Detailed Design and Characterization of All-Optical ... - IEEE Xplore

1582

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 46, NO. 11, NOVEMBER 2010

Detailed Design and Characterization of All-Optical Switches Based on InAs/GaAs Quantum Dots in a Vertical Cavity Chao-Yuan Jin, Osamu Kojima, Tomoya Inoue, Takashi Kita, Osamu Wada, Fellow, IEEE, Mark Hopkinson, and Kouichi Akahane

Abstract— We propose an all-optical switch based on selfassembled InAs/GaAs quantum dots (QDs) within a vertical cavity. Two essential aspects of this novel device have been investigated, which includes the QD/cavity nonlinearity with appropriately designed mirrors and the intersubband carrier dynamics inside QDs. Vertical-reflection-type switches have been fabricated with an asymmetric cavity that consists of 12 periods of GaAs/Al0.8 Ga0.2 As for the front mirror and 25 periods for the back mirror. All-optical switching via the QD excited states has been achieved with a time constant down to 23 ps, wavelength tunability over 30 nm, and ultralow power consumption less than 1 fJ/µm2 . These results demonstrate that QDs within a vertical cavity have great advantages to realize low-powerconsumption polarization-insensitive micrometer-sized switching devices for the future optical communication and signal processing systems. Index Terms— All-optical switch, quantum dots (QDs), ultrafast photonics, vertical cavity.

I. I NTRODUCTION

F

UTURE optical communication systems demand highspeed signal processing with a data transfer rate of 100 Gb/s–1 Tb/s. Ultrafast photonic devices, such as femtosecond light sources and all-optical switches, are essential components to build future optical networks [1]. To realize all-optical switching, it is important to access to the nonlinear operation region of photonic materials, which usually requires high-power excitation. This becomes a well-known problem of the “power/speed tradeoff” [2], [3]. To solve this problem, nanoscale materials such as self-assembled quantum

Manuscript received March 24, 2010; revised June 4, 2010; accepted June 13, 2010. Date of current version August 24, 2010. C. Y. Jin and O. Wada are with the Division of Frontier Research and Technology, Center for Collaborative Research and Technology Development, Kobe University, Kobe 657-8501, Japan. O. Wada is also with the Department of Electrical and Electronic Engineering, Graduate School of Engineering, Kobe University, Kobe 657-8501, Japan (e-mail: [email protected]; [email protected]). O. Kojima, T. Inoue, and T. Kita are with the Department of Electrical and Electronic Engineering, Graduate School of Engineering, Kobe University, Kobe 657-8501, Japan (e-mail: [email protected]; [email protected]; [email protected]). M. Hopkinson is with the Department of Electronic and Electrical Engineering, University of Sheffield, Sheffield S1 3JD, U.K (e-mail: [email protected]). K. Akahane is with the National Institute of Information and Communications Technology, Tokyo 184-8795, Japan (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/JQE.2010.2053916

dots (QDs) are particularly attractive due to their small volume as 3-D confined structures [4]. Atom-like carrier states in QDs with very high differential gain/absorption parameters are anticipated to generate high optical nonlinearity with ultralow power consumption. At present, considerable progress has been made on ultrafast QD devices, such as mode-locked QD lasers, QD semiconductor optical amplifiers (QD-SOAs), and QD saturable absorber mirrors [5]– [9] Nonlinear gain dynamics was initially mentioned to explain the ultrafast performance of QD-SOAs [10]. In the mean time, the use of nonlinear absorption dynamics has been proposed for a Mach–Zehnder interferometer based on QDs [11]. However, the small volume of QDs requires a very long waveguide structure to fulfill the interaction between dots and the light. Although this lateral geometry has provided the first demonstration of all-optical switches based on QDs [12], the size of such a device, which is of a few hundred micrometers, makes it high cost and difficult for integration. Alternatively, a vertical-cavity QD switch has been proposed based on optical Kerr effects inside QDs [13], [14]. Recently, we first reported a vertical-cavity QD switch using the absorption saturation of self-assembled InAs/GaAs QDs [15]. Such a vertical geometry could potentially provide lowpower-consumption polarization-insensitive micrometer-sized switching devices based on QDs. Since the detailed design and operation principles of the vertical-cavity QD switch have not reported in [15], two important issues should be further addressed for the design of QD switches: 1) the QD/cavity nonlinearity; and 2) the carrier dynamics inside the QDs. Historically, for vertical-geometry structures, an enhanced optical nonlinearity, was first observed by inserting a distributed Bragg reflector (DBR) below InGaAs/InAlAs quantum wells (QWs) [16]. Thereafter, structures with an asymmetric vertical cavity were developed for all-optical switches based on GaAs bulk materials [17]. Because a large absorption strength exists in the QW and bulk materials, high differential reflectivity can be easily approached with a weak-reflection front mirror, which corresponds to a low-finesse cavity. However, due to the dispersive distribution of self-assembled QDs both in the real and in the frequency space, the effective cross section and interaction length are extremely small. A relatively high-finesse cavity is therefore demanded for QDs. Both the front and back mirrors of the vertical cavity need to be optimized to fit into the QD case.

