Open Access
Add to BibSonomy
Abstract
Optical beam-induced current (OBIC) mapping is widely used to characterize semiconductor lasers, particularly for failure analysis, in which the reliability has been a critical issue to be resolved spectrally and temporally. OBIC microscopy is advantageous for its non-invasiveness, when compared with electron beam-induced current (EBIC) microscopy. However, for high-speed devices, conventional OBIC methods may be limited in observing the spectral responses adequately. In this work, we present a modified OBIC microscopy based on a tunable ultrafast laser, to address the need for spectral resolving for precision failure spot analysis in vertical-cavity surface-emitting laser (VCSEL) diodes. The spectral OBIC response of VCSEL diodes is investigated by varying the irradiation wavelengths. Importantly, the ultrafast mode-locked laser provides broadband wavelength range to investigate photocurrent responses of the VCSELs sample. Specifically, the OBIC, electroluminescence (EL) detection, and the reflectance of the normal and the electrostatic discharge (ESD) damaged VCSELs are compared. We have found the ESD damaged VCSELs showing a redshifted spectral response.
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Vertical-cavity surface-emitting lasers (VCSELs) have been widely adopted in short-reach (<300 m) transceiver and commercial sensing applications, attributing to its many significant advantages, including low power consumption, easy assembling, cost-effectiveness…etc. [1–4]. For the past years, many researches have focused on the study of reliability and failure analysis of VCSELs [5,6]. However, it is hard to determine failure spot location with the existing techniques such as current-voltage (I-V) characteristics and electroluminescence (EL) measurements. Thus, optical imaging with high spatial resolution can help to define the defect spot location to further check the failure root cause of VCSEL devices. Optical beam-induced current (OBIC) microscopy has been a novel and non-destructive imaging technique for inspecting semiconductor devices for failure analysis that is complimentary to EBIC [7]. The fundamental of OBIC is to detect electrical signal while irradiating a device with a focused and scanning light beam. The mapping with the optically induced electrical signal would show the device’s active area. If spectrally resolved absorption is implemented, the energy band structure of the devices may also be revealed. Direct current (DC) OBIC has found extensive applications in characterizing semiconductor devices and integrated circuits for failure analysis [8–14]. Furthermore, OBIC is able to probe the internal carrier dynamics of semiconductors by time-resolved measurements, which is usually implemented by combining a pulsed laser together with high frequency electronics for signal analysis. Ultrafast pulses from a mode-locked laser would not only generate extremely broad bandwidth for high frequency investigation, but also form frequency comb [15] that greatly facilitates electrical spectrum analysis. Previously, we have successfully developed two-photon (2-p) OBIC and radio frequency (RF) OBIC to investigate GaN LED [16] and PIN-based photodiode, respectively [17]. The spatial defects or anomalies in the active region of GaN LED are particularly sensitive to 2-p OBIC, while the uniformity of the temporal response of the active region of the PIN photodiode is clearly revealed by RF OBIC. In recent years, OBIC technique was applied to wide bandgap semiconductors with two different laser sources to determine minority carrier lifetime and ionization rates [18]. In addition, dynamic OBIC mapping with post-processing scheme also reveals the information of defect location [19]. Practically, all photosensitive devices can be investigated by OBIC methods. Additionally, there are various imaging techniques such as light-induced voltage alteration (LIVA) [20], optical beam induced resistance change (OBIRCH) [21] and thermally induced voltage alteration (TIVA) [22], which mainly uses laser beam to investigate the semiconductor devices. Among these scanning beam-based optical imaging techniques, OBIC system design leads greatly reduced instrumentation complexity. In this study, we have combined OBIC modality, spectrally-resolved irradiation analysis, to investigate the aluminum gallium arsenide (AlGaAs) based VCSEL diodes [23]. Note that VCSEL diodes are strongly affected by electrostatic discharge (ESD) and electrical overstress (EOS), which are the most common issues in reliability [24]. ESD is a transient discharge of static charge between two objects at different electrostatic potentials, either through direct contact or an induced electric field. An ESD damaged device may continue to function, albeit with compromised performance, as not all the active region is affected. The ESD induced damage is called the latent defect. The ESD damage can be analyzed easily above a certain threshold; however, low-level ESD is challenging to detect. The ESD damage significantly shortens the lifetime of the affected device. It can take place during the manufacturing and handling. EOS, on the other hand, would tend to affect the whole device and thus can be easily screened with common overall (global) device measurement without resorting to spatially resolved characterization.
