We present a practical experimental design for performing photoluminescence (PL) and photoreflectance (PR) measurements of semiconductors with only one PL spectroscopic system. The measurement setup is more cost efficient than typical PL-plus-PR systems. The design of the experimental setup of the PL–PR system is described in detail. Measurements of two actual device structures, a high-electron-mobility transistor (HEMT) and a double heterojunction-bipolar transistor (DHBT), are carried out by using this design. The experimental PL and PR spectra of the HEMT device, as well as polarized-photoreflectance (PPR) spectra of the DHBT structure, are analyzed in detailed and discussed. The experimental analyses demonstrate the well-behaved performance of this PL–PR design.
©2005 Optical Society of America
As a result of rapid development in semiconductor manufacture, optoelectronic devices such as light-emitting diodes (LEDs) and laser diodes (LDs), microwave devices such as high-electron-mobility transistors (HEMTs) and heterojunction bipolar transistors (HBTs), and optical detectors such as as p-i-n diodes and avalanche photodiodes (APDs) have been successfully fabricated by various expitaxial growth techniques . To improve device and material qualities, optical inspection of optoelectronic devices and materials by spectral characterization of luminescence, transmittance, and reflectance is more effective during manufacture. The line shape of transition features in the optical spectra of luminescence, transmittance, and reflectance can be analyzed in detail to yield related information about crystalline, energy state, and structural properties of optoelectronic materials and devices.
Among the various spectral characterization techniques, photoluminescence (PL) and photoreflectance (PR) spectroscopy are usually the most important characterization tools for optical investigation of semiconductors. Both PL and PR techniques offer a contactless and nondestructive probing method for studying optoelectronic materials and devices, and different kinds of optical information can be acquired by the two measurement techniques. PL is a popular characterization tool for studying optical radiation emitted from a material or a device structure upon external light excitation. PL emissions may result from various radiative recombination mechanisms, such as direct bandgap, intersubband energies, excitonic states, defect states, and impurity level mechanisms. The physical mechanism of carrier generation and recombination for PL makes this technique more powerful for studying the luminescent property of materials near and below the fundamental edge .
For the characterization of interband transitions above the band edge, modulation spectroscopy methods, such as PR, piezoreflectance (PzR), and thermoreflectance (TR) spectroscopy, have proved to be excellent for studying direct interband transitions seen in the derivative line shapes of optical spectra [3–6]. In a PR experiment the electric field is modulated in a sample through the creation of electron–hole pairs by a pump source (laser or other light source) chopped at some frequency. The periodic photoperturbation of the sample will result in a derivativelike spectral line feature that occurs at the critical-point transition of the reflectance spectrum. The feature’s derivativelike nature suppresses uninteresting background effects and emphasizes structures localized in the energy region near direct interband transitions of semiconductors. The well-resolved property of the PR spectra makes this experimental technique very popular for use in the characterization of semiconductor and actual device structures [7, 8]. Although the PL and PR techniques are powerful and can also be used as complementary tools for the optical characterization of semiconductors, each measurement system is expensive, and the cost can double when the two systems are built simultaneously.
In this paper we present a practical design for integrating PL and PR spectral measurements in one PL spectroscopic system. The measurement setup is cost efficient compared with the usual PL plus PR systems. In addition this PL–PR design has an enhanced ability to increase the signal-to-noise (S/N) ratio of signal detection when comparing with conventional PR experiments. The design of the experimental setup for the PL–PR system is described. Measurements of a GaAs/InGaAs grading-channel HEMT and an InGaP/InGaAsN/GaAs double heterojunction-bipolar transistor (DHBT) are carried out with this measurement design. The experimental PL and PR spectra of the HEMT device as well as the polarized-photoreflectance (PPR) spectra of the DHBT structure are analyzed and discussed. The intersubband energies, Fermi-level energy, and sheet carrier density of a two-dimensional-electron gas (2DEG) and the built-in electric field of the graded-channel HEMT are evaluated. The interband transitions and built-in electric fields of the collector-base and emitter-base junctions for the DHBT device are characterized. Well-resolved and easily analyzed properties of the experimental spectra for the HEMT and DHBT devices demonstrate the well-behaved performance of this PL–PR design.
