2.1

Motivation

First, let us look at the motivation of research based on the conception of VCSEL. Second, the mode behaviors and related applications, and lastly, the current progress of applications are discussed. After 43 years of research and development, a new optoelectronics field has been opened, which will be introduced here.

During the research on semiconductor lasers by our group starting in 1970 at Tokyo Institute of Technology, I was dissatisfied with the conventional semiconductor lasers from the viewpoint of their fabrication and device characterization. The cavity of the conventional semiconductor lasers was fabricated by cleaving wafers, i.e., by cutting an epitaxially grown wafer along a specific crystal plane (100) with the cavity length of approximately a few hundred microns. In the initial stage of production, the cleaving process was carried out using a kind of surgical knife. It was far from mass production, for example, on the order of 10⁹ pieces. As another disadvantage originating from the edge-emitting laser structure, device characterization was possible only after the cleaving process was completed, which made device characterization difficult, including the initial quality test on a wafer scale. This feature had also led to the difficulty in realizing two-dimensional semiconductor laser arrays.

Another consideration was how to realize a single-mode laser, that is, the lateral mode and longitudinal mode. Around 1976, we were interested in a short-cavity semiconductor laser and the possibility of maintaining in dynamic single-mode capability by making a short cavity to widen the free spectrum range (FSR). One of the conclusions made on the basis of computer simulation was that we must make the cavity length smaller than 50 μm. This result was presented at the IEEE Semiconductor Laser Conference in Nemunosato, Mie Prefecture and reported in a paper [7].

The third problem was how to control the lasing frequency, namely, how to guarantee the reproducibility of lasing frequency. It is almost impossible to precisely control the oscillation frequency of Fabry-Perot edge-emitting lasers lasing under the multimode condition. Also, the cutting position that determines the cavity length may have an error much larger than wavelength. Some grating filters are used to control oscillation frequency, as in distributed feedback (DFB) and distributed Bragg reflector (DBR) lasers. Since resonant frequency is determined by the grating period, the precise frequency should be dependent on waveguide structures that affect the equivalent index.

After critical consideration to solve the aforementioned issues associated with conventional edge-emitting lasers, I reached the conclusion that the laser cavity should be made vertical, not transverse to the semiconductor wafer surface. I thought that the cavity can be made of semiconductor layers and/or dielectric layers, which can be fabricated by semiconductor processes, not by manual cleaving processes.

2.2

Targets

The targets of its invention were to realize a laser with;

The targetss are illustrated in Fig. 2. The items inside the circle are those that were achieved in industries in the 1970s. We took a different way of fulfilling the aforementioned three conditions.

Fig. 2.

The targets of research leading to VCSEL invention

2.3

Edge-emitting lasers and VCSELs

The difference between stripe-shaped edge-emitting lasers and VCSELs for obtaining single-mode operation is shown in Fig. 3. The upper figure shows the behavior of edge-emitting stripe lasers. The horizontal axis is the stripe width and the vertical one shows the resonant frequency. f₀ means the central nominal frequency. Even in the case of a small stripe width on the order of microns, the lateral mode can be single as indicated in the figure by an ellipse. On the other hand, resonance occurs at multiple frequencies. To realize single-frequency operation, some filtering structure is required, such as distributed Bragg reflectors.

Fig. 3.

The difference between edge-emitting stripe laser and VCSEL

On the other hand, we consider the case of VCSEL shown in the lower figure. For small-diameter VCSELs, the lateral mode is single and the longitudinal mode can also be single, since the free spectral range (FSR) is large owing to small cavity length.

3.1

The first device demonstrated

On the basis of the initial idea of the surface-emitting semiconductor laser, we pursued analytical and experimental research on this new structure. The concept and the first experimental results on spontaneous vertical light emission from a GaInAsP/InP device were presented at the 25th Spring Meeting of Applied Physics Societies, Japan, in March 1978 [8]. The second report was presented at the 26th Fall Meeting of Applied Physics Societies, Japan, in November 1978. The paper consisted of results from the analysis of mirror reflectivity for vertical cavity laser and experiments on vertical cavity formation with a convex mirror surface[9].

In 1979, we demonstrated the lasing operation of a GaInAsP/InP VCSEL under the pulsed injection current condition at 77 K [10]. This is the first demonstration of current injected VCSEL lasing oscillation and showed the possibility of VCSEL as a practical-semiconductor laser.

