Conventional solid state lasers"pump"electrons into an excited state and emit light of a particular frequency when the excited electrons drop back to their ground state, thereby releasing the excess energy in the form of light. As all the electrons are"pumped"to the same degree they release the same amount of energy, which will be light of a particular frequency. The particular frequency of the emitted light can be determined from the equation E = hf, where f is the frequency of the emitted light, E is the energy of the photon, and h is Planck's constant.

Such solid state lasers have been used to generate electromagnetic radiation. It has, however, proven difficult to produce an efficient and powerful THz radiation source, as there is no good naturally occurring source of such radiation.

THz radiation, can be used for imaging samples and obtaining spectra at each pixel in an image. THz radiation penetrates most dry, non-metallic and non-polar objects like plastic, paper, textiles, cardboard, semiconductors and non-polar organic substances.

Therefore THz radiation can be used instead of x-rays to look inside boxes, cases etc.

THz radiation also has medical uses due to being non-ionizing.

Generally, continuous wave (CW) Terahertz radiation has been generated using p-doped Ge lasers and photomixers. However, p-doped Ge lasers have the disadvantage of requiring cooling at liquid Helium temperatures. Photomixers combine the difference frequency of two diode lasers. Although photomixers have the advantage of being able to operate at room temperature, the amount of power available using this technique is still confined to the u. W level.

Kohler et al, Nature 417,156 (2002) have reported Terahertz emission from a Quantum cascade laser (QCL).

Quantum cascade lasers were developed in 1994 by researchers at AT&T Bell Labs.

QCLs are a type of laser formed by a plurality of layers of different materials. In other words, the conduction band is made up of a number of sub-bands. In these lasers, electrons are again"pumped"but when they fall back to their ground state, the electrons effectively cascade down an energy staircase formed by the different sub-bands. At each step a photon of light is emitted. Therefore, instead of each electron emitting a single photon when falling to their normal state, as occurs with standard lasers, a number of photons are emitted. The amount of energy emitted and hence the wavelength for each photon can be controlled through the thickness of the layers. The radiation frequency is determined by the energy spacings of the sub-bands.

Recently, so-called bound-to-continuum QCLs, have been fabricated, for example, see Faist et al. Appl. Phys. Lett 78,147 (2001). Here, the upper state of the laser transition consists of an isolated subband inside a minigap. The lower state of the laser transition is given by the highest energy level of a miniband which consists of several subbands.

The upper state of the laser transition will be termed the"second state"and the said highest energy level of the miniband, the"first state". The lowest energy state of the miniband is the ground state.

To form the above structure, the QCL comprises a number of repeating periods, each period comprising a plurality of layers which are configured to primarily support a plurality of discrete energy states including the above described ground state, first state and second state. Adjacent periods are separated by an injection barrier. As the laser is brought towards resonance, the second state of one period starts to align with the ground state of an adjacent period such that carriers can tunnel through the injection barrier separating the adjacent periods. The states will never align properly, meaning that their energies will never be equal. Resonance is defined as the condition where the energy separation between the two states, also called"energy splitting", is minimum.

In a QCL operating in the THz frequency range, smaller splitting at resonance is required than in the case of Infra Red QCLs in order to avoid flow of carriers directly from the ground state of one period to the first and lower states of the following period, i. e. bypassing the second state. To achieve smaller splitting and thus ensure that more electrons are injected into the second state, it is necessary to increase the size of the injection barrier. This may be done by either varying the composition of the barrier to increase its height or increasing the thickness of the barrier. Increasing the size of the barrier has the effect of decreasing current flow which limits the current dynamic range and which can ultimately limit high temperature operation.

