JP3689038B2 - Semiconductor optical functional device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a semiconductor optical functional device that performs predetermined processing in the state of an optical signal, and in particular, a semiconductor optical amplifier and a nonlinear that controls the intensity, phase, or output destination of another signal light by an ultrashort optical pulse signal. The present invention relates to an optical semiconductor device.
[0002]
[Prior art]
Due to the rapid increase in the amount of information accompanying the spread of the Internet, a large amount of information of several hundred gigabits to several tens of terabits per second is expected to fly over the optical fiber in the near future in the backbone transmission system. In order to realize such an ultra-high-capacity optical network, high-performance optical amplifiers, optical functional devices such as optical switching elements capable of ultra-high-speed operation on the order of 1 Tb / s, and wavelength conversion elements are required.
[0003]
As optical amplifiers for optical communication, rare earth-doped optical fiber amplifiers and fiber Raman amplifiers are already widely used. On the other hand, introduction of traveling wave type semiconductor optical amplifiers is also being studied for applications such as optical LANs and in apparatuses. A traveling-wave semiconductor optical amplifier has a low-reflection coating on the end face of a semiconductor laser so that laser oscillation does not occur even when a high level of current is injected. In response to the stimulated emission gain, it is amplified and output. Traveling wave type semiconductor optical amplifiers are much smaller and consume less power than optical fiber amplifiers. In addition, the degree of freedom in wavelength is greater than that of rare-earth-doped optical fiber amplifiers.
[0004]
However, the fact that ASE (Amplified Spontaneous Emission) noise output over the entire gain band deteriorates the signal light-to-noise ratio is a constraint in application. There is also a problem that a sufficient output cannot be obtained because the gain saturation level is low. For this reason, the application range in which the semiconductor optical amplifier can be used as an optical amplifying element has been extremely limited.
[0005]
Semiconductor optical amplifiers are expected to be applied as nonlinear optical devices such as optical switches and wavelength conversion elements rather than as optical amplifiers. If a strong light pulse is incident, gain saturation and refractive index modulation occur, so that it can be used as various nonlinear optical devices. A nonlinear optical device using an optical amplifier having a gain is characterized by low switching power and high efficiency. However, since the gain change based on interband transition is limited by the carrier lifetime, the response speed is usually about several tens of Gb / s, and even when special measures such as assist light injection are used, it operates only up to several hundred Gb / s. I couldn't let you.
[0006]
On the other hand, the use of saturable absorption of transition between subbands of nitride semiconductors has been proposed as an operating principle of a 1 Tb / s class ultrahigh-speed optical switching element (for example, JP-A-8-179387, N. Suzuki and N. Iizuka, Japanese Journal of Applied Physics, Vol. 36, pp. L1006-1008, 1997). The relaxation time between subbands is about three orders of magnitude faster than the relaxation time between carriers (carrier life). In particular, if nitride semiconductor quantum wells are used, intersubband transitions in the near-infrared region, which are important for optical communications, can be realized, and the interaction between LO phonons and electrons is large, so the intersubband relaxation time can be reduced to that of InGaAs. Compared to one digit faster.
[0007]
For example, the intersubband relaxation time in a GaN quantum well is as short as 200 to 400 fs at room temperature and in the optical communication wavelength band. Therefore, if the saturation of the intersubband absorption is used, an ultra-high repetition bit of about 1 Tb / s.・ Mode light switching operation is also possible. In addition, since a semiconductor is used, it is needless to say that a small, light and stable optical switch can be realized and mass production is possible. Further, nitride semiconductors are particularly tough semiconductor materials and are resistant to temperature rise.
[0008]
Intersubband absorption occurs for the electric field component perpendicular to the well, so it is difficult to apply it to a planar optical switch in which light is incident perpendicularly to the substrate surface, and is mainly applied to optical waveguide nonlinear optical devices. . Various forms of optical waveguide type nonlinear optical devices are conceivable. For example, in the case of a saturable absorption optical gate switch using the saturation of intersubband absorption, signal light with a pulse energy that does not cause absorption saturation. A control light pulse having a pulse and a pulse energy large enough to cause absorption saturation is incident on the optical waveguide. Both pulses are assumed to be in a wavelength range where intersubband absorption occurs. When only the signal light pulse is incident, the signal light pulse is absorbed and not output. When the signal light pulse and the control light pulse are incident at the same time, the signal light is output because the saturable absorption by the control light reduces the absorption of the signal light pulse.
[0009]
In order to obtain a sufficient extinction ratio (switch ON / OFF ratio) with such an optical gate switch, a control light pulse energy of the order of a few pJ is necessary in the case of GaN, which is 1 THz. This corresponds to an average power of several W. Usually, the control light pulse is also a signal modulated with a certain pattern, and it is difficult to amplify it to such power.
[0010]
If the wavelength of the signal light is set at the end of the absorption band, it can also be used as an optical phase modulation element whose phase is modulated by saturable absorption by the control light. By incorporating this phase modulation element into an optical interferometer, it can be applied as an interferometer type optical gate switch or optical path switching switch. The present invention can also be applied to a four-wave mixing type optical switch, a wavelength conversion element, or the like that uses a change in intersubband carrier distribution due to beat components of light of two wavelengths in the absorption band. However, in any case, the need for strong control light and excitation light is the biggest problem that hinders practical use.
[0011]
As a method for realizing a nonlinear optical device having both the ultra-high speed of the intersubband transition and the high efficiency of the semiconductor optical amplifier, it is conceivable to use an optical amplifier using the intersubband transition. At a wavelength of 3 μm or more, a quantum cascade laser (QCL) is known as a semiconductor laser using intersubband transition. In QCL, an active layer in which a plurality of unit structures each composed of an electron injection region and an active region are stacked is provided in an optical waveguide structure so that carriers can be injected into the active layer by applying a voltage from the outside. QCL is disclosed, for example, in US Pat. No. 5,457,709 (inventor Capasso et al., Filing date March 27, 1997).
[0012]
FIG. 14 shows an example of the conduction band structure of the unit structure 100 constituting the active layer of the QCL. The unit structure 100 includes an electron injection region 101 and an active region 102, and both the electron injection region 101 and the active region 102 are formed of a quantum well or a superlattice including an InGaAs well layer 103 and an AlGaInAs barrier layer 104. The electron injection region 101 is composed of a chirped semiconductor superlattice layer, and a miniband 105 as shown is formed. The active region 102 is usually composed of a multiple quantum well layer or a semiconductor superlattice layer, but here it is represented by a single quantum well to simplify the discussion. The active region 102 has a ground level E ₁ , Excitation level E ₂ The two subbands are formed.
[0013]
A sufficient electric field is applied to the active layer via the electrode and the cladding layer. The electrons are excited from the left side of the figure through the miniband 105 of the electron injection region 101 and excited level E of the active region 102. ₂ Injected into. The ground level E of the active region 102 ₁ Time constant discharged from the right side to the electron injection region 101 and the excitation level E from the miniband 105 ₂ Any of the time constants in which electrons are injected into ₂ To E ₁ Therefore, if a sufficient current is passed, an inversion distribution can be formed in the active region 102, and a stimulated emission gain is generated. Laser oscillation can be obtained by forming a resonator by applying a highly reflective coating film to the end face.
[0014]
In principle, if low-reflection coating is applied to the end face of the QCL to suppress laser oscillation, it should be usable as an optical amplifier. By applying this, an ultrafast and highly efficient nonlinear optical semiconductor device can be realized. It should be. In the GaN / AlN system, InGaAs / AlAsSb system, ZnSe / BeTe system, etc., near-infrared wavelength intersubband transition is realized even in the wavelength band of 1.3 to 1.6 μm used for optical communication. However, it has been extremely difficult to realize an optical amplifier using the QCL or the QCL structure at a near infrared wavelength for the following reason.