0018–9197/$26.00 © 2010 IEEE

JIN et al.: ALL-OPTICAL SWITCHES BASED ON InAs/GaAs QUANTUM DOTS IN A VERTICAL CAVITY

On the other hand, the carrier dynamics inside QDs will further limit the switching performance. To introduce in defect channels by using impurity doping or low-temperature growth techniques was previously suggested for QW and bulk materials [18]. These methods would reduce the absorption strength and hence degrade the optical nonlinearity. In comparison with this, discrete energy states in QDs offer another routing mechanism to manipulate the switching dynamics via energy states higher than ground QD states (GS). The fast carrier relaxation between QD states could be utilized to enhance the device performance. In this paper, we investigate the design principles of verticalgeometry QD switches. The relation between the GaAs/AlAs mirrors and the QD/cavity nonlinearity is derived theoretically. Two switches based on the ground and excited QD states (ES) were fabricated, characterized, and compared. Our results show that QD materials are potentially suitable for compact all-optical switches in the future optical communication and signal processing systems.

II. P RINCIPLES OF THE V ERTICAL C AVITY QD S WITCH Fig. 1 schematically illustrates the working principle of a QD switching device with a vertical Fabry–Perot (FP) cavity. The FP cavity consists of two DBR mirrors, which further includes multiple periods of alternating high- and low-index layers. The two DBR mirrors are named the front and back mirrors in this paper. Each individual layer of the DBR mirror has a thickness of λ/4, where λ is the operation wavelength. The cavity region between two DBR mirrors has a thickness L equal to an integral multiple of λ/2, which is the so-called λ-cavity structure. The cavity reflectivity spectrum shown in Fig. 1 is numerically calculated by using a transfer matrix method (TMM) [19]. A narrow dip exists in the middle of the spectrum, which corresponds to the cavity resonant mode (or cavity mode in brief). Only the light in a narrow wavelength region at the cavity mode can penetrate the cavity. The high-reflectivity region with oscillating side lobes on both sides is the photonic bandgap, which is generated by the 1-D periodicity of the refractive index. When a control light pulse pumps the cavity mode, QDs inside the cavity are excited. The absorption saturation of QDs shifts the cavity mode and hence yields an optical switching process. Due to the Kramers–Kronig relation, if the pump light has a symmetric shape, the change of the refractive index is equal to zero at the central wavelength of the pump pulse. We have simply ignored the carrier heating effect because the switch is operated as a passive-type device and QDs are well separated spatially from each other. Only the absorption saturation is considered in the following theory. As mentioned in the introduction, energy states higher than GS in QDs are also utilized. In Fig. 1, two QD absorption spectra are depicted with dotted curves. By adjusting the QD size and composition, optical emission from either GS or ES can be selected. This provides a simple means to investigate the switching speed with respect to the intersubband relaxation of carriers.

1583

Light

Photonic band gap

R λ

Cavity mode

Side lobe

QD absorption GaAs Fig. 1. Schematic diagram for the working principle of a vertical-geometry all-optical QD switch.

A. Zero Reflectivity Condition At the cavity resonant mode, the reflection from the front mirror can be fully cancelled by the effective reflection from the back mirror. This condition is named the “zero reflectivity condition” [17]. The differential reflectivity is a key parameter to evaluate the switching performance, which corresponds to the reflectivity variation with and without the optical pumping. As discussed later, the maximum differential reflectivity slightly departs from the zero reflectivity condition in the case of QD switches. For a periodic structure, which comprises two consecutive layers of materials with different refractive indices, the total reflectivity at the cavity resonant wavelength λ is derived as [20] 2 n 0 − n H (n H /n L )2x (1) R= n 0 + n H (n H /n L )2x where n 0 is the refractive index of the incidence medium, n H and n L are the refractive indices of high- and low-index layers, respectively, and x is the period number of the alternating high- and low-index layers. For the proposed structure shown in Fig. 1, the reflectivity of the front and back mirrors is given by 2 1 − n H (n H /n L )2 p RF = (2) 1 + n H (n H /n L )2 p and RB =

n H − n H (n H /n L )2q n H + n H (n H /n L )2q

2 (3)

where p and q are the period of the front and back mirrors, respectively. When a λ-cavity is considered, the reflectivity at the cavity mode is expressed as [21] RCM = R F

1 − (R B /R F )1/2 e− 1 − (R B R F )1/2 e−

2 (4)

where = 2 α(l)dl is the total absorption in a cavity L

and α(l) is the absorption coefficient. is a dimensionless parameter, which comprises the total absorption strength inside


This equation presents the zero reflectivity condition. The reflectivity spectrum in Fig. 1 is working at the zero reflectivity condition. The reflectivity at the cavity mode after intensive pumping is derived from (4) by assuming that the absorption in the cavity is fully saturated with 0 = 0 2 1 − (R B /R F )1/2 0 RCM = RF . (6) 1 − (R B R F )1/2 The differential reflectivity is therefore defined by 0 RCM = RCM − RCM .