To address this and other problems, the present invention provides a laser comprising a quantum cascade layer structure, said quantum cascade layer structure comprising n repeating periods of a plurality of layers, wherein n is an integer of at least 2, said plurality of layers being configured to primarily support a plurality of discrete energy states, adjacent periods being separated by a barrier layer, the laser being configured such that electrons, injected into an mth plurality of layers in a mth period, descend through at least one energy state of said mth plurality of layers into an energy state of a m+lth plurality of layers in an m+lthperiod, and an overlap condition: ###2,m+1(x)#2##1,m(x)#2dx# 3. 1x10-5 nm is satisfied, where m is an integer between 2 and n, MJ2, m+1 iS the normalised wavefunction in nm of a second state of the m+lth period, MJI, m iS the normalised wavefunction in nm of a first state of the mth period and x is measured from the injection barrier which lies between the mth and m+lth periods, measured in the direction towards the mth period, the second state of the m+lth period is the state of the m+lth period which is closest in energy to the ground state of the mtl'period, the ground state of the mth period being defined as the state which has the largest amplitude of the states on the m+lth side of the barrier layer, and the first state of the mth period being defined as the state which forms the largest dipole within a period with the second state, wherein the dipole can be expressed mathematically as: <BR> <BR> <BR> <BR> <BR> #<BR> <BR> <BR> ##1,m(x)x#2,m*(x)dx.<BR> <BR> <BR> <BR> <BR> <P>-# In the above laser, the second state extends significantly through the injection barrier into the preceding period. This increase in the population density of the second state in the preceding period is believed to aid coupling between the second state and other states in the preceding period besides the grounds state and hence improve current flow through the laser. The extension of the second state into the preceding period is defined in terms of the overlap between the second state of one period and the first state of the preceding period.

The above laser has been found to have a lower threshold current density than those previously reported at the same or longer emission wavelength. Also, lasers fabricated in accordance with the above aspect of the present invention have been found to lase up to 95K, which is higher than current reported operating temperatures for an emission wavelength of greater than 100, m (103, m). CW lasing action was observed up to 70K.

The above overlap condition between the second state of one period and the ground state of a preceding period is stated when preferably the following condition holds: #E#2,m+1-E#g,m##2#E#2,m+1-E#g,m#resonance, where E#2,m+1 is the energy of the second state of the m+1th period and E#g,m is the ground state of the mth period.

The centres of the second state and the first state within a period should be preferably separated with the first state wavefunction, preferably occupying mostly the 4th and 5th quantum wells after the injection barrier, in order to limit direct scattering from the second state into the first state and from other states in the miniband into the first state.

Also, preferably, a large energy difference is introduced between the second state and the upper miniband of the same period in order to avoid re-absorption of the emitted radiation. Thus, preferably, this energy gap is larger than the laser transition energy. In use, an electric field will be applied across the active region or quantum cascade layer structure. To reduce the electric field across each period and to thus reduce Joule heating effects a period should typically be long. Preferably, the length of a period measured between injection barriers is at least 120nm. The use of a longer period also allows a smaller number of periods to be used for a given active region thickness thus leading to the use of a lower operating voltage, which also reduces Joule heating Thus, in a second aspect, the present invention provides a laser comprising a quantum cascade layer structure, said quantum cascade layer structure comprising n repeating periods of a plurality of layers, wherein n is an integer of at least 2, said plurality of layers being configured to primarily support a plurality of discrete energy states, adjacent periods being separated by a barrier layer, the laser being configured such that electrons, injected into an mth plurality of layers in a mth period, descend through at least one energy state of said mth plurality of layers into an energy state of a m+lth plurality of layers in an m+lth period, where m is an integer between 2 and n, and wherein the length of a period measured between injection barriers is at least 120nm.

Preferably, the length of a period measured between injection barriers is at most 250nm.

Also, in order to reduce Joule heating effects, the width of the states in the mth period in energy is preferably 20meV or less. The width of the miniband may be calculated by using a suitable modelling technique such as an effective mass approximation.

In a preferred embodiment, the laser further comprises a substrate, said substrate comprising a bulk region and a conducting layer; said quantum cascade layer structure being provided on a first surface of the substrate such that said active region is electrically connected to said conducting layer; first and second contacts provided to said conducting layer such that said first and second contacts are electrically connected to said quantum cascade layer structure, said first and second contacts being disposed on opposite sides of said quantum cascade layer structure and an active region contact provided to said quantum cascade layer structure such that a potential may be applied between said active region contact and said first and second contacts to cause said quantum cascade layer structure to lase.

Preferably the laser comprises one or more facets which are coated with a reflective material such as gold.

The conducting layer may comprise a highly doped semiconductor. Further, the conducting layer is preferably thin enough, such that in operation, the two surface plasmons present at the two interfaces of the conducting layer merge into a single mode.