2. Measurement design
A representative scheme of the experimental setup of the PL–PR measurement system is shown in Fig. 1. A Triax 320 imaging spectrometer equipped with three gratings of 600, 1200, and 2400 groves/mm acted as the optical dispersion unit. Two detecting elements, a photomultiplier tube (PMT) and a (TE) thermoelectric-cooled InGaAs detector were attached at the outside of two exit slits. The PMT and TE-cooled InGaAs photodetectors cover a wide measured spectral range, from 190 to 1650 nm. The operating temperature of the TE-cooled InGaAs detector was controlled by a TE-cooled driver with maximum cooling of -50°C. A data acquisition (DAQ) unit in the PL–PR system is capable of integrating the experimental details, such as implementing the analog-to-digital (A/D) conversion of signal detection from the PMT or the InGaAs detector, supplying high voltage to the PMT, and transferring the digital data to a personal computer (PC). The DAQ unit communicated with the PC via an RS-232 bus. To improve the S/N ratio of signal detection, ac phase-sensitive detection (PSD) is implemented by using a lock-in amplifier and an optical chopper. The chopper is utilized to cut the laser light into an ac-type pumping beam for both PL and PR measurements and to provide a reference signal for the chopping frequency for the lock-in amplifier. For PL measurements of the HEMT device, a frequency-doubled Nd–YAG laser (peak wavelength λp =532 nm) with an average output power of 100 mW was used as the pump light source. The PL emissions from the sample were collected and focused onto the spectrometer via a planar–convex lens. The PL measurements were performed by programming the control of the spectrometer and DAQ unit via the RS-232 bus connections.
For PR measurements, a dc 3-V tungsten halogen lamp acted as the white-light source. The white light was focused onto the sample via a planar–convex lens. The reflected light from the sample was collected and focused onto an imaging spectrometer by another planar–convex lens. A He–Ne laser (λp =632.8 nm) together with a neutral-density filter (ND 2.0) acted as the modulation light source of the HEMT sample. A 532-nm Nd–YAG laser combined with an ND 3.0 (0.1%) neutral-density filter was used for modulation of the DHBT device. A TE-cooled InGaAs detector operated at ~0°C was utilized for optical detection. A pair of visible-dichroic-sheet polarizers was utilized to perform the polarization-dependent measurements of the DHBT. The PR spectral measurements were implemented by first measuring and recording the spectral data of the white-light source reflected from the sample surface (i.e., R) and then detecting the change in reflection of the sample (i.e., ΔR) after photo perturbations. Note that this PL–PR system can avoid background-light interference arising from the pump laser because of the dispersion property of the imaging spectrometer, which prevents detection of the laser wavelength. It has an enhanced S/N ratio of signal detection with respect to a conventional PR system .
3. Experimental results and discussion
The functional performance of the PL–PR system design was tested by using two selected samples of a graded-channel GaAs/InxGa1−xAs HEMT and an InGaP/InGaAsN/GaAs DHBT at 300 K. Sample specifications and measurement conditions for testing the PL–PR design are summarized in Table 1. The graded-channel HEMT (sample 1) was grown by a computer-controlled LP-MOCVD . The epilayers of sample 1 were grown on a semi-insulating (100)-oriented GaAs substrate and were followed by an undoped 1 µm GaAs buffer layer, a 90 Å undoped graded InxGa1−xAs layer, an 80 Å undoped GaAs layer, a δ-doped GaAs layer, and finally a 400 Å undoped GaAs cap layer. The In composition of the channel layer in sample 1 was varied from x=0.15 to x=0.25. Figure 2 shows the representative energy band scheme of the graded-channel HEMT. A 2DEG was formed in the channel layer as a result of the electrons’ spilling over the band discontinuity from the N+δ-doped GaAs layer into the InxGa1−xAs grading channel, which also shifted the Fermi level above the conduction band edge inside the channel well.
PR and PL measurements have proved to be very powerful in the characterization of pseudomorphic HEMT structures . Figure 3 shows the experimental PR (solid curve) and PL (dashed curve) spectra of the graded-channel HEMT obtained with the PL–PR system design. There are a lot of transition features as well as a prominent spectral oscillation observed in the PR spectrum of the graded-channel HEMT. The transition features (denoted A, B, C, D, and E) at the lower energy side of the PR spectrum are closely related to the well intersubband transitions of the grading-channel layer. The spectral oscillations at the higher-energy side are the so-called Franz–Keldysh oscillations (FKOs), which frequently have been used to determine the built-in electric field inside the semiconductor device . Electromodulated line shapes of the PR spectra for semiconductors are expressed as [4, 5]
where α and β are the Seraphin coefficients and Δε 1, Δε 2 are, respectively, the modulated real and imaginary components of the complex dielectric function ε=ε 1+iε 2. For bound states, such as the intersubband transitions of a quantum well, it has been shown that electromodulation yields first-derivative spectroscopy . In the presence of a sufficiently dense 2DEG, the intersubband absorption function will be a broadened two-dimensional density of states multiplied by a Fermi level-filling function. From the one-electron theory, the imaginary part of the dielectric function can be expressed as 
where Dj is the amplitude of the jth feature, E is the photon energy, Γ j is the broadening parameter, and Ej (mn) is the inter-subband energy given by Ej (mn)=. and are the energies of the mth conduction and nth valence subband, respectively, referred to the jth feature. The Fermi function in Eq. (2) can be expressed as
where EF is the Fermi energy. The parameter λ in Eq. (3) is given by
where and are the electron and the in-plane heavy-hole effective masses, respectively, in units of the free electron mass.
