Open Access
Add to BibSonomyAbstract
We have demonstrated that, in elevated oxide-layer vertical-cavity surface-emitting lasers (VCSELs) where the spacing between 1–λ cavity and oxide layer is 9λ/4, a variety of turn-off-induced abnormalities such as secondary pulsations, bumps and tails is effectively suppressed. Compared are turn-off transient responses of conventional and elevated oxide-layer VCSELs with the oxide-aperture diameter of approximately 6.6 μm. The “on”- and “off”-current dependence of the turn-off-induced transient responses and bit rate dependence of the timing jitters show that the elevated oxide-layer structure effectively suppresses the turn-off-induced pulsations to less than half of the conventional one, which enables the VCSEL transmitter to operate without deleterious effects by the turn-off-induced pulsation.
©2010 Optical Society of America
1. Introduction
Vertical-cavity surface-emitting lasers (VCSELs) have been evolving toward an ideal light source featuring high-speed operation, low power consumption, compatibility with CMOS integrated circuits and low manufacturing cost. However, there are still unsolved issues in VCSELs: high sensitivity to electrostatic discharge (ESD), nearly parabolic temperature dependence of the threshold current and atypical modulation characteristics including spatial mode dependence. Particularly, in VCSELs under high speed modulation, various irregularities such as mode jumps, turn-off-induced pulsations [1–5], tails, bumps [5] and excessive jitters have been observed. Of these irregularities, the “off”-state turn-on originates from the spatial hole burning (SHB) [1,2] and severely affects the maximum attainable bit rate and minimum bit error rate (BER) [6] because it distorts output optical-signal waveform and eye diagrams [5–7].
Multifarious researches have been performed both theoretically and empirically on the turn-off transient of VCSELs to predict these effects on transmitter performance. Most of researches have focused on developing numerical models [1–9] and only a few researches have demonstrated effective approaches to reducing or eliminating the turn-on and turn-off-induced abnormalities such as transient oscillations and “off”-state turn-on [10, 11]. Moreover, the approaches are difficult to implement under present mass-production environment.
On the other hand, an elevated oxide-layer structure, where the spacing between the oxide layer and 1-λ cavity is greater than the usual spacing of λ/4, is a promising approach to suppressing turn-off-induced abnormalities. As the oxide layer elevates and the spacing d in Fig.1 increases, the carrier concentration gradient between the active region and its vicinity decreases owing to the decreased optical and current confinement. Therefore, the turn-off-induced abnormalities caused by the spatial hole burning are expected to be suppressed effectively in the elevated-oxide-layer structure.
In this paper, we have compared the turn-off transient responses and eye diagrams of the elevated oxide-layer VCSEL with those of conventional one. Experimental results have shown that turn-off-induced abnormalities are suppressed to less than half in the elevated-oxide-layer VCSELs with the oxide aperture diameter of approximately 6.6 μm.
Fig. 1. Schematic diagram of oxide VCSELs. The spacing d between the oxide layer and 1-λ cavity is λ/4 for type A and 9λ/4 for type B VCSELs.
2. Device structure and experimental setup
The oxide VCSELs used in this study consist of p-type distributed Bragg reflector (DBR) mirror layers, n-type DBR mirror layers and undoped 1-λ cavity sandwiched by the two DBR mirror layers as shown in Fig. 1. The upper p-type DBR mirror layer is comprised of 23.5 periods of the Al0.15Ga0.85As/Al0.92Ga0.08As layer pair, while the lower n-type DBR mirror layer is comprised of 39 periods. All the interfaces of the mirror layers are compositionally graded to minimize the series resistance. The p- and n-type DBR mirror layers are uniformly doped with carbon to 2×1018 cm-3 and silicon to 2×1018 cm-3, respectively. Also, on the top of the p-DBR mirror layer, a 55-nm-thick highly doped (>3×1019 cm-3) p-type GaAs layer is deposited for a low resistance Ohmic contact. Used contact metals are Cr/Ni/Au and AuGe/Ni/Au for p- and n-type Ohmic contacts, respectively. Studied are two different types of VCSELs differentiated by the spacing d in Fig. 1. Type A is a conventional VCSEL with the spacing d of λ/4 and type B device is the elevated oxide-layer VCSEL with d of 9λ/4. The diameter of the oxide aperture 2r 0 is 6.6 ± 0.1 μm and the diameter of p-metal aperture is 10 μm. Both VCSELs show multiple spatial mode behavior with lasing wavelength of 847.6 and 846.1 nm at room temperature for type A and B devices, respectively. Pigtail type transmitter modules with 1 m-long multi-mode fiber are prepared for a transmission experiment. The experimental setup for a transmission experiment consists of a pulse pattern generator, a driving circuit, pigtail-type VCSEL transmitter modules, a photoreceiver and an oscilloscope.