The present invention will now be described with reference to the following non- limiting embodiments in which : Figure 1 is a schematic of a laser in accordance with a preferred embodiment of the present invention; Figure 2 is a plot of the results from a model of the band structure of the laser described with reference to figure 1; Figure 3 is a plot of the Voltage and Power against current density for the device of figure 2, when the device is drive in pulsed mode, the inset shows JV curves of the laser and mesa at 4K ; Figure 4 is a plot of Power against current density for the device of figure 2, when the device is driven in continuous wave, the inset represent spectra at different current densities, the lower two spectra in a purged system, the upper in an unpurged system ; and Figure 5 are plots of the differential resistance at 4K against current density, figure 5 (a) for the laser and figure 5 (b) for the mesa.

Figure 1 schematically illustrates a QCL laser.

The QCL laser has an active region 1 provided on a surface of substrate 2. The substrate comprises a bulk insulating region 8 and conducting region 4 formed as a layer at the surface of said bulk region 8.

An active region 1 forms a strip or ridge 5 on the surface of said conducting layer 4.

The cross section of said strip 5 is substantially trapezoidal, and arranged so that the largest parallel side of the trapezium is adjacent conducting layer 4. The two sloping sides 9, 10 of the trapezium lean inwards towards the top surface of the active region 1.

First and second lower contacts 16 and 17 respectively are provided to the conducting layer 4 on either side of said active region 1. A top contact 31 is provided to the top of said active region 1. The contacts are arranged such that application of a bias between the top contact 31 and one or both of the lower contact 16 and 17 causes a bias to be applied vertically across the active region 1. Upon application of a suitable bias, the active region will lase.

The active region 1 comprises a structure constructed by periodically stacking many repeated elementary layers. The requirements of the active region will be discussed in more detail with reference to figure 2. In this particular embodiment, the active region 1 is a hetero-structure consisting of 90 repeated long-periods of GaAs/Al0 l5Gao 8sAs.

Each period comprises alternating layers of GaAs/Al0 15Gao. 8sAs, the layer thickness in nm of the layers of each period are, starting from the injection barrier 3.8 Alo. lsGao. ssAs /14. 0 GaAs/0.6 Alo. isGao. 8sAs/9. 0 GaAs/0.6 Alo. lsGao. ssAs/15. 8 GaAs/1.5 Alo. isGao. asAs/12. 8 GaAs/1.8 Alo. lsGao 8sAs/12. 2 GaAs/2.0 Alo. isGao. 8sAs/12. 0 GaAs/2.0 Alo. lsGao. 85As/11. 4 GaAs/2.7 Alo. isGao. 85As/11. 3 GaAs/3. 5 Alo. lsGao. 85As /11. 6 GaAs. The GaAs quantum wells which are 12.0 nm and 11. 4 nm thick are n doped at a level of 1. 6x1016 cm~3.

Alternative lasing structures may be used such as a GaAlInAs-InGaAs combination and a InAs/GaInSb/AlSb interband cascade laser.

The layer structure is finished off with a heavily n-doped (typically n equal or greater than 5 x 1018) GaAs layer.

As described above, the substrate 2 comprises a bulk region 8 and a conducting layer 4.

In this particular example, the conducting layer is 700 mu thick and doped with Si (n = 2 x 1018 cm~3). The total thickness of the bulk region 8 and conducting layer 4 is approximately 200pm.

In order to act as an efficient waveguide for the emitted photons, the dielectric constant of the conducting layer 4, SA is negative with respect to the dielectric constant of the surrounding semiconductor EB. During operation, two surface plasmons will exist at the two interfaces of the conducting layer 4. Layer 4 is made thin enough, so that these will merge into a single mode.

The laser structure of figure 1 may be fabricated using a growth technique such as molecular beam epitaxy (MBE).

The active region 1 is formed into a ridge 5 by wet chemical etching the active region 1 down to the conductive layer 4. After etching, the active region has a substantially trapezoid shape, so that a surface of the active region 1 above the surface level of the doped channel has a width which is less than the width of the active region 1 adjoining the conductive layer 4.