(7)

1.0

It should be noted that all the equations above are derived at the cavity resonant wavelength. By comparing (9) and (5), the maximum differential reflectivity requires a front-mirror reflectivity less than that of the zero reflectivity condition. Both (9) and (5) require an asymmetric geometry of the cavity. In the case of QW or bulk materials, a large value exists, which suggests a small reflectivity for the front mirror. A lowfinesse cavity therefore works well for QW and bulk materials [16], [17]. However, the value is extremely small for QD structures, normally on the order of 10−4 . A relatively highfinesse design needs to be addressed for QD switches. B. Back Mirror To study the back mirror design, the relation between the front and back mirror periods is fixed at the maximum differential reflectivity. In Fig. 2, the cavity reflectivity with different periods of the back mirror is simulated by using (4) and (9). When the period of the back mirror increases, the reflectivity of the mirror is significantly enhanced. After exceeding 20 periods, the back mirror reflectivity almost reaches 1, while the differential reflectivity of the cavity increases rapidly. In the figure, three test values of are used with = 1.0 ×10−4 (solid triangles), 3.0 × 10−4 (open circles), and 1.0 × 10−3 (solid squares), respectively. The differential reflectivity of the cavity exhibits great nonlinearity in the region of 20–30 periods of the back mirror. In our experiment, we chose 25 periods for the back mirror to ensure that the switching device was working at this highly nonlinear region. By assuming that the factor is on the order of 10−4 , we selected 12 periods for the front mirror.

High nonlinear region

0.8

0.4

0.6

0.3

0.4

0.2

0.2

0.1

0.0

0

5

10 15 20 Period of back mirror

25

0.0 30

Fig. 2. QD/cavity nonlinearity represented by the reflectivity of the back mirror (solid curve) and the differential reflectivity of the cavity with = 1.0 × 10−4 (solid triangles), 3.0 × 10−4 (open circles), and 1.0 × 10−3 (solid squares), respectively.

In the case of → 0 for QDs

1

100-fs laser pulse p = 12

Intensity (normalized)

∂ RCM . (8) RCM ≈ ∂ Assuming the front mirror reflectivity is arbitrarily varied, RCM reaches its maximum with ∂RCM /∂ R F = 0. Hence, the differential reflectivity has a maximum at √ 1/2 − 3R e − 1 1/2 1/2 R F = √ B 1/2 < R B e− . (9) − 3 − RB e

0.5 Differential reflectivity

the cavity with a certain electric field distribution. RCM reaches its minimum at R F = R B e−2 . (5)

Reflectivity of back mirror

1584

10 8 6 0 2 0 0.0

0.5

4

1.0 Time (ps)

1.5

2.0

Fig. 3. Optical pulse profile inside an asymmetric cavity with increased periods of the front mirror ( p = 0, 2, 4, . . . , 12). The period number of the back mirror is fixed at 25.

C. Front Mirror Ideally, an optical switch with high differential reflectivity is preferred. However, another limitation for QD switches is in the time scale. When a high-finesse cavity design is chosen, the optical pulse energy would be stored in the cavity for a certain duration, which obeys the universal inequality as follows: t · ω ≥

1 2

(10)

where the equality holds if a Gaussian optical pulse is under consideration. t is the pulse width and ω is the spectral bandwidth with the angular frequency. This inequality between the time and frequency intervals suggests that a narrower dip at the cavity mode corresponds to a longer storage of the optical pulse inside the cavity. The optical pulse duration is therefore extended when the cavity finesse value increases, which would prolong the switching response time. To illuminate this point, we have employed a 1-D finite difference time domain method to numerically calculate the field intensity profile of the optical pulse inside a cavity with a fixed 25-period GaAs/AlAs back mirror. As shown in Fig. 3, by using a 100-fs optical pulse (dashed curve) to excite the front mirror, the pulse profile

2.0 ps 100

10

10−1 0 0

5

10 15 Period of front mirror

20

25

Fig. 5. Schematic illustration for the designed structure of a QD switch described in this paper. The insert shows a TEM picture of QD layers and the cavity.

Fig. 4. Spectral linewidth of the cavity resonant mode (solid circles) and the optical pulse width inside the cavity (open squares) as a function of frontmirror periods. The dashed curve indicates the limitation estimated from the inequality between the time and frequency intervals.

III. S TRUCTURE AND FABRICATION Based on the principles discussed above, we have fabricated a cavity structure for the QD switch as shown in Fig. 5. Selfassembled InAs QDs were inserted into a vertical cavity which consists of 12 periods for the front mirror and 25 periods for the back mirror. Twenty percent Ga is added to the AlAs layer to prevent lateral oxidization of Al atoms. The thicknesses of the GaAs and AlGaAs layers are chosen to be 89 and 102 nm, respectively. The spatial distribution of the optical electric field intensity (|E|2 ) inside the cavity is calculated using TMM simulation, as shown by the solid oscillating curve in Fig. 5. A 22× enhancement is theoretically estimated for the optical field inside the cavity. The switching sample was grown by molecular beam epitaxy on a GaAs(001) substrate. InAs QD layers were prepared by the Stranski–Krastanow growth mode. To give a dot-in-a-well (DWELL) structure, 2.6 ML of InAs was deposited within an 8-nm In0.15 Ga0.85 As QW. Three layers of InAs QDs were placed at the antinode positions of the optical field. The QD ES emission peak is assigned to match the cavity resonant mode. The inset figure shows a photograph taken by transmission electron microscopy (TEM). High-quality QDs and a smooth interface in DBRs are observed. Another wafer was prepared for comparison, in which the InAs QDs were directly deposited in the GaAs matrix without the DWELL structure. The GS emission peak is located at the cavity mode.

