Quantum Well Infrared Detector

This invention concerns semiconductor infrared detectors.

These detectors have particularly advantageous applications in the range from 8 .mu.m to 12 .mu.m because these wavelengths correspond to a transparent window in the atmosphere. However, although this is a preferred sensitivity range, the invention is not limited to this particular range of values.

The material used most often for these semiconductor detectors is the alloy HgCdTe but this material is metallurgically extremely complex, which makes its industrial production difficult.

Several laboratories have recently proposed using a new type of infrared detector--of the generic type to which the detector of the invention pertains--based on the principle of absorption of photons associated with transitions between two quantum sub-bands appearing in the quantum wells created by stacking a very large number of alternating epitaxial layers of III-V semiconductor material.

The advantage of these detectors comes mainly from the fact that the use of III-V compounds in place of II-VI compounds such as HgCdTe involves matellurgy that is much more manageable, making it easier to effect industrial production of these detectors.

However, for various reasons which are explained in detail below, the performance of these quantum well detectors with III-V semiconductors has so far been inferior to that of HgCdTe detectors.

One of the objects of the invention is to overcome this limitation, by providing a quantum well detector with III-V semiconductors which has a high detection sensitivity.

To this end, the detector of the invention, which is of the type specified above, i.e. comprising a succession of alternating stacked layers of a III-V semiconductor material with a large forbidden band and a III-V semiconductor material with a small forbidden band with p-doping, defining a quantum well with sub-bands of HH and LH type in the region of the layer comprising the material with a small forbidden band in the valence band diagram of each corresponding heterostructure, is characterized in that the thickness of the material with a small forbidden band is essentially selected in such a manner that only two quantum sub-levels LH.sub.1 and HH.sub.1 appear in the well, and in that the energy difference between these two sub-levels corresponds to the energy of the photons to be detected, and the composition of the material with the large forbidden band is essentially selected in such a manner that the height of the barrier adjacent the quantum well is equal to or greater than the energy of the LH.sub.1 sub-band.

As to the material with a large forbidden band, its thickness is very advantageously so selected that the potential barriers defined by the layers of this material are sufficiently low for the resonant tunnel effect occurring through these barriers of light holes populating the sub-level LH.sub.1 to create for these light holes a state in which the wave function thereof is spread in the assembly of the quantum wells and potential barriers, while that of the heavy holes populating the HH.sub.1 sub-level is localized.

The material of this structure with a large forbidden band is preferably Al.sub.x Ga.sub.1-x As and the material with a small forbidden band GaAs. In this case, the thickness of the material with a small forbidden band lies in the range 1.5 nm to 2.5 nm approx., and the thickness of the material with a large forbidden band is about 8 nm approx. In a variant, the material with the small forbidden band can be In.sub.y Ga.sub.1-y As, with an indium content Y.sub.In.ltoreq. 0.05 approx.

The invention is now explained in more detail, with reference to the accompanying drawings. In all of the figures the same reference numerals designate like parts.

FIG. 1 is a schematic representation of the conduction band of a stack of AlGaAs/GaAs layers.

FIGS. 2a and 2b show the behavior of the conduction band of a prior art structure, respectively when quiescent and when biased, but with the thickness of the layers of GaAs reduced in such a manner that the energy difference of the quantum sub-bands corresponds to the wavelength of the light to be detected.

FIGS. 3a and 3b correspond to FIGS. 2a and 2b for the case in which the composition of the AlGaAs layer has been so selected that the higher quantum sub-level comes level with the edge of the quantum well.

FIG. 4 is a schematic representation of the valence band of a stack of AlGaAs/GaAs layers.

FIGS. 5a and 5b illustrate the behavior of the valence band of a prior art structure such as that of FIG. 4, respectively when quiescent and biased.

FIGS. 6a and 6b correspond to FIGS. 5a and 5b but for a structure in which the doping and dimensions of the layers have been determined in accordance with the teaching of the invention.

FIGS. 7a and 7b correspond to FIGS. 6a and 6b, for a modified embodiment of the invention.

The current state of quantum well detectors is firstly recalled, especially the mechanism whereby these detectors operate.

These detectors are essentially formed by a stack of heterostructures 1, as illustrated in FIG. 1, each formed by a layer 2 of GaAs and a layer 3 of AlGaAs. These different layers are deposited epitaxially one on the other and the complete stack can comprise up to about fifty heterostructures 1. This arrangement of layers creates a corresponding succession of discontinuities, alternately quantum wells 4 and potential barriers 5, in the conduction band E.sub.c, shown schematically in FIG. 1.