[0015]
Since the barrier layer for realizing the near-infrared intersubband transition tends to have a high barrier height and a large effective mass, the coupling due to tunneling is weak, and the time constant for electrons to propagate through the superlattice becomes long. Moreover, in the near-infrared wavelength intersubband transition, E ₂ Therefore, the energy level of the miniband 105 in the electron injection region 101 needs to be sufficiently high in the region in contact with the active region 102. However, the thickness of the well layer has a limit of one monolayer, and such a level cannot be realized unless the composition of the well layer in this portion is brought close to the composition of the barrier layer.
[0016]
A semiconductor layer having a composition close to that of the barrier layer material also has disadvantages such as an inability to increase the carrier density, a large effective mass, and a low mobility. Furthermore, the level of the miniband is changed to E ₁ E on the right side of the active region ₂ In order to increase the thickness of the electron injection region, the thickness of the electron injection region must be increased (the number of wells constituting the superlattice is very large). For this reason, the electron transport time in the electron injection region becomes longer, and E ₂ Electron injection time and E ₁ Will no longer meet the precondition that the time for electron emission from is short. Therefore, in the near infrared region, it is very difficult to form an inversion distribution between subbands.
[0017]
Further, since the voltage drop per unit structure is large (the voltage drop of photon energy + alpha is necessary, the voltage drop is close to 1 V per unit structure), and therefore a very high voltage is required as a whole. Furthermore, a thick electron injection layer means that the overlap between the active region and the guided light in the active layer is reduced. Since near-infrared light has a smaller waveguide mode diameter than mid-infrared light, the thickness of the active layer cannot be increased so much. Therefore, it becomes difficult to secure the number of stacked unit structures necessary to obtain a sufficient gain.
[0018]
As described above, the optical amplifier using the QCL structure cannot obtain a sufficient gain in the near-infrared region, and a required characteristic cannot be obtained even as a nonlinear optical semiconductor device.
[0019]
[Problems to be solved by the invention]
As described above, the semiconductor optical amplifier of the prior art using interband transition has a very limited application range because it has a large ASE noise and a low gain saturation level. In addition, a nonlinear optical semiconductor device using a conventional semiconductor optical amplifier cannot operate at an ultrahigh speed of 1 Tb / s or higher. On the other hand, a nonlinear optical device using intersubband transition such as GaN has a problem of high efficiency and low switching energy although it is high speed and wide band.
[0020]
In the optical amplifier using intersubband transition, which is a combination of the prior art, there is a problem that sufficient gain cannot be obtained in the near infrared region, and it is possible to obtain ultra-high speed and high efficiency characteristics as a nonlinear optical semiconductor device. could not.
[0021]
The present invention has been made in view of the above circumstances, and its object is to realize a semiconductor optical amplifier having a low noise and a large saturation level, a nonlinear semiconductor optical device having a very high speed and a low switching energy, and the like. An object of the present invention is to provide a semiconductor optical functional device.
[0022]
[Means for Solving the Problems]
(Constitution)
In order to solve the above problems, the present invention adopts the following configuration.
[0023]
That is, the semiconductor optical functional device of the present invention has a first quantum well (light amplification region) having at least two subbands and at least two subbands, and the energy difference between the subbands is the first quantum well. A second quantum well (electronic excitation region) larger than the energy difference between the sub-bands of the quantum well, and at least two quantum wells Is a barrier A unit-coupled quantum well structure coupled by layers is a semiconductor multilayer structure in which a plurality of layers are periodically stacked, an optical waveguide structure including the semiconductor multilayer structure, and a voltage in the stacking direction with respect to the semiconductor multilayer structure. Apply Means for injecting light from one end side of the optical waveguide structure, means for emitting light from the other end side of the optical waveguide structure, and intersubband absorption of the second quantum well in the semiconductor multilayer structure Means for injecting excitation light in a band, When a voltage is applied to the semiconductor multilayer structure in the stacking direction and excitation light in the intersubband absorption band of the second quantum well is injected, the second of the unit coupled quantum well structure The tunnel flows through the barrier layer from the excitation level of the quantum well, and carriers flow to the first quantum well of the unit coupled quantum well adjacent to the downstream side of the unit coupled quantum well structure. It is characterized by that.
[0024]
Further, preferred embodiments of the present invention include the following.
(1) The semiconductor multilayer structure is made of a nitride semiconductor.
(2) The light incident on the optical waveguide structure is signal light within the intersubband transition band of the first quantum well. This constitutes a semiconductor optical amplifier.
(3) The light incident on the optical waveguide structure is signal light and control light in the intersubband transition band of the first quantum well or the second quantum well. This constitutes a nonlinear optical semiconductor device.
(4) The ground level of the first quantum well is higher than the ground level of the second quantum well and is higher than the energy corresponding to the LO phonon energy of the material constituting the quantum well.
(5) The ground level of the first quantum well and the ground level of the second quantum well are strongly coupled, and the energy difference is 30 meV or more.
(6) The intersubband transition wavelengths of the first quantum well and the second quantum well are both 2 μm or less.
[0025]
(7) The first quantum well has a step-like or continuous composition gradient region on the opposite side to the second quantum well.
(8) The third quantum well is provided on the opposite side of the first quantum well from the second quantum well via another barrier layer thinner than the barrier layer between the first and second quantum wells. That.
(9) The excitation level of the second quantum well is a pseudo-binding level that is coupled by tunneling to an adjacent layer corresponding to the downstream side of the carrier flow.
(10) An electric field of 1 MV / cm or more is applied to the layer between the unit coupled quantum well structure and the unit coupled quantum well structure adjacent thereto.
[0026]
(11) The excitation light is continuous light.
(12) The excitation light is pulsed light and is incident prior to the arrival of control light and signal light.
(13) In the embodiment (12), the excitation light is introduced from the optical waveguide structure.
(14) Excitation light should be introduced from the side of the optical waveguide structure.
[0027]
Strictly speaking, in the coupled quantum well, a subband is formed across both the first quantum well and the second quantum well. Here, in order to simplify the discussion, subbands (base level, excitation level) in which the carrier existence probability of the first (second) quantum well is larger than the carrier existence probability of the second (first) quantum well. Is referred to as a “subband (base level, excitation level) of the first (second) quantum well”. When the first (second) quantum well has a superlattice structure, “subband” is read as “miniband”.
[0028]
(Function)
In the configuration of the present invention, a majority of carriers are distributed in the ground level of the second quantum well in a state where a voltage is applied and no excitation light is incident. When excitation light with sufficient intensity is incident, the carriers are excited to the excitation level of the second quantum well. The excited carriers are injected into the excited level of the first quantum well of the adjacent unit coupled quantum well structure in a short time due to the potential drop due to the applied voltage.
[0029]
The excitation level of the first quantum well is at a lower level than the excitation level of the second quantum well. Therefore, the lifetime of the excited level of the first quantum well is governed by the relaxation process to the ground level accompanied by the emission of LO phonon. The relaxation rate is the relaxation rate from the excitation level of the second quantum well to the excitation level of the first quantum well in the transition between subbands having a wavelength of 2 μm or less, and from the ground level of the first quantum well. It is smaller than any of the relaxation rates to the ground level of the quantum well of 2.
[0030]
Therefore, if the carrier is injected by applying a voltage and the excitation rate of the second quantum well by the excitation light is sufficiently large, the lifetime of the excitation level of the first quantum well becomes the longest, and the base of the first quantum well is obtained. An inversion distribution is formed between the level and the excitation level. As a result, a stimulated emission gain is generated in a wavelength band corresponding to the intersubband transition energy of the first quantum well. If signal light within that wavelength band is incident, it is amplified and emitted. The saturation gain can be increased by increasing the number of unit coupled quantum well structures and entering strong excitation light. Further, since spontaneous emission hardly occurs at the transition between subbands, the generation of ASE noise can be ignored and the noise is extremely low. (However, since the noise of the pumping light is picked up, pumping light with a small noise is required.) Since the transition energy between subbands depends on the well width, etc., the near infrared can be changed by changing the design of the well structure. To optical amplifiers of various wavelengths in the mid-infrared region.
[0031]
In addition, in the case of operating as a nonlinear optical device as in the embodiment (3), various combinations of the wavelengths of the signal light and the control light are conceivable depending on how to operate.
