GB2374201A

GB2374201A - Laser

Info

Publication number: GB2374201A
Application number: GB0108291A
Authority: GB
Inventors: Khaled Karrai; Richard J Warburton
Original assignee: Individual
Current assignee: Individual
Priority date: 2001-04-03
Filing date: 2001-04-03
Publication date: 2002-10-09
Also published as: GB0108291D0

Abstract

A laser comprises a chip structure having a surface 38, a first mirror 35 integrated with the chip structure 30 and a semiconductor gain medium 34 arranged in the chip structure between the surface and the first mirror 35. The laser has an optical fibre 50 at the end of which is arranged an integral second mirror 52. The end of the optical fibre is positioned above the surface of the chip structure to form a laser cavity with the semiconductor gain medium and the first mirror. The laser may have a piezoelectric tuning fork (60, Figure 9) attached to the optical fibre in order to vary the distance between the optical fibre and the chip structure. By varying the distance between the optical fibre and the chip structure the cavity length can be altered thereby tuning the laser. The first mirror may be a Bragg mirror having alternating semiconductor layers, alternating dielectric layers or may have multiple dielectric layers. The laser may be electrically or optically pumped. The laser may have a mesa structure 40. The method of tuning a VCSEL is also disclosed. The method may include using a control mechanism to identify a desired optical mode of the laser cavity, and a feedback control mechanism to measure a control parameter. The measured control parameter may be an interaction force between the end of an optical fibre and the surface of a chip structure.

Description

TITLE OF THE INVENTION LASER BACKGROUND OF THE INVENTION The invention relates to lasers, more especially but not exclusively to tunable lasers.

Lasers find a huge range of diverse applications. Some lasers are tunable, that is their lasing wavelengths can be varied. Typically, wavelength tuning is achieved by placing a gain medium in a cavity defined by two end mirrors and changing the cavity length. In this way lasing can be achieved across the wavelength range over which the gain medium is able to sustain sufficient stimulated emission for lasing action.

Semiconductor lasers have been in widespread commercial use for over a decade, for example in optical telecommunications and compact disc players.

Conventional semiconductor lasers are side-emitting and use the side facets of the chip structure as cavity mirrors. Another kind of semiconductor laser has been in development for several years, namely the so-called vertical cavity surface emitting laser (VCSEL).

As the name suggests, VCSEL's differ from conventional semiconductor lasers in that the optical axis of the laser cavity is vertical, that is the optical axis extends perpendicular to the plane of the semiconductor substrate. The laser mirrors in a VCSEL are usually Bragg reflectors formed by deposition of semiconductor epitaxial layers of alternating materials (e. g. GaAs and AlAs) when growing the semiconductor laser structure. The gain medium is formed by further semiconductor epitaxial layers deposited between the Bragg mirrors, typically by a multiple quantum well (MQW) structure.

Tunable lasers are important for a number of applications, for instance the testing of wavelength-division-multiplexing equipment in the telecommunications industry, environmental monitoring, printing and spectroscopy.

Figure 1 is a schematic drawing of a typical prior art tunable laser of the Litmann-Metcalf type in which a semiconductor diode 1 is arranged in an external cavity. The external cavity is formed of a thin film high reflection coating 2 on one

side of the semiconductor diode 1 and an external retroreflector 5. The cavity is folded by a diffraction grating 7 for wavelength selection. A collimating lens 8 is also provided in the cavity to collimate light output from the semiconductor diode 1. An anti-reflection coating 3 is arranged on the semiconductor diode 1 to prevent lasing in the semiconductor diode itself, forcing lasing on a single longitudinal mode of the cavity. The wavelength can be tuned by rotating the retroreflector 5 about a pivot point 6. Such devices are commercially available from a number of companies. These lasers typically have outputs of a few mW with sub-MHz linewidths. However, these lasers are quite bulky and the wavelength can only be changed slowly.

A monolithic and tunable device would be highly desirable for the telecommunications industry, preferably with vertical emission to allow integration into an array.

Two current approaches to provide such a device are now discussed. Both approaches have the disadvantage that they require formidably complicated processing.

One approach is to incorporate a voltage-tunable distributed feedback into the gain medium [7]. The linewidth tends to be large with this approach, and the tuning range is small (a few meV).

A second approach is to micromachine a distributed Bragg reflector which can be moved electrostatically with respect to the gain medium [1-4, 8]. The second mirror is buried in the sample or deposited as a dielectric Bragg stack on an etched back side of the substrate. This approach represents a processing tour-de-force. One of these lasers is now described in more detail with reference to the accompanying drawings.

Figure 2 of the accompanying drawings shows a wavelength tunable VCSEL of a type developed by M C Larson, J S Harris and co-workers at Stanford University [1-4]. Larson and Harris recognised that the short cavity lengths ofVCSEL's typically result in only a single longitudinal mode falling within the gain spectrum of the gain medium, making VCSEL's in principle capable of broad continuous wavelength tuning without mode hops, provided that some mechanism for moving one of the mirrors could be provided.

The basic idea of Larson and Harris is to dispense with the upper Bragg reflector of a conventional VCSEL and replace it with a mirror 14 formed in a membrane that is movably suspended over the surface of the semiconductor chip by an air gap 20. The upper mirror is formed of a semitransparent Au layer on a SiNxHy phase matching layer in turn on a GaAs supporting layer. (The GaAs may be substituted with InP for lasers operating at 1. 5 microns). The semiconductor chip, based on GaAs technology contains the lower Bragg mirror 10 (GaAs/AIAs layers) and a MQW gain medium 12 (5 x InGaAs quantum wells with GaAs barrier layers). The membrane upper mirror 14 is fabricated by sophisticated etching to leave it suspended by four cantilever struts 22 from contact pads 18 (Ti/Au) arranged on spacers 16 (AlAs) that are mesas on the upper surface of the semiconductor chip. The vertical position of the upper mirror 14 is controlled electrostatically by a bias voltage to tune the laser. The laser cavity is thus formed between the lower distributed Bragg reflector (DBR) mirror 10 and the upper membrane mirror 14, with the gain medium 12 being provided by the MQW structure arranged above the DBR mirror 10. The gain medium extends over a distance of two lasing wavelengths with the five quantum wells arranged at the five antinodes.

The technological complexity of fabricating lasers of the Harris/Larson design is immense. Working lasers have been fabricated using this design, but the tuning range of reported devices is modest. For example, it is believed that no single laser of the Harris/Larson design can provide tunability between 1. 3 and 1. 55 microns.