Subsequently, I and my group have been continuously conducting elaborate research and development of VCSELs and have achieved many basic results. Many papers that have contributed to the growth of VCSEL devices and related applications have been published.

3.2

Achievement for device realization

The primary breakthrough was the fabrication of a 6 μm cavity VCSEL that demonstrated a clear VCSEL mode even at 77 K, in 1982; i.e., single-mode, circular beam, and linear polarization[11]. In Fig. 4, we show an image of a cleaved cross section. The central portion corresponds to a laser cavity as narrow as 6 μm.

Fig. 4.

(Color online) Cross section of a 6 μm cavity ¹¹⁾.

A pulsed threshold current as small as 6 mA was reported at room-temperature in a AlGaAs/GaAs VCSEL with a cavity length of 7 μm [12]. The first room-temperature continuous wave (CW) operation was achieved in 1988 by Koyama and I using a GaAs system as shown in Fig. 5 [12][13]. This experiment demonstrated the possibility of VCSEL as an engineered semiconductor laser. We employed MOCVD crystal growth and multilayer mirror formation techniques to lower the threshold current density. We also achieved a near-room-temperature CW in a 1,300 nm VCSEL in 1993 [15].

Fig. 5.

(Color online) VCSEL that achieved the first room-temperature CW operation ^12),13).

3.3

Threshold current reduction of early VCSELs

A short cavity easily leads to a small laser cavity volume, which drastically lowers the VCSEL threshold current if sufficient photon confinement in the cavity is achieved. Low threshold currents are preferable for many applications in ecosystems. A strong research and development competition ensued in a worldwide scale from the end of the 1980s to the 1990s, where I and my group were almost always of the forefront. Aa described before, a pulsed threshold current as small as 6 mA was reported at room-temperature in a AlGaAs/GaAs VCSEL with a cavity length of 7 μm [12]. This was the first time that a threshold current lower than 10 mA was realized in VCSELs at that time. The injection current was confined in the active region of 6 μm diameter. Single-longitudinal mode oscillation was achieved up to 6.7 times the threshold current.

3.4

Various materials for VCSELs

We tried to find the possibility of fabricating VCSELs using various materials. Since 1977, basic studies have been carried out by us on semiconductor crystal growth, crystal evaluation, and applications in various emission wavelengths. The principal results are summarized as follows;

4.1

Quantum wells for VCSELs

Quantum well and quantum dot structures were verified to have higher gains than the bulk structure when applied to an active region of edge-emitting semiconductor lasers. It was thought questionable to use quantum wells in VCSELs, since the active thickness is too small, on the order of several nanometers.

The threshold current of surface-emitting lasers with a multi-quantum well (MQW) active region was theoretically estimated by Uenohara et al.[24] On the basis of the results, we introduced the MQW structure into the active region of surface-emitting lasers. A GaAlAs/GaAs MQW surface-emitting laser was grown by MOCVD and its first lasing operation was demonstrated by current injection at 77 K the under pulsed condition. After these pioneering trials, many researchers have actively utilized the MQW structure as the active gain material to reduce the threshold current of VCSELs, as described below.

4.2

Semiconductor Bragg reflectors

We first introduced a semiconductor multi-layered Bragg reflector as one of the mirrors for VCSEL²⁵⁾. This VCSEL cavity consisted of a pair of on n-Al_0.1Ga_0.9As/AlAs multilayer Bragg reflector and a Au/SiO2 mirror. With this structure, besides the satisfactory high reflectivity achieved, manual processes are not necessary for fabricating a laser cavity. Soon after this achievement, a VCSEL was realized, whose principal components fully consisted of semiconductors. Both upper and lower mirrors of the VCSEL were formed with semiconductor multilayered Bragg reflectors [25]. A drastic improvement in VCSEL fabrication productivity was accomplished. The semiconductor DBR is one important feature that I wanted to improve by eliminating the cleaving process in the formation of the cavity in conventional edge-emitting lasers. The semiconductor multilayered distributed Bragg reflector is utilized in almost all VCSELs today.

4.3

Mechanical frequency tuning

The reproducibility of laser frequency is one of the issues that I considered at the time of starting the semiconductor laser research. It is necessary to have some tunability to match the frequency to the system employed.