[0032]
For example, the control light is within the gain band of the optical amplification region and outside the absorption band of the electron excitation region, and the signal light is within the wavelength range where the gain band of the optical amplification region and the absorption band of the electron excitation region overlap. To do. In this case, when a strong control light pulse is not incident, the signal light pulse is affected by both gain and absorption when propagating through the optical waveguide structure. If the gain is set to be larger than the absorption, an output larger than the input can be obtained. When a control light pulse with sufficient pulse energy to saturate the gain in the optical amplification region propagates through the optical waveguide in synchronization with the signal light pulse, the gain saturation suppresses the amplification of the signal light and the absorption is greater. Therefore, the signal light is attenuated as the optical waveguide propagates and is not output. In this case, the inverter operation is performed with the sign reversed.
[0033]
If the relationship between gain and absorption is reversed by changing the wavelength arrangement, it can also be applied to absorption saturation type non-inverting operation. Even in this case, since the gain of the optical amplification region contributes to the control light or signal light, high efficiency or low switching energy can be achieved.
[0034]
Thus, the transmittance of the optical waveguide with respect to the signal light pulse can be changed depending on the presence or absence of the control light pulse. Alternatively, the phase of the signal light pulse can be changed along with this change. Since all the phenomena related to this change are high-speed, ultra-high speed optical control is realized. Unlike conventional nonlinear optical semiconductor devices that use only saturable absorption of intersubband transitions, there is gain for control light, signal light, or both, so switching energy can be reduced with high efficiency.
[0035]
In the nitride semiconductor of the embodiment (1), the intersubband relaxation time is several hundred fs even in the near infrared wavelength region, and the relaxation time between adjacent subbands corresponding to LO phonon energy or the intrasubband relaxation time. Is as short as several tens of fs. The carrier-carrier scattering time is also on the order of several tens of fs, which is one digit shorter than that of a GaAs system or the like. Accordingly, if the injection current and the excitation light intensity are sufficiently large, the carrier is replenished immediately from the adjacent level even if the excitation level of the carrier is insufficient due to the optical amplification, so that the gain saturation level is high. Also, the gain saturation level can be increased by increasing the number of unit quantum well structures. However, the greater the number of unit quantum well structures, the greater the required applied voltage and excitation light intensity. When applied to nonlinear optical devices, since the time constants of the related relaxation processes are all sub-picoseconds, an ultrafast optical switch of 1 terabit per second or more can be realized.
[0036]
In the case of the embodiment (4), the electrons in the ground level of the first quantum well are in the ground level of the second subband in a short time (about 100 fs in the nitride semiconductor) with LO phonon emission. To relax.
[0037]
In the state in which the two quantum wells are strongly coupled as in the embodiment (5), the squares of the envelope functions of the two base subbands are both applied to both the first quantum well and the second quantum well. Has a peak. Due to the phase difference between the electrons in the two subbands, a state in which carriers mainly exist in the first quantum well and a state in which carriers exist in the second quantum well are generated. The carrier is present in the first quantum well at the moment when it falls from the excitation subband of the first quantum well to the two base subbands by stimulated emission, but is determined by the energy difference ΔE between the two levels that resonate. It moves to the ground level of the second subband with a time constant that is half the period of the quantum beat (h / ΔE). Actually, the quantum beat is attenuated by the order of the phase relaxation time (10-30 fs for nitride semiconductor, 100 fs for semiconductors containing As and P as main constituent elements), but the carrier movement speed between wells is almost at this time. Can be estimated from a constant. If the coupling between wells is strong and ΔE is 30 meV or more, this time constant is at most 100 fs.
[0038]
The lifetime of the excitation level of the first quantum well becomes longer as the wavelength is shorter. Accordingly, when the intersubband transition wavelength is 2 μm or less as in the embodiment (6), it is easy to set the lifetime longer than other related time constants, that is, to form an inverted distribution.
[0039]
As in the embodiment (7), the first quantum well can be easily formed by providing a stepped or continuous composition gradient region (intermediate composition layer) on the opposite side of the first quantum well from the second quantum well. Can be made smaller than the intersubband energy of the second quantum well.
[0040]
Also in the structure of the embodiment (8), since the first quantum well and the third quantum well are strongly coupled, a subband whose origin is the ground level of the third quantum well is the first quantum well. This is the lowest excitation level of a coupled quantum well composed of a well and a third quantum well. If this coupled quantum well is regarded as the original first quantum well, the intersubband energy of the first quantum well (the coupled quantum well composed of the first quantum well and the third quantum well) can be easily determined as the first quantum well. It can be set smaller than the energy between subbands of the quantum well of 2.
[0041]
If the excitation level of the second quantum well is a quasi-bound level coupled by tunneling to the adjacent layer corresponding to the downstream side of the carrier flow as in the embodiment (9), the excitation level of the second quantum well The time for injecting the carriers excited in the first quantum well of the adjacent unit-coupled quantum well into the first quantum well can be shortened.
[0042]
Even if an electric field of 1 MV / cm or more is applied to the layer between the unit coupled quantum well structure and the unit coupled quantum well structure adjacent thereto as in the embodiment (10), the excitation level of the second quantum well is It is possible to shorten the time for injecting the carriers excited to the position to the excitation level of the first quantum well of the adjacent unit-coupled quantum well.
[0043]
As in the embodiment (11), when the excitation light is CW light, stable operation with small time fluctuation is possible.
[0044]
If the excitation light is pulsed light as in the embodiment (12), the average power of the excitation light can be kept small. However, in order to form an inversion distribution before the signal light pulse or the control light pulse arrives, it is necessary that the excitation light pulse arrives before the control light pulse or the signal light pulse. If the excitation light is synchronized with the control light pulse and is a periodic light pulse having a width wider than the control light pulse, the average power of the excitation light can be suppressed lower than in the case of continuous excitation, and A sufficient gain can be obtained when the control light pulse or the signal light pulse arrives.
[0045]
When the excitation light is pulsed light, if the excitation light pulse is introduced from the optical waveguide as in the embodiment (13), the pulse width can be suppressed to be slightly wider than the control light pulse, which is efficient.
[0046]
As in the embodiment (14), a method of introducing excitation light from the side surface of the optical waveguide is also conceivable. In this case, the second quantum well can be excited almost uniformly over the entire optical waveguide, and the optical waveguide can be lengthened to increase the interaction length between the control light pulse and the signal light pulse. The excitation light only needs to be irradiated while the control light pulse or the signal light pulse is propagating through the optical waveguide. If this condition is satisfied, the excitation light may be pulsed light or continuous light.
[0047]
DETAILED DESCRIPTION OF THE INVENTION
The details of the present invention will be described below with reference to the illustrated embodiments.
[0048]
(First embodiment)
FIG. 1 is a perspective view schematically showing main components of the semiconductor optical amplifier according to the first embodiment of the present invention. The optical semiconductor optical amplifier 20 includes a lower nitride semiconductor layer 2 stacked on a sapphire substrate 1 (0001) surface, a semiconductor stacked structure 3 made of a nitride semiconductor on the lower nitride semiconductor layer 2, and an upper nitride semiconductor. The mesa-shaped optical waveguide 5 in which the layers 4 are sequentially laminated, the polyimide film 6 formed on the lower nitride semiconductor layer 2 in contact with the optical waveguide 5, and the upper portion formed on the polyimide film 6 and the optical waveguide 5. The upper electrode 7 is electrically connected to the nitride semiconductor layer 4 and the lower electrode 8 is electrically connected to the lower nitride semiconductor layer 2. The element length is about 400 μm.
[0049]
The nitride semiconductor layers 2, 3 and 4 are wurtzite type, and the surface is a group III polar surface here. Depending on the conditions of crystal growth, a minute polarity reversal domain may be formed, but there is no significant effect as long as the group III polarity plane is dominant. In addition, when creating with the group V polarity, the direction of current and electric field may be reversed in the following description. There are many other defects such as dislocations in the nitride semiconductor layer. However, in a device using intersubband transition as in the present invention, a non-radiative recombination center in a light emitting device using interband transition is used. Such a fatal effect is not brought about. In the following explanation, it is assumed that most of the quantum wells are fabricated with fluctuations of about ± 1 monolayer, but the existence of the above-described crystal defects that do not substantially affect the operation and effect of the invention. Will be ignored.