The intersubband transitions of the PR spectrum in Fig. 3 can be analyzed by fitting the PR spectrum to Eqs. (1) and (2). The effective masses of the electron and in-plane heavy hole for the channel layer are assumed to be 0.058 and 0.326. The fitting parameters are Dj, Ej (mn), Γ j , and Ē j (m), respectively. The obtained intersubband energies of the well transitions Ej (mn) in Fig. 3 are indicated with arrows and are listed in Table 2. It must be noted that the line shapes of the PR spectrum of the graded-channel HEMT are uncommon for modulation spectroscopy from a quantum-well system in which 11H is usually the dominant feature. In Fig. 3 the feature associated with 11H (feature A) is small in comparison with the other transitions. The reasons for the weak 11H transition are the GaAs/InxGa1−xAs conduction band discontinuity and that the δ-doped layer introduces a large density of 2DEG into the InGaAs grading well. The Fermi level is lifted over the first conduction subband, and hence this subband is almost fully occupied by 2DEG. Consequently, the strength of the transition from the first valence subband to the first conduction subband is reduced dramatically, resulting in the weaker 11H feature. More evidence of the 11H feature in the graded-channel HEMT can be observed from the PL spectrum shown in Fig. 3. The maximum luminescent intensity of a quantum well for the PL measurement generally occurs in the ground-state recombination. In the PL spectrum of Fig. 3 the energy location of feature A has nearly the largest spectral amplitude of luminescence, which can be attributed to intersubband recombination from the 11H transition. The luminescence feature located at ~1.3 eV is the 21H transition arising from the recombination of the second conduction subband with the first valence subband (see Fig. 2). The energy value of the Fermi-level location EF can also be determined to be (6±2) meV for the graded-channel HEMT. The sheet density Ns , for the case of a broadened steplike 2D density of electron states [ρ2D(E,Γ)], can be expressed as 
where En is the photon energy of the nth extremum, E 0 is the bandgap, and χ is an arbitrary phase factor. The electro-optic energy ħΘ is given by
where F is the electric field and µ is given by
Here and are the effective masses of the electron and the hole, respectively, in units of the free electron mass. The relevant electron and heavy-hole effective masses for GaAs are 0.067 and 0.34, respectively. From the analysis of the FKOs shown in Fig. 3, the built-in electric field of the graded-channel HEMT at the InGaAs/GaAs interface is determined to be F=155±5 kV/cm.
The functional performance of the PL–PR system was also tested with PPR measurements of an InGaP/InGaAsN/GaAs DHBT. The PPR experiments were carried out with optical polarizations along  and [11̄0]. The DHBT was grown on a semi-insulating (001) GaAs substrate by metalorganic epitaxial chemical vapor deposition (MOCVD). The structure of the device’s epilayers is shown in Fig. 4. The DHBT comprised 5000Å n+ GaAs followed by 5000Å n- GaAs as the collector layer, 50 nm n-type InGaP as the emitter, and a 70 nm InGaAsN layer as the base. The contact cap layer was made of InGaAs. PPR spectra of the DHBT sample are shown in Fig. 5. The solid curve is the PPR spectrum of E‖ polarization, and the dashed curve is that of E‖[11̄0]. The polarization dependence of the PPR spectra at emitter region (InGaP) indicates that the E1 transition measured by E‖ polarization is lower in energy than the E2 transition of E‖[11̄0] polarization. The transition energies of E1 and E2 can be analyzed by using Eq. (1) and are determined to be E1=1.843±0.008 and E2=1.856±0.008 eV, respectively. The optical anisotropy in the emitter region is caused by the ordering dependency of the In–Ga sublattice inside the InGaP single crystal . For a more ordered InGaP crystal, the energy difference of E2-E1 is usually higher than a disordered one . There are also two obvious FKOs (denoted FKO1 and FKO2) found in both PPR spectra of Fig. 5. FKO1 and FKO2 are closely related to the built-in electric fields at the collector and emitter interfaces. The built-in electric fields F collector and F emitter of the DHBT can also be evaluated from Eqs. (8)–(10) by taking into account the electron and heavy-hole effective masses of 0.067 and 0.34 for GaAs (collector) and 0.118 and 0.66 for InGaP (emitter) , respectively. According to the evaluations, the electric fields in the collector and the emitter regions are determined to be F collector=74±6 kV/cm and F emitter=160±8 kV/cm by neglecting any dependence of the masses on ordering. The feature at ~1.2eV (EB) in Fig. 5 is attributed to the transition of InGaAsN from the base region, while the other transition at ~1.36 eV may come from the conducting cap layer of InGaAs. The transition feature at ~1.42 eV (EC) originates from the collector region of the GaAs layer.