0.8

PL Intensity (a.u.)

inside the cavity is extended when the period number of the front mirror increases. Fig. 4 presents the pulse duration as a function of the period of the front mirror (open squares). The pulse duration is defined as the time width at half-maximum. The dashed curve indicates a limitation governed by the inequality (10), in which the frequency interval ω is estimated by TMM calculation. Because the optical pulse profile inside the cavity obeys a nonGaussian shape, the curve with open squares departs from the limitation described by the dashed curve, as the inequality holds in (10). A cavity with 12/25 periods GaAs/AlAs gives a cavity pulse width around 2 ps, which is short enough for a 100 Gbits/s switch design.

1.0

0.6 0.4

Reflectivity

101

Distribution of the Electrical Field |E|2 GaAs substrate

20

1585

Pulse duration inside cavity (ps)

Linewidth of cavity mode (nm)


0.2

1100

1200 1300 Wavelength (nm)

0.0 1400

Fig. 6. Cavity reflectivity and PL emission of QDs as functions of wavelength. For cavity reflectivity spectra, the dashed curve presents the experimental results while the dot-dashed curve is the theoretical design [15].

IV. C HARACTERIZATION A. PL and Reflectivity Spectra For the optical characterization of the QD switch sample, two continuous wave techniques were employed: 1) edge PL spectra to reveal the QD electronic structure; and 2) reflectivity spectra to study the cavity reflection. In the edge emission PL measurement, the QD switch sample is vertically excited by a laser beam at the surface of the wafer. The PL signal is collected laterally from the sample edge. The emission peaks of InAs QDs are observed with GS at 1298 nm and ES at 1220 nm (solid curve). The second QD sample exhibits the GS emission at 1235 nm and the first ES emission at 1170 nm (dotted curve). Cavity reflectivity spectra of the ES switching sample are also shown in the Fig. 6. The reflectivity spectrum gives a cavity resonant mode with a spectral linewidth of 2.0 nm, corresponding to a cavity finesse value of ∼ 600. The cavity mode wavelength is at 1225 nm for the GS switching sample and 1238 nm for the ES switching sample. The dot-dashed curve in the Fig. 6 indicates a theoretical design given by the TMM simulation, which is well matched by the experimental result. B. Switching Dynamics Orthogonally polarized pump and probe beams were employed to measure the switching dynamics. Optical pulses of 130 fs with a typical bandwidth of 20 nm and a repetition rate of 80 MHz were generated by an optical parametric

1586


ECGaAs

Differential reflection

1.0 ECWELL

0.8

Mode 1

ES 0.6

τ0

ES 32 ps

0.2

0

ES

50

100

Fig. 7. Differential reflectivity as a function of delay time for both switching processes via GS and ES [15].

100

0º 50º

R

32

28

10−1 0

50

100

26 Experiment

24

Simulation 22 1210

1220

1230 Wavelength (nm)

GS

150

Time (ps)

30

Mode 2

GS

GS 80 ps

0.4

0.0

Switching time (ps)

Control light

1240

Fig. 8. Wavelength dependence of the switching time from measurement (open circles) and simulation (dashed curve). The inset gives two examples of switching curves with operation wavelength at 1240 and 1219 nm [15].

oscillator. When the pump beam excited the front cavity mirror at the wavelength of the cavity mode, the differential reflectivity was traced by the probe beam. A switching process with a time constant of 80 ps was demonstrated for the GSswitching sample as shown in Fig. 7. An ultrafast component within a few picoseconds was observed at the first part of the dynamic curve, in agreement with a previous report on QDs [22]. Another wafer, which was optimized for ES switching, showed a switching time of 32 ps, which is faster than the GS switching. The fast relaxation of carriers into the GS helps the absorption recovery of the ES, which explains the significant faster switching process in the ES switching sample. The detailed discussion can be found in [15]. The operation wavelength of the vertical-cavity QD switch is tightly limited by the spectral linewidth of the cavity resonant mode. However, a broad wavelength tunability can be still attained by simply verifying the incident angle of the optical pulses. A wavelength tunability of 30 nm is therefore demonstrated by the QD switch. The switching wavelength is varied near the ES emission peak from 1240 to 1210 nm with an incident angle between 0 to 50º . This degree of tuning is possible due to the inhomogeneous broadening of absorption spectra, which usually cover a range of ∼50 nm for QDs. As shown in Fig. 8, the angle-dependent switching

Fig. 9. Schematic diagram for the photon coupling mechanism due to the optical spectrum overlap between the GS and ES. The insert shows the intersubband relaxation of carriers in QDs.