[0050]
A negative voltage is applied to the upper electrode 7 relative to the lower electrode 8, and as a result, the upper nitride semiconductor layer 4, the nitride semiconductor multilayer structure 3, and the lower nitride semiconductor layer 2 of the optical waveguide 5 are formed from the upper electrode 7. Through this, electrons flow to the lower electrode 8.
[0051]
The light 9 incident on the optical waveguide 5 is signal light having a wavelength of 1.45 μm. Outgoing light 10 is extracted from the other end of the optical waveguide 5. The lower nitride semiconductor layer 2 is a slab optical waveguide. Low noise excitation light 11 having a wavelength of 0.98 μm is incident from the right side surface of FIG. 1 and guided to the optical waveguide 5. All the light is polarized in a direction perpendicular to the substrate (well layer) in which intersubband transition is allowed.
[0052]
FIG. 2 is a diagram schematically showing a cross-sectional layer structure of the semiconductor optical amplifier 20. The lower nitride semiconductor layer 2 includes an undoped AlN layer 22 having a thickness of 0.15 μm, a first n-type GaN layer 23 having a thickness of 0.3 μm, and a first AlGaN composition gradient layer 24 having a thickness of 0.05 μm. Become. The first AlGaN composition gradient layer 24 is formed so as to change stepwise from the composition close to GaN to the composition close to AlN from the bottom to the top. On the first AlGaN composition gradient layer 24, a nitride semiconductor multilayer structure 3 and an upper nitride semiconductor layer 4 are laminated in a mesa shape, and the ridge-type optical waveguide 5 is defined by this mesa structure. . The mesa width of the optical waveguide 5 is 1.5 μm. The upper nitride semiconductor layer 4 includes a second AlGaN composition gradient layer 25 having a thickness of 0.05 μm whose composition changes stepwise from the composition close to AlN to the composition close to GaN from the bottom to the top, and the thickness The second n-type GaN layer 26 is 0.1 μm. The n-type GaN layers 23 and 26 have Si of 1 × 10 ²⁰ cm ^-3 It is doped to some extent.
[0053]
The semiconductor multilayer structure 3 is formed by stacking 30 coupled quantum well structures 31 having substantially the same structure (a slight variation caused by crystal growth is allowed). FIG. 3 is a diagram for explaining a conduction band structure for three periods of the coupled quantum well structure 31. In Group III polarity, the left side is the front side and the right side is the substrate side. The reverse is true for group V polarity. This coupled quantum well structure 31 is an undoped Al layer formed between two first undoped AlN barrier layers 32 and 32R and having a thickness of about 0.75 nm. _0.65 Ga _0.35 An N-well intermediate layer 33, a first n-type GaN well layer 34 having a thickness of about 0.75 nm, a second undoped AlN barrier layer 35 having a thickness of about 1 nm, and a second n-type GaN well having a thickness of about 1 nm It consists of layer 36. The thickness of the first undoped AlN barrier layer 32 is 1.8 nm, and the thickness of one period is 5.3 nm. In the n-type GaN well layers 34 and 36, Si is 5 × 10 ¹⁹ cm ^-3 It is doped to some extent.
[0054]
Left first undoped AlN barrier layer 32, undoped Al _0.65 Ga _0.35 The first quantum well 37 including the N well intermediate layer 33, the first n-type GaN well layer 34, and the second undoped AlN barrier layer 35 serves as an optical amplification region. The second quantum well 38 including the second undoped AlN barrier layer 35, the second n-type GaN well layer 36, and the first undoped AlN barrier layer 32R on the right side serves as an electron excitation region. The first undoped AlN barrier layer 32R also functions as an electron injection region for injecting electrons from the coupled quantum well structure 31 on the left to the adjacent coupled quantum well structure 31R on the right.
[0055]
A large electric field is generated in the wurtzite nitride semiconductor heterostructure whose principal surface is the (0001) plane due to the piezoelectric effect and spontaneous polarization. This electric field is opposite for AlN and GaN. Further, an external voltage is applied between the upper electrode 7 and the lower electrode 8, and the first subband E of each of the first quantum well 37 and the second quantum well 38 is applied. ₂ , E ₁ Is adjusted so as to bind appropriately. At this time, AlN barrier layers 32 and 35, Al _0.65 Ga _0.35 The electric fields applied to the N well intermediate layer 33 and the n-type GaN well layers 34 and 36 are approximately −4.85 MV / cm, approximately −1 MV / cm, and approximately 6.15 MV / cm, respectively.
[0056]
As a whole, the potential of electrons in the coupled quantum well on the left side of FIG. 3 is higher, and the electrons flow from the left side to the right side of the figure. The voltage drop per cycle of the coupled quantum well 31 is about 0.35 V, and the voltage drop is about 10 V in 30 cycles.
[0057]
FIG. 3 shows the subband E formed in the coupled quantum well structure 31. ₁ , E ₂ , E _Three , E _Four And the shape of the square of each envelope function is schematically shown. Subband E ₁ And subband E ₂ Are strongly coupled, and both are ground levels in which electrons are distributed to the two quantum wells 37 and 38 with almost equal probability (hereinafter, E ₁ And E ₂ Are collectively considered as the “base level” common to the first quantum well and the second quantum well. ). Subband E ₁ And E ₂ The energy difference is about 30 meV (there is some variation due to nonuniformity depending on the location), and the quantum beat period is about 140 fs. Subband E _Three Is mainly an excitation level bound to the first quantum well (hereinafter referred to as “excitation level of the first quantum well”). Subband E _Four Is an excitation level weakly bound to the second quantum well (hereinafter referred to as “excitation level of the second quantum well”), and the three-dimensional quantum of the first undoped AlN barrier layer 32 on the right side. It is connected by state and tunneling.
[0058]
Ground level E ₁ , E ₂ And the excitation level E of the first quantum well _Three The transition wavelength between and is about 1.1 μm, the ground level E ₁ , E ₂ And the excitation level E of the second quantum well _Four The transition wavelength between and is about 1.41 μm. (The calculated values of the transition wavelength under the assumption that the composition change at the heterointerface is steep are about 1 μm and 1.3 μm, respectively. However, in the quantum well with a thin well layer, the calculated value is affected by the composition fluctuation of the interface. There is a tendency that the wavelength becomes longer.) The transition energy between subbands varies slightly depending on the location due to fluctuations in film thickness, etc., which causes the absorption spectrum width to be widened.
[0059]
In the present embodiment, electrons are injected from the second n-type GaN layer 26 into the first coupled quantum well layer 31 (corresponding to the upper end in FIGS. 1 and 2 and the left end in FIG. 3) of the semiconductor multilayer structure 3. This is realized by the composition gradient of the AlGaN composition gradient layer 25 and the potential gradient caused by the electric field. Of course, instead of using the second AlGaN composition gradient layer 25, its base subband (E ₁ Corresponding to the second quantum well coupled to the adjacent n-type GaN layer 26, and a quantum well having a structure similar to that of the second quantum well is provided. _Four Electron injection can also be performed by excitation.
[0060]
Excitation level E of the coupled quantum well structure at the right end of FIG. _Four Has a higher potential for electrons than the n-type GaN layer 23, and therefore the first AlGaN composition gradient layer 24 is unnecessary from the viewpoint of electron injection. In the present embodiment, the first AlGaN composition gradient layer 24 acts as a cladding of a slab optical waveguide that guides excitation light.
[0061]
Electrons injected from the first undoped AlN barrier layer 32 on the left side by the electric field into the first quantum well 37 have excited levels E _Three To be captured. Excitation level E _Three To ground level E ₁ And E ₂ The relaxation time with LO phonon emission into is about 500 fs. E ₁ And E ₂ Since the square of the envelope function of is almost the same shape, E ₁ And E ₂ Relax with almost equal probability. E ₁ And E ₂ Electrons cause so-called quantum beats that go back and forth between two quantum wells. E ₁ And E ₂ The period of the quantum beat (h / ΔE, where h is Planck's constant) when the energy difference ΔE of the first quantum well 37 is about 140 fs, the electrons relaxed in the first quantum well 37 are almost the second quantum after 70 fs. It has moved to the well 38.