The VCSEL can emit light with pure single wavelength characteristics owing to its short cavity length. Since the exact emission wavelength is defined by the effective cavity length, continuous wavelength tuning is possible by changing the cavity length.

Continuous wavelength tuning of 18 Å was demonstrated by Chang-Hasnain et al. [26]. Ebeling et al. reported tuning of 22 Å [26]. The principle concept was current injection through a newly added electrode into the n-DBR mirror to heat up.

To achieve a wide range of tunability, a mechanical tuning method was experimentally demonstrated by us with an InGaAsP/InP VCSEL[27]. One of the mirrors was replaced with a movable external mirror. It was driven mechanically by pushing a tuning rod, as shown in Fig. 6. The wavelength was changed continuously over 40-86Å at 77 K without any mode hopping. This indicated clearly the possibility of continuous wavelength tuning in VCSELs, which later attracted interest for various applications.

Fig. 6.

External-mirror-driven frequency tuning of VCSEL.

A monolithic version of the external mirror VCSEL was realized by Chang Hasnain et al and showed a record high wavelength tuning range of 15 nm [28]. Frequency tunable VCSELs are now becoming important light sources for spectroscopic sensing and Datacom with wavelength division multiplexing. Very recently, tunable VCSELs have been used for a ranging sensor in combination with a light beam deflector where the emitting beam angle depends on the wavelength. The wavelength-tunable VCSEL will open an extensive variety of applications.

5.1

Patents by Tokyo Institute of Technology and PARC

In the initial stage of VCSEL development, two patents for inventions were submitted. One was for the invention by Kenichi Iga, Yasuharu Suematsu, Katsumi Kishino and Haruhisa Soda, all from Tokyo Institute of Technology, Japan that was applied on January 9, 1980 and was named “Surface-emitting laser”. Hereafter, this is called the Titech Patent in this article [29]. Independently of the Titech Patent, R. D. Burnham, D. R. Scifres, and W. Streifer, all from Xerox PARC, U. S. A. applied for a patent on September 13, 1979. Hereafter, this is called the Xerox Patent in this article. The patent was named “Transverse Light-Emitting Electroluminescent Devices”[30].

The Titech Patent and Xerox Patent were granted as independent inventions. Thus, two parties were considered as having independent patents.

However, the Xerox patent had not been recognized by us or the academic community regarding VCSELs until I received the 2002 Rank prize together with Burnham and Scifres as co-recipients. Later, I received the Streifer Award in 1992 as the VCSEL initiator. This award was established by Xerox in memory of the late William Streifer who was one of the inventors named in the Xerox Patent. Aside from the patent, no notable research and development related to VCSELs has been seen from the Xerox group to the best of our knowledge, even though they performed outstanding work on high-power edge-emitting semiconductor lasers.

5.2

Bell Labs’ microlaser

Many microlasers with the cavities of various sizes (3-100 μm) were manufactured on a GaAs substrate by Jewell et al. of AT&T Bell Laboratories [31]. The active region was a GaInAs single-quantum well of 100 Å thickness. The micro-cavity was formed by a pair of semiconductor DBR mirrors consisting of AlAs/GaAs multilayers. Later, the threshold currents of 1.5 and 1.1 mA under CW and pulsed conditionsz, respectively, were reported for a 5 μm square microlaser [32]. They possibly intended to study the micropillar lasers as the component of optical computers, which was the mission of the department at Bell Labs at that time. The micropillar structure was not applied in practical systems, since it is incompatible with full planar processes

5.3

Threshold current reduction efforts

With promising results on VCSEL realized in our initial feasibility study, many organizations worldwide became interested in VCSELs and started their own research in the 1990s. To show the progress of VCSELs, we tracked the threshold current reduction achieved by our group at Tokyo Institute of Technology and by other organizations. The major players and their threshold current results are listed in Table II.

From the first half of the 1990s, the threshold current of VCSELs was further reduced from about 10 mA to 50 μA. Jewell et al. of AT&T Bell Laboratories achieved a threshold current of 1.3 mA (pulsed) with three GaInAs quantum wells in 1989[31]. Soon after that, they realized a CW threshold current as low as 1.5 mA (CW) with a GaInAs single quantum well and a 5 μm square micropost structure [32].