time (open circles) decreases with decreasing wavelength, and a minimum switching time of 23 ps is reached around the ES emission wavelength. For wavelengths shorter than 1220 nm, the switching time remains constant. The inset in Fig. 8 shows two switching dynamics curves with the operation wavelength at 1240 and 1219 nm, respectively. Therefore, the present QD switch using the ES has demonstrated a switching time down to 23 ps and a wavelength tunability over 30 nm. A photon coupling process that is induced by the spectral overlap between the GS and ES is utilized to explain the wavelength dependence of the switching time [23], [24]. For the 1.3 µm DWELL structure in the present case, a bimodal distribution of QDs has always been observed, which also appears in many other published reports [25]. A distribution of smaller dots presents a GS optical transition, which partially overlaps the ES transition peak, as shown in Fig. 9. The optical absorption of the pump light near the ES emission peak therefore is from a combination of different-size QDs, which may have either ES or GS transitions. The time-dependent absorption in QDs can be described by a summation of two exponential decays t t + A2 (λ) exp − α(t, λ) = A1 (λ) exp − τgs τes t t = A1 (λ) + A2 (λ) exp − · exp − τ0 τgs (11) where A1 and A2 represent the absorption intensity of GS and ES, respectively. They are described by Gaussian functions with the bimodal distribution Ai (λ) =

j =a,b

Ij

i·√ exp 2πσi

(λ − λij )2 σi2

i = 1, 2 (12)

where λa and λb is the central wavelength of the two modes, and σ is the standard deviation of the Gaussian function. If A2 A1 , (11) can be simplified by using second term


expansion of the Taylor series α(t, λ)

3

where

τ1 (λ) =

A2 (λ) −1 τ −1 + τgs A1 (λ) + A2 (λ) 0

ES switch GS switch

Simulation R/R (%)

t −A2 (λ) t ≈ [ A1 (λ) + A2 (λ)] exp · exp − A1 (λ) + A2 (λ) τ0 τgs t = [A1 (λ) + A2 (λ)] exp − τ1 (13)

1587

2

1

−1 (14)

with τ0 = (1/τes − 1/τgs )−1 . The time constant τ0 is related to the carrier relaxation time from ES to GS if we simply assume that the carrier escape rate via thermalization [22] and recombination is almost the same for both samples. Therefore, an intersubband carrier relaxation time of ∼32 ps can be estimated in QDs. τ1 is the angle-dependent switching time. In the simulation, we have assumed carrier decay times of the GS and ES to be 80 and 23 ps, respectively. The ratio between the dot densities of two modes is Ia : Ib = 3 : 2. The central wavelengths of two modes are at 1300 and 1260 nm, respectively. The standard deviations of two modes are assumed to be same, which is equal to 14 meV for QD ground states and 25 meV for excited states. As shown in Fig. 8, the simulation reproduces the experimental results very well. Due to increased violation from A2 A1 , the simulation curve slighted deviates at the longer wavelength side. C. Power Consumption Because of the small volume of QDs, ultralow saturation power is expected in the operation of QD switches [11]. In our measurement, the probe beam power is set at 1/100 of that of the pump beam. The optical pulse has a bandwidth of 20 nm, which is about 10 times of the cavity mode linewidth. To estimate the power consumption and differential reflectivity, we have simply assumed that 1/10 of the power has been used in the switching process equivalently. Fig. 10 shows the differential reflectivity as a function of the excitation power density. Ultralow power densities of less than 1 fJ/µm2 are obtained for both GS and ES switches. The saturation behavior can be simulated by using (4) by assuming that the QD absorption follows a saturation function as follows: 0 (15) = 1 + (P/Ps ) where P is the excitation power density and Ps is the saturation power density. Ps = 0.9 fJ/µm2 is used for the simulation of both curves with 0 = 4.0 × 10−4 and 2.0 × 10−4 , which provides good agreement with the experimental data. V. F UTURE W ORK In general, two possible routes can be considered for future improvement of the QD switches: one is to enhance the QD/cavity nonlinearity, and another is to improve the carrier dynamics inside QDs.

0 0.0

0.2

0.4 0.6 0.8 Power density (fJ/µm2)

1.0

Fig. 10. Differential reflectivity as a function of the power density of the pump pulses. The power of the probe beam is set at 1/100 of the pump power.

A. QD/Cavity Nonlinearity In the characteristics of the switching device, the differential reflectivity indicates 2–3% variation of the cavity reflectivity, e.g., R/R = 2–3%, as shown in Fig. 10. This small contrast ratio of the QD switch is caused by the small volume of dots. The total carrier states in the QD layer is about 1/100 of its counterpart such as the QWs or bulk materials. The small number of carrier states requires fewer carriers to be filled in to saturate the absorption. This helps the ultralow power consumption of QD switches, but at the same time results in a low contrast ratio. This situation can be optimized by increasing the optical enhancement inside the cavity, which would cause a tradeoff between the contrast ratio and the cavity response time as discussed in Section II-C, or by simply increasing the number of carrier states in the active region using high density dots or multiply layers of QDs, which would cause a tradeoff between the contrast ratio and the power consumption. By comparing with simulation results from (6) and (7), a contrast ratio of 2–3% corresponds to an absorption strength of ≈ 2.5 × 10−4 , which is close to one of the test values used in Fig. 2 (open circles). Based on the measured factor, the differential reflectivity as a function of the period number of the front mirror is calculated as shown in Fig. 11 (solid squares). The curve indicates that the optimized value for the front mirror is 15 periods. This could give only a 20% increase in the differential reflectivity. An increase of the QD density and layer number in the active region will further enhance the optical nonlinearity. If we could achieve a value of 1 × 10−3 , a 5× increase can be expected for the differential reflectivity, as shown in Fig. 11 by open circles. In this case, we would achieve R/R ≈ 10%, while the maximum differential reflectivity requires 14 periods for the front mirror. Also, as a more efficient means, the cavity nonlinearity could be significantly improved by simply increasing the period number of the back mirror while maintaining the maximum differential reflectivity. However, this may increase the growth time and may also delay the switching response since the optical pulse duration is prolonged in the cavity.