Here, we present spectrally resolved OBIC imaging technique on ESD induced defects under various testing models, including human body model (HBM) and machine model (MM), with normal VCSELs as a comparison [25,26]. The main purpose of this study is to precisely define the failure location, not to screen out the potential fail device in production. The advantages of OBIC allow precise defining of the damaged locations. OBIC imaging can find out the damaged region and define the failure reason which kind of damage, during handling or bad structure design.
2. Experiments
2.1 Working principles of OBIC and EL
The working principle of OBIC is straightforward. The scanning incident beam is irradiating the active region of a device. The electron-hole pairs thus generated in the active region, also known as the space charge region, are separated by the built-in electric field in the junction and then collected by the electrodes to form photocurrent and serve as the OBIC signal. For light emitting devices, electroluminescence (EL) provides a complementary characterization, in which a forward bias is applied to the active region. The recombination of carriers would generate emitted photons and indicate the conditions of the active region.
2.2 Materials and devices
We have investigated the multi-mode oxide-VCSELs with GaAs/AlGaAs quantum wells emitting at ∼ 850 nm sandwiched between the P-type and the N-type distributed Bragg reflectors (DBR). The MOVPE-grown laser structures were fabricated into deep-etched mesas followed by the formation of oxide apertures to confine the photons and electrons. The schematic of the VCSEL is shown in Fig. 1(a). Finally, VCSELs were mounted on TO headers for test. There are three types of VCSEL diodes prepared for our experiments, namely, the VCSEL without any treatment (normal), the VCSEL with ESD induced defects under machine model discharge (ESD-MM), and the VCSEL with ESD induced defects under human body model discharge (ESD-HBM). The diameter of the tested VCSEL devices (packaged) is 4.7 mm. The diameter of the active area is approximately 10 µm. The lasing threshold current of our VCSEL diodes is approximately 0.8 mA at room temperature.
Fig. 1. (a) The schematic of the sectional structure of oxide-confined VCSEL. QWAL: quantum well active layer; OCL: oxide confined layer. (b) Reverse (I-V) curve of the tested three VCSELs (normal, HBM and MM).
Figure 1(b) shows the measured reverse current voltage (I-V) curves of the three VCSELs (normal, HBM and MM). The test voltages for HBM and MM are −180 V and −40 V, respectively. Although we can observe significant changes in the I-V characteristics, the I-V measurement can only reveal the failure of VCSELs. For comparison, OBIC can reveal the defect locations with excellent imaging contrast, which greatly facilitates the manufacturing in performing cross-sectional TEM imaging to investigate the failure root causes of the VCSELs.