Moreover, it is believed that the laser powers are relatively low and the yields are also low, as would be expected owing to the complexity of the processing. Huge research and development resources continue to be invested to solve these problems and produce a commercial product.

SUMMARY OF THE INVENTION According to the invention, there is provided a much simpler tunable VCSEL laser which is amenable to mass-production with high yields, can provide a broad tuning range, and whose design presents no technological challenges, since it is based on technology used in current VCSEL's in combination with technology borrowed from the field of scanning probe microscopy (SPM). Moreover, the laser is capable of relatively high output powers, at least in the milliwatts range, possibly much higher.

According to the invention, there is provided a laser comprising: a chip structure having a surface, a first mirror integrated with the chip structure and a semiconductor gain medium arranged between the surface and the first mirror ; and an optical fibre having an end on which is arranged an integral second mirror, the end of the optical fibre being positioned above the surface of the chip structure to form a laser cavity with the semiconductor gain medium and first mirror.

The complex nanofabricated electrostatic mirror of the Harris/Larson design is thus replaced with a simple arrangement of optical fibre with integral mirror.

As well as the practical advantages that follow from the simplicity of the design of the present invention, there is an important inherent performance advantage over the Harris/Larson. Namely, the design of the present invention provides a cavity with high structural stiffness. The optical fibre carrying the second mirror is structurally very rigid along the optical axis. By contrast the membrane mirror in the Harris/Larson design is suspended with compliant spring elements (the cantilevers). This compliance is necessary in the Harris/Larson design since the mechanical force of the spring elements must not exceed the electrostatic force that can practically be generated by reasonably low voltages applied to the contacts in order to move the membrane mirror up and down.

The compliance in the Harris/Larson design is likely to be very problematic at modest to high powers where there is a substantial photon pressure. The photon pressure can push the membrane-mirror away from the structure, and out of resonance. The net result is an unstable laser.

Tuning in the present invention is provided simply by moving the optical fibre up and down relative to the chip structure. To effect this, the laser may be further provided with a positioning device for varying the distance between the optical fibre and

the chip structure, i. e. along the optical axis, thereby to vary the cavity length by moving the second mirror (fixed to the fibre) relative to the first mirror (part of the chip structure).

More specifically, according to another aspect of the invention there is provided a method of tuning a VCSEL comprising: (a) providing a chip structure having a surface, a first mirror integrated with the chip structure and a semiconductor gain medium arranged in the chip structure between the surface and the first mirror; (b) providing an optical fibre comprising an end provided with an integral second mirror ; (c) positioning the optical fibre a distance above the surface of the chip structure to form a laser cavity with the semiconductor gain medium and first mirror; (d) injecting carriers into the semiconductor gain medium to provide population inversion; and (e) varying the distance between the surface of the chip structure and the optical fibre to effect tuning of the laser cavity.

In a preferred embodiment, the step of varying the distance between the surface of the chip structure and the optical fibre includes using a control mechanism to identify a desired optical mode of the laser cavity under a feedback control mechanism having a measured control parameter.

Typically the scanning of the cavity will be provided by a piezoelectric actuator (high rigidity component) arranged under the chip structure (high rigidity component), with the optical fibre being fixed along the optical axis in a housing (high rigidity component). The whole laser package can thus be made very rigid, whereas this is inherently impossible with the electrostatically moved mirror of the HarrislLarson design. As already mentioned, this lack of rigidity of the membrane mirror mounting in the Harris/Larson design makes it prone to unstable oscillation which is highly damaging to stable laser operation.

The measured control parameter to identify near-contact of the fibre with the semiconductor chip and therefore to identify the optical mode may be an interaction force between the end of the optical fibre and the surface of the chip structure. In an embodiment, the interaction force is measured with a differential signal from a coupled

oscillator comprising first and second oscillators, one of which is mechanically linked to the optical fibre. More generally, any suitable interaction force known from SPM may be used.

The invention may also be applied to non-tunable lasers. For example, the fibre could be fixed relative to the chip structure at the time of fabrication during an optimisation procedure in which the cavity length is varied to provide a particular desired lasing wavelength in the finished laser.

The second mirror is preferably the output coupler of the laser. In other words, the second mirror is preferably very slightly semi-transparent to allow laser output through it from the cavity. This is particularly advantageous, since the laser output is then automatically coupled into the optical fibre. The optical fibre can form the start of an optical communications component, for example a transmitter. Alternatively, for instrumentation or medical applications, the optical fibre may be used simply for delivery of the laser light.

The cavity size (distance between first and second mirrors) is related to the lasing wavelength. In general, cavity size is preferably kept to a relatively small number of lasing wavelengths. The number of wavelengths of the cavity size should be kept small enough that the possibility of mode hopping is eliminated. In other words, the gain medium should be incapable of providing sufficient gain to support lasing at one wavelength higher or lower than the design number of wavelengths in the cavity. In practice, it is preferred that the cavity size is less than 50 lasing wavelengths, more preferably 10 lasing wavelengths or less, in particular cases 9,8, 7,6, 5,4, 3 or 2 lasing wavelengths or less. The cavity size in numbers of lasing wavelengths in many embodiments is likely to be 1,2, 3,4 or 5, or 3/2,5/2, 7/2 or 9/2. A cavity size of less than 1/2 lasing wavelength is clearly not possible. There are two important design criteria. First, the gain medium should be positioned more than about 50 nanometres from the semiconductor surface so that surface electric fields cannot degrade the performance. Secondly, the gain medium should be positioned at anti-nodes of the electric field distribution in the cavity to maximise the interaction between the gain and the optical field.

Clearly, such a short cavity requires that the optical fibre be placed in very close proximity to the surface of the chip structure. In particular, in the case of a tunable laser,

there is the requirement to be able to move the optical fibre up and down relative to the surface in a highly controlled manner over very short distances. This is where technology can usefully be borrowed from scanning probe microscopy (SPM), in particular scanning near-field optical microscopy (SNOM). All types of SPM involve approaching a tip to a surface and scanning the tip over the surface under some kind of feedback process. The standard technique for SNOM scanning uses a tuning fork with an optical fibre that tapers into a SNOM tip, the optical fibre being bonded along one of the prongs [5. 6].

Borrowing the SNOM technology, the laser preferably further comprises a tuning fork having first and second prongs, the optical fibre being secured along the first prong. From SNOM it is known that this arrangement can be reliably used to bring a delicate tip to within a few nanometres of a sample surface without the tip ever making contact with the sample. Furthermore, the distance between tip and surface can be controlled to sub-nanometre precision with a feedback mechanism up to tip-sample separations of about 10 nm with current technology.