[0062]
This electron is immediately excited by the excitation light E introduced into the optical waveguide 5 through the slab optical waveguide, and the excitation level E of the second quantum well. _Four Excited. Excitation level E _Four Are coupled to the three-dimensional quantum state of the first undoped AlN barrier layer 32R on the right side. Excitation level E _Four Since the effective barrier with respect to is low, the excited electrons escape to the first undoped AlN barrier layer 32R during about 100 fs, and excitation of the first quantum well 37R of the adjacent coupled quantum well 31R on the right side by a strong electric field Level E _Three Injected into.
[0063]
Excitation level E _Four Some of the electrons excited by the excitation level E of the first quantum well 37 _Three To relax. Further, some of the electrons injected from the first undoped AlN barrier layer 32 directly contribute to the excitation level E of the second quantum well 38. _Four To be captured. However, since the total thickness (159 nm in 30 periods) of the semiconductor multilayer structure 3 is sufficiently thin compared to the waveguide mode size (˜μm), the difference in light intensity for each coupled quantum well structure is small. The movement of electrons in the well structure 31 is balanced. Therefore, as a whole, electrons flow from the coupled quantum well structure on the left side to the coupled quantum well structure on the right side, and electrons are not accumulated or depleted in a specific coupled quantum well structure.
[0064]
When sufficiently strong excitation light is incident, the ground level E ₁ , E ₂ To excitation level E _Four And the excitation level E of the right first quantum well 37R. _Three The rate at which electrons are injected into each of these is from the excitation level E3 to the ground level E3. ₁ , E ₂ Greater than the rate at which electrons relax. Therefore, the excited level E _Three The electron density of the ground level E ₁ , E ₂ A state of inversion distribution greater than the electron density is created. As a result, the first quantum well (light amplification region) 37 has a gain in the vicinity of a wavelength of 1.4 μm. The light absorption of the second quantum well (electron excitation region) 38 is in a state where the absorption is partially saturated by the excitation light.
[0065]
FIG. 4 is a diagram for explaining the relationship among the gain spectrum of the optical amplification region 37, the absorption spectrum of the electron excitation region 38 when the excitation light is incident, and the wavelength arrangement of the incident light 9 and 11. FIG. The excitation light is absorbed by the electron excitation region 38 but is not absorbed by the light amplification region 37. The signal light is hardly absorbed by the electron excitation region 38 and is amplified by the gain of the optical amplification region 37. If the pumping light power and the injection current are sufficiently large and the level of the input signal light is sufficiently low, gain saturation does not occur. Since the spontaneous emission probability is very low at the transition between subbands, the generation of ASE noise can be ignored, and optical amplification with extremely low noise is realized.
[0066]
Since the voltage drop in each coupled quantum well 31 is suppressed to be small by utilizing photoexcitation, the applied voltage can be suppressed to be relatively small. (Even if such a function can be realized with a structure similar to a quantum cascade laser that does not perform optical excitation, the required voltage is more than three times.) Compared to a quantum cascade laser similar structure Since the thickness of one period can be reduced, the overlap with the optical field of the waveguide mode is large, and the influence of non-uniformity of the light intensity in the thickness direction is small, which is efficient.
[0067]
The semiconductor optical amplifier of this embodiment can be variously modified. For example, in order not to cause gain saturation on the output end side where the signal light power is increased, the pump light power may be adjusted to be increased on the output end side. Alternatively, in addition to the incident 11 from the side surface, excitation light can be injected from the exit end side. When the signal light is pulsed light, the average power of the pumping light can be reduced by using the pumping light as pulsed light synchronized with the signal light pulse. In this case, it is necessary to make the excitation light pulse incident at a timing that completely covers the pulse signal light propagating through the optical waveguide 5. If the design such as the well width is changed, the pumping light wavelength and the signal light wavelength can be changed.
[0068]
(Second Embodiment)
The semiconductor optical amplifier according to the first embodiment can be used as it is as a nonlinear semiconductor optical device (gain saturated optical gate switch). Except for the input light, the structure, band structure, function, etc. of each part are the same as those of the semiconductor optical amplifier of the first embodiment, and the description of the structure is omitted.
[0069]
The light 9 incident on the optical waveguide 5 of the nonlinear semiconductor optical device according to the second embodiment has a wavelength of 1.48 μm, a pulse width of about 200 fs, a repetitive 160 Gb / s control light pulse, a wavelength of 1.3 μm, and a pulse width of about This is a signal light pulse of 100 fs and 160 Gb / s repeatedly. The lower nitride semiconductor layer 2 is a slab optical waveguide. An excitation light pulse 11 having a wavelength of 0.98 μm, a width of about 3 ps, and a repetition of 160 GHz is incident from the right side surface of FIG. All the light is polarized in a direction perpendicular to the substrate (well layer) in which intersubband transition is allowed.
[0070]
When a control light pulse is not input to the optical waveguide 5 and a sufficiently strong excitation light intensity is incident, the ground level E ₁ , E ₂ To excitation level E _Four And the excitation level E of the right first quantum well 37R. _Three The rate at which electrons are injected into the excitation level E _Three To ground level E ₁ , E ₂ Greater than the rate at which electrons relax. Therefore, the excited level E _Three The electron density of the ground level E ₁ , E ₂ A state of inversion distribution greater than the electron density is created. As a result, the first quantum well (light amplification region) 37 has a gain in the vicinity of a wavelength of 1.4 μm. The light absorption of the second quantum well (electron excitation region) 38 is in a state where the absorption is partially saturated by the excitation light.
[0071]
This nonlinear optical semiconductor device operates at 160 Gb / s. The control light pulse is incident at a timing that overlaps the signal light pulse. The excitation light pulse is incident at a timing such that it overlaps the control light pulse at any position of the optical waveguide 5. The pulse width of the pumping light is 3 ps, which is wider than the time (about 2.8 ps) in which the signal light pulse and the control light pulse propagate through the optical waveguide 5 having a length of 400 μm. The control light pulse and the signal light pulse are modulated, but the excitation light pulse is periodically incident every 6.25 ps.
[0072]
FIG. 5 is a diagram for explaining the relationship among the gain spectrum of the optical amplification region 37, the absorption spectrum of the electron excitation region 38 when the excitation light is incident, and the wavelength arrangement of the incident light 9 and 11. The excitation light is absorbed by the electron excitation region 38 but is not absorbed by the light amplification region 37. The control light pulse is hardly absorbed by the electron excitation region 38. Although the front edge of the control light pulse is amplified near the incident end, the back side of the control light pulse is hardly amplified or attenuated due to gain saturation. Therefore, the pulse width of the control light is narrowed and the spectrum is broadened as it propagates through the optical waveguide 5. However, since the front edge of the control light pulse is amplified at any position, the control light pulse is amplified as a whole and propagates to the back of the optical waveguide 5. Further, the effect of gain saturation that follows is derived from the adjacent left coupled quantum well structure by E _Three The time constant in which electrons are supplied to the first quantum well and the time constant in which the electrons are ejected from the ground level of the first quantum well by the quantum beat and excitation light (both around 100 fs) continue. Therefore, the gain for the signal light pulse can be saturated in the entire optical waveguide 5.
[0073]
When there is no control light pulse, the signal light pulse is affected by both the gain of the optical amplification region 37 and the loss of the electron excitation region 38, but is amplified and emitted because the former is larger. The amplification factor strongly depends on the injection current and the intensity of the excitation light. In order to realize high-efficiency amplification, it is necessary to balance the injection current level and the excitation light level well. However, unlike an optical amplifier using interband transition, spontaneous emission hardly occurs in intersubband transition, so ASE noise in the gain spectrum band can be ignored.
