Electrostatic discharge(ESD) damage is known as a major source affecting the lifetime of oxide VCSEL. We investigated how ESD damage threshold voltage depends on the size, thickness, and composition of the oxide aperture by measuring the change of output power and reverse leakage current after ESD. ESD damage threshold voltage increased with the size of the oxide aperture, regardless of the thickness and the composition of the oxide aperture. However, damaged devices with thinner oxide layers showed relatively longer lifetime in the reliability test. The reliability data also showed that the VCSELs exposed to ESD have steeper power declines in reliability test than normal devices. This may be due to the defects formed in the active medium by ESD.
© 2006 Optical Society of America
Electro-static discharge(ESD) is a major source of damaging semiconductor devices. As the integration density increases, a device becomes increasingly sensitive to ESD . There have been several reports claiming that oxide vertical-cavity surface-emitting lasers(VCSELs) are significantly influenced by ESD like other semiconductor devices [2–5]. Especially, in the case of oxide VCSELs, a smaller aperture makes the VCSEL more sensitive to ESD, while the reduced aperture is inevitable for achieving high speed modulation [3–5]. The consequences of ESD damage are an immediate degradation of device performance and shortened lifetime. Therefore, research on ESD is essential for the study of the reliability and the performance of VCSELs. In addition, such research also provides important data which determines the guidelines for ESD control in the manufacturing processes and the storage of VCSELs. However, in spite of the importance of ESD, there are very few studies investigating the dependence of ESD damage on the various structural parameters of the oxide aperture in VCSELs. The current study measured the ESD damage threshold voltage and the reverse leakage current of the oxide VCSELs with variation of the structural parameters. Additionally, we performed a reliability test on the damaged devices and compared the degradation processes after damage to study its relation to the structural parameters of the oxide aperture.
2. Experiments and Results
We measured the ESD damage thresholds of various sizes of oxide apertures. The Human Body Model(HBM) was chosen to describe the discharge behavior of the ESD source. Human Body Model pulses are generated using ESS-606A of NoiseKen. The ESD damage threshold voltage is the voltage at which an abrupt change of output power begins to appear. In order to find the ESD threshold voltage, ESD is applied to a device being tested every 1 second for 0.5 second with forward bias and reverse bias in turn, three times each. The optical power is measured at 9 mA before and after the ESD. The results of power measurement are shown in Fig. 1.
The ESD threshold voltage increased with the size of oxide aperture, as seen in Fig. 1. The oxide apertures with the diameters of 2.7µm, 8µm, 12µm showed a significant change in the output power at ESD voltages of 40 V, 180 V, and 230 V respectively. It is noteworthy that the individual devices do not have exactly the same ESD damage threshold since each device has some deviation in the size of oxide aperture due to the inherent processing nonuniformity. In case of randomly selected samples, it is hard to say whether a VCSEL undergoes degradation when the fractional change of output power is less than 30% in lightcurrent(LI) curves. VCSEL usually shows 30% deviation in device performance even over the same wafer due to the deviation of DBR and oxide aperture. Therefore, in order to check the influence of ESD, the same device should be characterized before and after ESD.
The leakage current under reverse bias also provides information about the device damage. The increase in leakage current means that the device is damaged by ESD. We measured the reverse leakage currents of the VCSELs with the two different oxide apertures at a bias of -10 V. The measured reverse leakage current is plotted as a function of ESD voltage in Fig. 2.
In terms of reverse leakage current, the ESD damage threshold shows the same tendency as that in terms of output power change under ESD damage. The leakage current in Fig. 2 is actually the difference between one after ESD and the one prior to ESD. Normally, the leakage current before ESD is less than 1 nA. Therefore, the result without subtracting the value prior to ESD gives almost the same picture as the above one. In other words, ESD damages VCSELs and increases reverse leakage current significantly. The magnitude of the leakage current is chaotic above the threshold ESD voltage, but the measurement of reverse leakage current informs whether the VCSEL is influenced by ESD or not without measuring the value before ESD.
To investigate the effect of the thickness and composition of the oxide layer on ESD, we prepared three samples. The specifications of the groups are listed in Table 1. Sample A is the same as the one used for studying the effect of oxide aperture size on ESD.
