Optical interconnect - Applications of VCSELs

Chapter 1 Introduction and overview

1.4 Applications of VCSELs

1.4.2 Optical interconnect

The optical interconnect is considered to be inevitable in the computer technology. The performance of massively parallel computers is usually limited by the communication bottleneck between processors. Optic provides an effective mean to line these processors because of its high capacity, low crosstalk and attenuation, and the possibility to obtain three-dimensional architectures. Other potential applications include routers, switches and storage. The VCSEL is a strong candidate as the preferred optical light source for the emerging optical interconnect mass market, meeting the requirement of low cost, high density integration and low power dissipation. A 256-channel bi-directional optical interconnect using VCSELs and photodiodes on CMOS was demonstrated.

laboratories and corporations research in 10Gbps device, as Table1-1. In this paper, we focus on high-speed 850nm vertical cavity surface emitting lasers. First, we mention major VCSEL structure process, oxide and oxide-implantation, and explain high-speed test system in detail in chapter 2. In chapter 3, a simple introduction of phenomenon approach the rate equation is presented. The scattering coefficients are estimated by high-speed measurement, Transfer function is used in analyzing dynamic characteristics as relaxation frequency, damping rate. We measure static and dynamic characteristics of VCSELs and create an equivalent circuit for modeling in chapter 4. Finally, we give a brief conclusion in chapter 5.

Light output

Figure 1-2 A modal of vertical cavity surface emitting laser (VCSEL)

DBR DBR

Optical Confinement Current

Confinement Electrode: anode

Transverse Mode field

Active Layer D

Electrode: cathode

1.3~1.6

1.2~1.3

0.78~0.88

0.63~0.67

0.45~0.55

0.3~0.5

1.3 1.5

0.8 1.0 0.3 0.5

Wavelength (µm)

GalnAlN/GaAlN ZnSSe/ZnMgSSe GaAlInP/GaAs GaAlAs/GaAs GaInAs/GaAs GaInNAs/GaAs GaInAsP/InP AlGaInAs/InP

Figure 1-3 Material for VCSELs in wide spectral bands

Ring electrode Electrode

Active layer p-DBR

n-DBR

Substrate

H⁺ H⁺

Active layer p-DBR

n-DBR

Substrate

(a) Ring-electrode type (b) Proton-implant type

(d) Air-post type Substrate

n-DBR p-DBR Active layer Electrode

(c) Regrowth BH type Substrate

n-DBR Active layer Electrode

Oxide layer Electrode

Active layer p-DBR

n-DBR

Substrate Epitaxial regrowth

(e) Selective oxidation type

Figure 1-4 Structure for current confinement for VCSELs

Source: Laser Focus World, PIDA

Figure 1-5 Global consumption value of diode laser by the year 2002

Source: Laser Focus World, PIDA

Figure 1-6 Global consumption value of diode by the year 2002 (distinguished into wavelength)

Figure 1-7 Transmission spectrum for silica fiber

Parameter Symbol Material Wavelength Active layer thickness d 100 Ǻ ~ 0.1µm 80 Ǻ ~ 0.5µm

Active layer area S 3 x 300 µm² 5x5µm²

Active volume V 60 µm³ 0.07µm³

Cavity length L 300 µm ~1 µm

Reflectivity Rm 0.3 0.99 – 0.999

Optical confinement ξ ~ 3 % ~ 4 % Optical confinement

(Transverse) ξt 3 – 5 % 50 – 80 %

Optical confinement

(Longitudinal) ξl 50 % 2 x 1% x 3 (3 QWs) Photon Life time τp4 ~ 1 ps ~ 1 ps

Relaxation Frequency

( Low Current Levels) fr < 5 GHz > 10 GHz

Table 1-1 Comparison of parameters between stripe laser and VCSEL

Chapter 2 Theory

The chapter begins by developing a reservoir model for a flow of charge into double-heterostructure active regions and its subsequent recombination.

Recombination mechanism is determined by electron-hole recombination at quantum well generates photons for light emission. For this we describe the phenomenological approach to VCSEL first.