The tunable laser of a preferred embodiment comprises a tuning fork, one leg of which is bonded to the optical fibre, in order that the second mirror can be brought up to the semiconductor surface controllably, without damaging the mirror through contact with the semiconductor surface. The mirror can then be withdrawn by a certain distance, defined by applying a specific voltage to a piezoelement, so that the distance between the two mirrors can be carefully set. (The thickness in the semiconductor between the surface and the second mirror is known precisely from the epitaxial growth).

More generally, the tuning fork may be exchanged for any suitable oscillator, either pure mechanical like the tuning fork or hybrid mechanical/electronic, as described in detail in references [5] and [6], the contents of which are incorporated herein by reference. Still more generally, other distance regulation techniques could be borrowed from SNOM, scanning force microscopy, atomic force microscopy (AFM) or other SPM techniques.

Using a SNOM technology for the feedback control is advantageous in that: (i) it provides a large distance regulation range (in SPM terms) of several nanometres (at least in the case of the tuning fork approach) ; and (ii) it is automatically compatible with an optical fibre (since SNOM uses an optical fibre).

The second mirror is preferably a Bragg mirror arranged on the end of the optical fibre extending over the core of the optical fibre in the case of a simple optical fibre. In alternative embodiments, a double clad fibre can be used with the light coupled into the inner clad of the fibre, rather than the core. In that case the second mirror needs to extend over the inner clad. This arrangement may be used for a cladding pumped fibre laser. In either case, the Bragg mirror may be formed by deposition of multiple dielectric layers using conventional methods.

The optical fibre is preferably provided with chamfered or otherwise tapered cladding region. This reduces the risk of the dust adhering to the cladding preventing close approach of the fibre to the sample surface. In addition, if the chamfering is performed prior to formation of the second mirror, and the second mirror is formed by evaporation, this will reduce the area of the second mirror to something similar to the area of the optical fibre's core. This will increase losses associated with the higher order lateral modes, thereby encouraging the laser to operate on the fundamental lateral mode.

This is desirable as the fundamental mode is optimally coupled into the optical fibre.

The first mirror is preferably a Bragg mirror formed by deposition of multiple layers, also using conventional methods.

The semiconductor gain medium in one embodiment is formed by a twodimensional carrier gas, that is carrier confinement in one dimension, typically by quantum wells where the carriers are both electrons and holes.

The semiconductor gain medium in another embodiment is formed by onedimensional carrier gas, that is carrier confinement in two dimensions, typically by quantum wires.

The semiconductor gain medium in another embodiment is formed by a zerodimensional carrier gas, that is carrier confinement in three dimensions, typically by quantum dots (sometimes called boxes), rings or spheroids where again the carriers are both electrons and holes. The carriers may also be other particles or collective excitations such as polaritons, excitons or plasmons.

The quantum dots or other features are preferably formed by a self-assembly technique. Self-assembly can be used to provide a natural variation in the quantum confinement energies, thereby serving to broaden the gain spectrum of the gain medium.

Using quantum dots in the gain medium allows a very wide spectral region to be accessed. Emission at around 1. 1 microns is achievable with GaAs-based chip structures. This can be increased to 1.3 microns by changes in the growth. InAs dots in InP can be used for lasing around 1.5 microns, since they have broad emission in this region. InAs quantum rings can be used for lasing at around 950 run. InAlAs or InP dots can be used to provide lasing in the red. GaN dots can provide lasing in the blue.

In any of the above reduced dimensional carrier schemes, the semiconductor gain medium may advantageously be formed in a mesa. This also enhances the output from the fundamental lateral mode.

If a stripe or mesa is used, a positioning device capable of aligning the optical fibre over the stripe or mesa is preferably provided in any tunable laser embodiment. In practice, a positioning device is preferred which provides motion in three axes x, y & z, where z is used to denote the optical axis and x & y are perpendicular thereto.

A tilt mechanism by which the first and second mirrors can be brought into parallelism to within a milliradian is preferentially included. This enables the optimum cavity finesse to be achieved. In one embodiment, this is achieved by a tilt stage, containing high thread density adjustment screws for coarse alignment, and piezoelectric actuators for fine alignment, as is standard in high quality mirror mounts.

In one form, the laser is optically pumped. The pump radiation may be delivered through the optical fibre or independently of the optical fibre, for example by free space optics or another optical fibre. Preferably the pump wavelength is chosen and the second mirror designed so that the second mirror is effectively transparent, i. e. highly transmissive, to light of the pump wavelength. In addition the chip structure is designed having regard to the pump wavelength so that it has strong absorption at the pump wavelength. An optical pump laser could even be fabricated integrally with the chip structure.

In another form, the laser is electrically pumped using integrated components in the semiconductor heterostructure. These components can be based on existing VCSEL technology and designs. In one embodiment, the gain medium is formed in a mesa into which carriers are injected by contacts formed on the mesa.

The design may be implemented in a variety of different materials for the chip structure. The GaAs/AIGaAs/AIAs and InGaAs/AIGaAs systems are well developed.

InP-based devices could also be fabricated by avoiding semiconductor superlattice mirrors (in the InP material system, there are no two lattice-matched materials with a large enough refractive index contrast to make broadband Bragg mirrors).

A semiconductor lower cavity mirror can be avoided by using epitaxial (epi) liftoff techniques or back-etching of the semiconductor substrate, as described further below. A key advantage of both the back-etch and epi lift-off approaches for the present invention is that they decouple the lower cavity mirror design from the gain medium design, since fabrication of the lower cavity mirror is removed from the semiconductor heterostructure. This opens up the invention to any semiconductor heterostructures where suitable gain media can be provided, including wide and narrow band gap II-VI materials and a further variety of III-V materials. Specifically, the choice of semiconductor heterostructure materials system is no longer limited to those which allow a Bragg mirror to be fabricated. Inorganic semiconductor gain media may also be used. These approaches thus open up the present invention to tunable lasers operating over a very wide range of wavelengths from the infrared to the ultraviolet and beyond.

It is emphasised that the novel cavity designs described herein are relatively simple and inexpensive to implement. Moreover, no lateral processing of the sample is needed. Further, no arduous alignment of an optical fibre for coupling out the laser emission is needed, since the laser emission is automatically located at the fibre by virtue of the output cavity mirror being formed directly on the end of the fibre core.