If the GaAs layers are thin enough, in the order of a few nanometers, quantum sub-bands (quantum levels) E.sub.1, E.sub.2, E.sub.3, etc. appear from the quantum effect. By appropriate selection of the thickness of the GaAs layers, i.e. the width of the quantum wells 4, it is possible to adjust the energy positions of the E.sub.1 and E.sub.2 sub-bands in such a manner that the difference .DELTA.E.sub.1 -.DELTA.E.sub.2 between the energies of the quantum levels E.sub.2 and E.sub.1 will be in the order of 124 meV approx., as illustrated in FIG. 2a. If such a structure is illuminated by light with a wavelength of 10 .mu.m, i.e. by photons with energy h=124 meV, these photons will induce a resonant electronic transition from the sub-band E.sub.1 to the sub-band E.sub.2, symbolized by the arrow 7. If an electric field is applied to such a structure (FIG. 2b), that is to say if the component is biased, the electrons in the sub-band E.sub.2 can pass by the tunnel effect through the potential barriers 5 corresponding to the AlGaAs layers, as symbolized by the arrow 8, thus generating a measurable photo-current.

The first detectors implemented on this principle have been described by B. F. Levine et al., New 10 .mu.m Infrared Detector Using Intersubband Absorption in Resonant Tunneling GaA1As Superlattices, Appl. Phys. Lett., Vol. 50, No. 16, p. 1092 (1987). Account is taken there of the incidence of a parasitic tunnel current due to electrons in level E.sub.1 whose tunneling transparency, although less than that of the electrons in level E.sub.2 cannot be neglected. This tunnel current thus creates a high dark current which is detrimental to the performance of the detector.

In this respect, it is known that the tunneling transparency is expressed by an equation of the type:

where:

m* is the mass of an electron,

.DELTA.E is the height of the barrier for the electron in question (.DELTA.E.sub.1 for the sub-band E.sub.1 and .DELTA.E.sub.2 for the sub-band E.sub.2),

d is the thickness of the potential barrier of AlGaAs, and

V is the applied voltage.

It has then been proposed to increase the thickness d of the barrier and/or to increase its height .DELTA.E.sub.1. However, doing this also reduces the tunneling transparency of the electrons in the sub-band E.sub.2, i.e. those which generate the photo-current. A compromise appears to be difficult to achieve and the quantum well detectors with the best performance currently made are designed in such a manner that the E.sub.2 sub-band is located just at the edge of the well, as seen in FIGS. 3a and 3b,

It is thus possible to increase the thickness of the potential barrier 5 or in other words reduce the dark current without however affecting the photo-current. Reference may be made in this respect to the work of B. F. Levine et al., Bound-to-Extended State Absorption GaAs Superlattice Transport Infrared Detectors, J. Appl. Phys., Vol. 64, No. 3, p. 1591 (1988), which refers to a "detectivity" D*=10.sup.10 cm. Hz.sup.-1/2 /W at 77 K., a result close to but still less than that obtained with a conventional HgCdTe detector, which yields 3.times.10.sup.10 cm. Hz.sup.-1/2 /W at the same temperature.

A first limitation in the performance of this structure is the low lifetime of the electrons in the structure, which is related to the lifetime of electrons in the material (AlGaAs) forming the potential barrier in which the hot electrons move.

A second limitation is similar to the phenomenon of inter-sub-band optical absorption which should obey certain rules of quantum mechanical selection prohibiting light incident perpendicular to the plane of the layers.

Very recently, this second difficulty has been tackled by B. F. Levine et al., Normal Incidence Hole Intersubband Absorption Long Wavelength GaAs/Al.sub.x Ga.sub.1-x As Quantum Well Infrared Photodetectors, Appl. Phys. Lett., Vol. 59, No. 15, p. 1864 (1991), which proposes to use quantum wells created by discontinuities in the valence band, not in the conduction band.

Thus, in the latter case, the optical inter-sub-band absorption can be obtained with light incident perpendicular to the plane of the layers. The cited article thus refers to a detection sensitivity of 3.1.times.10.sup.10 cm. Hz.sup.-1/2 /W at 77 K. for the wavelength 7.9 .mu.m.

However, this proposal also suffers from limitations, due on this occasion to the low lifetime of the holes.