2.3 Experimental setup
The schematic of the experimental setup for the spectrally resolved OBIC and EL detection is shown in Fig. 2. The OBIC microscopy setup is based on an inverted confocal laser scanning microscope (IX71, Olympus, Japan). A Ti:sapphire femtosecond laser (Mira 900F, Coherent Inc., USA) pumped by a frequency-doubled solid state laser (Sprout-D12W, Lighthouse, USA) is used as the excitation source, with the repetition rate of ∼76 MHz and pulse width ∼200 fs. The wavelengths of the Ti:sapphire laser can be tuned from 780 to 900 nm to acquire spectral resolved OBIC images. The average power of the excitation laser on the surface of the sample is set at 0.6 mW. The energy per pulse of the excitation beam is 7.9 pJ. An air spaced 100X objective lens (UMPlanFl, 100x, 0.95 NA, Olympus) is used to achieve high spatial resolution for OBIC imaging. The beam spot diameter is approximately∼ 1-1.15 µm on the surface of the VCSELs. Note that the illumination wavelength of the excitation beam is tuned from 700-900 nm with 5 nm increment to acquire spectral resolved OBIC images. The spot size would slightly change with the incident wavelengths accordingly. The output photocurrent signal can be detected to reconstruct an OBIC image. For spectrally resolved OBIC, the signal is pre-conditioned by a low noise voltage preamplifier (SR560, Stanford Research System, USA) before being fed into the synchronized analog-to-digital (A/D) converter of the commercially available confocal galvano-mirror scanner (FV300, Olympus, Japan). The reflection images are taken by a photomultiplier tube. The VCSEL samples are point-scanned with the scan size 512×512 pixels for a field of view approximately 22 × 22 µm2, which corresponds to a scan step of ∼ 0.04 µm (beyond the optical resolution). The acquisition time of an image is approximately 1.6 s. So, the pixel dwell time is then 6 µs.
Fig. 2. The schematic of the experimental design for spectral resolved OBIC imaging and electroluminescence measurement. A wavelength-tunable ultrafast Ti:sapphire laser is used as the excitation light source. The two input channels allow simultaneous acquisition of OBIC and reflectance images. LSU: laser scanning unit; PMT: photomultiplier tube; DM: dichromatic mirror; M: mirror; GM: galvano-mirror; L: lens; OF: optical fiber; BS: beam splitter; NA: numerical aperture; VCSEL: vertical-cavity surface-emitting laser.
For EL detection, a source meter (Keithley2400, Tektronix, USA) is used as the DC power supply to light up VCSELs. The output optical signal is then detected by a spectrometer (QE65000, Ocean Optics, USA). The customized OBIC microscopy system is flexible and capable of acquiring OBIC and confocal reflectance images simultaneously.
3. Results and discussion
3.1 OBIC imaging
The low magnification confocal reflectance and OBIC images of three categories of samples are shown in Figs. 3(a)–3(c). It is not easy to identify any defects from the reflectance images. Note that the VCSELs are not biased during the measurement.
Fig. 3. The confocal reflectance and the corresponding OBIC images of (a) normal VCSEL, (b) ESD-HBM VCSEL, and (c) ESD-MM VCSEL. From the left to the right column, the images are presented as the reflectance, the OBIC, and the merged images of both, respectively. The scale bar is 20 µm.
3.2 Spectrally resolved OBIC imaging
The spectrally resolved OBIC images of normal VCSELs are shown in Fig. 4. The OBIC images are obtained by tuning the incident wavelengths at 5 nm interval from 780 to 900 nm. The photocurrent signal is conditioned by a voltage preamplifier before being fed into the A/D converter for mapping. The OBIC images obtained are presented by color coding. As shown, one can clearly identify the OBIC active region, which is confined by an oxide layer, thus resulted in a round shape pattern. In normal VCSELs, the photocurrent distribution in the active region is very uniform. The maximum intensity of photocurrent appears when the incident wavelength is at 780 nm. When the incident wavelength is longer than 860 nm, there is no photocurrent response.
Fig. 4. The spectrally resolved OBIC images of normal VCSEL diode presented by color coding. These images are acquired when the incident wavelength from the laser is tuned from 780 nm to 900 nm with 5 nm increment. The excitation power on the sample is controlled at approximately 0.6 mW. The acquisition time of a single OBIC image with 512×512 pixels is approximately 1.6 secs. The scale bar is 5 µm. The color bar besides the OBIC images represent photocurrent intensity in arbitrary units, where the 12-bit dynamic range (0-4095) of the A/D converter of the scanning platform is taken full advantage of.