1588


Differential reflectivity

0.10

R EFERENCES

0.05

0.00 −0.05

= 2.3 × 10−4 = 1.0 × 10−3

−0.10 0

5

10 15 Period of front mirror

20

25

Fig. 11. Simulation results of the differential reflectivity as a function of the period number of the front mirror, assuming 25 periods are used for the back mirror. The box shows the experimental results in this paper. The calculation is based on the experimental estimation of the factor.

B. Carrier Dynamics in QDs For the switching dynamics, a faster sweep out of remaining carriers from the GS or ES would be an effective approach to reduce the switching time and enable faster operation for the GS or ES switching devices. One possible means is to engineer a nonradiative channel. For this purpose, either the low-temperature growth method to introduce defects arising from poor stoichiometry or ion implantation to induce vacancies or interstitials could be employed. Impurity doping could be further utilized and, indeed, this has already shown its feasibility in the high-speed modulation of QD laser devices [24]. However, as mentioned in the introduction, a direct implantation of the ion or impurities into the dot may significantly reduce the optical nonlinearity. In that sense, a carrier tunnel injection structure, which was reported by our group recently, with a diluted nitride layer separated from QDs could be very useful to fast the carrier dynamics and in the mean time retain high optical nonlinearity of QDs [26]. VI. S UMMARY In summary, we have developed a vertical-geometry alloptical switch device based on self-assembled InAs QDs within a GaAs/Al0.8 Ga0.2 As vertical cavity structure. Pumpprobe measurements have shown a switching time of 80 ps via QD GS and 23 ps via QD ES, confirming that the fast intersubband transition of carriers inside QDs is an effective means to speed up the switching process. A comparison between theoretical design and the measurement shows that the crucial factor to yield an efficient QD switching device is to enhance the QD/light interaction (an increased value). These results indicate that the proposed vertical structure involving QD materials is potentially suitable for compact all-optical switches for future ultrafast telecommunication systems. VII. ACKNOWLEDGMENT The authors would like to thank Prof. Yasuda of Kobe University, Kobe, Japan, for the help on transmission electron microscopy.