The rate equations provide the most fundamental description of the laser in next section. It describes the time-evolution of carrier and photon densities in a laser cavity as a function of the pump rate, material gain and parameters associated with the material properties and laser construction. Dynamic characteristics of VCSEL are studied as rate equation, and it is produced for modulated current injection. Transfer function is relevant equation originated from rate equation which is explained next.

The last section is presented of V-parameter used in 2-D photonic crystal structure, the V-parameter evaluating the number of guided modes in cylindrical wave guides the rate equations for carriers and photons are found to be analogous to differential equations that describe the current and voltage in an RLC circuit.

Scattering parameter offers not only response as bandwidth but reflection coefficient which mirrored equivalent RLC circuit. Establishment of equivalent circuit is purposed to be found the limitation of bandwidth for VCSEL structure and making terminal impedance matching easily.

2.1 Recombination mechanism of VCSEL

The proposition considers the current injected into VCSEL, and suggests it is desirable to have all of it contributes to electrons and holes which recombine in the active region. Since the definitions of the active region and the internal quantum efficiency, ηi, are so critical to further analysis. Active region, evolved into lowest band-gap region, is where recombining carriers contribute to photon emission. Band diagram of active region, includes separate confinement hetero-structure (SCH)

band-gap region, illustrated in Figure 2-1. Internal quantum efficiency, ηi, is the fraction of terminal current that generates carriers in the active region. It is important to realize that includes all of the carriers that are injected into active region, not just carriers that recombine induce radiating at the desired transition energy.

The carrier density, n, in the active region is governed by a dynamic process. In fact, we could compare the process of a certain steady-state carrier density in the active region to that a reservoir analogy, which is being simultaneously filled and drained, in a certain water level. This is shown schematically in Figure 2-2(a). For the double heterostructure active region, the injected current provides a generation term and various radiative and nonradiative recombination processes as well as carrier leakage provides recombination term. Thus, rate equation is determined as

rec

gen R

dt G

dn = − (2-1)

where Ggen is the rate of injected electrons and Rrec is the rate of recombining electrons per unit volume in the active region. There are ηi I/q electrons per second being injected into the active region. V is the volume of the active region.

The recombination process is accompanied with spontaneous emission rate, Rsp, and a nonradiative recombination rate, Rnr, depicted in Figure 2-3. Carrier leakage rate, Rl, must be occurred at the transverse and/or lateral potential barrier are not sufficiently high. A net stimulated combination, Rst, is including both stimulated absorption and emission. That is, an increased injection results in an increased output, Rst, but no increase in carrier density as water level illustrated in Figure 2-2(b).

Total recombination rate is expressed as below

(2-2)

where the first three terms on the right refer to the natural carrier decay processes. It is common to describe the natural decay processes by a carrier lifetime, τ . In the absence of photon generation term, the rate equation for carrier decay is, dn/dt = n/τ , where n/τ = Rsp+ Rnr+ Rl, by comparison to Eq. (2-2). Thus, the carrier rate equation in equivalent be expressed as

i n Rst

The net stimulated recombination rate, Rst, in generating photons as well as effect of the resonant cavity in storing photons is investigated additionally. A rate equation for the photon density, np, which includes the photon generation and loss terms are constructed analogous to carrier rate equation. The main photon generation term above threshold is Rst. Every time an electron-hole pair is stimulated to recombine, another photon is generated. However, as indicated in Figure 2-4, the cavity volume occupied by photons, Vp, is usually larger than the active region volume occupied by electrons, V, the photon density generation rate will be (V/Vp)Rst

not just Rst. This electron-photon overlap factor, (V/Vp), is generally referred to as the confinement factor, Γ.

Photon loss occurs within the cavity due to optical absorption and scattering out of the mode, and it also occurs at the output coupling mirror where a portion of the resonant mode is usefully coupled to some output medium. These losses could characterize by a photon lifetime, τ p. The photon rate equation takes the form

where β is the spontaneous emission factor. For uniform coupling to all modes, β is just the number of spontaneous emission coupled into specific mode in the bandwidth of all spontaneous emission.