BRIEF DESCRIPTION OF THE DRAWINGS For a better understanding of the invention and to show how the same may be carried into effect reference is now made by way of example to the accompanying drawings in which: Figure 1 shows a commercially available tunable laser of the Litmann-Metcalf type according to the prior art; Figure 2 shows a tunable vertical emitting surface cavity laser (VCSEL) proposed in the prior art where tuning is carried out by an electrostatically movable mirror ; Figure 3 shows a tunable optically-pumped VCSEL according to a first embodiment of the invention comprising a chip structure incorporating a semiconductor gain medium and one cavity mirror, and an optical fibre incorporating a second cavity mirror; Figure 4 shows the structure and optical field of an example laser of the first embodiment having a cavity length of one-and-a-half wavelengths; Figure 5 shows an example of a semiconductor chip structure usable in the first embodiment based on a quantum well or quantum dot gain medium ; Figure 6 shows a tunable optically-pumped VCSEL according to a second embodiment of the invention where the optical fibre has chamfered sides; Figure 7 shows a tunable electrically-pumped VCSEL according to a third embodiment of the invention; Figure 8 shows an example of the semiconductor chip structure usable in the third embodiment based on a quantum well or quantum dot gain medium;

Figure 9 shows an arrangement for cavity tuning suitable for any of the first to third embodiments based on a tuning fork secured to the optical fibre ; Figure 10 shows an example range of the feedback signal from the tuning fork, plotting the shift in the resonance frequency'Af of the tuning fork as a function of distance'Zd'between the optical fibre tip and the adjacent surface of the chip structure; Figure 11 shows schematically a packaged tunable laser module in section; Figure 12 shows the packaged tunable laser module of Figure 11 together with ancillary control electronics ; Figure 13 shows a module combining a pump source as well as a packaged optically-pumped tunable laser module; Figure 14 shows an alternative chip structure using back-etching; Figure 15 shows an alternative chip structure incorporating an epitaxial lift-off semiconductor heterostructure; and Figure 16 shows an example epitaxial lift-off semiconductor heterostructure.

DETAILED DESCRIPTION Figure 3 shows a tunable optically-pumped VCSEL according to a first embodiment of the invention. The laser comprises two principal structural elements, a chip structure 30 and an optical fibre structure 50.

The chip structure 30 incorporates a semiconductor gain medium 34 and an underlying cavity mirror 35, both epitaxially grown on a semiconductor substrate 36.

The epitaxial growth may be performed by molecular beam epitaxy (MBE) or other suitable technique. In addition, the upper surface of the chip structure 30 is provided with an antireflection (AR) coating 33, as is conventional for tunable external cavity semiconductor lasers. The AR coating 33 is a dielectric multilayer deposited separately, for example by ion-assisted deposition at room temperature. The AR coating 33 serves to decrease the reflectivity of the surface from the standard 0.3 for a semiconductor-air interface. Reflectivities of 10-4 are routinely achievable with ionassisted deposition at room temperature. Room temperature deposition allows packaged lasers to be coated and tested immediately afterwards. The coatings are rugged enough to survive the heating and intense optical fields during laser operation.

In prior art tunable VCSEL lasers of the Litman-Metcalf type, reflectivities less than 10-3 are typically required in order to avoid instabilities and to suppress residual effects on the gain spectrum. However, in the presently proposed devices, higher reflectivities at the semiconductor-air interface can be tolerated, because the proposed cavity lengths are much smaller, only of the order of magnitude of the lasing wavelength, e. g. 1 to 10 lasing wavelengths. Especially for the shortest proposed cavity lengths of only one or a few lasing wavelengths, it may prove that an AR coating can be dispensed with altogether.

The optical fibre structure 50 comprises a single-mode optical fibre 51 having a core 51 a and a cladding 51 b. The optical fibre 51 terminates at an end on which is deposited a further cavity mirror 52 in the form of a Bragg reflector constituted by multiple dielectric layers of suitable materials. For ease of representation, the drawing is highly schematic in terms of the relative dimensions. In practice, the Bragg reflector will be between 1 to 3 microns thick. The optical fibre cladding 51b will

have an outside diameter of typically 125 microns. he optical fibre core will have a diameter of typically 5-10 microns.

The mirror 52 may be fabricated as a dielectric Bragg reflector by thin-film deposition on the optical fibre using ion-assisted deposition, which allows high quality thin-films to be produced at room temperature. No special preparation of the optical fibres is necessary. In situ monitoring may be carried out by measuring the broadband reflectivity of the fibre as deposition proceeds. The coatings can be optimised by adjustments to the thicknesses and compositions of the various layers in the stack. Reflectivities of up to 99.8% can be routinely achieved, and even higher reflectivities are possible by careful control of the deposition.

As illustrated in the figure, the vertical laser cavity is formed by positioning the mirrored end of the optical fibre 50 in close proximity to the surface of the chip structure 30 separated by a distance Zd, that will typically be from a few nanometres up to several hundred nanometres. In this way, a vertical cavity is created that has a length of only one, or a small number of wavelengths. The laser of this embodiment is optically pumped. Optical pumping may conveniently be carried out through the optical fibre 51, as indicated by the downwardly pointing arrow Bp in the figure. The pump wavelength (or pump wavelength range) is selected not only to be smaller than the lasing wavelength, but also in a region where the mirror 52 is transparent. The laser output is coupled directly into the core 51a of the optical fibre 51, as indicated by the upwardly pointing arrow bu in the figure.

Tuning of the laser is achieved simply by moving the optical fibre 50 towards or away from the chip structure 30 to vary the distance Zd, as indicated in the figure.

An advantage of this approach is that the cavity length can be made to be one or only a few lasing wavelengths. This means that the cavity length can be made so short that, across the entire gain spectrum of the gain medium where laser action is possible, there is only one longitudinal lasing mode that can arise, namely the fundamental. The laser can thus be highly stable, since there are no sustainable higher order longitudinal modes in the cavity. This is vastly superior to current commercially available tunable VCSEL's of the Litmann-Metcalf design where mode hopping is usual.

Another major advantage of this arrangement is that no complicated lateral processing or micro-machining of the chip structure is required.

A further significant advantage is that the laser output (upwardly pointing arrow ki in the figure) is automatically aligned with and coupled into an optical fibre. The separate procedure for alignment and bonding of the laser output to an optical fibre (so-called connectorisation or pigtailing) that is conventionally necessary, e. g. with the Harris/Larson design, is no longer needed. This is important, since the pigtailing procedure is quite difficult and prone to losses.

The design places no fundamental restriction on the tuning range. A wide range of tuning ranges can be obtained, limited only by the characteristics of the gain medium and the wavelength ranges over which the cavity mirrors and optical fibre are functional.