Thus, if we consider, as is illustrated in FIG. 4, the discontinuities created in the valence band of the structure, which alternates quantum wells 9 with potential barriers 10, it is apparent that in each quantum well 9 there are many sub-bands HH.sub.1, HH.sub.2, LH.sub.1, etc., (the designation HH referring to the heavy holes and LH to the light holes), Three of these sub-bands are illustrated in FIG. 4 but the situation can vary very considerably, depending on the width of the quantum wells 9 or the height .DELTA.E.sub.v of the potential barrier 10.

FIGS. 5a and 5b represent schematically the case described by Levine in the article last cited; it relates to quantum wells 9 which are 3 nm to 4 nm wide with a barrier height .DELTA.E.sub.v in the order of 160 meV, for an aluminum content of 0.30 of the material AlGaAs. (Here and in the following the "content" is understood to be the mole fraction x.sub.Al of Al.sub.x Ga.sub.1-x As). In this case, the sub-band HH.sub.2 is nearly at the edge of the well and the difference in energy between the edge of the well and the sub-band HH.sub.1 is 144 meV for a well of 3 nm or 157 meV for a well of 4 nm (FIG. 5a).

Under illumination (FIG. 5b), the photons with corresponding energy (144 meV and 157 meV respectively) are absorbed (arrow 11), thus causing movement of the holes in the continuum of the valence band (arrow 12) when an electric field is applied.

Since it is known that the effective mass of holes in the continuum of the valence band of AlGaAs (and GaAs) is very high, in the order of 0.4 m.sub.o (m.sub.o being the mass of the electron), it will be understood that their mobility will be reduced and that their mean free path will be short, in other words, their lifetime will be very short.

This prior art structure has another problem in the presence of the sub-band LH.sub.1 in the quantum wells 9. Thus this sub-band is populated with holes of low effective mass which are about 50 meV from the edge of the quantum well. Considering equation (1) above, the tunneling transparency of these light holes is thus high, the more so because the Fermi level approaches the level LH.sub.1, i.e. the layer of GaAs is p-doped--which applies in this case with the doping lying between 10.sup.17 cm.sup.-3 and 5.times.10.sup.18 cm.sup.-3 approx. Once again an irreconcilable compromise is encountered between a large barrier width, necessary to reduce the dark current which would otherwise be very high, and a concomitant reduction of the photo-current and thus of the sensitivity of the detector.

The invention, which will now be described, specifically seeks to overcome this problem. Its teaching lies essentially in an appropriate choice of the composition of the AlGaAs alloy and of the thicknesses of the layers of GaAs and AlGaAs.

The invention in particular used the phenomenon of inter-sub-band transition between HH.sub.1 and LH.sub.1 in the first place to eliminate the parasitic tunnel effect due to the light LH.sub.1 holes. In the second place, once this parasitic effect has been eliminated, it is possible to reduce the width of the AlGaAs barrier insofar as this is needed, because the holes populating the sub-level HH.sub.1 are very heavy, exhibiting a low tunneling transparency and therefore only having a small effect on the performance of the detector.

If equation (1) is applied, this reduction of thickness can be by a factor at least equivalent to:

and at most equivalent to

where m*.sub.HH1 and m*.sub.LH1 are taken to be 0.4 m.sub.o and 0.07 m.sub.o respectively (m.sub.o being the mass of the electron), and .DELTA.E.sub.HH1 and .DELTA.E.sub.LH1 are the heights of the respective barriers associated with the sub-bands HH.sub.1 and LH.sub.1, with .DELTA.E.sub.HH1 =190 meV approx. and .DELTA.E.sub.LH1 =66 meV approx.

For the detector to operate at the selected wavelength of about 10 .mu.m it is necessary that the energy between the two sub-bands LH.sub.1 and HH.sub.1 shall be in the order of 124 meV. It can be shown by quantum mechanics calculations that, in this case, the width of the hole should be in the order of 1.5 nm to 2.5 nm, this thickness depending on the height of the potential barrier, i.e. on the aluminum content in the AlGaAs. The calculations show that the height of the barrier should exceed about 230 meV, which requires an aluminum content greater than about 0.42.

A structure in accordance with the invention is illustrated schematically in FIG. 6a (at equilibrium) and FIG. 6b (under illumination and with an applied electric field).

The quantum well detector of the invention is thus formed from a succession of quantum wells 9 of very small width (around 1.5 nm to 2.5 nm), separated by potential barriers 10 which are themselves relatively thin.