Correspondingly, the spectrally resolved OBIC images of ESD-MM VCSEL are shown in Fig. 5. Note that there are spots with weaker photocurrent signal in the color coded images. Unlike in normal VCSELs, the ESD affected sample also shows photocurrent response for incident wavelengths longer than 860 nm, indicating either the leakage of the distributed Bragg reflectors (DBR) or the shift of the absorption bandgap.
Fig. 5. The spectrally resolved OBIC images of ESD-MM VCSELs presented by color coding. The image acquisition conditions are the same as in (Fig. 4.) Note that the ESD affected sample exhibits spots with weaker photocurrent response. Unlike in normal VCSELs, the ESD affected sample shows enhanced photocurrent response for incident wavelengths longer than 860 nm. The scale bar is 5 µm. The color bar besides the OBIC images represent photocurrent intensity in arbitrary units, where the 12-bit dynamic range (0-4095) of the A/D converter of the scanning platform is taken full advantage of.
Similarly, the spectrally resolved OBIC images of ESD-HBM VCSEL are shown in Fig. 6. There are two defective spots in the active area, a large one in the edge and a small close to the center. Interestingly, the central absorption wavelengths of these two spots are different, though both exhibit redshift. The HBM induced spots are generally larger in size.
Fig. 6. The spectrally resolved OBIC images of ESD-HBM VCSEL diode presented by color coding. The image acquisition conditions are the same as in (Fig. 4). As in the case of ESD-MM VCSELs, there is an enhancement in the longer wavelength regime. The HBM induced defective spots are in general larger in size. The scale bar is 5 µm. The color bar besides the OBIC images represent photocurrent intensity in arbitrary units, where the 12-bit dynamic range (0-4095) of the A/D converter of the scanning platform is taken full advantage of.
Fig. 7. (a) The spectral response of photocurrent from the selected region of interests (ROIs) of the three types of VCSELs. In which, the center region of normal VCSEL is represented by black line (solid square), the center of ESD-MM is represented by red line (open circle), the speckle region of ESD-MM is represented by green line (open up triangle), the center of ESD-HBM is represented by blue line (solid down triangle), and the speckle region of ESD-HBM is represented by brown line (solid star). (b) The enlarged spectral response in the long wavelength regime, from 870 nm to 900 nm [dashed black box in Fig. 7(a)].
Figure 7(a) shows the photocurrent spectra of the testing diodes as a function of the incident wavelengths, which also reflect the absorption spectra of VCSELs. We have measured the 5 locations of these three samples, namely, center of the normal sample, center of the ESD-HBM, center of the ESD-MM, the large defective spot of ESD-HBM, and the defective spot of ESD-MM. These spectra show a broad peak centered around 840 nm, which falls down around 860 nm, attributed to the working range of DBR. Note that the maximum absorption of these 850 nm VCSELs is located around 843 nm. In this range, the photocurrent of normal VCSEL is the highest, while the photocurrent of the ESD-HBM affected spot is the lowest. For comparison, the photocurrent response of the normal and the ESD damage spots in the long wavelength regime (> 870 nm) is completely opposite, showing the DBR structure is also changed. Figure 7(b) shows the photocurrent spectra in the long wavelength regime from 870 nm to 900 nm. The increase of photocurrent intensity above the wavelength 875 nm is related to change in transmittance in the DBR mirror.
3.3 Electroluminescence spectra
In the EL measurements, the source-meter (Keithley 2400, Tektronix, USA) provides a forward bias to power the laser diode. The lasing threshold current of these VCSELs is approximately 0.8 mA at room temperature. As expected, ESD defects would result in a lower quantum efficiency for VCSELs, shown in Fig. 8. By controlling the injection current, we measure both the spontaneous emission spectra and stimulated emission spectra from these VCSELs. Note that the normal VCSEL shows pronounced response when compared with the ESD affected ones.