Tuning range is important in many applications. Specifically, in optical telecommunications, it is presently not possible to use a single device to probe both 1.3 and 1.5 micron bands (second and third telecommunication windows). Such a device would also be very useful for dense wavelength division multiplexed (DWDM) systems. A large tuning range is also important for gas sensing as presently a separate laser must be used for each gas. Moreover, with an extended tuning range, semiconductor lasers could replace bulky, high-cost Ti: sapphire lasers for spectroscopy.

The tuning range of the laser of the first embodiment is limited by the semiconductor gain medium (semiconductor diode). As is known, semiconductor quantum wells have a relatively large tuning range, since the energy-independent density of states facilitates rapid band filling. If the well is too wide however, subbands higher than the first must be inverted, limiting the tuning range. If the well is too narrow, carriers spill out into the bulk-like barriers, also limiting the tuning range. A compromise exists at intermediate well widths, giving typically 50 meV of tuning on the first interband transition, almost independent of the centre wavelength.

This tuning range can be extended by using also lasing on the second interband transition of the quantum well, but at the expense of very large pumping powers.

Figure 4 is a more detailed schematic drawing of an example of the laser of the first embodiment, showing the nature of the laser cavity in more detail. The letters H and L are used to label elements of higher and lower refractive index respectively.

The cavity mirror in the semiconductor structure 35 is formed of a few tens of alternating layers of semiconductor material (e. g. Alo. lsGaossAs and AlAs), with the higher band gap material AlAs being of lower refractive index and the lower band gap material Alo 15Ga 8sAs being of higher refractive index. The upper cavity mirror 52 is a Bragg stack using alternate layers of silica (Si02 refractive index 1.45) and titania

(tri02 refractive index 2. 25) on the end of the silica glass fibre 51. The longitudinal laser mode (indicated by square of the photons'electric field E2 in the figure) is shown with the sine-curve. A standing wave is set up between the inner end surfaces of the cavity mirrors 35 and 52 and evolves to a peak centred in the gain medium 34, as indicated by the dashed line. As can be seen the total cavity length of the laser shown is 3V2 where k is the wavelength (three periods in E2). To ensure that the electric field maximum is centred in the gain medium 34, a gain medium spacer layer 37 of suitable thickness is provided (e. g. AIo3Gao7As). Similarly, a surface spacer layer 38 (e. g. Alo sGao.7As) is provided on the surface side of the gain medium below a thin passivating cap layer 39 (e. g. GaAs) which terminates at the surface of the chip structure. In this example no AR coating is shown. An AR coating may or may not be provided, but will generally not be deposited with the semiconductor layers. It will be appreciated that exact matching will not be provided owing to the tunable nature of the laser. The structure is merely designed to provide precise matching near the middle of the intended tuning range.

Finally, it will also be appreciated that the figure is highly schematic, not only in terms of dimensions, but also since it does not take account of the refractive index differences between the various materials and air which results in a distance defined by a particular fraction of a wavelength being quite different in air, glass or semiconductor.

Moreover, an alternative possibility (not shown) is to increase the gain by using two gain media, each placed at an anti-node of the electric field, simply by increasing the length of the cavity.

Figure 5 shows an example layer sequence for the chip structure of the first embodiment. The chip structure is deposited on a GaAs substrate on which is grown a conventional buffer layer 36a, which may incorporate a superlattice or the like as is conventional. The lower cavity mirror 35 is then deposited. It is made up of 25 pairs of AlAs and AloGaossAs layers of 71 nm and 61 nm thickness respectively (for operation at a wavelength of 850 nm), with a final further 71 nm AlAs layer on top.

These thicknesses correspond to a quarter of the intended centre laser wavelength in the respective materials. There then follows a spacer layer 37 of Alo. 3Gao 7As. The layers of the gain medium 34 then follow. In one example, the gain medium 34 is

constituted by a GaAs/Alc3Gao7As quantum well, or by two, three, four or five closely spaced quantum wells. In another example, the gain medium is constituted by 10 InAs/GaAs quantum dot layers. There then follows a further spacer layer 38 of Alo 3Gao 7As and a GaAs cap 39 (5 nm thick). The distances from the upper end of the lower cavity 35 to the centre of the gain medium and from the gain medium to the chip surface are both 125 nm, as shown.

Using semiconductor quantum dots for the gain medium is an attractive option. Quantum dots have ideal characteristics for laser applications because the discrete energy levels make it very easy to create a population inversion. Moreover, quantum dot lasers currently have the lowest threshold current and highest power of any semiconductor laser. In current quantum dot laser designs, performance is limited by fluctuations in dot dimensions. In the prior art, much design effort has been concentrated on reducing this variation in dot size so that large numbers of homogeneously dimensioned dots can be produced to improve performance.

However, with the present laser design the natural inhomogeneity of self-assembled quantum dot ensembles can be positively exploited to force lasing over the entire distribution of dots. This approach is based on the recognition that the natural inhomogeneity of the dot ensembles broadens the absorption and emission bands, which is highly desirable for a gain medium to be used in a tunable laser. The quantum dot embodiment of the invention thus provides a novel way of designing efficient and highly tunable semiconductor lasers. Other related embodiments can exploit the same effect in other zero or one dimensionally confined structures, such as quantum wires, quantum rings etc.

Figure 6 shows a tunable optically-pumped VCSEL according to a second embodiment of the invention. In this embodiment, the optical fibre has chamfered sides 5 c to reduce the flat area at the end of the fibre 51 including the end of the core 5 a. but is otherwise the same as the first embodiment. For the sake of brevity, like components are not described again, but referred to using the same reference signs.

The chamfering may be performed by mechanical polishing. Alternatively etching could be used, adopting techniques used for SNOM tip fabrication, such as preferential etching. The chamfering is performed prior to deposition of mirror.

Chamfering may reduce the diameter of the flat end of the fibre containing the core to a diameter of below one of 50, 40, 30,20 or 10 microns and above the diameter of the core, e. g. above 5 or 10 microns.

The chamfering reduces the risk of the optical fibre crashing into the surface of the chip structure, for example as a result of any protruding defects on the core. This risk can be understood when it is recalled that standard optical fibres will have a diameter of 125 microns and the tip-surface distance Zd will only be of the order of tens of nanometres.

In addition, by depositing the mirror 52 onto the chamfered tip, the area of the mirror 52 is reduced to something more similar to the area of the optical fibre's core 5 a.

This reduction in area will reduce any stray capacitance or other electrical interference effects induced by the second mirror, which may be important if an electrical feedback mechanism is used for positioning the optical fibre in proximity with the chip surface, as described further below.

Furthermore, reducing the lateral extent of the mirror 52 will aid suppression of any higher order lateral modes, so that only the fundamental lateral cavity mode can lase with the fundamental longitudinal cavity mode.