[0074]
On the other hand, when a control light pulse is incident, E _Three The electrons accumulated in the ground level E ₁ , E ₂ Since the optical amplification is performed by falling to, the gain of the optical amplification region 37 is saturated. When the control light pulse is strong, the light is instantaneously transferred to the loss side, but the gain is small on average. Accordingly, the signal light pulse is absorbed while propagating through the optical waveguide 5 due to transient absorption in the optical amplification region 37 and absorption in the electron excitation region 38, and is not output from the optical waveguide 5.
[0075]
According to the nonlinear optical semiconductor device of this embodiment, since at least the front edge of the control light pulse is amplified, switching is performed with a pulse energy (<1 pJ) that is one digit or less smaller than that of an optical gate using saturable absorption. (The sign is inverted, so that the inverter operates.) When there is no control light pulse, the signal light pulse is amplified and output from the input, and no ASE noise is added. (Because the spectrum of the control light is broadened, the wavelengths of the control light and the signal light need to be separated to some extent in order to avoid crosstalk of the control light pulse.) The signal light pulse has a gain or a gain depending on the presence or absence of the control light. Since loss is incurred, the extinction ratio can be increased even if the device length is short. Further, since only a high-speed phenomenon having a time constant of the order of 100 fs is used, high-speed performance is not impaired.
[0076]
FIG. 6 shows an example of a configuration in which a strained InGaAs pulsed light source 50 in the 0.98 μm band is integrated in order to uniformly inject excitation light from the side surface. This pulse light source 50 is an integrated type of a mode-locked laser 51 and a tapered optical amplifier 52. The mode-locked laser 51 includes a gain region 53, a saturable absorption region 54, a tuning region 55, and the like. An electrode 56 is formed in each region, and a common electrode is formed on the back surface. Synchronous light for subharmonic synchronous mode synchronization can be incident from the rear end face, and an excitation light pulse synchronized with a control light pulse or a signal light pulse can be output.
[0077]
In the example of this figure, the pulse light source 50 is arranged perpendicular to the optical waveguide 5, but in consideration of the pulse propagation delay in the optical waveguide 5, the excitation light pulse arrives earlier toward the incident end side of the optical waveguide 5. It is also possible to arrange them at an angle. Further, an optical branching section is provided between the mode-locked laser 51 and the optical amplifier 52, an optical delay waveguide and an optical amplifier having different delay amounts are provided for each branch, and the incident end side of the optical waveguide 5 is earlier from a plurality of optical amplifiers. It can also be configured to emit an excitation light pulse.
[0078]
The present embodiment can be variously modified and applied without departing from the spirit of the invention.
[0079]
(Third embodiment)
FIG. 7 is a schematic diagram showing a main configuration of a Mach-Zehnder interferometer type ultrahigh-speed optical branching switch using a nonlinear optical semiconductor device 74 according to the third embodiment of the present invention. Since the configuration of the nonlinear optical semiconductor device 74 used here is the same as the configuration of the semiconductor optical amplifier 20 of the first embodiment shown in FIGS. 1 and 2, detailed description using the drawings is omitted. . However, there is no injection of excitation light from the side. Although the layer structure is slightly different, the details will be described later.
[0080]
This Mach-Zehnder interferometer type optical switch is formed by integrating optical elements connected by a passive optical waveguide 62 on a quartz optical waveguide 61. There are five optical input / output ports: a signal light input port 63, a pump light / control light input port 64, a first signal light output port 65, a second signal light output port 66, and a pump light / control light output port 67. is there. A signal light pulse having a wavelength of 1.35 μm, a pulse width of 100 fs, and a repetition of 1 Tb / s is input from the signal light input port 63. From the control light input port 64, a control light pulse having a wavelength of 1.55 μm, a pulse width of 200 fs, and a repetition of 1 Tb / s and an excitation light pulse having a wavelength of 1.2 μm, a pulse width of 500 fs, and a repetition of 1 THz are incident substantially simultaneously. . The control light pulse is set to be incident at a timing that overlaps the signal light pulse. The excitation light pulse is incident at an earlier timing than the control light pulse so as to overlap the control light pulse and the signal light pulse. The control light pulse and the signal light pulse are modulated. The excitation light pulse is a periodic clock pulse.
[0081]
The signal light pulse input from the signal light input port 63 is split by the optical coupler 68 into the first branch 71 and the second branch 72 at a predetermined branching ratio. The signal light pulse branched to the first branch 71 is combined with the control light pulse input from the control light input port 64 by the wavelength coupler 73 and guided to the nonlinear optical semiconductor device 74.
[0082]
The nonlinear optical semiconductor device 74 is designed such that when a control light pulse with a predetermined power is input, the phase is shifted by π with almost no change in the transmittance of the signal light pulse. The operation will be described later.
[0083]
The signal light pulse output from the nonlinear optical semiconductor device 74 is separated from the control light pulse and the pump light pulse by the wavelength coupler 75 (almost no pump light pulse is transmitted), and the second branch 72 in the 1: 1 optical coupler 69. It is made to interfere with the reference light pulse that has passed through. The optical path lengths of the two branches 71 and 72 are set so that the propagation delay of the signal light pulse is the same. The branching ratio of the input-side optical coupler 68 is set so that the two pulse intensities are approximately balanced in the 1: 1 optical coupler 69, but is slightly reduced by the light intensity modulator 76 provided in the second branch 72. Adjustment can be made. The optical phase modulator 77 provided in the second branch 72 is adjusted so that the signal light pulse is output only from the first signal light output port when the control light pulse is not incident. When the signal light pulse undergoes π shift by the light pulse, it is output from the second signal light output port 66.
[0084]
The control light pulse separated by the wavelength coupler 75 is output from the control light output port 67. The leakage component of the control light wavelength output to the signal light output port without being completely separated can be removed by the wavelength filter.
[0085]
The nonlinear optical semiconductor layer (corresponding to the semiconductor multilayer structure 3 of the first embodiment) of the nonlinear optical semiconductor device 74 is formed by laminating 10 periods of the coupled quantum well structure 81 having substantially the same structure. FIG. 8 is a diagram for explaining a conduction band structure for three periods of the coupled quantum well structure 81. This coupled quantum well structure 81 has an undoped Al thickness of about 0.75 nm formed between two first undoped AlN barrier layers 82 and 82R. _0.65 Ga _0.35 An N-well intermediate layer 83, a first n-type GaN well layer 84 having a thickness of about 0.75 nm, a second undoped AlN barrier layer 85 having a thickness of about 0.75 nm, and a second n-type having a thickness of about 1 nm It consists of a GaN well layer 86. The thickness of the first undoped AlN barrier layer 82 is 1.25 nm. In the n-type GaN well layers 84 and 86, Si is 5 × 10 ¹⁹ cm ^-3 It is doped to some extent. The thickness of one cycle is 4.5 nm, and the total thickness of 10 cycles is 45 nm.
[0086]
Layers 82, 83, 84, 85 from the first undoped AlN barrier layer on the left side to the second undoped AlN barrier layer constitute a first quantum well 87 serving as an optical amplification region, and the second undoped AlN The layers 85, 86, and 82R from the barrier layer to the first undoped AlN barrier layer on the right side constitute a second quantum well 88 that functions as an electron excitation region. The first undoped AlN barrier layer 32 also functions as an electron injection region. AlN barrier layers 82 and 85, Al _0.65 Ga _0.35 The electric fields applied to the N-well intermediate layer 83 and the n-type GaN well layers 84 and 86 are approximately −5.9 MV / cm, approximately −2.05 MV / cm, and approximately 5.1 MV / cm, respectively. The voltage drop per cycle of the coupled quantum well 81 is about 0.44 V, and the voltage drop is about 4.4 V in 10 cycles.
[0087]
This embodiment differs from the first embodiment in that the ground level E of each of the first quantum well 87 and the second quantum well 88 is different. ₂ , E ₁ The difference in energy ΔE is 89 meV, which is set to resonate with the LO phonon energy. Subband E ₂ And E _Three Is mainly bound to the first quantum well 87, and the subband E ₁ And E _Four Is mainly bound to the second quantum well 88. Since the second AlN barrier layer 85 is thinner than the coupled quantum well structure 31 of the first embodiment, the ground level E ₁ , E ₂ , Excitation level E _Three , E _Four Both of these have a considerable existence probability in the other quantum well.