Rst represents the photon stimulated net electron-hole recombination which generates more photons. This is a gain process for photon. An increased photon is proportional to an increased injection carriers overfill the reservoir, which is critical condition to generate stimulated emission, shown in Figure 2-3. The proportion is defined as gain coefficient, g(n). That is,

(2-5) to laser oscillation condition with regenerated amplifier approach due to multiple reflection, as shown in Appendix (1). ntr is a transparency carrier density, and vg is the

group velocity of the mode of interest including both material and waveguide dispersion. vg still can be expressed as C/nr, where C is 3x10⁸ m/s and nr is index of cavity medium. Thus Rst is replaced with g(n)np, and now the carrier and photon density rate equations can be written as

Equations (2-6) and (2-7) are two coupled equations that can be solved for the steady-state and dynamic responses of a diode laser in next section.

2.2 Transfer function

Under small signal modulation, the carrier and photon density rate equation, Eq.

(2-6) and (2-7), are used to calculate relaxation resonance frequency and its relationship to laser modulation bandwidth.

Consider the application of an above-threshold DC current, I0, carried with a small AC current, Im, to a diode laser. Illustration is, shown in Figure 2-5, under basic L-I characteristics (Light output power versus current). The small modulation signal with some possible harmonics of the drive frequency, ω. Small signal approximation, assumes Im<<I0 bias and spontaneous emission term, β , is neglected, is expressed as

Before applying these to Eq. (2-6) and (2-7), the rate equation is rewritten for the gain. Assumption under DC current is sufficiently above threshold that the spontaneous emission can be neglected. Without loss of generality, we suppose full overlap between the active region and photon field, Γ =1; furthermore, internal quantum efficiency, η_i, is neglected. That is,

substitute Eq. (2-8) into Eq. (2-9)

)

for this, it is similarly expressed modulation terms as

The small signal terms in frequency domain of carrier and photon are given by

substitute into Eq. (2-11) and (2-12), the equations become

Carrier modulation term in frequency domain is simplified as

Photon modulation term in frequency domain is simplified as response of two arranged equations as below

where

……Damping constant (decay rate)

With the Eq. (2-15) and (2-16) we observe the coupling between the small signal photon, npm, and carrier, nm. Small signal carrier injection induces photon achieved oscillation. This phenomenon produces a natural resonance in the laser cavity which shows up the output power of the laser in response to sudden changes in the input current. The natural frequency of oscillation associated with this mutual dependence between nm and npm. Modulation response is expanded the small signal modulation relationship to steady-state. From Eq. (2-15) and (2-16), the modulation response is denoted as

The general behavior of M(ω) is shown in Figure 2-6. Modulation bandwidth is determined as cutoff frequency, fc, which is the position with half response written as

( )

²¹

for ωr2Ω² << (ωc2-ωr2)², the cutoff frequency, ωc, is approximated to 3 ωr.

Transfer function, H(ω), is the identical term in Eq. (2-11) and (2-12) respectively obtained with Cramer’s rule [3]. It is similar to modulation response, M(ω), describing the response of the laser intensity to small variations in the drive current through the active region. That is,

π γ similar to Eq. (2-18), and C is a constant. Accounting for additional extrinsic limitations due to carrier transport and parasitic elements related to the laser structure results in an extra pole in the small signal modulation transfer function

⎟⎟

where fp is the cutoff frequency of the low pass filter characterizing the extrinsic limitations. It is crucial for microwave applications that the modulation bandwidth of the VCSEL is sufficiently large so that efficient modulation is achieved as the modulation frequency.

2.3 V-parameter

Lateral-mode control of vertical-cavity surface-emitting lasers (VCSELs) is one of the key issues in realizing high performance optical communication systems, in which single-mode operation is necessary for long and short wavelength regions.

High-power single-mode operation is also required for free-space data communication applications. Recently, a two-dimensional photonic crystal (2-D PhC) structure formed on a VCSEL surface has been investigated as a lateral-mode control method. The most attractive feature of this structure is the enlargement of the emission area, thereby permitting higher power output. The large area can be realized because of strong wavelength dependence of the refractive index in the 2-D

photonic crystal structure, analogous to the situation in a photonic crystal fiber.