The laser of the second embodiment is optically pumped, similar to the laser of the first embodiment.

Figure 7 shows a laser according to a third embodiment which is electrically pumped. The optical fibre structure 50 is the same as for the second embodiment, so is not described further. The chip structure has the same basic layer sequence as the first and second embodiments, but differs in order to accommodate the contacts necessary for electrical injection of carriers into the gain medium. To accommodate

electrical contacting, the gain medium 34 is arranged in a mesa 40 upstanding from the surrounding chip surface. This allows metal or metallic material 41 to be deposited around the gain medium using silicides, metallisation and other conventional semiconductor fabrication techniques, thereby to provide electrical connection to a p-type contact above the gain medium 34. All the layers above the gain medium are p-type, specifically the upper spacer 38 and cap 39 are p-type. The electrical contact to the semiconductor may be formed by a ring contact (not shown) on top of the p±type cap layer 39, as is known in the art. An n-type contact is formed from underneath, by choice of an n substrate 36 and n-doping of the layers making up the lower cavity mirror 35 and lower spacer 37. The lower cavity mirror 35 is arranged below the mesa (but could be inside the mesa in an alternative embodiment).

The lower spacer 37 extends from the main buried part of the substrate up into the mesa 40. An AR coating 33 is also shown on top of the mesa 33.

Electrical pumping enhances the simplicity of the packaged device, since the fibre coupler required for introducing the pump beam into the fibre in the optically pumped designs, is omitted.

Figure 8 shows an example layer sequence of the chip structure for the third embodiment. The chip structure is deposited on an n±GaAs substrate on which is grown a conventional buffer layer 36a, which may incorporate a superlattice or the like as is conventional, which is also n-type. The lower cavity mirror 35 is then deposited. It is made up of 25 pairs of n-AlAs and n-Alo 15Gao 85As layers of 71 nm and 61 nm thickness respectively, with a final further 71 nm n-AlAs layer on top.

These thicknesses correspond to a quarter of the intended centre laser wavelength (850 nm in this example) in the respective materials. There then follows a first portion of spacer layer 37a of n-AIo3Gao 7As which in one area leads to the mesa 40, but otherwise terminates in a surface 43. The spacer then continues into the mesa 40 with a second portion of spacer layer 37b. The spacer layer portions 37a and 37b may be made of different materials to allow stop etching to be used when defining the mesa 40. The layers of the gain medium 34 then follow. These are in one example a GaAs/ Alo 3Gao 7As quantum well. In another example, these are 10 InAs/GaAs quantum dot layers. There then follows a further spacer layer 38 of p-Alo 3Gao 7As and a p±GaAs cap 39 (5 nm thick). Electrical contact is made to the upper end of the mesa 40 with a

ring contact 42 arranged on the top surface of the mesa 40. The distances from the upper end of the lower cavity mirror 35 to the centre of the gain medium and from the centre of the gain medium to the chip surface are both 125 nm. The centre of the gain medium 34 is indicated with a horizontal dashed line. The mesa is 10 microns in diameter in the illustrated example.

Having now described the basic structure of tunable lasers according to several embodiments of the invention, a mechanism for providing cavity tuning is now described.

Figure 9 shows an arrangement for cavity tuning suitable for any of the first to third embodiments. To tune the laser, it is necessary to provide an arrangement for controllably varying the cavity length. The arrangement shown is based on a piezoelectric stack with a force feedback scheme that is widely used for SNOM. The optical fibre structure 50 is secured by adhesive bonding along one leg, prong or tine of a quartz tuning fork 60 (an example leg length is 6 mm) so that the mirrored end of the optical fibre structure 50 protrudes by a small amount (e. g. 0.5 mm) from the end of the tuning fork leg. The two legs of the quartz tuning fork 60 collectively form a coupled oscillator.

During use, the tuning fork is driven to oscillate so that its legs move towards and away from each other. The total lateral leg excursion is sufficiently small so as to have no substantial effect on the lasing action. For example, the total excursion will typically be sub-nanometre and may be in the picometre to angstrom range. When the mirrored end of the optical fibre structure 50 approaches the surface of the chip structure 30, the oscillating leg to which the optical fibre is attached experiences a damping force from the interaction between the fibre end and chip surface. The physical origin of the interaction force is well understood. It arises from an interaction between the fibre and semiconductor surface mediated by an ever-present surface layer. As the separation Zd between fibre end and sample surface reduces, the interaction force increases. The interaction force is picked up electrically by monitoring a differential signal between pick-up electrodes (not shown) placed on respective ones of the tuning fork legs. In order to have an even better defined z=0 position, it is possible to coat the fibre and/or surface with a hydrophobic coating.

The tuning fork and optical fibre system described above is well known from SNOM. The original description of this system can be found in reference [5], the contents of which are incorporated herein by reference. The preferred arrangement for the present laser is that described in reference [6], the contents of which is incorporated herein by reference. The development of reference [6] is to use a custom-designed pre-amplifier circuit located close to the tuning fork to enhance the sensitivity of the measurement of the interaction force. This allows the interaction force to be measured for a separation z of up to several tens of nanometres.

As described in references [5] and [6] there are many variations to the coupled oscillator theme, not all of which use tuning forks. Any of these variations may be used in the present laser to control the separation Zd between fibre end and chip surface. In addition, other SNOM techniques for controlling the fibre-surface separation may also be used, for example those described in reference [9] and [10].

Furthermore, depending on the tuning range (and thus separation range) required, other SPM techniques may be used, for example techniques developed for atomic force microscopy (AFM) or scanning tunnelling microscopy (STM). Once the semiconductor surface has been located by the fibre-tuning fork assembly, the fibre can be withdrawn into the operating position by a piezo acctuator which also serves to tune the wavelength. If the operating position is some distance away from the surface, the Fabry-Perot etalon that is formed between the first and second mirrors can be used to locate the position. Fringe counting is used to controllably vary the fibre-to-surface separation. This techniques is referred to as interferometric tuning and is known from AFM and STM.

Figure 10 shows an example range of the feedback signal from the tuning fork, plotting the change in the resonance frequency'Af of the tuning fork as a function of distance'Zd'between the optical fibre end and the adjacent surface of the chip structure. The tuning fork used had a resonance at approximately 32 kHz. The resonance frequency shifts appreciably for separations Zd less then about 10 nm.