[1] O. Wada, “Femtosecond all-optical devices for ultrafast communication and signal processing,” New J. Phys., vol. 6, no. 1, p. 183, Nov. 2004. [2] P. W. Smith, “Application of all-optical switching and logic,” Phil. Trans. R. Soc. Lond. A, vol. 313, no. 1525, pp. 349–355, Dec. 1984. [3] O. Wada, “Femtosecond semiconductor-based optoelectronic devices for optical-communication systems,” Opt. Quantum Electron., vol. 32, nos. 4–5, pp. 453–471, May 2000. [4] Y. Arakawa and H. Sakaki, “Multidimensional quantum well laser and temperature dependence of its threshold current,” Appl. Phys. Lett., vol. 40, no. 11, pp. 939–941, 1982. [5] E. U. Rafailov, M. A. Cataluna, and W. Sibbett, “Mode-locked quantumdot lasers,” Nature Photon., vol. 1, no. 7, pp. 395–401, 2007. [6] X. Huang, A. Stintz, H. Li, L. F. Lester, J. Cheng, and K. J. Malloy, “Passive mode-locking in 1.3 µm two-section InAs quantum dot lasers,” Appl. Phys. Lett., vol. 78, no. 19, pp. 2825–2827, 2001. [7] T. Akiyama, O. Wada, H. Kuwatsuka, T. Simoyama, Y. Nakata, K. Mukai, M. Sugawara, and H. Ishikawa, “Nonlinear processes responsible for nondegenerate four-wave mixing in quantum-dot optical amplifiers,” Appl. Phys. Lett., vol. 77, no. 12, pp. 1753–1755, 2000. [8] P. Borri, W. Langbein, J. M. Hvam, F. Heinrichsdorff, M.-H. Mao, and D. Bimberg, “Time-resolved four-wave mixing in InAs/InGaAs quantum-dot amplifiers under electrical injection,” Appl. Phys. Lett., vol. 76, no. 11, pp. 1380–1382, 2000. [9] E. U. Rafailov, S. J. White, A. A. Lagatsky, A. Miller, W. Sibbett, D. A. Livshits, A. E. Zhukov, and V. M. Ustinov, “Fast quantum-dot saturable absorber for passive mode-locking of solid-state lasers,” IEEE Photon. Technol. Lett., vol. 16, no. 11, pp. 2439–2441, Nov. 2004. [10] T. Akiyama, H. Kuwatsuka, T. Simoyama, Y. Nakata, K. Mukai, M. Sugawara, O. Wada, H. Ishikawa, “Nonlinear gain dynamics in quantum-dot optical amplifiers and its application to optical communication devices,” IEEE J. Quantum Electron., vol. 37, no. 8, pp. 1059–1065, Aug. 2001. [11] R. Prasanth, J. E. M. Haverkort, A. Deepthy, E. W. Bogaart, J. J. G. M. van der Tol, E. A. Patent, G. Zhao, Q. Gong, P. J. van Veldhoven, R. Nötzel, and J. H. Wolter, “All-optical switching due to state filling in quantum dots,” Appl. Phys. Lett., vol. 84, no. 20, pp. 4059–4061, May 2004. [12] H. Nakamura, Y. Sugimoto, K. Kanamoto, N. Ikeda, Y. Tanaka, Y. Nakamura, S. Ohkouchi, Y. Watanabe, K. Inoue, H. Ishikawa, and K. Asakawa, “Ultrafast photonic crystal/quantum dot all optical switch for future photonic networks,” Opt. Exp., vol. 12, no. 26, pp. 6606–6614, Dec. 2004. [13] T. Kitada, T. Kanbara, K. Morita, and T. Isu, “A GaAs/AlAs multilayer cavity with self-assembled InAs quantum dots embedded in strainrelaxed barriers for ultrafast all-optical switching applications,” Appl. Phys. Exp., vol. 1, no. 9, pp. 092302-1–092302-3, 2008. [14] K. Morita, T. Takahashi, T. Kitada, and T. Isu, “Enhanced optical Kerr signal of GaAs/AlAs multilayer cavity with InAs quantum dots embedded in strain-relaxed barriers,” Appl. Phys. Exp., vol. 2, no. 8, pp. 082001-1–082001-3, 2009. [15] C. Y. Jin, O. Kojima, T. Kita, O. Wada, M. Hopkinson, and K. Akahane, “Vertical-geometry all-optical switches based on InAs/GaAs quantum dots in a cavity,” Appl. Phys. Lett., vol. 95, no. 2, pp. 021109-1–0211093, Jul. 2009. [16] R. Takahashi, Y. Kawamura, and H. Iwamura, “Ultrafast 1.55 µm alloptical switching using low-temperature-grown multiple quantum wells,” Appl. Phys. Lett., vol. 68, no. 2, pp. 153–155, Jan. 1996. [17] H. S. Loka and P. W. E. Smith, “Ultrafast all-optical switching with an asymmetric Fabry-Pérot devices using low-temperature-grown GaAs: Material and device issues,” IEEE J. Quantum Electron., vol. 36, no. 1, pp. 100–111, Jan. 2000. [18] S. Gupta, M. Y. Frankel, J. A. Valdmanis, J. F. Whitaker, G. A. Mourou, F. W. Smith, and A. R. Calawa, “Subpicosecond carrier lifetime in GaAs grown by molecular beam epitaxy at low substrate temperatures,” Appl. Phys. Lett., vol. 59, no. 25, pp. 3276–3278, Dec. 1991. [19] M. Born and E. Wolf, Principles of Optics, 6th ed. New York: Plenum, 1980. [20] H. A. Macleod, Thin-Film Optical Filter. New York: Macmillan, 1986. [21] E. Garmire, “Criteria for optical bistability in a lossy saturating FabryPérot,” IEEE J. Quantum Electron., vol. 25, no. 3, pp. 289–295, Mar. 1989. [22] D. B. Malins, A. Gomez-Iglesias, S. J. White, W. Sibbett, A. Miller, and E. U. Rafailov, “Ultrafast electroabsorption dynamics in an InAs quantum dot saturable absorber at 1.3 µm,” Appl. Phys. Lett., vol. 89, no. 17, pp. 171111-1–171111-3, Oct. 2006.


[23] C. Y. Jin, H. Y. Liu, K. M. Groom, Q. Jiang, M. Hopkinson, T. J. Badcock, R. J. Royce, and D. J. Mowbray, “Effects of photon and thermal coupling mechanisms on the characteristics of self-assembled InAs/GaAs quantum dot lasers,” Phys. Rev. B, vol. 76, no. 8, pp. 085315– 085326, 2007. [24] C. Y. Jin, T. J. Badcock, H. Y. Liu, K. M. Groom, R. J. Royce, D. J. Mowbray, and M. Hopkinson, “Observation and modeling of a room-temperature negative characteristic temperature 1.3 µm p-type modulation doped quantum dot laser,” IEEE J. Quantum Electron., vol. 42, no. 12, pp. 1259–1265, Dec. 2006. [25] H. Y. Liu, I. R. Sellers, M. Gutierrez, K. M. Groom, W. M. Soong, M. Hopkinson, J. P. R. David, R. Beanland, T. J. Badcock, D. J. Mowbray, and M. S. Skolnick, “Influences of the spacer layer growth temperature on multilayer InAs/GaAs quantum dot structures,” J. Appl. Phys., vol. 96, no. 4, pp. 1988–1992, Aug. 2004. [26] C. Y. Jin, S. Ohta, M. Hopkinson, O. Kojima, T. Kita, and O. Wada, “Temperature-dependent carrier tunnelling for self-assembled InAs/GaAs quantum dots with a GaAsN quantum well injector,” Appl. Phys. Lett., vol. 96, no. 15, pp. 151104-1–151104-3, 2010.