3. Results and discussion
Figure 2 shows the measured room-temperature L-I and V-I characteristics of types A and B VCSEL chips. The threshold current of type A VCSEL is as low as 0.4 mA while that of type B VCSEL is 1.5 mA, and the difference comes from the weaker current confinement in the latter. The average slope efficiency of both type A and B VCSELs is about 0.4 W/A up to the operating current of 5 mA. The operating voltage of type B VCSEL is nearly 0.07 V lower than that of type A VCSEL over the measured current range. The differential resistance of type A and B VCSELs at 3 mA are 220.4 and 135.7 Ω, respectively. The output power of type A VCSEL begins to saturate at 6 mA.
Figure 3 shows the “on”-current dependence of the turn-off-induced pulsations measured at room temperature when the “off”-currents are Ith + 0.05 and I,h + 0.1 mA. Slow tails [5] and “off”-state turn-on [1–6] are clearly observed in type A VCSEL [Fig. 3(a) and 3(b)] while the “off”-state turn-on is considerably lowered in type B VCSEL [Fig. 3(c) and 3(d)]. When the “off”-current is Ith + 0.05 mA, the measured optical power variation ΔP0, defined as the difference of the peak power of the turn-off-induced pulsations and steady-state optical power in the “off” state, is 30.4 μW at the “on”-current of 1.8 mA, and it increases to 46.8 μW as the “on”-current increases to 4.0 mA as shown in Fig. 3(a). As the “off”-current increases to Ith + 0.1 mA, ΔP0 increases as can be seen in Fig. 3(b). It is 41.6 μW at the “on”-current of 1.8 mA and increases to 60.8 μW as the “on”-current increases to 4.0 mA. The increase in ΔP0 with the “on”-current is attributable to deeper spatial holes at higher “on”-current.
Fig. 3. Turn-off transient responses of type A and B VCSELs measured at room temperature with the “off”-current of Ith + 0.05 and Ith + 0.1 mA. The “on”-currents of type A and B VCSEL are adjusted for the chip level optical power of 0.67, 1, 1.33 and 1.67 mW. Zero-levels of turn-off transient curves for each “on”-current are shifted for visual aid.
In contrast with type A VCSEL, type B VCSEL shows much smaller ΔPo as shown in Fig. 3(c) and (d). When the “off”-current is Ith + 0.05 mA, ΔP0 is 12.4 μW at the “on”-current of 3.1 mA and increases to 16 μW as the “on”-current increases to 5.2 mA [Fig. 3(c)]. At the “off”-current of Ith + 0.1 mA [Fig. 3(d)], slightly greater ΔP0 of 15.6 μW is observed, and ΔP0 increases to 20.8 μW as the “on”-current increases to 5.2 mA. The “on”-currents of 2.4 and 3.9 mA correspond to 1 mW chip-output power in type A and B VCSEL, respectively. At this power level, ΔP0 values of type B VCSEL at the “off”-current of Ith + 0.05 and Ith + 0.1 mA are 41% and 33% of those of type A VCSEL, respectively.
The delay time td is defined as the time spacing between the peak of the “off”-state turn-on and the starting point of the turn-off transition as shown in Fig. 3(b). In type A VCSEL, when the “off”-current is Ith + 0.05 mA, td slowly decreases from 1.25 ns to 1.17 ns as the “on” current increases from 1.8 mA to 4.0 mA as shown in Fig. 3(a). When the “off”-current is Ith + 0.1 mA [Fig. 3(b)], td becomes slightly shorter showing 1.22 and 1.05 ns at the “on”-current of 1.8 and 4.0 mA, respectively. The delay time td is shorter in type B VCSEL as shown in Fig. 3(c) and (d). When the “off”-current is Ith + 0.05 mA, it decreases from 1.15 to 1.08 ns as the “on”-current increases from 3.1 to 5.2 mA. When the “off”-current is Ith + 0.1 mA [Fig. 3(d)], it decreases further from 1.00 to 0.86 ns as the “on”-current increases from 3.1 to 5.2 mA.
Both the optical power variation δP0 and delay time td are smaller in type B VCSEL because the spatial hole-burning effect is weaker in the elevated oxide-layer VCSEL. Since there is plenty of room for the carriers to spread laterally outside the active region in type B VCSEL, carriers do not pile up at the edge of the oxide aperture. This is in sharp contrast with the significant carrier pile-up in type A VCSEL. Also, weaker optical confinement in type B VCSEL leads to weaker spatial hole-burning effects. The weaker current and optical confinements bring about smaller optical power variation and delay time.