Figure 11 shows schematically in section a packaged tunable laser module 100 according to any of the first to third embodiments using a tuning fork force feedback arrangement as described with reference to Figures 9 and 10. The optical fibre

structure 50 is fed through into the interior of a housing through a top flange 94 having an inner surface on which the tuning fork 60 is mounted as well as a pre amplifier circuit board 70 as described in reference [6]. The top flange 94 seats in an upper opening of a cylindrical housing 92 with integral base. The housing 92 and top flange 94 thus provide a sealed environment for the moving parts of the tunable laser. On the outside of the top flange 94 the optical fibre structure 50 is cemented in with cement to provide stress relief. The optical fibre structure then continues with protective outer covering 98 as a packaged optical fibre cable.

In order for optimum laser performance, the first and second mirrors must be parallel to each other. This is achieved with a tilt stage incorporated as part of the positioning stage 90. The tilt state comprises both a coarse and a fine angular adjustment as is standard in mirror mounts allowing adjustment of the first mirror with respect to the second to milliradian precision.

To provide cavity tuning, that is variation of the fibre-chip separation, the chip structure 30 is mounted on a piezoelectric stack, which provides a fine z-tuning element 80. Any suitable piezoelectric or other actuator with sub-nanometre resolution could be used. The piezoelectric stack 80 is in turn mounted on a coarse actuator also providing actuation in the z-direction, but with a coarser resolution. The coarse z-tuning element is preferably an inertial positioner as described in reference

[I I], the contents of which are incorporated herein by reference. Specifically the ztuning element of the best mode is that of the third embodiment of reference [11].

This z-tuning element moves in steps. The coarse and fine z-tuning elements are chosen so that the travel of the fine tuning element is greater than the step size of the coarse tuning element so that a full range of z positioning can be attained. However, it will be understood that there is a very wide range of positioning elements available that could be used. In some cases, a single positioning element may provide both coarse and fine z-tuning.

In some embodiments, it may be either necessary or advantageous also to provide coarse positioning in one or both of the x and y directions, i. e. in the two mutually perpendicular axes to z. For example, if the gain medium is arranged in a mesa, as in the third embodiment, x-and y-tuning elements may be incorporated into the coarse positioner 90. An xyz-stack of inertial positioners may be used, as

described in reference [11]. The ability to position in x and/or y may also be useful to move the lasing region around the chip structure.

This facility could be used in case bum-out or other failure occurs as a result of local damage to the chip structure in order to move to a fresh undamaged part of the chip structure.

Another use of this facility could be to allow lateral variations in the chip structure to be exploited. For example, as a result of the epitaxial growth, there may be gradual variations in the layer thicknesses across the chip structure. This may arise from directionality of molecular beam sources. Such variations may occur without deliberate intent, or may be deliberately incorporated into the chip structure, for example by not rotating the substrate during MBE growth. Similar effects may be produced by metaloorganic deposition techniques such as metalloorganic MBE (MOMBE) or metaloorganic chemical vapour deposition (MOCVD). In addition, if multiple mesas are fabricated on a single chip structure, lateral x and y positioning may be used to switch between different mesas. The quantum states in the gain media in the different mesas may be made to vary deliberately from mesa to mesa, e. g. by variations in lateral confinement induced by the mesa sides.

Natural statistical variations in ensembles of self-organised quantum dots, rings, etc. may also result in lateral performance variations that can be probed with x and y positioning.

Figure 12 shows the packaged tunable laser module 100 together with ancillary control electronics shown schematically as boxes. As previously stated, the fibre-chip separation is controlled under feedback by an electronic controller 110. The controller 110 may be microprocessor controlled, for example by a computer. The principal control inputs/outputs to/from the controller 110 are: (i) the interactionforce-defining input from the tuning fork 60 via the pre-amplifier 70 which is received by the controller 110 via detection electronics 120, for example using lock-in amplification techniques (see references [5] and [6]) ; (ii) the output for driving the fine z-tuning element 80 which is supplied via an appropriate voltage supply 130 (standard item for driving piezo-actuators); (iii) the output for driving the coarse tuning element 90 which is supplied via an appropriate pulse generator 140 (see reference [II]) ; and (iv) an output for controlling carrier injection into the gain

medium of the laser. In the system shown in the figure, it is assumed that the laser is electrically pumped (third embodiment) in that the control line is shown as being from the controller 110 to a current source 150 that biases the p-type and n-type contacts to the chip structure 30. Alternatively, for optical pumping (first and second embodiments), the control line would be to drive electronics for a pump laser used for generating the pump signal.

For high volume production, the electronics can be implemented with simple, custom-designed circuits, for example including a suitably programmed digital signal processor (DSP). Much or all of the electronics may be positionable as part of the module 100, either inside or outside the sealed interior formed by the housing 92 and top flange 94. The whole device can be packaged into a standard TO9-sized holder.

Figure 13 shows a device combining a pump source as well as a packaged optically-pumped tunable laser module. The laser module 100 is arranged in a common package 190 with a pump source 160. An optical fibre leading from the pump source 160 is connected to the optical fibre leading from the laser module 100

by a directional coupler 170 so that the pump light Bp is transmitted to the laser module 100 and the laser output kl is passed from the laser module 100 to an output 180.

In the above examples, essentially the entire chip structure (with the exception of the optional AR coating) is a semiconductor heterostructure grown by MBE or other epitaxial deposition technique on a semiconductor substrate. Two alternative ways of fabricating the chip structure are now described.

Figure 14 shows a back-etched chip structure according to an alternative embodiment. The semiconductor layer sequence of the chip structure is similar to that of any of the previously described example, except that the lower cavity mirror layers are omitted. Starting from the top surface, there is thus still the cap 39, upper spacer 38, gain medium 34 (e. g. quantum well (s) or dots) and lower spacer 37. However, instead of making a semiconductor Bragg mirror as part of the heterostructure growth, an aperture 126 is back-etched into the substrate 36 to terminate at the lower surface 127 of the lower spacer layer 37. An example diameter for the aperture is 500 microns. Termination of the back-etching may be controlled by including a stop-etch

layer. A Bragg mirror 135a is then deposited on the termination 127 of the aperture 126 using similar techniques to those used for depositing the Bragg mirror on the end of the fibre, e. g. ion-assisted deposition. For example, a Bragg mirror may be formed by deposition of alternating layers oft02 and Si02.

Figure 15 shows another alternative chip structure. This chip structure is based

on a glass substrate 136 on which is deposited a Bragg mirror 135b using similar techniques to those used for depositing the Bragg mirror on the end of the fibre, e. g. ion-assisted deposition. For example, a Bragg mirror may be formed by deposition of alternating layers of TiO2 and Silos. Then an epitaxial (epi) lift-off semiconductor heterostructure 145 is arranged on the Bragg mirror 135b. The epi-lift off semiconductor heterostructure 145 incorporates the cap 39, upper spacer 38, gain medium 34 (e. g. quantum well (s) or dots) and lower spacer 37 familiar from the previous examples. Epi lift-off is a technique which allows a semiconductor heterostructure to be removed from its underlying substrate by etching away a sacrificial layer. The technique has been developed for and applied to various different semiconductor heterostructure systems, for example GaAs/AlGaAs [12] and InP/InGaAs [13].