In the present embodiment of the invention, the ridge is 220pm wide at its widest along the edge adjoining conducting layer 4. Although this distance may readily vary, and a range of 20 to 500 microns could be expected.

Top contact 31 is created over at least a portion of the active region. With reference to Figure 1, the top contact 31 lies only on the upper surface of the active region 1. The top contact 31 is formed by depositing a metal strip on the n-GaAs layer on the upper surface of the active region 1. The metal strip, such as a Pd/Ge strip, may be defined using optical contact lithography. Preferably this strip is between 20 and 500 microns wide and with a length of more than 20 microns.

The contact is formed by annealing 25 mn of Pd and 75 nm of Ge at 350 degrees Celsius to form an ohmic contact. Although any suitable metallic materials may be used for the strip, a Pd/Ge alloy is useful as it has low diffusion into the semiconductor layers forming the active region 1. Typically the diffusion for Pd/Ge is limited to a few tens of nm.

Therefore, in effect, the active region 1 is embedded between a top contact 31 and the conducting layer 4. In the present example the top contact 31 is a 80nm thick GaAs layer and is doped at n = 5 x 1018. This top contact 31, together with the conducting layer 4, act as electrical contacts and also provide guiding layers from the emitted radiation.

Two lateral contacts 16, 17 are created on the conducting layer 4 on either side of the ridge 15 of the active region 1. These lateral contacts 16,17 may be formed by any means, such as by evaporation. Preferably these contacts are alloyed Ti/Au, but may also be provided by NiAuGe or other contacts which make a good ohmic contact with n-type material.

The lateral contacts 16,17 are spaced from the ridge 15 at a distance of 50 n. m.

The laser of figure 1 has back 21 and front 22 facets. One or both of these facets may be coated in order to increase its reflectivity. Gold and nickel or a Ti/Au combination are materials suitable for such coatings, other materials may be used such as ZnSe or ZnAu. Where. Ti/Au is utilised, appropriate thicknesses of the materials are 10 and 100nu respectively.

By coating at least one of the facets 21, 22 with a suitably reflective material the coating can act as a perfectly reflecting mirror, which allows the threshold current density to be reduced. This therefore assists in the ability of the laser to achieve CW operation. This is apparent by considering the expression relating to mirror loss: aM = 1/2L x ln (R1R2) (1) where L is the length of the ridge and Ri, R2 are the reflectivities of the facets. For an uncoated facet, the reflectivity is approximately Ru = 0.34, whereas for a coated facet, the reflectivity is approximately Rc = 1. Therefore, at sub-mm wavelengths, an Au or Ni layer can act as a perfectly reflecting mirror.

Therefore, from this equation it is apparent that by coating one fact of the laser, a 50% reduction of the mirror losses can be obtained, from (1/L) ln (Ru) to (1/2L) ln (Ru). This is also equivalent to doubling the length of the cavity. So an effective cavity length LEFF can be defined as L for an uncoated laser and 2L for a coated one.

The reflective coating may be preceded by the evaporated of a Si02 layer to prevent electrical shorting. Preferably this layer is 200nm thick.

Figure 2 is a band diagram of two periods in the active region or quantum cascade layer structure region of the device of figure 1. The modulus squared of the wave functions under an electric field of 2. lkV/cm are shown. Area 101 and shaded area 103 represent two adjacent injector mini-bands. In period one, 105, two layers of the mini-band are shown. However, in reality, there are many states as shown in the mini-band 101 of second period 107.

Two states are shown in first period 105, the ground state 109 and the first state 111.

The ground state 109 can be recognized because it is the state of the mini-band 103 which has the highest population distribution on the left-hand side of the injector barrier 113. Injector barrier 113 lies between the first period 105 and the second period 107.

How the first state 111 is determined will be explained with reference to the second period 107.

In the second period 107 two states are shown in bold, the first state 115 and the second state 117. It is the transition between these states 115 and 117 which forms the laser transition. Second state 117 is determined by its relationship to ground state 109 of the preceding period 105. The second state of period 107 and the ground state of period 105 should be the two states which are closest in energy. When the laser is brought close to resonance, electrons are believed to tunnel from the ground state 109 of first period 105 through injection barrier 113 into the second state 117 of second period 107.