4. Conclusion and outlook
The combination of tunable ultrafast laser and high-speed measurement electronics has enabled unprecedented approaches and capacity in better characterizing high speed VCSELs through OBIC. Specifically, we have demonstrated how the ESD induced defects in VCSELs can be functionally visualized and analyzed spatially and spectrally. In the spectrally resolved measurements, we have found the ESD affected VCSELs exhibits enhanced photocurrent at wavelengths longer than 860 nm, indicating DFB reflectors may be compromised since the relative response (ratio) between the normal and the defective region remains the same. It is difficult to determine the failure spot from EL spectral measurement. For comparison, OBIC can identify specific spots through multiple dimensional imaging, by sweeping the incident wavelengths. Note that MM model has a much faster discharging time constant when compared with the HBM one. As a result, the defect spots attributed to MM model are much smaller than the HBM induced ones (Figs. 5 and 6). Additionally, the quantum efficiency response of ESD affected VCSELs also show significant reduction, which would then suggest the concurrent degradation of the active layer. This observation matches the fact that the highest electric field takes place across the active layer. In summary, functional OBIC enabled by a tunable ultrafast laser has proven to be a very powerful tool for investigating high-speed photonics devices such as VCSELs. The presented spectrally resolved OBIC imaging technique can be applied to the VCSEL research community, especially in the industry, where VCSEL reliability and failure modes and effect analysis are of vital importance.
Funding
Ministry of Science and Technology, Taiwan (MOST 105-2112-M-010-001-MY3, MOST 108-2112-M-010-001).
Acknowledgments
The authors thank TrueLight Corporation, 21, Prosperity Road 1, Hsinchu Science Park, Hsinchu, 30078, Taiwan for providing the vertical-cavity surface-emitting lasers (VCSELs) as the samples used throughout this study.
Disclosures
The authors declare no conflicts of interest.
References
1. K. Iga, “Surface-emitting laser-its birth and generation of new optoelectronics field,” IEEE J. Sel. Top. Quantum Electron. 6(6), 1201–1215 (2000). [CrossRef]
2. F. Koyama, “Recent advances of VCSEL photonics,” J. Lightwave Technol. 24(12), 4502–4513 (2006). [CrossRef]
3. R. Michalzik, VCSELs: fundamentals, technology and applications of vertical-cavity surface-emitting lasers (Springer, 2012).
4. A. Liu, P. Wolf, J. A. Lott, and D. Bimberg, “Vertical-cavity surface-emitting lasers for data communication and sensing,” Photonics Res. 7(2), 121–136 (2019). [CrossRef]
5. R. Herrick, Y. Cheng, and P. Petroff, “Failure analysis of 850 nm proton-implanted VCSELs,” in 1997 Digest of the IEEE/LEOS Summer Topical Meeting: Vertical-Cavity Lasers/Technologies for a Global Information Infrastructure/WDM Components Technology/Advanced Semiconductor Lasers and Application (IEEE, 1997), pp. 77–78.
6. O. Ueda and R. W. Herrick, “Failure analysis of semiconductor optical devices,” in Materials and Reliability Handbook for Semiconductor Optical and Electron Devices (Springer, 2013), pp. 19–53.