Figure 16 shows an example epi lift-off semiconductor heterostructure suitable for the chip structure of Figure 15. The cap 39, upper spacer 38, gain medium 34 (e. g. quantum well (s) or dots) and lower spacer 37 are made of the same materials and have the same thicknesses as the corresponding layers shown in the example of Figure 5. However, under the lower spacer layer 37, there is provided a sacrificial layer 141 of AlAs (e. g. 70 nm thick), under which is situated a buffer layer 136a and substrate 136. To perform the epi lift-off the AlAs layer is etched away with HF, the overlying layers 39,38, 34,37 removed from the GaAs substrate 136 and buffer 136a and placed on the TiO2/SiO2 Bragg mirror 135b, as shown in Figure 15.

It will be appreciated that although a glass substrate is shown, any material may be used for the substrate which is suitable for forming a lower cavity mirror of the desired design.

A key advantage of both the back-etch and epi lift-off approaches for the present invention is that they decouple the lower cavity mirror design from the gain medium design, since fabrication of the lower cavity mirror is removed from the

semiconductor heterostructure. This opens up the invention to any semiconductor heterostructures where suitable gain media can be provided, including wide and narrow band gap II-VI materials and a further variety of III-V materials. Specifically, the choice of semiconductor heterostructure materials system is no longer limited to

those which allow a Bragg mirror to be fabricated. Inorganic semiconductor gain 41 media may also be used. These approaches thus open up the present invention to tunable lasers operating over a very wide range of wavelengths from the infrared to the ultraviolet and beyond.

REFERENCES [1] Larson and Harris : Applied Physics Letters, volume 67, pages 590-592 (1995) [2] Larson and Harris: Applied Physics Letters, volume 68, pages 891-893 (1996) [3] Sugihwo, Larson and Harris: Applied Physics Letters, volume 70, pages 547 549 (1997) [4] Sugihwo, Larson and Harris: Applied Physics Letters volume 72, pages 10-12 (1998) [5] US 5,641, 896 (Karrai) [6] US 6,006, 594 (Karrai and Manus) [7] Amann and Buus, Tunable Laser Diodes, pages 103-133 (Artech House, 1998) [8] E. C. Vail, G. S. Li, W. Yuen, and C. J. Chang-Hasnain: Electronics Letters, volume 32, page 1888-1889 (1996) [9] Toledo Crow et al : Applied Physics Letters, volume 60, pages 2957-2959 (1992) [10] Betzig et al: Applied Physics Letters, volume 60, pages 2484-2486 (1992) [11] US 5,912, 527 (Karrai) [12] Yablonovitch et al: Applied Physics Letters, volume 51, pages 2222-2224 (1987) [13] Shumacher et al : Electronics Letters, volume 25, pages 1651-1654 (1989)

Claims

CLAIMS 1. A laser comprising : a chip structure having a surface, a first mirror integrated with the chip structure and a semiconductor gain medium arranged in the chip structure between the surface and the first mirror; and an optical fibre having an end on which is arranged an integral second mirror, the end of the optical fibre being positioned above the surface of the chip structure to form a laser cavity with the semiconductor gain medium and first mirror.
2. A laser according to claim 1, further comprising a positioning device for varying the distance between the optical fibre and the chip structure, thereby to tune the laser.
3. A laser according to claim 2, wherein the laser cavity has a design length of between one half and 50,10, 9,8, 7,6, 5,4, 3 or 2 lasing wavelengths.
4-A laser according to claim 3, wherein the design length of the laser cavity is 1,2, 3,4 or 5, or 3/2,5/2, 7/2 or 9/2 lasing wavelengths.
5. A laser according to any one of claims 1 to 4, wherein the first mirror is a Bragg mirror comprising multiple alternating semiconductor layers.
6. A laser according to any one of claims 1 to 4, wherein the first mirror is a Bragg mirror comprising multiple alternating dielectric layers.
7. A laser according to any one of the preceding claims, wherein the second mirror is a Bragg mirror comprising multiple dielectric layers.
8. A laser according to any one of the preceding claims, wherein the optical fibre tapers towards the end.
9. A laser according to any one of the preceding claims, further comprising a coupled oscillator having first and second oscillators, wherein at least the first oscillator is formed of an elongated piezoelectric member to which the optical fibre is secured.
10. A laser according to claim 9, wherein the coupled oscillator is a tuning fork having first and second legs which form the first and second oscillators respectively.
11. A laser according to any one of claims 1 to 10, wherein the semiconductor gain medium comprises a heterostructure providing one-dimensional carrier confinement.
12. A laser according to any one of claims 1 to 10, wherein the semiconductor gain medium comprises a heterostructure providing two-dimensional carrier confinement.
13. A laser according to any one of claims 1 to 10, wherein the semiconductor gain medium comprises a heterostructure providing three-dimensional carrier confinement.
14. A laser according to any one of the preceding claims, wherein the semiconductor gain medium is arranged in a mesa.
15. A laser according to any one of claims 1 to 14, further comprising an electrical contacting arrangement for electrical pumping of the laser.
16. A laser according to any one of claims 1 to 14, further comprising a pump source for optically pumping the laser.
17. A method of tuning a VCSEL comprising: (a) providing a chip structure having a surface, a first mirror integrated with the chip structure and a semiconductor gain medium arranged in the chip structure between the surface and the first mirror; (b) providing an optical fibre comprising an end, a core and a clad, with an integral second mirror arranged at the end over the core,

(c) positioning the optical fibre a distance above the surface of the chip structure to form a laser cavity with the semiconductor gain medium and first mirror. (d) injecting carriers into the semiconductor gain medium to provide population inversion : and (e) varying the distance between the surface of the chip structure and the optical

fibre to effect tuning of the laser cavity. c
18. A method according to claim 17, wherein step (e) includes using a control mechanism to identify a desired optical mode of the laser cavity under a feedback control mechanism having a measured control parameter.
19. A method according to claim 18, wherein the measured control parameter is an interaction force between the end of the optical fibre and the surface of the chip structure.
20. A method according to claim 19, wherein the interaction force is measured with a differential signal from a coupled oscillator comprising first and second oscillators, one of which is mechanically linked to the optical fibre.