Electrons in second state 117 of first period 107 then drop to first state 115 emitting a photon of the desired energy. State 115 is determined as it is the state with the largest dipole with second state 117. The size of the dipole may be calculated mathematically as: <BR> <BR> <BR> <BR> <BR> #<BR> <BR> <BR> <BR> ##1,m(x)x#2,m*(x)dx.<BR> <BR> <BR> <BR> <P> -# In order to enhance the current density, the second state 117 of period 107 extends into first period 105. It is believed that the extension of this state into the first period 105 aids coupling of this state with the other states of the mini-band 103. The extension of this state into first period 105 is designed with relation to the overlap of the second state of the second period 107 with the first state 111 of the first period 105. Ideally, this should be greater than 3. lux10-5 nom when not at resonance. The overlap is defined mathematically as: ###2,m+1(x)#2##1,m(x)#2 dx#3.1x10-5nm The energy difference between the upper state of the laser transition, 117, and the superlattice miniband is 11 meV. Oscillator strengths off21 = 1 1 andf2l = 4, where 1 and 1'are the top states of the miniband were computed. In agreement with superlattice sum-rules, the oscillator strength to the remaining states is negligeable. It follows that the main radiative transition is between states 2 and 1.

The energy splitting at resonance between the second state and the ground state of the injector miniband, g, is A = 1.7 meV. Together with the lifetime of the upper state, the magnitude of A controls the resonant tunnelling from g to the second state. Compared to QCLs operating in the mid-infrared, THz QCLs are characterized by substantially smaller As, leading to a lower tunnelling rate. This is necessary in order to selectively inject electrons into the upper state and not in the miniband, which is more critical for smaller transition energies. However, small values of A typically lead to small maximum current densities, which limits the current dynamic range and ultimately can limit the high temperature operation. This detrimental effect is compensated by deliberately allowing the upper state wavefunction to penetrate deeply inside the miniband on the right side of the injection barrier. This way coupling with states other than g should be enhanced.

Also, the centroids of upper and lower state wavefunctions are more spatially separated, with state 1 wavefunction occupying mostly the 4th and 5th quantum wells after the injection barrier. For low transition energies, this is believed to be a crucial requirement to limit direct scattering into 1, from state 2 and in general from all the states of the miniband.

To minimise joule-heating effects, the miniband was designed to be only 18 meV wide and extending over 125 nm, i. e. slightly longer compared to previous superlattice designs. This latter characteristic not only reduces the electric field, but also allows using a smaller number of periods for a given active region thickness, which leads to a lower operating voltage.

Finally, care was taken to avoid re-absorption from states 2 and g of the emitted radiation. To this end, the energy difference between the upper miniband starting with subband 119 and state 2,117, was made substantially larger (17.5 meV) than the transition energy. Also, dipole matrix elements between g and the higher lying states of the miniband are negligeable.

Figure 3 is a plot of current density vs voltage (JV) and light vs current density characteristics (LJ) from a 3 mm-long device with one facet back-coated with gold. The laser was driven in pulsed mode, with a 100 kHz repetition rate and a 1% duty cycle. In order to avoid atmospheric-water absorption, radiation was collected with the optical path under vacuum. This was achieved by means of a wide-area room temperature thermopile detector mounted on the window of the cryostat, and placed at a distance of 1 mm from the laser facet.

The threshold current density at 4K is Jth = 110 A/cm2, and a maximum emitted peak power of 50 mW is reached at 280 A/cm2. On the right axis of Figure 3, the scale results from calibrated power measurements. At 60K a current density of Jtl, = 155 A/cm2 is measured. Raising further the heat sink temperature results into a rapid increase of the threshold current density up to 300 A/cm2 at T = 95K, where the limit of the power supply is reached.

From the JV curve at 4K a threshold voltage of 2 V was measured, corresponding to an electric field of 17. 3 kV/cm, as for the band diagram of Figure 2. At approximately 3V, the end of resonant injection into the upper state of the laser transition produces a dramatic increase of the differential resistance, which ultimately limits the high current and high temperature operation of the laser. On the other hand, an abrupt decrease of the differential resistance at J = Jth is visible up to high temperatures. This indicates good injection efficiency and a small ratio Tl/T21, where il and i21 are the lower state and the upper-to-lower state lifetimes. To better illustrate the effect, in the inset of Figure 3 demonstrates JV characteristics at 4K from the laser and from a 75 J. m- diameter mesa, where the effect of optical feedback is absent.