The three types of oxide layers are used to fabricate the three groups of VCSELs with the oxide aperture of 12 µm. The VCSELs go through the ESD experiment by checking the reverse leakage current. The characteristics of the devices after ESD damage are presented in Fig. 3. The ESD damage threshold of sample A, B, and C are 230 V, 250 V, and 240 V, respectively. The threshold voltages do not show a meaningful difference between the types of samples. The samples vary in the thickness and the composition of oxide layer, not in the size of the oxide aperture. This suggests that the ESD damage threshold mainly depends on the size of the oxide aperture instead of on the thickness and the composition of oxide. It is well known that the AlAs layer experiences 15% contraction in the volume after oxidation, which is greater than that of AlGaAs layer. The larger volume reduction causes greater strain between neighboring layers. The contraction and strain increases with the thickness of oxide layer since the amount of contraction is accumulated by addition of atomic layers . Therefore, a thicker oxide layer maintains a larger strain against the neighboring layers even when the composition is the same as that of the thinner oxide. However, the current experiment demonstrated that the ESD damage threshold does not have much dependency on the thickness and the composition of oxide layer within the range of variation in oxide parameters. Therefore, it is inferred that the ESD damage threshold of oxide VCSEL is determined more by the area of current injection rather than by the amount of strain accumulated in the oxide layer.
A reliability test is performed on the ESD damaged devices shown in Fig. 3 in order to investigate the effect of the oxide structure on device reliability. The reliability data of VCSELs without ESD damage and VCSELs with application of 50% of ESD threshold voltage are presented in Fig. 4. The devices were tested at 90° and 20 mA operation. The sample operated normally after about 260 hours of the test. The similar behavior in reliability test indicates that sub-threshold ESD voltage rarely affects the reliability characteristics of the devices.
As shown in Fig. 4, the control VCSELs were not exposed to ESD and 50% of the ESD threshold voltage was applied to the ESD-exposed devices. On the other hand, Figure 5 shows the test results from the samples treated with 150% ESD damage threshold voltage. Sample C failed earlier than Sample A and Sample B. Although the reverse leakage current measurement gives similar values for the damage threshold voltage regardless of the oxide thickness, the reliability test indicates some dependency on the thickness. The sample including the thin oxide layer(Sample A and B) survived about 100 hours, but the one with thick oxide(Sample C) degraded in output power within about 10 hours. This result does not agree with the reference  reporting that ESD-damaged devices with an oxide aperture of 10 µm in diameter suffer 2 dB reduction of output power within 72 hours under 100°, 5 mA operation. The test condition used in reference 3 is equivalent to 10 hours at 90°, 20 mA operation. Therefore, sample A and B in our experiment showed much longer lifetime than those tested in the preceding report.
The degradation due to ESD can be attributed to heat generation inside the device. The heat creates defects at the interface between the oxide layers and its surrounding layers because of the difference in lattice constants and thermal expansion coefficients [3–5]. The defects at the interface propagate into the active layer and those defects increase the reverse leakage current. Heat generation by ESD is proportional to the cross-section area of the current path or oxide aperture. Actually, the ESD threshold voltage in our measurement agrees with those seen in other reports [3–5]. However, as shown in Fig. 5 the apparent difference in reliability depends on the oxide thicknesses. The distinctness may stem from the number of defects formed at the interface between the oxide region and non-oxidized region. The number of defects is proportional to the amount of strain accumulated at the interface. The amount of strain also influences the propagation speed of dislocation. Since the oxide layer in sample A and B is thinner than sample C, it may generate fewer defects. In this way, sample A and B show better reliability than Sample C. As a result, a thin oxide layer can slow the degradation speed of VCSELs considerably when exposed to ESD.
As observed in Fig. 5, one peculiarity is that the device failed abruptly at a certain point with a sharp decrease in its output power. It is rare to observe such an abrupt decrease in other reliability tests. In general, since the starting point of degradation, the device degrades gradually for a long time and the output power reaches zero after a substantial amount of time [6, 7]. According to other reports, the sudden death of a device can be caused by defects formed in the active layer which activates 1/2 <100> type dislocation in the active medium [4, 8, 9]. The increase of reverse leakage current in the active medium can also be traced to defects in the active medium. The sudden failure shown in Fig. 5 belongs to the case that dislocations are formed by ESD directly in the active medium and are spread more quickly by the high density of minority carriers and photons in the quantum well compared to dislocations formed near the oxide layer. The rapid propagation of the dislocations eventually leads to a sudden failure of the device.