Geels and Coldren of U. C. Santa Barbara reported a threshold current of 0.7 mA in a VCSEL with a strained InGaAs single-quantum-well active region [33]. This is the first report of a threshold current of less than 1 mA. Then, a sub-mA vertical-cavity surface-emitting laser with a 2.2 nm continuous tuning was demonstrated at Ulm University [34]. In 1993, Numai et al of NEC reported a threshold current as low as 190 μA with a pulsed condition at room-temperature with an InGaAs/GaAs strained single-quantum-well active region and an air post structure of 5 μm [35].

Deppe et al. of the University of Texas, Austin, achieved a room-temperature CW threshold current of 225 μA in a VCSEL with an InGaAs single-quantum-well active region [36]. The current was effectively confined by a 8 μm square native oxide aperture. The threshold current became as low as 91 μA by reducing the buried oxide aperture to 2 μm square [36].

These results were followed by a 70 μA threshold current achieved by our group in 1995 [18]. We employed an oxide AlAs/GaAs DBR to effectively confine the injection current and optical mode. A couple of months later, Yang et al of the University of Southern California published a paper claiming the achievement of a threshold of 8.7 μA [37], but this value has not been reproduced.

Besides the organizations described here, many others participated in the research on VCSELs and achieved low threshold currents of sub-mA, including the University of Ulm, the University of Würtzburg, Technical University of München, Technical University of Berlin, Sandia National Laboratories, Stanford University, and many corporations [4] ⁴⁾.

5.4

Innovations of VCSELs

Since the beginning of the 1990s, many research sectors have taken on the development of VCSELs and very low threshold devices reaching several μA have been reported. The turning points of the early stage of research are shown in Table I.

Table 1

Turning Points in Early Stage of Research (Our group, unless otherwise specified)

Since 1992, 850, 780, and 980 nm devices in lightwave systems have been commercialized. Aiming at exploring applications, 1300-1550-nm-wavelength, red-emitting AlGaInAs, and blue-emitting GaN devices are being developed as mostly cited in Ref. 4.

Our group has proposed some key concepts such as quantum-well VCSEL, multiquantum barrier (MQB), 1200 nm GaInAs/GaAs VCSEL, modulation schemes, phased-array VCSEL, Talbot-cavity VCSEL, tunneling injection, and tandem VCSEL[4].

Also, high-efficiency and high-power VCSEL were developed to demonstrate a wall-plug efficiency >60% [38] and arrayed output power >1 kW [39][40].

6.1

Advantages of VCSELs and possible applications

With the progress of the device performance, the features of VCSELs, such as high productivity, high efficiency, low power consumption, a circular beam, a single frequency, and a two-dimensional array, are becoming well recognized by the society and their application areas have been expanding explosively since the mid 1990s.

In Fig. 7, we show the oscillation frequency vs diameter of VCSEL shown in Fig. 3. We distinguish the single-mode and multimode version of VCSELs, and their application areas. In the case of single-mode devices, we can achieve single transverse mode operation by reducing the diameter of the device to a few microns. A circular beam and single oscillation frequency can be obtained. On the other hand, in the case of multimode devices that have larger diameters, such as several microns or larger, we can attain high optical output and high-speed modulation capability.

Fig. 7.

Single-mode vs multimode device categories and related applications

From 1999, the VCSEL was applied to high-speed LANs, such as Gigabit Ethernet and its mass production started. To date, VCSELs have been applied to various IoT fields, as shown in Table II and Fig. 8. The market size of production is forecast to reach US$ 2,500 M in 2020.

Fig. 8.

Some applications of VCSELs

Table 2

Possible Applications of VCSELs

6.2

Some commercial applications

6.3

Future developments

We show the possible wavelength bands of VCSELs and applications in Table III. The naming of the 1300-nm-wavelength band “O” follows the expression for silica optical fibers. The others are tentatively named by myself for convenience.

Table 3

Wavelength Bands of VCSELs and Applications

Abbreviations : com: communication, POF: plastic optical fiber

After I retired from Tokyo Institute of Technology, many value-added technologies have been proposed by Koyama and his group, such as continuous tuning [43], wavelength-temperature insensitivity (zero temperature coefficient), MEMS integration, phased arrays, beam steering, high frequency modulation (>40 GHz) [44], and hollow-waveguide integration (modulator/slow light).