Fig. 4. The turn-off response of type A and B VCSELs measured at room temperature with fixed “on”-current and varied “off”-current of Ith - 0.2 mA to Ith + 0.6 mA with a 0.1 mA-step. The “on”-current is fixed at 2.4 and 3.9 mA for type A and B VCSELs, respectively, for the chip level optical power of 1 mW. The zero-levels of the curves for each “off”-current are shifted for visual aid.
The “off”-current dependence of the turn-off-induced abnormalities measured at room temperature is shown in Fig. 4. We have fixed the “on”-current of type A and B VCSELs to 2.4 and 3.9 mA, respectively, for the chip’s output power of 1 mW, and changed the “off”-current from Ith - 0.2 to Ith + 0.6 mA with a 0.1 mA step. In type A VCSEL, the “off”-state turn-on is observed clearly when the “off”-current is above the threshold as shown in Fig. 4(a), and the optical power variation ΔP0 is largest owing to the “off”-state turn-on when the “off”-current is Ith + 0.1 mA that corresponds to 1.25Ith. With the “off”-current higher than Ith + 0.2 mA, the “off”-state turn-on are observed clearly, however δP0 decreases slowly as the “off”-current increases. The delay time td decreases as the “off”-current increases, although the rate of decrease slowly decreases as the “off”-current increases. The off-current dependence of type B VCSEL is similar to that of type A VCSEL, but the optical power variation δP0 is significantly smaller. The carrier concentration gradient of type B VCSEL at a given operating condition is smaller compared to type A VCSEL so that the recovery of the spatial holes is not sufficient to generate a clear pulsation, which results in less prominent “off”-state turn-on in type B VCSEL in agreement with [4]. For the same reason, the delay time td of type B VCSEL is shorter compared with type A VCSEL.
Fig. 5. (a) Measured optical power variation δP0 in the “off” state and (b) delay time td as a function of Ioff - Ith for type A (‖) and B (◇) VCSELs when the “on”-current is 2.4 and 3.9 mA for type A and type B VCSELs, respectively.
Figure 5 shows the measured δP0 and td as a function of the difference between the “off” and threshold currents, Ioff - Ith, so-called over-drive current. As shown in Fig. 5(a) δP0 of type A VCSEL peaks at Ioff = Ith + 0.1 mA, and is as large as 50 μW, which could result in increased bit error rate. The optical power variation also peaks at Ioff = Ith + 0.1 mA in type B VCSEL, but the magnitude is significantly smaller. Experimentally obtained δP0 in this study is in good agreement with the calculated results of [4], except for a peak at Ioff - Ith = 0.1 mA. When the over-drive current Ioff - Ith is smaller than 0.1 mA, the “off”-state turn-on is barely possible and δP0 is small. On the other hand, when Ioff - Ith is sufficiently large, the “off”-state turn-on disappears as can be seen in Fig. 4 because the fundamental mode will be never turned off. Between these two extremes, the optical power variation δP0 shows a peak. When Ioff - Ith is larger than 0.4 mA, relaxation oscillation dominates over the “off”-state turn-on. Since clear “off”-state turn-on starts to appear when the “off”-current is higher than the threshold current, the delay time td is plotted as a function of Ioff - Ith in Fig. 5(b). When the “off”-current is low, the delay time of type A VCSEL is longer because of the stronger spatial hole-burning in this VCSEL. As the “off”-current increases, the delay time decreases due to the faster recovery from the hole-burned state at higher “off”-current, which is consistent with the theoretical results of [4]. Also, the delay times of type A and B VCSELs get close together as the “off”-current increases.
Fig. 6. Unfiltered eye diagrams of type A VCSEL transmitter module measured at room temperature with fixed “on”-current of 2.4 mA and “off”-current of 0.5 mA at (a) 500 Mb/s, (b) 700 Mb/s, (c) 900 Gb/s, (d) 1.1 Gb/s, (e) 1.3 Gb/s, (f) 1.5 Gb/s, (g) 1.7 Gb/s and (f) 1.9 Gb/s, respectively. The vertical scale is 50 μW/div. and time scale is 200 ps except for (a), (b) and (c) with 500 ps.
In Fig. 6 we have shown the unfiltered eye diagrams of type A VCSEL transmitter module driven by a pulse pattern generator using a non-return-to-zero pseudorandom bit sequence of 27-1. We have fixed the “on”- and “off”-current to 2.4 and 0.5 mA, which correspond to 6Ith and 1.2Ith, respectively. As shown in Fig. 6(a) and (e) the “off”-state turn-on at 500 Mb/s and 1.3 Gb/s is observed near the center of the bit-time-slot, which would result in increased “off” state intensity noise and bit-error-rate. At other bit rates, the “off”-state turn-on occurs at or near transition edges, and we can observe increased timing jitter. The “off”-state intensity noise has been increased more than 200% over the bit rate from 500 Mb/s to 1.9 Gb/s due to the “off”-state turn-on. The “off”-state intensity noise by the “off”-state turn-on may be filtered out at a low bit rate such as 500 Mb/s when a filter is used. However, complete elimination of the “off”-state intensity noise at a higher bit rate such as 1.9 Gb/s is unattainable, and the unfiltered “off”-state intensity noise will deteriorate transmission performance.