In conclusion, a practical PL–PR experimental design for performing photoluminescence and photoreflectance measurements of semiconductors by using only one PL spectroscopic system is demonstrated. The functional performance of the measurement design was tested on two semiconductor devices, a GaAs/InGaAs grading-channel HEMT and an InGaP/InGaAsN/GaAs DHBT. From analyses of the PR and PL spectra of the graded-channel HEMT, the intersubband energies, Fermi-level location, and 2D sheet density of the graded InxGa1−xAs channel well were determined. The InGaP/InGaAsN/GaAs DHBT was characterized by using PPR measurements with polarizations along  and [11̄0]. The experimental results clearly show two FKOs and some transition features present in the PPR spectra of the DHBT. From analyses of the PPR spectra, the built-in electric fields near the emitter and collector regions are evaluated, and the transition energies in the base, collector, and emitter layers are determined. The well-resolved and easily analyzed properties of the experimental PR and PL spectra of the HEMT and DHBT devices indicate the well-behaved function of this PL–PR design.
The authors would like to acknowledge the project fund supported from the National Science Council of Taiwan under the grant No. NSC 93-2215-E-259-002.
References and links
1. P. Bhattacharya, “Elemental and compound semiconductors,” in Semiconductor Optoelectronic Devices, 2nd ed. (Prentice-Hall, London, 1997), Chap. 1, pp. 2–58.
2. D. K. Schroder, “Optical characterization,” in Semiconductor Material and Device Characterization1st ed. (Wiley, New York, 1990), Chap. 9, pp. 490–494
3. M. Cardona, Modulation Spectroscopy (Academic, New York, 1969).
4. F. H. Pollak and H. Shen, “Modulation spectroscopy of semiconductors: bulk/thin film, microstructures, surfaces/interfaces and devices,” Mater. Sci. Eng. R10, 275–374 (1993).
5. D. E. Aspnes, “Modulation spectroscopy/electric field effects on the dielectric function of semiconductors,” in Handbook on Semiconductors, Vol. 2., M. Balkanski, ed. (North Holland, New York, 1980), p. 109.
6. C. H. Ho, “Optical study of the structural change in ReS2 single crystals using polarized thermoreflectance spectroscopy,” Opt. Express 13, 8–19 (2005), http://www.opticsexpress.org/abstract.cfm?URI=OPEX-13-1-8. [CrossRef] [PubMed]
7. C. H. Ho, P. C. Yen, Y. S. Huang, and K. K. Tiong, “Photoreflectance study of the excitonic transitions of rhenium disulphide layer compounds,” Phys. Rev. B 66, 245207 (2002). [CrossRef]
8. D. Y. Lin, F. C. Lin, Y. S. Huang, H. Qiang, F. H. Pollak, D. L. Mathine, and G. N. Maracas, “Piezoreflectance and photoreflectance study of GaAs/AlGaAs digital alloy compositional graded structures” J. Appl. Phys. 79, 460–466 (1996). [CrossRef]
9. W. C. Hsu, C. M. Chen, and R. T. Hsu, “A δ-doped GaAs/graded InxGa1-xAs pseudomorphic structure grown by low-pressure metal organic chemical vapor deposition,” Appl. Phys. Lett. 59, 1075–1077 (1991). [CrossRef]
10. D. Y. Lin, S. H. Liang, Y. S. Huang, K. K. Tiong, F. H. Pollak, and K. R. Evans, “Room-temperature photoreflectance and photoluminescence characterization of the AlGaAs/InGaAs/GaAs pseudomorphic high electron mobility transistor structure with varied quantum well compositional profiles,” J. Appl. Phys. 85, 8235–8241 (1999). [CrossRef]
11. Y. Yin, H. Qiang, F. H. Pollak, D. C. Streit, and M. Wojtowicz, “Two-dimensional electron gas effects in the electromodulation spectra of a pseudomorphic Ga0.78Al0.22As/In0.21Ga0.79As/GaAs modulation-doped quantum well structure,” Appl. Phys. Lett. 61, 1579–1581 (1992). [CrossRef]
12. H. Shen and F. H. Pollak, “Generalized Franz-Keldysh theory of electromodulation,” Phys. Rev. B 42, 7097–7102 (1990). [CrossRef]
13. C. J. Lin, Y. S. Huang, N. Y. Li, P. W. Li, and K. K. Tiong, “Polarized-photoreflectance characterization of an InGaP/InGaAsN/GaAs NpN double-heterojunction bipolar transistor structure,” J. Appl. Phys. 90, 4565–4569 (2001). [CrossRef]
14. A. Lindell, M. Pessa, A. Salokatve, F. Bernardini, and M. Paalanen, “Band offsets at the GaInP/GaAs heterojunction,” J. Appl. Phys. 82, 3374–3380 (1997). [CrossRef]