Although good single mode operation has been reported for a specific structure, the optimized design of 2-D photonic crystal structure was not clear, especially when considering a finite etching depth. Since the mode control mechanism utilized in this technology is the effective refractive index control achieved by forming a 2-D photonic crystal structure, a parameter representing this control must have a strong dependence on the etching depth. We have investigated the etching depth dependence, both theoretically and experimentally, of the effective index change in a VCSEL structure.

The 2-D photonic crystal structure with finite etching depth incorporating a single point or a seven-point defect is formed in the top DBR. It is known that the normalized frequency or V-parameter is useful in evaluating the number of guided modes in cylindrical wave guides, an important example being step index optical fibers. The cutoff condition of the first higher mode leads to Veff = 2.405, and thus a wave guide with Veff < 2.405 is considered to be single mode. In a photonic crystal VCSEL, the effective V-parameter can be expressed by

2 2 2 (2-23)

( )

eff eff eff

V πr n n γ

= λ − − n∆

where λ is an operating wavelength, r is an equivalent defect radius, neff is the effective refractive index of the VCSEL cavity [6] without a photonic crystal structure present, Dn is the refractive index reduction introduced by the photonic crystal structure, and gis the hole depth dependence factor that accounts for finite etching depth of the photonic crystal holes in actual photonic crystal VCSELs. The γ factor can be understood qualitatively as proportional to the spatial overlap between the photonic crystal structure and the longitudinal optical power distribution inside the VCSEL structure. Thus, γ = 0 for vanishing etching depth, γ = 1 for infinite etching depth, and γ = 0.5 for holes reaching the middle of the cavity. In the following discussion, the equivalent defect radii of a single point defect and seven-point defect structures are assumed to be Λ-d/2 and 3 Λ-d/2, respectively, where L is a lattice constant and d is the hole diameter of a circular hole of the 2-D photonic crystal

structure. Since we need to investigate larger d/L ratios than those of PCFs,[5] Veff is slightly modified from its appearance in Ref. 5, with the introduction of the Veff - parameter and the modified r in Eq. (2-23). The refractive index variation Dn can be obtained from the photonic band diagram of an out-of-plane propagation mode,[7,8]

calculated by assuming that the photonic crystal structure is infinite both in lateral and vertical directions.

Electron

SCH MQW SCH Cladding Cladding

Photon

Hole

Figure 2-1 Band diagram of active region of VCSEL

Rth

Ggen=ηi(I/qV)

Rnr

Current leakage

I/qV

Rsp

(a)

Rth

Rst

Ggen=ηi(I/qV)

Rsp

Rnr

Current leakage

I/qV

(b)

Figure 2-2 Reservoir analogy (a) under threshold (b) above threshold

Hole Electron

Photon

(b) Stimulated generation (absorption) (a) Spontaneous emission

(c) Stimulated recombination (emission)

(d) Nonradiative recombination

Figure 2-3 Basic electronic recombination/generation mechanism

Light output

Mirror 2 Mirror 1

Active region: N, Np, V Cavity: Vp

Figure 2-4 Schematic of VCSEL illustrating active region and cavity volumes

Light wer (mW

po )

Optical signal

Injection current (mA) Modulating

small signal

Figure 2-5 Conversion from electrical small signal train into optical signal

Frequency (Hz) Relaxation frequency

Cutoff frequency -3

0 Response

(dB)

Figure 2-6 Small signal modulation response of a VCSEL

Chapter 3 VCSEL fabrication and measurement setup

In this chapter, we presented the fabrication method of VCSEL, instrument setup for electrical and optical characteristics measurement. The fabrication techniques for VCSEL such as air-post, regrowth, proton-implantation, selective oxidation and reactive ion etch (RIE) had been employed to define the current path, gain region, carrier and the optical confinement. With in these techniques, the VCSEL fabricated by proton-implantation and reactive ion etch (RIE) had some superior properties such as simple and stable process. The design principles of VCSEL wafer structure were based on the method described as follow section. Finally, probe station system, spectrum measurement system , the establishments of eye diagram measurement system and far field pattern measurement system were described as follows.