[0088]
Excitation level E _Three To ground level E ₂ The relaxation time with LO phonon emission into is about 500 fs. Excitation level E of first quantum well 87 _Three To ground level E ₂ The electrons transitioned by relaxation or light emission emit one LO phonon in a time of about 10 fs, and the ground level E of the second quantum well 88 ₁ To relax. In the coupled quantum well structure 31 of the first embodiment using a quantum beat, E _Four There is a risk that electrons that could not be fully excited will return to the first quantum well 37 after 70 fs, but there is no such fear in the coupled quantum well structure 81 of this embodiment. Since the energy difference of 89 meV is greater than the thermal energy, ₂ The probability of returning to is low. Therefore, compared with the coupled quantum well structure 31 of the first embodiment, E ₂ Discharge of electrons from the ₂ , E _Three An inversion distribution between them becomes easier to form. In other respects, the operation of the coupled quantum well structure 81 is generally the same as that of the coupled quantum well structure 31 of the first embodiment, and thus the description thereof is omitted.
[0089]
Before the excitation light pulse is incident, most of the electrons are accumulated in E1, and almost no current flows. When an excitation light pulse is incident, E ₁ More than half of the electrons accumulated in E _Four Through E _Three Move to. By injecting the excitation light pulse slightly earlier than the control light pulse, an inversion distribution is formed before the control light pulse arrives. In this embodiment, since the pumping light pulse propagates simultaneously with the control light pulse and the signal light pulse, the pulse width of the pumping light can be shortened. As a result, the operation speed can be repeatedly increased as compared with the case where the excitation light is incident from the side surface. However, strong excitation is required to propagate the excitation light through the entire optical waveguide.
[0090]
FIG. 9 shows that the gain of the optical amplification region 87, the absorption of the electron excitation region 88, and the incident light immediately after the current having a level sufficient to obtain a sufficient gain in the optical amplification region 87 is injected and the excitation light pulse is incident. It is a figure explaining the relationship of wavelength arrangement. The gain peak wavelength of the optical amplification region 87 is 1.45 μm, and the absorption peak wavelength of the electron excitation region 88 is 1.25 μm. E compared to the first embodiment _Four Since the confinement of the electron excitation region 88 is weak, the absorption spectrum width of the electron excitation region 88 is wider than in the case of the first embodiment (FIGS. 4 and 5). When there is no control light pulse, the signal light pulse is output with substantially the same power as the input because the gain of the optical amplification region 87 and the loss of the electronic excitation region 88 are substantially balanced. Since the excitation level of the electron excitation region 88 is weakly confined, the spectrum is wide. For this reason, when the control light pulse is incident, gain saturation of the optical amplification region 87 and absorption saturation of the electron excitation region 88 occur simultaneously. As a result, the gain and loss with respect to the signal light pulse are reduced, and the light is output with almost the same power as the input. Since the control light wavelength has a net gain, switching can be performed with a smaller pulse energy than a conventional saturable absorption light gate using intersubband absorption.
[0091]
FIG. 10 is a diagram for explaining the wavelength dependence of the refractive index change when excitation light is incident. When the control light pulse is incident, the refractive index decreases at the signal light wavelength in both the optical amplification region and the electron excitation region, so that a large phase shift is obtained. As a result, the phase of the signal light pulse can be shifted by π without significantly changing the intensity of the signal light pulse by the incidence of the control light pulse. Therefore, in the interferometer type configuration shown in FIG. Can be switched at a ratio. Of course, it can also be used alone as a light control type optical phase modulator.
[0092]
The nonlinear optical semiconductor device according to the third embodiment of the present invention can be variously modified and applied. For example, FIG. 11 shows a structure in which the intermediate layer 83 in FIG. 8 is replaced with a third GaN well layer 89. Parts having the same function are given the same numbers. The first well layer 84 and the third well layer 89 are strongly coupled via a third barrier layer 85b that is thinner than the second barrier layer 85, and the first well layer 84 and the third well layer A coupled quantum well structure composed of 89 and the third barrier layer 85b functions as a first quantum well (light amplification region). FIG. 11 also plots the square of the envelope function of each subband, but is almost the same as in FIG. 8 and functions in exactly the same way. This structure facilitates crystal growth because no intermediate composition is used.
[0093]
The optical amplification region may be a superlattice instead of a single quantum well. In this case, E in the above description ₂ , E _Three Can be read as the level of a mini-band with a wide range of energy, not the level of a sub-band. Similarly, the electron excitation region may be formed of a superlattice. In this case as well, E ₁ , E _Four Can be read as mini-band level. Alternatively, the electron excitation region may be composed of a plurality of quantum wells.
[0094]
FIG. 12 is a diagram schematically showing a conduction band structure for two periods of the unit coupled quantum well structure 90 in a modified example having an electron excitation region in which two quantum wells are coupled. This unit coupled quantum well structure 90 includes an electron injection region 91, a light amplification region 92, and an electron excitation region 93. The optical amplification region 92 includes a first quantum well 94 having a two-stage composition intermediate layer on the upstream side. The electron excitation region 93 includes a second quantum well 95 and a third quantum well 96 having substantially the same structure.
[0095]
In FIG. 12, the energy level of the subband is represented by a thick line only in a portion where the existence probability of electrons is high. Ground level E of the first quantum well 94 ₂ And the ground level E ′ of the second quantum well 95 ₁ Difference in energy, ground level E ′ of the second quantum well 95 ₁ And the ground level E of the third quantum well 96 ₁ And the excitation level E ′ of the second quantum well 95 _Four And the excitation level E of the third quantum well 96 _Four Are set to be substantially equal to the LO phonon energy. Therefore, E or E _Three To E ' ₁ The electrons that have fallen into the E due to relaxation and absorption of excitation light in a short time. ₁ Or E ' _Four , And E _Four Through the electron injection region 91 and the excitation level E of the light amplification region 92 on the downstream side. _Three To be transported to. In this configuration, E ₂ The electrons are more easily discharged from the case of FIG. In addition, since the composition intermediate layer of the first quantum well is also in two stages, E _Three The efficiency of electron injection into can be increased.
[0096]
FIG. 13 shows an example in which the wavelength arrangement of excitation light, control light, and signal light is changed and applied to a non-invertible saturable absorption gate. Since the excitation light pulse is also amplified by the optical amplification region 87, it can propagate to the depth of the optical waveguide with almost no change in level. Due to the excitation light pulse, the gain of the optical amplification region 87 and the absorption of the electron excitation region 88 are partially saturated. When a control light pulse that is stronger and narrower in pulse width is incident on this state, the gain of the optical amplification region 87 and the absorption of the electron excitation region 88 are completely saturated. The signal light pulse is in the absorption band of the electron excitation region 88 but at a wavelength outside the gain band of the optical amplification region 87. Therefore, if there is no control light pulse, it is absorbed, and if the absorption is saturated by the control light pulse, it is transmitted.
[0097]
In this case, the nonlinear optical operation is realized mainly by conventional absorption saturation, and the optical amplification region is mainly used for loss compensation of the control light pulse. The control light pulse can propagate through the optical waveguide without much change in pulse energy or waveform due to the balance between the gain of the optical amplification region 87 and the absorption of the electron excitation region 88. In addition to this effect, combined with the fact that absorption saturation is about to occur to some extent by the excitation light pulse, the control light pulse energy required for switching can be kept low. Further, since the control light pulse can be propagated even if the optical waveguide is long, the extinction ratio of the signal light output can be increased.
[0098]
Alternatively, if the wavelength of the signal light pulse is set in a region where the net absorption is small, it can be used as a phase modulator. In this case, since the control light is propagated without being attenuated over a long distance, a large phase shift can be obtained with relatively small control light energy. That is, the modulation efficiency is improved. Since the excitation light pulse acts as a kind of bias, it can also be used for fine adjustment of the phase bias during non-modulation.
