672 for a mask pattern with Wo = 100 μm. In the theoretical analysis, it was found that Δλdim has a dominant effect on the wavelength shift at a relatively large Wo, and as Wo decreases, the effect of Δλcom increases quickly; as a result, a cross point of both curves exists. For this SAG pattern, the cross point appears at a No of 0.5, where a total wavelength shift, Δλtot, of about 83 nm was obtained. Both No and Δλtot at this cross point depend highly on pattern parameter M at a given P. When M is decreased, No is increased and Δλtot is decreased. Next, we took C-V and I-V measurements of the SAG MQW LD array and extracted the growth rate enhancement and bandgap energy from these measurements according to a simple parallel-plate junction capacitor model and PN junction model, respectively. Through this work, the extracted results were compared with the well-fitted simulation results from the PL-peak wavelength shifts. They both show good agreement with the simulation results. As a result, it was confirmed that this technique is very useful in evaluating the dimensional and compositional change for the SAG MQW structure.
Acknowledgment
This work was supported by the “Energy Efficient Power Semiconductor Technology for Next Generation Data Center” IT R&D Project (No. 10038766) and “Deep-UV Laser Diode development for medical & precision processing applications (No. N0000482).
References and links
1. A. K. Dutta, N. K. Dutta, and M. Fujiwara, WDM Technologies: Active Optical Components (Academic, 2002), Chap. 10.
2. M. Aoki, M. Suzuki, H. Sano, T. Kawano, T. Ido, T. Taniwatari, K. Uomi, and A. Takai, “InGaAs/InGaAsP MQW electroabsorption modulator integrated with a DFB laser fabricated by band-gap energy control selective area MOCVD,” IEEE J. Quantum Electron. 29(6), 2088–2096 (1993). [CrossRef]
3. T. Fujii, M. Ekawa, and S. Yamzaki, “Growth pressure dependence of selective area metalorganic vapor phase epitaxy on planar patterned substrates,” J. Cryst. Growth 156(1-2), 59–66 (1995). [CrossRef]
4. J. H. Song, K. Kim, Y. A. Leem, H. J. Kim, and G. Kim, “Strain-controlled selective-area growth of InGaAsP films on InP,” Jpn. J. Appl. Phys. 46(33), L783–L785 (2007). [CrossRef]
5. N. Dupuis, J. Décobert, P.-Y. Lagrée, N. Lagay, F. Poingt, C. Kazmierski, A. Ramdane, and A. Ougazzaden, “Mask pattern interference in AlGaInAs selective area metal-organic vapor-phase epitaxy: Experimental and modeling analysis,” J. Appl. Phys. 103(11), 113113 (2008). [CrossRef]
6. Y. Morishima, J. Yaguchi, A. Mukai, T. Ohgoh, and H. Asano, “Longitudinal bandgap modulated broadband (>150 nm) InGaAs/GaAs MQWs superluminescent diodes by selective area MOVPE growth,” Electron. Lett. 45(10), 521–522 (2009). [CrossRef]
7. Y. Shimogaki, M. Sugiyama, and Y. Nakano, “Selective area MOVPE of InGaAsP and InGaN systems as process analytical and design tools for OEICs,” IPRM2013, May 19 - 23, 2013, Kobe, Japan MoD4–1. [CrossRef]
8. C. Zhang, S. Liang, H. Zhu, L. Han, and W. Wang, “Multichannel DFB laser arrays fabricated by upper SCH layer SAG technique,” IEEE J. Quantum Electron. 50(2), 92–97 (2014). [CrossRef]
9. O. K. Kwon, Y. A. Leem, Y. T. Han, C. W. Lee, K. S. Kim, and S. H. Oh, “A 10 × 10 Gb/s DFB laser diode array fabricated using a SAG technique,” Opt. Express 22(8), 9073–9080 (2014). [CrossRef] [PubMed]
10. Y. Suematsu and A. R. Adams, Handbook of Semiconductor Lasers and Photonic Integrated Circuits (Chapman & Hall, 1994), Chap. 3.
11. I. Vurgaftman, J. R. Meyer, and L. R. Ram-Mohan, “Band parameters for III-V compound semiconductors and their alloys,” J. Appl. Phys. 89(11), 5815–5875 (2001). [CrossRef]
12. S. L. Chuang, “Efficient band-structure calculations of strained quantum wells,” Phys. Rev. B Condens. Matter 43(12), 9649–9661 (1991). [CrossRef] [PubMed]
13. D. Ban, E. H. Sargent, K. Hinzer, St. J. Dixon-Warren, A. J. SpringThorpe, and J. K. White, “Direct observation of lateral current spreading in ridge waveguide lasers using scanning voltage microscopy,” Appl. Phys. Lett. 82(23), 4166–4168 (2003). [CrossRef]