Figure 7 shows nearly periodic change in the timing jitter with the bit rate depending on the relative position of the “off”-state turn-on. The root-mean-square (RMS) timing jitter shows peaks whenever the “off”-state turn-on occurs at transition crossing points in eye diagrams. Turn-off pulsation can bring about more than two times larger timing jitter, and rising edges are more vulnerable to the turn-off-induced timing-jitter. As expected, peak timing jitter is much smaller in type B VCSELs.
Fig. 7. Data-rate dependence of the RMS timing jitter of type A (■) and type B (◆) VCSELs measured at room temperature. The “on”-current is fixed at 2.4 and 3.9 mA for type A and B VCSELs, respectively, for the chip level optical power of 1 mW. The “off”-current is fixed at Ioff = Ith + 0.1 mA for both types.
4. Conclusion
We have shown that turn-off-induced pulsation, especially the “off”-state turn-on, not only increases intensity noise but also timing jitter in conventional oxide VCSELs, and experimentally clarified the “on”- and “off”-current dependence of the intensity and timing noises. The intensity noise due to the “off”-state turn-on is most severe when the “off”-current is 0.1 mA above the threshold nearly irrespective of the “on”-current, and timing jitter is most problematic when the “off”-state turn-on occurs at transition crossings of eye pattern. Recognizing that the fundamental origin of the excess intensity and timing noises is the spatial hole-burning, we have suggested the elevated oxide-layer VCSEL as an effective solution to the problem. The elevated oxide-layer VCSEL has been shown to suppress the intensity and excess timing noise to less than half of the conventional VCSELs, where the increased spacing between 1-λ cavity and oxide layer result in weaker current and optical confinements that greatly relieve the spatial hole-burning effects.
References and links
1. A. Valle, J. Sarma, and K. A. Shore, “Spatial holeburning effects on the dynamics of vertical cavity surface-emitting laser diodes,” IEEE J. Quantum Electron. 31, 1423–1431 (1995) [CrossRef]
2. A. Valle, J. Sarma, and K. A. Shore, “Secondary pulsations driven by spatial hole burning in modulated vertical-cavity surface-emitting laser diodes,” J. Opt. Soc. Amer. B 12, 1741–1746 (1995) [CrossRef]
3. A. Valle and L. Pesquera, “Turn-off transients in current-modulated multitransverse-mode vertical-cavity surface-emitting lasers,” Appl. Phys. Lett. 79, 3914–3916 (2001) [CrossRef]
4. A. Valle and L. Pesquera, “Bias level dependence of turn-off oscillations in vertical-cavity surface-emitting lasers,” J. Opt. Soc. Amer. B 23, 2148–2156 (2006). [CrossRef]
5. J. J. Morikuni, P. V. Mena, A. V. Harton, K. W. Wyatt, and S. M. Kang, “Spatially independent VCSEL models for the simulation of diffusive turn-off transients,” J. Lightwave Technol. 17, 95–101(1999). [CrossRef]
6. J. Tatum, D. Smith, J. Guenter, and R. Johnson, “High speed characteristics of VCSELs,” Proc. SPIE 3004, 151–159 (1997). [CrossRef]
7. J. Park, T. Kim, S.-H. Kim, and S.-B. Kim, “A passively aligned VCSEL transmitter operating at fixed current over a wide temperature range,” Opt. Express 17, 5147–5152 (2009). [CrossRef] [PubMed]
8. M. S. Torre and H. F. Ranea-Sandoval, “Influence of the carrier diffusion process on the transient response of vertical-cavity surface-emitting lasers,” Int. J. Numer. Model. 16, 29–39 (2003). [CrossRef]
9. A. Gholami, Z. Toffano, A. Destrez, M. Pez, and F. Quentel, “Spatiotemporal and thermal analysis of VCSEL for short-range gigabit optical links,” Opt. Quantum Electron. 38, 479–493 (2006). [CrossRef]
10. C. W. Tee, S. F. Yu, R. V. Penty, and I. H. White, “Transient response of ARROW VCSELs,” IEEE J. Quantum Electron. 41, 140–147 (2005). [CrossRef]
11. S. Riyopoulos, “Elimination of transient vertical-cavity surface-emitting laser oscillations using photoactive feedback,” Appl. Phys. Lett. 75, 3057–3059 (1999) [CrossRef]