Another way to study the damage of VCSELs is to observe the surface of the output aperture. The near-field pattern analysis provides information about the light intensity distribution and the corresponding carrier distribution. In the current experiment, a near-field observation was performed at 0.3 mA. The near-field pattern right after ESD was not much different from that before ESD when the applied voltage exceeded the ESD threshold voltage and thus the reverse leakage current was increased. The near-field pattern of the device Aa(+50%) is displayed in Fig. 6-a. If many defects are formed near the oxide layer right after ESD, the light intensity near the defects should be reduced, as seen from Fig. 6-b. However, no distinguishable change is observed in Fig. 6-a. The defect density of the observed device did not yet reach the detectable level right after ESD. For comparison, we present a surface image (Fig. 6-b) of a damaged device under a harsh aging condition of 90°, 30 mA for 2 hours. The part indicated by an arrow underwent a reduction in light intensity due to the propagation of dislocation. Fig. 6-c shows a near-field image of the failed device(Aa[+50%]) after the reliability test. The image was taken while the device operates at 10 mA. No light is emitted from the device in spite of current injection. It seems that most of the active layers were already destroyed by dislocation.
In the current study, the ESD damage threshold voltage of VCSEL increased with the size of oxide aperture. The measured ESD damage threshold voltages are 40 V, 180 V, and 230 V for oxide apertures of 2.7µm, 8µm, 12µm in diameter, respectively. Measuring the reverse leakage current of a device makes it possible to decide whether the device is damaged or not without the necessity of measuring the device before ESD. Measurement of reverse leakage current demonstrates that the thickness and composition of oxide layer do not affect the ESD damage threshold. However, reliability tests reveal that the thickness and composition of oxide layer can influence the lifetime after the damage. Samples exposed to ESD with 20 nm oxide layer survive four times longer than those with 46 nm oxide. This may be attributed to the fact that a thicker oxide layer generates denser defects and dislocations under the same ESD condition due to more accumulated strain. In addition, the reliability data from the VCSELs exposed to ESD show that the devices suffered a steeper power decline and sudden device failure than normal devices. It is hypothesized that the defects formed in the active medium by ESD propagate more rapidly and kill the device more quickly as a result.
References and links
1. C. Duvvury, “ESD protection device issues for IC designs,” Proc. of the IEEE custom integrated circuits conference41–48 (2001).
2. H. Neitzert, A. Piccirillo, and B. Gobbi, “Sensitivity of Proton Implanted VCSELs to Electrostatic Discharge Pulses,” IEEE J. Sel. Topics Quantum Electron 7, 231–241 (2001). [CrossRef]
3. J. Krueger, R. Sabharwahl, S. A. McHugo, N. Tan, N. Janda, M. S. Mayonte, M. Heidecker, D. Eastley, M. R. Keever, and C. P. Kocot, “Studies of ESD-related failure patterns of Agilent oxide VCSELs,” SPIE Proc. 4994, 162–172 (2003). [CrossRef]
4. S. A. McHugo, A. Krishnan, J. Krueger, Y. Luo, N-X. Tan, T. Osentowski, S. Xie, M. Mayonte, B. Herrick, Q. Deng, M. Heidecker, D. Eastley, M. Keever, and C. P. Kocot, “Characterization of Failure Mechanisms for Oxide VCSELs,” SPIE Proc. 4994, 55–66 (2003). [CrossRef]
5. B. M. Hawkins, R. A. Hawthorne III, J. K. Guenter, J. A. Tatum, and J. R. Biard, “Reliability of Various Size Oxide Aperture VCSELs,” Electronic Components and Technology Conference. Proc. 52, 540–550 (2002).
6. T. Takeshita, T. Kagawa, K. Tateno, O. Tadanaga, Y. Tohmori, and C. Amano, “Degradation behavior of 850-nm vertical-cavity surface-emitting lasers with an air-post index-guide structure,” J. Lightwave Technol. 13, 722–729 (2002). [CrossRef]
7. J.A Tatum, Andrew Clark, J,K, Guenter, R.A Hawthorne, and R.H Johnson, “Commercialization of Honeywell’s VCSEL Technology,” SPIE Proc. 3946, 2–13 (2000). [CrossRef]
8. S. Xie, G. De Brabander, W. Widjaja, U. Koelle, A. N. Chen, L. Giovane, F. Hu, B. Herrick, M. Keever, and T. Osentowski , “Reliability of Oxide VCSELs in Non-Hermetic Environments,” Conference Proceedings of IEEE LEOS 15th Annual Meeting 2, 544–545 (2002).
9. S. Xie, R. W. Herrick, D. Chamberlin, S. J. Rosner, S. McHugo, G. Girolami, M. Mayonte, S. Kim, and W. Widjaja, “Failure mode analysis of oxide VCSELs in high humidity and high temperature,” J. Lightwave Technol. 21, 1013–1019 (2003). [CrossRef]