[0099]
The present invention can be variously modified and applied in addition to the above-described three embodiments and modifications thereof without departing from the spirit of the invention. For example, the layer structure of the first and second embodiments is applied to the phase modulator and the interferometer type optical branch switch as in the third embodiment, or the layer structure as in the third example is applied to the first. It can also be applied to the semiconductor optical amplifier of the embodiment and the gain saturation type optical gate switch of the second embodiment. Moreover, you may use as an optical arithmetic element which inputs excitation light and control light. Alternatively, more complex operations can be realized by using a plurality of wavelengths of excitation light, control light, and signal light.
[0100]
It is also possible to use these optical switches as wavelength conversion elements. For example, in the second embodiment, if the signal light pulse is a clock pulse and the control light pulse is an input signal, the pattern of the input signal is transferred to a clock pulse having a different wavelength (the sign is inverted).
[0101]
In addition, various functions can be realized by combining or integrating with other elements. As the material system, II-VI group semiconductors, InGaAs / AlAsSb systems, and the like can be used.
[0102]
【The invention's effect】
As described above in detail, according to the present invention, it is possible to realize a semiconductor optical amplifier having a very low noise and a high saturation level. In addition, it is possible to switch the intensity, phase, optical output destination, etc. of signal light with high efficiency and high speed with a smaller pulse energy than conventional saturable absorption optical gates using intersubband absorption. . When used as a saturable absorption gate or an optical branching switch, the extinction ratio can be improved as compared with the prior art.
[Brief description of the drawings]
FIG. 1 is a perspective view schematically showing the structure of a semiconductor optical amplifier according to a first embodiment.
FIG. 2 is a sectional view schematically showing a layer structure of the semiconductor optical amplifier according to the first embodiment.
FIG. 3 is a band diagram corresponding to three cycles of a coupled quantum well structure constituting the semiconductor multilayer structure of the semiconductor optical amplifier according to the first embodiment.
FIG. 4 is a view for explaining the wavelength arrangement of the semiconductor optical amplifier according to the first embodiment.
FIG. 5 is a view for explaining the wavelength arrangement of a nonlinear optical semiconductor device according to the second embodiment.
FIG. 6 is a diagram schematically illustrating a configuration example in which a nonlinear optical semiconductor device according to a second embodiment is integrated with a pumping light source.
FIG. 7 is a diagram schematically showing a configuration of an ultrafast optical branch switch using a nonlinear optical semiconductor device according to a third embodiment.
FIG. 8 is a band diagram corresponding to three cycles of a coupled quantum well structure constituting a nonlinear optical semiconductor layer of a nonlinear optical semiconductor device according to the third embodiment.
FIG. 9 is a view for explaining the wavelength arrangement of a nonlinear optical semiconductor device according to the third embodiment.
FIG. 10 is a view for explaining a change in refractive index due to control light of the nonlinear optical semiconductor device according to the third embodiment.
FIG. 11 is a conduction band structure diagram illustrating a modification of the optical amplification region of the nonlinear optical semiconductor device according to the third embodiment.
FIG. 12 is a conduction band structure diagram illustrating a modification of the electron injection region of the nonlinear optical semiconductor device according to the third embodiment.
FIG. 13 is a view for explaining a modification of the wavelength arrangement of the present invention.
FIG. 14 is a diagram schematically showing a conduction band structure for two periods of a unit structure constituting an active layer of a quantum cascade laser according to the prior art.
[Explanation of symbols]
1 ... Sapphire substrate
2,4 ... Nitride semiconductor layer
3 ... Nonlinear optical semiconductor layer
5 ... Mesa optical waveguide
6 ... Polyimide
7, 8 ... Electrode
9 ... Incident light
10 ... outgoing light
11 ... Excitation light
20 ... Semiconductor optical amplifier / nonlinear optical semiconductor device (gain saturated optical gate switch)
22 ... Undoped AlN layer
23: First n-type GaN layer
24. First AlGaN composition gradient layer
25: Second AlGaN composition gradient layer
26: Second n-type GaN layer
31, 81... Coupled quantum well structure having a single-layer period of a semiconductor multilayer structure
32, 82 ... first undoped AlN barrier layer (electron injection region)
33, 83 ... Undoped AlGaN well intermediate layer
34, 84 ... first n-type GaN well layer
35, 85 ... second undoped AlN barrier layer
36, 86 ... second n-type GaN well layer
37, 87 ... first quantum well (light amplification region)
38, 88 ... second quantum well (electron excitation region)
50. Semiconductor photoexcitation element
51. Mode-locked laser
52. Tapered semiconductor optical amplifier
53, 54 ... Gain region
55. Saturable absorption region
56 ... Electrode
57 ... Synchronous light input
61 ... Quartz optical integrated circuit
62: Passive optical waveguide
63 ... Signal light input port
64 ... Control light input port
65: First signal light output port
66 ... Second signal light output port
67 ... Control light output port
68 ... Optical coupler
69: 1: 1 optical coupler
71 ... 1st branch
72. Second branch
73, 75 ... wavelength coupler
74: Nonlinear optical semiconductor device (optical phase modulation element) of the second embodiment
76: Light intensity modulator
77 ... Optical phase modulator
89 ... Third GaN well layer
90. Unit-coupled quantum well structure of modification
91 ... Electron injection region
92: Light amplification region
93 ... Electronic excitation region
94: First quantum well
95 ... Second quantum well
96 ... third quantum well
101: Electron injection region
102 ... Active region
103 ... well layer
104 ... barrier layer
105 ... Mini band

Claims (11)

A first quantum well (light amplification region) having at least two subbands, and an energy difference between the subbands of the first quantum well, the energy difference between the subbands having at least two subbands. A semiconductor multilayer structure in which a unit coupled quantum well structure in which at least two quantum wells of the second quantum well (electron excitation region) are coupled by a barrier layer is periodically stacked.
An optical waveguide structure including this semiconductor laminated structure;
Means for applying a voltage in the stacking direction to the semiconductor stacked structure;
Means for entering light from one end of the optical waveguide structure;
Means for emitting light from the other end of the optical waveguide structure;
Means for injecting excitation light in the intersubband absorption band of the second quantum well into the semiconductor multilayer structure;
Comprising a result, the
The second quantum well of the unit coupled quantum well structure when a voltage is applied to the semiconductor multilayer structure in the stacking direction and the excitation light in the intersubband absorption band of the second quantum well is injected. A semiconductor optical functional device, characterized in that carriers flow through a first quantum well of a unit coupled quantum well that is tunneled through a barrier layer from the excited level of and adjacent to the downstream side of the unit coupled quantum well structure .   2. The semiconductor optical functional device according to claim 1, wherein the light incident on the optical waveguide structure is signal light within a transition band between subbands of the first quantum well.   2. The semiconductor optical function according to claim 1, wherein the light incident on the optical waveguide structure is signal light and control light in an intersubband transition band of the first quantum well or the second quantum well. device.   The semiconductor optical functional device according to claim 1, wherein the semiconductor multilayer structure is made of a nitride semiconductor.   4. The semiconductor optical functional device according to claim 1, wherein an intersubband transition wavelength of each of the first quantum well and the second quantum well is 2 μm or less. The semiconductor optical functional device according to claim 1, wherein the first quantum well has a step-like or continuous composition gradient region on the opposite side to the second quantum well.   The third quantum well is provided on the opposite side of the first quantum well from the second quantum well via another barrier layer thinner than the barrier layer. 4. The semiconductor optical functional device according to any one of 3 above.   The semiconductor optical functional device according to claim 1, wherein the excitation light is introduced from a side surface of the optical waveguide structure. When a voltage is applied, the excitation level of the second quantum well is a pseudo-binding level coupled by tunneling with a three-dimensional quantum state of an adjacent layer corresponding to the downstream side of the carrier flow. The semiconductor optical functional device according to claim 1. The ground level of the first quantum well is higher than the energy corresponding to the LO phonon energy of the material constituting the quantum well than the ground level of the second quantum well. The semiconductor optical functional device according to claim 1, wherein: The ground level of the first quantum well and the ground level of the second quantum well are coupled, and the energy difference between the two levels due to the resonance is 30 meV or more. 4. The semiconductor optical functional device according to any one of 3 above.

Images