3.1 Fabrication of tapered and blunt oxide VCSEL

A vertical cavity surface emitting laser consisted of multi-quantum-well sandwiched between two highly reflective distributed Bragg reflectors (DBRs). The 850nm GaAs / AlGaAs VCSEL devices studied here were grown by MOCVD on the Si-doped GaAs substrate. The bottom DBR had 39.5 pairs of alternating Al0.15Ga0.85As / Al0.9Ga0.1As pairs with quarter-wavelength-thick layer. The active region was consisted of three GaAs / AlGaAs quantum wells and cladding layer. The top DBR had 22 pairs of alternating Al0.15Ga0.85As / Al0.9Ga0.1As layer. A high Al-content selective oxidation layer, Al0.98Ga0.02As, was inserted just three layers above the active region. The VCSEL structure was trenched a ring of mesa, indicated in Figure 3-1, in order to execute oxidation process, which still improved current spreading effect. A bridge connected with contact ring and bond pad for current injection. Mesa structure benefited reducing current path and inducing optical confinement. Passivation layer coating, used of SiNx, formed an isolation layer reducing leakage current below metal contact. The detailed process of oxide-implant VCSEL was desired as Figure 3-2.

was placed into center of furnace and purge by N2. The furnace annealed approach 420^oC then imported the invariable flow-rate of steam into furnace for fixed time.

Oxidation time was crucial for the process for forming different oxide apertures. The oxide-extent was almost linear relative to oxidation time, shown in Figure 3-4. For blunt oxide VCSELs, a 30-nm-thick Al0.98Ga0.02As layer was placed in node position.

For tapered oxide VCSEL, the tapered oxide layer was formed of a 10-nm-thick Al0.99Ga0.01 layer adjacent to a 200-nm-thick Al0.98Ga0.02As layer, which is similar to Ref. [9]. Multiple proton implantations with a dose in the range of 10¹³ ~ 10¹⁵ cm^-2 and four different proton energy ranges between 200 to 420 keV were adapted according to simulation results of the stopping and range of ions in matter (SRIM). The implantation region was kept away from the mesa to prevent damage to the active region and consequent voiding of triggering reliability issues. A series VCSELs with different type of oxide layer were fabricated in the same implant process condition and their static and dynamic characteristics in were presented in chapter 4.

3.2 Fabrication of photonic crystal VCSEL

3.2.1 Fabrication of oxide photonic crystal VCSEL

The epitaxial layers of the PC-VCSEL’s wafer structure were grown on the n⁺-GaAs substrate by metal-organic chemical vapor deposition (MOCVD). The bottom distributed Bragg reflector (DBR) consists of a 30.5-pair n-type (Si-doped) quarter-wave stack of Al0.12Ga0.88As/AlAs. The top DBR consists of 22-period p-doped (carbon-doped) Al0.12Ga0.88As/AlAs quarter-wave stack. Above that is a heavily doped p-type GaAs contact layer. The graded-index separate-confinement heterostructure (GRINSCH) GaAs/AlGaAs active region has an undoped three-quantum-well (3QWs) GaAs/Al0.3Ga0.7As, a lower linearly-graded undoped AlxGa1-xAs (x = 0.6->0.3) waveguide layer and an upper linearly-graded undoped AlxGa1-xAs (x = 0.3->0.6) waveguide layer.The process was similar to oxide-confined VCSEL. After that, hexagonal lattice patterns of photonic crystal with a single-point defect were defined within the p-contact ring using photolithography and etched through the p-type DBR by using a reactive ion etch (RIE). The lateral index around a single defect can be controlled by the hole diameter (a) to lattice constant (Λ) ratio

and etching depth [10]. This ratio (a/Λ) is 0.5; the lattice constant Λ is 5 mm in the PC-VCSEL and the etching depth of the holes is about 17 pairs out of 22 pairs of the top DBR layers. The device structure, top view image of the PC-VCSEL, and the

在文檔中 850nm高速和光子晶體面射型雷射之特性量測與分析 (頁 16-0)