7. W. Hergert, S. Hildebrandt, and L. Pasemann, “Theoretical investigations of combined EBIC, LBIC, CL, and PL experiments,” Phys. Stat. Sol. (a) 102(2), 819–828 (1987). [CrossRef]
8. E. Zanoni, G. Spiazzi, G. Dalla Libera, B. Bonati, M. Muschitiello, and C. Canali, “Detection and localization of gate oxide shorts in MOS transistors by optical-beam-induced current,” IEEE Trans. Electron Devices 38(2), 417–419 (1991). [CrossRef]
9. H. Komoda and K. Shimizu, “Optical beam induced current techniques for failure analysis of very large scale integrated circuits devices,” Jpn. J. Appl. Phys. 33(Part 1, No. 6A), 3393–3401 (1994). [CrossRef]
10. C. Xu and W. Denk, “Comparison of one-and two-photon optical beam-induced current imaging,” J. Appl. Phys. 86(4), 2226–2231 (1999). [CrossRef]
11. V. R. Daria, J. J. Miranda, and C. Saloma, “High-contrast images of semiconductor sites via one-photon optical beam–induced current imaging and confocal reflectance microscopy,” Appl. Opt. 41(20), 4157–4161 (2002). [CrossRef]
12. J. J. Miranda and C. Saloma, “Four-dimensional microscopy of defects in integrated circuits,” Appl. Opt. 42(32), 6520–6524 (2003). [CrossRef]
13. V. J. Cemine, B. Buenaobra, C. M. Blanca, and C. Saloma, “High-contrast microscopy of semiconductor and metal sites in integrated circuits by detection of optical feedback,” Opt. Lett. 29(21), 2479–2481 (2004). [CrossRef]
14. T. Takeshita, M. Sugo, T. Sasaki, and Y. Tohmori, “Failure analysis of InGaAs/GaAs strained-layer quantum-well lasers using a digital OBIC monitor,” IEEE Trans. Electron Devices 53(2), 211–217 (2006). [CrossRef]
15. U. Keller, “Recent developments in compact ultrafast lasers,” Nature 424(6950), 831–838 (2003). [CrossRef]
16. F.-J. Kao, M.-K. Huang, Y.-S. Wang, S.-L. Huang, M.-K. Lee, and C.-K. Sun, “Two-photon optical-beam-induced current imaging of indium gallium nitride blue light-emitting diodes,” Opt. Lett. 24(20), 1407–1409 (1999). [CrossRef]
17. F.-J. Kao, J.-C. Chen, S.-C. Shih, A. Wei, S.-L. Huang, T.-S. Horng, and P. Török, “Optical beam induced current microscopy at DC and radio frequency,” Opt. Commun. 211(1-6), 39–45 (2002). [CrossRef]
18. H. Hamad, D. Planson, C. Raynaud, and P. Bevilacqua, “OBIC technique applied to wide bandgap semiconductors from 100 K up to 450 K,” Semicond. Sci. Technol. 32(5), 054001 (2017). [CrossRef]
19. M. Thor, S. Goh, B. Yeoh, H. Hao, Y. Chan, and Z. Lin, “Dynamic optical beam induced current variation mapping: A fault isolation technique,” Microelectron. Reliab. 107, 113603 (2020). [CrossRef]
20. E. I. Cole Jr and J. M. Soden, “Light-induced voltage alteration for integrated circuit analysis,” (Google Patents, 1995).
21. J. Y. Liao, H. L. Marks, and F. Beaudoin, “OBIRCH analysis of electrically stressed advanced graphic ICs,” Microelectron. Reliab. 47(9-11), 1565–1568 (2007). [CrossRef]
22. E. I. Cole Jr, “Thermally-induced voltage alteration for integrated circuit analysis,” (Google Patents, 2000).
23. M. Larson, M. Kondow, T. Kitatani, K. Nakahara, K. Tamura, H. Inoue, and K. Uomi, “GaInNAs-GaAs long-wavelength vertical-cavity surface-emitting laser diodes,” IEEE Photonics Technol. Lett. 10(2), 188–190 (1998). [CrossRef]
24. H. Mawatari, M. Fukuda, K. Kato, T. Takeshita, M. Yuda, A. Kozen, and H. Toba, “Reliability of planar waveguide photodiodes for optical subscriber systems,” J. Lightwave Technol. 16(12), 2428–2434 (1998). [CrossRef]
25. D. Mathes, J. Guenter, B. Hawkins, B. Hawthorne, and C. Johnson, “An atlas of ESD failure signatures in vertical cavity surface emitting lasers,” in Proc. 31st International Symposium for Testing and Failure Analysis (ISTFA) (2005).
26. V. Kuznetsov, “HBM, MM, and CBM ESD ratings correlation hypothesis,” IEEE Trans. Electromagn. Compat. 60(1), 107–114 (2018). [CrossRef]