It is known that the LH.sub.1 holes, which have a low effective mass, can pass through thin barriers by the tunnel effect, even if the height of the barrier .DELTA.E.sub.v is above their energy level, and this the more easily when these levels present the same energy, thus creating the well-known "resonant tunnel effect". The holes populating the sub-band LH.sub.1 thus come to move freely, perpendicular to the plane of the layers, in a periodic potential, sometimes also known as a "superlattice".

It is thus possible to be released from too precise a definition of the thickness of the AlGaAs by using barrier heights such that the sub-band LH.sub.1 is located right at the edge of the quantum well, as shown in FIGS. 7a and 7b; in this case, the height .DELTA.E.sub.v of the barrier adjacent the quantum well is no longer greater than but just equal to the energy of the sub-band LH.sub.1. This configuration is close to that shown in FIGS. 5a and 5b (corresponding to the proposal of Levine in the cited article) but with the essential difference of the presence, in the prior art situation, of the LH.sub.1 sub-band located in the quantum well below the energy of the potential barrier, contributing thus to an increase in the dark current.

It is also known that although the sub-band LH.sub.1 is quantized, it has its wave function completely spread over the whole structure, by the resonant tunnel effect, which leads to an increase .delta.E.sub.LH1 in the energy of the sub-band, as can be seen at 13 in FIGS. 6a and 6b. This increase .delta.E.sub.LH1 depends on the width of the well. For wells 2 nm wide, the increase is in the order of 50 meV, 15 meV and 5 meV for barriers 5 nm, 7.5 nm and 10 nm wide respectively (the height of the barrier also having an effect on these values).

As for the holes populating the sub-band HH.sub.1, which do not only have a large effective mass but which are also located deep in the hole, these do not experience the tunnel effect and their wave function also remains localized.

In summary, the light holes LH.sub.1 moving in the periodic potential have a lifetime longer than that of the heavy holes moving in the continuum of the valence band, as explained with reference to FIG. 5.

Moreover, although the movement of holes in the band LH.sub.1 will be favored by a small barrier width (a small thickness of the AlGaAs layer), it is desirable not to reduce this value, because too large an increase in .DELTA.E.sub.LH1 will result in a reduction in the energy difference between the sub-bands HH.sub.1 and LH.sub.1, so increasing the dark current.

Knowing however that infrared detectors operate at the temperature of liquid nitrogen, namely 77 K., and that the thermal energy is 6.6 meV at this temperature, it is only important that the increase .DELTA.E.sub.LH1 shall be less than 6.6 meV. It is thus possible to find an optimum barrier width, defined in such a manner that .DELTA.E.sub.LH1 =6.6 meV approx. For wells 2 nm wide, this optimum corresponds to a thickness of AlGaAs of about 8 nm (a value which can vary with the aluminum content of the AlGaAs).

Taking into account the very low values of the width of the quantum wells which have been indicated, a variation of an atomic monolayer, i.e. in the order of 0.2 nm to 0.3 nm involves an relative variation in thickness of 10%. It is thus necessary to use techniques such as molecular beam expitaxy (MBE) or metal organic chemical vapor deposition (MOCVD) which are compatible with such precision and ensure good uniformity of the epitaxy over the whole surface of the semiconductor chip.

Moreover, the invention is not limited to a GaAs/A1GaAs heterojunction and its teaching can apply equally to other heterostructures formed on the basis of III-V alloys.

In particular the alloy InGaAs can be used in place of GaAs to realize the quantum wells. In particular, this alloy InGaAs having a lattice constant greater than that of GaAs, it will be subjected to a uniaxial strain which will have the effect of pushing the sub-band LH.sub.1 further away in energy, in other words of increasing the energy difference between the sub-bands HH.sub.1 and LH.sub.1. It is possible, for the same performance, to enlarge the quantum well by this phenomenon, relative to a similar structure using GaAs, and thus to facilitate practical implementation by virtue of a greater thickness of the layer to be deposited.

This effect is particularly marked: thus, for an indium content (i.e. a mole fraction Y.sub.In of In.sub.y Ga.sub.1-y As) of only 0.03, thus for a composition very close to that of GaAs, the increase in the energy is in the order of 15 meV. It is however noted that alloys with an indium content greater than 0.05 cannot really be used, because the excessive strain then risks creation of interfacial dislocations, given the total thickness of the quantum well structure.