Chao-Yuan Jin received the B.Sc. degree in physics from Nanjing University, Nanjing, China, in 2000, the M.Sc. degree in microelectronics and solidstate electronics from the Chinese Academy of Sciences, Beijing, China, in 2003, and the Ph.D. degree in electronic and electrical engineering from the University of Sheffield, Sheffield, U.K., in 2008. He is currently working in the Division of Frontier Research and Technology, Center for Collaborative Research and Technology Development, Kobe University, Kobe, Japan. His current research interests include ultrafast all-optical switching using self-assembled quantum dots, 1.3-µm lasers and amplifiers based on InAs quantum dots, and GaInNAs quantum wells.

Osamu Kojima is an Assistant Professor in the Department of Electrical and Electronic Engineering, Graduate School of Engineering, Kobe University, Kobe, Japan. His current research interests include ultrafast dynamics in lowdimensional semiconductor material systems.

Tomoya Inoue received the B.E., M.E., and Ph.D. degrees in electrical and electronic engineering from Kobe University, Kobe, Japan, in 2005, 2006, and 2009, respectively. He is currently working in the Department of Electrical and Electronic Engineering, Graduate School of Engineering, Kobe University, as a Japan Society for the Promotion of Science Post-Doctoral Research Fellow. His current research interests include molecular beam epitaxial growth of selfassembled InAs quantum dots and optical amplifiers based on InAs quantum dots.

Takashi Kita received the Ph.D. degree in electrical engineering from Osaka University, Osaka, Japan, in 1991. He is currently a Professor in the Department of Electrical and Electronic Engineering, Graduate School of Engineering, Kobe University, Kobe, Japan, where he works on epitaxial growth and material physics of III–V materialbased quantum heterostructures and semiconductor nanostructures for optical amplifiers, ultrafast optical switches, new generation solar cells, and novel light sources.

1589

Osamu Wada (F’97) received the B.E. degree from the Himeji Institute of Technology, Himeji, Japan, and the M.E. degree from Kobe University, Kobe, Japan, both in electrical engineering, and the Ph.D. degree in electronic and electrical engineering from the University of Sheffield, Sheffield, U.K. He joined the Department of Electrical and Electronics Engineering, Graduate School of Engineering, Kobe University, in 2001, where he is currently a Professor. Earlier, was with Fujitsu Laboratories Ltd., Kawasaki, Japan, in 1971. From 1971 to 1976, he was engaged in research of Gunn-effect logic ICs. From 1976 to 1978, he was an independent research worker supported by the Science and Engineering Research Council of U.K. at the University of Sheffield (on leave from Fujitsu) to carry out research on InP materials and devices. He was a Research Fellow at Atsugi Laboratory, Atsugi, Japan, from 1989 to 1996. In 1996, he was engaged as a Group Leader at the Femtosecond Technology Research Association for the The Ministry of Economy, Trade and Industry/New Energy and Industrial Technology Development Ogranization project on Femtosecond Technology, in charge of the research of ultrafast all-optical switching devices. Since 1978, he has been engaged in research on various optoelectronic devices including InP-based LEDs, pin-PDs and APDs, and optoelectronic integrated circuits on both GaAs and InP material systems for their application to optical communications and optical interconnections. Since 2003, he has also been the Director of the Division of Frontier Research and Technology, Center for Collaborative Research and Technology Development (CREATE), Kobe University. Since 2010, he has been a Visiting Professor at CREATE. His current research interests include quantum dots and ultrafast photonic device applications. Prof. Wada is a Fellow of the Optical Society of America, the Institute of Electronics, Information, and Communication Engineers, and the Japan Society of Applied Physics.

Mark Hopkinson received the B.Sc. degree from the University of Birmingham, Birmingham, U.K., in 1986, and the Ph.D. degree from the University of Sheffield, Sheffield, U.K., in 1990. Currently, he is with the Department of Electronic and Electrical Engineering, University of Sheffield. Following work as a Post-Doctoral Researcher at Warwick and Sheffield Universities, U.K., he took a position of Senior Process Engineer at Marconi PLC, before returning to Sheffield for the second time to take up his current role in 2002. He was awarded a Chair in Electronic Engineering in 2007. He is the leader of a research group involved in the development of III–V epitaxial nanostructures by molecular beam epitaxy, with emphasis on novel optoelectronic devices. He has 17 years of experience in the field, with over 450 research publications and a wide range of research interactions. His current research interests include III–V quantum dot materials and novel quantum well systems, including antimonide-based structures and dilute nitride materials.

Kouichi Akahane received the B.E., M.E., and Ph.D. degrees in materials science from the University of Tsukuba, Tsukuba, Japan, in 1997, 1999, and 2002, respectively. He joined the National Institute of Information and Communications Technology, Tokyo, Japan, in 2002. He is now a Senior Researcher at the National Institute of Information and Communications Technology. His current research interests include semiconductor photonic devices.