The weak temperature dependence of the slope efficiency, decreasing by only 15% from 4 to 77K, is a further confirmation of a small ratio Tl/T2l. From the LJ at 4K the slope efficiency at threshold is 70 mW/A. Using waveguide losses of 10 cri 1, a slope efficiency of 170 mW/A is calculated, assuming unit injection efficiency and il = 0.

The latter, however, is a very crude approximation.

Figure 4 demonstrates LJ characteristics in CW. Compared to pulsed operation, there is virtually no degradation of the threshold current density and average power up to- 50K. This is believed to be a consequence of the low operating voltages and currents, considerably reducing the effect of joule-heating. Above 50K, Jth increases very rapidly with temperature and lasing stops at 70K. The inset demonstrates emission spectra obtained by FTIR (0.25 crri resolution) and a liquid-He cooled bolometer (same set-up used for the CW LJs). Starting from the bottom, the first two spectra were obtained with the optical path purged with nitrogen. Up to ~ 170 A/cm2 the laser emits in a single mode centred at 96.7 cru 1 (2. 9 THz), in good agreement with the band diagram of Figure 2. Multimode emission in the range 94-98.5 cri 1 (2. 8-2.95 THz) is observed at higher current densities as shown in the central panel. In the top panel we display the same multimode spectrum collected with the optical path un-purged. The pronounced deep at 96.2 cm~2 is due to an atmospheric-water absorption line. Despite the purging, water absorption is responsible for the ripples in the CW LJs.

Figure 5 (a) is a plot of the differential resistance, Rd, of the laser as a function of the current density at 4K. The latter initially increases up to J = 40 A/cm2, corresponding to a bias voltage of 0.75 V. Below this voltage transport results from electrons being scattered from one miniband directly into the adjacent one, without being injected into the upper state. By augmenting the bias, this component of the current is reduced owing to adjacent minibands being separated by an increasing energy. Above approximately 0.75 V, electrons start to be injected into the upper state, yielding a fast decrease of the differential resistance. Population of the upper state of the laser transition begins therefore at- 40 A/cm2. Such"leakage current"limits the minimum achievable threshold current density and should therefore be further reduced.

After the abrupt decrease at threshold (1 10 A/cm2), Rd increases steadily. This is not a consequence of a reduction of the injection efficiency. Indeed, from band diagram calculations, states g and 2 anticross at Vbias = 2. 7 V, corresponding to saturation of the output power in the LJ curves of Figure 3. This point is highlighted in Figure 5 (a) by the vertical dotted line, and should correspond to a maximum tunnelling rate between states g and 2, i. e. to a minimum differential resistance. This is confirmed in figure 5 (b), displaying the differential resistance of the 75 pm diameter mesa, where a clear absolute minimum is observed at the anticrossing (Vbias = 2.7 V, J = 180 A/cm2 ; see the JV curve in the inset of Figure 3).

Therefore, the increase of Rd for J > Jth is believed to be produced by an increase of the ratio'Ul/'r2l and is believed to be a result of a cunent-induced heating of the electronic distribution. Electron-electron scattering is believed to be very effective in increasing the temperature of the electrons. The magnitude of the effect depends strongly on carrier density and it was shown that inside the miniband of a mid-infrared QCL it becomes important for sheet densities greater that 1 x 1010 cm-2. In the present laser the nominal sheet doping density is 6.4 x 101° cm~2, and contrary to a mid-infrared QCL, there is no efficient LO-phonon emission to cool the electron gas. This could lead to strong carrier heating with increasing current density, with the following detrimental consequences: (i) decrease of 321 via activated LO-phonon emission from the upper state; (ii) effective increase of cl owing to hot carriers filling the lower state of the laser transition. By inspection of Figure 5 (a) it appears that Rd increases more rapidly above- 200 A/cm2. This is consistent with the fast deterioration of Jth for temperatures higher than 70K (Figure 3).